Shape Casting: Fourth International Symposium 2011 (in honor of Prof. John T. Berry)

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry

*<.

140th Annual Meeting & Exhibition Check out these new proceeding volumes from the TMS 2011 Annual Meeting, available from publisher John Wiley & Sons: 2nd International Symposium on High-Temperature Metallurgical Processing Energy Technology 2011 : Carbon Dioxide and Other Greenhouse Gas Reduction Metallurgy and Waste Heat Recovery EPD Congress 2011 Friction Stir Welding and Processing VI Light Metals 2011 Magnesium Technology 2011 Recycling of Electronic Waste II, Proceedings of the Second Symposium Sensors, Sampling and Simulation for Process Control Shape Casting: Fourth International Symposium 2011 Supplemental Proceedings: Volume 1: Materials Processing and Energy Materials Supplemental Proceedings: Volume 2: Materials Fabrication, Properties, Characterization, and Modeling Supplemental Proceedings: Volume 3: General Paper Selections To purchase any of these books, please visit www.wiley.com. TMS members should visit www.tms.org to learn how to get discounts on these or other books through Wiley.

eMM«Y

, ■ V^

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry Proceedings of a symposium sponsored by the Aluminum Committee of the Light Metals Division and the Solidification Committee of the Materials Processing & Manufacturing Division of TMS (The Minerals, Metals & Materials Society) Held during the TMS 2011 Annual Meeting & Exhibition San Diego, California, USA February 27-March 3, 2011 Edited by Murat Tiryakioglu John Campbell Paul N. Crepeau

WILEY

A John Wiley & Sons, Inc., Publication

TIRAIS

Copyright © 2011 by The Minerals, Metals, & Materials Society. All rights reserved. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of The Minerals, Metals, & Materials Society, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http:// www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Wiley also publishes books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit the web site at www.wiley.com. For general information on other Wiley products and services or for technical support, please contact the Wiley Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Library of Congress Cataloging-in-Publication Data is available. ISBN 978-1-11802-937-4 Printed in the United States of America. 10 9 8 7 6 5 4 3 2 1

©WILEY

A John Wiley & Sons, Inc., Publication

THUS

TABLE OF CONTENTS Shape Casting IV In Honor of Professor John T. Berry Foreword Editors Session Chairs/Lead Reviewers

ix xi xiii xv

Shape Casting IV Modeling The History of Casting Process Simulation C. Heisser, E. Flender, andJ. Stern

3

State of the Art Review of Modelling Entrainment Defects in the Shape Casting Process 13 C. Reilly, N. Green, and M. Jolly Physical and Computational Models of Free Surface Related Defects in LowPressure Die-Cast Aluminum Alloy Wheels 21 J. Duan, D. Maijer, S. Cocker oft, C. Reilly, K. Nguyen, and D. Au Solidification Model Coupling Lattice Boltzmann Method with Cellular Automation Technique H. Yin, L. Wang, andS. Felicelli

29

Physical Characterization of the Permeability of Equiaxed Eutectic Structures in Hypoeutectic Aluminum Alloys 37 E. Khajeh, and D. Maijer Foam Filters Used in Gravity Casting F. Hsu, and H. Lin

45

Simulation of Macrosegregation during Directional Solidification using Mesh Adaptation 53 U. Sajja, andS. Felicelli

v

A Mathematical Model for Simulating the Microporosity of Squeeze Casting of Aluminum Alloy 61 Z Han, J. Li, W. Yang, and B. Liu

Solidification Review of Defect Behavior in Ni-Based Superalloys J. Campbell Premium Quality Super Duplex Stainless Steel Castings without Secondary Refining B. Puhakka

71

79

In Situ High Speed X-Ray Observation of the Solidification of AL15CU with and without A1203 Composite Addition 87 R. Hamilton, A. Phillion, A. Leung, T. Connolley, P. Lee, and P. Rockett In-Mold Thermal Analysis of Ductile Cast Iron M. Onsoien Modeling of Hot Tearing and Its Validations in Metal Castings J. Quo, J. Zhu, andS. Scott

95 103

Effect of Alloying Elements (Magnesium and Copper) on Hot Cracking Susceptibility of AlSi7MgCu Alloys S. Bozorgi, K. Haberl, C. Kneissl, T. Pabel, and P. Schumacher

113

Hydrogen and Cooling Rate Effects on Microporosity Formation in the Production of Defect-Controlled Fatigue Specimens R. Squatrito, I. Todaro, L. Ceschini, L. Tomesani, and A. Morri

121

Effects of Gravity on the Columnar to Equiaxed Transition in Directional Solidification W. Mirihanage, and D. Browne

129

Properties Fracture Surface Facets and Fatigue Life Potential of Castings M. Tiryakioglu, J. Campbell, and C. Nyahumwa

139

Effect of Holding Time before Solidification on Double-Oxide Film Defects and Mechanical Properties of Aluminium Alloys 149 M El-Sayed, H. Salem, A. Kandeil, and W. Griffiths vi

Weibull Analysis of Thin A356 Plates Cast with an Electromagnetic Pump Green Sand Process R. Lett, S. Felicelli, J. Berry, R. Cuesta, R. San Jose, andJ. Antonio Maroto

157

Guidelines for 2-Parameter Weibull Analysis For Castings M. Tiryakioglu, and D. Hudak

165

Melt Cleanliness, Hydrogen Content and Tensile Properties of A356 D. Dispinar, A. Nordmark, and F. Syvertsen

173

The Origin of Griffith Cracks J. Campbell

181

The Use of the Weibull Statistical Method to Assess the Reliability of Cast Aluminum Engine Blocks Made from Different Casting Processes 191 G. Byczynski, and R. Mackay Ultra-High Strength Sand Castings from Aluminum Alloy 7042 O. Senkov, A. Druschitz, S. Senkova, K. Kendig, andJ. Griffin

199

Relationship between Structure and Properties of Al-Cu Alloys A. Ares, L. Gassa, C. Schvezov, andS. Gueijman

207

Microstructure Characterization of Magnesium Control Arm Castings L. Wang, R. Lett, S. Felicelli, andJ. Berry

215

Methods and Systems The Effect of Reduced Molecular Weight of the Pattern on the Properties of Al Alloy Castings Made by the Lost Foam Casting Process 225 K. Siavashi, C. Topping, and W. Griffiths Classical Nondestructive Testing Techniques Do Not Correlate with Strength as Does Process Compensated Resonant Testing 233 R. Nath, M. Johnson, C. Grupke, and C. Leonard Advanced Methoding Concepts for the Gravity Casting of Steel Alloys B. Puhakka

241

Advanced Casting Mold Design Technology of the LCS Waterjet Inlet Tunnel Entry Edge Components 249 L. Nastac, andJ. Romanelli

vu

Evaluation of the Distortion of a Hydro Turbine Blade during Heat Treatment Process 257 J. Kang The Capability Enhancement of Aluminium Casting Process by Application of the Novel CRIMSON Method 265 X. Dai, M. Jolly, and B. Zeng Optimization of the Process Parameters and Tooling Improvement for the Rheocasting of High Quality Aluminum Components Using the SEED Process C. Zheng, E. Samuel, and F. Laplume

273

Shaped Castings and Machining J. Wyatt, andJ. Berry

281

The Estimable Value of'Clever' Experiments

289

J. Berry Author Index

311

Subject Index

313

vin

in honor of Prof. John T. Berry

Foreword It was the spring of 1990 when paths of Prof. John T. Berry and mine crossed for the first time. I was looking for research opportunities in castings in graduate schools. I had written a letter to the Ph.D. supervisor of my father, the late Dr. Voya Kondic, who had then forwarded my request to one of his many former Ph.D. students, John Berry. Prof. Berry was the department head of the Metallurgical Engineering program at the University of Alabama at that time when he sent me a letter and kindly offered me a graduate assistantship. It was an agonizing decision to decide to pursue my M.S. degree elsewhere and it was even more agonizing to send a letter to Prof. Berry informing that I would have to decline his offer. In the letter, I wrote that he would always be my friend and that I hoped to meet him one day. I thought that I had offended this kind man by deciding not to join his program. I could not have been more wrong. Since 1990,1 have had the pleasure of getting to know Prof. Berry, as a scholar, a mentor, a teacher, a leader. His academic accomplishments in many prestigious universities in the United States speak for themselves. The success of his former students, some of whom serve as faculty members and academic leaders, including a university president is even more impressive. Prof. Berry is a role model for any scholar; he has a passion for learning and even a greater passion to share his knowledge with others. He has a unique character that combines a high level of energy and patience for others. Although I never had the pleasure to work with Prof. Berry at the same institution, I feel privileged to know him. He was there for me whenever I needed something from him; data, a recommendation letter, an endorsement, a candid review of a paper. I learned a lot from his published literature on castings. I think that I learned even more from him on how to help others. This symposium, the fourth in the series, has been organized to celebrate the accomplishments of Prof. Berry as a scholar in castings and solidification, a mentor, an advisor and a friend. The papers included in this volume were recruited to reflect the broad research interest of Prof. Berry. John, many thanks for all that you have done for the casting world and those who were fortunate enough to have met you. You have made the world a better place. Murat Tiryakioglu December 21, 2010 P.S. The picture on the cover is from the Ph.D. thesis of my late father, Dr. Ergin Tiryakioglu (University of Birmingham, UK, 1964) who used the results from Prof. Berry's thesis (1954). This picture, which shows an optimum size feeder, was selected as a tribute to John's many contributions to feeder dimensioning and was cast in the same laboratory where Prof. Berry conducted his Ph.D. research.

xi

Symposium Organizers/Editors Murât Tiryakioglu is the director of School of Engineering and a Professor of Mechanical Engineering at the University of North Florida. He received his B.Sc. in Mechanical Engineering from Bogaziçi University, M.S. and Ph.D. in Engineering Management from the University of Missouri-Rolla, and another Ph.D in Metallurgy and Materials from the University of Birmingham, England. Dr. Tiryakioglu grew up in his family's foundry, which continues to thrive in Istanbul, Turkey. This has led to research interests in process design for high quality castings, aluminum heat treatment modeling and optimization, processstructure-property relationships in metals, statistical modeling and quality and reliability improvement, on which he has written over 100 papers, technical reports and book chapters, and edited 6 books. Dr. Tiryakioglu is the recipient of the inaugural John Campbell Medal awarded by the Institute of Cast Metals Engineers in UK. He was selected a TMS Young Leader in Light Metals and was awarded the SME Eugene Merchant Outstanding Young Manufacturing Engineer. He is a member of TMS, and a senior member of ASQ, and is an ASQ Certified Quality Engineer. John Campbell has retired from his post at The University of Birmingham and his editorship of the International Journal of Cast Metals Research. He keeps in touch, retaining an Emeritus Status at the University. Since these moves he has mainly occupied himself with practical work in foundries around the world, applying, testing, and extending the latest technology to upgrade quality and reduce costs. Despite some casting successes in which the author is grateful and proud, he has also accumulated a few scrapped castings that confirm the technology of filling castings is still not fully developed. Thus globetrotting activity constitutes valuable education. However, longer-term updating of existing books and the writing of further books has been delayed, but Castings, 2ed (2003) and Castings Practice (2004), in addition to the proceedings the previous Shape Casting Symposia, all remain bargains! Complete Castings Handbook should appear in 2011, and will be required reading. xin

Paul N. Crepeau is a GM Technical Fellow in Advanced Materials Engineering at General Motors Powertrain Group in Pontiac, MI USA. He supports aluminum intensive engine programs and leads a multidisciplinary team merging CAE and Materials Engineers to advance structural FEA of automotive components. Dr. Crepeau received his B.S. in metallurgical Engineering at the University of Alabama (1978) and, after a 5-year respite at an iron foundry, both M.S. in Metallurgy (1985) and Ph.D. in Metallurgical Engineering (1989) from the Georgia Institute of Technology. He has published in the areas of fracture mechanics, molten metal processing, quantitative metallography and image analysis, aluminum heat treatment, Monte Carlo simulation of fatigue test methods, and material property database strategy. Dr. Crepeau is a registered professional engineer and former chairman of both the AFS Aluminum Division and the TMS Aluminum Committee. Dr. Crepeau was editor of Light Metals 2003.

xiv

Session Chairs/ Lead Reviewers Modeling Session Daan Maijer received his B.A.Sc. and Ph.D. in metals and materials engineering from The University of British Columbia in 1994 and 1999, respectively. He is currently the Director of the Integrated Engineering Program and an Associate Professor in the Department of Materials Engineering at UBC. His undergraduate teaching is focused on engineering design taught through project-based learning where groups of 3 - 5 students propose, design, build, and test multidisciplinary projects. As one of the principal researchers in the Materials Processing Group, his research aims to develop insight into the industrial processes used to transform metals; in particular, casting processes, to improve product quality and process productivity. This research often involves the development of mathematical models that capture the complex physical phenomena active in these processes and relies on laboratory experiments and\or plant trials to provide the data necessary for model development and validation. This research is industrially oriented and has led to collaborations with companies within Canada (Alcan International Ltd., Canadian Autoparts Toyota Inc., and Timminco Ltd.) and abroad (Corus, Titanium Metals Corp. and The Timken Co.). Mark Jolly is a Senior Lecturer and Director of Industrial Liaison in the School of Mechanical Engineering at the University of Birmingham, UK. He has run the Castings Centre at the University since 1995. He runs the Process Modeling Group within the school and has been Principal Investigator on more than 15 funded programs in the last ten years valued at over £5M. He was awarded the University of Birmingham Josiah Mason Award for Business Advancement in 2010 and the Institute of Cast Metals Engineers' Oliver Stubbs award in 2008. Mark is on the Solidification Committee of TMS and a key reader for Met&Mat Trans B and a previous chair of the board of key readers. He also sits of a number of committees for the UKs Institute of Materials, Minerals and Mining namely, Sustainable Development Committee (Vice Chair), Light Metals Board and Materials Science xv

and Technology Division. He is also on the Institute of Cast Metals Engineers Membership Committee and Education and Training Committee. Mark graduated from the University of Sheffield in 1978 with a Bachelor of Metallurgy and continued his studies at Cambridge University to obtain a PhD in 1982. He then started worked in industry for 15 years for a number of companies in the automotive and foundry sectors in the UK and abroad before moving back into academia in 1995. He has over 280 publications including 4 in energy. These include: 2 Patents, 2 book chapters, editor of 2 conference proceedings, over 50 invited seminars & lectures and over 100 technical reports for industry. Mark is a Chartered Engineer, a Chartered Environmentalist, a Fellow of the Institute of Materials, Minerals and Mining and a Fellow of the Institute of Cast Metals Engineers.

Solidification Session W. D. Griffiths is currently Senior Lecturer in the School of Metallurgy and Materials in the University of Birmingham, UK. His research interests involve the study of interfacial characteristics of the metal casting processing process. Research topics to date have included, (i). the prediction of metal-mould interfacial heat transfer coefficients for better modelling of the casting process, (ii). the study of oxide film defects in light alloy castings and (iii). the study of the liquid-metal pattern interface in the Lost Foam casting process. Most recently, research has concentrated on the application of radioactive particle tracking techniques to the study of the behaviour of inclusions during mould filling, (Positron Emission Particle Tracking). To date, over 70 journal and conference papers have been published on these topics. Also, Dr. Griffiths was National President of the Institute of Cast Metals Engineers for 2009-2010, and is currently Chair of the Institute of Cast Metals Engineers Technical Board.

xvi

Peter Schumacher is Managing Director of the Austrian Foundry Research Institute and holds the Chair of Casting Research at the University of Leoben. He obtained his Dipl. Ing. in 1989 at the University of Braunschweig Germany to continue his education in the UK. He received his by Alcan sponsored Ph.D. in 1994 in Metallurgy and Materials Science at the University of Cambridge U.K and held an Advanced EPSRC Research Fellowship at Oxford University from 1997 to 2002. In his 21 year career in metal casting he has co- and chaired numerous conferences and received the Cook Ablett Award U.K. and the TMS Magnesium Technology Award for his work on grain refinement.

Process Session Glenn Byczynski is Manager of Nemak Engineering Centre in Windsor Ontario, Canada. He received his Ph.D. in Metallurgy and Materials Science from the University of Birmingham in U.K. in 2002. His Masters (Materials Science in 1997) and Bachelor's (Mechanical Engineering in 1994) were conducted at the University of Windsor. In his 18 year career in metal casting he has held several R&D and Engineering positions within Nemak and Ford Motor Company including Research and Development Manager for Nemak's European Business Unit, based in Germany and Engineering Manager at Nemak's Windsor Aluminum Plant. He was Chairman of the Detroit-Windsor Chapter of the American Foundry Society (AFS) in 2006-2007, is a director and regional chairman of the Foundry Educational Foundation and is a registered Professional Engineer in the Province of Ontario. He enjoys spending time with his wife and two sons.

xvii

Sergio Felicelli has a Nuclear Engineering degree from Institute Balseiro (Argentina) and a Ph.D. degree in Mechanical Engineering from the University of Arizona. He worked 17 years for the Argentine Atomic Energy Commission, where he was head of the Computational Mechanics Division, and 2V4 years for the Crystals Division of Saint-Gobain High Performance Materials in Northborough, Massachusetts. In August 2004, he joined the faculty of the Department of Mechanical Engineering of Mississippi State University, where he is currently a tenured Endowed Professor and a faculty member of the Center for Advanced Vehicular Systems (CAVS). Dr. Felicelli has worked over 25 years in developing numerical models for applications in solid and fluid mechanics, heat transfer, transport processes, and solidification of alloys. He is the author of some of the pioneer works in computer modeling of freckle segregation during solidification, having written 35 journal articles in the area of macrosegregation and porosity defects in solidification processes and a total of 80 peer-reviewed publications during his career.

Methods & Systems Session Alan Druschitz is the Director VT-FIRE and an Associate Professor in the Department of Materials Science and Engineering at Virginia Tech. He received his PhD in Metallurgical Engineering in 1982 from the Illinois Institute of Technology, Chicago, IL. He was previously a Research Professor at the University of Alabama at Birmingham, a Staff Research Engineer General Motors Research Laboratories and the Corporate Director of Materials R&D Intermet Corporation. He is a co-founder of BAC of VA, LLC, a small company that provides design support and castings for the military and specialty vehicle market. He is an SAE Fellow, Chairman Alabama Section of SAE, a past president of the Ductile Iron Society, former Vice-Chairman of the Governors Board of Transportation Safety for the Commonwealth of Virginia and a current member of AFS, ASM International and AIST.

xvin

Derya Dispinar graduated from Metallurgical and Materials Engineering Department, Istanbul University in 1996. He started working as a Research Assistant at the same department and gained an MSc degree in 1999. He earned a PhD in Metallurgy and Materials, University of Birmingham, UK in 2002 after which he returned to Istanbul University and worked as an Assistant Professor. He has worked as a researcher in SINTEF Materials and Chemistry, Casting Group in Trondheim, Norway between 2007 and 2010 while doing a Post-Doc at Norwegian University of Science and Technology (NTNU). Since October 2010, he has been working as a Senior Researcher in Aluminium Group at TUBITAK (The Scientific and Technological Research Centre of Turkey).

xix

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry

Modeling Session Chairs: Mark Jolly Daan Maijer

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

The History of Casting Process Simulation Christof Heisser MAGMA Foundry Technologies, Inc. 10 N. Martingale Road, Suite 425, Schaumburg, Il 60173 Erwin Flender and Jörg C. Sturm MAGMA Giessereitechnologie GmbH Kackertstr. 11, 52072 Aachen, Germany Abstract Not many developments in recent decades have changed the understanding of the metalcasting process as fundamentally as casting process simulation has. The main intention of this paper is to provide an easy to read and attractive overview for foundrymen addressing the development, current state, and future of casting process simulation. Keywords: simulation, filling, solidification, modeling, autonomous optimization, thermo-physical properties, microstructure, heat treatment, stress, distortion Introduction The description of the metal casting process in a physical-mathematical model and its simulation in a computer demanded the quantification of process parameters and process steps as they impact the casting quality. The idea of utilizing numerical models to predict the filling and solidification of castings came from physicists, mathematicians, and mechanical engineers. Today, casting process simulation is utilized to develop a technical knowledge database, as a management tool to provide training and education to foundry personnel, and to facilitate communications both within a corporation as well as with customers. The History The theoretical fundamentals of heat conduction in solid matter were developed by Jean Baptiste Jospeh Fourier at the Ecole Polytechnique in Paris. His thesis, "The Analytical Theory of Heat", received awards in 1822. It has provided the basis for all following calculations of heat conduction and transfer in solid materials. The French physicist and engineer Claude-Louis Navier, and the Irish mathematician and physicist George Gabriel Stokes, subsequently provided the basics of flow dynamics. The differential equations describing fluid flow are now known as the Navier-Stokes equations. The basic equations describing diffusion were developed by Adolf Fick, who worked during the 19th century at the University of Zurich and published them in 1855. In the 1950s, Paschkis used analog computers to predict the movement of a solidification front in one or two dimensions. With the development of the first digital computers, Fursund was the first who used computers to solve casting process related problems (penetration of steel into mold sand), in 1962. Three years later, Hentzel and Keverian

3

published their ground-breaking work about two dimensional simulation of steel casting solidification. They utilized a program developed by General Electric to simulate heat transfer. In 1968, Vestby developed a 2-D model to evaluate temperature distributions during welding, using, for the first time, the finite difference method. Two years later, V. de Lange Davies used Vestby's program to simulate feeding distances in plate-like castings. P. N. Hansen published his thesis describing his work to predict hot tears in steel castings (Figure 1) in 1975. In the preparation of this thesis, a 3-D model was programmed for the first time. Starting at the beginning of the 1980s, the research and development activities around the topic of casting process simulation increased substantially in multiple locations. In addition to the activities at the Technical University of Denmark around Hansen (Figure 2), work groups were established world-wide, including Berry and Pelke in the United States, Niyama in Japan, and Kurz in Lausanne, Durand in Grenoble and most notably, Sahm in Aachen at the Foundry Institute (Figure 3). Important milestones were the introduction of the term, "criteria function" by Hansen and Berry (1980), the introduction of a criteria function to depict centerline porosities by Niyama (1982), as well as the proposal of a criteria function to detect hot tears in steel castings by Flender and Hansen (1984). By the end of the 1980s, the first solutions to simulate the mold filling were provided. In the 1990s, development activities focused on the simulation of stresses and distortions in castings (Hattel and Hansen, 1990), as well as the first steps were taken to predict microstructures and mechanical properties by Svensson and Wessen in Sweden.

Figure 1: The first results of a temperature distribution. (Hansen, 1975)

„.

„ „ , f r<; 2= ^ p l a y ° f a l o w , P ™ wheel). (Sahm and Hansen, 1984)

F

dle cast

The Methods Numerical simulation is the process of solving a physical model through mathematical (differential) equations and the display of the calculated domain (the casting and the mold) through discrete single elements. In order to calculate the differential equations, several methods were developed (FEM, FDM, FVM, BM, MM, etc.), which will not be discussed in detail here. In 1924, Schmidt developed a graphical method to solve 1-D heat conduction problems. In 1949 and 1959, important contributions regarding the analytical solution of heat transfer problems were provided by Ingersol, Zobel and Ingersoll, as well as by Carslaw and Jaeger.

4

The finite element method (FEM) was developed in 1945, to solve special load calculations. In 1956, the first structural simulations were conducted on airplane wings at Boeing. In 1967, the reference book, "The Finite Element Method", was published by Zienkiewicz. Hansen performed 2-D- and 3-D solidification calculations for the first time in 1975 utilizing the finite volume method (FVM). Each method has specific benefits and drawbacks and can yield good qualitative results depending on its area of application. The finite element methods have their roots in load simulations. The finite difference and finite volume methods come from the fluid flow simulation and show benefits in the description of heat and material transport phenomena.

Figure 3: A steel casting hammer was simulated by the CASTS program developed by the foundry institute of the RWTH Aachen (Sturm, Schäfer and Sahm, 1988)

The choice of which numerical method and mesh is used is driven by finding the best compromise between the quality (accuracy) of the calculation, optional automatic enmeshment and calculation time. The first steps of describing the process in virtual terms were taken by focusing on heat transfer calculations and focused on the solidification process. The mold filling is an integral part of the process and therefore must be considered. This is not only important for the gating layout but for the detection of filling related defects as well. Indeed, the inhomogeneous temperature distribution in the melt caused by the filling process has in many cases an impact on the solidification process (Figure 4). Even today, the dynamics of the mold filling process are often underestimated by practitioners. Key words like "quieescent filling" and "laminar flow" are frequently used, but from a physical point of view, all filling processes from sand castings to high pressure die castings are highly turbulent. This fact is based on the rheological properties of metal melts. The energy that is created by the flowing melt is so high that it cannot be eliminated through foundry technological efforts. Therefore, strong turbulences and eddy currents are found inside the melt even when the melt surface appears to be rising quietly (Figure 5). Many casting defects result from these under-surface movements, as well as reactions between melt and mold material. These defects include mold defects, air entrapments, oxidation defects, slag entrainments or metallurgical challenges (Figure 6).

5

Figure 4: Prediction of fluid flow combination with the temperature loss. / / /

in Figure 5: The effects of turbulence can be depicted in simulation tools by using virtual particles. /2/

Figure 6: Current fluid flow simulation can predict the creation of oxides and inclusions. (Carlson, and Beckermann, 2005) As soon as a 3-dimensional geometry of a casting is available, a basic solidification and cooling simulation can be performed in minutes (Figure 7). The prediction of hot spots and areas of final solidification do not only help the metal caster in the engineering department, but also support the designer in evaluating the designs. The knowledge of temperatures and solidification behavior leads to a quantitative prediction of the local thermal Modulus in the casting, as well as solidification times, cooling rates, temperature gradients and shrinkage defects (Figure 8).

6

The Core Question in Engineering Feeding and Shrinkage Defects Prediction of feeding related problems is still one of the most important uses of casting simualtion software Depending on the alloy poured, different feeding behaviors and self-feeding capabilities need to be considered to provide a defect free casting. Solidification simulation has to be combined with density and mass transport calculations in order to evaluate the impact of the solidification morphology on the feeding behavior, as well as to consider alloy dependent feeding ranges.

Figure 7: The solidification and cooling process of complex castings can now be predicted within minutes. /3/

Figure 8: Examples displaying the accuracy of shrinkage prediction in gray and ductile iron castings. The Multitude of Materials Even if the fundamental physics for filling, solidification, stress development, and cooling process are the same for all alloys, the specific material behavior makes a difference, as displayed in Figure 9 for aluminum alloys. As well as the process conditions, the nominal composition and metallurgical parameters (grain refinement) are defined. Based on this information, the program calculates the potential equilibrium phases, which are impacted by the accelerated cooling conditions they experience (phase kinetics). The inhomogeneous solubility of alloying elements in the solid and the liquid phase leads to segregations and thereby to the potential creation of new, and sometimes undesired, phases. Only in the final step, based on this information, the solidification progress and the resulting temperature distributions are calculated in a time step. These steps are repeated for every location and every point in time before microstructures and mechanical properties are predicted. The key to the development of material-specific simulation models was the specific feeding behavior of cast alloys and their strong dependence on the chosen metallurgy. A calculation of the feeding behavior based solely on temperature distributions was not

7

sufficient. For example, large hotspots in iron castings can potentially completely feed themselves, but small hot spots can lead to shrinkage defects. The local shrinking and expansion behavior of a casting can only be calculated under the consideration of the locally developing phases (graphite, austenite, cementite) and their respective contribution to the local shrinking and expansion behavior. The nucleation and growth kinetics of each phase is therefore considered throughout the entire progression of the solidification. This means that for cast iron not only is the dominant impact of the alloying elements considered, but also the inoculation and melt quality. Metalcasters use the impact of the inoculation or alloying elements for the creation or avoidance of white iron. These are overlaid by the local cooling conditions inside a casting. A simulation solely of the macroscopic solidification and cooling behavior cannot describe this interaction. Therefore, this so-called micromodeling is performed on many materials, considering the amount of any new phase created at any time based on the phenomena described above (Figure 10).

Figure 9: Overview of input parameters, calculation steps, and results for the prediction of microstructures in aluminum alloys.

Figure 10: Differences between macroscopic and microscopic simulation (micromodeling) on simulated cooling curves.

Stresses and Distortion The developments regarding the prediction of hot tears, as well as the creation of residual stresses and distortion behavior created much more transparency (Figure 11). Additionally, the stress simulation has to consider not only the part itself, but also the impact of mold and cores, as they are often predominantly responsible for stress-related defects (Figure 12). In addition to residual stresses, hot tears, crack formation and the shrinkage and warpage of the castings, dies and permanent molds are moving into focus. Due to the high costs for permanent molds and dies, maintenance and repair efforts are very often the deciding factor if a process is profitable. As the temperature behavior and the resulting stress development can easily be simulated, this application of simulation provides additional value and cost reduction potential (Figure 13).

Figure 11 : Distortion of a structural part in a diecasting die. /4/

Figure 12: Prediction of potential cracks after machining. / 5 /

Figure 13: Die-life prediction,

Simulated Properties The final goal of casting process simulation is the prediction of casting properties. The base for such predictions is the quantitative information about local microstructures and potential casting defects. Due to its specific solidification behavior, cast iron was the pioneer in this area. The local microstructure and mechanical properties, including hardness, can today be simulated for all common graphite morphologies and compositions in a quantitative manner (Figure 14). Similar developments are meanwhile available or will be available shortly for aluminum alloys.

Figure 14: Prediction of locale hardness (HB) for a gray iron engine block. /6/ From Casting to Mold - Sand Simulation The core making process is, from a physical point of view, a multi-phase fluid problem. After opening the valve, sand is engulfed by air and thereby accelerated. Air transports the sand into the core box. At the end of the filling process, the air needs to be separated (vented) from the sand. Recently steps have been taken into this new complex world with its demanding physics. One challenge is the proper description of those physics. A bigger challenge is the multitude of internal and external boundary conditions (how many different valves and vents are there in the market?). Initial successes have motivated continuing development of simulations for the core room (Figure 15). The final goal of the current development is the simulation of the entire process chain, with the

9

consideration of shooting, gassing, venting, and degradation of binders during the casting process.

Figure 15: Simulation of coremaking and comparison to experiments. /7/ Optimization Casting process simulation always displays the status quo of its expert user. The user decides if the rigging system or process parameter set led to an acceptable result. Additionally, proposals for optimized solutions have to come from the operator. Multi-objective autonomous optimization (Figure 16) uses the simulation tool as virtual experimentation field and changes pouring conditions, gating designs or process parameters and tries this way to find the optimal route to fulfill the desired objective (Figure 17). Several parameters can be changed and evaluated independently from each other. Autonomous optimization tools take the classic approach of foundry engineers, to find the best compromise, and use validated physics.

_. , Λ . . Figure 16: Optimization Principle.

Figure 17: Optimization process of riser optimization for a steel casting.

10

Summary Considering the long tradition of casting, the history of casting process simulation is a small cultural revolution within the industry. The 5,000 year-old art of casting through trial and error is transformed into a transparent, reproducible process where process parameters can be not only predicted, but also manipulated. Still, not all questions can be answered through simulation, but the current acceptance of this tool in foundries and at casting users confirms that casting process simulation is one of the key innovations of the last thirty years in the foundry world. Simulation unlocks an enormous potential for cost reduction and leads thereby to an increased competitiveness of foundries. Figure Sources III with kind permission of DANA Spicer, Argentina 111 with kind permission of Voith Paper, Brazil 13/ with kind permission of vonRoll Casting, Switzerland IAl with kind permission of Volkswagen AG, Kassel, Germany 151 with kind permission of Coupe Foundries, Great Britain 16/ with kind permission of Ford Forschungszentrum, Aachen and Eisenwerk Brühl, Germany ΠΙ with kind permission of BMW, Landhut, Germany /8/ with kind permission of Ford Forschungszentrum, Aachen, Germany References 1. P. N. Hansen, P. R. Sahm,"A 3-D Geometric Modeler-Implicxit FDM Solver Package for Simulation of Shaped Casting Solidification","Modelling of Casting and Welding Process II", Metall. Soc. AIME (1984), New England College Henniker, New Hampshire, p. 243-247, (1983) 2. P.R. Sahm und P.N. Hansen, Numerical Simulation and Modelling of Casting and Solidification Processes for Foundry and Cast-House, CIATF (1984), 3. E. Flender, P. N. Hansen, Peter R. Sahm,"Rechnerisches Simulieren und Modellieren des Warmrißverhaltens warmfester Stahlgußsorten bei der Erstarrung",Giessereiforschung 39, Heft 4, S.137-149,1987 4. J. C. Sturm, W. Schäfer und P.R. Sahm, Modelling the mold filling and solidification of a steel hammer Casting by use of the Computer Aided Solidification Technologies (CASTS) Software system, Modeling and Control of Casting and Welding Processes IV, S. 845 , Herausgeber A.F. Giamei and G.J. Abbaschian, verlegt bei TMS (1988), 5. Egner-Walter:„Berechnung der Entstehung von Spannungen beim Gießen" Hoppenstedt, Gussprodukte '99 (1999) 6. I.L Svensson und M. Wessen: " Foundry Of Cast Irons : Processing and Simulation ", Numerical Simulation of Foundry Processes, S. 87-145,.(2001) 7. G. Hartmann und R. Seefeldt, "Die zweite Generation von Simulationswerkzeugen: Praktische Anwendung der rechnerischen Optimierung im Druckguss", Giesserei, Nr.2/2004, S. 38-42 (2004)

11

8. J. C. Sturm, " "Stand der Simulation für Gusseisen" ", Giesserei 91, Nr. 6, Special Simulation von Giessereiprozessen S. 4, (2004) 9. G. Hartmann, A. Egner-Walter, H. Dannbauer, Simulation of Local Properties of Metal Cast Engine and Suspension Parts, Virtual Product Creation 2004, Konferenz (2004) 10. J. C. Sturm, „Vorhersage lokaler Eigenschaften von Gussteilen im Motorenbau", VDI Fachtagung Gießen im Motorenbau, Magdeburg, (2005) ll.J. Hattel (Herausgeber), Fundamentals of Numerical Modelling of Casting Processes, Polyteknisk Forlag (2005) 12. K.D. Carlson, and C. Beckermann, "Modeling of Reoxidation Inclusion Formation During Filling of Steel Castings," in Defect Formation, Detection, and Elimination During Casting, Welding, and Solidification (Proceedings of a Symposium Sponsored by Materials Science & Technology 2005), eds. M.L.C. Clemens et al., TMS, Warrendale, PA, pp. 35-46. (2005) 13. C. Midea, M. Burns, M. Schneider I.Wagner ,„Advanced thermo-physical data for casting process simulation -the importance of accurate sleeve properties", „Foundry Research/Giessereiforschung" Volume 59, No. 1 page 34-43 (2007) 14. R.J. Menne et al.: Implementation of Casting Simulation for Increased Engine Performance and Reduced Development Time and Costs - Selected Examples from FORD R&D Engine Projects, Wiener Motoren Symposium (2007) 15. Pawlowski, R. Seefeldt, J.C. Sturm "Aus Eins mach Zwei", Sichere Übertragung einer bewährten Füllcharakteristik von einem Einfach- auf ein Zweifachdruckgusswerkzeug, GIESSEREI 94, Nr. 4, S. 34-42, (2007) 16. M. Schneider et al. Experimentation, Physical Modeling and Simulation of Core Production Processes, AFS Transactions, (2008)

12

Shape Casting: The 4'" International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011 STATE OF THE ART REVIEW OF MODELLING ENTRAINMENT DEFECTS IN THE SHAPE CASTING PROCESS C Reilly1·3, N.R Green2 and M.R. Jolly1 'School of Mechanical Engineering, University of Birmingham, UK School of Metallurgy and Materials, University of Birmingham, UK 3 School of Materials Engineering, University of British Columbia, Canada 2

Keywords: Modelling, Casting, Defects, Entrainment, Optimisation Abstract The entrainment of oxide films and bubbles into the bulk liquid material has been shown to have a detrimental effect on casting integrity when solid. A number of mechanisms have been shown to initiate the entrainment of oxide films, including: returning waves, plunging jets, bubble trails and fountains. The use of computational fluid dynamics (CFD) software packages, which are now widely available to the foundry engineer, has allowed him/her to improve casting system design by using qualitative parameters. Optimisation software is now an economically viable option for many foundries. However, optimisation for casting integrity requires a quantitative casting integrity assessment technique which allow the modelling and quantification of defects. Therefore, modelling and quantification of defects is becoming an ever more important research area to allow the optimisation software manufacturers to meet the needs of industry. The current methods of modelling surface film and bubble generated casting defects have been described and critically reviewed shedding light on the qualities and issues currently associated with the present available methods. It is clear that further investigations and development is still required to allow the accurate and efficient modelling of casting defects. Introduction With competition within the foundry industry becoming fiercer and customers demanding higher quality components, shorter development times and more complex geometry, the use of computational simulation has become essential to stay competitive. In more recent times the economic viability and increased ease of use has encouraged many larger foundries to use computational optimisation software. The modelling of defects is essential to allow the optimisation of casting systems for component integrity. Optimisation can only occur if "the right optimisation criteria to formulate the objective functions are available" [1]. Therefore to optimise a casting system for casting integrity knowledge of defect location and quantity is required. This is the challenge facing modellers. As these optimisation software become even more user friendly and computer hardware more powerful the requirement for quantitative defect assessment criteria will become more acute. The entrainment of oxide films into the bulk fluid has been shown to have detrimental effects on cast component integrity [2]. When the surface oxide film is entrained into the bulk fluid it acts as a crack initiation site upon solidification. This is because the film does not bond correctly within the metallic structure and is therefore a potential crack initiation site. 13

Campbell's 2006 paper[3] summarised most of the methods researched for the modelling of defect entrainment in castings thus far developed. Recent work has both proposed new methods and further developed, and assessed, previously proposed methods of modelling defects. This has given further insight onto this important topic. An overview of the currently available methods for modelling filling related (oxide film) defects both quantitatively and qualitatively are discussed below. Review of Modelling Entrainment A variety of approaches to the modelling of entrainment defects have been investigated. These can be largely split into two groups, those modelling discrete defects and indiscrete modelling of entrainment. The available models are briefly described below, a thorough description and discussion is outside the scope of this paper. Indiscrete Modelling of Entrainment Cumulative Entrained Free Surface Area - Work by both Lai et al[4] and Sun et al[5] investigated using the free surface area to describe the magnitude of entrainment. Little is known of the work by Sun et al. due to commercial sensitivities, although the authors reported positive results using this technique. The work by Lai et al. [4] takes the instantaneous free surface area and plots it against time. This is then compared to the proposed instantaneous free surface area assuming the mould had been filled in a tranquil manner, to allow the excess of free surface area to be calculated. This technique, although easily understood and requiring minimal computational power has one major drawback, namely; how to define the minimum free surface area should the mould fill quiescently. For very simple geometries comparison between differing geometries is possible, however time consuming. For complex geometries however, this could prove near impossible. Therefore this technique is unsuitable to use for optimisation except for instances where direct comparison can be made between two or more models (i.e. for models of identical geometry). This technique gives no distribution of defects but is felt to be nevertheless highly informative. The lack of ability to track the motion of the entrained defects also proved detrimental to the usefulness of this technique. Vorticitv - MAGMAsoft, Flow-3D and CFD post processing software such as FieldView, CEI and Tecplot have developed techniques to identify and assess vortices within the bulk fluid during flow simulation of mould filling. These analysis tools allow the vortex core location and axis and vortex magnitude to be defined. The problem arises however in filtering of the data. The bulk fluid flow in the casting scenario is usually in a highly turbulent regime, producing many vortices, filtering the data to only show those relevant to free surface entrainment can be highly problematic. The authors are currently unaware of any work which has been undertaken directly relating vortex assessment using a computational model to defect entrainment or casting integrity. Cumulative Scalar Technique - A cumulative near surface scalar technique has been developed by a number of the commercial casting software manufacturers [6, 7]. The technique works by assuming that oxide defects accumulate upon the fluids free surface at a constant rate, this oxide accumulation is described by a scalar parameter. This scalar once entrained into the bulk fluid at the free surface is allowed to gradually diffuse throughout the fluid and advect with the flow of the bulk fluid. This allows a final defect probability to be obtained. This is a simple and robust approach which neglects the physics involved in bi-film 14

entrainment. However the approach was seen to yield results in accord with more sophisticated models and experimental data. An almost identical technique is also utilised in smoothed particle hydrodynamics (SPH) [8], As stated by the Barkhudarov and Hirt[9], the cumulative scalar technique does have some drawbacks in the casting scenario, namely; adhesion of oxide film to mould walls is not accounted for, no oxide film strength is modelled, no buoyancy of oxide film is modelled and without any experimental results the significance of the absolute values of the scalar are meaningless. However the defect location patterns are still valid. MAGMAsoft Air Entrainment Model - An air entrainment model has been developed by MAGMAsoft as a mechanism to track small air bubbles transported by the bulk flow. The model is made of two main constituents; a venting model and an air entrapment model. The criteria MAGMAsoft use to define the quantity and threshold of air entrainment at the free surface into the bulk fluid is proprietary. The main mechanism for tracking air pockets is the venting model, this tracks changes in topology of air pockets and calculates their thermodynamic parameters. Using this, the number of discrete air pockets and each pockets location is known, along with their density, volume, mass, temperature and pressure. Air pocket can collide with other air pockets and can merge or split. The venting (permanent moulds/dies) or permeability (consumable moulds) of the mould is modelled to allow accurate modelling of vented regions. The venting model can operate only on air pockets that are resolved by at least several mesh elements. The air entrapment is a model that enables tracking air pockets that are too small to be tracked by the venting model. Air entrapment operates only on the air volume transporting it with the bulk melt velocity field. The model is valid for small bubbles. The air entrapment models give the user a contour map of air distribution within the melt volume. Flow-3D Air Entrainment Model - Flow-3D have developed an algorithm to model the turbulent entrainment of air at a free surface [10]. The model works by assessing whether the turbulent energy at the free surface is enough to overcome the restraining effects of the surface tension and gravity. If the magnitude of surface turbulence is able to overcome these restraining effects then a series of equations are used to calculate a quantity of entrained air. This air is then entrained into the fluid and allowed to advect, dissipate and escape at the free surface. The bulking of the fluid with the volume of entrained air can be modelled. The model has been validated on data collected by researchers in the hydraulic engineering fields. Dimensionless Number Criteria - The use of dimensionless numbers has been previously proposed for use in assessment of defect entrainment by Campbell [11] among others. Previous studies utilising the Weber Number (We) [12] (ratio of surface tension to inertial pressure) and Froude Number (Fr) [12] (ratio of gravitational pressure to inertial pressure) include that of Cuesta et al. [13] and Isawa [14] respectively. Reilly et al. [15] have used dimensionless numbers to create criterion functions with which to interrogate computational models for quantification of entrainment. The Froude number, Weber number and Hsu number[16] (ratio of inertial to gravitational and surface tension forces) were used for the assessment of returning wave forms in horizontal runner bars. This work attempted to validate computational models against quantitative experimental data. The results showed that the Froude number does have the ability to quantify entrainment. However the difficulty arises when trying to define the entrainment threshold and quantify entrainment magnitude for a given magnitude of the dimensionless parameter. The Weber and Hsu numbers did not

15

correlate with the experimental data in the cases investigated. Thus the use of dimensionless numbers for optimisation is currently limited at this current state of development. Multi-Phase Modelling - The accurate modelling of bubbles; their entrainment, advection and coalescence is an important element of the modelling of casting entrainment. Bubble trails are highly detrimental to casting integrity. By modelling both the bulk fluid and surrounding gas (two phase modelling) it has been possible to describe the entrapment, advection a coalescence of bubbles within the melt[17]. Initial results show correlation of bubble motion, coalescence and separation with experimental data. These software are, as expected, computationally highly intensive when compared to single phase flow modelling due to the substantial additional complexities of modelling the second phase. However, modelling both the liquid and gas phase appears to be the only way to correctly model highly aerated flows. It appears that currently the developers of two-phase-focused software are concentrating on developing the flow modelling rather than the addition of models for the quantitative modelling of casting defects. At the current time the authors are unaware of any two-phasefocused software incorporating quantitative defect prediction models. However, this does not mean that they have not been successfully validated and applied qualitatively in the application of defect prediction and process optimisation[18]. Modelling of Discrete Defects Methods have been developed to model the entrainment and advection of discrete defects. This is obviously very challenging and usually requires greater computational expense than the indiscrete methods described previously. However there are some considerable advantages associated with this approach, namely; entrainment mechanisms can often be identified and the final defect location can be obtained. However, many of the techniques described below have had to make assumptions about; both physical characteristics of the defects and their behaviour, critical entrainment thresholds and interaction of defects with both mould materials and each other. It is often not the practical modelling, but determining exactly what mechanism or physical situation to model which is the greatest challenge facing modellers. For this reason, modellers will have to work closely with experimentalists for effective progress to be made within this field. Many of the following techniques used to discretely model film entrainment utilise particles to represent entrained defects. This comes with some currently inherent issues, often caused by not having understanding of the physical behaviours of oxide films in the real world. Many of the particle models within the software have had no experimental validation. It is only current work by Griffith et al. [19] which have allowed the possibility of validating a simulation software's particle tracking model. Quantification of the amount of oxide film entrained for a given entraining event is not understood. It is incorrect to assume that the number of particles directly correlates with the number of oxide films which would be created experimentally from the same flow phenomena. Therefore it is not possible to categorically state that the number of particles entrained correlates with the area of oxide film entrained and thus damage to the material. However, the greater the numbers of defects present, the greater the probability of a highly damaging defect which initiates failure being present. Oxide films are individually unique, varying in size, shape and density. However, many models are not capable of modelling this. The particles are commonly specified as spheres of either constant density and varying size, or varying density and constant size.

16

Bubble Entrainment - A major omission from many casting software is the ability to model correctly the entrainment of air into the bulk fluid during casting. Defects caused by the entrainment of air into the bulk fluid include bubble trails, splash defects and entrapped bubbles. Bubble trails are hollow cracks (tubes) which create leak paths through the casting[ll]. The direct simulation of the bubble entrapment and subsequent effects of the bubble have currently been avoided due to the computational expense of the meshing requirements which create runtimes deemed too long for industrial use. Work by Ohnaka et al. has used particles to represent and track entrained bubbles for the prediction of porosity[20, 21]. This technique has been developed to remove the need to use extremely small mesh element sizes to track small bubbles. To allow the tracking of these small bubbles Ohnaka et al. place particles when the void region becomes too small to be defined by the mesh. These particles are then tracked and their final locations defined. These are then used to define heterogeneous nucleation of gas porosity[20, 21]. This technique is an adaptation ofthat developed previously by Tomivama et al. [22]. This allows the tracking of the bubbles without the computational expense of small mesh cell sizes. They found the technique gave results which correlated well with experimental results. However the technique was found to be extremely sensitive to the particles buoyancy force. Modelling Of Oxide Film Deformation - Work by Pita et al. [23] modelled the transport and deformation of a single oxide film within a fluid volume. The technique has shown that it is possible to accurately model large scale deformation of a solid film within a fluid, and the effect this solid deformation has upon the fluid motion: i.e. coupling of a fluid and deformable thin film. The work therefore shows promise for the applications in modelling the unfurling of oxide films in castings, which is believed to be one of the mechanisms of porosity formation[ll]. Pita et al. state that they aim to further develop this technique by simulating a more realistic model and including a breakage criterion for the film, and solidification elements (phase change and solute transport). With the present state of computational hardware it seems unlikely that this technique can be applied to large scale castings in the near future due to the computational intensity of modelling numerous films (which require the micro flow to be simulated) alongside the macro flow and solidification. However, the technique may play an important role in gaining insight into the physical behaviour of oxide films, of which surprisingly little is definitively known. Modelling Of Oxides In Steels - A method based upon the formation of oxides from nuclei was used as a methodology to model defects in steels. A team from Iowa University lead by Beckermann[24] introduced particles into the melt and allowed them to grow when upon the free surface. The particles were either added to the incoming fluid stream, or placed upon the free surface so as to give a minimum free surface particle density. When the particles are sub-surface, only their advection is modelled, their growth is not permitted. The particle motion is tracked until solidification ends and agglomeration of colliding particles (defects) is permitted. The final location of these particles and their size give probabilistic representation of the likelihood of entrainment defects being present. Carlson and Beckermann have further developed and undertaken validation work of this steel inclusion modelling technique showing good correlation with an number of experimental validations and has been successfully used in industrial applications[25]. They have created a very elegant and seemingly robust method of modelling oxide inclusions in steels, and shown it to give reliable results in industrial applications.

17

Modelling The Folding Mechanism - This method, used by Lin et al and Dai et al. [26, 27], models the entrainment of bi-films through the folding of the free surface. It is therefore only able to model certain entraining phenomena such as returning waves and folding surfaces. The methodology used by both Lin and Dai is based upon placing particles on a fluid's free surface to represent the oxide film. Particles are added if the surface is expanding, and when a particle is added to the model then all particles are re-labelled. The particles have to be replaced onto the fluid's surface at every time step, should the free surface form have changed. This suggests that the technique may be computationally intensive. These techniques have also only currently been applied in two dimensions, expansion into three dimensions would severely complicate the programming required and further increase the computational effort. Lin assumes the oxide films to be present between neighbouring particles and calculates the strain the film is under by tracking the movement of neighbouring particles. Should the strain exceed the strength of the film, further particles are added as the film is assumed to have torn and immediately new oxide film has been formed. The model is able to assess the entrainment of air bubbles into the bulk liquid by surface turbulence. Once a film is entrained in the bulk material the tracking points are no longer adjusted to fit the free surface and it is assumed that there is no atmosphere for oxidisation within the bulk material. Therefore no new particles are added should the film break due to excessive stress. The location and number of these entrained films are tracked. Dai's approach varies slightly by assessing the surface normals of the films. Should they be pointing towards each other and their velocity vectors obey a predetermined mathematical rule (meaning that the films would converge) then entrainment is deemed to occur. This model was then compared to experimental data and deemed to be qualitatively consistent with the modelled results. One major drawback with this model is that it is currently only implemented on the OFET 2D CFD (computational fluid dynamics) software. The authors are unaware of any shape casting simulations undertaken with this software, presumably due to its current 2D limitation. Further investigations and development are required to exploit its full potential. Modelling Of Oxide Entrainment - The work by Ohnaka et al. on modelling bubbles in single phase flow[20, 21] was further extended to include the modelling of oxide entrainment in aluminium castings[28]. Making the assumption that the aluminium surface which is exposed to the atmosphere instantly forms an oxide film, the free surfaces are then assessed using defined physical rules[28] to see if they collide, thus entraining oxide films. If entrainment occurs, marker particles are placed to represent the entrained films. Their advection within the flow is then calculated to define their final locations upon solidification. The number of entrained oxides per unit area is estimated as a function of collision velocity and alloy composition the parameters of which were defined through unpublished experimental work. The average surface area of the broken oxides is estimated using a function of the collision surface area, the surface area of a broken oxide and the number of entrapped oxides. The function means that at larger collision velocities, more but smaller oxide films are entrained. The judgment of a free surface collision is classified by assessment of the velocity vectors, distances between particles and the time period. This methodology allows the entrainment caused by fluid jets, bubbles, colliding fronts, impinging flows and return waves to be modelled.

18

Research by Reilly et al. [29] used a very similar methodology to that developed by Ohnaka et al, however the implementation varied due to factors associated with the differences in software used. The model uses Boolean logic criteria to define entrainment. Once an entraining event has been detected a particle of determined size and density is placed to represent the defect. The particles motion is then modelled. Upon solidification the particles within defined critical volume(s) or the whole casting can be counted to allow quantitative analysis of the casting system. For validation the number of particles present in the gauge length of tensile test samples were compared to the Weibull modulus of the experimental data of Green and Campbell[2]. The modelled results agreed with those found experimentally. However, further investigations to give a much larger data set and using a variety of running system designs which emphasise different entraining flow phenomena are required for conclusive validation of the technique. Summary The modelling and quantification of defect entrainment in the casting scenario is in its infancy and is an extremely difficult proposition due to a number of complex problems which have to be addressed. One of the most difficult of these is not the actual modelling of the defect but instead acquiring the knowledge of what to model. For example; do oxide films agglomerate if they collide, do oxide films stick to the mould surfaces and or under what conditions, what are the characteristics of oxide films created through different entrainment mechanisms and how do oxide characteristics affect the motion of defects within the melt? These problems require experimentalists to work alongside modellers to make further progress in the modelling of entrainment defects in castings. Conclusions 1. The topic of modelling entrainment defects in casting has received little attention in recent times despite its obvious commercial significance. 2. There are numerous models available to the foundry engineer to quantify entrainment. However, understanding of the techniques limitations and applicability are required for correct implementation. 3. The development of quantitative defect modelling techniques is difficult and complex, but also of great industrial significance, and therefore further research is urgently required. References 1. Kokot, V. and P. Burnbeck, "What is a good gating system? or Quantifying quality- but how?", Modelling of casting, welding and advanced solidification process XI, (2006), 119126. 2. Green, N.R. and J. Campbell, "Influence in oxide film filling defects on the strength of Al-7Si-Mg Alloy Castings", Transactions of the American foundry society, 114, (1994X341 -347. 3. Campbell, J., "The Modeling of entrainment defects during casting", TMS Annual Meeting, v 2006, Simulation of Aluminum Shape Casting Processing: From Alloy Design to Mechanical Properties, (2006), 123-132. 4. Lai, N.W., W.D. Griffiths, and J. Campbell, "Modelling of the potential for oxide film entrainment in light metal alloy castings", Modelling of casting, welding and advanced solidification process X., (2003),415-422. 5. Sun, W., et al., "Modeling, model verification, and defect formation in ductile iron castings", Ductile iron society 2003 Millis symposium 11, (2003),

19

6. Barkdudarov, M.R. and C.W. Hirt, "Tracking Defects", www.flow3d.com/pdfs/tp/castJp/FloSci-Bib9-98.pdf. (1998), 7. MAGMASOFT, V4.4 Manual. 8. Prakash, M., et al., Preliminary SPH modeling of oxide formation during the mouldfilling phase in DC casting of extrusion billets, in Fifth international conference on CFD in the minerals and process industries. 2006: Melbourne, Australia. 9. Barkhudarov, M.R. and C.W. Hirt, "Tracking Defects", 1st international Aluminium casting technology symposium, (1998), 10. Hirt, C.W., "Modeling turbulent entrainment of air at a free surface", Flow Science Technical Note, FS1-03-TN61, (2003), 11. Campbell, J., Castings 2nd Edition,(Oxford, Butterworth Heinemann, 2003), 12. Massey, B.S., Mechanics of fluids 6th edition. 6th ed,(London, Chapman & Hall, 1992), 13. Cuesta, R., et al., "Numerically modelling oxide entrainment in the filling of castings: the effect of the Webber Number", Journal of Materials, 58, (2006),62-65. 14. Isawa, T., The control of the initial fall of liquid metal in gravity filled casting systems, in Department of Metallurgy and Materials. 1994, The University Of Birmingham: Birmingham. 15. Reilly, C, et al., "Using criterion functions to quantify entrainment in castings", (2010), 16. Hsu, F.-Y., Further Developments of Running Systems for Aluminium Castings. 2003, The University of Birmingham. 17.Jakumeit, J., K. Goodheart, and M. Albers, "Influence of gas phase on mould filling for sand casting", Modelling of casting, welding and advanced solidification process XII., 12, (2009),427-434. 18. Wang, H., et al., "Modelling of the tilt casting process for the tranquil filling of titanium alloy turbine blades", Modelling of casting, welding and advanced solidification process XII, (2009),53-60. 19. Griffiths, W.D., et al., "The application of positron emission particle tracking (PEPT) to study the movements of shape castings ", Shape casting: 3rd international symposium, (2009),231-238. 20. Kimatsuka, A., et al., "Mold filling simulation for prediction gas porosity", Modelling of casting, welding and advanced solidification process XI, (2006),603-610. 21.0hnaka, I., et al., Porosity formation mechanism in Al and Mg alloy castings and its direct simulation, in Melting of casting and solidification processes VI (6th pacific rim conference). 2004. 22.Tomiyama, A., et al., "A three-dimensional particle tracking method for bubbly flow simulation", Nuclear Engineering and Design, 175, (1997),77-86. 23. Pita, CM. and S.D. Felicelli, "a fluid -structure interaction method for highly deformable solids", Computers and structures, 88, (2010), 24. Blair, M., et al., "Predicting the Occurrence and Effects of Defects in Castings", Journal of Materials, 57, (2005),29-34. 25. Carlson K C and C. Beckermann, "Modeling of reoxidation inclusion formation during filling of steel castings", 58th technical and operational conference, steel foundry society of america, (2004), 26. Lin, J.T., M.R.A. Sharif, and J.L. Hill, "Numerical simulation of the movement, breakup and entrapment of oxide films during aluminum casting", Aluminum transactions, 1, (1999),71-78. 27. Yang, X., et al, "Numerical modelling of entrainment of oxide film defects in filling aluminium alloy castings", International journal of Cast Metals Research, 17, (2004),321-331. 28. Sato, Y., et al., Modeling of oxides entrapment during mouldfillingof Al-alloy castings, in 7 th Asian foundry congress. 2001, The Chinese foundrymens association: Tapei. 29. Reilly, C, N.R. Green, and M.R. Jolly, "Investigating surface entrainment events using cfd for the assessment of casting filling methods", Modelling of casting, welding and advanced solidification process XII., 12, (2009),443-450.

20

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

Physical and Computational Models of Free Surface Related Defects in LowPressure Die-Cast Aluminum Alloy Wheels Jianglan Duan1, Daan M. Maijer1, Steve L. Cockcroft1, Carl Reilly1, Ken Nguyen2, Dominic Au2 'The Department of Materials Engineering, The University of British Columbia, Vancouver, BC, V6T 1Z4, Canada 2 Canadian Autoparts Toyota INC, 7233 Progress Way, Delta, BC, V4G 1E7, Canada Keywords: LPDC, A356 wheel, Flow-related defect, Mould filling, Water analogue experiment, Mathematical modeling Abstract A water analogue model has been used to simulate the free surface behavior during mould filling of a low-pressure die-cast (LPDC) wheel. The water analogue has been used to validate a mathematical model of the filling process. A transparent die was manufactured with a 2D profile extracted from a typical production die, and the experimental parameters used were based upon the production process. The flow occurring in the water model was recorded with cameras. A mathematical model of the filling process was developed to reproduce the behavior observed in the water model. Both the experimental and modeled results have shown a relative tranquil fill of the sprue and the rim, while persistent returning waves were developed in the spoke. The results highlight the significant effects of venting, both in the water model and computational model. Future work is required to improve the accuracy of venting modeling of the mould cavity. Introduction The quality of mould filling, which can be characterized by the free surface flow behavior, has been shown to have a significant effect on the integrity of cast components [1-3]. The rationale is that surface turbulence developed during filling can cause the liquid surface to fold and entrap oxide films or air pockets, which will then freeze into the casting and act as crack initiation sites, impairing the component's mechanical integrity [1, 4]. Take the mould filling of low-pressure die-cast of automobile wheels for example, an undesired filling condition with excessive turbulence can cause portions of the surface oxide film to be entrained within the bulk liquid resulting in defects such as cosmetic, hot tear, porosity and rim-leak defects[5]. In order to study defects caused by free surface turbulence, there is a need to view the free surface flow pattern during the filling process. However, due to the opaqueness of the steel wheel die cavity, it is not possible to directly observe the liquid flow. The complex 3D geometry of the wheel cavity also makes obtaining meaningful and instructive flow information troublesome using techniques such as real time X-ray imaging. As a preliminary investigation, this work has explored alternative means of studying the free surface flow - water analogue experiments and mathematical modeling using the computational fluid dynamics (CFD) software CFX. Water analogue experiments refer to using water to simulate the liquid metal flow [6, 7]. This methodology has previously been used to validate mathematical models and optimize casting systems [8]. The primary focus of the work at this stage is to validate the computational model against data obtained via water modeling. Once validated, the computational model will be then used, with

21

the appropriate properties for aluminum, to optimize the production mould geometry and process variables to reduce the occurrence of flow related defects. Experimental Procedure A 2D profile of the wheel and sprue extracted from the full 3D geometry of the production die, shown in Figure 1 (a), was used in the experimental work. It was decided at this preliminary stage not to use a 3D water model of the LPDC process due to the expense and complexity of obtaining meaningful data.

(a) (b) Figure 1. Water model design: (a) side view of a die-tooling assembly for typical LPDC process; (b) a transparent water model made of plexiglas The 2D profile was extruded to a thickness of 18 mm, and cut in a transparent Plexiglas plate (refer to Figure 1(b)). The plate was sandwiched between two Plexiglas plates, bolted together and sealed by O-rings. 1 mm holes were drilled at the hub, the outboard flange, and the inboard flange respectively, as these locations correspond to where the ejection pins and the mould parting lines are located in the production die (in the production die the gaps present in these areas serve to vent air from the die cavity during filling). The riser area is removed after casting, therefore it is not necessary for the region to fill completely, however the material contained in the riser is used to feed the rim section to prevent shrinkage porosity. Therefore inadequate filling of the riser due to lack of air venting in this area would result in insufficient feeding of liquid metal to offset the volumetric shrinkage associated with solidification in the neighboring areas. Thus an extra vent was added on the top of the riser in the water model to test its influence on filling. Drain cocks were installed on the vents, so different venting resistances could be studied. A schematic of the pressure control setup used in the experiments is illustrated in Figure 2. A water tank supplying a static head pressure of 2 m to an electronic valve was used to supply the fluid inlet condition to the mould. The electronic valve was controlled with an automatic PID control program to regulate the pressure downstream to follow a predefined pressure curve. The

22

predefined pressure curve used in the water analogue tests was a replication of that used in production, but which has been normalized to account for the different material densities and equipment configuration. The real time pressure information is collected and fed back to the control system by a pressure sensor located beside the bottom of the transition pipe. Five tests with the same predefined pressure curve, but different venting conditions (Table 1) have been carried out. Detailed characterization of the vents has not been conducted at this stage, thus the vent cocks were left either fully closed or fully open. Two cameras were used to record the fill process - one for a macro view and the second for close up viewing of specific areas of interest.

Figure 2. Schematic of the experiment setup Table 1. Venting Conditions for the Different Tests

Computer Simulation of Experiments The filling process of the aforementioned water experiments has been simulated in a commercial CFD program, ANSYS CFX, Version 12.0. Both the water and air are modeled so that the effect of venting can be properly characterized. A two phase inhomogeneous model, which calculates the liquid and gas phases discretely [9], was used to simulate the filling process. The material property data for these two fluids are specified from the CFX V12.0 materials database. The free surface sub-model and the surface tension model were used with the aim of modeling a realistic fluid interface and to allow for the accurate tracking and entrainment of gas bubbles. Heat transfer was not considered in the model due to the experimental process being isothermal. A geometry with the same profile as the experimental cavity was simulated, however, to reduce the computational size of the problem, the thickness of the computational domain was limited to 1 mm as compared to the experiment, which was 18 mm. As CFX is unable to compute 2D simulations a single element was employed through the thickness of the domain, and a symmetry

23

boundary condition was applied to the plane perpendicular to the extrusion direction and bisecting the sprue. This was undertaken after a computational study that found the wall effect to be negligible. The geometry was discretized by 19124 hexahedra elements, with a 1 mm element size for the wheel and sprue, and 1/3 mm for the vents. The boundary condition at the metal inlet was based on the real time pressure data detected by the pressure transducer located at the bottom of the transition pipe in the water analogue tests. To maintain as much similarity between the numerical model and the water model an attempt was made to match the vent area-to-die cavity volume ratio of the two. One challenge with using a 1 mm thick geometry instead of the actual 18 mm geometry is that similarity would require the vent in the computational model to be 1/18 mm χ 1/18 mm. As this would require an impractical mesh size the vent geometry within the model was set to 1 mm χ 1 mm χ 1 mm. The resulting excess vent capacity was addressed by adding a momentum source term to each vent comprised of a viscous loss term and an inertial loss term. The values for the losses were generated through a regression analysis. A series of computational models of the actual 3D vent geometry from the water model were used to characterize the pressure drop through the vents in the experimental apparatus. The pressure drops through the vent for nine different velocities ranging from 0.004 to 0.03 m s"1 were calculated. The momentum source across a vent within CFX is approximated by Equation 1, where AP is the pressure drop across the vent (Pa), a is the viscous loss coefficient and b is the inertial loss coefficient. These investigations allowed the viscous and inertial loss coefficients to be generated for the experimental apparatus used in this work. The simulation time for the model is approximately 25 hours when running on 2 Intel Xeon processors at 2.33 GHz AP = ax2+bx + c

(1)

Results and Discussion Experimental Results Analysis of the filling video of the water model has shown that the sprue is the first part to be filled and the flow is quiescent (Figure 3(a)). After that the fluid flows down to fill the outboard flange and a returning wave is generated, traveling backward from the outboard flange to fill the spoke and the hub (Figure 3(b)). One to three bubbles at the diameter of ~ 0.5 - 1.5 mm were seen to be entrapped below the free surface of the returning wave (Figure 3(c)). An obvious liquid height difference between the hub and the rim was observed, with the hub not fully filling until the fluid has risen up to approximately half the height of the rim (Figure 3(d)). In order to evaluate the effect of venting on the fill pattern, the time taken to fully fill discrete domains of the water model has been compared among all experimental tests. Lines are set at different locations of the die, as shown in Figure 4, to define the discrete domains. To be counted as fully-filled the volume fraction of water in the discrete domain has to be 1, small gas bubbles were neglected. When the free surface passes the line at the bottom of the sprue, with notation 0, the time frame is set to zero. Figure 5 shows the time variations between different test conditions. There are one or more columns missing at some locations, due to the fact that these domains did not fill under certain venting conditions. It can be seen that when the hub and riser vents are open (#1 and #3), regardless of the condition of the other vents, the mould can be fully filled, and the time taken for

24

the water to fill all regions is quite close. This indicates that the hub and riser vents play a significant role in the venting of the water model. However, in the production mould there is no riser vent, which is simulated by Test #4. In this condition, the fluid stops at the inboard flange because the riser fails to fully fill and a solidification defect may occur due to insufficient liquid metal supply.

Figure 3. Video images showing examples of (a) quiescent flow in the sprue; (b) returning wave in the spoke; (c) air entrainment; (d) filling sequence of the hub and rim. Each grid square shown in the background is equal to 5 mm

Figure 4. Definition of discrete domains within the wheel profile

Figure 5. Time taken to fully fill certain locations of the water model For Tests #1, #3 and #4, no considerable time variations between different venting conditions have been observed. The conjecture is that the automatic control has counteracted the effect of vents, by trying to maintain the static pressure in different tests to follow the same predefined

25

pressure curve. It is anticipated that without the automatic control, that is, leaving the electronic valve open, the time variation will be much more appreciable. Test #2 appears to be an exception, as it takes significantly less time to fill half of the rim than the other conditions, but then it fails to reach the inboard flange position within the defined pressure cycle. However, a check of the recorded pressure values shows that this phenomenon was largely an effect of the automatic control as well. In the lower half rim, the rate of pressure increase (height increase) due to the fact that the hub is not being filled, is larger than the changing rate in the predefined pressure curve. This results in the automatic control system over-closing the electronic valve as the upper part of the rim is being filled. Experimental Validation

(a)

(b)

Figure 6. Free surface level at different times: (a) experimental results of test 1; (b) modeling results of Test 1 The initial model results and equivalent real time free surface shape and position at different times in Test #1 are plotted in Figure 6. It can be seen that the free surface profiles between the simulated and experimental results are comparable. In addition, there is relatively good agreement in the fill times. The predicted liquid levels up to 17.8 s are consistent with the experimental results; however, after the lower spoke is filled and the air in the mould is separated into two isolated volumes around 19.4 s, the modeled free surface level deviates from the experiment. This is when the effect of the vents becomes obvious, since the remaining air is isolated in two regions and there is only one vent for each region. This result may indicate that the momentum loss applied to the vents does not accurately represent the actual momentum loss through the vents in the experiment. Thus, a set of computational studies with the momentum loss at the vents, either increased 75% or decreased 75% were carried out, to improve the accuracy of the model and also to obtain a preliminary perception of how sensitive the flow is to the venting conditions. A summary of the flow's sensitivity to the venting condition can be found in Figure 7.

26

Figure 7. Model prediction of the effect of venting condition on liquid height at areas of: (a) Sprue; (b) Spoke; (c) Rim Figure 7(a) shows negligible difference in the sprue, which is expected since the flow in the sprue is slow and is not strongly affected by the venting conditions. However, in the hub and the rim, a 75% increase or decrease in the momentum loss in the vent results in a height increase of around 50 mm in the hub and over 100 mm in the rim, which represents a significant change in the free surface level. All curves end at the time when the mould is fully filled, which are 22.6s, 21.8s, and 24s respectively for the original, decreased and increased momentum loss conditions. It clearly shows that with less momentum loss applied to the vent, there is less resistance to the mould fill, and thus the liquid advances faster and less time is required to fill the mould. This agrees with previous work on sand casting [10] that shows the importance of venting on flow patterns in casting cavities. The results of the sensitivity analysis on vent resistivity suggested that it may be possible to achieve better agreement between the water and numerical model with some additional tuning. In terms of free surface turbulence both the water model and the numerical model have shown a relative smooth and tranquil flow in the hub and rim, and a persistent returning wave in the spoke. Few bubbles have been shown in the simulated result; however, there is an appreciable fold-in of free surface while filling the hub. No noticeable difference in free surface turbulence severity was found when venting conditions were changed, which indicates that within 75% change range, the free surface turbulence magnitude is not affected by the venting condition.

27

Conclusion A water analogue model has been developed to conduct a preliminary investigation of the free surface behavior during mould filling of a low-pressure die cast (LPDC) wheel, and also to validate a mathematical model of the filling process. The filling video has revealed a relatively tranquil flow except within the spoke where a persistent returning wave occurs. Tests with different combinations of vents show that the hub and the riser vents are more important in venting than the flange vents. The interplay of the pressure control system and the vents appears to be complex and it would appear that in some of the cases examined the control algorithm is able to offset the changes imposed to the venting. The simulated free surface pattern is generally in good agreement with experimental results; however there is a trend to increasing difference between the two for times greater than 17.8 s. Changing the momentum loss in the vents by 75% was shown to result in a significant difference in the time taken to fill the mould and the height of fluid in all domains at a given time. The sensitivity analysis results clearly show that with less momentum loss applied to the vents, there is less flow resistance thus filling time is reduced. It is expected that with a proper momentum loss selected, the numerical model will be able to reproduce the experimental results for times greater than 17.8 s more accurately.

Reference 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

J. Campbell, Castings (Oxford, Boston: Butterworth-Heinemann, 1991), 1-26 R. A. Harding, "Towards more reliable investment castings," International Journal of Cast Metals Research, 19(5) (2006), 289-301. J. Mi, R. A. Harding, and J. Campbell, "Effects of the entrained surface film on the reliability of castings," Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, 35A(9) (2004), 2893-2902. J. Campbell, "Entrainment defects," Materials Science and Technology, 22(2) (2006), 127-145. B. Zhang, et al., "Casting defects in low-pressure die-cast aluminum alloy wheels," JOM Journal of the Minerals, Metals and Materials Society, 57(11) (2005), 36-43. R. Cuesta, et al., "Water analogue experiments as an accurate simulation method of the filling of gravity castings," Trans. Am. Foundry Soc, 114 (2006), 137-150. T. Nguyen and J. Carrig, "Water analogue studies of gravity tilt casting copper alloy components," Trans. Am. Foundrymen's Soc, 94 (1986), 519-28. R. M. McDavid and J. A. Dantzig, "Fluid flow in casting rigging systems: modeling, validation, and optimal design," Metallurgical and Materials Transactions B Process Metallurgy and Materials Processing Science, 29B(3) (1998), 679-690. ANSYS CFX 12.0 Documentation: ANSYS, Ine, 2009. J. Jakumeit, K. Goodheart, and M. Albers, Influence of the Gas Phase on Mold Filling for Sand Casting, 2009), 427-434.

28

Shape Casting: The 4,h International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SOLIDIFICATION MODEL COUPLING LATTICE BOLTZMANN METHOD WITH CELLULAR AUTOMATON TECHNIQUE Hebi Yin, Liang Wang, Sergio D. Felicelli Department of Mechanical Engineering and Center for Advanced Vehicular Systems Mississippi State University, Mississippi State, MS 39762 Keywords: Dendrite Growth, Modeling, Aluminum Abstract A two dimensional model combining the lattice Boltzmann method (LB) and the cellular automaton technique (CA) was developed to simulate dendrite growth during solidification. The LB method was used for the coupled-calculation of temperature, composition, and velocity fields while the liquid/solid interface was tracked by the CA method. The LB-CA model was validated by comparing tip velocity and equilibrium composition with analytical solutions. Single and multi- dendrite growth was simulated with energy and solute transport not only by diffusion but also by convection. In addition, the simulation morphology and computational time obtained with the LB-CA model was compared to that of a finite element - CA model. Introduction The solidification process is an important step in the manufacturing of components. In most cases, the mechanical properties depend on the solidification microstructure. Furthermore, the solidification process is usually accompanied by melt convection, which may significantly alter the pattern formation of dendritic microstructure. The emergence of new simulation methods enables prediction of grain structure and morphological evolution. Various types of numerical methods have been applied to characterize the dendritic growth in convection, including front tracking (FT) [1-3], level set (LS) [4. 5], phase field (PF) [6-11], and cellular automaton (CA) [12-15] methods. In the above solidification models, the calculation of transport phenomena, including mass, momentum, and energy, is based on solving of the equations of transport numerically, using finite difference method (FDM), finite volume method (FVM) or finite element method (FEM). Since all these solvers imply continuum-based approaches, it is not easy to properly handle the discontinuity of flow velocity at the moving S/L interface. Over the last two decades, the lattice Boltzmann method (LB) has rapidly emerged as a comparatively powerful technique with great potential for numerically solving fluid flow phenomena. The LB method has also been extended to model energy and solute transport coupled with fluid flow.

29

Some solidification models have been built with the lattice Boltzmann method for the calculation of transport phenomena. Miller et al. [16, 17] constructed a PF-based model for crystal growth with convection in the framework of the LB method. The two-dimensional growth of a crystal with different growth directions has been studied for a pure metal with buoyancy convection. Medvedev et al. [18] adopted a PF-LB scheme to simulate dendrite growth from a supercooled melt, including heat transport by both diffusion and convection. The phase-transition part of the problem was modeled by the PF approach, whereas the flow of the liquid was computed by the LB method. Chatterjee et al. [19] proposed an enthalpy-based hybrid LB technique for simulating transport phenomena during phase-transition processes. The newly developed method was subsequently applied to crystal growth during equiaxial solidification of an undercooled melt. Sun et al. [20] developed a 2D LB-CA model to simulate dendrite growth. In their model, the LB method was adopted to simulate the solute distribution and fluid flow, while a CA technique was used to predict the dendrite growth, but the model did not consider the calculation of temperature field. In this work, a solidification model is developed to simulate dendritic growth under convection. The CA technique for interface tracking is coupled with a transport model (LB) for calculating heat and solute transfer by both convection and diffusion during solidification. Single and multidendrite growth is simulated with various boundaries conditions, showing good performance of the model and potential for large scale calculations. Model Description LB model for heat/mass transfer and fluid flow In the present work, the LBM is adopted to numerically calculate the fluid flow, solute transport and heat transfer in a 2D domain with Cartesian coordinate system. The governing equations and boundary conditions for transport phenomena and fluid flow can also be found in detail in Refs. [20-25], The primary variables in the LB formulation are the so-called fluid density distribution functions or directional densities, each relating to the probable number of fluid particles moving with velocity ea(a = 0,... 8), along the ach direction at each node. Following the single relaxation time formulation, the evolution of density distribution functions at each time step is given by: faOC + e.At, t + At) = /„(*, t) - Wi-jr™

(1)

where X + eaAt is the nearest node to X along the direction α, τν is the relaxation time (τν = 3v + 0.5) with v kinematic viscosity, and / j f is the equilibrium distribution function.

/."») = wap(X) [l + 3ψ

30

+ 1<ψ.-£]

(2)

where wa is the weight for "a" direction and c is the lattice speed [20]. ( (0,0) ea = j (cos[(a - 1)π/2], sin[(a - 1)TT/2])C ((cos[(o - 1)π/2], sin[(a - l > / 2 ] ) c

a=0 a = 1-4 a = 5- 8

(3)

The macroscopic fluid variables, density p and velocity u, can be obtained from the moments of the distribution function fa as: Ρ = Σ2=ο/β

(4)

U =^Σα=0 /αβα

(5)

A separate distribution function fTa is used for the heat transfer problem, whose evolution follows the equation: fTA(X + eaàt, t + At) = fTA(X. t) - ^ ^ 0 ; ^ 1 * ·

0

(6 )

where τ α = 3α + 0.5, a is thermal diffusivity. The macroscopic energy is defined as e = Σα Τα/Ρ, and tne equilibrium distribution function for the thermal energy distribution / / ' can be written as:

(

feq

_

IT.O

-

2peu2 3

fl^P JT,a ■"^0

c2

+^ 9 Î2 36 L

+^

2 c

2

C2

-

m 2

(a = 1,2,3,4)

(7) v

2

c"

2 C J

v

2

c4

2c!J

v

'

'

· · · J

Similarly, the distribution function ga for the solute transport satisfies the equation: ga(X + eaAt, t + At) = g a (*. 0 - " - ^ - ^ ^ (8) where TC = 3Dt + 0.5, D; is the solute diffusivity. The solute concentration is ca = Σα9α> w ' t n the equilibrium distribution given by: • gea" = wacatt + 3ea-u) (9) Kinetics parameters for the CA model The solute distribution ahead of the interface is used as a driving force to simulate the dendrite growth in the CA model. The process of dendrite growth is predominantly controlled by the difference between the local interface equilibrium solute concentration and the local actual liquid

31

solute concentration. The changing rate of solid fraction determines the velocity and morphology of solid growth. Based on the calculation of the local actual liquid concentration ct and the interface equilibrium composition c\, the increase of solidfraction,Δ/5 at the interface cells can be obtained as [26]: ΔΛ = (c,* - c,)/(c; ■ (1 - k))

(10)

The interface equilibrium composition is calculated by: C

' = C° +

7"-7f+ΓΚ·/(φ,8|,)

,...

<M>

„, .

where c0 is the initial solute concentration, T°q is the equilibrium liquidus temperature at the initial solute concentration, ml is the liquidus slope, Γ is the Gibbs-Thomson coefficient, and K is the curvature of the S/L interface. The function accounting for the anisotropy of the surface tension is denoted by f(ip, θ0) where φ is the growth angle between the normal to the interface and the x-axis, and θ0 is the angle of the preferential growth direction with respect to the x-axis. The interface equilibrium temperature is denoted by Γ*. For cubic crystals, the function f(q>, 0O) exhibits a four-fold anisotropy [27]: /(
(12)

Where δ is the anisotropy coefficient, and the growth angle can be calculated from the following equation: hir-cos^f *-^ ) l Z7r LOb \ws/dx)*Hafs/eyW2j

^<0 ay ^υ

The interface curvature of a cell with solid fraction fs can be obtained by [26]:

« = f(^)2 + (T)T/2 [Vox/

\dyj

J

x W-rirt- ίΨΪτ^- [ψ)τ% [ dx dy dxdy

\dxj

dy2

\dy/

dx2J

Verification The energy and solute transport simulated by the LB method was validated by comparing the simulation results of temperature and composition distribution to analytical solutions. Also, the LB-CA model was validated by modeling the free growth of a single dendrite and comparing the tip velocity and equilibrium liquid composition to those obtained by the analytical LGK model [28] with various undercoolings. For the latter validation, a two-dimensional case of single

32

^

dendrite growth was simulated, without consideration of heat transfer and fluid flow. At the beginning of the simulation, a solid seed with composition kC0 and a preferential crystallographic orientation of 0 degrees with respect to the horizontal direction was placed at the center of the domain. The cells surrounding the seed were assigned as interface cells. Lipton, Glicksman and Kurz [28] developed an analytical model (the LGK model) which describes the steady-state free dendrite growth under a given melt undercooling. Fig. 1 shows the tip growth velocity and equilibrium concentration of the steady-state growth of an Al-3wt%Cu dendrite as a function of undercooling. As observed, a good agreement is obtained between LB-CA and the values calculated by the LGK theory as well as those obtained by a Finite Element (FE) - CA model previously developed by the authors [29].

(a)

(b)

(c)

Figure 1. (a) Schematic of dendrite growth model and boundary conditions (changed to achieve different undercoolings). Comparison between numerical simulations and LGK model predictions of the steady-state (b) tip velocity and (c) tip equilibrium liquid composition for Al-3.0wt%Cu alloy Simulation results All the performed simulations of dendrite growth were done for a binary Al-3wt% Cu alloy. In the example of Figure 2, a single nucleus with 0-degree preferential direction was placed at the center of a calculation domain of 90χ90μηι discretized with a 300x300 grid.

Figure 2. Dendrite morphology and solute map with various boundaries conditions, (a) solute diffusion only; (b) solute + heat ; (c) inflow at left side (Uin = 0.0023m/s).

33

Three cases were studied for different boundary condition: Case 1 - constant undercooling and no convection; Case 2 - constant temperature gradient (400 K/m) imposed on the four boundaries and no convection; Case 3 - inflow imposed on the left wall (Uln = 0.0023m/s). Figs. 2(a-c) show the single dendrite morphologies and composition fields for simulation cases 1 - 3 by the LB-CA model. By comparing the dendrite morphologies shown in Figs. 2(a) and (b), it is observed that the cooling rate due to the heat extraction from the boundaries enhances the side branching and thus the formation of secondary arms. By comparing the dendrite morphologies shown in Figs. 2(a) and (c), it is noticed that the growth of dendrite arms is enhanced in the upstream side. As the dendrite grows, solute rejected to the liquid ahead of the solid/liquid interface is washed away by the upstream fluid flow, which leads to an asymmetrical solute distribution and dendrite morphology according to Eq. (10). A multiple dendrite case is considered in the next example. Fig. 3 presents the simulated evolution of equiaxed multi-dendrite growth. The domain is assumed to have uniform initial temperature and composition. Seven nuclei with an initial composition kC0 and random preferred growth orientations ranging from 0 to 90 degrees with respect to the horizontal direction were randomly distributed in the calculation domain.

Figure 3. Equiaxed growth of multiple dendrites of Al-3.0 wt% Cu alloy solidified with (a) solute transport only, (b) a constant heat flux at boundaries, (c) constant inflow velocity at left side, and (d) inflow velocity and constant heat flux at boundaries. Fig. 3(a) shows the simulated dendrite morphologies considering only solute transport. It is observed that the dendrites develop the main primary arms along their crystallographic orientations but without secondary arms. The growth of some primary arms is suppressed by nearby dendrites. A cooling rate due to heat flux imposed at the boundaries enhances the side branching as shown in Fig. 3(b). Fig. 3(c) is similar to the simulation of Fig. 3(a) but with convection added, due to an imposed uniform inflow on the left boundary. Compared to Fig. 3(a), the primary dendrite arms shown in Fig. 3(c) are coarser and longer in the upstream direction than those in the downstream direction. Finally, Fig. 3(d) shows the simulated dendrite morphologies when both solute and heat transfer with convection are included in the model. The melt flow washes away the interdendritic composition and enhances the dendrite growth and the merging between dendrites. It is observed that convection promotes the removal of solute from the solid/liquid interface in the upstream side and thus increases the interface stability, resulting in coarsening of the dendrite morphology.

34

In the last example, Figs. 4(a) and (b) show the simulated dendrite morphologies, solute profiles, and flow fields obtained with the LB-CA model of this work and the FE-CA model developed in Ref. [29]. A uniform inlet flow velocity of Uin = 0.0023m/s is imposed on the left side. For both FE-CA and LB-CA models, the growth of the dendrite arms is promoted in the upstream side. The LB calculation yields a higher velocity around the dendrite (and consequent enhanced growth) than the FE model, probably due to the differences in both schemes in treating boundary conditions at the S/L interface. Fig. 4(c) shows the computational time needed to obtain the dendrite morphologies shown in Figs. 4(a) and (b) with different grid sizes adopted in this simulation for FE-CA and LB-CA models. The simulation results show that the LB-CA model has much higher computational efficiency compared to the FE-CA model. This is particularly evident for solidification simulations including fluid flow. As the size of the problem increases, so does the size of the assembled matrices resulting from the finite element equations. The LBCA is an all-local scheme, where results in one cell are calculated from information on neighbor cells; no assembly of matrices or solution of algebraic equations is required. Consequently, the method has a high potential for large scalability and parallelization.

(a)

(b)

(c)

Figure 4. Dendrite morphology and solute map by (a) FE-CA and (b) LB-CA with inflow at left side (Uln = 0.0023m/s), (c) CPU time by FE-CA and LB-CA models for various grids Conclusions This work presented a new solidification modeling technique to simulate dendrite growth that uses lattice-Boltzmann to solve the transport equations and cellular automaton to track the interface. Simulations of single and multiple-dendrite growth with the binary Al-3.0wt%Cu alloy were performed showing good stability and accuracy. When compared with a finite element - cellular automaton model, the LB-CA model showed significant improvement in scalability for problems involving solidification under convection.

35

Acknowledgements This work was funded by the National Science Foundation through Grant Number CBET0931801. We appreciate the generous sharing by Prof. M.F. Zhu of her Lattice Boltzmann code, from which our LB-CA model was developed. References 1. 2. 3. 4. 5. 6. 7. 8. 9.

Al-Rawahi N and Tryggvason G. Journal of Computational Physics 2004; 194:677. Al-Rawahi N and Tryggvason G. J. Comput. Phys. 2002; 180:471. Li CY, Garimella SV, and Simpson JE. Numerical Heat Transfer, Part B 2003; 43:143. Tan L andZabarasN. Journal of Computational Physics 2006; 211:36. Tan L and Zabaras N. Journal of Computational Physics 2007; 221:9. Beckermann C, Diepers HJ, Steinbach 1, Karma A, Tong X. J. Comput. Phys. 1999; 154:468. Bansch E and Schmidt A. Interf. Free Boundaries 2000; 2:95. Jeong JH, Dantzig JA, and Goldenfeld N. Metall. Mater. Trans. A 2003; 34:459. Lu Y, Beckermann C, and Karma A. Convection effects in three-dimensional dendritic growth, in: Proceedings of the 2001 Fall MRS Meeting in Boston, MA, 2001. 10. Tong X, Beckermann C, Karma A, and Li Q, Phys. Rev. E 2001; 63(6):061601. 11. Lan CW, and Shih CJ. J Cryst Growth 2004; 264:472. 12. Zhu MF, Lee SY, and Hong CP. Phys. Rev. E 2004; 69(61):061610. 13. Shin YH and Hong CP. IS1J Int. 2002; 42(4):359. 14. Zhu MF, Hong CP, Stefanescu DM, and Chang YA. Metall. Mater. Trans. B 2007; 38:517. 15. Mullis AM. Acta Mater. 1999; 47(6): 1783. 16. Miller W, Succi S, and Mansutti D. Physical Review Letters 2001; 86(16):3578. 17. Miller W, Rasin I, and Succi S. Physica A 2006; 362:78. 18. Medvedev D and Kassner K. Journal of Crystal Growth 2005; 275:el495. 19. Chatterjee D and Chakraborty S. Physics Letters A 2006; 351:359. 20. Sun D, Zhu MF, Pan S, and Raabe D. Acta Materialia 2009; 57:1755. 21. Peng Y, Shu C, and Chew YT. Physical Review E 2003; 68:026701. 22. Dixit HN and Babu V. International Journal of Heat and Mass Transfer 2006; 49:727. 23. Liu CH, Lin KH, Mai HC, and Lin CA, Comput. Math. Appi. 2010; 59(7):2178. 24. Okaltun SG and Dulikravich GS, Comput. Math. Appi. 2010; 59(7):2431. 25. Feng YT, Han K and Owen DRJ. Int. J. Numer. Meth. Engng. 2007; 72:1111. 26. Zhu MF and Stefanescu DM. Acta Mater. 2007; 55(5):1741. 27. Trivedi R and Kurz W. Int. Mater. Rev. 1994; 39(2):49. 28. Lipton J, Glicksman ME and Kurz W. Mater. Sei. Eng. 1984; 65(1):57. 29. Yin H and Felicelli SD. Acta Mater. 2010; 58(4): 1455

36

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

PHYSICAL CHARACTERIZATION OF THE PEREMABILITY OF EQUIAXED EUTECTIC STRUCTURES IN HYPOEUTECTIC ALUMINUM ALLOYS Ehsan Khajeh and Daan M. Maijer Department of Materials Engineering, The University of British Columbia, 6350 Stores Road, Vancouver, BC, Canada V6T 1Z4 Keywords: Permeability, Eutectic, Cellular automaton, X-ray Microtomography, Al-Cu Abstract The permeability of hypoeutectic aluminum alloys during equiaxed eutectic solidification has been determined through physical modeling. The 3D geometries of primary phase obtained from X-ray microtomography (XMT) scans of solidified aluminum alloys have been used to generate computational domains for use in simulating the eutectic transformation. By applying the Cellular Automaton technique to simulate the growth of eutectic colonies, the evolution of the liquid channels has been modeled. Large-scale analogues of the simulated structures were produced by rapid prototyping for use as physical models. A glycerin-based solution was passed through the physical models and the permeability was calculated from measurements of the discharge flow rate and pressure drop. This work presents an alternative technique to determine permeability compared to conventional permeameters and shows a deviation from the conventional Carman-Kozeny expression at high eutectic grain density. Introduction Most commercially relevant casting alloys form significant amounts of eutectic phase during solidification. Differences in the size, shape and distribution of the eutectic grains influences the permeability, and thus will have a large impact on the magnitude of the pressure drop that develops during the last stages of solidification. Knuutinen et al.[l] showed that there is a strong correlation between the amount and distribution of porosity and the eutectic solidification mode in aluminum alloy A356. They characterized the effect of eutectic modifier additions on porosity and rationalized their results through comparisons to the eutectic solidification mode and its effect on permeability in the mushy zone. Experimental studies to measure permeability have predominantly focused on samples with primary phase equiaxed dendritic microstructures [2]. The permeability of these samples is measured with the aid of a permeameter where the sample is held in an isothermal environment at a temperature in the mushy zone and liquid, typically of eutectic composition, is forced through the sample. Although the natural tendency of the microstructure to change during the test is a major source of error in this type of experiment [3], this method is still an effective means to measure permeability. For eutectic solidification, this measurement methodology cannot be employed because eutectic solidification is an isothermal transformation and it is not possible to hold the eutectic structure at a point mid way through the eutectic transformation. Physical models may provide a suitable technique to address the difficulty associated with measuring the permeability in a developing eutectic phase. Despois and Mortensen [4] measured the permeability of open-pore microcellular materials (open foam) by passing glycerin mixed

37

with water through aluminum open-pore foams produced using a replication process. From this work, they proposed a permeability equation for open-pore materials. James et al. [5] passed oil through a large scale physical model of repeating slotted plates to measure permeability. Khajeh and Maijer [6] have also developed a physical model to measure the permeability of dendritic structures by passing a glycerin-based solution through large-scale analogues of interdendritic structures produced by a rapid prototyping. The results of these studies [4-6] suggest that experiments employing physical models to characterize permeability are a suitable alternative to permeameter measurements. In this study, the 3D, near eutectic, primary phase structure of Al-20wt%Cu aluminum alloy, obtained by XMT, has been used to generate a computational domain in which the eutectic transformation takes place. By modeling the nucleation and growth of eutectic grains, the evolution of the eutectic phase has been predicted using a Cellular Automaton (CA) technique. Large-scale analogues of the simulated microstructure at various points in the eutectic transformation were then produced by rapid prototyping for use as physical models for permeability determination. Procedures X-ray microtomographv (XMT) An experimental test casting, with a cylindrical shape, was produced from Al-20wt%Cu alloy (prepared from commercially pure Al and Al-50wt%Cu master alloy). The casting, poured with a melt temperature of 650 °C, was solidified with a low cooling rate (measured above the eutectic temperature) of 0.19 C/s. From the centerline of this casting, a small diameter cylinder (~ 3mm dia x 15mm) was extracted for XMT analysis. The sample was scanned, with a resolution of 1.85μπι and energy of 35KeV, on the TOMCAT X-ray microtomography facility located at the Swiss Light Source. The different X-ray attenuation of the primary and eutectic phases allowed the geometry of the eutectic phase to be separated by phase contrast. A computational domain of the eutectic structure in a 1 χ 1 χ 1 mm3 cube was extracted from the scan data for use in the present eutectic growth model. Eutectic Solidification Model The eutectic solidification of hypoeutectic Al-20wt%Cu has been modeled using the CA technique. The model has been employed to predict the evolution of the equiaxed eutectic phase that forms in the domain, which was extracted from the XMT data described in the previous section. Instantaneous nucleation of eutectic grains at the beginning of eutectic transformation is assumed and the locations of the nucleation sites within the calculation domain are determined randomly. The nucleation density of eutectic grains, Nn depends on the cooling rate and impurity level and based on previous assessment by the authors [7], a value of 90 mm was employed for the current cooling rate and melt impurity level.. The calculation domain is a cube, with 1 mm edge length, that is divided into a regular network of uniform cubic cells with 5.55 μπι edge length (180 cells in each direction). The state of each cell is characterized as either liquid, primary solid, or a eutectic grain number. A periodic boundary condition is applied to all six faces of the cubic domain. Although the dendritic structure is initially non-symmetric due to the stochastic nature of the solidified dendritic structure, the application of periodic boundary conditions allows the effects of eutectic grains that grow into the domain to be considered. For the sake of brevity, the reader is referred to [7] for a complete description of the CA model employed in this study.

38

The evolution of the eutectic microstructure for a eutectic grain density (Nv) of 90 mm1 is shown in Fig. 1 at two stages during eutectic solidification corresponding to solid fractions of 0.70 to 0.85. The permeability determination has been performed using large scale replicas of the simulated microstructures.

Figure 1. Predicted microstructure for Nv=90 mm'' at solid fraction of a) 0.70, and b) 0.85. The cube edge length is 1 mm. Physical determination of permeability A physical modeling approach has been developed to measure the permeability during eutectic solidification. Scaled replicas of the solid phase (primary Al phase and the solidified eutectic grains) predicted by the CA model for Nv=90 mm'3 were constructed using a rapid prototyping technique. For this study, the voxel-based simulated microstructures were first manipulated to identify the isosurface representing the solid/liquid interface [6]. The corresponding geometries were then scaled by a factor of 50 and used to construct scaled physical models via the Selective Laser Sintering (SLS) rapid prototyping technique. Fig. 2 shows an example of the surface-based (triangulated interface) microstructure along with the corresponding scaled replica.

Figure 2. (a) Surface-based representation of the solidified microstructure for Nv=90 mm'3 at solid fraction of 0.78, and (b) polyamide replica of (a).

39

The permeability of each microstructure replica was measured using the purpose-built setup shown in Fig. 3. In this apparatus, a pressure drop is established across the sample to force the working fluid through the sample and the flow rate is measured at steady state. To achieve similitude with the flow conditions (Re < 1) occurring in a solidifying alloy, it was necessary to use a liquid with a high viscosity. Therefore, laboratory-grade pure glycerin was selected as the working fluid used in these measurements. A 40L, pressurized tank, connected via a 30mm diameter pipe, was used as a fluid reservoir for the apparatus. Due to the low discharge rates of glycerin solution during the measurements and the large diameter of the reservoir tank (-40 cm), the fluid level in the reservoir and the resulting supply pressure were effectively constant during each test.

Figure 3. Layout of the apparatus built to measure the permeability of replica samples, cross-section area surrounding the sample holder (section A-A), and close-up image of sample inside the holder (section B-B).

To adapt the cube-shaped microstructure replicas to the round pipe cross-section, an insert was fabricated from silicone rubber and bonded to the inside of a section of 4" pipe. During a test, a replica sample was mounted inside the cubic cavity of the holder and the side-surfaces were sealed with silicone gel sealant. To assess whether the working fluid could by-pass the interior of the replica by flowing along the outside surface of the replica and/or along the insert/pipe interface, a leak check was performed with a solid cube. At pressures corresponding to the maximum pressure used during testing, no flow was detected. Once a replica sample had been loaded and the piping resealed, the following steps were performed during each test:

40

Step 1: Valves #1 and #3 were closed and valves #2, #4 and #5 were opened. A vacuum source, connected on the outlet near valve #2, was activated to extract the air in the system from both sides of the sample. The evacuation continued until a vacuum pressure of-97 kPa was reached. Step 2: Valves #2 and #3 were closed and valves #1, #4 and #5 were opened. The working fluid entered the system from the fluid reservoir and filled around and into the replica sample from both sides. The vacuum applied during the previous step ensures proper filling of the system including the replica sample and eliminates the possibility of air bubble formation during filling. Step 3: Valves #2, #4 and #5 were closed and valves #1 and #3 were opened. Once valve #3 was opened, the outlet pipe of the system (left side of valve #3) filled with liquid. Once steady state flow conditions were achieved based on timed flow measurements, the permeability measurement was performed. The pressure of the liquid supplied to the apparatus was set at gauge pressure values of up to 200 kPa and was controlled by adjusting the pressure in the fluid reservoir with compressed air. The pressure at the entry of the replica sample, Ptn, was measured by a calibrated pressure transducer, with an accuracy of ±1.5 kPa. The outlet pressure, P0uh was assumed to be equal to atmospheric pressure or a gauge pressure of zero. For each replica sample, the volumetric flow rate, Q, was determined for an applied pressure drop by measuring the time to fill a volume-calibrated container after steady state flow was achieved. The permeability was then calculated using Darcy'slaw [4]:

Q. q^EmZfouL)

(1)

A μ L where A is the cross-sectional area of the replica sample, q is the superficial velocity, K is the permeability, L is the length of the replica sample, and μ is the viscosity of the fluid. The viscosity of the fluid was measured at test temperature using a rotational viscometer (Brookfield DV-E viscometer) with the accuracy of ±1% FS. The permeability tensor was measured by sequentially reorienting the replica sample inside the holder. To assess the consistency of the measurements, 4-10 successive measurements of the pressure and the flow rate were performed for each replica sample orientation. Results and Discussion The non-zero components of the experimentally measured permeability tensors along with the inverse of the specific surface area (S„";) and effective solid fraction (real solid fraction plus the fraction of isolated liquid regions) are summarized in Table I for the microstructure with a eutectic grain density of Nv=90 mm'3. The absolute values of the physical permeability increase with decreasing solid fraction consistent with the ease of flow through the microstructures. Table I. Physical permeabilities for Nv=90 mm3. Reported values are for non-scaled microstructure. Sample # 1 2 3 4

Effective Solid fraction 0.524 0.626 0.732 0.831

S v '(pm) 25.0 34.8 52.8 87.4

K„ 7.96 3.70 2.51 1.22

41

Physical permeability (x 10 K,v Κ,ν. κ,, 8.56 9.40 8.64 4.60 6.39 4.90 2.82 2.60 2.47 1.09 1.01 0.74

uK 0.60 0.83 0.36 0.16

Uncertainty analysis The measured variables in the experiments were the pressure drop (ΔΡ), the volumetric flow rate ( 0 , the viscosity of the working fluid (M), and the dimensions of sample (L and A). By estimating the uncertainty in the measurements of each of these variables, the overall uncertainty of permeability measurement can be analyzed. Neglecting the uncertainty of sample dimensions, i.e. L and A, the Root Sum Square (RSS) of the estimated uncertainties has been used to calculate the propagated uncertainty as:

By applying equation (2) to equation (1), uK is given by:

uK=L

{&LU

A\hP

f+{JLu

ΔΡ

Q

Y+{Q-U f AP

μ

(3)

u^p is the uncertainty of pressure measurement and for the pressure transducer used in this study was ±1500 Pa. For the combination of selected spindle and rotational speed used to measure the viscosity of the working fluid, the accuracy of the viscosity measurement was ^=±0.03 Pa.s. The volumetric flow rate, Q, was measured based on the time to fill a volume-calibrate container. Neglecting the uncertainty of filling time (upO), UQ can be expressed as: "Q=-UV

(4)

where uv is the uncertainty of container volume and was equal to ±10 cc. Using the elemental uncertainties and equation (3), the calculated relative uncertainty is 7-15% (refer to Table I). Another factor causing uncertainty is the accuracy of SLS replicas. The SLS machine is not able to accurately reproduce features of sizes equal to the minimum accuracy (0.15 mm for the machine used in this study) leading to growth, shrinkage, or smearing of features in the cured solid, which affects the solid fraction and specific surface area. Comparison with related studies Experimentally measured permeabilities were compared with numerically determined permeabilities [8], as well as, values predicted through a recently proposed mathematical expression [8]. Numerical permeabilities were calculated by solving the continuity and momentum equations for the corresponding unstructured meshes of the 3D geometries used for the physical models. The mathematical expression was obtained by considering the eutecticdendritic network as a dual-structured porous medium. As presented in [8], the permeability of hypoeutectic alloys during equiaxed eutectic solidification, K, is: K=

O-fJ* ^_tanh(g) 60(/ t -/ ff ) 2 V Θ'

42

(5)

e_

(\-fs)D \\-fa 7.75(/,-/e)V Kd

V

'

where fs is the solid fraction, fa is the fraction of primary phase, D is the diameter of eutectic grains, and Kd is the near-eutectic permeability. For comparison, the measured permeabilities have been also compared with the widely used Carman-Kozeny expression [9]:

kc^vfs where kc is a constant and Sv is the inverse of microstructural length scale defined as solid/liquid interfacial area per unit volume of solid. Fig. 4 shows the measured (average of K„, Kyy and Kzz) and calculated (refer to [8]) permeabilities versus the effective solid fraction for Nv equal 90 mm'3 and lines representing equations (5) and (7). As shown in Fig. 4, the physically determined permeabilities decrease with increasing solid fraction but deviate form Carman-Kozeny expression with constant kc. The physical permeabilities are in good agreement with predictions through equation (5) as well as the numerically calculated permeability. Deviations are attributed to the previously discussed uncertainties. Additionally, the boundary condition applied to the outer walls (free-slip) in the numerical permeability calculations was different than the conditions present in the physical permeability tests where it was no-slip. Khajeh and Maijer [6] showed that applying a no-slip condition to the outer walls in the numerical model results in a maximum 7% decrease in the calculated permeability compared with the free-slip outer walls predictions.

Solid fraction

Figure 4. Absolute permeability versus solid fraction for Al-20wt%Cu eutectic with Nv=90 mm'3. Closed circles with error bars represent measured permeabilities. Open triangles represent the numerically determined permeabilities from [8]. Lines indicate permeability based on the equations (5) and (7).

43

Summary and Conclusion A physical model for measuring the eutectic permeability of Al-20wt%Cu has been developed. The permeability was measured on large-scale analogues of the solidified structures constructed using an SLS technique. The solidified microstructure was obtained from CA simulation of eutectic grain growth starting from primary grain structure based on X-ray microtomography scans of a cast sample. A glycerin-based working fluid was then passed through the scaled replicas and the permeability was calculated from measurements of the discharge flow rate and pressure drop. Characterized permeabilities were then compared with the calculated permeabilities from two available mathematical expressions. The results of this study showed a good agreement between numerical and physical permeabilities but deviation from Carman-Kozeny expression. It was found that the recently developed mathematical expression based on dual-structured porous medium approach is accurate for predicting the permeability during equiaxed eutectic solidification. References 1. A. Knuutinen et al., "Porosity formation in aluminum alloy A356 modified with Ba, Ca, Y and Yb", Journal of Light Metals, 1 (2001), 241-249. 2. O. Nielsen et al., "Experimental determination of mushy zone permeability in AluminumCopper alloys with equiaxed microstructure", Metallurgical and Materials Transactions A, 30 (1999), 2455-2462. 3. O. Nielsen, and L. Arnberg, "Experimental difficulties associated with permeability measurements in aluminum alloys", Metallurgical and Materials Transactions A, 31 (2000), 3149-3153. 4. J.F. Despois, and A. Mortensen, "Permeability of open-pore microcellular materials", Ada Materialia, 53 (2005), 1381-1388. 5. J.D. James et al., "Experimental apparatus for validation of computer models of the permeability of metallic alloys in the mushy zone", Modeling of Casting, Welding and Solidification Processes - XI. Opio, France, TMS, (2006), 1189-1196. 6. E. Khajeh, and D.M. Maijer, "Physical and numerical characterization of the near-eutectic permeability of aluminum-copper alloys", Ada Materialia, 58 (2010), 6334-6344. 7. E. Khajeh, and D.M. Maijer, "Inverse analysis of eutectic nucleation and growth kinetics in hypoeutectic Al-Cu alloys", Metallurgical and Materials Transactions A, 2010, doi: 10.1'007/sl 1661-010-0489-7 8. E. Khajeh, and D.M. Maijer, "Permeability of hypoeutectic aluminum alloys during equiaxed eutectic solidification", 2010, submitted. 9. P.C. Carman, "Flow of gases through porous media", Butterworth Scientific, London, 1956, 11-13.

44

Shape Casting: The 4 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

FOAM FILTERS USED IN GRAVITY CASTING Fu-Yuan Hsu and Huey-Jiuan Lin Department of Materials Science and Engineering, National United University, No. 1 Lein-Da, Kung-Ching Li, MiaoLi, 36003 Taiwan, R.O.C. Keywords: ceramic foam filter, runner system, aluminum gravity casting, critical gating velocity. Abstract Ceramic foam filters are normally used for reducing the velocity of liquid metal in the design of runner system. In this study, four designs of runner systems with various orientations of foam filters were explored and their apparent velocities were estimated by casting experiment and computational modeling. In the casting experiment, trajectory and metal weighing methods were employed for measuring apparent velocity and flow rate respectively. Using Forchheimer's equation, a porous material such as a foam filter could be simulated in the modeling. The modeling result was validated by the casting experiment. For high efficient usage of a foam filter, through which liquid metal with high flow rate and low velocity was transformed, the optimized design is recommended in all of runner systems. Introduction Campbell [1] proposed that the whole running system should be designed so that the velocity of the metal in gates is below a certain critical value. The value varies from one alloy to another. For aluminum-based alloys, it should be within the safe range of 0.25-0.5 m/s. High reliable aluminum alloy castings were obtained by installing the foam filters in a bottom gating system at an initial gate velocity less than or equal to 0.5 m/s [2, 3]. Consequently, the liquid aluminum has a critical ingate velocity of less than 0.5 m/s. Thus, one of the desirable functions of a good runner system design is to decelerate the liquid metal to a speed below this critical ingate velocity. This could explain why a ceramic foam filter has played an essential role for reducing flow velocity in aluminum gravity casting. Except the structures of extruded channels and planar meshes, the reticulated structure of a ceramic foam filter provides high tortuosity for liquid metal flowing through it [4], A commercial ceramic foam filter is made by the replica technique, which is a polymeric foam skeleton coated with ceramic slurries (e.g., SiC and/or AI2O3 slurry). A subsequent heat treatment is applied to cure the coated ceramic and to burn out interior polymeric materials. Consistent specifications of foam filters are obtained by controlling the size of bubbles during the foaming process of manufacturing polymeric foam. The type of foam filters has been classified by the count of pores per linear inch (denoted as ppi). However, the uniformity of the pore size in the filter is varying; and, it changes from time to time even at the same production in one company. Although same ppi value of foam filters is used, their efficiency would be different from that produced by other manufacturers. In this study, a

45

procedure to quantify the efficiency of a foam filter was proposed. Four runner systems with different orientations of foam filters were explored. The flow velocity and flow rate of the liquid metal flowing through these filters were also estimated in each runner system. Method With different orientations of foam filters, four runner system designs were shown in Figure 1. In each runner system, common geometries with the same sizes were included, such as a pouring basin, a sprue, a L-shaped runner [5], and a horizontal runner. For all runner systems, the constant total head height of 300mm was intentionally designed for simulating a normal condition in a real casting. Figure 1 also illustrated four orientations of foam filters, which are the systems of (a) runner across filter (RAF), (b) runner zigzag with filter (RZF), (c) runner tangent to filter (RTF), and (d) expanded runner tangent to filter (ERTF).

Figure 1 Four runner systems with different orientations of foam filters: (a) runner across filter (RAF), (b) runner zigzagging with filter (RZF), (c) runner tangent to filter (RTF), and (d) expanded runner tangent to filter (ERTF). In this study, a commercial ceramic foam filter was used. Initially, the specification (ppi) of this filter was not decided, until a quantification procedure for a foam filter was implemented. In this procedure, both the computational modeling and the experimental data from literatures were utilized. Consequently, the modeling result of a foam filter was validated by the casting experiment. After the liquid metal flows through a foam filter, its flow velocity was estimated by the trajectory method and its flow rate was recorded by the metal weighing machine. Computational modeling A computational fluid dynamics (CFD) code, Flow-3D™, has been used for these investigations. Since the flow phenomena in running systems were mainly considered here, one fluid (i.e. liquid metal) with sharp interface tracking (i.e. free surface boundary tracking) of a VOF algorithm was employed [6]. Isothermal conditions were assumed since the change of viscosity of the liquid could be ignored as approximately only 1 per cent of heat loss was expected [7]. The boundary conditions of all models are the same and the runner systems with total head height of 300 mm are applied. Cubic cell with the size of 2mmx2mmx2mm are meshed. The total numbers of cells are around 1,257,000. In 2001, Midea [8] measured the pressure drop of water flowing through a ceramic foam filter with similar filter orientation in RAF runner system (cf., Figure 1). In Figure 2, the flow velocity correlated strongly when pressure drop is divided by filter thickness. All of correlations conform to the second order polynomial format of Forchheimer's equation, which is a quadratic function for the pressure drop normalized by filter thickness versus to flow velocity.

46

In Forchheimer's equation, the pressure drop (ΔΡ) through a rigid and homogeneous porous medium, such as a ceramic foam filter, is given by: ΔΡ μ p 2 Equation L k, L Where u is the flow velocity, L is the filter thickness, μ is viscosity, and p is flow density; kt and k2 are constants known as the Darcian and non-Darcian permeabilities respectively. Table 1 showed the permeability coefficients of commercial ceramic foam filters utilized in the modeling. Through the calculation in Forchhemier's equation, the permeability coefficients derived from the water experiment (e.g., Table 1) could be adjusted in an isothermal state of a liquid metal. Table 1 Darcian and non-Darcian permeability coefficients of commercial ceramic foam filters deriving from Figure 2. Foam filter type ΙΟρρί 20ppi 30 ppi

Darcian permeability coefficient. MlO'm2) 2.62 2.14 1.79

Non- Darcian permeability coefficient. k2 ( 10'1 mi 2.66 1.69 1.48

Figure 2 experimental data of the correlation of pressure drop and flow velocity [8]. Based on Ergun's equation, Innocentini et al [9] proposed equations to describe the permeability coefficients of a ceramic foam filter. That is: kx =aedc Equation 2 k2 = β ■ ε1 ■ dc Equation 3 where a and β are constants, ε is the porosity, and dc is the mean pore diameter of a ceramic foam filter. Also, the nominal pore count (i.e., ppi) is equivalent to the reciprocal value of this characteristic length dc. As the pore count increases, the mean pore diameter (dc) and the permeability coefficients (ki and k2) decrease. In their experimental results, a linear assumption is best fit to the data in the graph of the ppi versus to permeability coefficients(k! and k2). Trajectory method When the flow exits in the end of the horizontal runner, its trajectory has two components, horizontal and vertical (cf., Figure 3). If the air resistance during the fall is neglected, the horizontal distance (Dh) of this trajectory is: D„ = V„-1 Equation 4 where V0 is the launch velocity at the end ofthat channel. In addition, the distance of vertical component (Dv,) is therefore, 1 2 Equation 5 -gt then:t = 2 where g is gravitational acceleration. Substituting the value t in Equation 5 into that in

D,

Equation4,

thus: V0

Equation 6

47

Figure 3 The schematics of the trajectory method for determining the launch velocity V0. Metal weighing method The method of measuring flow rate is to measure the weight W of metal passed by a runner system in a given time /. Therefore, the actual flow rate in the exit of the runner system can be calculated as following equation: W Actual flow rate = A„V m = Equation 7 pxt where Ag is the cross-sectional area of actual flow stream, Vm is mean velocity in runner exit and p is the density of liquid metal. Casting experiment Sand mold material completely mixed with Phenolic-Urethane Cold Box (PUCB) binders was used in the casting experiment. The use of PUCB binders is around 1.8% of total sand weight for the resin and the curing agent, respectively. Aluminum alloy, A356 (Al-7%Si-Mg), was melt at 720±20°C. The molten alloy was then poured into a pouring basin, where a ceramic-stopper was placed at the entrance of sprue. The molten alloy was poured until the basin was full, signalled by the flow of the liquid overflowing from it. Subsequently, the stopper was abruptly lifted clear of the basin, permitting the start of filling. Additional molten alloy was poured continuously in order to maintain a constant head height of 300mm, designed in the runner system. The pouring lasted approximately 2.5s. Quantification procedure for an unknown foam filter Only RAF runner system (cf., Figure 1(a)) was used to quantify an unknown foam filter. In the quantification procedure, the following steps were carried out: 1. Using the permeability coefficients in Table 1, various specifications (ppi) of foam filters were modeled. 2. Using the trajectory method, the launch velocity for each foam filter was derived from the modeling result. 3. Plot three diagrams of the ppi value against to the reciprocal values of Darcian and nonDarcian permeabilities (1/ki and l/k2), and against to the launch velocity V0 predicted by modeling (called modeling velocity), respectively. Linear relationships were assumed in these three diagrams. 4. In the real casting of RAF system, the launch velocity (V0) is also estimated by the trajectory method.

48

5. Using the value of V0 in real casting, a specific ppi value for the unknown foam filter could be interpolated by the linear relationship between ppi and V0 in the modeling result. 6. Using this ppi value, a specific value of the reciprocal Darcian (1/ki) and non-Darcian permeability coefficients (l/k2) could be also obtained from the other two diagrams. 7. Using these coefficients, RAF system with this unknown filter could be modeled. Then, the launch velocity V0 in modeling result could be validated by the real casting experiment. Result Based on the procedures described previously for quantifying an unknown foam filter, three diagrams with linear relationships of reciprocal Darcian and non-Darcian permeabilities, and launch velocity of RAF system in terms of ppi value were plotted in Figure 4, respectively.

Figure 4 The linear relationships in terms of ppi values against to: (a) reciprocal Darcian permeability, (b) reciprocal non-Darcian permeability, and (c) the launch velocity of RAF system.

Figure 5 The trajectory and its launch velocity of RAF system at various time frames: (a) the casting experiment (note: the grid of 50mmx50mm), (b) the modeling result, and (c) the history data of the launch velocity V0 for these trajectories of modeling and casting experiments. In the real casting of RAF system with this unknown foam filter, the launch velocity was estimated by the trajectory method. Figure 5(a) and (b) showed the sequence of the flow issuing from the end of the runner at various time frames in casting and modeling experiments. Their launch velocity, V0, against to the time was also illustrated in Figure 5(c). It shows the averaged V„ is around 1.14 (m/s) in this real casting, as the trajectory flows uniformly at the period from 1.0 to 2.5 seconds (cf., Figure 5(c)). If the modeling velocity equal to that in the real casting (i.e., 1.14 m/s) were considered, the predicted ppi value of 28.6 could be interpolated from the linear relationship in Figure 4(c). Using this ppi value, the reciprocal Darcian and non-Darcian permeabilities for this foam filter then could be derived from the linear relationships in Figure 4 (a) and (b), respectively. Table 2 presented these reciprocal values for the 28.6 (ppi) foam filter obtained by interpolating the linear equation in Figure 4(a) and (b). Using these values of the 28.6 (ppi) filter in Table 2, the modeling result of the launch velocity of RAF system at various time frames was also shown in Figure 5(c). As a result, the average

49

modeling velocity is around 1.15 (m/s), which is very close to that in the real casting (i.e., 1.14 m/s). Therefore, this unknown foam filter could be predicted and validated by this quantification procedure. In the rest of modeling experiments, the 28.6 ppi foam filter was used. For all runner systems, the trajectory results of modeling and casting experiments were demonstrated in Figure 6. Averaged launch velocity of the trajectories from all runner systems in modeling and casting experiments were listed in Table 3. Table 2 The reciprocal values of the Darcian and non-Darcian permeability coefficients for the predicted 28.6 (ppi) foam filter. ^_^^ ^^ Foam filter type 28.6 ppi

reciprocal Darcian permeability, 1/k] (10' m*) 5.53

reciprocal Non- Darcian permeabiliry,l/k2 (m"1) 678.59

Table 3 the experimental results of averaged launch velocity, averaged volumetric flow rate, and actual cross-sectional area of the trajectory in all runner systems.

Figure 6(b) also showed the volumetric flow rate at various times recorded by an electrical weighing machine during the trajectory in casting experiment (note that the density of 2430 kg/m3 is considered at the melt temperature of 720°C, for the calculation of the flow rate in Equation 7). Averaged volumetric flow rate for all runner systems could then be found in Table 3. Comparing to RAF system, the volumetric flow rate normalized by that of RAF system was also presented in this table. (a)

!b)

Figure 6 The history data for: (a) the launch velocity V0 estimated by the casting experiment and the modeling; (b) the volumetric flow rate recorded by an electrical weighing machine. Discussion Using same ppi value of ceramic foam filters, the volumetric flow rate of ERTF system is 1.87 times ofthat of RAF system and this is the largest in Table 3. And, its averaged launch velocity of 0.86 m/s is the lowest. It implies that the foam filter in ERTF system is used in the most

50

efficient way in this study. In runner system design, it is very important to obtain the optimized use of a foam filter, which provides the greatest flow rate and at the same time reduces the flow speed. Thus, ERTF system could be recommended for designing a runner system in future.

Figure 7 The dead zones within the foam filters in (a)RAF, (b)RZF, (c)RTF, and (d) ERTF systems. Moreover, Figure 7 showed the dead zones within foam filters during pouring in the modeling. Because of various orientations in the runner systems, the size and the location of the dead zone in each filter are different. In Figure 7, ERTF system contains the smallest region of dead zone while RAF has the largest. As the volume and the velocity quantities of the dead zone increases in each system, its volumetric flow rate decreases (cf., Table 3). If die volumetric flow rate divided by its launch velocity is considered, the actual cross-sectional area of a trajectory (or named the trajectory area) could be derived. Table 3 showed the trajectory area and the ratio of this area over its original inlet cross-sectional area of the horizontal runner in each system. In all of systems, their original inlet area is the same (i.e., 2.25 xlO"4 m2). In Table 3, the actual trajectory area of ERTF system is 1.40 times greater than its original inlet area, because there is a special expanded region in the inlet area of the filter in this system. In contrary, there is no extra region in RTF system. The scope of the dead zone in RTF system is larger than that in ERTF (cf., Figure 7). In a result, the volumetric flow rate of RTF is low. However, in ERTF system, the designed outlet cross-sectional area of the exit runner is 40mmxl5mm (i.e., ö.OxlO"4 m2), which is nearly twice larger than its actual trajectory area (i.e., 3.14x 10"4 m2 in Table 3). Therefore, an empty region could be found at the top of the exit runner. For avoiding air entrapment during casting filling, it is necessary to cut down the excess region of this empty area. Also it is needed to re-produce the same trajectory area as designing the exit runner. This can insure that liquid metal contacts the walls of the exit runner fully. Table 5 the ratio of the original inlet area over the actual outlet area of the trajectory.

In this unknown ceramic foam filter, the ppi value of 28.6 was predicted by the quantification procedure. In a result, the launch velocity of the trajectory in the modeling is very close to that in

51

real casting experiment. However, the actual size of this foam filter is around 10-15 ppi if it is carefully tested. Considering the same effect on launch velocity, the predicted pore size (i.e., 28.6 ppi) is much smaller than its actual pore size (i.e., 10-15 ppi). This probably is because the liquid metal is solidified in some part of this filter. The viscosity of the liquid metal next to the solidifying regions would increase. This brings an extra viscous shear to the melt. Therefore, the total effect of this foam filter with smaller ppi (i.e., 10-15 ppi, the large pore size) is similar to that with greater ppi (i.e., 28.6 ppi, the small pore size). This kind of solidification effect on ceramic foam filter could not be detected from the water analogy experiment, where the data of the Darcian and non-Darcian coefficients are obtained (cf., Table 1). For that reason, some modifications are needed for utilizing the result of a water analogy experiment, when liquid metal is used as a medium and the thermal effect on a foam filter should be also considered. Conclusion 1.

2.

3. 4.

The quantification procedure could be used for identifying an unknown foam filter in terms of ppi value having the same effect on the launch velocity of the trajectories in the casting and the modeling experiments respectively. Launch velocity predicted by the modeling was validated by the casting experiment. Using the permeability coefficients obtained from water analogy experiment, the predicted pore size (i.e., 28.6 ppi) of this unknown filter is smaller than its real one (i.e., approximately 10-15 ppi). This is because extra drag derived from the solidification effect on the foam filter is not considered in the modeling. In all of four runner systems, the optimized and recommended runner system, providing the largest volumetric flow rate (i.e., 2.70X10"4 m3/s) and the lowest velocity (i.e., 0.86 m/s), is ERTF system. Actual cross-sectional area of a trajectory could be calculated from volumetric flow rate divided by launch velocity. For avoiding air entrapment during casting, it is necessary to tailor the cross-sectional area of the exit runner equal to its actual trajectory area.

Acknowledgements F-Y Hsu acknowledges the help of Mr. Hsin-Wei Wang and Mr. Po-Sen Chen, and the sponsorship of Lien-Ho (LCTC) Foundation in Taiwan (R.O.C.), with the project no. of 99NUU-06. 1. 2. 3. 4. 5. 6. 7. 8. 9.

References J. Campbell, Castings, (Butterworth-Heinemann, 1991), 32. N.R. Green and J. Campbell, "Statistical distributions of fracture strengths of cast A1-7SÌMg", Materials Science and Engineering, A173 (1993), 261-266. H. Hashemi and R. Raiszadeh, "Naturally-Pressurized Running Systems: the role of ceramic filters", Journal of applied Sciences, 9(11) (2009), 2115-2122. J.-C. Gebelin and M. R. Jolly, "Modelling filters in light alloy casting processes (or "What really happens when aluminium flows through a filter")", AFS transactions, (2002), 02-079. F.-Y. Hsu, M. R. Jolly and J. Campbell, "A multiple-gate runner system for gravity casting", Journal of Materials Processing Technology, 209 (2009), 5736-5750. Flow Science, manual, http://www.flow3d.com D.S. Richins and W.O. Wetmore, "Hydraulics applied to molten aluminum", ASME Trans., July, (1952), 725-732. A. C. Midea, "Pressure drop characteristics of iron filters", AFS Transactions, (2001), 01042. M. D. M. Innocentini et al, "Prediction of Ceramic Foams Permeability Using Ergun's Equation", Mat. Res., 2(4) (1999), 283-289.

52

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SIMULATION OF MACROSEGREGATION DURING DIRECTIONAL SOLIDIFICATION USING MESH ADAPTATION Udaya K. Sajja and Sergio D. Felicelli Department of Mechanical Engineering and Center for Advanced Vehicular Systems Mississippi State University, Mississippi State, MS 39762, U.S.A. Keywords: Macrosegregation, Freckles, Fractional Step Method, Mesh Adaptation Abstract Modeling the formation of macroscopic segregation channels during directional solidification processes has important applications in the casting industry. Computations that consider thermosolutal convection involve different length scales ranging from the small solute boundary layer at the dendrite tips to the characteristic size of the casting. In general numerical models of solidification in the presence of a developing mushy zone are computationally inefficient due to nonlinear transport in an anisotropie porous medium. In the present work, mesh adaptation with triangular finite elements is used in conjunction with an efficient fractional-step solver of the momentum equations to predict the occurrence of channel-type segregation defects or freckles. The triangulations are created dynamically using an unstructured grid generator and a refinement criterion that tracks the position of the channel segregates. The efficiency of mesh adaptation is illustrated with simulations showing channel formation and macrosegregation in directional solidification of a Pb-Sn alloy. Introduction The rejection rate of metal casting components due to macrosegregation defects is still a major problem for the casting industry. In the production of directionally solidified single crystal superalloy turbine blades for high temperature applications, the occurrence of equi-axed grain channels commonly known as freckle defects severely affects the yields. Since the pioneering work of Giamei and Kear [1], several experimental studies [2-4] provided the conclusive evidence that buoyancy driven thermosolutal convection is responsible for the formation of freckles. In the last four decades, numerous mathematical models were developed to predict macrosegregation in cast alloys. Earlier modeling efforts did not include the effect of thermosolutal convection and mainly concentrated on solute redistribution only. In late 1980s, models that include double diffusive convection and capable of capturing the channel segregates were developed [5-9]. In these so called single domain models, the mushy zone of the solidifying alloy was treated as a continuum porous medium with anisotropie permeability and a unique set of conservation equations that are valid in all the regions (solid, liquid and mushy) was developed based on either volume averaging or mixture theory. Numerical simulations showing the development of freckles by solving a set of mass, momentum, energy and solute conservation equations in a fixed numerical grid were reported [10-12].

53

Most of the previously used models solve the momentum equation as a coupled system of velocity and pressure which made them computationally inefficient. Solidification models that use fractional step or projection method for the solution of momentum equation in binary alloy solidification problems were presented in [13-16]. In particular, [15, 16] presented the use of fractional step method for the simulation of freckle defects that occur in directional solidification processes. A large three dimensional calculation presented in [16], showed evidence that considerable improvement in the efficiency of macrosegregation computations is possible with the use of fractional step method for the solution of momentum equation. Macrosegregation computations that consider double diffusive convection involve different length scales; a region close to the solidification front where the fluid velocity is significant and the characteristic size of the entire solidifying domain. To counter the difficulties associated with different length scales, adaptive discretization of space is needed. Recently, articles on the use of adaptive mesh refinement to predict channel segregation began to appear in the literature. An adaptive domain decomposition method that results in non-conforming rectangular finite element meshes was employed in [17]. The initial coarse mesh was refined based on the volume fraction of liquid at the nodes. All the elements having a node for which the fraction of solid falls within an arbitrarily chosen range were refined. The governing equations were then solved on different meshes using an iterative method and non-conforming discretizations were dealt with the mortar technique. In [18], an unstructured mesh with linear triangular finite elements together with a mesh refinement criterion based on the norm of the gradient of solid fraction was used, while the momentum equation was solved using Galerkin Least Squares (GLS) method which requires the coupled solution of velocity and pressure. In the present work, mesh adaptation with linear triangular finite elements in conjunction with an efficient fractional-step solver of the momentum equations is used to predict freckle defects that can occur during directional solidification processes. The meshes are generated using the unstructured mesh generator AFLR2 [19, 20], which is based on advancing front type iterative point placement and local reconnection technique. Simulations showing the occurrence of freckles during the directional solidification of a binary Pb-Sn alloy are presented and the performance of the approach is analyzed. Mathematical Model The mathematical model presented below is a continuum model capable of simulating solidification of binary alloys. The basic assumptions are: (1) The liquid is Newtonian and incompressible, the flow is Darcy type and laminar. (2) Only solid and liquid phases are present, no pores form. (3) Solid phase is stationary and there is no solute diffusion in the solid. (4) Density and specific heat can be different in the solid and liquid phases, but constant in each phase. Boussinesq approximation is made in the buoyancy term of the momentum equation. (5) The mushy zone is treated as a porous medium with anisotropie permeability.

54

With these assumptions, the volume averaged conservation equations of mass, momentum, energy and solute components can be derived [8, 9]. In non-dimensional form they are: Continuity equation: H

(i)

dt

Momentum equation: 5u 1 _ βδφ _ 1 f_2 β_δφ — + - u V u + — — u = -éVp + — V u + —V — dt φ φ dt Re^ 3 dt -Da u + É J_ T + Re Fr

{signßs){St-X) (2)

Energy equation: pcP-

δΤ +

P (cp-l)(TH-T)

+ -

ÊÊ. + u.VT dt

=

J—v*r

PrRe

(3)

Solute conservation equation:

>-£—°·->ψ·*ύ?-™

(4)

In the above equations, u denotes the velocity vector, ß = ——— is the contraction coefficient, A

with ps and p( as the solid and liquid densities, respectively; φ denotes the volume fraction of liquid and t is time. The pressure is denoted by p ; 7" and S, denote local temperature and local solute concentration in the liquid respectively. pcp = pcp{\ — φ) + φ is the non-dimensional heat capacity of the solid plus fluid mixture, where p P(

Cpt

, with c„.. and c„

the solid and liquid specific heats respectively. The solid-liquid mixture concentration is pS = pSs(l —φ) + S$, where Ss is the local solute concentration in the solid phase. The equations have been non-dimensionalized using a reference length H related to the primary dendrite arm spacing; the reference flow speed U = JßsgS0H , where /?, is the solutal coefficient of volumetric expansion, g is the magnitude of the gravitational constant and S0 is H the initial and reference solute concentration; the reference time is τ = — ; the reference U

55

pressure is p0= p(U , the reference temperature is T0, the reference enthalpy temperature is T„ and the reference temperature gradient is G . The model is closed with the liquidus curve of the phase diagram which is assumed to be the linear relation T = mS(+Tm. With the assumption of no diffusion in the solid, the solid solute concentration is given by 1 M „ tnSa ύ< = \ ,Κο,αφ, where m = is the non-dimensional slope of the liquidus line and k s l \-φ>* GH is the partition coefficient. The non-dimensional parameters are: π

Reynolds number Re =

VH

v

,

1 Darcy coefficient Da = —rK, H2

Froude number rr = Stefan number St =

v

U2

r-,

ßTgGH2

GHcp( — L v

Prandtl number Pr = —, Schmidt number Sc = a Ds In the above expressions v is the kinematic viscosity of the fluid phase, /?, is the thermal coefficient of volumetric expansion, K is the permeability, L is the latent heat at the reference temperature TH, a is the thermal diffusion coefficient and Ds is the solutal diffusion coefficient. The permeability tensor K has been assumed to be diagonal and has been obtained from empirical data and numerical calculations as a function of the volume fraction of liquid φ and the primary dendrite arm spacing rf, [21, 22], The fractional step method is used for the solution of the momentum equation. In this method, the velocity components are solved explicitly and only the pressure equation is solved implicitly. This semi explicit formulation of the momentum equation results in improved computational efficiency and less memory requirements. Full details of the formulation are given in [16], and will not be repeated here. Adaptive Meshing Solidification models that consider thermosolutal or double diffusive convection and channel formation, involve different length scales. A very small solute boundary layer develops ahead of the solidification front due to large (several orders of magnitude) differences in thermal and solute diffusivities. Hence, for accurate macrosegregation computations, proper resolution of the fluid flow close to the tip of the dendrites in the mush zone and in the liquid just ahead of the solidification front is very important. This necessitates the use of very fine meshes in those critical regions. The use of mesh adaptation enables accurate prediction of macrosegregation while reducing the computational cost. In the present work, an adaptive meshing scheme based on linear triangular elements is used. In the mushy zone, only the regions where the liquid metal can still flow leading to the formation of channels and those close to the solidification front are

56

discretized with fine meshes. Deep in the mushy zone, the flow velocities are one or more orders of magnitude smaller compared to the velocities in critical regions and can easily be captured by the coarse mesh. It is known that the channel regions are the last regions to solidify in the whole solidification domain. Hence the gradient of the volume fraction of liquid is also used as an additional criterion for identifying the regions at which finer discretization is needed. In this work, the meshes are generated using an efficient unstructured grid generator AFLR2 developed at Mississippi State University [19, 20]. In this mesh generator, the points (nodes) are created iteratively at the desired spacing using an advancing front type point generation algorithm and the connectivity is optimized locally based on a quality criterion such as minimizing the maximum angle or maximizing the minimum angle. In order to generate the solution-adapted meshes, this mesh generator allows the use of mesh adaptation sources. At these adaptation sources, the point distribution function (or length scale) of the standard automatic point generation algorithm can be modified to a smaller desired value. The mesh is allowed to grow from these adaptation sources, with the value of the point distribution function varying between the small and a predefined large value with a specified growth rate. This algorithm produces a smooth adaptive mesh. In the present problem, initially a coarse mesh with a uniform spacing is generated by prescribing a larger point distribution function (or length scale) for the whole computational domain. Once the solution with the coarse mesh is obtained, the nodes that satisfy the following criteria are chosen for identifying the regions where fine space discretization is needed: 1) the nodes in the mushy zone at which the volume fraction of liquid is greater than a prescribed value and less than one; 2) the nodes at which the gradient of volume fraction of liquid is higher than a prescribed value. All these nodes act as the adaptation sources for the mesh generation algorithm and a smooth solution adapted mesh will be generated. The generated meshes are updated periodically at fixed intervals of solidification time chosen as a function of the solidification speed determined by the cooling rate. Whenever the mesh is updated, the solution variables at the new mesh nodes are interpolated from the corresponding old mesh data. Numerical Results The solidification model with the projection method for solving the momentum equation is implemented by means of a stabilized Petrov-Galerkin formulation based on solution adapted linear triangular finite elements. Numerical simulations for directional solidification of a binary Pb-Sn (23wt.pct) alloy were performed in a two dimensional domain of 30 mm x 50 mm. The domain is enclosed by solid walls with no-slip boundary conditions on all surfaces. Initially, the alloy is all liquid with temperature varying linearly from 546.5 K at the bottom to 596.5 K at the top. The lateral walls are insulated and a constant temperature gradient of 1000 K/m was imposed at the top. At the bottom boundary, the temperature varies with time according to the prescribed cooling rate 1.0 K/min. The properties of the alloy are listed in Table 1. The simulation starts with a coarse mesh of uniform spacing 1 mm. To perform mesh adaptation, the nodes in the mushy zone with a volume fraction of liquid greater than 0.9 and also the vertices of the elements for which the gradient of volume fraction of liquid exceeds 500 m"' are identified as the adaptation sources described above. At these adaptation sources, a value of 0.2 mm is prescribed as the length scale which is the desired smallest elements' size. The mesh starts

57

Figure 1: Directional solidification of Pb-23wt%Sn alloy with adaptive finite elements. Finite element mesh, contours of volume fraction of liquid and mixture solute concentration of Sn after 1165s.

Figure 2: Detail of mesh refinement showing contours of volume fraction of liquid and solute concentration of Sn after 1165s. to grow smoothly from these nodes with a specified growth rate such that the maximum spacing does not exceed the initial coarse spacing of 1 mm at the farthest regions. With this criterion for adaptation, the channel regions and the regions near the solidification front are discretized with fine elements. These channel regions are enriched in solute concentration and are surrounded by solute depleted regions. Figure 1 shows the adapted mesh together with the contours of volume fraction of liquid and solute concentration after 1165 s of solidification time. The result shows the development of the channels along each wall of the domain. In Figure 2, magnified views of the critical regions where the mesh is refined are shown along with the contours of volume fraction of liquid and solute concentration with superimposed discretization. It can be observed that the refinement follows the evolution of freckles quite well.

58

Conclusions A numerical model that can predict the occurrence of freckle defects in directionally solidified castings has been presented. Adaptive re-meshing with linear triangular elements has been used in conjunction with Galerkin finite element method. A simple mesh adaptation strategy that can produce a fine mesh in the critical regions and coarser mesh in other regions has been implemented. The fluid flow equations are solved using an efficient fractional step formulation. Simulation results of the binary Pb-Sn alloy solidification demonstrate the ability of the method to capture the evolution of freckle defects. The method is currently being improved with the goal of capturing freckles in simulations of large castings with complex geometries. Table 1: Thermodynamic and transport properties of Pb-23 wt%Sn alloy used in calculations Reference concentrations (wt pet): Sn=23

Specific heat of liquid (J kg"1 K"1): cpt = 190

Reference temperature (K): TR = 546.494

Specific heat of solid (J kg"1 K"1): cps = 160

Eutectic temperature (K): TE=456

Latent heat of fusion (J kg"1): Z, = 3.76xl0 5 Density of liquid (kg m"3): p ; = 8800

Temperature at which latent heat is given (K) = 528 1

Density of solid (kg m"3): ps = 9700

Thermal expansion coefficient (K" ): βτ = -1.2x10^" 1

Solutal expansion coefficients (wt pet" )ßc = -5.15x10 Thermal conductivity of liquid (W K"1 m"1): κι = 18.4 Thermal conductivity of solid (W K"1 m"1): Ks = 36.8

3

Viscosity (Ns m"2): μ = 2.1736χ10~3 Equilibrium partition ratio: 0.31 Slope of liquidus: -2.32633

Solute diffusivity in liquid (mV 1 ): D = 3x10"9 Melting temperature of the pure substance (K): 600 :200 Acknowledgements This work was funded by the National Science Foundation through grant number CTS-0553570. Support with the AFLR mesh generation software by Prof. David Marcum at Mississippi State University is gratefully appreciated. References 1. A.F. Giamei and B.H. Kear, "On the Nature of Freckles in Nickel Base Superalloys," Metallurgical Transactions, 1 (1970), 2185-2192. 2. S.M. Copely, A.F. Giamei, S.M. Johnson, and M.F. Hornbecker, "The Origin of Freckles in Unidirectionally Solidified Castings," Metallurgical Transactions, 1 (1970), 2193-2204. 3. J.R. Sarazin and A. Hellawell, "Channel Formation in Pb-Sn-Sb Alloy Ingots and Comparison with the System NH4-C1-H20," Metallurgical Transactions A, 19 (1988), 1861-1871. 4. S.N. Tewari, R. Shah, and M.A. Chopra, "Thermosolutal Convection and Macrosegregation," Metallurgical Transactions A, 24 (1988), 1661-1669.

59

5. W.D. Bennon and F.P. Incropera, "A continuum model for momentum, heat and species transport in binary solid-liquid phase change systems-I. Model formulation," International Journal of Heat and Mass Transfer, 30 (1987), 2161-2170. 6. W.D. Bennon and F.P. Incropera, "A continuum model for momentum, heat and species transport in binary solid-liquid phase change systems-II; applications to solidification in a rectangular cavity," Internationaljournal of Heat and Mass Transfer, 30 (1987), 2171-2187 7. C. Beckermann and R. Viskanta, "Double diffusive convection during dendritic solidification of a binary mixture," PhysicoChemical Hydrodynamics, 10 (2) (1988), 195-213. 8. S. Ganesan and D.R. Poirier, "Conservation of mass and momentum for the flow of interdendritic liquid during solidification," Metallurgical Transactions B, 21 (1990), 173-181. 9. D.R. Poirier, P.J. Nandapurkar, and S. Ganesan, "The energy and solute conservation equations for dendritic solidification," Metallurgical Transactions B, 22 (6) (1991), 889-900. 10. S.D. Felicelli, J.C. Heinrich, and D.R. Poirier, "Simulation of Freckles during Vertical Solidification of Binary Alloys," Metallurgical Transactions B, 22 (1991), 847-859. 11. H. Combeau and G. Lesoult, "Simulation of freckles formation and related segregation during directional solidification of metallic alloys," Modelling of Casting, Welding and Advanced Solidification Processes IV. The Minerals, Metals & Materials Society, Warrendale Pennsylvania (1993) 201-208. 12. J. Guo, and C. Beckermann, "Three-dimensional Simulation of Freckle Formation during Binary Alloy Solidification: Effect of Mesh Spacing," Numerical Heat Transfer A, 44 (2003), 559-576. 13. G. Amberg, "Computation of macrosegregation in an iron-carbon cast," International Journal of Heat Mass Transfer, 34 (1) (1991), 217-227. 14. D. Xu and Q. Li, "Numerical method for solution of strongly coupled binary alloy solidification problems," Numerical Heat Transfer A, 20 (1991), 181-201. 15. D. G. Westra, "Simulation of directional solidification in a binary alloy using the fractional step method" (PhD. Dissertation, The University of Arizona, Department of Aerospace and Mechanical Engineering. 2003). 16. J.C. Heinrich, U.K. Sajja, S.D. Felicelli and D.G. Westra, "Projection method for flows with large local density gradients: Application to dendritic solidification," International Journal for Numerical Methods in Fluids, 57 (2008), 1211-1226. 17. U.T. Kämpfer and M. Rappaz, "Modelling of macrosegregation during solidification processes using an adaptive domain decomposition method," Modelling and Simulation in Material Science and Engineering, 11 (2003), 575-597. 18. W. Liu, C. Xie, M. Bellet, and H. Combeau, "2-Dimensional FEM modeling of macrosegregation in the directional solidification with mesh adaptation," Ada Metallurgica Sinica (English Letters), 22 (2009), 233-240. 19. D.L. Marcum and N.P. Weatherill, "Unstructured grid generation using iterative point insertion and local reconnection," AIAA Journal, 33 (1995), 1619-1625. 20. D.L. Marcum and N.P. Weatherill, "A procedure for efficient generation of solution adapted unstructured grids," Computer Methods in Applied Mechanics and Engineering, 127 (1995), 259-268. 21. S. Ganesan, C.L. Chan and D.R. Poirier, "Permeability of flow parallel to dendrite arms," Material Science Engineering A, 151 (1992), 97-105. 22. M.S. Bhat, D.R. Poirier and J.C. Heinrich, "Permeability for cross flow through columnardendritic alloys," Metallurgical and Material Transactions B, 26 (1995), 1049-1056.

60

Shape Casting: The 4Ih International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

A Mathematical Model for Simulating the Microporosity of Squeeze Casting of Aluminum Alloy Zhiqiang Han1, Jinxi Li1, Wen Yang1, Baicheng Liu'· 2 1 Key Laboratory for Advanced Materials Processing Technology (Ministry of Education), Department of Mechanical Engineering, Tsinghua University, Beijing 100084 2 State Key Laboratory of Automotive Safety and Energy, Department of Automotive Engineering, Tsinghua University, Beijing 100084 Keywords: Aluminum Alloy, Squeeze Casting, Microporosity, Modeling and Simulation Abstract A mathematical model for simulating the microporosity of squeeze casting of aluminum alloy has been developed, in which the heat transfer, solidification shrinkage, feeding flow, pressure transfer, and hydrogen conservation were taken into account. The shrinkage induced flow and the pressure drop in the mushy zone were calculated by solving mass and momentum conservation equations. A mechanical model was solved for obtaining the pressure transferred into the central area of the casting. By coupling the pressure drop with the pressure transferred into the central area, the pressure in the mushy zone was calculated. Based on the hydrogen conservation equation, the microporosity volume fraction was estimated by referring to the pressure in the mushy zone. The squeeze casting processes of aluminum alloy under different process conditions were simulated and the simulation results agree well with experimental results. Introduction Squeeze casting is an advanced metal processing technology where solidification is promoted under a high pressure to produce castings with compact interior and excellent mechanical properties [1-3]. It is important to appropriately control the process to avoid the formation of porosity in the casting as the existence of porosity significantly reduces the tensile strength, elongation, fatigue strength, and toughness of the castings [4-6], which fails the attempts to improving casting quality through the squeeze process. Porosity is resulted from volume shrinkage caused by cooling and phase change as well as the precipitation of dissolved gas in the liquid metals. In the squeeze casting, the pressure transferred into the casting decreases with the increasing of the thickness of the solidified shell, and when the pressure inside the casting drops to a certain extent, porosity may form due to insufficient feeding and gas precipitation. The scope of this research lies in modeling and simulation on microporosity of squeeze casting to understand the effect of process parameters on the formation of microporosity. Some efforts have been made on the prediction of microporosity [7-19]. Lee et al [7] and Stefanescu [8] made comprehensive reviews on the research work of predicting microporosity. Kubo and Pehlke [9] proposed an interdendritic flow model, and then many authors made contributions in this field [10-14]. Lee et al [15] developed a pore growth model based upon the diffusion limited growth of pores. Then, they developed a model of pore formation coupled with microstructure simulation using Cellular Automaton method [16, 17]. Backer et al [18] combined the interdendritic flow model with the pore growth model to improve the precision of microporosity prediction. Carlson et al [19] analyzed microporosity nucleation and growth by a

61

volume-average model considering the local, finite-rate diffusion of dissolved hydrogen in the liquid towards the pores, and with that calculated the volume fraction of microporosity. However, the reported work mainly focused on the solidification process under normal pressure. There is few work reported on the microporosity modeling of squeeze casting process. We conducted research on modeling and simulation of microporosity in squeeze casting, aiming at developing a simulation tool facilitating the analysis and optimization of the process design. Mathematical Model Microporosity forms due to volume shrinkage and gas precipitation during solidification. Local volume shrinkage resulted from cooling and solidification in the mushy zone induces feeding flow, and a local pressure drop develops when the volume shrinkage cannot be fully fed. As a result of local pressure drop, the solubility of hydrogen in the liquid reduces and hydrogen precipitation takes place. Moreover, with the solidification carrying on, hydrogen concentration in the liquid of mushy zone increases since the solubility of hydrogen in solid is far lower than that in liquid. It is well understood that the volume shrinkage and hydrogen precipitation are the primary factors resulting in microporosity defect in aluminum alloys. In squeeze casting process, the applied pressure is fully or partially transferred into the casting and creates a pressure distribution inside the component, which has a crucial effect on the microporosity. Hence, the following physics must be taken into account in the modeling: (1) the local pressure drop in the mushy zone induced by the volume shrinkage and feeding difficulties, (2) the transfer of the applied pressure during the solidification, and (3) the conservation of hydrogen and the formation of microporosity. 2.1 Conservation Equations The following assumptions are introduced into the present model: (1) the densities of solid and liquid phase are constant, (2) the liquid flow is lamellar flow and the viscosity of liquid phase is constant, (3) the effect of flow on the thermal field is not considered, and (4) the precipitated gas phase does not move. Base on the above assumptions, the conservation equations of energy, mass, momentum and hydrogen are as follows. The energy conservation: riH p— = V.(iNT) (1) at where T is temperature, κ is thermal conductivity, t is time, H is enthalpy, P = P,g,+ Ptëi+ Ppëpls the average density, where g is volume fraction and the subscripts s,l,p

mean solid, liquid and gas phases, respectively. The mass conservation:

^ + V.(p,F) = 0 (2) ot where V = g,Vl is superficial velocity, g, is the volume fraction of liquid, V, is the velocity of liquid. The density of hydrogen is far less than that of the liquid metal, so we have P = P,g,+P,g, This equation can be further simplified as follows, V.F = - 0 ^ L + - ^ (3) at dt

62

whereß = (p3 - p , ) / p , gives the solidification shrinkage rate. The momentum conservation:

p,^+p,v.i-^J=//iv 2 K+Iv(v.F)j-g / |-F-g / v/'+pg,i

(4)

where P is pressure, μ is viscosity, g i s gravity acceleration, K is permeability described as follows Λ

=-ΐ—^-,-

(5)

180 (1-g,) 2 where λ^ is the secondary dendrite arm spacing. The hydrogen conservation: P,g,C'H + P,g,C'„ + P.g.Cf, = pCl (6) where C°H is the initial hydrogen concentration, C'H, C\, and CH are the concentration of solid, liquid and gas phases, relatively. 2.2 MicroDorositv Formation The formation of pores is judged by using the following criterion: where PG is the gas precipitation pressure, P is the local pressure in the casting during squeeze process, Plhr is reduced pressure caused by solidification shrinkage, Ργ = 2γ/τρ is the additional pressure caused by interfacial energy, where y is the gas-liquid interfacial energy and r is the pore radius. The relationship between the gas precipitation pressure and the hydrogen concentration is described by the Sievert law, (8) C'H=Kiy[p^ where K, is a coefficient depending on alloy composition and temperature and can be calculated by using the following equation [20]: K,=K.lfH (9) where Ke is a parameter depending upon temperature. For aluminum alloy [20]: .og,^=-^*-1.32 / „ is a parameter depending on alloy composition [20]: log,„/ H =Z[ e ;iC; v +^(C, A -) 2 ] Λ'

(„)

where the superscript X denotes alloying element, Cf is the concentration of X, ef,, r^ are the impact factors of X on hydrogen solubility. When the pore has formed, the pressure inside the pore can be described by Pa=P^-P,*r+P,

63

(12)

Normally, microporosity forms in the dendritic array, so it is assumed that the size of the pores equals to the secondary dendrite arm spacing, ! ■ „ = —

(13)

Λ=[(Λ°)3+Λ^

(14)

' 2 where the secondary dendrite arm spacing Xj can be described by the following coarsening model:

where λ°2 is the initial secondary dendrite arm spacing, M is the coarsening coefficient and tf is the solidification time. Numerical Algorithm 3.1 Finite Element Equation The model was solved by using finite element method. The details of finite element discretization and solution method of the energy equation can be found elsewhere [21]. Standard Galerkin method was used for the discretization of the mass and momentum conservation equations. In order to ensure the stability and convergence of the equations, mixed interpolation method was used, where quadratic interpolation function was used for velocity and linear interpolation was used for pressure. The space-discretized equation may be written in the following matrix form ~MU

0

0

0 0

Mv 0

0 0

u V

~Km Km + Km K„

P

cl cl

c„~ II

c. 0

V

P

\F"

= Fv

U

(15)

where M is the mass matrix, K is the velocity stiffness matrix which contains the advection and viscous terms, CT is the divergence matrix and F denotes the force vector. A fully implicit (backward Euler) method was used for the temporal discretization, finally yielding a system of nonlinear equations. A detailed description of the discretization and solution can be found in [22], By solving the equation, the pressure difference in the mushy zone as well as the feeding flow can be obtained. 3.2 Coupling with the Mechanical Model During the squeeze casting process, a thin solid shell forms immediately after the pressure has been applied on the metal by a punch or an upper die. The punch or the die squeezes the casting and keeps the solid shell continuously deformed, and a static pressure develops in the liquid or mushy core of the casting. At the same time, volume shrinkage happens near the outer part of the casting and feeding flow from the center to the outer part takes place. The static pressure inside the casting was calculated by using a mathematical model developed earlier [21, 23] for describing the deformation and stress of squeeze casting. On the other hand, the pressure difference in the mushy zone was calculated by solving the mass and momentum conservation equations. The absolute pressure value in the mushy zone was determined by referring both the

64

static pressure and the pressure difference, which was used for judging whether a pore forms and for calculating the volume fraction of porosity. 3.3 Calculation of Volume Fraction of Microporositv When the porosity has not formed ( gp = 0 ), the volume fraction of solid phase can be calculated directly by g =

J

i— =

K+V,

=

· ^'

f./p.+f./p,

il!-L

(16)

f,P,+f,P,

where Vs and V, denote the volume of solid and liquid phases. The mass fraction of solid phase fs can be calculated according to temperature. The hydrogen redistributes during solidification: C'H=kHC'„

(17)

where kH is the partition coefficient of hydrogen. By using Eq. (6) and Eq. (17), the hydrogen concentration of liquid phase can be calculated (18)

C'„ = pC°H/(kHp,g,+plgl)

By using the Sievert law, the hydrogen precipitation pressure can be calculated and used to judge whether the porosity forms based on the criterion defined by Eq. (7). When the porosity has formed, the initial volume fractions of solid and liquid phases, g] and g* can be calculated by .

V

1

V,

ν,+ν,'>gi = v,,s+v,,.

(19)

The relationship between gt andgj, as well as g, andg*, is as follows:

ν. + ν, + ν,

'

v,+v, + vP

(20)

The density of porosity can be calculated based on the ideal gas assumption: P p=a— p Ύ

(21)

r

where a is a constant. As only hydrogen is considered, the hydrogen concentration in the pore is CH = 1. By substituting Eq. (6) for Eq. (8), Eq. (20) and Eq. (21), the volume fraction of microporosity can be calculated as follows "

{p,g',+p,g',)c°H -{kHp,gl '

apG/T+(PX

+psg;)c°H-(kHp,g;

+p,gl)K,4^ +p,gl)K,4FG

(22)

When the solid fraction reaches a certain extent, the dendrites grow into a close skeleton, which makes the liquid and solid phases separated from each other. As a result, the fraction of microporosity calculated using pressure drop in the mushy zone has a large deflection from the

65

experimental data. In this paper, it is assumed that the feeding flow is cut off when the solid fraction is higher than a critical value g and the volume fraction of porosity can be calculated as follows gP=g'P+{\-gAi-g'P)ß

(23)

where g' is the volume fraction of microporosity wheng 5 = glc. Simulation Examples A casting of A3 5 6 aluminum alloy with symmetric geometry was simulated. Fig. 1 shows the casting geometry and simulated results. The vertical edge at the left is the symmetric line of the casting. In the figure shows the calculated velocity of feeding flow and the pressure distribution in the castings with a pressure of 40MPa applied. The pressure distribution is not shown in the area that has already solidified in order to display the pressure change in an appropriate scale for the mushy zone. As shown in Fig. 1, when the pressure has been applied on the casting for 10s, the upper and right part of the casting has almost solidified, and the liquid metal flows from the central area to the mushy zone adjacent to the solidified part. The casting with low die temperature solidifies more quickly than the casting with high die temperature. The pressure in the center of the casting with low die temperature is approximately 9.5MPa while in the casting with high die temperature is 18.5MPa, as shown in Fig. 1 (a) and (b). At the time of 15s after the pressure is applied, most part of the casting with low die temperature has solidified, and the pressure in the center is around 0, which implies that the applied pressure cannot be transferred into the casting effectively whereas in the casting with high die temperature, a large portion of the central area has not solidified yet, where a pressure of 14MPa still exists, as shown in Fig. 1 (c) and (d).

Fig. 1 The calculated flow velocity and pressure distribution in the casting at 10 and 15s after the pressure is applied. The process parameters: applied pressure 40MPa, punch temperature 100"C, die temperature 150'C and ejector temperature 150'C for (a) and (c); applied pressure 40MPa, punch temperature 200°C, die temperature 250°C and ejector temperature 250°C for (b) and (d). Fig. 2 shows the calculated feeding flow velocity and pressure distribution in the castings with pressures of 40 and 60MPa applied. The pressure in the central area of the casting with a pressure of 60MPa applied is approximately 18MPa while that with a pressure of 40MPa applied is about lOMPa. It is shown in Fig. 2(c) that at the time of 22s after the pressure is applied the pressure of the unsolidified area is inadequate for preventing the porosity from being formed for the casting with a pressure of 40MPa applied. In contrast, in the casting with a pressure of 60MPa applied, a pressure of 8MPa in the unsolidified area still exists. Fig. 3 shows the calculated results of microporosity in the casting with different process parameters. In the case with low die temperature, a pressure of 40MPa is not sufficient to ensure a sound casting. There is some microporosity in the hot spot part of the casting, see Fig. 3(a). The casting with higher die temperature has less microporosity, as shown in Fig. 3(b), indicating

66

that an appropriate increasing of the die temperature is helpful to the pressure transfer and further to microporosity reduction. When the applied pressure increases to 60MPa, there is almost no obvious porosity in the whole casting. These results agree well with experimental results.

Fig. 2 The calculated flow velocity and pressure distribution in the casting at 18 and 22s after the pressure is applied. The process parameters: applied pressure 40MPa, punch temperature 200°C, die temperature 250°C and ejector temperature 250'C for (a) and (c); applied pressure 60MPa, punch temperature 200'C, die temperature 250°C and ejector temperature 250'C for (b) and (d).

(a) (b) (c) Fig. 3 The calculated volume fraction of microporosity in the castings. The process parameters: (a) applied pressure 40MPa, punch temperature 100 "C, die temperature 150"C and ejector temperature 150'C; (b) applied pressure 40MPa, punch temperature 200°C, die temperature 250 'C and ejector temperature 250*C; (c) applied pressure 60MPa, punch temperature 200'C, die temperature 250'C and ejector temperature 250°C. Conclusions Based on the understanding on the formation mechanism of microporosity, a mathematical model for simulating the microporosity of squeeze casting of aluminum alloy has been developed, in which the heat transfer, solidification shrinkage, feeding flow, pressure transfer, and hydrogen conservation were taken into account. The shrinkage induced flow and the pressure drop in the mushy zone were calculated by solving the mass and momentum conservation equations. A mechanical model was solved for obtaining the pressure transferred into the central area of the casting. By coupling the pressure drop with the pressure transferred into the central area, the pressure in the mushy zone was calculated. Based on the hydrogen conservation equation, the microporosity volume fraction was estimated by referring to the pressure in the mushy zone. The squeeze casting of aluminum alloy under different process conditions was simulated and the simulation results agree well with experimental results. Acknowledgement The research work is funded by the National Natural Science Foundation of China (No. 50675113 and No.50875143). One of the authors ZQHAN also would like to appreciate the support of the

67

Scientific Research Foundation for the Returned Overseas Chinese Scholars, Ministry of Education of China, and the support of State Key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and Technology. References [1] M. R. Ghomashchi, A. Vikhrov, Journal of Material Processing Technology, 2000, Vol.101 (l),ppl-9. [2] P. X. Qi, Special Casting and Nonferrous Alloys, 1998, Vol.4, pp 32-36. [3] S. J. Luo, B. G. Chen, P. X. Qi, Liquid Forging and Squeeze Casting Technology, Beijing: Chemical Industry Press, 2007, pp 1. [4] J. F. Major, AFS Transactions, 1998, Vol.105, pp 901-906. [5] A. M. Samuel, F. H. Samuel, Metallurgical Transactions A, 1995, Vol.26, pp 2359-2372. [6] C. D. Lee, Materials Science and Engineering A, 2007, Vol.464, pp 249-254. [7] P. D. Lee, A. Chirazi, D. See, Journal of Light Metals, 2001 (1), pp 15-30. [8] D. M. Stefanescu, International Journal of Cast Metals Research, 2005, Vol.l8(3), pp 129143. [9] K. Kubo, R. D. Pehlke, Metallurgical Transactions B, 1985, Vol.16, pp 359-366. [10]D. R. Poirier, K.. Yeum, A. L. Maples, Metallurgical Transactions A, 1987, Vol.18, pp 19791987. [11]S. Shivkumar, D. Apelian, J. Zou, AFS Transactions, 1990, Vol.98, pp 897-904. [12]A. S. Sabau, S. Viswanathan, Metallurgical and Materials Transactions B, 2002, Vol.33(2), pp 243-255. [13]C. Pequet, M. Gremaud, M. Rappaz, Metallurgical and Materials Transactions A, 2002, Vol.33, pp 2095-2106. [14]H. D. Zhao, C. Z. Wu, Y. Y. Li, I. Ohnaka, Acta Metallurgica Sinica, 2008, Vol.44(ll), pp 1340-1347. [15]R. C. Atwood, S. Sridhar, W. Zhang, P. D. Lee, Acta Metallurgica, 2000, Vol.48(2), pp 405417. [16] P. D. Lee, A. Chirazi, R. C. Atwood, W. Wang, Materials Science and Engineering A, 2004, Vol.365, pp 57-65. [17]J. S. Wang, P. D. Lee, International Journal of Cast Metals Research, 2007, Vol.20(3), pp 151-158. [18]G. Backer, Q. G. Wang, Metallurgical and Materials Transactions B, 2007, Vol.38, pp 533540. [19]K. D. Carlson, Z. P. Lin, C. Beckermann, Metallurgical and Materials Transactions B, 2007, Vol.38, pp 541-555. [20] P. N. Anyalebechi, Acta Metallurgica, 1995, Vol.33, pp 1209-1216. [21]Z. Q. Han, W. Zhu, B. C. Liu, Acta Metallurgica Sinica., 2009, Vol.45, pp 356-362. [22] R. W. Lewis, Z. Q. Han, D. T. Gethin, Comptes Rendus Mécanique, 2007, Vol.335, No.5-6, pp 287-294. [23] W. Zhu, Z. Q. Han, B. C. Liu, Acta Metallurgica Sinica, 2009, Vol.45, pp 363-368.

68

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry

Solidification Session Chairs: William Griffiths Peter Schumacher

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

REVIEW OF DEFECT BEHAVIOR IN Ni-BASED SUPERALLOYS John Campbell Emeritus Professor, Metallurgy and Materials, University of Birmingham, B15 2TT, UK, je @ campbelltech.co.uk. Abstract The Ni-base superalloys, normally melted and cast in vacuum, entrain their surface oxide film during turbulent pouring of the melt, which unfortunately at this time, is universally practiced for investment castings of these materials. The entrained film automatically becomes a bifilm crack, so that cast alloys have a large population of cracks that controls their failure behavior. The problems of the growth of single crystals, and the welding of polycrystalline alloys are reviewed to illustrate the central role of bifilms in the cracking of turbine blades, the heat affected zones of welds and the reliability of properties. It has been demonstrated that improved gravity pouring systems can significantly reduce these problems, but only counter-gravity filling of molds is expected to result in defect-free castings. Key words: Ni superalloys; cracks; single crystals; welds; HAZ. Introduction There is a growing body of evidence that Ni base superalloys harbour cracks because of the poor casting techniques that are currently used to shape these materials [1-3]. These defects arise naturally during the turbulent pouring of metals (Figure 1). The importance of this subject was confirmed to the author in the tragic case of the last failed turbine blade that he examined that had caused the plane to crash, costing lives. While it is acknowledged that in other cases engine manufacturers go to great lengths to ensure a 'blade off event does not cause the engine to fail, it seems not helpful to continue to put huge efforts into metallurgical research on alloy development while ignoring the major casting defects necessarily introduced by current manufacturing techniques. The defects in metallic liquids and final castings are principally bifilms. They appear to give rise to a wide spectrum of phenomena including porosity, hot tearing, cold cracking, stray grain initiation, fatigue initiation and corrosion. This is a formidable list. Bifilms are a serious issue that so far has been overlooked or ignored. Bifilms are created easily and rapidly during casting. The surface of the melt oxidises rapidly (whether in air or in so called 'vacuum') so that when folded in, or when experiencing collisions between droplets, the surface oxide contacts dry-face-to-dry-face when impinging against other masses of liquid. The resulting unbondable interface, as a double film called a 'bifilm', is then entrained in the bulk liquid as a crack. Our existing turbulent pouring systems fill the liquid metal with cracks. The defects remain in suspension sufficiently long to become frozen into the casting.

71

At this time, the preparation of superalloys and the manufacture of the majority of turbine blades involves pouring of the liquid metal from considerable heights into molds designed with poor filling systems; the melt experiencing energetic turbulence, guaranteeing that most if not all Ni-base feed stock, and Ni-base blades will contain large populations of defects, some of serious size. For instance nearly the whole of the fracture surface seen in Figure 2 is covered with an oxide bifilm (EDX spectra from this film are added in Figure 5). Although many blades are cast by bottom-gated designs of filling system, these systems are not necessarily free from problems, and in any case the damage by the trauma of the prior pour event usually cannot be reversed. Similarly D/S and single crystal blades are grown in a temperature gradient with a relatively quiescent, planar front, but they too suffer irreversible damage caused by the severely turbulent filling of the mold. The general belief that melting and casting in vacuum avoids the problems of oxidation during casting is seen in general to be an unfortunate and serious error, (even though there is recent evidence [4] that in some conditions a high quality vacuum might avoid this problem). The design of vacuum melting and casting furnaces used for practically all investment cast Ni-base alloys enhances the turbulence problem because of the significant fall of metal from the lip of the melting crucible to the mouth of the mold. This height is often in excess of a meter, whereas it is known that any fall distance greater than about 10 mm for heavy metals can cause entrainment of the surface film and the consequent formation of bifilm cracks [1]. For wrought Ni-base alloys, the situation is even worse, with the melts falling in air through heights of several meters, via poorly designed ceramic channels that ensure the mixing of large quantities of air into the melt during the casting of bottomfilling of tonnage-sized ingots. The ingots are destined for subsequent hot plastic working such as forging, rolling or extrusion. These processes ensure that the entrained bifilms may become tightly closed, making them more difficult to detect, but the bifilms will be expected to be resistant to welding. The working processes will extend the length of the defect, but residual air trapped in folds and creases of the bifilm is likely to continue oxidation or nitridation of freshly extended surfaces, preventing welding until the reservoir of air is finally consumed. The bifilms introduced by turbulence during casting are usually invisible, or at least difficult to see, as a result of their extreme thinness, often measured in nanometres. This contrasts with their impressive surface areas, sometimes measured in square millimetres or even square centimetres. These extensive natural cracks are probably the most important defects in both cast and wrought metals. Evidence for bifilms in metals in general is summarised elsewhere [1-3]. 'Brittle' intermetallics and second phases. The wetted outer interfaces of oxide bifilms (those surfaces from which the oxides grew, atom by atom, and so in perfect atomic contact with the melt) appear to be favoured substrates for the precipitation of second phases in a wide variety of different matrices, including Al, Cu, Ni and Ti alloys [3].

72

In particular, in Ni-base alloys the association of cracks with so-called 'brittle' grain boundary phases has been observed by many workers [5,6]. It has been observed [6] that the outside surfaces of the bifilms act as favoured substrates for the precipitation and growth of carbides containing Cr, Ti, W and Mo. The cracks in the carbides are the visual reminder of the presence of the bifilm that initiated the formation of the carbide. The presence of the bifilm is usually not detectable by optical inspection and therefore unsuspected, but becomes completely clear at higher magnifications of the fracture surface observed in the scanning electron microscope (SEM). Rashid found that the bifilms themselves were oxides rich in Cr and Al (Figures 3 and 5). This study was carried out on an industrially cast blade for power generation made under similar conditions to an aero-engine blade. Exactly similar features were found in blades for aero engines cast in China in 2005 [7]. Furthermore, it is worth emphasising that the carbides themselves are not expected to crack, since it is likely that they will be extremely strong. The tensile failure of the casting would take place by a crack that followed the unbonded bifilm interfaces, apparently following the carbide cracks, giving the appearance that the carbides have caused the failure by their brittleness. D'Souza [8] observes linear features in the microstructure of his CMSX4 on which Nb, Zr and Cr rich phases have formed. He indicates that the composition in which these phases occur is the most likely to form hot tears. Both these observations are consistent with the presence of oxide bifilms. Furthermore, the stray crystals seen in his single crystal castings may have origins associated with the presence of bifilms acting as barriers to the advance of grains as described below. In polycrystalline superalloys Qin et al. [9] observed that long term thermal exposure created chains of carbides that formed pathways for the spread of cracks, and the subsequent deterioration of properties. Once again, the carbides would be expected to have precipitated on oxide bifilms, and so effectively would be pre-cracked; the long term thermal exposure merely gradually opening these features, possibly by the increased pressure caused by the expected precipitation of hydrogen into the cracks, and the creep of the surrounding solid to allow some slight opening. Sidhu et al. [10] also note intergranular cracks associated with continuous films of M23C6 and MC carbides in their welded and heat treated Inconel 738LC alloy. Similar interdendritic carbides are seen associated with film-like defects on fracture surface of a Co base alloy [11]. The report by Mälzer and colleagues [12] on the creep failure of single crystal superalloy LEK94 clearly shows films (assumed to be oxides, but possibly nitrides) on fracture surfaces and microcracks associated with as-cast pores. One would expect pores and bifilm-type cracks to be associated, because both initiate from entrainment mechanisms during casting [1]. In fact bubbles and bifilms are hard to differentiate at times; both are entrained defects; the major difference being the amount of gas that each contains. However, the difference in their gas contents is sometimes not clear, blurring the distinction between them. Also, in passing, it is worth noting that pores and cracks are not solidification defects but casting defects.

73

Grain Boundary Phenomena The commonly accepted reason for the approximately three orders of magnitude benefit of creep life for directionally solidified Ni base alloy structures compared with conventional equiaxed structures is the absence of transverse grain boundaries, the assumption being that these boundaries are weak, resulting eventually in decohesion. Since bifilms are to be expected in gravity poured castings, and because their preferred siting will be between dendrites and grains (cannot penetrate the microscopic air layer between the films so that they are mainly pushed ahead into interdendritic and intergranular spaces) the presence of bifilms as invisible unbonded interfaces easily explains the rupture of transverse boundaries. Moreover, as in the case of intermetallics, there is little reason to suppose that grain boundaries are actually weak; it is almost certain that they are extremely strong, even if not quite as strong as the matrix. Crack formation along the longitudinal grain boundaries of directionally solidified Ni base superalloys during solidification has been attributed to so called grain boundary decohesion in the absence of any really consistent explanation resulting from many studies over past decades [7]. Once again, the presence of bifilms is to be expected, and can be predicted to result in grain boundary cracking [1,3]. Furthermore, these authors found that stray grain formation was increased, once again likely to be the result of a higher density of bifilms, or possibly mechanically stronger bifilms, that mechanically obstruct the advance of the desired single crystal. The blocked advance of the dendrite front, while the withdrawal of the mold continues, will ensure that the liquid above the blockage will progressively undercool as it is withdrawn down the furnace temperature gradient. Eventually the undercooling will become sufficient to nucleate a new grain. The bifilm, with its internal layer of air, will ensure that the new grain will have no benefit of contact with the blocked original crystal, with the result that the new grain will have a totally independent growth orientation. In agreement with this proposed mechanism, Carney and Beech [13] identified oxides at the root of most of the stray grains in single crystals. Furthermore, the incidence of stray grains was reduced by filtering the metal. Welding of Ni-Base Superalloys Up to this time it has been understandable that most authors studying welding have overlooked the probability that oxide bifilms will be present in their alloys, so that the materials that they study are already effectively pre-cracked. Welding provides the opportunity for the cracks to open and become visible. The author [14] has suggested a mechanism for the damaging effect that incipient grain boundary melting has in the heat affected zone (HAZ). For instance, if a grain boundary phase melts in the heat of the weld, but subsequently re-solidifies, why should the properties not be fully recovered, if not improved, as a result of the rapid freezing and consequent fine structure? If a bifilm occupies the grain boundary, the melting of a

74

nearby phase will be associated with (usually) an expansion, with the necessary plastic yielding of the matrix. However, on re-freezing, the volume contraction will merely open the bifilm, creating an open crack, and thus lowering properties (a closed bifilm can at least support some shear stress as a result of friction between the surfaces, the nonplanarity provided by jogs and folds). The differing phases on either side of the crack are another feature to be expected of a bifilm crack; many bifilms are asymmetrical, with one thick side consisting of an older film, often having a spinel structure, whereas its opposing side consists of a pure oxide with a different structure. These differing sides favor the precipitation and growth of different phases during the solidification of the surrounding alloy. Wang at al. [5] present data comparing the behavior of laser beam welding of two single crystal alloys, CMSX-4 and -486, finding increased cracking in the high Zr and Hf -486 alloy. This seems likely to be a result of the higher reactivity of Zr and Hf with oxygen, strengthening the oxide film and enhancing its damage potential during entrainment. Unreliability of tensile properties The statistical chance of oxides being folded in by chance events of turbulence to create scatter in the tensile properties of Ni-base superalloys has been clearly shown by the work of Cox et al [15] who compared top-filled with bottom-gated molds (Figure 4) for alloy IN939 in the hipped condition. Interestingly, both casting techniques show some scatter at lower strength levels, indicating the filling system designs used in this work could be improved. This seems typical of gravity filling where it is difficult or impossible to suppress the formation of all damage. Concluding Remarks Nearly all the above work has been carried out on alloys poured freely under gravity into molds of various kinds and so is expected to contain a generous quantity of bifilm cracks. The rules for the design of gravity casting techniques to avoid the entrainment of the oxidised surface during the filling of the mold have been developed over recent years [2]. Different authors [15,16] have demonstrated that the application of these rules for gravity pouring can reduce the number of casting defects by a factor of 10. At first sight the achievement of a reduction in defects by a factor of 10 might seem impressive. However, if metals were cast with a good counter-gravity technique the factor would be expected to approach infinity. This is because the number of entrained defects can, in principle, fall to zero [2]. It is to be hoped that both researchers and industry will convert to either improved gravity pouring systems, or better, countergravity for the future casting of Ni base and other alloys. Finally, it is worth emphasising that bifilms in cast metals and their remnants in wrought metals probably permeate nearly all our engineering materials. However, they need not be present. Although there are now improved designs of gravity filling systems for castings, any system of alloy production or casting manufacture that involves the pouring

75

of metals involves the production of damage to the liquid metal. A degree of unreliability in properties and performance in the final cast material seems then unavoidable. Our technology for the production of such safety critical components urgently requires to be changed. Only the avoidance of pouring, by proper design of tilt techniques, or preferably full counter-gravity handling of melts and filling of molds, will ensure perfectly reliable metals. This technology is already available and proven [2]. The transformation of many of our engineering materials into materials without bifilms has potential to bring a revolution in properties and performance. References 1. J. Campbell: 'Castings', 2nd edn; 2003, Elsevier, Oxford, UK. 2. J. Campbell: 'Castings practice - the 10 rules for casting' 2004, Elsevier, Oxford, UK 3. J. Campbell: Mater. Sci. TechnoL, 2006, 22, 127-145; discussion 2006, 22, 999-1008 4. D Giuranno, E Ricci, E Arato, P Costa; Acta Materialia 2006 54 2625-2630 5. Y. L. Wang, O. A. Ojo, R. G. Ding and M. C. Chaturvedi: Mater. Sci. TechnoL, 2009, 25, 68-75. 6. K. M. B. Rashid and J. Campbell: Metall. Mater. Trans., 2004, 35A, 2063-2071. 7. Huang Aihua: Proc. 68th World Foundry Cong. 2008, 215-218 8. N. D'Souza: Mater. Sci. TechnoL, 2009, 25, (2), 170-185. 9. X. Z. Qin, J. T. Guo, C. Yuan, C. L. Chen and H. Q. Ye: Metall. Mater. Trans., 2007 38A, (12) 3014-3022. 10. R. K. Sidhu:, N. L. Richards and M. C. Chaturvedi; Mater. Sci. TechnoL, 2007, 23, (2)203-213 11. Montero-Ocampo, M. Talavera and H. Lopez: Metall. Mater. Trans., 1999, 30A, 611-620 12. G. Malzer, R. W. Hayes, T. Mack and G. Eggeier: Metall. Mater. Trans. 2007, 38A, (2), 314-327 13. A. Carney and J. Beech: Proc. Solidification Processing Conf., Sheffield, UK, 1997, University of Sheffield, (ed. J. Beech and H. Jones), 33-36. 14. J Campbell; Mater Science & Technol 2000 25 (1) 125-126 15. M. Cox, M. Wickins, J. P. Kuang, R. A. Harding and J. Campbell; Mater. Sci. TechnoL, 2000, 16, 1445-1452 with additional personal communications from Cox reported in [1] pp 57-61. 16. Z. Li, J. Campbell and Y. Y. Li: J. Processing TechnoL, 2004, 148, (3), 310-316.

Figure 1. Entrainment of a bifilm in a liquid metal.

76

Figure 2. Fracture surface of a turbine blade casting.

Figure 3. Close up of fracture surface showing line scan confirming O, Al and Cr in oxide region. 'Brittle' carbide seen on right, precipitated on oxide [6].

Figure 4. Two-parameter Weibull plot of vacuum melted and cast top poured (squares) and bottom filled (circles) of IN939 test bars into investment molds [15].

77

78

Shape Casting: The 4' International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

P r e m i u m Quality Super Duplex Stainless Steel Castings Without Secondary Refining Bob Puhakka Alloy Casting Industries, New Hamburg, Ontario Canada Keywords: super duplex stainless steel, oxide bifilms, naturally pressurized fill system Abstract The ASTM A890/ A995 (25Cr 7Ni 3.5Mo) super duplex stainless steels are a popular family of high alloy cast steels used extensively in the power generation and energy sectors. The manufacturing steps for these alloys have up to now required the use of costly secondary refining processes such as Argon Oxygen Decarburization (AOD). This paper will describe and validate a set of process parameters that obviate the need for secondary refining. Initial results indicate that the principle embrittling features, carbides and sigma phases, are not found, confirming the proposal that these phases form on bifilms entrained by surface turbulence during pouring. Other significant benefits from the absence of bifilms appear to include the elimination of cracking and leakage defects. Introduction Over the past five years there has been a great deal of truly first-rate research performed toward the goal of understanding the behavior of the superduplex alloy family [1]. The variety in this report targets the nominal composition 25Cr 7Ni 3.5Mo. We now have a detailed comprehension of the chemical and thermal processing requirements needed to produce quality components. However, the processing of the alloys used in such studies has been lamentably poor, ensuring the cast material studied has been polluted with air bubbles, entrained oxide films and reoxidation inclusions. In addition to previous observations that sigma phase can nucleate at ferrite-ferrite boundaries, ferrite-ferrite-austenite triple points [1] and at precipitated carbides [6], it seems likely that sigma phase and carbides precipitate on oxide bifilms, often at grain boundaries, and hence displaying cracks that give these phases the appearance of brittleness [4]. The resulting 'embrittled' matrix has served as the most significant challenge for casting facilities working with these alloys assumed to be the result of the presence of intermetallic phases and carbides at boundaries. The view presented in this work is that the sigma and carbide phases are strong and would therefore be expected to be resistant to cracking. The cracks are present simply because the phases form on the pre-existing oxide bifilms which are effectively cracks; the intermetallics and carbides are not believed to be 'brittle' of themselves. One might speculate that in the absence of doubled oxide films, the sigma phase would probably not precipitate (or not nearly to the degree traditionally observed), since it may find no other suitable substrate. In this case the sigma phase constituents would simply remain in supersaturated solution (it is just possible that they might precipitate later during a heat treatment or slow cooling cycle as an extremely fine phase, possibly now of a completely different composition and structure, and contributing to strength rather than embrittlement).

79

The molten metal is likely to be practically defect-free when it leaves the furnace because of the large density difference between oxides and steel, leading to rapid flotation of oxides and assimilation into the surface slag layer. It is the pouring of the metal into the ladle, and the following pour into the mold that can create so much damage. The prior use of a secondary refining process such as AOD (argon oxygen decarburization) treatment probably helps to eliminate much of the damage introduced by the pour into the ladle. Such treatment adds significant cost to the production of super duplex stainless steel. However, of course, it can do little to assist with the damage caused by a turbulent filling system in the mold. Initial results presented in this report indicate the naturally pressurized filling systems for super duplex stainless steel castings appear to obviate the need for secondary refining. The implication of this result is that the damage introduced during the pour from the furnace into the ladle floats out quickly, prior to the arrival at the mold, and the pouring into the mold, in agreement with expectations [4]. Processing Logic for Super Duplex Stainless Steel For the ASTM A890/ A995 family of alloys the utilization of a custom-batch induction melting process, constituted from washed, dry, virgin raw materials, alloy- specific ingot and controlled, internal, alloy- specific returns provides the foundation for a clean, workable base melt chemistry. Without introducing moisture, carbon-containing liquids and surface-oxidized charge material the melt will be clean and free from undesirable contaminants. Even with all of these precautions, however, the process metallurgist can still expect to find a heavy population of precipitated carbides and nucleated intermetallics such as sigma[l]. The mechanism for this phenomenon is the deterioration of a metastable super saturated ferrite phase that decomposes into a secondary austenite phase and alloy-rich intermetallics. The out-of-mold microstructure is certain to be frilly damaged and unusable without a proper solution anneal. However, even in taking all these precautions, many metal casters continue to experience the serious difficulties resulting from castings that crack during these initial and critical processing steps. As the result of an intensive twenty-four month long foundry trial it is the experience of the author that the difficulties encountered w;ith casting this family of alloys - difficulties that have been traditionally identified as solely solidification phenomenon - are in fact the result of oxide bifilms and air bubbles generated and introduced during the filling of the casting. Taylor reflects the traditional metallurgical view in his statement [2] "The gating design must also deliver the metal to the mold cavity with a minimal degree of turbulence. With the high nitrogen content of the metal, this smooth flow is particularly important to prevent subsurface gas indications. Pouring of duplex stainless steels has been compared to pouring a bottle of beer into a glass. The beer has a high quantity of dissolved gas. If the beer is poured slowly, very little foam develops. When the beer is poured quickly, a large amount of foam develops. Turbulent gating systems can produce gas indications in the casting. A gas problem may first appear as a slight mushrooming of the risers. The majority of the indications will be found on the cope in the highest parts of the castings, either after heat treatment, or when the skin of the casting is broken."

80

-Traditionally the use of the phrase "turbulence" when referring to the fill system has almost universally been associated with the bulk turbulence of the fluid system as quantified by the Reynolds number. The entrainment mechanism described here, however, is the results of surface turbulence resulting in a stainless steel. folding and mixing-in of the surface oxides, air, and other debris such as molding material. Understanding the difference between bulk and surface turbulence is essential for grasping the defect mechanisms described within this paper. The author describes elsewhere [3] explicitly why it is that the current industry-standard filling system design theory is incapable of insuring an adequately tranquil and aspiration-free delivery of the molten metal into the mold cavity. In fact, it has become clear that the consequences of the current fill system design practices are responsible for the embrittlement and porosity mechanisms occurring within these alloys. The use of a naturally pressurized filling system, however, completely eliminates the troublesome failures experienced when processing this family of alloys. Briefly, the naturally pressurized fill system is a system in which the areas of the filling channels are calculated by finding the velocity, V, at each fall distance, h, from the melt level in the pouring basin, assuming no friction. The approach is therefore a simple balance between potential energy, mgh, and kinetic energy, mV2/2. The approach is well known to casting method engineers. The significant difference in this application is to accept these areas and provide the filling system with only these calculated areas at every point throughout the downsprue and runners. Only the gates would be increased in size to reduce the velocity of entry to the mold to the critical 0.5 m/s if possible (on occasions this would be raised to 1.0 m/s if necessary, but not beyond this already 'stretched' limit to the Rule). A typical 'sprue exit/runner/gate' ratio for such a system might vary from 1:1:4 to 1:1:20. It must be stated however that the pre-selection of such a 'ratio' has no part to play in the design of a proper fill system; the ratios simply happen, occurring as a result of the design process. Confirmation by Results Empirical validation of the newly introduced processing concepts took place over a twenty-four month period as an extended foundry trial. The compilation of testing below was performed using test specimens sampled from a typical production-run pour. The test specimens were solution annealed with the castings immediately following shakeout and shot blasting. The alloy is an ASTM A890 Grade 6A super duplex stainless steel, air induction melted without secondary refining.

81

Table 1- Results of Chemical Analysis of Exemplar Heat ASTM A890 Grade 6A Required, \vt. percent Carbon 0.03 max 1.00 max Manganese Silicon 1.00 max Phosphorus 0.030 max Sulphur 0.025 max Chromium 24.0-26.0 Nickel 6.5-8.5 Molybedenum 3.0-4.0 Copper 0.5-1.0 Tungsten 0.5-1.0 Nitrogen 0.20-0.30

Results, wt. percent 0.014 0.79 0.79 0.019 0.007 25.49 6.86 3.62 0.758 0.581 0.22

Mechanical Testing Testing was carried out to meet the specification NORSOK M-630 MATERIAL DATA SHEET MDS D56 Rev. 3 TYPE OF MATERIAL: Ferritic/Austenitic Stainless Steel, Type 25Cr 'fable 2 - Results of Mechanical Testing on Exemplar Heat Results Required NORSOK M-630 MDS D56 Rev 3 Yield Strength, MPa 514 450 min 779 Ultimate Tensile Strength, MPa 700 min Elongation percent 18 min 36 301 max 233 Hardness Testing, BHN 45 min 110.6 Charpy Impact Testing -46 C, Joules Corrosion Resistance AST M A923 Ferric Chloride corrosion Test, Method C, 24 Hours Table 2 - Results of Corrosions Testing on Exemplar Heat Results Required Corrosion Rate, mdd* 10 max 0.94 *mdd = mass loss(mg) per total exposed surface area (dm2) per day Non-Destructive Assessment Pressure cover castings poured from the exemplar heat were subjected to 100% radiographie inspection. The castings were found to free of all defects, requiring no upgrading or rework. The elimination of defects traditionally identified as nitrogen gas porosity and microslinnkage is, frankly, nothing short of amazing. Furthermore the use of liquid penetrant inspection was dreaded for the defects that would be revealed as the nonn. The norm now has become a relaxing test expecting and confirming a routinely defect-free appearance.

82

Microstructural Assessment

Figure 2 ASTM A890 6A KOH electrolytic etch x3100 SEM, the microstructure appears perfectly clean; free from all carbides, intermetallics or other inclusions.

Figure 3 ASTM A890 6A KOH electrolytic etch x775 SEM, the microstructure appears perfectly clean; free from all carbides, intermetallics or other inclusions.

The specification NORSOK M-630 Material Date Sheet MDS D56 Rev. 3 requires the following: "The ferrite content shall be determined according to ASTM E 562 or equivalent and shall be within 35 - 55 %. The microstructure on a suitably etched specimen shall be free from intermetallic phases and precipitates." Naturally, the microstructure is normally checked using optical microscopy. However the author elected to examine the microstructure using scanning electron microscopy (SEM) to be certain that the examination was as stringent as possible. SEM not only has the benefit of greater resolution but has better phase differentiation than traditional reflected light microscopy for these alloys. As observed in Figures 2 and 3, the results are truly outstanding. There is an ideal ferrite-austenite phase balance, and a complete lack of nucleated intermetallics, precipitated carbides or other inclusion matter. It bears repeat for the sake of complete certainty, that this test specimen was air melted in an induction furnace with no covering shrouds and not subjected to any secondary refining. Furthermore, although there is a continued wide debate on whether the use of traditional degassing practices such as the addition of a calcium-silicon additive is necessary for these alloys, in fact, during the developmental stages of this research the author was able to demonstrate clearly that no degassing additives are required. In fact, it was clearly shown that the addition of degassing materials can result in the creation of dispersed inclusions. [5] Other benefits accruing from these new procedures is the experience of zero leakage defects or cracking during the production of approximately three hundred castings over the past 24 months. Once again this is perhaps to be expected in the absence of bifilm defects that would be expected to form excellent cracks and leak paths [4].

83

Validation The introduction of a new set of operating concepts accompanied by bold claims is necessarily subject to scrutiny. In the spirit of the scientific method, new concepts and theories must possess (as rightly suggested by Popper) the opportunity for falsifiability. The empirical evidence presented here demonstrates that the concepts introduced produce excellent results. These results have been repeatedly verified, effectively over 300 times. What the research does not answer, however, is what the out-of-mold microstructural differences are between castings poured using traditional methods and those poured using the new concepts. What is the quantified difference in the presence of sigma and carbides? This study has not yet been attempted so far, so that further work is clearly required. Thus although no direct 'before and after' comparison is currently available for microstructures of eastings by traditional filling and new techniques (as is common for industrial developments, in contrast to laboratory research), the current general appearance, properties and freedom from defects of the improved castings are strongly suggestive that the microstructures are uniquely free from undesirable embrittling phases. Conclusions The use of a precisely designed naturally pressurized fill system coupled with stringent melting practices allows for the production of premium quality super duplex stainless steel castings with the following unique list of properties. 1. No necessity for secondary refining processes such as AOD. 2. Mechanical properties, corrosion resistance and microstructural condition of castings far exceed industry guideline requirements. 3. The loss of castings due to cracking during subsequent thermal processing is completely eliminated. 4. Sub-feeder cracking following cut-off is completely eliminated. 5. Leakage of castings is eliminated. 6. Surface finish is uniquely good. 7. Upgrading of castings by weld repair is eliminated. 8. The study confirms the recent view that entrainment defects (air bubbles and bifilms) are the major defects in super duplex stainless steels, and that bifilms in particular appear to constitute the substrates for the precipitation of the major undesirable constituents such as sigma phase and carbides. Acknowledgements The author would like to thank John Campbell for his advice and assistance during the development stages of this project. The fill system concepts proposed by Campbell served as the foundation upon which these processing steps were built. The application of the naturally pressurized fill system to the family of high alloy steels has yielded phenomenal results not previously believed attainable. Additionally, the group of projects of which this paper was a part were conducted in a production metal casting facility with the all the pressures therein contained. Without the great support and patience of Steve Blenkhorn, such developments would never have happened.

84

References 1. Metallurgical Evaluation Of Cast Duplex Stainless Steels And Their Weldments, U.S. DEPARTMENT OF ENERGY Award Number - DE-FC07-00 ID13975, Songqing Wen, Carl D. Lundin, Greg Batten 2. Duplex Stainless Steel Production, Taylor, Steel Founders' Society of America Technical & Operating Conference Chicago, IL November 1994 3. Advanced Methoding Concepts for the Gravity Casting of Steel Alloys, Bob Puhakka, TMS 2011, San Diego CA. 4. Castings (2003), J. Campbell, Elsevier Butterworth-Heinemann 5. http:/fàobpuhakka.blogspot.com/2010/09/using-energy-dispersive-x-ray.html 6. Duplex Stainless Steels, A State-of-the-Art Literature Review, SFSA, March 2001.

85

Shape Casting: The 4 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals <S Materials Society), 2011

IN SITU HIGH SPEED X-RAY OBSERVATION OF THE SOLIDIFICATION OF AL15CU WITH AND WITHOUT AL 2 0 3 COMPOSITE ADDITION R. W. Hamilton1", A. Leung1, A. B. Phillion2, P. Rockett1, T. Connolley3, and P. D. Lee1, 'Department of Materials, Imperial College London Prince Consort Road, London SW7 2BP, UK 2 School of Engineering, The University of British Columbia 3333 University Way, Kelowna, BC, Canada VIV 1V7 3 Diamond Light Source Ltd, Harwell Science & Innovation Campus, Didcot, OX11 ODE, UK Keywords: X-ray tomography, aluminium, porosity Abstract The use of a novel high temperature furnace in combination with a high speed camera capable of acquiring a complete 3D tomographic image in 3 s at DLC (Diamond Light Source) has enabled in-situ imaging of the formation of porosity during solidification, using monochromated X-rays. This study focuses on the comparative differences between Al-15wt.Cu with and without composite addition, and shows clearly the final stages of pore growth in 3D. Post-solidification high-resolution X-ray tomography was also performed to characterize the final structure, helping to explain the full mechanism of pore growth. Although at relatively coarse spacial resolution, the in-situ results demonstrate the effectiveness of the technique, and allow for the quantification of the rapid changes that occur during solidification. Introduction The development of high-speed X-ray tomographic capabilities on Beamline 112 at Diamond Light Source+ has enabled an in-depth study of the final stages of pore evolution and growth in aluminium alloys. This final stage is generally complex and is of key interest in many industrial applications. The ability to resolve pore development in short time steps (~1.5 sec) allows for direct observation of the complex interactions between the evolving primary Al, the appearance of the eutectic and the influence of third phases (e.g. particulates, intermetallics and contaminants). Although there is plenty of literature covering the characterisation of porosity by X-ray tomography[l-3], until recently it has been confined to post solidification observation. In the present work, novel results are presented comparing in-situ pore evolution in a simple Al 15%Cu binary alloy both with and without AI2O3 reinforcement. Experimental methods Material Two aluminium alloy cylindrical specimens, ~ 2 mm in diameter by 3 mm in height were examined in this study, Specimen 1 was a high-purity Al-15wt.%Cu alloy. Specimen 2 was the same Al-15wt.%Cu binary but containing 9.6vol.% AI2O3 particulate with a particle size range of 10-30 μιη. Corresponding author: [email protected] Diamond Light Source Ltd, Harwelî Science & innovation Campus, Didcot, 0X11 ODE, UK

87

Apparatus The test apparatus, Figure 1, consisted of a small furnace with an internal bore of 13 mm surrounded by six alumina coated resistance elements and capable of maintaining a constant temperature of up to 900 °C. A 9 mm square hole in the furnace, covered with a thin layer of Al foil to prevent heat loss, allowed the X-rays to pass through the specimen from the source to the detector. The furnace was fixed in space above a rotation table, and held a specimen mount that was aligned perpendicular to the beam. This specimen mount could rotate freely within the furnace while acquiring the radiographs needed for computed X-ray tomography. Specimens were held in a boron nitride (BN) cup with a wall thickness of 0.5 mm. Two K-type thermocouples were mounted within the furnace, on opposite sides and as close as possible to the specimen-holder and at the same level as the specimen (and the observation window). There was a small air gap (~1 mm) between the holder and the thermocouple to ensure that there was no contact during rotation. Temperature was controlled via a National Instruments interface using a NI cRIO module. 3D X-ray tomographic scans were collected on Beamline 112 at DLC using monochromated Xrays (55keV), and a high speed Phantom camera with module 3 optics* collecting 120 frames per s at a voxel size of 12 μπι. Combined with the stage rotating continuously at 60° s"1, this enabled the capture of one complete tomograph every 3 s, as only 180° rotation is needed for tomography with monochromated X-rays. The on-board buffer had space for 17000 frames, enabling the capture of 47 complete scans.

Figure 1: Schematic of the test apparatus Test Methodology Each sample was heated in the furnace to 700°C and then allowed to stabilize at this temperature. During heating and stabilization, the samples were continuously observed using X-ray radiography to ensure that füll melting had taken place, and in the case of sample 1, all porosity had dissipated. In the case of sample 2, it was not possible to remove entirely the pre-existing bubbles due to the particulate addition, and so the experiment was run with the residual bubbles. After stabilization was completed, a cooling rate of 1 °C/s was applied, and the behaviour of the samples was observed in-situ. The camera continuously captured radiographs, storing the images in the on-board circular buffer, with the addition of a time stamp. Once the buffer was filled with data, the newest images automatically overwrote the earliest ones. At the point when solidification was observed to have completed, the capture function in the camera was triggered, www.diamond.ac.uk

88

then the buffer was downloaded to a hard disk after one final tomograph was acquired. The timestamp in the radiographs was used to synchronise them with the temperature data recorded from the thermocouples. One challenge in the apparatus design was the fact that the thermocouples were not mounted directly in the sample, but were instead recording the temperature within the furnace. These thermocouples were therefore separated by a small air gap, ~lmm, from the outer surface of the BN cup. Although this design would not be an issue in isothermal experiments, prior experience in cooling tests has shown that there may be an offset in temperature between the thermocouple and the sample during heating or cooling. Reconstruction and image processing A code written specifically for JEEP was used for the tomographic scan reconstruction^]. In total, 47 tomographic images were acquired for each specimen during the solidification process. Data processing was kept to a minimum by selecting scans at suitable time intervals, based on observed changes in the microstructure. Image processing was conducted using Avizo5. An initial 3x3 median filter was applied, and the porosity identified using a simple grey-level threshold. The porosity was quantified using internal sub-routines, and checked manually by reference to arbitrarily selected representative pores. In particular, apparent pores of less than 9 voxels in volume were excluded as noise. In the case of sample 2, regions between the large residual pores were selected to analyze the newly nucleated pores while excluding much of the residual porosity. Results and Discussion The reconstructed and rendered images of typical pores in the Al-15wt.%Cu are provided in Figure 2 and show clearly pore evolution with time at a cooling rate of 1 °C/s. Due to the limited memory buffer in the camera, it was not possible to capture the entire duration of the cooling process, so the pores are already visible in the first image at a size of several voxels. However, over the subsequent image sets at time steps corresponding to temperatures of approximately 598, 574, and 544 °C, Fig 2a,b,c, the pores can be seen to grow and develop complex shape. A multitude of smaller pores can be seen in the final image of the sequence (Fig 2c). The nucleation of these pores is due to the lack of feeding at high fraction solid, when the final phases solidify[3, 5] and hence fluid-flow cannot compensate the additional volume created by solidification shrinkage. Similar images were obtained from sample 2 containing reinforcing paniculate; however, due to the nature of the residual porosity previously mentioned, the overall images are not particularly informative. By focusing on areas which initially contained no porosity, as shown in Figure 3, it was possible to observe pores growing that did not exist at the onset of solidification.

Avizo v6.2, www.vsg3d.com

89

Figure 2.3D rendering of pores in Al-15wt.%Cu in real time; a) 598 °C, b) 574 °C, c) 544 °C In order to characterize the distribution and size of the pores, a volume of 6 mm3 was cropped from the reconstructed tomographic images and analyzed to quantify typical microstructural features, i.e. equivalent spherical radius, volume and surface area as shown in Figure 4, along with the determination of overall pore number density and porosity fraction. The graph of total porosity, P%, and pore number density, Nv, as a function of temperature in the Al-15wt.%Cu alloy shown in Figure 5 indicates that P% increased continuously during cooling, with a rapid increase in P% at high fractions solid. In contrast, the number of pores, Nv, remained low and relatively constant until a high fraction solid was reached at which point there was also a rapid increase in Nv. This agrees well with previous observations[6] and the concept of initial pores growing steadily due to relatively free diffusion of hydrogen in the liquid, until the final solidification when the remaining liquid solidified rapidly. At this point, the rapid reduction in liquid flow caused by the narrowing intergranular regions and increased liquid viscosity is accentuated by the presence of particulate, and results in reduced pressure within the liquid, void nucleation, and void growth. The decrease in Nv seen at the end of solidification cannot currently be accounted for, but it could be an effect of the relatively coarse resolution, which could have caused nearby pores to be counted as separate when they are actually connected. In the final stages of solidification, shrinkage allows them to grow enough to be identified as one interconnected pore.

Figure 3. 3D rendering of pores in Al-15wt.%Cu with 9.6vol.% AI2O3 particulate; a)618°C, b)565°C

90

Figure 4. Typical microstructural features of as-cast micro-porosity Also shown in Figure 5 is the development of P% and JVv in the Al-15wt.%Cu alloy with 9.6 νο1.%Αΐ2θ3. As can be seen in the figure, the initial percentage porosity was larger compared to the alloy without particulate, due to the residual porosity, but the increase in Nv during cooling occurred at a similar rate. It also appears that there was a tendency for more pores to grow above the observable threshold, rather than relatively few pores to grow at the expense of others. In comparison the Al-15wt.%Cu showed a similar magnitude of porosity, with markedly fewer (1/3) actual pores. This hypothesis is also partially validated by the fact that the sharp increase in pore nucleation occurred at ~590'C in this alloy, at least 20' C above the start of pore nucleation in the alloy without particulate. Thus, the particulate phase can significantly restrict the growth of the pores, as well as acting as sites for further heterogeneous pore nucleation, leading to a larger number of smaller pores. It should be noted that the presence of residual porosity would also deplete the amount of hydrogen available for newly nucleated pores to grow.

Figure 5. The P% and Nv for the Al-15wt.%Cu alloy and Al-15wt.%Cu with 9.6vol.% A1203 particulate as a function of temperature.

91

By isolating an individual, but representative pore in sample 1 and sample 2, it was possible to examine the development of specific pores starting from the point at which the resolution threshold was exceeded (in this case at least 9 voxels in volume) until final solidification. In the Al-15wt.%Cu alloy both with and without paniculate, the pore growth, shown in Fig 6 as pore equivalent radius, was predominantly as expected from the literature[7]: 1. Initial nucleation and initial growth (which cannot be observed here) 2. A rapid growth phase with a rate which then decreases with decreasing temperature. A plot of the ratio of the pore surface area as compared to the expected surface area of a sphere containing the same volume (i.e. a measure of sphericity), also provided in Fig 6, indicates that initially pores in both samples grew reasonably spherically. It might be expected that some impingement occurs during the final stages of solidification, but this detail was below the resolution of this in-situ tomographic experiment. However, it can be seen that, in the presence of paniculate, the pore stopped growing sooner and smaller, suggesting that growth became restricted by the paniculate. During the final stages of solidification the pore in sample 2 grew considerably less, and becomes more tortuous as it follows the intergranular boundaries[8].

Figure 6. Equivalent radius and surface area ratio for a single pore in Al-15wt.%Cu with and without paniculate. Conclusions High-speed tomographic images were successfully captured of the solidification of an Al15wt.%Cu alloy both with and without reinforcing A1203 paniculate, allowing for visualization and quantification of porosity growth as a function of temperature in the semi-solid regime.

92

Despite limitations in the optics available giving a relatively large spatial resolution of 12 μιτι, and the fact that it was not possible to monitor the exact temperature of the sample due to the high speed rotation, the results are surprisingly clear and show good agreement to the theories of pore nucleation, growth, and coalescence. There is good agreement with the expectation, that the growth of porosity will be constrained where there is a significant addition of paniculate, resulting in a larger number of smaller pores. In particular, the final evolution of shrinkage porosity that occurs at final solidification is observed, and shown to be more tortuous (less spherical) in the paniculate containing sample. These are significant initial results, which will be strengthened by future planned work at higher time and spacial resolution. Acknowledgements The authors would like to thank all the staff at Diamond Light Source, Beamline 112; (EE2071) and the EPSRC (EP/FOO1452/1). References [1] J.Y. Buffiere, S. Savelli, E. Maire, "Characterisation of MMCp and cast Aluminium alloys," X-ray Tomography in Material Science, ed. Baruchel, Buffière, Maire, Merle, Peix, Ed. Hermes, Paris, Fr. (2000), 103. [2] P.D. Lee and J.D. Hunt, "Hydrogen porosity in directional solidified aluminium-copper alloys: In situ observation," Acta Materialia, 45(10) (1997), 4155. [3] O. Lashkari, L. Yao, S.L. Cockcroft, D.M. Maijer, "X-ray microtomographic characterization of porosity in aluminum alloy A356," Metallurgical and Materials Transactions A, 40A(4) (2009), 991. [4] V. Titarenko, S. Titarenko, P.J. Withers, F. De Carlo, and X. Xiao, "Improved tomographic reconstructions using adaptive time-dependent intensity normalization," Journal of Synchrotron Radiation, 17 (2010), 689. [5] J.D. Zhu, S.L. Cockcroft, D.M. Maijer, "Modeling of microporosity formation in A356 aluminum alloy casting," Metallurgical and Materials Transactions A, 37A(3) (2006), 1075. [6] M.A. Easton and D.H. St.John, "The effect of grain refinement on the formation of casting defects in alloy 356 castings," International journal of Cast Metals Research, 12(6) (2000), 393. [7] R.C. Atwood, S. Sridhar, W. Zhang, P.D. Lee, "Diffusion-controlled growth of hydrogen pores in aluminium-silicon castings: In situ observation and modelling," Acta Materialia, 48 (2) (2000), 405. [8] R. Sasikumar, M. Rettenmayr, S. Savithri, H.E. Exner, "A model for the formation of interdendritic cavities from pores pre-existing in the melt," Zeitschrift für Metallkunde, 92 (2) (2001), 158.

93

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011 IN-MOLD THERMAL ANALYSIS OF DUCTILE CAST IRON Morten I. Onsoien SINTEF Materials and Chemistry, N-7465 Trondheim, Norway Keywords: Ductile cast iron, In-mold melt treatment, In-mold thermal analysis, Microstructure

Abstract The objective of the present work is to study characteristic solidification data of three experimental ductile irons, produced using a flow through in-mold melt treatment technique, by means of in-mold thermal analysis. All produced irons were in addition subjected to chemical analysis, quantitative metallography, as well as an evaluation of shrinkage porosity and carbide forming propensity. Applied melt treatment alloys were based on FeSiMg with small contents of cerium, lanthanum or misch-metal. Based on the obtained results it was concluded that in-mold thermal analysis provides an effective way of monitoring the graphite nucleation throughout the solidification. Furthermore, the selected melt treatment resulted in ductile irons having very high nodule counts, up to 1162 nodules mm"2, very low shrinkage porosity and very low carbide forming propensity. Introduction In-mold melt treatment of ductile iron with ferrosilicon based treatment alloys was first developed and introduced to the foundry practice in mid-sixties [1]. In-mold processes for making ductile iron has gained increased popularity over the years due to advantages such as consistently high magnesium recovery, an almost complete absence of glare and fume, simultaneous nodularization and inoculation, and avoidance of problems related with magnesium fade [2-7]. Conditions for a successful in-mold procedure require a balanced amount of ferrosilicon-magnesium (FeSiMg) based alloy in the reaction chamber under the runner to last throughout the pour. A filter in the runner system is often used in combination with in-mold melt treatment to avoid problems with dross related defects in the casting [2], Thermal analysis is commonly used to study phase transformations, and this technique is frequently used in establishing equilibrium phase diagrams [8]. Under given kinetic conditions, measured deviations from equilibrium reaction temperatures can be used to study nucleation and growth processes [9]. Solidification characteristics can be investigated using a simple cooling curve analysis applied to standardized castings, and thus, together with chemical analyses, provide rapid characteristic information that can be used to tailor the cast iron melt [10-12]. In the current work use of thermal analysis in material and process control of ductile iron produced using a flow through in-mold treatment has been explored. The major goal was to study the characteristic solidification data of the ductile iron within the mold immediately after melt treatment. In addition the produced irons were characterized by means of chemical analyses, quantitative metallography, relative shrinkage porosity measurements and carbide forming propensity.

95

Experimental procedure Figure 1 shows schematically the applied casting procedure using a flow through in-mold system. An experimental ductile cast iron charge of 80 kg was made based on re-melting of iron returns, pig iron, high purity ferrosilicon and graphite in an induction furnace. The molten iron reached a temperature of 1510 °C prior to pouring the melt into the holding ladle. The melt was kept in the holding ladle until it reached a temperature of 1400 °C. Three batches of 20 kg each were then poured into three different sand molds containing a reaction chamber and a runner system for nodularization using a flow through in-mold melt treatment technique. Prior to pouring, the first reaction chamber mold was charged with a commercial lanthanum containing FeSiMg alloy (La-FeSiMg), the second with an experimental cerium containing FeSiMg alloy (Ce-FeSiMg) and the third with an experimental FeSiMg alloy containing misch-metal (MMFeSiMg). The amount of treatment alloy corresponded to 1 wt.% of the melt. All FeSiMg alloys contained 5.6-5.7 wt.% magnesium and 0.3 to 0.4 wt.% rare earth elements. The melt was allowed to flow through the reaction chamber mold and into four different sample molds (Casting A-D). The sample molds were filled, by passing them continuously on a conveyor, under the liquid iron flowing from the reaction chamber mold (sequence A-B-C-D) at a filling rate of around 1.8 kg s"1. The sample molds consisted of a 20 mm thick plate, a 5 mm thick plate, a chill wedge sample, and a hemispherical sample for shrinkage porosity evaluation. In addition a single thermocouple Quik-Cup for thermal analysis was embedded in the drag part of the mold. This experimental setup allowed sample extraction for "in-situ" thermal analyses, chemical analyses and metallographic analyses at different times after the magnesium reaction start without any risk of extended mixing of the melts.

Figure 1. Flow diagram of casting procedure using the flow through in-mold system. Analysis of the collected temperature data was done using a commercially available software for thermal analysis. The following parameters were extracted; undercooling, recalescence, graphite factor 1 (GRF1) and graphite factor 2 (GRF2). GRF1 is an indicator of eutectic graphite precipitation rate, while GRF2 is an expression of the amount of eutectic graphite formed during the last stage of solidification [13]. Samples for chemical analysis were extracted from the 20 mm thick plates. Metallographic examination of the cast irons were performed on samples extracted from a cross section cut at the centre of both the 5 mm and the 20 mm thick plates. The metallographic samples were prepared according to standard metallographic techniques, i. e., polished to a Ιμτη diamond spray finish. The graphite nodules were then characterized using an automated image analysis system. For analyses of characteristic graphite data such as nodule count, nodule diameter, nodule shape factor and nodularity, only graphite nodules larger than 5 μιη were measured. Nodule diameter was defined as the diameter of a circle with the same area as the graphite nodule under consideration, whereas the shape factor S is an area to perimeter function expressed as: S^iTtA/P2, where A is the nodule area and P is the perimeter. Nodularity was defined as the percentage of nodules with a shape factor larger than 0.65. The polished

96

samples were then etched in 2 % nital for quantification of the microstructure constituents such as ferrite, pearlite and carbide by means of counting minimum 1000 grid points at a magnification of 200 X using an optical microscope. The cast hemispherical body was used to determine the shrinkage tendency of the different irons by means of density measurements using Archimedes principle of density measurement, while carbide forming propensity was evaluated by means of a standard chill wedge sample. Results and discussion Chemical composition The chemical composition of the produced ductile irons is summarized in Table 1. Within each series the magnesium level increases from the first mold to be filled to the last one, i.e. from mold A to mold D. Magnesium treatment using the Ce-FeSiMg alloy seems to give higher magnesium yield than magnesium treatment using the two other FeSiMg alloys, especially in the firstfilledmold. The rare earth levels in the samples are, as expected, quite low as a result of the low amount of treatment alloy added. Still the levels are not to far away from the optimum levels of rare earth metals reported by Lalich [14], such that a positive effect on nodule count is expected. Table 1. Chemical composition of experimental ductile irons. Sample La-FeSiMg A La-FeSiMg B La-FeSiMg C La-FeSiMg D Ce-FeSiMg A Ce-FeSiMg B Ce-FeSiMg C Ce-FeSiMg D MM-FeSiMg A MM-FeSiMg B MM-FeSiMg C MM-FeSiMg D

C 3.81 3.70 3.89 3.81 3.82 3.73 3.94 3.75 3.87 3.83 3.83 3.80

Si 2.46 2.40 2.51 2.55 2.55 2.64 2.60 2.59 2.40 2.55 2.37 2.55

0.23 0.22

Elements P 0.023 0.022 0.022 0.022 0.023 0.021 0.022 0.023 0.022

0.23 0.23 0.24

0.023 0.022 0.023

Mn 0.23 0.25 0.23 0.24 0.23 0.24 0.24

(wt.%) S 0.011 0.011 0.011 0.012 0.011 0.010 0.010 0.010 0.011 0.010 0.010 0.011

Mg 0.021 0.028 0.029 0.030 0.027 0.028 0.027 0.031 0.018 0.027 0.030 0.030

Ce <0.003 <0.003 <0.003 O.003 0.003 0.003 0.003 0.003 <0.003 O.003 <0.003 <0.003

La 0.005 0.006 0.005 0.005 <0.005 <0.005 <0.005 <0.005 O.005 <0.005 <0.005 <0.005

Thermal analyses The increase in peak temperature as pouring progresses from casting A to D as seen in Figure 2, is probably caused by heating of the reaction chamber and runner system as pouring progresses. As a consequence of holding ladle heat loss to the surroundings and sequence of pouring, the peak temperature decreases from pouring of the first series of castings (La-FeSiMg) to the last series poured (MM-FeSiMg). The lower melt temperature obtained for the MM-FeSiMg series probably results in slower reaction rate, and consequently causes the lower magnesium concentration found in the first mold (MM-FeSiMg A).

97

Figure 2. Measured peak temperatures in the mold The eutectic undercooling and recaleseence are highest (average values) for the iron produced using the Ce-FeSiMg alloy, Figure 3a and b. Thus, the nuclei for graphite in the Ce-FeSiMg treated iron requires high undercooling to be triggered which is in agreement with the findings reported by Strong [12] and Prinz et al [15]. Furthermore, the high recaleseence show that a high number of nuclei are triggered more or less simultaneously at the onset of the solidification, resulting in early formation of graphite during solidification, which in practice increase the risk of shrinkage porosity. High average GRF1 values and low average GRF2 values are obtained for the irons treated with the La-FeSiMg alloy, Figure 3c and d. A high GRF1 is desirable because it indicates more continuous graphite expansion and less overall risk for shrinkage. GRF2, which is determined from the first derivative of the cooling curve, describes the degree of late graphite formation. A low GRF2 is desirable and indicates late-forming graphite that counteracts shrinkage in the last metal to solidify. Thus, low shrinkage porosity should be expected in the irons produced using the La-FeSiMg alloy, while some porosity may be expected in the irons produced using the CeFeSiMg alloy. Microstructure Figure 4 show examples of typical microstructures found in the experimental irons. Eutectic carbide was not observed in any of the samples, even though high undercooling, in the range from 15 to 20 °C, was measured in several samples, Figure 3a. The amount of graphite was around 10 vol.% for all samples. Figure 5 show the amount of pearlite in the analyzed specimens. As expected the amount of pearlite is lower in the 20 mm thick plates than in the 5 mm thick plates, due to the slower cooling rate, and thus improved conditions for carbon diffusion, in the thicker sections [16]. In the 5 mm thick plates the nodule count is very high, from 851 to 1162 nodules mm"2. The shape factor and nodularity is also very high above 0.87 % and above 96 %, respectively, while the nodule size is very low in the range from 10.3 to 11.8 μπι. This shows that all of the used FeSiMg alloys provide very good graphite nucleation conditions in the 5 mm thick plates. There

98

is however a tendency for the nodule count to dropfromthe first to the last casting poured within each series, Figure 6a.

(c) (d) Figure 3. Characteristic data from thermal analysis of the produced in-mold magnesium treated ductile irons, a) Undercooling, b) Recalescence, c) Graphite factor 1 and d) Graphite factor 2.

(a) (b) (c) Figure 4. Examples of microstructure in examined 5 mm thick plates of produced experimental ductile irons using: a) La-FeSiMg, b) Ce-FeSiMg and c)MM-FeSiMg. In the 20 mm thick plate the nodule count is in the range from 326 to 597 nodules mm"2. Also in these plates the shape factor and nodularity is high, above 0.77 % and above 87%, respectively. Due to the slower cooling rate in the 20 mm thick plates, the average nodule size is larger than in

99

the 5 mm plates, i.e., in the range from 13.0 to 18.9 μπι. Highest average nodule count was found for the ductile irons produced using the Ce-FeSiMg. This alloy also showed the highest undercooling and recalescence during solidification, see Figure 3.

(a) (b) Figure 5. Amount of pearlite in a) 5 mm thick plates, b) 20 mm thick plates of produced experimental ductile irons.

(a) (b) Figure 6. Nodule count in a) 5 mm thick plates and b) 20 mm thick plates of produced experimental ductile irons. Relative shrinkage porosity and carbide formation Highest density was obtained for one of the samples produced using the La-FeSiMg. Thus, this alloy was used as basis for calculation of relative shrinkage porosity of the other ductile irons produced. Even though the relative porosity level is low, there is a tendency for irons produced using the Ce-FeSiMg alloy to have highest shrinkage porosity while the irons produced using the La-FeSiMg alloy have lowest shrinkage porosity, Figure 7a. These findings support the interpretation of the GRFl and GRF2 data from the thermal analysis, where a high GRFl and a low GRF2 is desired for minimum shrinkage formation tendency, see Figure 3. Furthermore, these observations are in agreement with the work of Dunks [17] who reported that lanthanum

100

containing in-mold treatment alloys had a beneficial effect on the shrinkage characteristics of ductile iron compared to cerium containing in-mold treatment alloys. The carbide forming propensity of the different ductile irons produced is fairly low, and there are no significant differences in measured chill lengths between the different irons, Figure 7b. Based on experiments with ductile iron samples having similar chemistry and geometry as in the current investigation, but produced using a conventional sandwich method for melt treatment, Skaland [18] reported that the use of La-FeSiMg resulted in lower carbide formation in the 5mm thick plates, compared to FeSiMg treatment alloys containing either cerium or misch-metal. In the present work the metallographic investigation revealed that no carbide was formed in any of the 5 mm thick plates. Moreover, the nodule count in Skalands [18] investigation was also significantly lower than that obtained in the present work. The carbide formation is associated with the nucleating abilities of the melt treatment alloys, i. e., the number of potent nuclei formed as a result of the treatment alloy addition. Formation of a high number of potent nuclei promote the formation of a large number of graphite nodules which gives rise to the high nodule count found, Figure 6, and suppress the carbide formation. Thus, the flow through in-mold magnesium treatment technique is effective in producing a large number of potent nuclei for graphite formation, which in turn results in a very high nodule count as well as low porosity and carbide forming propensity.

(a) (b) Figure 7. Measured (a) relative shrinkage porosity and (b) chill lengths in the produced experimental ductile irons. Conclusions In-mold thermal analysis was performed to characterize the solidification behavior of experimental ductile iron samples produced using a flow through in-mold melt treatment technique applying lanthanum, cerium or misch-metal bearing FeSiMg based treatment alloys. The produced irons were subjected to chemical analysis, quantitative metallography, as well as an evaluation of shrinkage porosity and carbide forming propensity. Based on the conducted work the following main conclusions can be drawn: • In-mold thermal analysis is an effective approach to optimize the amount of in-mold treatment alloy in a process that virtually eliminates fading of the melt treatment agents. • In-mold thermal analysis provides an effective way of monitoring the graphite nucleation such that optimal melt treatment alloys can be selected to ensure a high number of potent

101

• •

nuclei, giving rise to a high nodule count, and graphite nucleation towards the end of solidification as a means to reduce or completely avoid problems associated with shrinkage porosity. Very high nodule counts can be obtained using in-mold melt treatment, in the current investigation nodule counts up to 1162 nodules mm"2 was obtained in 5 mm thick plates, and up to 597 nodules mm"2 in 20mm thick plates. The high potency of the graphite nuclei favors the formation of graphite and prevents the formation of carbide. Carbide was not found in any of the 5mm thick plates produced. References

[1]. W. J. Dell, R. J. Christ: Chill elimination in ductile iron by mold inoculation, Modern Casting, 46 (1964) 408-416. [2] Y. S. Lerner, L. S. Aubrey, D. Craig, T. Margaria: In-mould treatment processes in iron foundry practice, Foundry Trade Journal, 175, 3598 (2002) 24-27 [3] G. Hauck: In-Mould Treatment Processes in Iron Foundry Practice, Foundry Trade Journal, 177,3599(2003)28-31. [4] R. W. Thomas: In the mold process for horizontally parted molds, AFS Transactions, 88 (1980)501-502. [5] C. Loper, Jr., R. Heine, C. Wang and L. Janowski: Fading of magnesium treatment in ductile cast irons, AFS Transactions, 84 (1976) 203-214. [6] I. Henych: Some metallurgical aspects of producing ductile iron and holding treated metal, AFS Transactions, 94 (1986) 609,620. [7] K. Q. Davis, R. K. Buhr and J. G. Magny: Dissolution of MgFeSi alloy during inmold treatment, AFS Transactions, 86 (1978) 379-384. [8] R. Ferro, A. Saccone: Thermal analysis and alloy phase diagrams, Thermochimia Acta, 418 (2004) 23-32. [9] C. Labrecque, M Gagné: interpretation of cooling curves of cast irons: A literature review, AFS Transactions, 106 (1998) 83-90. [10] A. Dioszegi, I. L. Svensson: On the problems of thermal analysis of solidification, Materials Science and Engineering A, 413-414 (2005) 474-479. [11] R. V. Sillen: Optimizing iron quality through artificial intelligence, Modern Casting, 11 (1996), 43-45. [12]. G. R. Strong: Thermal analysis as a ductile iron molten metal processing evaluation tool, AFS Transaction, 91 (1983) 151-156. [13]. ATAS Verifer Users Guide, NovaCast Aß, Ronneby, Sweden. [14]. M. J. Lalich: The influence of rare earth elements on magnesium treated ductile cast irons, in Proceedings from the 2nd International Symposium on the Metallurgy of Cast Iron, Geneva, Switzerland, May 1974, Eds.. B. Lux, I. Minkoff, F. Mollard, 561-581. [15]. B. Prinz, K. J. Reifferscheid, T. Schulze, R. Döpp: Zur Anwendung der termischen Analyse bei Gusseisen mit Kugelgraphit, Giesserei-Praxis, Nr. 9/10 (1992) 146-159. [16]. D. A. Porter, K. E. Easterling: Phase transformations in metals and alloys 2nd ed., Chapman & Hall, London (1991). [17]. C. M. Dunks: In-the-mold worldwide - Today and tomorrow, AFS Transactions, 90 (1982) 551-556. [18]. T. Skaland: A new method for chill and shrinkage control in ladle treated ductile iron, in Proceedings from the 66th World Foundry Congress, Istanbul, Turkey, September 6-9 (2004), pp. 975-987.

102

Shape Casting: The 4"1 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Pau! N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

Modeling of hot tearing and its validations in metal castings J. Guo, J.Z. Zhu, and S. Scott ESIUSR&D, 6851 OakHallLane, Suite 119, Columbia, MD21045, U.S.A. Hot tearing is one of the most serious defects for a metal casting. Hot tearing can result in permanent defect in the casting, which may cause the final cast product to be unusable. It is believed that the imposed strains and stresses on the solid framework in the mushy zone by the solidification shrinkage and thermal contraction create the conditions for hot tearing. To predict the susceptibility of hot tearing in a casting, a hot tearing indicator is proposed which is based on Gurson's constitutive model. In this paper, the numerical implementation of the hot tearing indicator is described and the numerical modeling of the hot tearing formation is validated by experiments. 1. INTRODUCTION Previous studies have revealed that hot tearing occurs in the late stage of solidification when the volume fraction of solid is close to 100 percent and the solid phase of casting is formed by a continuous network of grains. The formation and propagation of hot tearing have been found to be directly affected by the cooling history, the chemical composition, and mechanical properties of the alloy, as well as the geometry of the casting [1]. Hot tears initiate partially as the result of the formation of shrinkage porosity when an alloy is in the semisolid state, and propagates during solidification. Its evolution is often the direct consequence of strains induced by the solidification shrinkage and thermal contraction imposed on the continuous solid network. Therefore, in the present study, castings are assumed to be solid undergoing elastoplastic or viscoplastic deformation. A hot tearing indictor is proposed which is based on Gurson's constitutive model. The proposed hot tearing indicator has been implemented in the finite element simulation software ProCAST [2]. 2. HOT TEARING INDICATOR Much research has been given recently to the study and understanding of hot tearing phenomena [3-11], as it is one of the most serious defects encountered in castings. Various theories have been proposed in the literature on the mechanisms of hot tearing formation. Detailed reviews on the theories and experimental observations of the formation and evolution of hot tearing can be found in references [3-4] and the references therein. Most of the existing hot tearing theories are based on the development of strain, strain rate, or stress in the semi-solid state of the alloys. For strain-based theories, the premise is that hot tearing will occur when the accumulated strain exceeds the ductility [5-7] of the material. The strain rate-based theories suggest that hot tearing may form when the strain rate reaches a critical limit during solidification [8-9], The stress-based criteria, on the other hand, assume that hot tearing will start if the thermal induced stress in the semi-solid exceeds some critical values [10-11]. Although these theories were proposed independently as distinct theories, they can, indeed, be considered as somehow related due to the fact of the relationship between strain, strain rate and stress. It is such a relationship that motivates the

103

development of a hot tearing indicator which uses the accumulated plastic strain as an indication of the susceptibility of hot tearing. This considers the evolution of strain, strain rate and stress in the later stage of solidification. A Gurson type of constitutive model, which describes the progressive micro-rupture in the ductile and porous solid, is adopted to characterize the material behavior in the semi-solid state. Unlike other existing hot tearing indictors that are designated for a particular type of casting process, the current proposed hot tearing indicator is general in theory and applicable to all types of casting processing. The constitutive model used to describe the material behavior in the semisolid state is a modified form of Gurson's model [12-15], which was originally developed for studying the progressive micro-rupture through nucleation and growth of micro-voids in the material of ductile and porous solids. Denoting the displacement field at time / as u (x, t), assuming strain in the casting to be small as observed in experiments, the strain tensor is written as £

= Vsu = -[Vu + (Vu)T]

(1)

The additive decomposition of the rate representation of the total strain is given by έ = έ'+έ"+έ"· (2) where é ' ^ ' a n d έ" are the elastic strain rate, inelastic strain rate, and thermal strain rate respectively. The stress rate is written as σ(χ,ί) = Ό:(έ-έρ-έώ)

(3)

and D, the isotropie elasticity tensor. When a casting material is considered to be elastoplastic or viscoplastic with a yield function 0, the yield condition can be written as
- έ " :έ"ατ

(5)

with associated flow rule: é"=r^ da

104

(6)

where fis the plastic flow parameter, and it will be determined by solving the constitutive equations. When a casting is considered to be viscoplastic, γcan be an explicit function of^in the form of [16] (7)

γ = -Ψ(φ) 1 where η =η (Τ) is the viscosity parameter and

ψ

(8)

Η^ί

with material parameter m = m(T), yield stress ΪΌ= YéJ). The isotropie hardening has the form of (9)

/c=Y0(T) + H(T)ë"

with H(T) as the plastic modulus. The evolution of the back stress for the kinematic hardening is of the Armstrong and Frederick form [ 17] i = cèp-bì"x

(10)

Here c and b are temperature dependent material parameters, and the effective plastic strain rate is given by

*'=^7'=t^

(H)

G„ in Equation (4) is referred to as Gurson's coefficient. It is defined to be G „ = - 2 / * 9 l c o s h ( ^ ) + {l + ( 9 ,/*) 2 }

(12)

where q\ is a material constant and / ' = /,

f-f / * = /* TJFT 7 -JC( / v -f'} +

f°r

f^L

for f

>f

"

*

(13)

Here,/; = \lq\,fc is the critical void volume fraction, and,//.- is the failure void volume fraction. Their values should be different for different materials. In current hot tearing indicator calculation, constant

105

values are used in [18]. Gurson's coefficient characterizes the rapid loss of material strength due to the growth of void volume fraction/,. When/. =//.-, then / * = / , =l/<7,, and G„ = 0 forzerò stress, i.e., the stress carrying capacity of the material vanishes. The evolution of the void volume fraction is described by the nucleation of new voids and the growth of existing voids, Jv

J nucleation

V'^/

J growth

with the rate of void growth defined as

Λ— = 0 - f*Mé') = m - / * ) ( ^ ^ ) s i n h ( ^ ì )

(15)

In this study, the nucleation of a void is assumed to be strain controlled and is defined as J nucleation ~

e

I'

lit

0

/

where ^=K=ii^i'-è'dT

K
(17)

is defined as the hot tearing indicatoiiHTl), tc represents the time when the coherency temperature is reached, and ts denotes the time when the solidus temperature is reached. It is observed that the hot tearing indicator is in fact the accumulated plastic strain in the semi-solid region that corresponds to the void nucleation. It provides a good indication for the susceptibility of the hot tearing during solidification. Gurson's constitutive model and the proposed hot tearing indicator are implemented in the finite element software ProCAST, a comprehensive numerical simulation tool for modeling and studying solidification and metal casting, following the standard procedures [19-20]. It is observed from the integration schemes used for the Gurson's constitutive model that the hot tearing indicator is computationally very efficient. In fact, no additional computation is needed beyond that required for the integration of the constitutive model. 3. EXPERIMENTAL VALIDATIONS There are many experimental works have been completed to study the formation of hot tearing during casting solidification. In order to validate the model proposed above, two sets of experimental results are compared. 3.1 Application to Mg alloys Cao et al performed some experiments to study hot tearing formation during solidification of binary MgAl and ternary Mg-Al-Ca alloys in a steel mold [21-22]. A cross section of the casting and mold is shown in Figure 1. A hot cracking susceptibility (HCS) was introduced, which is a function of the

106

maximum crack width, crack length factor, and the crack location. It was found that it is more likely to have cracks at the sprue end than at the ball end, and it is less likely to have crack in the middle of the rod. Also, the longer rod is easier to crack. Figure 2 shows the simulated results of the hot tearing indicator for a Mg-2%A1 alloy casting, which is assumed to be elastoplastic. The computed hot tearing indicator agrees very well with the experiments.

Figure 2. HTI for a Mg-2%A1 alloy casting

Figure 1. Steel mold for constrained rod casting

Figure 3 shows the experimental results of hot tearing at the sprue end of the rods for three different alloys. The calculated hot tearing indicators are shown in Figure 4 accordingly. It can be seen that the hot tearing at the same location is less severe when the Al content increases from 2% to 4% and then to 8% for the same casting with the same casting conditions. The increasing of Al reduces the temperature range between fraction of solid at 0.9 and the end of solidification, hence the reduction of hot tearing susceptibility. Again the simulated hot tearing indicators agree well with the observations.

Figure 3. close-up views of hot tearing (cracks) in the bottom rods near the sprue: (a) Mg-2% Al;(b) Mg-4% AI; (e) Mg-8% Al

Figure 4. Hot tearing indicator in the bottom rods near sprue: (a) 2% Al; (b) 4% Al; (e) 8% Al

5.2 Application to Al alloys Experimently Li systematically studied the effects of various casting conditions, such as mold temperature, pouring temperature, and grain refinement, on hot tearing of different cast aluminum alloys

107

[23], In Li's study, an instrumented constrained rod mold was developed such that it can be used to simultaneously measure the load/time/temperature during solidification for a restrained casting. The setup is shown schematically in Figure 5. Figure 6 gives the dimension of the studied casting. A detailed description of the experimental setup for quantitatively measuring hot tearing onset and contraction during solidification of aluminum alloys can be found in [23]

Figure 5. Experimental set up

Figure 6 Dimensions of the studied casting

In one of the experiments, three mold temperatures were used, 200°C, 300°C, and 370°C. The cracks in the hot spot region of alloy M206 for those three mold temperatures are shown in Figure 7.

Figure 7 hot tears in neck region for different mold temperature a) 200°C, b) 300°C, and c) 370°C The results clearly suggest that the mold temperature has a significant effect on hot tearing susceptibility of this 206 alloys. The hot tearing susceptibility decreases with increasing mold temperature. This is because of the reducing solidification range with higher mold temperature. Based on the experimental casting condition, the simulations were performed. Figure 8 shows the hot tearing indicator distribution for those three mold temperatures. Table 1 lists the actual values of the indicator. Both average and maximum hot tearing indicators in the hot spot region decrease with the increasing of mold temperatures. Increasing mold temperature can reduce hot tearing formation for this situation. From the simulations above, it can be concluded that the current modeling can predict the hot tearing susceptibility for this aluminum alloy casting very well.

108

Figure 8. Predicted hot tearing indicator for mold temperature from top to bottom of a) 200°C, b) 300"C, and c)370°C Table 1 Values of the hot tear for different mold temperature Mold Temperature (°C) Average Hot Tear Indicator Maximum Hot Tear Indicator 200

.0166

.1207

300

.0154

.0567

370

.0129

.0497

There are more experimental results in Li's work [23]. More simulations are on under investigation to validate the current model, such as the effect of melting temperatures, grain refinement etc. 4. CONCLUSIONS A hot tearing indicator for numerical simulation of metal casting is presented. The hot tearing indicator is based on Gurson's constitutive model that describes the internal damage of solids by considering the initiation and growth of the voids in the mushy zone of the casting. The validation of the computed hot tearing indicator is given by experiments. It is shown that the susceptibility of hot tearing can be predicted accurately for different alloys and casting processes. Therefore, the proposed hot tearing indicator can be effectively used in the numerical simulation of metal casting. REFERENCES

109

1. Guo J, Samonds M. Modeling of casting and solidification processing Fundamentals for Metals Process Modeling, Volume 22A, ASM Handbook, 2010, to appear. 2. ProCAST 2009.0 User Manual, ESI Group Inc., 2009. 3. Guo J, Zhu JZ. Prediction of hot tearing during alloy solidification. 5th Decennial International Conference on Solidification Processing SP07, University of Sheffield, UK. 2007; pp. 549-553. 4. Eskin DG, Katgerman L. A Quest for a new hot tearing criterion. Metallurgical and Materials Transaction A 2007, 38:1511 -1519. 5. Stangeland A., Mo A., M'Hamdi M., Viano D., Davidson C. Thermal strain in the mushy zone related to hot tearing. Metallurgical and Materials Transactions A: Physical Metallurgy and Materials Science 2006; 37(3): 705-714. 6. Eskin DG, Suyitno, Katgerman L. Mechanical properties in the semi-solid state and hot tearing of aluminium alloys. Progress in Materials Science 2004; 49:629-711. 7. Monroe C., Beckermann C. Development of a hot tear indicator for steel castings. Materials Science and Engineering A, 2005; 413-414:30-36. 8. Magnin B, Maenner L, Katgermann L, Engler S. Ductility and rheology of an Al-4.5% Cu alloy from room temperature to coherency temperature. Mater Sci Forum 1996; 1209:217-222. 9. Zhang J., Singer RF. Hot tearing of nickel-based superalloys during directional solidification. Ada Materialia, 2002; 50(7): 1869-1879. 10. Rappaz M, Drezet JM, Gremaud M. New hot-tearing criterion. Metall Mater Trans A, 1999; 30:449-455. 11. Eskin DG., Suyitno, Katgerman L. Mechanical properties in the semi-solid state and hot tearing of aluminum alloys. Progress in Materials Science 2004; 49(5):629-711. 12. Gurson AL. Continuum Theory Of Ductile Rupture By Void Nucleation And Growth: Part 1: Yield criteria and flow rules for porous ductile media. Journal of Engineering Materials and Technology. 1997;99:2-15. 13. Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a round tensile bar. Acta Metall. 1984;32:157-169. 14. Needleman A, Tvergaard V. An analysis of ductile rupture modes at a crack tip. J. Mech. Phys. Solids 1987;35:151-183. 15. Zhang ZL., Thaulow C , Odegard J. Complete Gurson model approach for ductile fracture. Engineering Fracture Mechanics 2000; 67(2): 155-168. 16. Perzyna P. The constitutive equations for rate sensitive plastic materials. Quart. Appi. Math. 1963; 20:321-332. 17. Armstrong PJ, and Frederick CO. A Mathematical Representation of the Multiaxial Bauschinger Effect. GEGB: Report RD/B/N731, 1966. 18. Hartmann S, Luhrs G, Haupt P. An efficient stress algorithm with applications in viscoplasticity and plasticity. International journal for numerical methods in engineering 1997; 40:991-1013. 19. Zienkiewicz OC, Taylor RL, Zhu JZ. The Finite Element Method: Its Basis and Fundamentals. Elsevier, 2005. 20. Zienkiewicz OC, Taylor RL. The Finite Element Method: For Solid and Structural Mechanics. Elsvier, 2005. 21.Cao G, Kou S, Chang YA. Hot cracking susceptibility of binary Mg-Al alloys. Magnesium Technology. Edited by Luo AA. TMS. 2006; pp. 57-61. 22. Cao G, Kou S, Hot tearing of ternary Mg-Al-Ca alloy castings. Met. and Mat. Trans. A 2006; 37:3647-3663.

110

23. Shimin Li , PhD thesis, "Hot Tearing in Cast Aluminum Alloys: Measures and Effects of Process Variables", Worcester Polytechnic Institute, 2010

111

Shape Casting: The 4"1 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

EFFECT OF ALLOYING ELEMENTS (MAGNESIUM AND COPPER) ON HOT CRACKING SUSCEPTIBILITY OF AlSi7MgCu-ALLOYS Salar Bozorgi1, Katharina Haberl1, Christian Kneissl2, Thomas Pabel2, Peter Schumacher1'2 'Chair of Casting Research, Metallurgy Department, University of Leoben, Franz-Josef-Str. 18, 8700 Leoben, Austria 2 Austrian Foundry Research Institute, Parkstr. 21, 8700 Leoben, Austria Keywords: hot cracking susceptibility, terminal freezing range, cracking susceptibility coefficient, hot cracking index Abstract Hot cracking during solidification can be a serious problem in aluminium casting alloys under certain conditions. This feature is well known but still insufficiently investigated in shape casting. This study gives a brief overview of the factors influencing hot cracking during shape casting. Five different AlSi7MgCu-alloys with varying Mg and Cu contents were examined. Theoretical models including the cracking susceptibility coefficient (CSC) from Clyne and Davies have been considered. Thermodynamic calculations of the behaviour of the fraction solid during solidification have been compared to an experiment based hot cracking indexing (HCI) method. Scanning electron microscopy (SEM) was used to compare existing microstructure and precipitated thermodynamic phases. Furthermore, SEM was used to investigate crack surfaces initiated by a dog bone shaped mold during casting. A good correlation between theoretical models and the experimental hot cracking index method was observed. Introduction AlSi7MgCu-alloys find wide application in many castings especially in the automotive industry. Complex thin walled components, such as cylinder heads, can be achieved. One serious problem in shape casting can be hot cracks which are fundamentally influencing the quality characteristics of a casting. In general the hot cracking susceptibility of AlSi-alloys is lower than in other Al-alloys such as AlZn, AlMg, or AlZnMg(Cu) [1-3]. However, various amounts of alloying elements can affect the hot cracking susceptibility of AlSi-alloys. In grain refined alloys hot cracks occur when insufficiently feeding by two phase flow and liquid flow between grains cannot accommodate the deformation caused by a hindered shrinkage [4]. At the point of rigidity bridges are formed between grains which do not permit further two phase flow. Subsequent micro feeding between grains cannot compensate shrinkage, stresses and strains occur, so that hot cracks can be generated in the final stage of solidification [5-7]. These cracks remain in the solidified casting. However, the exact mechanism nucleating a hot crack is still under discussion. Theoretical Background Influencing Factors. The most important factor on hot cracking is the chemical composition affecting freezing range, grain size, fraction of eutectic and segregation for a given casting process.

113

Freezing Range. In general as the freezing range increases the hot cracking susceptibility also increases. Depending on cooling conditions, a long freezing range leads to the formation of complex dendrites which interlock at relatively low fraction solid to form rigid bridges. Subsequently, feeding at the late stages of solidification is greatly hindered. Because pure metals and eutectic alloys have little to no freezing range, they show no hot cracking susceptibility [7-9]. The chemical composition is the main influencing factor on the freezing range. Impurities and their segregations which increase the freezing range are deleterious [9]. Furthermore, the final freezing range, the so-called terminal freezing range (TFR), is of major importance. A large TFR is objectionable; it causes a higher risk of hot cracks in the last stage of solidification [9]. If in an eutectic system a large amount of dendrites is formed already well above the solidus (i.e. at high temperature), the alloy possesses a high strength during final solidification of the remaining liquid, resisting contractional stresses. For alloys close to eutectic composition, large amounts of liquid freeze isothermally at the eutectic temperature (i.e. at low temperature) and shrinkage stresses are kept small [9]. It has been suggested by Djurdjewic et al. [10] to define TFR in temperature intervals of mass fraction solid 88-98%, 85-95% or others. In this study the solid fraction for TFR is defined as 95-99.5%. The very last percentage is neglected because of susceptibility to errors [10]. Grain Size. A fine grain size causes better feeding and uniform distribution of eutectic phases. When eutectic is present at grain boundaries, it has the maximum effect on permitting free movement of grains to accommodate contraction of the casting by two phase flow [11]. Bishop [12] and Lees [13] considered the effect of grains on hot tearing. They suggested that coarse grains result locally in a high thermal concentration of strain per grain boundary and, therefore, to hot cracking. In contrast a fine grain size results in a decrease in strain concentration accompanied by a decrease in hot cracking tendency [12,13]. However, the deformation of a granular structure should be considered as a movement within a network of grains and not of individual grains. The most common way to obtain fine grains is the addition of grain refiner or to increase the cooling rate. In this study the grain size was kept constant for die cast samples (~ 250 μιη) and sand cast samples (~ 350 μπι). Fraction of Eutectic Phase. A high fraction of eutectic phase in the microstructure and an eutectic phase with sufficient wettability results in a decreasing susceptibility for hot cracking. The eutectic surrounds the entire primary crystalline grains. Furthermore, a sufficient eutectic film between grains eases the movement of the granular system. If contraction and stresses occur, developing cracks are healed by backfilling [7,8]. It is important to note for Sicontaining alloys that Si exhibits a volumetric expansion during solidification and thus helps micro feeding. Small amounts of impurities which exist in the melt can form low melting eutectics. If more strain is imposed the tendency towards hot cracking increases markedly [12]. The reason for this is the weak bridging between dendrites. When tensile stresses occur weak bridges degrade, a hot crack may form between the grains [14,15]. Theoretical Models. There are various theoretical models for the calculation of the hot cracking tendencies. The most commonly used is the cracking susceptibility coefficient (CSC) model from Clyne and Davies for shape casting. [16]. However, the model describes only the material properties based on Gulliver-Scheil assumption and not the casting process condition. Other models are e.g. from Katgerman [17], Feurer [18] or Rappaz et al. [19]. However, all the mentioned models are not always applicable to different casting processes such as continuous, direct, chill, shape casting or welding. The CSC model correlates the susceptibility-composition relationship based on the consideration of the time during which processes related to crack production may take place and the structure is most vulnerable to cracking (critical time interval during solidification). The CSC is defined as = tv/tR; t v is the

114

vulnerable time period and is calculated as the time difference between mass fraction of liquid 10% and mass fraction of liquid 1%. tR is the time available for stress relief processes and is calculated as the time difference between mass fraction of liquid 60% and mass fraction of liquid 10%. A comprehensive study on the hot cracking susceptibility was performed to compare theoretical and practical techniques. Therefore CSC was examined semi-empirical and HCI was examined experimental. In this present work five different AlSi7MgCu-alloys with varying Mg and Cu content were investigated. Experimental Five different AlSi7MgCu-alloys with varying Mg and Cu-content, AlSi7Mg0.1Cu0.05, AlSi7Mg0.1Cu0.5, AlSi7Mg0.3Cu0.05, AlSi7Mg0.6Cu0.05, and AlSi7Mg0.6Cu0.5, were examined by using subsequently mentioned methods. The experimental tests were performed in sand and in die casting to evaluate the effect of the casting process. TFR. The TFR was calculated by the software ThermoCalc Classic (TCC) (Stockholm, Sweden), the database used was TTA15. For simulation of the solidification process existing phases and their fraction at the different temperatures were calculated for non equilibrium using Gulliver-Scheil. For the forecast of precipitated phases in the as-cast microstructure at room temperature equilibrium conditions were chosen. CSC. CSC was calculated semi-empirically using TCC for the evaluation of temperatures and mass fractions combined with practical thermal analysis in a permanent die mold (die temperature 250°C) and a sand mold for evaluation of associated times for tv and tR. The thermocouple used for thermal analysis was a type K-element. HCI. For HCI examination experimental casts in dog bone shaped die mold (die temperature 250°C) and sand mold were performed. The molds were identical in shape apart from the gating system. Fig. 1 shows the dog bone shaped sand casting. HCI is defined as = £(NOC*WF)/NOF; NOC is the number of cracks, WF is the weighting factor, depending on the observed level of hot cracking (see Fig. 2) and NOF is the number of castings [11,20,21].

Figure 1. 3D-picture of dog bone shaped sand casting for HCI evaluation.

Figure 2. WF for various hot cracking levels [21,22].

115

The HCI can be defined as follows [22]: • < 0.5 no hot cracking susceptibility • 0,5 - 1.25 small cracking susceptibility • 1.25-2.25 moderate cracking susceptibility • 2.25-3.5 high hot cracking susceptibility • >3.5 very high hot cracking susceptibility Microscopy. SEM examination was performed at 20 kV in BSD-mode to compare the as-cast microstructure with results from TCC and to investigate fracture surfaces. Results As-Cast Microstructure. Existing phases in the as-cast microstructure of various alloys were calculated by TCC (equilibrium conditions) and are shown in Fig. 3. Microstructure examination with SEM confirmed the theoretical predicted results. Alloy AlSi7Mg0.6Cu0.5 is given as an example in Fig. 4 to compare forecast phases by TCC and detected phases by SEM. Qualitatively, it is apparent from 50 EDX point analysis that in the sand mold a higher fraction of Mg2Si can be found.

Figure 3. As-cast phases at room temperature, calculated by TCC in equilibrium.

Figure 4. SEM, BSD, AlSi7Mg0.6Cu0.5, as-cast phases, (a) die mold, (b) sand mold. 116

Crack Surfaces. Crack surfaces initiated during casting of the HCI-samples in the dog bone shaped die were investigated by SEM. Samples with a small hot cracking level, i.e. samples not completely separated by a crack, were mechanically opened to subsequently observe the crack surface. Fig. 5 shows three SEM pictures of various hot cracking levels. SEM results indicate that at areas next to hot cracks no or insufficient eutectic phase exists. Furthermore, detailed SEM investigation of the fracture surfaces revealed no presence of bifilms as these may act as crack initiation sides within interdendritic liquid.

Figure 5. SEM, fracture surfaces, (a) dendrites in fully broken sample, WF=1, (b) dendrites and eutectic phase in sample with modest crack, WF=0.5 - mechanically opened, (c) eutectic in sample with hair crack, WF=0.25 - mechanically opened. TFR. Table 1 shows the TFR of all alloys. It is evident that the Cu-content has the dominating influence on TFR over that of Mg-content. Firstly, a high Cu-content results in a large TFR. Secondly, a low Mg-content results also in large TFR. Hence, the largest TFR is obtained in the alloy AlSi7Mg0.1Cu0.5 (see Fig. 6), the smallest TFR is obtained in the alloy AlSi7Mg0.6Cu0.05 (see Fig. 7). Table 1. TFR of evaluated alloys, calculated with TCC Alloy TFR [°C] AlSi7Mg0.1Cu0.5 46.0 27.0 AlSi7Mg0.6Cu0.5 AlSi7Mg0.1Cu0.05 17.0 AlSi7MgO.3CuO.05 9.5 AlSi7Mg0.6Cu0.05 4.0 CSC. Table 2 shows the CSC of three evaluated alloys. Again Cu has the dominant influence on the CSC. A high Cu-content results in a high CSC, a low Mg-content results also in a high CSC. Furthermore, the CSC results show that the CSC is much lower in sand casting than in die casting. The reason for this is a longer solidification time in sand casting and the larger amount of eutectic present which may induce a healing process for cracks. Table 2. CSC of evaluated alloys. CSC [-] Alloy Die Mold Sand Mold 0.69 7.3 AlSi7Mg0.1CuO,5 0.36 4.5 AlSi7Mg0.6CuO,5 0.33 3.7 AlSi7Mg0.1Cu0.05

117

Figure 7. TCC, calculation of TFR (4°C), AlSi7Mg0.6Cu0.05. HCI. Table 3 shows the HCI and subsequent resulting hot cracking susceptibility. For every alloy five hot cracking samples were investigated (NOF=5). Again Cu has a dominant effect on HCI. A high Cu-content results in a high HCI, a low Mg-content results also in a high HCI. Furthermore, all hot cracking susceptibilities for alloys in sand casting are negligible. Table 3. HCI and hot cracking susceptibility of evaluated alloys. HCI [-] HCI [-] Hot Cracking Hot Cracking Alloy Susceptibility Susceptibility Die Mold Sand Mold 0.8 0.01 AlSi7Mg0.1Cu0.5 no susceptibility small susceptibility AlSi7Mg0.6Cu0.5 0.6 no susceptibility 0.01 small susceptibility AlSi7Mg0.1Cu0.05 0.3 0.01 no susceptibility no susceptibility AlSi7Mg0.3Cu0.05 0.22 no susceptibility no susceptibility AlSi7Mg0.6Cu0.05 0.01 no susceptibility no susceptibility Summary of Results. Figure 8 shows in a summary of results the theoretical models and the experimental hot cracking index method for different AlSi7MgCu-alloys. On the left y-axis TFR values are plotted. On the right y-axis CSC and HCI values are plotted, the HCI values are multiplied by 10 so that it was possible to show both measurement values on one axis.

118

Figure 8. Trend lines of TFR, CSC and HCI for different AlSi7MgCu-alloys for theoretical and experimental methods for measuring hot cracking susceptibility. Discussion A brief overview of influencing factors on hot cracking was given. Five different AlSi7MgCu-alloys with varying Mg and Cu content were evaluated with three methods: theoretical TFR (Gulliver-Scheil condition), semi-empirical CSC model (Gulliver-Scheil condition) and experimental HCI examination. In contrast to the review for DC casting by Eskin et. al [4] all three performed examinations indicate the same trend (see also Fig.8): The Cu-content has a dominating influence on hot cracking susceptibility in AlSi7MgCu-alloys. A high Cu-content results in a large hot cracking susceptibility (large TFR, high HCI and high CSC), a high Mg-content results in small hot cracking susceptibility (small TFR, low HCI and low CSC). Furthermore, theoretical predicted phases were also found in SEM investigations. At higher Cuconcentrations Cu-phases segregate in form of Al2CuMg, Al5Cu2SÌ6Mg8 and Al2Cu during solidification; this has a negative effect and depletes the alloy of eutectic available for micro feeding. Despite the fact that the grain size in sand casting is larger, in general a lower hot cracking susceptibility is observed in sand casting. The amount of precipitated Mg-containing phases in the eutectic in as-cast alloys is higher in sand casting than in die casting. Moreover, the soft sand mold can accommodate shrinkage strains. For AlSi7MgCu-alloys of similar grain size a good correlation between theoretical models and the experimental hot cracking index method was observed as a material property. Especially for the development of new casting alloys a theoretical tool to forecast the hot cracking susceptibility is of major interest. Experimental evaluation of hot cracking tendency is intricate. TCC calculations are an adequate method of predicting the hot cracking susceptibility qualitatively. Acknowledgement Part of this work was financially supported by the Austrian Research Promotion Agency FFG.

119

References [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16]

[17] [18] [19] [20] [21] [22]

F. Matsuda, K. Nakata, K. Tsukamoto, S. Johgan, "Combined Effect of Current Pulsation and Zr Addition on Improvement of Solidification Cracking of Al-Zn-Mg Alloy Weld Metal," Transactions ofJWRl, 14, No. 2 (1985), 99-104. F. Matsuda, K. Nakata, and Y. Shimokusu, "Effect of Additional Element on Weld Solidification Crack Susceptibility of Al-Zn-Mg", Transactions ofJWRl, 12, No. 1 (1983), 81-87. G.L. Petrov, A.G. Makarov, "The sensitivity of Al-Zn-Mg Alloy to Hot Cracking During Welding," Avtomaticheskaya Svarka, No. 9 (1961), 18. D.G. Eskin, L. Katgerman, "A Quest for a New Hot Tearing Criterion," Metallurgical and Materials Transactions Λ, 38 (2007), 1511-1514. E. Cicala, G. Duffet, H. Andrzejewski, D. Grevey and S. Ignat, "Hot cracking in Al-Mg-Si alloy laser welding - operating parameters and their effects," Materials Science and Engineering A, 395 (2005), 1-9. E. Schubert, M. Klassen, J. Skupin, G. Sepold, "Effect of filler wire on process stability in laser beam welding of aluminium-alloys," Proceedings of the 6th International Conference on C1SFFEL, Toulon, France (1998), 195-203. T.W. Clyne, G.J. Davies,"The influence of composition on solidification cracking susceptibility in binary alloy systems," The British Foundryman, 74 (1981), 65-73. E. Brunhuber, Giesserei-Lexikon (Berlin: Schiele & Schön, 14. Auflage, 1988), 1100-1102. A.A. Gokhale, "Solidification Cracking: A Review," Transaction of the Indian Institute of Metals, 39 (1986), 153-164. M.B. Djurdjevic, R. Schmid Fetzer, "Thermodynamic calculation as a tool for thixoforming alloy and process development", Material Science and Engineering A, 417 (2006), 24-33. S. Lin, "A study of hot tearing in wrought aluminum alloys" (Ph.D. thesis, University of Quebec, 1999), 7-68, 69-90. H.F. Bishop, CG. Ackerlind, W.S. Pellini, "Investigation of metallurgical and mechanical effects in the development of hot tearing", Trans. AFS, 65, 1957, 247-258. D.C.G. Lees, "The Hot Tearing Tendencies of Aluminium Casting Alloys," The Journal of the Institute of Metals, 72 (1946), 343. J.A. Spittle, A.A. Cushway, „Influences of superheat on grain structure on hot-tearing susceptibilities of Al-Cu alloy castings," Metals Technology, 10 (1983), S. 6-13. J.A. Dantzig, M. Rappaz, Solidification (Lausanne: EPFL Press, CRC Press, 2009), 519565. T.W. Clyne, G.J. Davies, "Comparison between experimental data and theoretical predictions relating to dependence of solidification cracking on composition," Proceedings of the Conference on Solidification and Casting of Metals, Metals Society, London (1979), 274-278. L. Katgerman, "A Mathematical Model for Hot Cracking of Aluminum Alloys During D.C.Casting," Journal of Metals (1982), 46-49. U. Feurer, "Mathematisches Modell der Warmrissneigung von binären Aluminium Legierungen," Giesserei Forschung, 28 (1976), 75-80. M. Rappaz, J.M. Drezet, M. Gremaud, „A New Hot-Tearing Criterion," Metallurgical and Materials Transactions A, 30A (1999), 449-455. B. Lenczowski, H. Koch, K. Eigenfeld, "Neue Entwicklungen auf dem Gebiet der warmfesten Aluminium-Gusswerkstoffe," Gießerei, 8 (2004), 32-38. A. Franke, „Design of new high-performance aluminum casting alloys" (Ph.D. thesis, University of Leoben, 2006), 50-61. C. Kneissl, T. Pabel, G. Dambauer, P. Schumacher, "Formenkonzept und Ergebnisse gießtechnologischer Versuche zur Legierungsentwicklung im Niederdruckkokillenguss," Giesserei-Rundschau, 56 (2009), 120-125.

120

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

HYDROGEN AND COOLING RATE EFFETCS ON MICROPOROSITY FORMATION IN THE PRODUCTION OF DEFECT-CONTROLLED FATIGUE SPECIMENS Rosario Squatrito1, Ivan Todaro1, Lorella Ceschini2, Andrea Morri2, Luca Tomesani1 ΌΐΕΜ (Department of Mechanical Engineering Constructions) Viale Risorgimento 2,40126 Bologna, Italy 2 SMETEC (Department of Metallurgy) Viale Risorgimento 4,40126 Bologna, Italy Keywords: Gravity casting, Aluminum Alloy, Fatigue Specimen, Microporosity Abstract In experiments aimed at the production of fatigue specimens, the increased number of nearly identical specimens needed for each processing condition, together with the high sensitivity to pore size, call for very strict requirements of both the casting tool and the processing conditions. An experiment for producing aluminium alloy fatigue specimens by gravity casting with controlled microstructure and defects is presented here. The main requirements to be obtained on a set of specimens (extracted from a single casting block) were to have the near identical microstructure and gas porosity content. The main process parameters were the hydrogen level of the melt, the addition of oxides for improving the number of pore nucleation sites and the cooling rate within the casting mould. The distribution of the relevant properties (SDAS, %area of porosity) was measured throughout the casting plates in order to validate the design criteria of both the experiment and the mould. Introduction Al-Si casting alloys, such as A356/A357, find extensive applications in the transport field, due to their excellent castability, corrosion resistance, and especially their high strength-to-weight ratio which increases performance and fuel economy. However, the casting process inevitably introduces solidification defects, which can significantly reduce the mechanical properties, mainly the elongation to failure and, above all, the fatigue strength of the final cast component. Several studies have shown that fatigue resistance of cast Al-Si alloys, is dominated by casting defects, such as gas and shrinkages pores, which considerably decrease the fatigue life with respect to defect-free cast components [1]. Only when the porosity is negligible (as in the case of castings subjected to hot isostatic pressing, HIP), the negative effect on the fatigue life of others solidification defects (such as oxide films) becomes dominant [2-3]. In order to evaluate the effect of solidification defects and other microstructural features on fatigue strength, the production of fatigue specimens with controlled microstructure and defects is crucial. In fact, when setting a tool for casting experiments, many sources of defects generate at both the filling and solidification stages have to be considered. To this aim, it is not only important to follow all the rules for good mould design, but it is also essential to carefully evaluate the specimen casting process by numerical analysis as if it was a process itself.

121

To meet these needs, the experimental activity presented in this paper was aimed to produce several sets of nearly identical casting specimens with different microstructure and amount of porosity. This goal was reached by means of a permanent gravity casting device, properly studied to ensure a sufficient control of the casting process, and by implementing a precise melt treatment procedure. 2. EXPERIMENTAL The aims of the experiments were: 1. Producing simple castings of A356 aluminium alloy, that would allow the extraction of two sets of nearly identical specimens in terms of microstructure and defects distribution. 2. Producing different castings, each one with different microstructure and porosity. 3. Producing a gradient of microstructure and porosity throughout each casting cross section. To achieve these aims it was necessary to have a high level of control of process conditions in terms of local cooling rate and preliminary melt treatment. 2.1 Design of the casting mould The general features for the casting mould, were set as follows: 1. Possibility to vary the cooling conditions in different casting experiments 2. Forcing of heat flux direction during solidification inducing planar isothermal surfaces, thus obtaining a microstructure gradient inside the casting 3. Filling conditions to avoid metal flow surface turbulence and trapping of surface oxides. These requirements resulted in the tool geometry outlined in Fig. 1. The global mould concept follows the classical arrangement of a gravity die tool, based on a vertical tapered sprue, followed by a 90° curve, a horizontal runner, a single gate connected at the bottom side of the casting and having the same section length of cast plate, which is filled from the bottom up. The casting geometry is a simple vertical plate, 30 mm in thickness, 300 mm in height and 250 mm in width. The mould was made of two elements (A and B in Fig.l) machined from two blocks of CK45 steel. In order to control the cooling conditions in different castings and the initial temperature distribution of the mould, the element (A) was crossed by a system of cooling channels positioned at 20 mm from the casting/mould interface, to be fed either by water or air. The element A could be heated by oxyfuel combustion as well. In order to obtain a gradient in the cooling rate throughout the casting thickness, a 30 mm layer of insulating material (C) was placed in a pocket machined in the element (B), preventing the casting from cooling in that direction. In order to check the initial temperature distribution of the mould, and to record the evolution of temperature with time, eight holes 1.5 mm in diameter were drilled through the back of element (A) to allow thermocouples embedded at 5 mm and 10 mm from the casting/mould interface. The filling system dimensions (runner sections reduction, gate length and depth) were studied by means of numerical simulation with PROCAST v. 2008, in order to avoid early metal flows into the plate during the filling of the runner (metal stops) and to fill the mould evenly.

122

Fig. 1: Outline of the casting tool A critical limit around 0.5 m/s of melt velocity inside the gate was assumed as an optimization criterion to define the dimensions of the inlet system, in order to avoid any disturbance coming from the unevenness of structure that could be related to the filling phase: in particular gas entrapment, turbulence, excessive speed, filling stops [4- 5]. To meet these requirements two devices were used: 1. a cylindrical filling reservoir placed above the filter (positioned horizontally) to absorb the peaks of metallostatic pressure due to the fluid-dynamics behaviour of metal during the filling of the basin and the downsprue; 2. a trap at the end of the rise channel, to avoid the flow into the plate cavity of oxides that form on the metal free surface in the first steps of riser filling. The filling analysis of the adopted single gate solution showed the formation of a large fluid recirculation (Fig. 2) in the vertical plane of the casting that could force convective flows during cooling and solidification. The optimal inlet-gate casting dimensions were then evaluated by minimizing the momentum of the recycled fluid structure, taking into account the massive dimensions of the cast and the presence of an imposed thermal gradient along the plate section. Numerical results showed that the ratio between first cooling time (above T|jq) and insurgence time of convective flow meant that the temperature differences inside the casting at the end of filling phase were negligible. In all of the experimental conditions, the maximum temperature difference between the top and bottom of the casting was below what was considered to be consistent with the requirement of 'near identical' specimens production (Fig. 3). The numerical thermal analysis performed with this mould showed that a planar solidification front proceeds from the cold side to the hot side of the mould. The predicted extensions of isothermal surfaces and thermal gradient along the cast depth, suggested the capability to obtain two casting portions where specimens can be extracted, thus producing two sets of specimens (one of faster solidification, the other of lower solidification rate) (Fig. 4), each one characterized by nearly identical cooling rate conditions.

123

Fig. 2: Flow Streamlines during rilling

Fig. 3: Temperature distribution at the end of filling

2.2 Melt Treatment In order to achieve different amounts of porosity in the castings the initial hydrogen content in the melt was varied. To get different and controlled hydrogen content in the molten material, it was necessary to build a simple gassing feeding system to the rotary impeller and to formalize a melt treatment procedure. In order to always have the same initial conditions, each charge of molten material in the furnace underwent the same preliminary treatment, consisting of Sr addition for eutectic Si modification ,Ti-B addition for grain refinement and dross removal. Several Ar degassing cycles by means of a rotary impeller (time of up to 45 minutes) and then the measurement of hydrogen content through Foseco Alspek H probe completed the start-up procedure of the experiment. The initial targeted hydrogen content was 0.08 - 0.1 mlHi/lOOg. After the preliminary treatment, it was possible to reach the desired hydrogen levels (Tab.l) simply by up-gassing the melt through the rotary impeller fed by a 10% II2- 90% Ar mixture . Table 1 Hydrogen Level Hydrogen Level (HL) Val.(mlH2/100g)

HL0 - very low

HL1 -low

HL2 -medium

HL3-high

0.08

0.13-0.16

0.23-0.24

0.30-0.33

Another aim of the experimentation was to study the effect of oxide content on porosity nucleation. For each hydrogen level, the oxide content was varied by introducing dry air through the gassing system. The oxidation level was determined by the amount of gassing treatment time (Tab. 2): Table 2 Oxide Level Oxide Level (OL) Val (minutes)

OL0 0

OLI 1

OL2 3

OL3 5

2.3 Microstructural characterization To evaluate the correspondence between the targets of the experiment and the actual achieved the central zones of plates were investigated, corresponding to the part of the casting designed for the extraction of fatigue-test specimens (Fig. 4).

124

Fig. 4: Cast plate and fatigue specimen Fig. 5: Metallographic specimen extraction For this purpose , 5 mm thick slices were cut from each cast plate at 130 mm from the gate (Fig. 5) then 30x10 mm metallographic specimens were extracted in the center of every slice to perform the microstructural analysis along their whole thickness. The specimens were prepared using standard metallographic techniques, according to ASTM E3 [6]. Qualitative and quantitative metallographic analyses were carried out, using an optical microscope (OM) and an image analysis software (Image pro-Plus®), on approximately 30 optical micrographs for each specimen. The microstructural characterization was focused on the evaluation of: • secondary dendrite arm spacing (SDAS [μηι]), • percentage area fraction of defects (DAF%) • number of defects per cm"2 (Nd [cm"2] ) • maximum Feret diameter (largest side of the rectangle enclosing the defect [μηι]) • roundness R of every defect, defined as: (Perimeter d e f e r t ) 2 4-rtAreadefect Where R=l means a circular shape and higher values means more irregular shape =

All of the microstructural data were evaluated as a function of the distance from the uninsulated side of the mould, the hydrogen content, the oxide level and the different cooling system of the mould (air or water cooling). 3. RESULTS AND DISCUSSION SDAS trends on the casting thickness The analysis of all of the average experimental SDAS data highlighted an increase with the distance from the mould side (Fig. 6), testifying to the actual gradient of cooling rate achieved. In Fig 6. average SDAS values are separated into two classes depending on the way the mould was cooled. Water cooling induced a difference in the SDAS values of up to 10 μτη in the central part of the plate while no variation was obtained close to the mould and insulator sides due to their starting temperature. Effect of hydrogen content Hydrogen content in the melt proved to have a high influence on porosity. Low hydrogen level (HL0) castings showed no porosity but in the side near to the insulator (Fig.7). Moreover the

125

number of pores was low (Fig.8), and they had a high R value (Fig.9) so that they appear to be "mainly shrinkage driven".

Distance from Che not insulated side of the castings, mm

Fig.6: SD AS as a function of the distance from the not insulated side of the mould, for air /water cooled castings.

Fig. 7: Average values of the percentage area fraction of defects, as a function of the distance from the mould, at the different hydrogen levels.

Fig.8: Average values of the number of defects as a function of the distance from the mould, at the different hydrogen levels

Fig. 9: Average R values of defects, as a function Fig. 10: Average values of the max ferèt of the distance from the mould, at the different diameter of defects, as a function of the hydrogen levels. distance from the mould, at the different hydrogen levels.

126

As the hydrogen level went up Percentage Area Fraction of defects, Maximum Feret and Number of Pores/cm2 increased accordingly (Figg. 7,8,10). A little decrease of the R values was observed as well, (the lower values for higher hydrogen level castings (HL3)), testifying a trend towards more rounded shape, thus an effect of hydrogen content on pore growth. As expected a dependence of porosity on cooling rate (i.e. the distance from the cold side of the mould) was observed. Fig.7 and Fig.8 show an increase of Percentage Area Fraction of pores and Number of Defects going from the cold to the hot side of the casting. Effect of oxide content Oxides are supposed to act as nucleation sites for porosities. The higher the amount of oxides in the melt the higher the number of pores. Moreover, considering the same amount of hydrogen in the melt, it is supposed to get an increase of the number of pores with a decrease of pore dimensions. This effect could be useful in the production of fatigue specimen to "drive" the porosities dimension. Owing to the high standard deviation of measurments, the expected behavior could not be consistently inferred, nevertheless Fig. 11(a) suggests a general increase in the number of pores nucleated and grown, in function of the oxidation time. Moreover, Fig 11 (b) suggests a trend of Feret max decreasing with oxidation time, for the highest hydrogen level only. As a final remark it must be underlined that the control of the melt oxidization by dry air addition in the melt was really a challenging issue: the fine and dispersed bubbles of dry air delivered to the melt with the rotary impeller, several times removed the hydrogen from the melt and cleaned the melt itself from the largest part of oxides, making them float towards the surface.

(a) (b) Fig. 11: Pore Number per cm2 (a) and max Feret of the pore (b) in function of oxide level 4. CONCLUSIONS In this work, an experimental method to obtain sets of fatigue specimens with nearly identical microstructure and defects content was proposed. First, a casting tool was designed and simulated in order to allow the production of a set of specimens in nearly identical processing conditions.

127

Second, a casting experiment was performed by controlling all of the process variables that have an influence on the generation of microporosity: cooling rate, hydrogen level, oxide presence. For this aim, a specific melt treatment procedure based on the set up of hydrogen content and oxide generation was developed and validated. Metallurgical analysis on cast material confirmed the influence of increased hydrogen level on pores growth [7]-[8] and showed the increment of porosity percentage area fraction, pores dimensions and the evolution towards more rounded pore morphology. The control of the melt oxidation was difficult to achieve. This is probably due to the particular way to increase the oxide level that adopted the use of dry air. Despite this, the experience showed a good repeatability of experimental results, suggesting its usefulness as a prospective evaluation of numerical models for the prediction of gas porosity defects. REFERENCES [1] Wang QG, Apelian D, Lados DA. Fatigue behavior of A356-T6 aluminum cast alloys. Part I. Effect of casting defects. J Light Metals 2001;1:73. [2] Wang QG, Apelian D, Lados DA. Fatigue behavior of A356/357 aluminum cast alloys. Part II. Effect of microstructural constituents . Journal of Light Metals 1 (2001);85:97. [3] Ceschini L, Morri, A, Sambogna G. The effect of hot isostatic pressing on the fatigue behaviour of sand-cast A356-T6 and A204-T6 aluminum alloys. Journal of materials processing technology 2008;204:231-238. [4] J.Campbell and R.A. Harding, "TALAT: The freezing of Castings", European Aluminum Association, (1994) Lecture 3204. [5] J. Campbell, Casting Practice - the 10 rules of castings, Elsevier, Oxford, 2004 [6] ASTM E3-01 (2007) Standard Practice for Preparation of Metallographic Specimens [7] J-Y. Buffiere, S. Savelli, P.H. Jouneau, E. Maire, R. Fougères, "Experimental study of porosity and its relation to fatigue mechanisms of model Al-Si7-Mg0.3 cast Al alloys", Material Science and Engineering A316 (2001) 115-126 [8] D. Dispinar, A. Nordmark, J. Voje, and L. Amberg, Influence of Hydrogen Content and Bifilm Index on Feeding Behaviour of A1-7SÌ Alloy, Shape Casting: 3rd International Symposium 2009, pp. 63-70 [9] J. A. Dantzig, M. Rappaz, Solidification, EFPL Press, 1st edition, 2009, Lausanne Acknowledgements The authors would like to acknowledge Ferrari SpA and in particular Eng. Gianluca Pivetti for general collaboration, Fonderie Scacchetti and Eng. Lorenzo Pivetti, FOSECO-Italy for providing technical facilities. Great acknowledges also to Franco Iorio (Modelleria CPC), for the mould production.

128

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbeil, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

EFFECTS OF GRAVITY ON THE COLUMNAR TO EQUIAXED TRANSITION IN DIRECTIONAL SOLIDIFICATION. Wajira U. Mirihanage, David J. Browne School of Electrical, Electronic and Mechanical Engineering, University College Dublin, Belfield, Dublin 4, Ireland Keywords: CET, Grain transport, Solidification Abstract In industrial casting processes microstructure plays a major role in determining the properties of the final cast product. Columnar to equiaxed transition (CET) is a frequent result of the evolving grain structure during alloy solidification. In this contribution, we analyze CET in directional solidification via numerical simulations. The numerical model employs front tracking to track columnar growth and a volume average approach to account for the evolution of the equiaxed zone. The effects of gravity, thermal natural convection and dendrite transport were integrated into the model. Simulations of vertical directional solidification of an Al-7%wt.Si alloy both in and opposite the direction of gravity were conducted for different cooling conditions. Here, we present a preliminary analysis of these numerical simulations and a comparison of the predictions with previously published theoretical and experimental work. Introduction Dendritic microstructures are the most common type in cast alloys. Columnar and equiaxed grains are normally present in the macrostructure of alloy castings. When both types co-exist, a distinguishable change, known as the Columnar-to-Equiaxed Transition (CET), is often visible. Much research attention has been paid to analytical, experimental and computer modeling of CET phenomena. Various CET models have been presented at the macro/mesoscopic scale [1-8] and most of this work has recently been reviewed by Spittle [9]. However, many CET models ignore the effects of gravity during solidification. Only a few CET models (e.g. ref. [6,7]) consider the effects of gravity on solidification, but they lack explicit tracking of the columnar dendrite front. In the present contribution, a Front Tracking (FT) algorithm [8,10-12] is used to track the columnar growth under the effects of natural convection. A volume average model of equiaxed solidification [13,14] which considered the effects of gravity is used to model the equaxed dendrite evolution during solidification. Both the models are then combined together via considering macroscopic energy, momentum and mass conservation. The CET model was used to simulate vertical directional solidification of an Al-7%wt.Si alloy both in and opposite the direction of gravity, with different cooling rates. The simulation results are analyzed by contrasting the effects of gravity and previous experimental results [15,16].

129

The Computational Model The CET prediction model presented here is a combination of the columnar FT model [10] and the volume averaged equiaxed solidification model [13]. Descriptive details of these two models can be separately found in the previous literature [8,10-14]. Previously, combined CET models were presented for the pure thermal diffusive condition [17] and for the thermal diffusive/convective condition, but ignoring equiaxed dendrite sedimentation [18]. Hence, only the basic details of the model are outlined here. For the columnar front tracking and equiaxed volume averaging model, the transient conservation equations govern the transport of energy, momentum and mass.

ar dt

:

d(uT) t d(vT) dx

dy

du du1 d(uv) — + + ——'- = dt dx dy dv d(uv) dv2 σί dx dy du dv — + — dx dy

-

k pC 1 dp p dx 1 dp pdy

d2T dx2

+

d2T dy2 d2u dx2

+

(1) d2u dy2

μ a^v as, p\_dx2

dy2

(2)

+ {T-Tre/)ßgy+ sy + p

(3) (4)

= 0

Where, u, v ,T, t, p, μ, p, Cp, k and β are velocity in x direction, velocity in y direction, temperature, time, pressure, dynamic viscosity, density, specific heat, thermal conductivity and volumetric thermal expansion coefficient, respectively; Tre/ is the reference temperature; S, P and E are source terms relating to flow resistance from the porous medium, momentum effects from dendrite sedimentation, and latent heat, respectively; g is an external force (gravity). Heterogeneous nucleation at the mould wall is considered for the columnar grains. So at a given nucleation undercooling, columnar grains start to grow from the domain boundaries. For equiaxed grains, free growth from inoculant particles is computed as explained in rei [13], considering a log-normal distribution of commercial inoculant particle diameters [19]. In the current work, growth of dendrite envelopes is computed using a simplified dendrite tip velocity relationship [20]. A low thermal gradient is assumed and the following modified equation is used to calculate dendrite tip velocity v, of Al-7%Si at undercooling AT, (5)

C AT" where, C and n are alloy dependent constants.

130

The transportation of nucleated equiaxed grains with melt flow is performed by considering the system as a slurry [13], but subjected to sedimentation setting. Therefore, dendrite velocity in the vertical direction is the vector sum of the flow velocity and the settling velocity. So, solid crystals do not necessarily stay on the same transport path as the molten metal flow. For this deviation, a momentum difference to the single phase conservation equation arises and is accounted for via a source term {equation no.(3)}. The source term, f i s defined as [14], P

= / , £ at

(6)

For the slurry, viscosity is a function of the volume fraction solid volume/; modified viscosity is obtained using Thomas' empirical approximation, as cited in [21]. The equiaxed dendrites are free to move until dendrites form a coherent network locally. Instead of using a pre-defined coherency fraction, the coherency point is defined as the point where equiaxed envelopes first fill the local space (control volume), i.e. when equiaxed envelope volume fraction reaches unity [14]. The detailed consideration of prediction of dendrite coherency fraction and comparison with experimental observation is the subject of a separate article under preparation. The average dendrite sedimentation/settling speed w is calculated using averaged characteristic dendrite parameters. The complete mathematical description of these calculations is presented in ref. [22]. Average and equal constant density for both solid and liquid (p=ps=pi) is assumed, except in the settling calculations where appropriate densities are considered (p*s>p*i). Where, Ps, pi, P*S and p*i are nominal density of the solid, nominal density of the liquid, actual density of the solid and actual density of the liquid, respectively. The columnar zone and coherent equiaxed zone are treated as a porous medium. In porous media, the phase interaction forces are proportional to the liquid velocity (liquid velocity relative to the medium), and source terms Sx and Sy are given by [11,12], S , = — P

(7)

Ku

Sy = -Ü2. K v

(8)

where AT is a component of the permeability tensor - a physical property of a porous medium. It is common practice to simplify these equations by the assumption that the mush is isotropie [11,12]. Therefore, the permeability tensor K is defined by the Blake-Carman-Kozeny model using a morphological constant and the local solid volume fraction. The source term E in the energy equation accounts for the latent heat effects. Latent heat effects are incorporated by considering changes in the grain volume and local solid fraction [8,10-14]. These changes include the latent heat release during growth as well as absorption of latent heat during re-melting of the solidified grains.

131

*-[«■£♦"

(9)

Here, g s is the local solid fraction, and is calculated by assuming no diffusion in the solid phase and complete mixing in the liquid phase between secondary dendrite arms; so the Scheil approximation is used to calculate internal solid faction. The evolution of equiaxed and columnar solid fractions in each CV are separately but simultaneously calculated by average equiaxed calculation and columnar front tracking, respectively. The columnar front is free to progress until it meets a control volume that is fully occupied by the equiaxed grains. At this point, no further growth of the columnar front is possible and CET is predicted [17,18].

Results and Discussion Model simulations that include gravity-induced natural thermal convection and sedimentation effects were carried out for directional solidification of Al-7%Si. For these simulations, a mould 10cm high and 3cm wide was chosen and the dissipation of heat (cooling) from either top or bottom surface was investigated. All of the other surfaces were treated as adiabatic walls. A mixed boundary condition controls heat extraction from this cold surface, primarily set by choice of value for the heat transfer coefficient h. The chilled surface temperature was kept at a constant 473K. The total number of grain refiner particles in each melt was set to 0.3 x 106 particles per m2 (for 2D) and the initial melt temperature was set at 896K for all cases. Two different heat transfer coefficients (ft = 1,000 Wm"2K~'and h = 3,000 Wm~2K~') were applied for the different bottom and top cooled solidification simulations. For all of these simulations, primary dendrite arm tip velocity (for both columnar and equiaxed growth) was calculated according to the empirical relation given for Al-7%Si in ref. [20]. In these simulations, the columnar solidification front is either moving against the direction of the gravity vector or moving with it. According to basic casting experience, equiaxed zone formation is sensitive to the heat extraction rate from the solidifying melt [23]. Predicted equiaxed and columnar volume fractions are shown in Figure 1. These four simulation results can be compared and analyzed against the previously published experimental results. The possible engulfment of very small equiaxed dendrites in between primary columnar dendrite arms was considered in these simulations. For these computations, columnar primary interdendritic space was calculated according to the dimensionless relationship of Hunt and Lu [24]. It was assumed that some of the equiaxed dendrites are captured in between the columnar dendrite trunks during their growth. This is possible when the columnar front reaches an undercooled region and if equiaxed dendrites are not coherent and are smaller than this columnar spacing. Those equiaxed dendrites with growth orientations that are not parallel to the columnar

132

growth can be blocked by the columnar dendrites. Only equiaxed crystals with growth orientations that are aligned (parallel) with the columnar growth direction can still grow. Such equiaxed dendrites can grow further in an elongated manner, and may look like columnar grains in any post-mortem macrostructure analysis. Thus, no further growth can be expected for such small dendrites in an equiaxed manner. Therefore in the model computations, further growth of these encapsulated equiaxed dendrites is halted. According to the current simulations, the diameters of these dendrites are well below the average grain diameters of the fully equiaxed zone. Similar equiaxed islands encapsulated between columnar dendrite trunks were predicted in previous Cellular Automata -finite difference (CA-FD) directional solidification simulations by Dong and Lee [5]. Recently reported experimental observations on directionally solidified Al3.5%Ni alloy [16] also provide very good experimental evidence of the presence of such equiaxed grains stuck in between the columnar channels.

Figure 1 : As-cast columnar and equiaxed volume fractions (al) bottom cooled and h = 1000 Wm"2K_1 (a2) bottom cooled and h =3000 Wm"2K"' (bl) top cooled and ft = 1000 Wm"2K"' and (b2) top cooled and h =3000 Wm"2K"'. According to the simulation results in Figure 1, one can observe that low cooling rates (low h) contributed to increase the equiaxed zone size and shorten the columnar length, irrespective of the direction of solidification. This simulation outcome agrees with the indirect predictions [23] of formation of the equiaxed zone ahead of a columnar front. A similar experimental outcome was found [15] from a large number of directional solidification experiments. Furthermore, the simulation results agree with general casting experience [23], A snapshot view of the solid and liquid flow transport processes during bottom and top chilled directional solidification (h = 1,000 Wm"2K"'), 30 seconds after the cooling begins, is shown in Figure 2. Here liquid flow and sedimentation velocities are shown to the same vector scale. With the temperature gradient, we can see the liquid metal is circulating due to convection currents

133

and free solid equiaxed dendrites are falling towards the growing columnar front. In comparison to the top chilled solidification, convection currents seem to be relatively weak in the bottom chilled simulation (average flow speeds in the centre: top down - 1 . 0 mras'1 , bottom up ~ 0.2mms"'). In this bottom chilled simulation, the columnar front has progressed slightly further (2.7cm from the chill) than in the counterpart top chilled case (2.6cm from the chill). In the top chilled solidification, hot liquid is rising towards the columnar front and a resulting slight reduction in the undercooling is assumed to be the main reason for this difference. This could be a significant cause of the difference between the CET positions (see Figure 1) for top and bottom cooled solidification, where the top chilled solidification always has slightly lower columnar length. The shape of the columnar front can be seen to deviate slightly from planar, and the horizontal temperature gradient created due to the natural convection may be the cause of this. However, equiaxed dendrite sedimentation toward the columnar front can set conditions that favour early CET in the bottom chill solidification if they can become coherent. If not and they are small enough, free dendrites can sediment between columnar trunks and this can delay CET further. (a)

(b)

Figure 2 : Bottom (a) and top (b) chilled directional solidification (i) temperature and fluid flow (ii) solid fraction and equiaxed dendrite transport, 30 seconds after solidification starts, for h = 1000Wm"2K"' As shown in figure 1, the simulated CET has occurred through a small transition zone rather than at a sharp boundary. In this mixed region, equiaxed dendrites co-exist with columnar trunks. This nature of CET with a mixed zone rather than a sharp boundary between the columnar and equiaxed zone was experimentally observed in the directional solidification experiments conducted by Ares et al. [15]. A similar mixed region between the columnar and equiaxed zones was also reported from the directional solidification experiments with grain refined Al-7%Si and

134

Al-3.5%Ni [16]. The appearance of such a mixed region is understood to be dependent on the equiaxed dendrite nucleation, undercooling and dendrite transport conditions present ahead of the columnar front. There are a number of factors that can cause the rate of equiaxed growth to decrease ahead of the columnar front. These include (i) fewer nucleation events due to low undercooling, (ii) lower dendritic growth rate due to low undercooling, and (iii) high rate of equiaxed dendrite transport towards the superheated liquid. At a high cooling rate, the columnar front can progress quickly and can reduce the undercooled zone. A high rate of cooling can increase convective flow also. Therefore, one or more of these factors can delay the formation of a coherent equiaxed network ahead of the columnar front. However, in instances where no interconnected equiaxed network is present, columnar trunks are expected to grow through the growing equixed dendrites without facing any effective mechanical barrier, and thus a mixed region can be formed. The blocking fraction, the equiaxed envelope volume = 1.0, used here is different to the established mechanical blocking fraction of 0.49 [1], But, notably no direct physical explanation was given for choice of the extended volume fraction value of 0.66, which yields the volume fraction 0.49. Biscuola and Martorano [25] reviewed this mechanical blocking fraction value with deterministic and stochastic models. According to their analysis, an equiaxed volume fraction value of 0.2 for the mechanical blocking criteria closely agreed with their stochastic CAFD model with a solutal blocking criterion [4]. But it should be noted that the equiaxed envelope that was used to define the equiaxed volume fraction in the stochastic CA-FD model was different to that of the Hunt model [1]. According to Hunt, a circular equiaxed envelope was assumed, and the total equiaxed envelope volume was considered as the extended volume. In the stochastic CA model (in ref. [25]), the equiaxed envelope was based on a fine mesh. The latter part of the same ref. [25] describes the way that the columnar front is blocked: via fully impinged equiaxed grain envelopes. This is equivalent to the blocking of columnar front by the justcoherent equiaxed dendrites. Here it would seem that the most important factor is the presence of a continuous mechanical barrier to the growing columnar grains. Such a continuous barrier exists when equiaxed dendrites form an interconnected network.

Conclusions A model of CET in alloy solidification is presented and directional solidification of Al-7%Si is simulated. The CET model is a combination of a columnar front tracking model and an equiaxed volume average model. The combined model considers the effects of gravity on solidification such as thermal natural convection and equiaxed dendrite transport. A modified mechanical blocking criterion is employed in the model. Upward and downward directional solidification cases were simulated with different cooling rates. Simulations show gradual CET transition rather than a sharp boundary between columnar and equiaxed zones. According to the simulations, the effects of gravity promote equiaxed solidification. The simulations results are in agreement with previous directional solidification experiment results found in the literature.

135

Acknowledgements The authors wish to acknowledge the support of the European Space Agency (ESA) via PRODEX funding (contract number 90267). This work is part of the ESA-MAP (Microgravity Applications Promotion) project CETSOL.

References 1. J.D. Hunt, Mater. Sei. Eng., 1984, 65, pp. 75-83 2. S.C. Flood and J.D. Hunt, Journal of Crystal Growth,, 1987, 82, pp.552-560 3. C.Y. Wang and C. Beckermann, Metallurgical and Materials Transactions A, 1994, 25, pp. 1081-1093 4. M.A. Martorano, C. Beckermann, Ch.-A. Gandin, Metall. Mater. Trans. A, 2003, 34, pp.1657-1674 5. H.B. Dong, P.D. Lee, Acta Mater., 2005, 53, pp.659-668 6. A.Ludwig, M. Wu., Mater. Sei. Eng. A, 2005,413-414,ρρ.109-114 7. M. Wu, A. Ludwig, Metall. Mater. Trans. A,, 2006, 37, pp.1613-1631 8. S. McFadden, D.J. Browne, App. Mathematical Modelling, 2009, 33, pp. 1397-1416 9. J.A. Spittle, Inter. Materials Reviews, 2006, 51(4), pp 247-269 10. D.J. Browne, Hunt J.D., Num. Heat. Trans. B, 2004, 45, pp 395-419 l l . J . Banaszek, D. J. Browne , Mater. Trans., 2005, 46(6), pp 1378-87 12. J. Banaszek, S. McFadden, D.J. Browne, L. Sturz, G. Zimmermann, Metal. Mater. Trans. A.., 2007, 38A, pp 1476-84 13. W.U. Mirihanage, D.J. Browne, Comput. Mater. Sci., 46(4), 2009, pp.777-784 14. W.U. Mirihanage, D.J. Browne, Proceedings of Global Innovations in Manufacturing of Aerospace Materials: The 11th MPMD Global Innovations Symposium at the TMS Annual Meeting, Seattle, USA, Feb 14-18, 2010, pp. 249-256 15. A.E. Ares, S.F. Gueijman, R. Caram, CE. Schvezov, J. Crystal Growth, 2005, 275, pp. e319-e327 16. H. Jung, N. Mangelinck-Noël, H. Nguyen-Thi, B. Billia, J. Alloys and Compounds, 2009, 484, pp. 739-746 17. W.U. Mirihanage, S. McFadden, D.J. Browne, Materials Science Forum, 2010, 649, pp.355360 18. W.U. Mirihanage, S. McFadden, D.J. Browne, Proceedings of 3rd International Symposium on Shape Casting held at the 2009 TMS Annual Meeting, TMS, Warrendale, PA, USA, pp. 257-263 19. T.E. Quested, A.L. Greer, Acta Mater., 2005, 53, pp. 4643-4653 20. Ch.-A. Gandin, Acta Mater., 2000, 48, pp. 2483-2501 21. R.S. Qin, Z. Fan, Matter. Sci. Technol. 2001, 17, pp 1149-52 22. W.U. Mirihanage, D.J. Browne D.J., Comput. Mater. Sci., 2010, 50 (1), pp.260-267 23. D.J. Browne, ISIJ International, 2005, 45(1), pp.37-44 24. J.D. Hunt, S.Z. Lu, Metall. Mater. Trans. A, 1996, 27, pp.611-623 25. V. B. Biscuola, M.A. Martorano, Metall. Mater. Trans. A, 2008, 39A, pp.2885-2895

136

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry

Properties Session Chairs: Glenn Byczynski Sergio Felicelli

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

Fracture Surface Facets and Fatigue Life Potential of Castings Murat Tiryakioglu School of Engineering University of North Florida Jacksonville, FL 32224 USA e-mail: [email protected] John Campbell Department of Metallurgy and Materials University of Birmingham Edgbaston, B15 2TT, UK e-mail: [email protected] Christian Nyahumwa Department of Mechanical Engineering Dar es Salaam Institute of Technology Dar es Salaam, Tanzania e-mail: [email protected] Keywords: Facets, casting defects, Bifilms, Fatigue potential Abstract Fatigue potential has been studied in cast aluminium alloys with regard to fatigue crack initiation mechanism at casting defects, particularly surface and subsurface defects. The significance of facets are interpreted as the presence of defects in the interior of castings. Furthermore, two varieties of facets have been identified, one originating as a dendrite-straightened bifilm, and the other originating from a slip plane mechanism. Implications of the findings are discussed in terms of the fatigue potential of castings in the absence of defects. Introduction Fatigue failure in metals accounts for 90% of all in-service failures due to mechanical causes [1]. Consequently, much effort has been made to determine the mechanisms of fatigue failure, namely crack initiation and propagation before the final rupture. In castings, cracks initiate almost always from defects, such as inclusions and pores. However facets have been observed on fatigue fracture surfaces in cast alloys, including Al alloys [2,3,4,5], cast irons [5,6,7,8], steels [5], magnesium alloys [9] and Ni-base superalloys [10,11,12,13], These facets have been interpreted as "persistent slip bands" [2,3] and assumed to represent the metallurgical or ideal fatigue failure in the absence of defects. In one of these studies, Nyahumwa et al. [2] investigated the effect of casting techniques and hot isostatic pressing (HIP) on fatigue life on A356 aluminum alloy castings. For each fatigue fracture, the type of fatigue crack initiators was determined through extensive fractography. Nyahumwa et al. found that a majority of the specimens failed at fatigue cracks initiated at oxide bifilms and pores, which degraded fatigue life significantly. In the specimens with the highest fatigue lives, however, facets, originally interpreted as persistent slip bands, were reported to be involved in the process of fatigue crack initiation and Stage I crack growth. Assuming that facets

139

represented defect-free failure, Nyahumwa et al. [14] stated that the fatigue life potential of cast aluminum alloys is several orders of magnitude higher than the values usually obtained in fatigue tests, due to the presence of casting defects. However, there is considerable evidence in the literature that facets form around casting defects in fatigue testing. Therefore assumptions and interpretations about facets in fatigue failure, in particular the assumption that the appearance of facets represents an 'ultimate' metallurgical limit to the fatigue life of castings, need to be revisited. The present study is intended to review the observations about facets in fatigue failure of castings and build a case for the mechanism of their formation and their implications on fatigue life potential in castings. Fatigue Crack Initiation Mechanism for Formation of Facets The fatigue crack initiation mechanism that leads to the formation of facets on fatigue fracture surfaces of aluminium alloy castings has been a source of speculation. One of the authors [15] interpreted facets found by Jang et al. [4] in aluminum castings as oxide bifilms straightened by the advance of the dendrites during solidification. This dendrite pushing process creates large planar areas, which are separated by the planar unbonded interface between the two films, thus forming extensive transgranular, and sometimes intergranular, cracks. However, Wang et al. [16] conducted energy dispersive spectrometry (EDS) on the facets in A356 castings presented in Figure 1. They found only Al with small amounts of Si and Mg, no Fe, and a trace of O. Hence the facets in Figure 1 are planes through a primary Al dendrite. However, Wang et al. also identified oxide bifilms associated with the some facets, including the ones in Figure 1, and provided two scenarios: (i) casting defects (bifilms) cause slip concentrated in the resulting shear plane, or (ii) no significant defect present initially and the crack initiates from a shear plane. Wang et al. indicated that there is no totally convincing evidence to favor either hypothesis, and it seems quite likely from their observations that both cases occur. Evidence for two quite different facet forming mechanisms is presented in Figures 1 and 2. In Figure 1 the clean, mirror-smooth facets with sharply delineated steps and edges are expected to be typical of those generated by a slip plane mechanism. Figure 2 shows a facet that does not display mirror smoothness, and is characterised by what appear to be a distribution of pore fragments across its surface. Interestingly, close examination reveals its surface to flow seamlessly in places into the surrounding oxide folds. In particular, the oxide that surrounds the sand grain, necessarily present as a result of its entrainment via the oxidized liquid surface, is seen to connect smoothly and become contiguous with the facet surface, suggesting the facet has an origin as an oxide film. It is suggested that this variety of facet is formed by the advance of dendrites, straightening oxide bifilms. The bifilms would contain minute irregular bubbles from its own entrainment event, explaining the features observed on the face of the facet. In addition, the facet is not particularly smooth, and will clearly depend on the degree of perfection of supply of feed metal. If the supply is poor liquid will be sucked away from the bifilm into the dendrite mesh, leaving only dendrites clearly revealed. Alternatively if feeding is over-generous liquid will exude slightly between dendrite arms. Figure 3 illustrates the situation commonly observed by Nyahumwa [17] that the two varieties of facet can occur together. In fact in Figure 3 most of the observed surface consists of the Type 2 (bifilm) facets, with a small central region of Type 1 (slip plane) facet, and the remaining area in the bottom left hand comer appears to be a non-straightened area of bifilm. Kunz et al. [10,11] investigated the high cycle fatigue failure in cast Inconel Ni-based superalloys and found facets in areas adjacent to large casting defects, which is consistent with the observations

140

of Price [13] in another Ni-based superalloy. Kunz et al. attributed the formation of facets to the stress concentration effect of casting defects. They increase the local stress amplitude, promote the slip on adjacent slip planes and contribute to the decohesion process. Kunz et al. also observed that facets (i) enclose the casting defect, which was identified as the crack initiator, (ii) are in the vicinity of the crack initiator, or (iii) intersect with the casting defects, which "evokes impression that the development or growth of facets is not influenced by this porosity."

Figure 1. Facets formed in a cast A356 alloy near an oxide bifilm [16]

Figure 2. Faceted fatigue fracture in a specimen with a sand inclusion [17].

Fatigue studies by Reed [18] found facets on fracture surfaces of single crystal U720 alloy. She argues that facets are the result of a slip-band decohesion process that can operate under the conditions of cyclic stress which seems likely for this special failure mode. Boyd-Lee [19] investigated the formation of facets in a forged Ni-based superalloy and found the fatigue crack growth rate in the facet to be an order of magnitude higher than in the other parts. Boyd-Lee suggested the following mechanism for the formation of facets:

141

Figure 3 : A faceted trans-granular appearance on fatigue fracture surface indicating that persistent slip bands (PSBs) were involved in the process of crack initiation and Stage 1 crack growth in an unfiltered and HIPed Al-7Si-Mg alloy casting. Nf = 4.4 x 105 cycles at a maximum stress of 240 MPa and stress ratio R =+0.1 [17]

1. Active slip bands form to dissipate stress concentrations. 2. During load cycling, strains in active slip bands become decreasingly reversible. 3. Stress concentrations increase until slip bands form on parallel planes. This and initial slip band formation result in the observed 'intense' slip bands. 4. Eventually there occurs a zone where the dislocation density has built sufficiently resulting intense plane-normal stresses exceed the bond strength of the material. 5. Hence, a ligament in the slip band unzips, and part of the released energy contributes to surface area increase, resulting in the exposure of fresh microstructure. The suggestion that slip bands form to dissipate stress concentrations is noteworthy. For persistent slip bands to appear, stress range and crack length need to be small [20]. However there has to be a preexisting crack that causes the stress concentration. This argument implies that a casting defect needs to be present to initiate the formation of a facet but the resultant stress concentration should be low. Such a defect has to be subsurface (internal) so that stress concentrations will be low as shown in detailed finite element studies. For instance Borbely et al. [21] showed that internal failure originating from pores should be limited to the ones in which defects lie very close to the surface, as observed by Staley et al. [22] in HIPped A206 castings. Such a defect is presented in Figure 2 [17] which shows facets formed around a sand inclusion in an A356 casting. As it happens, this particular facet does not appear to originate from a slip band mechanism but from a bifilm source. However, the behavior is essentially the same as discussed further below. The observed fatigue faceted failures were perhaps originated at the interface between aluminium matrix and the subsurface oxide film defects at which stress concentration generated persistent slip bands (PSBs) during cyclic loading. The stress intensity at the interface between a subsurface defect which is considered embryo crack and aluminium matrix is normally raised by a portion of material close proximity to a free surface. During the cyclic loading, the fatigue crack possibly starts growing towards the free surface of the specimen due to high stress concentration at the vicinity of the subsurface defect edge before it propagate in other directions. The slip bands were probably involved in Stage I crack growth at the material confined between the free surface and the oxide film defects as shown in Figures 2 and 3.

142

Effect of Subsurface Crack Initiators on Fatigue Potential of Castings When attempting to understand the action of slip plane or bifilm facets on failure during fatigue, it is perhaps a surprising simplification to realize that their actions will be expected to be nearly identical. This follows because the slip plane and the straightened bifilm can both be viewed as acting like cracks. Furthermore, their size is closely similar because both are initially limited to the grain size. Subsurface crack initiation in fatigue has received considerable attention recently [23,24,25,26,27,28,29], The S-N curve for the subsurface defects is different from the one for those on the surface, as shown schematically in Figure 4. The location of the S-N curve for internal defects will depend on the size of the internal defect. [30]. Separate S-N curves for surface and subsurface defects were found [27] for a cast Al-10%Si-4%Cu-0.6%Mg alloy when specimens were subjected to a surface treatment. In several studies, cracks were observed to grow from structural defects at or shortly after the first stress cycle [31,32,33]. However in cases where these is no defect on the surface large enough to initiate a propagating crack, fatigue crack initiation from a subsurface defect takes significantly longer, resulting in increased fatigue life. Results in a fatigue experiment may come from castings, some of which have defects on the surface while the rest have subsurface defects, we can expect to have two fatigue life distributions, as shown schematically in Figure 5. The fatigue life data for A356 castings from literature [17] are presented in Figure 6. The logaverage fatigue life of castings failed from subsurface defects is 11 times that of the castings failed at surface defects at a maximum stress of 150 MPa, and 7 times at 240 MPa. Moreover, the fatigue life results of HIPed castings were observed to have even higher fatigue life than those failed from subsurface defects [22]. The fatigue life of the HIPed castings ranged from 1.7 * 107 to 7.6 χ IO7 cycles for the specimens suspended from the experiment (run-outs). The castings were assumed to be defect free. This is a proof that fatigue lives of castings observed to fail from persistent slip bands emanating from subsurface oxide film defects in the same study are several orders of magnitude lower than the fatigue life potential. Hence, one would expect that the real fatigue potential of aluminium alloy castings in the absence of defects will be above 7.6 χ IO7 cycles.

S

Nf Figure 4. The schematic illustration of S-N curves corresponding to surface and internal (subsurface) defects.

143

s

Figure 5. The schematic illustration of S-N curves showing how a sample of fatigue specimens tested at a stress level may have data with different defect locations resulting in two distinct statistical populations. The existence of faceted fracture in specimens that fail from internal defects were observed in magnesium and titanium alloys as well. Tokaji et al. [34] found in an AZ31 magnesium alloy that in subsurface fracture, facets were always present at the crack initiation site, which were located mostly close to the surface, and the facet sizes were nearly the same as, or smaller than, the average grain size. The same authors [35,36] observed subsurface fracture in beta titanium alloys and indicated the presence of smooth facets at the crack initiation site. Implications of Faceted Fatigue Failure in Castings Based on the discussion presented above, we can state the following about faceted failure in castings: 1. Type 2 (Bifilm) facets exist in the as-cast matrix, and are common defects to initiate Type 2 (slip plane) facets. 2. In the absence of surface defects, subsurface casting defects cause local stress concentrations under cyclic loading, leading to the formation of slip bands around the casting defects. 3. With increasing number of fatigue cycles, the material along the slip bands work-hardens, reducing ductility and therefore the energy needed to separate the atoms is also reduced. 4. When the minute deformations in the material around the oxide defect accumulate and separate the two halves of the bifilm sufficiently, the material ruptures along the slip band and facets and a propagating fatigue crack form. This explains why the facets will follow crystallographic planes [10]. 5. During propagation the crack follows similar slip bands around other casting defects to minimize energy, but crack growth rate is higher in facets. Consequently, the fatigue life data of Nyahumwa et al. [2] can be reinterpreted as presented in Figure 7 [37]. The presence of the two fatigue life distributions is due to the location of defects when the specimens are excised which also determines whether facets form on fracture surfaces. In either case, the failure is due to the presence of casting defects, namely bifilms. Therefore the concept of fatigue life potential of aluminum castings proposed by Nyahumwa et al. [14] needs to

144

be revised. It is the authors' opinion that fatigue life of aluminum castings will be much higher than the data collected so far, all of which seem to have come from castings with defects.

a. E

σ

150

100 10"

IO5

IO6

Nf

Figure 6. S-N curves illustrating life of fatigue specimens tested at maximum stress levels 150 and 240 MPa at stress ratio +0.1 and failed at surface and subsurface defects resulting in two distinct SN curves (Data from [17]). Conclusions • Type 2 facets (flattened bifilms) are formed during casting and solidification. • Type 1 facets (slip planes) form around casting defects which are subsurface when there are no defects on the surface large enough to generate a propagating crack. • Types 1 and 2 facets act similarly to promote fracture. • A fatigue crack initiation mechanism at the subsurface defects for the formation of facets has been proposed. • Improvement in fatigue life of castings with subsurface defects is attributed to a longer crack initiation period. The average fatigue life of castings failed from subsurface defects is 14 times that of the castings failed at surface defects at 150 MPa, and 6 times at 240 MPa. • The fatigue life potential of aluminum castings is much higher than the data in the literature.

145

3

ff

0

-2

-5 11

12

13

14

15

16

17

ln(Nf) Figure 7. Weibull probability plot [37] of the fatigue life data by Nyahumwa et al. [2], References 1. G. E. Dieter. Mechanical Metallurgy.McGraw Hill, p. 375, 1986. 2. C. Nyahumwa, N.R. Green, and J. Campbell, "Influence of Casting Technique and Hot Isostatic Pressing on the Fatigue of an Al-7Si-Mg Alloy," Metallurgical and Materials Transactions A, Vol. 32A, (2001), pp. 349 - 358. 3. Q.G. Wang, D. Apelian, and D.A. Lados, "Fatigue Behavior of A356-T6 Aluminum Cast Alloys: Effect of Casting Defects - Part 2", Journal of Light Metals, vol. 1, issue 1, pp. 85-97, 2001. 4. Y.H. Jang, S.U. Jin, Y.I. Jeong, S.S. Kim: Metall. Mater. Trans. A, 2009, vol. 40A, pp. 157987. 5. E. Bayraktar, I. M. Garcias, C. Bathias: International Journal of Fatigue 28 (2006) 1590-1602. 6. J.H. Bulloch: Theoretical and Applied Fracture Mechanics 24 (1995) 65-78 7. J. Yang, S.K. Putatunda: Materials Science and Engineering A 393 (2005) 254-268. 8. G.L. Greno, J.L. Otegui, R.E. Boeri: International Journal of Fatigue 21 (1999) 35^13. 9. M.F. Horstemeyer, N. Yang, K. Gall, D.L. McDowell, J. Fan, P.M. Gullett: Acta Materialia 52 (2004) 1327-1336.

146

10. L. Kunz, P. LukâS, R. Koneöna: "Initiation and propagation of fatigue cracks in cast IN 713LC superalloy", Engineering Fracture Mechanics 77 (2010) 2008-2015 11. L. Kunz, P. Lukas, R. Konecnâ: "High-cycle fatigue of Ni-base superalloy Inconel 713LC", International Journal of Fatigue 32 (2010) 908-913. 12. D. Gelmedin, K.-H. Lang: Procedia Engineering 2 (2010) 1343-1352. 13. C E . Price: Metallography, Volume 17, Issue 4, November 1984, Pages 359-370 14. C. Nyahumwa, N.R. Green, and J. Campbell, "The Concept of the Fatigue Potential of Cast Alloys," J. of the Mechanical Behavior of Materials, Vol. 9, No. 4, (1998), pp. 227 - 235. 15. J. Campbell: Metall. Mater. Trans. A, 18—VOLUME 41A, JANUARY 2010 16. Q.G. Wang, C.J. Davidson, J.R. Griffiths, P.N. Crepeau: "Oxide Films, Pores and Fatigue Lives of Cast Aluminum Alloys", Metall. Mater. Trans. B, v. 37B, pp. 887-895, 2006. 17. C.W.M. Nyahumwa: Influence of Oxide film filling defects on fatigue properties of cast A1-7SÌMg Alloy, PhD. Thesis, University of Birmingham, UK, 1997. 18. P. A. S. Reed: Mater. Sci. Technol., 2009, 25, (2), 258-270. 19. A.D. Boyd-Lee / International Journal of Fatigue 21 (1999) 393^105 20. C. Laird: in: "Physical Metallurgy", ed. R.W. Cahn, P. Haasen, Elsevier, Volume 3, p.2379, 1996. 21. A. Borbely, H. Mughrabi, G. Eisenmeier, H.W. Höppel: International Journal of Fracture, v. 155, pp. 227-232,2002. 22. J.T. Staley, Jr., M. Tiryakioglu, J. Campbell: Materials Science and Engineering A 465 (2007) 136-145. 23. O. Umezawa, K. Nagai and K. Ishikawa: Internal crack initiation in high cycle fatigue for Ti-5 At-2.5 Sn ELI alloy at cryogenic temperatures, Tetsu to Hagane, 75(1) (1989), 159-166. 24. H. Mughrabi: "On Multi-Stage Fatigue Life Diagrams and the Relevant Life-Controlling Mechanisms in Ultrahigh-cycle Fatigue", Fatigue and Fracture of Engineering Materials and Structures, v. 25, pp. 755-764. 25. K. Sadananda, A.K. Vasudevan, N. Phan: International Journal of Fatigue 29 (2007) 2060-2071 26. C. Przybyla, R. Prasannavenkatesan, N. Salajegheh, D. L. McDowell: International Journal of Fatigue 32 (2010) 512-525 27. Y. Nakamura, T. Sakai, H. Hirano, K.S. Ravi Chandran: International Journal of Fatigue 32 (2010)621-626 28. C. Bathias: Fatigue Fract Engng Mater Struct 22, 559-565, 1999. 29. Q. Y. WANG, J. Y. BERARD, 1 A . DUBARRE, 2 G. BAUDRY, 2 S. RATHERY1 and C. BATHIAS: Fatigue Fract Engng Mater Struct 22, 667-672, 1999. 30. O. Umezawa, K. Nagai: "Subsurface Crack Generation in High-cycle Fatigue for High Strength Alloys", ISIJ International, Vol. 37 (1 997), No. 12, pp. 1170-1 179 31. B. Skallerud, T. Iveland and G. Härkegärd: Eng. Fracture Mech., 1993, v. 44, 857-874.

147

32. S.A. Barter, L. Molent, N. Goldsmith and R. Jones: J. Eng. Failure Analysis, 2005, v. 12, pp. 99-128. 33. B.R. Crawford, C. Loader, A.R. Ward, C. Urbani, M.R. Bache, S.H. Spence, D.G. Hay, W.J. Evans, G. Clark, A.J. Stonham: Fatigue Fract. of Eng. Mater. Struc, 2005, v. 28, pp. 795-808. 34. K. Tokaji, M. Kamakura, Y. Ishiizumi, N. Hasegawa: "Fatigue behaviour and fracture mechanism of a rolled AZ31 magnesium alloy", International Journal of Fatigue 26 (2004) 1217-1224. 35. Tokaji K, Bian JC, Ogawa T, Nakajima M. The microstructure dependence of fatigue behaviour in Ti-15Mo-5Zr-3Al alloy.Mater Sei Eng 1996;A213:86-92. 36. Tokaji K, Ohya K, Kariya H. Subsurface fatigue crack initiation in beta titanium alloys. Fatigue Fract Eng Mater Struct 2000;23:759-66. 37. M. Tiryakioglu, J. Campbell: Metall. Mater. Trans. A, v. 41A, pp. 3121-3129, 2010.

148

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

EFFECT OF HOLDING TIME BEFORE SOLIDIFICATION ON DOUBLEOXIDE FILM DEFECTS AND MECHANICAL PROPERTIES OF ALUMINIUM ALLOYS M. El-Sayed1, H. Salem2, A. Kandeil1, W. D. Griffiths3 'Arab Academy for Science and Technology and Maritime Transport; Abu Qir, P.O. Box 1029; Alexandria, 21599, Egypt. 2

American University in Cairo; AUC Avenue, 5th district, P.O. Box 74; New Cairo, 11833, Egypt. university of Birmingham, Edgbaston; Birmingham, United Kingdom. B15 2TT. Keywords: Double oxide film defects, Aluminium, casting, mechanical properties Abstract

Double oxide films (bifilms) have been held responsible for the variability in mechanical properties of aluminium castings. It has been suggested that the air entrapped inside a bifilm can react with the surrounding melt leading to its consumption, which might improve the mechanical properties of the castings. In this work, the effect of the holding time of the melt before solidification on the distribution of mechanical properties, and by implication, on entrained double oxide films, was investigated for different aluminium alloys. The Weibull moduli of the plate castings were determined under tensile conditions, and their fracture surfaces examined for evidence of oxide films. The results suggested the occurrence of two competing mechanisms during the holding treatment. The consumption of air inside the bifilms due to reaction with the surrounding molten metal may lead to improvements in mechanical properties, but this may be accompanied by hydrogen passing into the bifilms, which has a deleterious effect on properties. Introduction One of the most important casting defects affecting the reproducibility of mechanical properties of aluminium castings is the double oxide film defect [1, 2], created due to surface turbulence of the liquid metal, a common feature during metal transfer and pouring in the shape casting process. When the liquid metal surface is exposed to air, a surface oxide film forms. As a result of surface disturbance, the liquid metal surface can be folded over onto itself, causing the oxidised surfaces of the folded-over metal to come together but not to fuse, trapping a layer of the local atmosphere between them, and creating a double oxide film defect or "bifilm" [1, 2] which can be entrained into the bulk metal, as shown in Figure 1. Such entrained double oxide film defects represent one of the easiest possible initiating features for cracks, since their unbonded inner surfaces can be separated with minimal effort. Also, gas dissolved in the liquid metal can precipitate inside the bifilm gap initiating porosity [3]. In addition, double oxide films are favourable sites for the nucleation and growth of intermetallic

149

compounds. These effects not only reduce the elongation, tensile strength and fatigue limit of aluminium alloy castings, but also increase their variability.

Figure 1: The formation of a double oxide film defect (1) surface turbulence leads to a breaking wave on the metal surface, and (2) the two unwetted sides of the oxide films contact each other leading to the submerging of the bifilm into the bulk liquid metal. Nyahumwa et al. [4] suggested that, due to the transformation of the oxide layer from γ-Αΐ2θ3 to α- ΑΙ2Ο3, ( a process thought to take about 5 hours), cracks are introduced into the oxide which allows the liquid aluminium to come into contact with, and react with, the air inside the oxide film defect (mainly oxygen and nitrogen). This mechanism could result in the consumption of the atmosphere inside the bifilm and possibly lead to its deactivation. The rate of consumption of the internal atmosphere has been examined by Raiszadeh and Griffiths [5], who trapped an air bubble inside liquid Al and monitored its change in volume with time using real-time x-ray radiography. Their results showed that the oxygen in the trapped air should be consumed first, to form AI2O3, then the nitrogen would react to form A1N. These reactions started immediately, (with no need for an initiating phase transformation). Also, if the initial hydrogen content of the melt was higher than the equilibrium associated with the ambient atmosphere, hydrogen diffused into the trapped air bubble, increasing its volume, which supported the idea that double oxide film defects could act as initiation sites for hydrogen porosity during the solidification of Al castings. The reaction rates of the trapped air with the melt were utilized to build a semi-empirical mathematical model capable of predicting the duration of the atmosphere inside the double oxide film defect, which suggested that the consumption of oxygen and nitrogen inside the defect would not take more than about three minutes. The aim of this work was to study the effect of the holding time of the melt before solidification on entrained double oxide films, and the corresponding change in the mechanical properties of aluminium alloy castings. Understanding these issues could lead to the development of techniques by which the effect of double oxide film defects might be reduced or eliminated in aluminium castings.

150

Experimental Procedure The experimental procedure involved the production of castings (by the investment casting technique), which contained oxide films of different ages; 0, 10 and 20 minutes. Three different aluminium alloys were considered in this work, commercial purity Al, Al-7wt.%Si-0.3wt.%Mg (2L99 alloy) and Al-5wt.%Mg aluminium alloy, so as to involve different oxide films which might have different behaviours, (AI2O3, MgAl2C>4 and MgO, respectively). In each experiment about 10 kg of the alloy was melted and held at about 800°C under a vacuum of about 80 mbar for one hour, a procedure intended to remove most, or all, previously introduced oxide films from the melt [6]. The liquid metal was then poured into preheated ceramic shell moulds, which were then placed in an induction furnace and stirred using a power setting of 7.5 kW and frequency of 2350 Hz, for one minute. This led to splashing of the liquid metal surface, and the creation and entrainment of new double oxide film defects, and their introduction into the melt. One casting was then allowed to solidify immediately, while two further castings were maintained in the liquid state by placing the filled ceramic shell mould in a furnace for 10 and 20 minutes, respectively, before removal and solidification. During holding, the hydrogen content of the melt was evaluated using a Hyscan H-measuring device. After solidification, each of the castings were machined into fifteen tensile test bars with the shape and dimensions shown in Figure 2, and tested using a Zwick 1484 tensile testing machine, with a strain rate of 1 mm min"'. Tensile results were evaluated using a Weibull statistical analysis approach to assess the influence of the holding treatment on the variability of the mechanical properties of the castings. Finally, SEM with EDX analysis was used to investigate the fracture surfaces of the test bars.

Figure 2: Sketch of the casting and the tensile test specimens taken.

151

Results Table 1 shows the results of the Weibull analysis of both of the UTS and percentage elongation values obtained from the different aluminium alloy castings, together with the position parameter and R2 values of the linear fits to the Weibull plots, as well as the results from the hydrogen measurements. Figure 3 illustrates the effect of the holding time before solidification on (a) Weibull Moduli of the UTS, (b) Weibull Moduli of the % Elongation and (c) the amount of H in solution with the liquid metal. Table 1: Results of the Weibull analysis for the test bars of different Al alloys that contained oxide films of age 0, 10 and 20 minutes.

H(cm 3 /100g)

UTS (MPa)

%

Elongation

Weibull modulus Position parameter (MPa) R2 Weibull modulus Position parameter (%E1.) R2

Commercial purity Al alloy 0 10 20 min. min. min.

Al-7Si-0.3Mg alloy

0.10

0.15

0.28

33

36

52

0 min

20 min.

0.08

10 min. 0.10

30

37

55

53

0.96

0.95

7.95

Al-5Mg alloy 10 min.

0.15

0 min. 0.91

1.0

20 min. 1.22

39

34

22

31

24

139

142

125

193

192

187

0.94

0.95

0.93

0.96

0.93

0.97

0.96

9.06

7.06

7.2

9.35

8.4

9.7

13.8

8.4

34

37

30

3.08

2.51

3.21

26

28

27

0.87

0.92

0.94

0.87

0.88

0.88

0.92

0.97

0.99

In all three alloys both the UTS and %Elongation Weibull Moduli were a maximum in the case of the casting held for 10 minutes in the liquid state before solidification, although in the case of the pure aluminium and Al-7Si-0.3Mg alloys the differences in Weibull Moduli were slight. In the case of the Al-5Mg alloy the increase in Weibull Moduli was most marked. The table also shows that the hydrogen content of the alloy consistently increased with holding time, but the Al-5Mg alloy possessed a greater hydrogen content due to the greater solubility of hydrogen in this alloy [7]. Figure 4(a) shows an SEM image inside a pore on the fracture surface of a specimen of Al-5Mg alloy. Many oxide fragments were visible inside the pore, and EDX analysis of the fragments indicated the presence of MgO, suggesting that the origin of this pore lay with a double oxide film defect. Also, Figure 4(b) shows an SEM image with the corresponding EDX analysis for the fracture surface of a specimen of Al-7Si-0.3Mg alloy, in which an iron intermetallic is associated with a spinel substrate, suggesting a nucleation relationship. This would be an indication of the role played by double oxide films in creating other defects such as porosity and intermetallics.

152

(e) Figure 3: Plot of the holding time versus (a) Weibull Modulus of UTS, (b) Weibull Modulus of % Elongation and (c) H content of the melt.

(a) (b) Figure 4: Consequences of entrainment of bifilms inside Al castings, (a) oxide-associated porosity and (b) oxide-related intermetallic formation.

153

Figure 5 shows whisker-like oxides found within pores of castings held for 20 minutes in the liquid state before solidification, (in the case of commercially-pure Al and Al-5Mg alloys). The pores were also associated with oxide films. The delicacy of these features suggests they formed during solidification, rather than earlier, say, within an oxide film defect floating within the liquid metal. Although the interconnections are too small to influence mechanical properties, their presence may be informative about conditions inside the pores during their formation.

(a) (b) Figure 5: Interconnections between adjacent oxide layers in castings containing 20-min. old oxide films, (a) commercially pure Al, (b) Al-5Mg alloy. Discussion As described in Table 1, the Weibull Moduli of the commercial purity Al alloy and A1-7SÌ0.3Mg alloy consistently exceeded 30, a value of Weibull Modulus often associated with a casting made with a well-designed running system, with reproducible properties [8]. Also, oxide films, as demonstrated by EDX analysis in the SEM, were mostly associated with pores, (as shown in Figure 4(a), rather than lying on the fracture surfaces. This could be an indication of the effect of the holding treatment on the elimination of oxide films. The Al-5Mg alloy exhibited the largest differences between the Weibull Moduli with different holding periods. This may be associated with the high solubility limit for H in this alloy, and also the permeability of its oxide film, MgO, that could lead to a more rapid reaction of O and N with the surrounding melt, as well as enhance the diffusion of H into the oxide film defect interior. These results may therefore be an indication of the significance of the H content in controlling the scatter of mechanical properties, as well as oxide film defects.

154

The consequences of the entrainment of double-oxide films were illustrated in Figure 4. They not only act as cracks in the solidified casting. Gas dissolved in the liquid metal can precipitate inside the bifilm gap initiating porosity, and in addition, double oxide films are favorable sites for the nucleation and growth of a wide variety of intermetallics. This might minimize the effect of the holding treatment on the mechanical properties of the castings, especially in the case of the commercial purity Al alloy and Al-7Si-0.3Mg alloy. The whisker-like structures within pores, shown in Figure 5, demonstrated by EDX to be an oxide, occurred mostly at holding periods of 20 minutes. The whisker-like growths are suggestive of ceramic structures grown from a vapour phase, and suggest the formation of oxide structures within porosity that contains an atmosphere. This in turn suggests that at 20 minutes holding time any double oxide film defects may have still contained an atmosphere, which may be separating their internal surfaces and preventing any bonding. The defects, therefore, would still be expected to have a deleterious effect on mechanical properties. In this work it has been found that, for all three alloys, (with their different oxides), the Weibull Moduli representing the UTS and % Elongation rose to a peak at holding times of 10 minutes, although the H content of each alloy rose progressively with holding time, as illustrated in Figure 3. The moduli decreased significantly when the holding period was increased to 20 minutes. This is suggestive of competing mechanisms affecting the distribution of mechanical properties. The first mechanism may be related to the consumption of air inside the bifilms due to reaction with the surrounding molten metal, while another mechanism may be related to the amount of hydrogen picked up by the liquid metal from the furnace atmosphere (and hence the porosity size in the resulting casting). Previous experiments had suggested that the interior atmosphere inside double oxide films could be consumed within a few minutes [5]. Hence some bonding of the two layers of the bifilm might then take place, or some reduction in the size and hence impact of the oxide films could occur, which might improve the mechanical properties of the castings. This mechanism would tend to increase Weibull Moduli with time. But on the other hand, the H content increased as the liquid metal was held in the furnace for longer periods, and this would lead to a decrease in overall mechanical properties due to increased porosity, and lead to a decrease in the Weibull Modulus with time. The Weibull Moduli for the castings allowed to solidify immediately represented the scatter of mechanical properties derived from castings that contained the lowest hydrogen contents, but that perhaps also contained oxide films then in the process of losing their initial atmospheres. The morphology of these defects seems to have been most harmful in the Al-5Mg alloy. The castings held for 10 minutes in the liquid state may therefore have possessed the best Weibull Modulus (narrowest distribution of properties) because at this time the defects may have lost all or most of their initial internal atmospheres, and may only have absorbed a little hydrogen from the surrounding melt. Under these circumstances the oxide film defects may have had a morphology that was least harmful to the properties of the castings. The mechanical properties of all castings subsequently declined with further holding up to 20 minutes, as the H content of the melt increased with holding time, increasing the deleterious effect of the double oxide film defects and perhaps leading to increased oxide-related porosity.

155

Conclusions 1. SEM examination of the fracture surfaces detected many oxide films, which demonstrated a role for such defects in influencing the failure of Al castings. 2. SEM examination also showed that the interior of porosity associated with double oxide film defects might develop whisker-shaped interconnections between them, apparent in the castings held for 20 minutes before solidification. The interconnections would not have a significant effect on mechanical properties, but could perhaps be indicative of chemical reactions resulting in deposition of ceramic whiskers, which in turn suggests an atmosphere present in the pores during solidification despite the (relatively) long holding periods. 3. The mechanical properties of the castings were most improved after a holding period for the liquid metal of 10 minutes, perhaps due to the consumption of air inside the bifilms and a reduction in their size and effect on casting properties. 4. Increasing the holding period to 20 minutes increased the H content of the alloy. This may have resulted a greater effect of porosity, which would reverse the initial enhancement in the distribution of mechanical properties (due to air consumption), leading to a decrease in the Weibull Moduli to values that were close to those of the castings that were solidified immediately after pouring. 5. The holding treatment may reduce the effect of double oxide film defects in Al melts, but would not prevent them from playing a role in the formation of other defects, such as serving as sites for the formation of hydrogen porosity or intermetallic phases. Acknowledgements The authors would like to thank Mr. Adrian Caden (of the University of Birmingham), Eng. Hanadi Hussein, Eng. Khalid Iraqi and Eng. Ramy Wasfi (of the American University in Cairo) for their technical support during the laboratory work. Also, the authors wish to acknowledge the sponsorship of this study by the Arab Academy for Science and Technology, Alexandria, Egypt. References 1. Campbell, J., Castings. 2nd. ed. 2003: Butterworth-Heinemann. 2. Campbell, J., Entrainment defects. Mat. Sci. Technol., 2006. 22: p. 127-145. 3. Griffiths, W. D. and Raiszadeh, R., Hydrogen, porosity and oxide film defects in liquid Al, J. Mat. Sci., 2009, 44, 3402-3407. 4. Nyahumwa, C , Green, N. R, and Campbell J., Influence of casting technique and hot isoslatic pressingon the fatigue ofanAl-7Si-Mgalloy, Met. and Mat. Trans. A, 2001, 32A: p. 349-358. 5. Raiszadeh, R. and Griffiths, W.D., A Semi-empirical Mathematical Model to Estimate the Duration of the Atmosphere within a Double Oxide Film Defect in Pure Aluminum Alloy. Met. and Mat. Trans. B, 2008, 39(2): p. 298-303. 6. Raiszadeh, R. and Griffiths, W.D., The behaviour of double oxide film defects in liquid Al alloys under atmospheric and reduced pressures. J. Alloys and Compounds, 2010. 491(1-2): p. 575-580. 7. Anyalebechi, P.N., Analysis of the effects of alloying elements on hydrogen solubility in liquid aluminum alloys. Scripta Met. et Mat., 1995. 33(8): p. 1209-1216. 8. Nyahumwa, C , Green N. R., and Campbell J., Effect of Mold-Filling Turbulence on Fatigue Properties of Cast Aluminum Alloys. AFS Trans., 1998. 58: p. 215-223.

156

Shape Casting: The 4lh International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

WEIBULL ANALYSIS OF THIN A356 PLATES CAST WITH THE ELECTROMAGNETIC PUMP GREEN SAND PROCESS Ratessiea Lett1, Sergio Felicelli1, Rafael Cuesta2, John Berry1, J. Antonio Maroto2, Ruth San José2 'Department of Mechanical Engineering and Center for Advanced Vehicular Systems, Mississippi State University, Mississippi State, MS, 39762, USA foundation for the Research and Development in Transport and Energy (CIDAUT), Parque Tecnologico de Boecillo,Valladolid, SPAIN, 47151 Keywords: Aluminum A356, Weibull, four point bend, electromagnetic pump Abstract Four different gating systems were used to produce plates by means of an electromagnetic pump linked up with a green sand molding machine, promoting a counter-gravity of the mould. Three of the systems were multiple gated, whereas one was single gated. Eight cast plates were examined, two from each gating design. The method of four point bend testing was used to obtain information about the mechanical properties of the castings, as this method produces a uniform distribution of the bending stress within the central span of the test specimens. From these results, a Weibull statistical analysis was performed in order to quantify specimen failure rate for each of the configurations. The specimen cross sections were then examined using optical microscopy. This project is funded by the National Science Foundation (NSF) under the International Research and Education in Engineering (IREE) program. Introduction During conventional casting processes, liquid metal is often subjected to agitation and movement at high velocities, producing a final melt with a considerable amount of inclusions and defects which affect mechanical properties of the material, thus resulting in premature failure[l-3]. As aluminum alloys are highly susceptible to oxidation during casting, there exists the opportunity for the formation of a thin film-like structure to form on the surface of the metal. These structures, referred to as oxide films, may result in an unsuccessful joining or "weld" of opposing streams of metal in the mold cavity, producing a region known as a confluence weld when the streams meet[l]. In previous works [4-5] the production of confluence welds amongst multipleingate configurations of varying geometries and their effect on the casting structure was studied. The focus of this work is the determination of mechanical properties for various bottom-filled gated systems (referred to as Al, A2, B, and BO). Experimental Procedures Two A356 10 mm-thick casting plates for each configuration were produced by the Electromagnetic Pump Green Sand (EPGS) process in the foundry Aleaciones Ligeras Aplicadas S.L. (Applied Light Alloys), in Valladolid, Spain. The EPGS process, which was developed by CIDAUT, employs the DISAMATIC to produce high speed vertically parted green sand molds, linked up with an electromagnetic pump for controlled dosing of the molten metal inside the

157

mold. A more quiescent filling of the casting is produced in this manner, thus minimizing the introduction of oxides into the melt and preventing defect formation during solidification in comparison with other casting processes [6]. The inlet is located on the side of the mold near the bottom. Cuesta, Martin, and Maroto [7] provide a thorough explanation of the development of the process and corresponding parameters. From the cast plates, information about mechanical properties was gathered to gain insight on the quality of the castings in relation to the different gating geometries (see Figure 1).

(a) (b) (c) (d) Figure I. The four gating configurations usedfor this research: (a) Configuration Al; (b) Configuration A2; (c) Configuration B; (d) Configuration BO. Configurations Al and A2 both have tapered bottom ingates while configuration B has tapered side ingates. For the BO configuration, the same geometry is used as that of the B configuration, however, only one of the ingates is utilized, whereas in the B design metal flow is permitted through both side ingates. These are the major geometrical differences amongst the molds. Each consists of a runner connected to side ducts which serve as filling channels for the Bconfiguration. At the top of all plate cavities feeders have been incorporated, in order to assist the feeding of the cavity during solidification. Four Point Bend Tests In order to characterize the flexural properties of simply supported beams, it is common practice to conduct three- or four-point bend tests. For this research, the method of four point bend testing was chosen over that of the three point, as the maximum load is evenly distributed between the top forces (referred to as the center span hereafter), whereas in the three-point test the maximum load is concentrated at the point of application of the top force [8-9]. With the even distribution of the load, a more accurate analysis of the cause of failure can be performed on the specimen, as fracture will generally not occur at the same location for each test. For this study, a pouring temperature of 730C was used. Sixteen bend test specimens were cut from each plate, each having dimensions of about 90 mm long, 20 mm wide, and 10 mm thick. The bend tests were conducted on an Instron 5869 testing apparatus with a default loading rate of .001in/s (,0254mm/s). Figure 2 shows the bend test specimen locations for all plate configurations. All specimens were tested in the as-cast condition.

158

(a) (b) Figure 2. Locations of the bend test specimens for each of the plates (not drawn to scale): (a) Entire plate with overall dimensions and reference to outer region excluded for edge effects; (b) Specimen area with dimensions and location of the 16 test specimens. For all samples, two ASTM Standards for testing of flexural properties of non-metallic materials were used (see Figure 3) [8-9], as there are currently no standards for the testing of metallic specimens.

Figure 3. Bend test configurations as described in the ASTM Standards [8-9] with central span length divided into sections of length L/3, where the load is represented as F. Weibull Analysis A Weibull statistical analysis was used in order to show the distribution of mechanical properties about mean values in relation to failure for each of the gating configurations. One form of the two parameter Weibull distribution is shown in the following Equation [10]: P = l-exp

f

σ σ„

(1)

where P is the probability of failure at the variable being measured (ultimate bending strength σ ), m is the Weibull modulus, and σ o is the scale parameter, which is a characteristic stress value at which approximately 63.2 percent of the specimens failed [10-11]. After taking the natural logarithm of both sides of Equation 1 twice, the linear regression equation is given as [11-12]:

159

F = l n lIn n

(

1

il= mlna-m\na -'JJ

1-

0

(2)

The Weibull modulus, m, is the slope of the regression line between Y and the natural logarithm of the ultimate bending strength (UBS), namely X for simplicity, and -mlnfa o) is the Y-intercept in the X-Y plane. The probability of failure, Pf„, for each value of UBS ranked in ascending order, can be evaluated using various methods [11-12]. The method chosen for this work is given by the equation:

n Where /' is the ranking of the result when UBS values are arranged in increasing order, and n is the total number of results. Results and Discussions As expected, test samples fractured within the central span. Most of the specimens tended to do so in the center of the span, confirming the soundness of the procedure.

Figure 4. Bend test of one of the specimens taken from an Al plate prior to complete fracture in the Instron 5869 load cell. Microscopy After bend tests were conducted, the cross sections of the samples were examined using Optical Microscopy. Eight samples were submitted to observation. Following the code [plate type_plate number-specimen number], these samples were: A l l - 3 , A12-12, A21-16, A 2 2 - 7 , B l - 5 , B2-10, B01-14, and B 0 2 - 1 . Samples were chosen from these locations in order to have a comparison of the microstructures throughout the specimen area from the regions closest to the feeder and to the bottom of the plate.

160

The surfaces of the samples tended to have similar dendritic features and apparent areas of porosity, however, preliminary inspection of the fine pore sizes would generally indicate a relatively quiescent filling compared with typical gravity systems [1]. Figure 5 illustrates characteristic microstructural images of the samples analyzed. Additional analysis will be performed on the samples in order to characterize mechanical properties in relation to porosity grain size and microstructure.

(c) (Φ Figure 5. Characteristic optical microscopy images ofspecimens: (a) A1J2-12; (bj A21-16; (c) B2-10; (d) B02-1 showing a largely free of porosity dendritic structure. Weibull Distributions Table I summarizes the results obtained from the four point bend tests. As is shown, values ranging from approximately 226MPa to 327MPa were obtained. The high magnitude of these figures for the as-cast A356 material used reinforces the idea that, in general terms, no major entrainment events (macroscopic bubbles) took place during the filling of the plates. Table I. Characteristic UBS Values (MPa) for each of the Gating Systems System Minimum Maximum Average Al 264.42 252.89 226.49 A2 253.44 269.04 285.81 296.69 B 259.14 327.43 323.84 286.22 BO 232.78

161

The results of the Weibull analysis (see Table II and Figure 6) allows us to make comparisons between the different gating configurations. As it can be seen, the Weibull moduli of the bottom gated plates (Al and A2) are widely higher, (and thus the span of the corresponding UBS values is widely smaller) than those filled laterally. This generally agrees with the specification given by Campbell of limiting the horizontal flow to a minimum extent in order to avoid the incorporation of the oxide film into the bulk melt [13]. Congruently, some indications of turbulent fold-in of the oxide upon the joining of the two metal streams inside the B plate (featuring relatively long horizontal flow paths) have been assessed by computer simulation [5]. In relation to this, it is also interesting to note that the maximum value of Weibull modulus was found for the plates for which the horizontal flow at the early stage of their filling is theoretically of least importance (A2) even when the fracture area of the test specimens lies most precisely in the region located in between the ingates, and therefore it is most likely to contain oxide films at the end of the casting process.

Configuration Al A2 B BO

Table II. Weibull Parameters for All Plate Configurations Weibull Distribution Weibull Modulus Scale Parameter Equation y = 35.46x- 196.72 35.46 256.83 y = 41.04x- 230.17 41.04 272.68 304.85 y = 19.56X- 111.85 19.56 y =19.25x-109.41 294.22 19.25

3

2 1 0 ? -1 Έ £

-J

-2

-3

-4

5.2

5,3

5,4

5,5

5,6

5,7

5,8

Ln (UBS)

Figure 6. Weibull plot of UBS distribution for all gating configurations.

162

5,9

6

Conclusions • • •

A series of A356 plates were produced by means of the EPGS process using several bottom and side filling systems prone to produce oxide entrainment into the bulk melt. The Weibull analysis of the ultimate bending strength of multiple test specimens preliminary indicates that oxide entrainment is basically related with the extent of horizontal flow at an early stage of the filling of the plates. Future work is planned in order to assess the effect of flow velocities, porosity, microstructure and grain size over mechanical properties. Acknowledgements

This work was funded by the National Science Foundation through Grant Number CTS-0553570 and supplemental funding by the International Research and Education in Engineering (IREE) program. The authors are thankful to Aleaciones Ligeras Aplicadas S.L. of Valladolid (Spain) for providing the castings used for this research. The assistance of Prof. E. William Jones with the Weibull analysis and Jacob Coleman and Hunter Cole with Optical Microscopy and porosity analysis is gratefully acknowledged. References 1

1. John Campbell, Castings 2" Edition - The New Metallurgy of Cast Metals fButterworthHeinemann, Oxford, UK, 2003). 2. M. Barkhudarov and C.W., Hirt, Paper presented at the Proc. Materials Solutions Conference on Aluminum Casting Technology, Illinois, October 1998. 3. J.T. Berry and R. Luck, "Porosity criteria functions revisited", Paper presented at the 2006 World Foundry Congress, Harrogate, United Kingdom, 4-7 June 2006. 4. R.L. Lett, S.D. Felicelli, R. Cuesta, J.T. Berry, A. Rivas, and M.E. Alcalde , "Confluence welds in aluminum castings - Part Two", AFS Transactions, 118 (2010), 29-38 5. R.L. Lett, S.D. Felicelli, R. Cuesta, J.T. Berry, and D. Losua, "Confluence welds in aluminum castings", AFS Transactions, 117 (2009), 131-138. 6. C. Nyahumwa, N.R. Green, and J. Campbell, "Effect of mold-filling turbulence on fatigue properties of cast aluminum alloys", AFS Transactions, 106 (1998), 215-223. 7. R. Cuesta, J. Martin, and J.A. Maroto, "New casting process: the EPGS process: the science and technology of casting production", Foundry Trade Journal, (March 2008), 66-70. 8. "Standard Test Method for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials by Four-Point Bending," ASTM International, 08.03, D627202(2008), 513-518. 9. "Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature." ASTM International, 15.01, Cl 161-02c (2008), 217-226. 10. J. Mi, R.A. Harding, and J. Campbell, "Effects of the Entrained Surface Film on the Reliability of Castings," Metallurgical and Materials Transactions A, 35 (2004), 2893-2902. ll.W.D. Griffiths and N.W. Lai, "Double Oxide Film Defects in Cast Magnesium Alloy," Metallurgical and materials transactions A, 38 (2007), 190-196. 12. D. Wu, Y. Li, J. Zhang, L. Chang, D. Wu, Z. Fang, and Y. Shi, "Effects of the Number of Testing Specimens on the Weibull Parameters of Solid Catalysts", Chemical Engineering Science, 56 (2001), 7035-7044.

163

13. Campbell J. "The 10 Casting Rules Guidelines for the Reliable Production of Reliable Castings", First International Conference on Gating, Filling and Feeding of Aluminum Castings, October 1999.

164

Shape Casting: The 4"1 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

GUIDELINES FOR 2-PARAMETER WEIBULL ANALYSIS FOR CASTINGS Murat Tiryakioglu School of Engineering University of North Florida Jacksonville, FL 32224 USA e-mail: m.tirvakiogluiS?.unf.edu David Hudak Department of Mathematics Robert Morris University Moon Township, PA 15108 USA e-mail: hiidakiS!rmu.edu Keywords: Weibull modulus, confidence interval, hypothesis testing Abstract 2-parameter Weibull statistics used in the analysis of mechanical data from castings are reviewed. Guidelines to estimate Weibull parameters by the linear regression technique are provided. Moreover goodness-of-fit tests for Weibull fits and calculating confidence intervals for the estimated Weibull modulus are discussed. A new hypothesis test for comparing two estimated Weibull moduli is introduced. The use of these guidelines is demonstrated by using data from the literature. Introduction Wallodi Weibull [1,2] introduced an empirical distribution based on the "weakest link", using the theory developed by Pierce [3], which has since been widely applied to the fracture-related mechanical properties of ceramics and metals. The cumulative probability function of the Weibull distribution is expressed as: — (1) σ0) where P is the probability of failure at a given stress (strain, fatigue life, etc.), σ, or lower. The term, σο, is the scale parameter, and m is the shape parameter, alternatively referred to as the Weibull modulus. Green and Campbell [4] showed that the tensile strength (ST) of A356 castings alloys follows a 2-parameter Weibull distribution and that the filling system design has a strong effect on the Weibull modulus. Hence the reliability of castings could be measured with the Weibull modulus. Since the results of Green and Campbell, the 2-parameter Weibull distribution has been used extensively in the casting literature to characterize fracture-related mechanical properties such as tensile strength (ST) [5], elongation and fatigue life (Nf) [6,7,8], According to Campbell [9], m is often between 1 and 10 for pressure die castings, and between 10 and 30 for many gravity-filled castings. For good quality aerospace castings, m is expected to be between 50 and 100. Hence Weibull modulus has been used as a measure of the casting quality.

165

Recently, Tiryakioglu and Campbell [10] provided guidelines for interpreting Weibull probability plots including the 3-parameter Weibull distribution and Weibull mixtures. They stated that the 2-parameter Weibull distribution is applicable only when castings have defects with large sizes that come from the same, single distribution, i.e., one source of damage during melt preparation and mold filling. The present study is intended to provide additional guidance when the 2-parameter analysis is warranted and provides a step-by-step procedure for Weibull analysis with the linear regression technique. Procedure for 2-Parameter Weibull Analysis Step 1. Assign probability to each data point: There are a number of probability estimators (alternatively referred to as plotting position formulae) in the literature. These formulae can all be written in the form (2)

n+b

where ; is the rank in ascending order, n is the sample size and a and b are empirical constants. Montecarlo simulation studies [11,12,13,14] showed that all probability estimators produce biased estimates of the Weibull modulus, i.e., the average of estimated Weibull moduli is different from the true Weibull modulus used in these studies. The magnitude of the bias depends on the values of a and b as well as the sample size. The authors [15] determined the values of a and b that yield unbiased estimates of σο and m, which are listed in Table 1 for sample sizes (n) between 5 and 100. It is strongly recommended that the values of a and b listed in Table 1 be used in lieu of other plotting position formulas in the literature. Table 1. The values of a and b for Equation 2 that yield unbiased probability estimators for the two Weibull parameters. n a b n a b 5 6 7 8 9 10 11 12 13 14 15 17 20 22

0.173

0.5

0.243

0.39

25

0.28

0.31

0.309

0.251

0.322

0.21

0.348

0.19

0.367

0.16

0.371

0.13

0.382

0.11

0.388

0.1

0.394

0.08

0.407

0.05

0.417

0.03

0.43

0

27 30 32 35 40 45 50 55 60 65 70 75 80 90

0.443

0

100

0.448

0.518

0 0 0 0 0 0 0 0 0 0 0 0 0 0

0.519

0

0.455 0.46 0.465 0.472 0.481 0.486 0.499 0.503 0.509 0.518 0.522 0.516

166

Step 2. Obtain linear regression fit to 1η(σ) versus ln(-lnd-P)) One of the most commonly used methods of presenting the Weibull fits to data is the Weibull probability plot. After rearranging, Equation 1 can be written as 1η[-1η(ΐ-Ρ)] = »ι1η(σ)-»ι1η(σ 0 )

(3)

Note that Equation 3 has a linear form when the left-hand side of the Equation is plotted versus 1η(σ) with a slope of m and an intercept of-m 1η(σο). Hence the best fit line to this probability plot represents the linear regression method for estimating the two Weibull parameters. Step 3. Conduct Goodness-of-Fit test: Usually the trend of the data on the Weibull probability plot is used to judge whether the data indeed come from a Weibull distribution. The use of a probability plot, however, is subjective and insufficient and therefore it is strongly recommended that probability plots be always augmented by other goodness-of-fit tests [16]. The coefficient of determination, R2, has been commonly used as a measure of the goodness-offit, especially in the metal casting literature. The higher the value of R2, the higher the confidence that data follow the distribution being tested. However clear guidelines for using R2 as a goodness-of-fit test have been recently developed by Tiryakioglu et al. [17] who ran Monte Carlo simulations to determine the critical points of R2 at a=0.05 (R2o 05) and reported that the following formula can be used for. sample sizes between 5 and 100: R6.05 = 1 . 0 6 3 7 — ^ 3 -

n°

(4)

The hypothesis that the dataset follows the tested distribution is rejected when p-value for calculated R2 is less than a specified value for Type I error (a), which is typically prescribed as 0.05. If the calculated R2 is higher than R2oos, then it can be concluded that the data indeed come from a 2-parameter Weibull distribution. The authors recommend that Equation 4 be used to evaluate goodness of fit to the Weibull distribution with R2 instead of other general guidelines provided in literature. It should be noted that the 2-parameter Weibull analysis is not valid if R2 < R2o.05, in which case steps 5 and 6 should not be taken. Step 4. Calculate Confidence Interval for the estimated Weibull Parameters It is important to realize that estimated Weibull parameters have their own statistical distributions. Consequently, it is necessary to calculate confidence intervals for the two Weibull parameters, especially m. The distribution of the estimated Weibull modulus was found [14] not to follow any formal distribution. Therefore the use of percentage points is necessary to calculate the confidence intervals. Percentage points for the distribution of estimated m when unbiased probability estimators in Table 1 are used are presented in Table 2 for sample sizes between 5 and 100. For instance if the sample size is 30, then two-tailed 95% confidence intervals are calculated by first finding the percentage points of 0.025 and 0.975 for n=30 (95% will be

167

included between percentage points of 0.025 and 0.975). In Table 2, these points are 0.712 and 1.499, respectively. Finally, the 95% confidence limits are found by multiplying the estimated Weibull modulus by 0.712 and 1.499. Table 2. Percentage points of the distribution of estimated m (after normalization) obtained by using the unbiased probability estimators in Table 1. n

0.005

0.01

0.025

0.05

0.1

0.9

0.95

0.975

0.99

0.995

S

0.274

0.316

0.407

0.497

0.613

1.984

2.304

2.625

3.077

3.425

6

0.328

0.371

0.457

0.540

0.643

1.821

2.105

2.392

2.793

3.125

7

0.363

0.406

0.492

0.572

0.666

1.727

1.996

2.247

2.558

2.809

8

0.388

0.436

0.513

0.587

0.676

1.661

1.887

2.128

2.445

2.674

9

0.422

0.470

0.545

0.606

0.693

1.618

1.835

2.049

2.342

2.551

10

0.435

0.488

0.558

0.623

0.705

1.567

1.776

1.984

2.237

2.475

11

0.456

0.499

0.573

0.635

0.715

1.534

1.718

1.890

2.151

2.331

12

0.490

0.529

0.592

0.648

0.725

1.511

1.686

1.855

2.066

2.294

13

0.505

0.544

0.608

0.662

0.734

1.477

1.639

1.818

2.049

2.222

14

0.518

0.558

0.613

0.670

0.741

1.456

1.618

1.786

1.988

2.179

15

0.526

0.565

0.624

0.683

0.750

1.439

1.600

1.761

1.988

2.165

17

0.538

0.576

0.635

0.690

0.755

1.412

1.555

1.684

1.894

2.037

20

0.568

0.608

0.661

0.708

0.770

1.374

1.506

1.637

1.795

1.912

22

0.583

0.621

0.676

0.723

0.780

1.340

1.464

1.575

1.739

1.866

25

0.609

0.638

0.691

0.734

0.790

1.326

1.435

1.543

1.675

1.770

27

0.620

0.650

0.699

0.746

0.797

1.312

1.425

1.534

1.664

1.789

30

0.639

0.670

0.712

0.753

0.806

1.294

1.397

1.499

1.637

1.736

32

0.651

0.676

0.723

0.763

0.810

1.280

1.372

1.462

1.580

1.689

35

0.657

0.688

0.730

0.769

0.816

1.266

1.355

1.449

1.550

1.634

40

0.673

0.702

0.745

0.784

0.829

1.253

1.335

1.412

1.511

1.582

45

0.694

0.720

0.758

0.794

0.836

1.230

1.309

1.383

1.493

1.580

50

0.704

0.728

0.767

0.800

0.840

1.220

1.294

1.362

1.462

1.529

55

0.711

0.738

0.775

0.807

0.845

1.205

1.274

1.337

1.429

1.493

60

0.725

0.751

0.783

0.814

0.852

1.200

1.266

1.335

1.410

1.479

65

0.735

0.757

0.790

0.821

0.856

1.189

1.253

1.312

1.389

1.447

70

0.742

0.762

0.794

0.826

0.863

1.183

1.244

1.300

1.368

1.422

75

0.742

0.767

0.799

0.829

0.864

1.176

1.235

1.289

1.361

1.420

80

0.754

0.776

0.808

0.835

0.870

1.170

1.224

1.282

1.351

1.403

90

0.767

0.786

0.818

0.845

0.876

1.160

1.214

1.266

1.325

1.383

100

0.777

0.800

0.826

0.853

0.883

1.149

1.199

1.242

1.300

1.348

In contrast to the distribution of the modulus m, the distribution of estimated scale parameter is normal [15]. There is no need therefore to use percentage points tables. The standard deviation of the estimated scale parameter (after normalization), Soo, calculated by using the probability estimators listed in Table 1 is found by

168

\

=_

0.359

_

r~

(5)

Therefore confidence intervals for the true scale parameter ( σ0,ιηιί ) can be found by 0 359

1.000 + Z, _, ^ 4 ^ · ^ n

σ 0|<me

V«

Taking o=0.05 (95% confidence), Z

'Ά

Q 35Q

οΐ,™, ^ 1.000 + Z/ 2a / ττττκ ^X- (6) V"

, and Z

7

Δ

are 1.96 and -1.96, respectively.

Step 5. Comparing two Weibull Moduli Because Weibull modulus has been used as a measure of reliability of castings, a formal procedure is necessary to compare the Weibull moduli from two sets of castings. Such a hypothesis test, to the authors' knowledge, is not available in the literature. The authors [18] have recently provided the results of their Monte Carlo simulations for comparing two Weibull moduli for sample sizes between 10 and 100. The results are provided in Table 3, which shows the 2.5 and 97.5 percentile of the distribution of mi/ni2 where ni>n2. The values in Table 3 can be used to test the hypothesis that the two Weibull moduli are equal at a=0.05. Table 3. The 2.5 and 97.5 percentiles for the distribution of the ratio of two estimated Weibull moduli.

Application to Datasets Three datasets were considered from the study by Green and Campbell [19] who showed that the tensile strength (ST) of cast Al-7%Si-Mg alloys is affected to a great extent during the mold filling stage. If the mold is filled quiescently, tensile strength is not only higher but also has less variability, i.e., higher reliability. Conversely, tensile strength has a lower average and higher variability when the mold filling occurs turbulently. The datasets represent these two types of mold filling: top-filled (TF) which is quite turbulent, and bottom-filled (BF), which is quiescent.

169

Also included in the analysis are the castings which were modified by Sr followed by quiescent filling (BFmod). The sample size is 45, 36 and 37 for TF, BF and BFmod, respectively. For the plotting position formula (Equation 2), a has a value of 0.481, 0.466 and 0.468 for TF, BF and BFmod, respectively, and b=0. The Weibull probability plot for the three datasets is presented in Figure 1 and the estimated parameters as well as goodness of fit measures are given in Table 4. For TF, R2 < R2o 05, indicating that the 2-parameter Weibull fit has to be rejected. Note in Figure 1 that the slope for the lowest five points seems to be less than that for the rest of the data. This change in slope is indicative of the presence of multiple defect distributions and therefore the data may come from a mixture of two Weibull distributions [10,20], For BF and BFmod, R2 values are in excess of R2oo5, indicating that the Weibull fit is acceptable. The 95% confidence limits for the two parameters were established for BF and BFmod castings, as presented in Table 4. Note that the values to establish the confidence limits for m for n=36 and 37 are not listed in Table 2 and therefore they were found by interpolation.

ln(ST(MPa)) Figure 1. Weibull probability plot for the tensile strength of Al-7%Si-Mg alloy castings filled turbulently (TF) and quiescently (BF), as reported by Green and Campbell [4].

170

Table 4. Statistical results on the Weibull fits to the three datasets.

If one wishes to determine whether the reliability of BF and BFmod castings is different, i.e., whether Sr modification has an effect on reliability, the estimated Weibull moduli of the datasets can be compared. The ratio of the estimated Weibull moduli is 1.321 (50.71/38.40). Note that the Weibull modulus of BFmod is written in the numerator because the sample size of that dataset is larger than that of BF. This ratio of 1.321 falls within the values listed in Table 3 for sample sizes of 30 and 40. Hence we do not have enough evidence at this confidence level to support the hypothesis (at a=0.05) that Sr modification increases reliability. Conclusions •

A step-by-step procedure was introduced for the 2-parameter Weibull analysis of casting data by the linear regression method, in cases where such analysis is warranted.

•

Goodness-of-fit tests should be conducted on the Weibull fits to determine whether the data do come from a 2-parameter Weibull distribution.

•

If the distribution is indeed Weibull, confidence limits on the two parameters should be established.

•

A new hypothesis test for the comparison of estimated Weibull moduli has been introduced.

References 1. Weibull, W. (1939). A statistical theory of the strength of materials. Proc. The RoyalSwedish Institute for Engineering Research. Nr. 151. 2. W. Weibull, The phenomenon of rupture in solids, Royal Swedish Institute of Engineering Research (Ingenioersvetenskaps Akad. Handl.), Stockholm, Vol. 153, 1-55, 1939. 3. F. T. Pierce: J. Textile Inst. 1926 vol. 17, pp. T355-T368. 4. N. R. Green, J. Campbell: Mater. Sei. Eng. A, 1993, vol. A137, pp. 261-266. 5. G. E. Byczynski, J. Campbell: In Advances in Aluminum Casting Technology II. Edited by M. Tiryakioglu and J. Campbell; ASM International, 2002, pp. 65-74.

171

6. C. Nyahuniwa, N. R. Green, J. Campbell: Metall. Mater. Trans. A, 2001, vol. 32, pp. 349358. 7. D. Casellas, R. Pérez, J. M. Prado: Mater. Sei. Eng. A, 2005, vol. A398, pp. 171-179. 8. Q. G. Wang, D. Apelian, D. A. Lados: J. Light Metals, 2001, vol. 1, pp. 73-84. 9. J. Campbell: "Castings", 2nd Edition, p. 303, London, Elsevier, 2003. 10. M. Tiryakioglu, J. Campbell: Metall. Mater. Trans A., in press. H.A. Khalili, K. Kromp, J. Mater. Sci. 26 (1991) 6741. 12. R. Langlois, J. Mater. Sci. Lett., 10 (1991) 1049. 13. K. Trustrum, A. de S. Jayatilaka, J. Mater. Sci. 14 (1979) 1080. 14. M. Tiryakioglu, D. Hudak: J. Mater. Sci. (2007)42:10173-10179 15. M. Tiryakioglu, D. Hudak: J. Mater. Sci. (2008)43:1914-1919 16. S.S. Shapiro, C.W. Brain, in "Statistical Distributions in Scientific Work", v.5, eds. C. Taillie, G.P. Patii, B.A. Baldessari, D. Reidel Publishing, 1981, p.l. 17. M. Tiryakioglu, D. Hudak, G. Ökten: Mater. Sei. Eng. A, 2009, vol. 527, pp. 397-399. 18. D. Hudak, M. Tiryakioglu, submitted to Mater. Sei. Eng. A. 19. N.R. Green, J. Campbell: AFS Trans. 102 (1994) 341. 20. C.A. Johnson: J. Frac. Mech. Ceramics 5 (1983) 365.

172

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals <£ Materials Society), 2011

MELT CLEANLINESS, HYDROGEN CONTENT AND TENSILE PROPERTIES OF A356 Derya Dispinar, Arne Nordmark, Freddy Syvertsen SINTEF, Materials and Chemistry, 7465 Trondheim, Norway Keywords: bifilms, hydrogen, porosity, degassing Abstract Degassing of aluminium melts is one of the most important stages in a casting operation. Main reason for this treatment is the removal of hydrogen from the melt, and thereby the ultimate goal is the achievement of a pore-free casting. However, it was shown that the interaction between the bifilms and the hydrogen plays a significant role on pore formation. Therefore, series of degassing experiments were carried out with the commercially available alloy A356. Two melts were prepared and one melt was upgassed gradually and the other was degassed gradually. Tensile samples were collected and bifilm index measurements were compared at each treatment sequences. Weibull analysis was used and it was found that the turbulence and vortex (increase in bifilm index) during rotary degassing caused an increase in the scatter irrespective of the hydrogen content. Introduction The decrease in the solubility of hydrogen in liquid aluminium with temperature (Fig 1) has been believed to be a major source of porosity. This decreased solubility of hydrogen in the solid phase can result in the precipitation of hydrogen gas, which may cause porosity. Campbell [1-3] had shown that homogeneous nucleation of pores was difficult and a nonnucleation process was required for pore formation whether gas or shrinkage. He also suggested that if there were no more liquid left to feed the shrinkage; then the solid would contract to compensate. Thus, the non-nucleation process for porosity in aluminium and its alloys would simply be aided by bifilms [4-8].

Figure 1: Solubility of hydrogen in aluminium [9] The conventional and the typical application that is called "degassing" is carried out in order to minimise the pore formation due to the solubility problem of hydrogen. This takes place by bubbling an inert gas through a melt of aluminium to reduce the amount of undesirable hydrogen. Hydrogen is removed from the melt by its diffusion into the degassing bubbles as a result of the difference in partial pressure between the melt and the rising bubbles. For this 173

purpose, different types of degassers are developed [10]. The most common one is called the "rotary degasser" where a rotating rod is place inside the melt with a diffusor at the bottom that distributes fine bubbles through the melt. This method is known to be the most efficient type of degassers [10]. Besides its advantages, the rotary degassers actually suffer from severe surface turbulence which may lead to entrained surface oxide films (i.e. bifilms) into the liquid metal. Therefore in this study, two casting experiments were carried out with commercial A356 alloy with different degassing sequences in order to check the metal quality correlation with mechanical properties. Experimental Work 75 kg of A356 alloy was melted at 750°C in a resistance furnace. The composition of the commercially available A356 is given in Table 1. Table 1. Chemical composition of A356 used in the experiments Si 6.9

Mg 0.32

Mn 0.002

Fe 0.116

Ti 0.11

Na 0.0012

Sr 0.0005

P 0.0002

Al rem.

Hydrogen level of the melt was measured by A1SPEK and a rotary degasser (Foseco XSR Rotor 0135, at 500 rpm) was used to control the melt hydrogen level. Three hydrogen levels were selected: 0.1, 0.2 and 0.4 mL/lOOg Al. Argon was used for degassing, Ar+10%H2 mixture was used to reach 0.2 level and Ar+water vapour mixture was used to obtain 0.4 levels of hydrogen in the melt. In the first series of testing, the melt was degassed and then upgassed to three different levels. In the second series, the melt was first upgassed and then degassed gradually. Ten reduced pressure test samples were collected for bifilm index [6] measurement at each hydrogen level and 20 cylindrical bars were cast for tensile testing. Results The bifilm index changes of the melts that are treated to achieve three different hydrogen levels are shown in Fig 2. It is important to note that the x-axis is selected to show the treatment sequence. The corresponding values are as follows: for the upgassed melt, bifilm index values are 10, 15 and 95 mm for 0.1, 0.2 and 0.4 mL/lOOg Al hydrogen levels respectively. For the degassed melt; the increase in the bifilm index is 75, 90 and 120 mm for 0.4, 0.2 and 0.1 mL/lOOg Al hydrogen levels.

174

(a) (b) Figure 2: Bifilm index of the melts at different hydrogen levels (a) degassed and then upgassed melt; (b) upgassed and then degassed melt (note: x-axis shows the treatment sequence) The tensile test results were chosen to be presented as Weibull distributions for same hydrogen levels. The comparison of the ultimate tensile properties and elongation at fracture is given in Fig 3 and Fig 4 respectively.

Figure 3: Weibull distribution of ultimate tensile strength for different hydrogen levels (a) 0.1 mL/100 g Al, (b) 0.2 mL/100 g al, (e) 0.4 mL/100 g Al

175

Figure 4: Weibull distribution of elongation at fracture for different hydrogen levels (a) 0.1 mL/100 g Al, (b) 0.2 mL/100 g al, (e) 0.4 mL/100 g Al A typical example of a folded double oxide film is given in Fig 5. These SEM pictures were taken from the two halves of the facture surface of the tensile bar.

Figure 5: SEM picture showing the two halves of a bifilm on the fracture surface A graph between the bifilm index and the mechanical properties was also plotted which is given in Fig 6.

176

(a) (b) Figure 6: Tensile property change with bifiim index (a) ultimate tensile strength, (b) elongation at fracture Discussion There has been a long discussion about the linear relationship between hydrogen and porosity; and the inverse relationship between mechanical properties and hydrogen. The purpose of this work was to investigate the effect of different hydrogen levels on metal quality and mechanical properties by means of altering the hydrogen levels in two melts with opposite treatment sequences. One melt was upgassed first and degassed to low levels and sample collection was carried out at selected hydrogen levels. The other melt was degassed first and then upgassed and same sample collection procedure was followed. In the upgassed melt (Fig 2a), as the hydrogen level was increased, the bifiim index was also increased. However, the interesting finding was the increased bifiim index with decreasing hydrogen content in the degassed melt (Fig 2b). The reason for the increase in bifiim index in both cases with the treatment sequence is mainly due to the entrainment of surface oxide during the high speed rotation of the rotary degasser. The effect of the entrained surface oxides with vortex in the rotary degassing operation is much more clearly seen in Figs 3 and 4 where the mechanical test results are shown as Weibull distributions. The graphs are chosen to display the same hydrogen levels. In addition, the bifiim indexes of the melts are given on the top left corner. As can be seen from the Figures 3-4; the melt with the lower bifiim index has less scatter than the melt with high bifiim index. This is same for both ultimate tensile properties and elongation at fracture values. These results indicate that as long as there are bifilms in the melt, regardless of the hydrogen content, the mechanical properties will fail. A typical example of a bifiim in its folded state can be seen in Fig. 5, which shows mirror images of the bifiim on the two fracture surfaces of a test bar. As seen in Fig. 6, as the bifiim index is increased, both the ultimate tensile strength and elongation at fracture are in a decreasing trend. It is important to note that the bifiim index seems to have a more pronounced effect at higher values; i.e. lower tensile properties. On the other hand, at low bifiim index values, it is not clear to conclude that the properties will be high; due to the scatter of the results.

177

Conclusion Uncontrolled rotary degassing of the melt may lead to entrainment of bifilms from the surface which decreased the tensile properties. The tensile properties are more sensitive to the bifilm index than to the hydrogen content. Decreased hydrogen content does not significantly indicate increased properties. As the bifilm index increases, mechanical properties decrease. ACKNOWLEDGEMENT The European Integrated Project NADIA (New Automotive components Designed for manufactured by Intelligent processing of light Alloys - Contract NMP2-CT-2006-026563) and FREMST0T project (Norway) is gratefully acknowledged for financial support. Authors would also like to acknowledge the help of Mr. Kurt Sandaunet during the experiments. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Campbell, J., Shrinkage pressure in castings (The solidification of a Metal Sphere). Transactions of the Metallurgical Society of AIME, 1967. 239(February): p. 138-142. Campbell, J., Hydrostatic tensions in solidifying materials. Transactions of the Metallurgical Society of AIME, 1968. 242(February): p. 264-267. Campbell, J., Hydrostatic tensions in solidifying alloys. Transactions of the Metallurgical Society of AIME, 1968. 242(February): p. 268-271. Campbell, J., Castings. 2nd ed. 2003: Buttonworths. Dispinar, D., S. Ahktar, A. Nordmark, M. Di Sabatino, L. Arnberg, Degassing, hydrogen and porosity phenomena in A356. Materials Science and Engineering: A, 2010. 527(16-17): p. 3719-3725. Dispinar, D. and J. Campbell, Critical assessment of reduced pressure test. Part 2: Quantification. International Journal of Cast Metals Research, 2004. 17(5): p. 287294. Dispinar, D. and J. Campbell, Critical assessment of reduced pressure test. Part 1: Porosity phenomena. International Journal of Cast Metals Research, 2004. 17(5): p. 280-286. Dispinar, D. and J. Campbell, Effect of casting conditions on aluminium metal quality. Journal of Materials Processing Technology, 2007. 182(1-3): p. 405-410. Ransley, C.E. and H. Neufeld, The Solubility of Hydrogen in Liquid and Solid Aluminium. Journal of Institute of Metals, 1947-48. 74: p. 599-620. Gruzleski, J.E. and B. Closset, Treatment of Liquid Aluminum-Silicon Alloys. 1990: American Foundry Soceity.

178

Authors' Reply for the Reviewer comments: 1) A more insightful explanation is needed of the results shown in Fig. 2. Fig. 2(a) is understandable: as we increase the hydrogen content, bifilms introduced by a previously degassing operation will grow and cause an increase of the Bl. A plot of the number ofbifllmsfor each gas content would help to confirm that the increase ofBI is due mostly to an increase in average bifilm size while the number ofbifilms remains basically constant. On the other hand, the increase in Bl in Fig. 2(b) is, according to the authors' hypothesis, due to the entrained bifilms produced by the rotary degasser. Again, a plot of the number ofbifilms would not only confirm this hypothesis, but it would also help determine if the starting number of defects in Fig. 2(a) and Fig. 2(b) are actually comparable.

o 0

0,1

0,2

0,3

0,4

0,5

hydrogen content Number of bifilms is increasing linearly in the upgassed melt. On the other hand, in the degassed melt, it drops a little but increases again in the final treatment (as can be seen above) 2) If the findings of Fig. 2(b) are correct, then one wonders on the usefulness of degassing (at least with this method), if the properties are actually worsened. Is this finding particular to the case studied? Are there cases in which the rotary degasser reduces the bifilm index? •

These findings are particular for this experimental work. Authors have recently published a work where they could establish reproducible results for 5 consecutive samplings after a careful controlled rotary degassing.

3) In reference to Figs. 3-4, the authors affirm (third paragraph in Discussion) that melts with lower Bl show less scatter than melts with higher Bl. This is true for the upgassed melts (black triangles); however, the opposite trend seems to hold for the degassed melts (gray triangles). An explanation is needed of why this might happen. Note that in both Fig. 3 and 4, the x-scale of part (c) is not the same as parts (a) and (b). •

It was an interesting finding that as the melt was degassed from high levels, the bifilm index increased and mechanical properties decreased. Although the hydrogen content was decreasing.

4) Were the samples heat treated?. If not, are the results of any practical use? Would heat treatment affect the conclusions?

179

•

Samples were not heat treated. As-cast conditions were investigated. The heat would probably increase the elongation (and UTS), however the Weibull distributions would probably follow a similar trend, as it is the bifilms that are responsible for failures.

Other minor observations/suggestions follow: •

All the minor suggestions were corrected in the text

180

Shape Casting: The 4"1 International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

"The origin of Griffith cracks." J Campbell Emeritus Professor, Department of Metallurgy and Materials, University of Birmingham, B15 2TT, UK. Abstract The presence of any pre-existing Griffith crack or a pore is not necessarily to be expected in metals as a result of the extremely high interatomic forces. It seems likely that pores and cracks may not be created intrinsically by the atomic mechanisms involved in the formation of a solid by solidification from liquid, or condensation from vapour phases nor, probably, by mechanisms of plastic deformation. It is proposed here that initiation sites for pores and cracks for most, if not all failures of metals, can only be introduced into metals via extrinsic (entrainment) mechanisms resulting from production processes, particularly melting and casting, but also spraying and powder metallurgy processes. It seems probable that only entrainment processes can create unbonded interfaces that can explain microstructures containing cracked (apparently 'brittle') intermetallics and decohered phases, and the initiation of tensile fracture and fatigue. Key Words: Griffith crack, defects, brittle, ductile, fatigue, fracture. Introduction This short note examines the problem of the initiation of fracture in metals by the initiation of cracks and pores. Our well-established theories of fracture require a pre-existing crack to initiate brittle failure, as elegantly proposed by Griffith [1, 2]. Similarly, ductile failure requires a preexisting population of pores or cracks [3, 4]. Often, such cracks are sited in a dispersion of brittle second phase particles, or particles easily decohered from their matrix, so that under tensile stress the opened void can initiate either cracking to link up cracks from neighboring particles, or initiates failure by plastic flow, the matrix shearing to create knife-edged cusps surrounding conical dimples containing at their bases the original fractured or decohered particles. The Problem This paper attempts to address the question as to how failure can occur if the pre-existing Griffith crack, or the pre-existing population of pores did not exist. This is not a trivial question, since even a cursory overview of solidification (and of other bulk forming processes such as condensation from a vapour) gives strong pointers that such defects are not to be expected. It is easy and quick to demonstrate that solidification cannot produce defects such as a pore: the well-known equation for the mechanical stability of a spherical pore is P = 2T / r where P is the internal pressure (or external hydrostatic tensile stress),

181

T is the surface tension, and r is the radius of the pore. For liquid aluminium we can take T as approximately 1 N/m and for nucleation of a bubble of radius r = 0.28 nm, approximately one atom diameter, corresponding to a pore size of approximately 8 vacancies, P is immediately seen to be approximately 7 GPa. For liquid iron the equivalent value is approximately 16 GPa. These high stresses are confirmed to within a factor of about 2 by a number of more sophisticated estimates [5, 6]. The theoretical strength of solid metals is expected to be even higher than those of their liquid phases because the interatomic distances are slightly closer. Clearly, both liquid and solid metals are expected, with good reason, to have high strengths. Whereas many texts now conclude that some pre-existing pore must now be postulated, such as a pocket of gas trapped in a recess in an inclusion, such assumptions pre-suppose the very problem we are attempting to explain. How could a void or gas pocket occur in a solid produced by solidification? The atomic movements during the reorganisation of the liquid metal into a solid metal are only small fractions of an atomic diameter; the structure of the liquid is that of a randomly close packed solid, and the structure of the solid is, of course, regular, but otherwise very similar, with quite similar interatomic spacings. The high forces that keep the atoms together effectively forbid the opening of a void. These extremely high stresses for the 'homogeneous' nucleation of pores or cracks might, of course, be reduced in the presence of a poorly-wetted substrate that would allow 'heterogeneous' nucleation. However, for conditions of the worst possible wetting, assuming the highest contact angles ever recorded, in the region of 160 degrees, the nucleation stress is predicted to be reduced by a factor of nearly 20. Thus the fracture stresses for both liquid and solid metals is somewhat reduced, but remains high [5]. Significantly, the stresses remain in the range 103 to 104 times higher than can be met during solidification, since, as every foundry person knows, a poorly fed casting can collapse forming external sinks under only atmospheric pressure (0.1 MPa), indicating the limit to which internal tensile stress can be supported. Thus the tensile stresses sufficient to create volume defects in castings cannot be generated, simply because interatomic forces are too high to allow pores to open and hot castings are in general too weak to support such stresses. Even in some solid metals at room temperature there has been direct evidence for over 40 years that cracks and pores cannot form [7]. Transmission electron microscope observations of the condensation of a supersaturation of vacancies in a lattice might be expected to form cavities in the same way that condensation of supersaturated solutes can form second phases. However, for all metals studied so far, this is not true. TEM observations of quenched fee metals, including Al, Ag and Au, indicates that condensation of vacancies does occur, but instead of the formation of vacancy discs or three-dimensional voids, the lattice collapses under its own interatomic forces, consolidating to create dislocation rings or stacking fault tetrahedra. In more recent electron radiation studies of vacancy condensation in Fe [8], Mo [9], Zr [10] and U [11] voids were never reported; only dislocation loops were observed. Recent MD simulations

182

confirm this behaviour [12, 13, 14] showing how clusters of up to 45 vacancies collapse unstably to stacking fault tetrahedra. MD studies by Milstein [15] indicate that a tensile stress of over 15GPa is required to stabilize a void in Ni, causing it to grow explosively to promote failure. Void growth studies by Meyer et al [16] used reflected shock waves in Cu but found voids formed at grain boundaries only when the tensile stress exceeded 37 GPa. Clearly, we can conclude that solidification cannot form volume defects. The interatomic forces can only be overcome to create voids at huge stresses. This has the interesting consequence that during the normal solidification of metals there can be no formation of features such as Griffith cracks to initiate failures by cracking in castings [3]. Similarly there will be no porosity or cracks (and no decohering phases as discussed below) to initiate ductile failure [4]. The absence of failure initiation mechanisms will necessarily result in tensile tests resulting in either in very high tensile strengths, or extensive plastic flow, necking down to 100 % reduction in area. It follows that classical physical (intrinsic) metallurgy would predict that an Al-Si eutectic alloy undergoing a tensile test would not exhibit a failure of a single Si particle. If the Si particles were sufficiently strong they would cluster together as the Al matrix plastically flowed until the Si particles impinged. Alternatively, the Si particles themselves might at some stage start to plastically flow until the whole specimen finally parted by necking down to 100 % reduction in area. Practical Defect Generating Mechanisms From the above considerations, conditions for failure appear to be clear and logical: in general, failure can initiate in a metal only from interfaces that are unbonded (since atomic bonds are too strong to be broken). Since unbonded surfaces cannot be formed by intrinsic processes such as solidification or vapour phase deposition, such interfaces have to be introduced from outside the metal. These are necessarily extrinsic features. There are three main extrinsic defects: pores, bifilms and extrinsic (exogenous) inclusions [16]. All are effectively introduced in to the matrix by the impingement of surfaces during consolidation. Thus in power metallurgy routes such defects are necessarily incurred. In casting the impingement of apparently liquid surfaces (usually in fact coated with a solid oxide film of course) during turbulence is unfortunately common, but not necessary, as we shall discuss. During casting, the analogous defects are bubbles, bifilms and extrinsic inclusions. Again, the entrainment mechanism is similar, consolidation now occurring as the result of impinging droplets, or a folding over of a breaking wave. The entrainment actions and the nature of the defects are illustrated in Figures 1 to 3. Such defects, especially the bifilms, are introduced into the melt at every stir and every pour event. Also, a succession of such handling traumas adds its contribution to the total population of suspended defects in the liquid. Although the near-neutral density of the alumina bifilms in liquid aluminium ensures that these defects have a long life in suspension in the melt, severe bifilm

183

problems can also be experienced in a wide variety of cast metals, including cast irons, stainless steels and Ni-based superalloys even when cast in so-called vacuum. These defects are, of course, subsequently frozen in to the solid. Despite its extensive unbonded interface (diameters are typically in the range micrometers to centimeters) the bifilm is generally overlooked because it is often so thin (usually in the range of nanometers to micrometers) as to be invisible to casual observation. The bifilm is, of course, usually an oxide, but can on occasions be a film of carbon, or nitride etc. Its folded or collided origin necessarily results in its structure characterised by a double film with (i) unbonded inner faces, entrapping traces of residual air, and (ii) perfectly wetted exterior faces (originally the underside of the surface oxide film). All entrained oxide films necessarily have this double structure. These features explain their pivotal roles affecting the mechanical and metallurgical properties of castings and their wrought products. The third variety of entrained defect, the extrinsic inclusion, has to enter the melt through the surface oxide, necessarily carrying with it a wrapping of the surface film, and thus isolated from the melt by the oxide and its entrapped layer of air (Figure 2). In effect it enjoys no bond with the matrix, and contrasts therefore with the in-situ intrinsic inclusion that has grown atom-by-atom from the melt, and thus remains in perfect atomic contact. The population of bifilms in most engineering metals is estimated to be often in the range 106 to 109 m"3 [16]. These high numbers may appear surprising in view that these features are not generally reported. The realization that such a dense population of defects is the norm in metals makes a re-interpretation of much accepted metallurgy highly desirable. The bifilms in cast metals appear to survive considerable plastic working. Thus these casting defects influence the behaviour of many wrought products. Al alloys retain their unbonded regions even after the severe extrusion required to produce window frames, as is evident from the filiform corrosion, clearly seen by eye on unprotected extrusions, in which tens or hundreds of corrosion sites per square centimetre follow the elongated unbonded bifilms that tunnel through the metal, happening to intersect the surface from time to time to create a corrosion site. The survival of the bifilms during plastic working is probably the result of the reservoirs of air that remain trapped in the rucks and folds between the oxide surfaces. Thus any extension of the area of the interfaces by working is accompanied by simultaneous oxidation and nitridation of the freshly-created surfaces, preventing bonding until all the air is consumed. It is proposed that these unbonded surfaces are probably common in most wrought alloys, and constitute the Griffith cracks necessary for brittle failure. The Role of 'Brittle' Intcrmctallics and Second Phases Every primary intermetallic and primary second phase so far investigated appears to have formed on the wetted outer surfaces of a bifilm [16]. An image of beta-Fe particles and Si particles in Al-Si alloys is shown in Figure 3. The central cracks in both types of particle denote the location of the originating bifilm (the short transverse cracks on one side of

184

the main crack are rucks and folds in one of the components of the main bifilm, since when folding together in a turbulent event, one component will always have a larger area than the other, and will therefore be forced to adopt additional wrinkles and creases, whereas its adjoining film will be mainly flat). The bifilms are expected to be present in every beta-Fe and silicon particle in the alloy, but are not always obvious. In this case the bifilms have been opened, appearing as cracks, because of inflation, usually by some shrinkage or gas, during solidification. It seems likely that in the absence of such favoured substrates both beta-Fe and Si will not precipitate as primary phases but will be forced to appear at lower temperatures as constituents of eutectic phases. This is the proposed mechanism of modification of Si by Na and Sr, both of which are proposed to deactivate bifilms as substrates [17,18]. In general, it seems likely therefore that the appearance of cracked intermetallics and Si particles (Figure 4) is not an indication of brittleness. Intermetallics are known to be strong, and the forces involved during solidification of metals are generally a factor of 105 to 106 too small to cause fracture. (An intermetallic in an Al matrix is probably mechanically equivalent to a pebble in butter. On would not expect the pebble to fracture if the butter was deformed.) The cracks merely denote the presence of an unbonded interface that is an integral feature of their favoured substrate. These considerations are corroborated by measured fracture strengths of Si particles in Al-Si alloys that have been shown to be as low as 200 MPa [19] compared to expected theoretical strengths of at least 30GPa [20-22]. Decohesion Decohesion is possible for those phases that have precipitated from solution, but which happen to have formed on only one side of a bifilm. The other side, now consisting of only the flimsy unbonded film is easily separated from the other half of the bifilm that is now firmly attached to the precipitate. Thus it appears that the precipitate is capable of nucleating a pore or crack. The ability of a surface to resist decoherence from a matrix in perfect molecular contact was nicely demonstrated as long ago as 1867 by Gernez [24]. He showed that crystalline solids which had been grown in the liquid, and which had never been allowed to come into contact with air, were incapable of inducing effervescence in a liquid supersaturated with gas. Solids which had been allowed to dry but which were otherwise identical always caused effervescence. In this experiment the decohering forces were relatively weak, but the principle is sound. By analogy, the contact between intermetallics and the matrix from which they were formed will be atomically perfect, and thus will be strong. The famous observations on NaCl crystals [25], brittle when crushed in air, but deforming in a ductile mode when compressed under water because surface flaws are dissolved away, is analogous to the condition of an intermetallic that had been formed insitu in a melt, having an atomically smooth interface with the liquid that enjoys essentially perfect atomic contact, and so enjoying extreme resistance to decohering or fracturing.

185

Emamy and Campbell [26] compared the bonding between second phases and the matrix in two commercial metal matrix composites (MMCs) by solidifying a casting under the modest hydrostatic tension induced by the lack of feeding of a cylindrical shape. It was clear that the MMC formed by introducing SiC particles through the liquid surface, even when this was conducted under high vacuum to reduce oxide problems, exhibited significant decoherence from the matrix, creating a dispersion of fine pores. In contrast the MMC containing TiB2 particles which had been formed in situ by reaction in the matrix were resistant to decoherence from the matrix despite being subjected to even higher hydrostatic tension (the opening of pores by decoherence of a large proportion of SiC particles to some extent relieved the tension experienced by the surviving SiC/matrix interfaces). Bifilms, with their central unbonded interface, are pushed by dendrites, and therefore often finally reside in grain boundaries. Thus those boundaries containing bifilms will easily decohere (being effectively pre-cracked) during creep or superplastic flow. From observations of decohesion of such boundaries it is usually concluded that boundaries in general are weak, even though most boundaries do not decohere. Clearly, boundaries that do not contain bifilms will be expected to be strong; it may be appropriate to remind ourselves that virtually all of our major engineering alloys are full of grain boundaries: it is inconceivable that such features are weak; in fact strength is enhanced in many alloys by increasing the density of boundaries (i.e. making grains smaller). Conditions for avoidance of failure The theoretical prediction of 100 % reduction of area of metals free from extrinsic defects is supported on a microscale, as is evident in SEM images of ductile failure (Figure 6). The knife edged cusps at some locations between the cups and cones are regions that typify the failure of metals without defects. The failure mode exhibits necking to zero in this tiny cusp area. This microscale example appears similarly true on a macroscale in 'cup and cone' failure, where ductile necking to failure is only interrupted by the nucleation and growth of a central cracking region. Without the nucleation of the central crack, the test piece would continue to extend until 100 per cent reduction in area. Accepting for a moment the high density of defects commonly present to initiate failure, if, despite their presence, conditions are applied to prevent the defects opening and propagating, then high tensile performance would be expected as though the defects were absent. For instance, when tensile tests are carried out under pressures approaching the ultimate failure strength of the material, elongation continues to 100 % reduction of area. Bridgeman [27] showed that as the hydrostatic pressure on a steel specimen approached 2.67 GPa the RA rose to 100 per cent. This would be expected if cavities were prevented from opening, thus artificially simulating conditions for a metal without initiating sites. In conclusion it needs to be stated that modern melting technology is now capable of producing bifilm-free melts in most metals and alloys. Similarly, such very clean melts can be now cast without turbulence that would re-introduce a second population of

186

bifilms. These practices are beginning to be adopted, and substantial achievements in terms of improved cast products are beginning to be reported [28]. Conclusions It is postulated that mesoscopic or macroscopic defects entrained during manufacture dominate the failure of metals, and that these features constitute the main, if not the only, source of Griffith cracks in both cast and wrought metal. Thus 1. Powder metallurgy and other particulate processes such as spray forming etc necessarily result in extrinsically introduced permanent damage in the form of oxide bifilms. 2. Unnecessarily poor casting techniques cause similar extrinsic crack defects. 3. Solidification does not initiate the creation of extrinsic damaging defects 4. Porosity or cracks have an extrinsic origin in the casting process and can constitute points of initiation of failure of metals. 5. Probably most if not all types of failure initiate only from entrained unbonded interfaces provided by the immense population of bifilms to be expected in all our current engineering metals. 6. Failures that appear to have initiated by decoherence or fracture of foreign inclusions will in fact have originated on its associated bifilm acquired during entrai nment. 7. Extrinsic defects introduced during the manufacture of metals appear to survive significant plastic working processes, so that defects tend to be permanent, becoming usually merely elongated or fragmented along the working direction. In an appropriately cast and solidified metal, within the scope of current technology, a Griffith crack could not be formed and could not subsequently be generated by, for instance, plastic flow, and thus in general would not exist. Thus metals cast without such defects would be expected to exhibit the following microstructural behaviour: 1. Intermetallics should never crack because they are in general extremely strong (furthermore, most would never form in the absence of the favourable entrained bifilm substrate). 2. Inclusions formed in-situ in the melt or matrix should never decohere from the matrix. 3. Grain boundaries should never decohere in creep or superplastic forming. 4. Failure of metals should therefore occur by (i) brittle fracture only at extremely high stress close to the theoretical strength of the metal, or (ii) by ductile failure at extremely high elongations associated with 100% reduction of area. In summary; it is proposed that many, if not most, metal failures originate from casting defects, particularly bifilms, and are preventable. The Griffith crack need not exist. Premature failure of metals need not occur.

187

References 1. A. H. Cottrell: "The Griffith Centenary Meeting" 23-25 March 1993 pp 4-15 Institute of Materials, London. 2. A. H. Cottrell: Proc. Roy. Soc. A 1963 279 1-10. 3. P. F. Thomason; "Ductile Fracture of Metals" Pergamon, Oxford and NY, 1990 4. J. F. Knott: in 'Recent Advances in Fracture' (Editor R. K. Mahidhara et al) 1997 TMS Warrendale, PA, USA. 5. J. Campbell; Iron & Steel Inst. Special Report "The Solidification of Metals" 1968 vol 110 pp 18-26. 6. J. C. Fisher; Journal of Applied Physics 1948 vol 19 pp 1062-1067. 7. D Kuhlmann-Wilsdorf: in 'Lattice defects in quenched metals' 1965 Academic Press, New York. P. Moser, C. Corbel, P. Lucasson, P. Hautojarvi; Materials Sci. Forum 1987 vols. 15-18 pp 925-930 8. A. Sen Gupta, P. Moser, A. Bourret, C. Corbel, S. V. Naidu, P. Sen and P. Hautojarvi; Ibid pp 931-936. 9. C. Hellio, C. H. de Novion, A. Marraud, L. Boulanger; Ibid pp 937-942. 10. C. Weiberg and Y. Quere; Ibid pp 943-948. 11. B. P. Uberuaga, R. G. Hoagland, S. M. Valone, A. F. Voter; Phys Rev Lett 2007 vol 99 p 135501-3. 12. M. J. Sabochick, S. Yip, N. Q. Lan; Mater Sci Forum 1987 vols 15-18 pp 857-862 13. S. Traiviratana, G. M. Bringa, D. J. Benson, M. A. Meyers; Acta Mat. 2008 in press. Doi:10.1016/j.actamat.2008.03.047. 14. F. Milstein, J. Zhao, D. Maroudas; Phys Rev B 2004, vol 70, 184102, pp 1-6 15. M. A. Meyers, S. Traiviratana, V. A. Lubarda, D. J. Benson, E. M. Bringa; J of Metals 2009 vol 61 (No 2) 35-41. 16. J. Campbell; Materials Sci & Technology 2006 vol 22 (no 2) 127-145 and (no 8) pp 999-1008. 17. J. Campbell; Met & Mat Trans A 2009 vol 4-A pp 1009-1010 18. J. Campbell and M. Tiryakioglu; Materials Sci & Technol 2010 26 (3) 262-268. 19. J. Griffith, E. C. Oliver, M. E. Fitzpatrick, T. R. Finlayson, D. Viano and Q. Wang; Shape Casting: The 2nd Internat Symp edited by P. N. Crepeau, M. Tiryakioglu and J. Campbell; TMS (The Minerals, Metals and Materials Soc) 2007 pp 127-134. 20. E. Orowan "Fracture and Strength of Solids" Reports Progress Physics 1948-1949 vol 12 pp 185-232. 21. A. Kelly "Strong Solids" Clarendon Press 1966. 22. K. Gall, M. F. Horstemeyer, M. van Schilfgaarde and M. I. Baskes: J Mech Phys Solids 2000 vol 48 pp 2183-2218. 23. R. K. Govila and D. Hull; Acta Met. 1968, vol. 16, pp. 45-52 24. M Gernez; Phil Mag 1867 33 (4)79 25. S. Mendelson: J Appi Phys 1962 vol 33 (no 7) pp 2182-2186. 26. M. Emamy and J. Campbell: Cast Metals 1995 vol 8 pp 115-122 27. P. W. Bridgeman; "Studies in Large Plastic Flow and Fracture (with special emphasis on the effects of hydrostatic pressure)." McGraw-Hill 1952 pp. 38-86. 28. J. Campbell "Casting Practice; The 10 Rules for Casting" Elsevier 2004.

188

Figure 1. Creation of bifilms of large size by surface turbulence.

Figure 2. The entrainment of the surface oxide as an extrinsic inclusion penetrates the surface of a melt.

Figure 4. SEM image of a fracture surface of an Al-Si eutectic phase (Courtesy M. Tiryakioglu).

Figure 3. Beta-Fe particle (upper left) and Si particles (centre) in an Al-Si alloy casting, showing central and transverse cracks, and apparent decoherence from the matrix (Courtesy X. Cao).

189

Shape Casting: The 4th International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

The Use of the Weibull Statistical Method to Assess the Reliability of Cast Aluminum Engine Blocks made from Different Casting Processes Robert Mackay & Glenn Byczynski Nemak of Canada Corporation 4655 GN Booth Drive, Windsor ON N9C 4G5, Canada Keywords: Weibull Statistics, High Pressure Die Cast, Precision Sand, Tensile Properties Abstract The use of aluminum cast engine blocks has grown considerably within the automotive industry due to their lighter weight when compared to traditional materials such as cast iron. In this investigation three aluminum cast engine block processes - High Pressure Die Cast, Precision Sand Cast ProcessZircon and Precision Sand Cast Process-Silica/Chill, are evaluated in terms of reliability (Weibull Plots) of the mechanical strength determined from tensile test samples extracted from the bulkhead region. Metallographic analysis was performed on the tensile test samples to provide interpretative feedback on the statistical analysis of the mechanical test results. Introduction Today cast aluminum alloys are the materials of choice for automotive cylinder blocks and cylinder heads. They have proven themselves to be a lightweight alternative to the much heavier ferrous based versions of the past. The reduction in vehicle weight remains as one of the largest single improvements to overall vehicle fuel economy. Despite having firmly established themselves in this niche, aluminum castings are poised for yet future growth, notably in the diesel engine cylinder block area where many designs today are still based on iron. Close to 60% of all aluminum engine blocks are cast using the High Pressure Die Cast (HPDC) process, with the balance made using either Gravity Semi-permanent Mould (GSPM) or Precision Sand Casting Process (PSCP). PSCP's in recent years have begun to integrate a crank positioned chill for improved tensile and fatigue performance. Table 1 summarizes much of the differences in key process parameters associated with Table 1: Key Metallurgical & Process Characteristics for Aluminum Engine Block Castings

Low

PSCP Head deck: 40-50 μηι Bulkhead: 55-65 μηι Medium

PSCP - Chill Head deck: 60-70 μηι Bulkhead: 20-30 μπι High

Typical Scrap

5- 10%In-house 0.2 - 2% Customer

3 - 5%In-house 0.10 -0.30% Customer

3 - 5% In-house 0.10-0.30% Customer

Process Element Typical Microstructure Soundness

10-15 μπι

HPDC

Rationale Inherent to process

Casting Yield

70-75%

73%

55%

Heat treatment

T5

T5

T5&T7

Alloy Impregnation Durability Casting Temp Modifier / Grain Refîner

AlSil0Cu2.5(1.0Femax) 10-100% of product Low 680-700X

AISi7Cu3,5(0.4Femax) 0% of product Medium 730-760°C

AISi8Cu3(0.6Femax) 0% of product High 730°C

Property Increase Inherent process (warm up scrap included) Design necessity Property Increase & Dimensional stability Customer Requirement As required Customer requirement Inherent to process

None

None

Sr/TiB

Improved Castability

191

casting structure, required thermal processing, typical scrap rates, cast melt temperatures and finally the use of metal enhancers such as A1-5TÌ-1B additions. HPDC processes for engine blocks can be very complex, requiring significant capital investment, complex dies with cooling circuits, controlled filling characteristics (gate velocities of close to 30 meters/sec), and pressures of 1,400 bar in the die cavity. The stages for a complete die cast cycle are: (1) the die preparation (die spray, die blow off, cast iron liner insertion & finally die closing), (2) metal delivery (shot chamber filling followed by high pressure injection), (3) phase transformation (liquid to solid) and then finally (4) casting extraction. It is commonly believed that the most critical stage of the die cast cycle is the metal delivery phase, where three main steps are involved. The slow shot (initial introduction of metal) is where the liquid metal is brought into the gating section of the die. This ensures that there is a minimal amount of trapped air in the shot sleeve, and ensures that liquid metal turbulence doesn't occur, preventing entrainment defects being formed. The injection stage is where the liquid metal in the shot sleeve is pushed into the die cavity by the plunger at very high velocity. The fill rates are typically between 80 and 100 milliseconds; for reference the blink of a human eye is between 300 to 400 milliseconds. The third and final stage is intensification, where the semi-solid metal in the die is placed under incremental pressure to reduce both internal porosity in the casting and compensate for volumetric shrinkage. After solidification is complete the casting is removed, mechanically degated, and then quenched in 90°C water to lock in solute within the aluminum matrix, and cool sufficiently for handling. The casting is then typically thermally processed using a T5 treatment, machined, quality checked & leak tested, impregnated if necessary and ultimately shipped to the customer's engine plant. For PSCP, a room temperature mould is employed. The mould is composed of individual interlocking sand cores that can be produced in a variety of core making methods. One of the most popular is the phenolic urethane cold box method employing silica or zircon sand and a two part amine cured binder sytem. The assembled sand mould may be filled using gravity or counter gravity methods e.g. Low Pressure or Electromagnetic Pump. Both counter gravity casting processes are capable of producing a controlled quiescent filling of the mold that reduces the formation of bifilms in the casting [1,2]. Bifilms are known to impair casting quality and properties most notably fatigue and tensile performance. Typical metal front velocities are below 0.5 meters/sec, and the typical fill times can range from 20 to 30 seconds depending on the casting size; this is in strong contrast to the HPDC fill process. While the virtues of zircon sand as a mould material have been enjoyed in the past by metal casters its prohibitive cost and availability have caused a major trend towards the more affordable silica sand. When higher property performance is required, a chill can be integrated with the core package to provide a strong extraction of heat, yielding a refined microstructure in the cast section adjacent to the chill. In cylinder blocks it is typically the bulkhead or main bearing saddle area that warrant these improved material properties. Thermal processing of the filled moulds is usually a Thermal Sand Removal (TSR), followed by T5, T6 or T7 heat treatment. Typically for dimensional stability a T7 (over-aged) process is used. Following the thermal process, cubing or hyper cubing is performed, and after the appropriate quality checks it is then shipped to the customer engine plant.

192

Experimental Procedure The typical chemistry of the casting process studied in this work for HPDC, PSCP-Zircon and PSCPSilica/Chill are given in Table 2. Table 2: Alloy Chemistries for Engine Block Casting Processes (wt.%) Cu

Mg

Mn

1.00

0.50

0.34

0.27 0.34

Fe

Casting Type

Si

HPDC

10.00

2.75

PSCP-Zircon

7.40

3.40

PSCP-Silica/Chill

8.60

2.85

0.50

Zn

Ti

Ni

0.50

3.00

0.25

0.50

0.24

0.15

0.13

0.04

0.35

0.50

0.11

0.06

The tensile test bars were fabricated and tested as indicated by the ASTM B557 protocol. The tensile test sample geometry is shown in Figure 1. To ensure the exact placement of the tensile test bar blanks consistently from a bulkhead section of an engine block casting, a template was used to define the exact location for the blanks before sectioning. Due to microstructural anisotropy even small variations of location of the tensile test bar (particularly in the case for PSCP-Silica/Chill) have been found to produce statistical noise in the test bar property population. The total number of test samples extracted from the bulkheads of each cast type was 30. Previous work has shown this to be an acceptable number of tensile test samples needed to support the statistical Weibull method [3].

--ΕβΞί G

G - Gauge Length: D - Diameter: R-Radius of Fillet: A - Length of Reduced Section:

45.0 ±0.1 mm 9.0 ± 0.1 mm 8.0±0.1mm 54.0 ± 0.1 mm

R

f

Figure 1: Dimensions of the cylindrical tensile test samples used in this study. They are in accordance with ASTM-577 standard. All dimensions are in millimeters. Secondary Dendrite Arm Spacing (λ2), Mean Porosity and Largest Pore Found are measured just 5 mm below the fracture surface for a representative test bar from the HPDC, PSCP-Zircon and PSCPSilica/Chill data sets. The Metallurgical Laboratory that performed all tensile and metallographic work in this article is register under the ISO/IEC 17025 standard, and was accredited to this registration by A2LA (certificate #2478.1). The Weibull analysis begins with rearrangement of the data in ascending order. A frequency plot can then be constructed where the occurrence of a specific data point is quantified within narrow range sets [3]. In this work both the Ultimate Tensile Strength (UTS), and total Elongation (Ε1.ΤΟτ) have been treated using this approach.

193

A cumulative frequency distribution of a random variable x can be represented using the three parameter form of the Weibull equation: (1)

/>(*) = 1-exp

P(x) is the cumulative frequency distribution of a random variable x, in this case either UTS, or El.TOT; X„ is the minimum allowable value of the analyzed variable, which in this case is 0 for both stress and strain data; Θ is the scale parameter, also referred to as the characteristic value; and m is the shape parameter, also known as the Weibull modulus. Equation (1) can be converted into a linear representation by applying the 'In' operator to both sides of the equation to yield the following form:

In Ini

1

\-P(x)

"

ηι\η(χ)-ηι\η(θ)

(2)

In this form the equation can be interpreted by the linear form 'y = mx + b\ where a variable y is plotted in the domain x to give a line with slope m, and a ^-intercept at b. By applying this approach a plot of (v vs. x) was constructed for both the UTS data and the ΕΙ.τοτ data sets, and a liner best-fit was made to calculate the slope m in each case. In order to interpret the cumulative frequency (which represents the .y-axis) of the variable x, the P(x), a commonly applied method was used:

W =i ^ n

(3)

In this equation the reorganized V data points (ascending order) are assigned a ranking of y in the range of (/-«), such that each point has a ranking P(j) on the cumulative frequency plot. A combination of this ranking for P(x) data set with the previously described equation (2) yields the Weibull plot. The usefulness of this analysis comes from the realization that different distributions of data of variable 'x' will yield different amounts of scatter in the data sets. The Weibull modulus 'm' in equation (2) is a quantitative measure of this scatter, such that the greater the value of '/«', the steeper the slope of the Weibull plot, and the smaller the scatter in the data. This translates into a narrower range of the analyzed property. The smaller the value of 'm' on the other hand, the shallower the slope of the Weibull plot, which implies a large scatter in the analyzed data. This in turn makes accurate predictions of the material property that much more difficult. As a result, it becomes evident that the 'm' parameter in the Weibull analysis is a powerful tool in the tensile data analysis, and material reliability in particular. Results Figures 2a, 2b and 2c represent the microstructure found below the fracture surface of the representative tensile test samples for the HPDC, PSC-Zircon, and PSCP - Silica/Chill. Included in the mosaic images are the measured value of Secondary Dendrite Arm Spacing (λ2), Mean Porosity

194

and Largest Pore Found. Seen from the micrographs the fracture surface traverses specifically through the interdendritic regions which contains the Al-Si eutectic and other secondary phase constituents (transgranular fracture) for the HPDC, PSC-Zircon, and PSCP - Silica/Chill tensile test samples. The preferred fracture path is governed by the stress concentration effects of brittle secondary phase constituents, oxides and porosity, ahead of the advancing crack. From the porosity results seen in Figures 2a, 2b and 2c the HPDC test bar contains the highest value in porosity, while the PSCP-Silica/Chill has the lowest. It should be noted that the maximum pore size of 0 μηι merely reflects that fact that the Image Analysis software doesn't measure below pores below 50 μιτι. The PSCP-Zircon has porosity values close to that of PSCP-Silica/Chill, but has λ2 values that are much higher than for HPSC and PSCP-Silica/Chill. The origin of the porosity found in HPDC and PSCPZircon is very different from one another. The porosity seen in the HPDC test sample of Figure 2b is due to metal turbulence upon filling, whereas in Figure 2a for the PSCP-Zircon, the porosity evolves from very slow solidification rates after non-turbulent filling. A slower solidification rate promotes a coarser microstructure and longer liquid metal feeding ranges [4]. The lowest porosity seen in the PSCP-Silica/Chill is due to non-turbulent or quiescent filling, followed by rapid solidification imposed by the chill. Rapid solidification conversely promotes shorter liquid feeding ranges [4]. Figures 3a and 3b shows the histogram plots for Elongation and UTS for the three castings processes studied for this work. These histogram plots clearly show not only the range of values produced for the three processes, but demonstrates the clear separation in the population groups. For the HPDC there exists a wide range of UTS and Elongation values, whereas for PSCP - Zircon, there exists a narrow range in UTS and Elongation values. For PSCP-Zircon the distribution in UTS and Elongation does overlap with HPDC. Weibull probability plots for UTS and Elongation are given in Figures 4a and 4b respectively. The value of the slopes was determined by a best fit to the data in the Weibull plots. HPDC has a low Weibull Modulus in both UTS and Elongation, reflecting a much larger scatter in values, or a lower reliability. PSCP-Silica/Chill has the highest Weibull Modulus in both UTS and Elongation, or a high reliability. PSCP-Zircon has reliability in between HPDC and PSCP-Silica/Chill. Discussion The turbulent fill process associated with HPDC produces a high density of oxides (bifilms) that consequently lead to higher amounts of porosity formation, and ultimately lead to a higher scatter in tensile test results. This contrasts with PSCP process where the scatter is lower. However when considering the values themselves, the Histogram Plots show notable overlap in Elongation and UTS between the PSCP Zircon and HPDC distributions. The PSCP-Silica/Chill process has a quiescent filling regime (leading to a lower oxide generation), followed by rapid solidification (shorter feeding ranges) imposed by the chill. This leads to lower porosity levels and consequently to more consistent tensile test value range. However, there are secondary factors that contribute to the difference in reliability seen in Figures 4a and 4b. HPDC and PSCP-Zircon are thermally processed in a T5 condition, where PSCP-Silica/Chill is thermally processed in a T7 condition, and HPDC has elevated Fe concentrations in the melt, which leads to Fe-based-sludge formation (see Figure 2b) that can be found in the interdendritic regions of the cast structure. The value of SDAS (λ2) is be a another secondary factor when comparing directly PSCP-Zircon and PSCP-Silica/Chill, contrasting once again with HPDC, where there is sufficient quantity of porosity and the presence of Fe-bearing sludge particles to override effect of smaller of Xj and finer sized defect structure.

195

Figure 2a: Micrograph Mosaic of the PSCP-Zircon Tensile Test Bar, along with High Magnification Micrograph Image.

Figure 2b: Micrograph Mosaic of the HPDC Tensile Test Bar, along with High Magnification Micrograph Image.

Figure 2c: Micrograph Mosaic of the PSCP-Silica/Chill Tensile Test Bar, along with High Magnification Micrograph Image

196

Figure 4a: Weibull Plot (Elongation)

Figure 4b: Weibull Plot (UTS)

197

Nonetheless, considering the overlap in HPDC and PSCP-Zircon seen in the Histogram Plots of Figures 3a and 3b, it is possible to get a HPDC test samples that have a low enough density of oxides/porosity to achieve a tensile test results in the range to that found in the quiescent filled PSCPZircon test samples. PSCP-Silica/Chill however shows a clear separation due to low porosity, low oxides, and a refined microstructure imposed by the chill. The T7 helps develop the mechanical properties to a higher level than that achievable by the T5 alone. Conclusions From the data studied in this investigation it was found that castings produced by the HPDC process tend to demonstrate lower reliability compared with PSCP process. This is thought to be due to the inherent contrasting nature of the cavity filling phases of the two processes. In terms of Elongation however, the HPDC values measured are on par with PSCP-Zircon produced castings in terms of performance and scatter. This is likely due to the relatively slow cooling rate in PSCP-Zircon process that allows (albeit fewer) bilfilms sufficient time to unfurl and expand to larger size defects which consistently limit the elongation potential of the area of the casting tested. Weibull analysis is a useful tool to discriminate the reliability of tensile strength performance in aluminum casting processes. The Weibull moduli of UTS values clearly demonstrate the relative reliability of the processes tested in this study. The higher modulus in the PSCP-Silica/Chill values is distinct in both UTS and Elongation from HPDC and PSCP-Zircon. These benefits are attributed to the favourable and additive affects of the lower density of bifilms in the material and the size suppression of bifilms (lack of unfurling) due to the rapid solidification of the chilled metal. References 1) O.E. Byczynski, The Strength and Fatigue Performance of 319 Aluminum Alloy Castings, Ph.D. Dissertation, School of Metallurgy and Materials - University of Birmingham, 2002. 2) N. R. Green, J. Campbell, Influence of Oxide Film Filling Defects on The Strength of Al-7Si-Mg Alloy Castings, AFS Transactions, v. 102, p. 341, 1995. 3) W. Weibull, A Statistical Distribution of Wide Applicability, Journal of Applied Mechanics, v. 18, p. 293, 1951. 4) R.I. Mackay & J.H. Sokolowski, "Alloying with Silicon and Copper in Aluminum Alloys", Journal Materials Science Forum, Vols. 539-546, pp.392-397 (March 2007). 5) G. Shabestari & J. E. Gruzleski, "Modification of Iron Intermetallics by Strontium in 413 Alloys", AFS Transactions, Vol. 103, pp. 285-293, 1995.

198

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

ULTRA-HIGH STRENGTH SAND CASTINGS FROM ALUMINUM ALLOY 7042 O.N. Senkov1·2, A.P. Druschitz3, S.V. Senkova1·2, K.L. Kendig1, J. Griffin3 1

Air Force Research Laboratory, Materials and Manufacturing Directorate, Wright-Patterson AFB, OH 45433 2 UES, Inc., Dayton, OH 45432 3 University of Alabama at Birmingham, Birmingham, AL 35294

Keywords: cast aluminum alloy, solidification under pressure, ultra-high strength ABSTRACT Ultra high strength aluminum alloy castings with good ductility have been successfully produced using the combination of bonded sand casting, an Al-Zn-Mg-Cu-Sc alloy (AA 7042) and solidification under pressure. INTRODUCTION Currently available commercial foundry aluminum alloys have much lower strength than wrought aluminum alloys and, therefore, their applications are limited to low-load structures [1]. High-strength Al-Zn-Mg-Cu (7xxx series) alloys are difficult to cast materials. Among the most common problems that occur during casting of these materials are intergranular porosity, hot tears and cold cracks. Hot tears form in the mushy zone due to the interplay between two main mechanisms: deformation of the partially coherent solid and lack of interdendritic liquid feeding [2,3]. A foundry version of the high strength 7XXX series alloy that can be cast and provide combination of strength and ductility of wrought alloys, but without additional working, would be desirable. This paper reports results of our recent work on successful sand casting of Sccontaining high strength aluminum alloy 7042. Hydrostatic pressure was applied during solidification to reduce intergranular porosity and prevent hot tearing. EXPERIMENTAL PROCEDURES 178-mm diameter billets of the 7042 aluminum alloy, which was developed by UES, Inc. under several Air Force contracts [4,5] and was formerly known as SSA0X8, were produced by Universal Alloy Corporation, Anaheim, CA using direct chill (DC) casting. The chemical composition of the produced billets is given in Table 1. To study the effect of hydrostatic pressure on casting quality, alloy charges of ~11.5 kg were melted in an electrically heated crucible furnace at 760°C, degassed with argon gas for 3 minutes, poured into chemically bonded sand molds, and solidified under 1 or 10 atmospheres of hydrostatic gas pressure. The pressure vessel was 1.2 meter in diameter and 1.8 meter in height and the pressure was applied via a nearequal mixture of dry compressed air and compressed nitrogen gas. Peak pressure was achieved about 90 seconds after pouring.

199

Table 1. Chemical composition (wt%) of the 7042 aluminum alloy used in this study Zr Ti Al Zn Mg Mn Sc Others Cu Si Fe balance 7.12 0.17 0.28 <0.05 <0.12 2.01 1.62 0.33 0.01 0.03 A schematic of the wedge shape casting and gating system are shown in Figure 1. The wedge thickness increased from 6.35 mm at the bottom to 57.15 mm at the top, the wedge height was 254 mm and the wedge width was 235 mm. As no separate risers were used, the top of the wedge casting acted as the riser by feeding the lower regions with liquid alloy. The variation in the wedge thickness produced various solidification rates, with the fastest cooling rate at the bottom and decreasing cooling rates as the thickness increased. To determine the actual cooling rates, four thermocouples were placed at different wedge thicknesses and designated A, B, C and D, Figure 1.

Figure 1. Schematic of the wedge casting and gating system. Microstructures of the sand cast wedges were studied with the use of scanning electron microscopy and analyzed with the use of Fovea Pro image processing software. Hot isostatic pressing (HIP) of the wedge casting was conducted at Bodycote North American HIP, Princeton, KY, by ramping at 4°C/min to 460°C with a simultaneous increase in the pressure to 207 MPa, holding at 460°C for 2 hours, ramping to 500°C, holding for 1 hour and furnace cooling as the process pressure was released. Prior to tensile testing, the samples were heat treated to the T4, T6 or T7 tempers. The solution treatment and aging conditions are given in Table 2. The solution treated samples were water quenched prior to aging. Table 3. ieat treatment conditions. HT ID Temper Solution Treatment T4 T4 460°C for 2hrs + 480°C for 1 hr T6-1 T6 441°Cfor4hrs T6-3 T6 480°C for Ihr T7 T7 441°Cfor4hrs

200

Aging 23°C for 7 days 121°Cfor24hrs 120°Cfor24hrs 150°Cfor 18hrs

Tensile tests of samples extracted from different sections of as-cast and HIPed wedges were conducted at room temperature, in accordance with ASTM E8-04 [6] and ASTM B557M-10 [7], at both the University of Alabama at Birmingham (UAB) and the Air Force Research Laboratory (AFRL) using MTS 810 servo-hydraulic materials testing systems. The UAB tensile samples had a diameter of 9 mm and gage length of 36 mm. The AFRL tensile samples had a rectangular cross-section of 2.5 x 3.6 mm and length of 20 mm and were tested at a constant ramp speed of 0.02 mm/sec (the initial strain rate of 0.001 s"1). RESULTS & DISCUSSION Macrostructure of As-Cast Wedge Samples Large spherical pores were present at the side surfaces of the sample that was solidified at 1 atm pressure. The size of these pores increased with increasing sample thickness, i.e. decreasing solidification rate. A rather large number density of smaller pores evenly distributed throughout the cross-section could also be seen by the naked eye. Contrary to this, no visible porosity was seen in the wedge sample that was solidified at 10 atm pressure. Instead, a large shrinkage cavity was observed at the top of this wedge, outlining the wedge region that solidified at the slowest cooling rate and that played the role of a riser feeding liquid to the regions that were faster to solidify. Photographs of the transverse cross-sections of sand cast wedge samples solidified under 1 or 10 atm pressure are shown in Figure 2.

Figure 2. Polished and macro-etched cross sections of sand cast wedge samples solidified at (a) 1 atm and (b) 10 atm pressure. Microstructure of As-Cast Wedge Samples The microstructures in two sections of the as-cast wedges solidified at 1 or 10 atm pressure plus the 10 atm HIPed sample are shown in Figure 3. Wedges had a relatively fine, equiaxed grain structure, with the grain size increasing from 50-58 μιη to 72-80 μηι with an increase in the wedge thickness from -15 mm to 51 mm. Solidification under pressure and HIPing had almost

201

no effect on the grain size. This result indicated good resistance of the 7042 Al alloy to grain growth during exposure to elevated (solution treatment) temperatures.

Figure 3. SEM backscatter images of the as-cast microstructure in (a) Section A and (d) Section D of the sample solidified at 1 atm pressure and (b) Section A and (e) Section D of the sample solidified at 10 atm pressure and (c) Section A and (f) Section D of the sample solidified at 10 atm pressure followed by HIPing. The images were taken from the middle of the corresponding wedge sections. Second phase particles and shrinkage-induced pores were mainly located at grain boundaries in both wedges. The volume fraction of second phase particles in the as-cast samples depended upon the solidification pressure and the sample thickness. After solidification at 1 atmosphere pressure, the volume fraction of these particles was ~ 6.8% in the thin Section A and continuously decreased from -6.8% to 2.1% as the wedge thickness increased from -12 mm (Section A) to 51 mm (Section D). HIPing decreased the volume fraction of second phase particles (see Figure 4). The decrease in the particle volume fraction (relative to the as-cast condition) was higher in Sections B and C, followed by Section A, and was the lowest in Section D. After solidification at 10 atmosphere pressure, the volume fraction of the second phase particles was ~ 3.5% in Sections A, B, and C and decreased slightly to -2.8% in Section D. Within the experimental error, the volume fraction of second phase particles did not depend on the applied

202

pressure in sections with the thicknesses greater than -25 mm, however, 10 atmosphere pressure applied during solidification definitely suppressed formation of the second phase particles in the thin Section A, which agreed with cooling curve data that showed no eutectic formation in Section A. One possible explanation is that some eutectic was dissolved during continuous cooling as the alloy was still at high temperatures resulting in some level of homogenization. However, the cooling time interval of the sections between 475°C and 400°C was approximately the same and ~570 sec in the 1 atm wedge and -540 sec in the 10 atm wedge. Another possible explanation is that the most of the second phase particles at grain boundaries formed not during the eutectic reaction but at higher temperatures, due to the presence of higher melting point elements, such as Fé, Cu, Si, Zr and Sc. The volume fraction of pores in the wedge solidified at 1 atmosphere continuously increased from -1.7% to - 4 % with increasing wedge thickness from 12 mm to 51 mm. Application of 10 atmospheres during solidification led to a considerable decrease in the volume fraction of pores in the sections with thicknesses below 40 mm. For example, the amount of pores was reduced from 3.4% to 1.5% in Section C and from 1.7% to 0.2% in Section A. At the same time, the volume fraction of pores in Section D (thickness -51 mm) was noticeably larger in the 10 atm wedge, which was evidently due to formation of the large shrinkage pore (see Figure 4). HIPing closed almost all of the pores in Sections A and B, however, there was no decrease in the porosity in Sections C and D, Figure 4, indicating that the porosity was connected to the sample surface.

Sample Thickness (mm)

Sample Thickness (nm>

Sample Thickness (mm)

Figure 4. Variation of (a) grain size, (b) second phase particles and (c) pore volume as a function of wedge thickness. Measurements are for the as-cast condition for the 1 atm and 10 atm samples, and for the 10 atm sample after additional HIPing. Mechanical Properties As-Cast Condition. The positive effect of the pressure on the properties was very strong. For example, in 13 to 36 mm thick sections, application of 10 atm pressure during solidification increased the yield strength (YS) by 23-28% and the ultimate tensile strength (UTS) by 21-33% compared to the conventionally cast alloy. In the wedge sections of -33 mm and smaller, the samples solidified atlO atm pressure exhibited YS = 489-522 MPa and UTS = 529-592 MPa. Such high strength values have never been reported for sand cast aluminum alloys before. In both 1 atm and 10 atm cast wedges though, YS and UTS continuously decreased with increasing wedge thickness. Room temperature tensile properties of the T6 and T7 tempered samples extracted from as-cast wedge sections of different thicknesses are given in Tables 3&4 for wedges solidified at 1 and 10 atmosphere pressure, respectively.

203

Table 3. Tensile properties of samples extracted from as-cast wedge regions of different thicknesses solidified at 1 atm pressure and heat treated to the T6 and T7 tempers. Wedge Thickness (mm)

14 33 21 25 30 40 43 47 18 37

Heat Treatment ID T6-1 T6-1 T6-2 T6-2 T6-2 T6-2 T6-2 T6-2

T7 T7

YS (MPa)

UTS (MPa)

Elongation

465 396 453 450 428 360 343 327 446 370

473 402 453 480 450 360 355 332 446 370

1.0 0.9 0.8 1.5 1.2 0.7 0.9 0.8 0.8 0.8

(%)

BHN 144 124 138 140 142 132 132 120 145 126

Table 4. Tensile properties of samples extracted from as-cast wedge regions of different thicknesses solidified at 10 atm pressure and heat treated to the T6 and T7 tempers. Wedge Thickness (mm)

13 30 16 19 23 27 33 37 39 42 16 33

Heat Treatment ID T6-1 T6-1 T6-2 T6-2 T6-2 T6-2 T6-2 T6-2 T6-2 T6-2

T7 T7

YS (MPa)

UTS (MPa)

Elongation

522 504 505 519 519 511 489 443 344 313 544 507

584 547 592 592 596 585 529 480 363 326 573 533

7.0 5.4 5.1

(%)

10.3

8.4 6.9 1.9 1.8 1.0 0.9 5.6 2.3

BHN 154 151 152 159 163 165 145 145 132 113 164 158

Tensile ductility of the conventionally cast alloy was very low, 0.8 -1.5%, and it was practically unaffected by the wedge thickness. Probably, the amount of intergranular porosity presented in the thinnest section was sufficient to trigger brittle intergranular fracture, so that further increase in the volume fraction of the pores with an increase in the wedge thickness had little effect on the tensile ductility but reduced the strength considerably. Solidification at 10 atm pressure led to a noticeable increase in the tensile ductility of the cast alloy in the wedge thickness regions up to about 30 mm. In thicker regions, the ductility rapidly decreased and samples extracted from -39-42 mm thick wedge sections showed almost brittle behavior. It is interesting to note that all samples showing brittle-like behavior had > 1 vol% porosity, which may indicate the critical amount of intergranular porosity for this alloy to become brittle is - 1 % . After solidification at 10 atm pressure and solution treatment, the alloy showed good response to natural aging. For example, after holding at room temperature for 7 days, 10-20 mm thick sections had YS = 350-360 MPa, UTS = 520-540 MPa and elongation above 13%, Table 5. The values of these three properties rapidly decreased to YS = 239 MPa, UTS = 296 MPa and El = 2.5% with an increase in the wedge thickness to 41 mm; however, the T4-tempered samples

204

were much more ductile than To-tempered samples in the studied thickness range. The finer GPI zones formed during natural aging (T4 temper) probably resulted in more homogeneous deformation of grains and reduced stress concentrations at grain boundaries than the coarser GPII zones and η ' particles formed during artificial aging (T6 and T7 tempers). Table 5. Tensile properties of samples extracted from as-cast wedge regions of different thicknesses solidified at 10 atm pressure and heat treated to the T4 temper. UTS Elongation Wedge Thickness Heat Treatment YS BHN (mm) ID (MPa) (MPa) (%) 13.1 123 13 T4 360 539 13.9 120 19 T4 351 521 125 23 T4 512 11.8 347 9.8 115 26 T4 338 492 7.8 115 29 T4 335 472 9.9 114 36 T4 321 469 5.8 102 37 T4 281 380 2.5 77 41 T4 239 296 Brinell hardness of the cast alloy increased with increasing pressure during solidification and decreasing wedge thickness (see Tables 3-6). After T6 and T7 tempers, the alloy solidified at 1 atm pressure had a hardness range of 138-145 BHN, while the alloy solidified at 10 atm pressure hadahardness range of 151-165 BHN in 13-30 mm thick section sizes. The T4-tempered samples solidified at 10 atm pressure had a hardness range of 114-125 BHN in 13-36 mm thick section sizes. Effect of HIPing. The YS slightly decreased, while UTS, ductility and hardness increased after HIPing and heat treatment to the T6 or T7 tempers for the wedge solidified at 10 atm pressure in the thickness range of 13 to 30 mm compared with the same thickness, non HIPed samples, (compare Table 6 with Table 4). The slight decrease in the YS was probably due to coarsening of the Al3(Sc,Zr) nano-particles during HIPing, while the increase in the UTS was most likely associated with the strain hardening and higher ductility of the HIPed samples. Table 6. Tensile properties of samples extracted from cast wedge regions of different thicknesses solidified at 10 atm pressure then HIPed and heat treated to the T6 and T7 tempers. Elongation UTS YS Heat Treatment Wedge Thickness BHN (MPa) (%) (MPa) (mm) ID 506 612 7.7 165 14.5 T6-1 500 604 28.5 T6-1 11.1 160 436 513 39.5 T6-1 2.0 136 298 313 50.5 T6-1 0.4 131 549 7.3 T7 514 14.5 585 6.7 515 T7 28.5 487 1.1 39.5 T7 439 357 1.0 344 50.5 T7

205

CONCLUSIONS 1. Ultra-high strength aluminum alloy castings have been produced using the combination of solidification under pressure, bonded sand molds and Sc-containing aluminum alloy 7042. 2. Hydrostatic pressure applied during solidification significantly reduced the amount of porosity. 3. Hydrostatic pressure applied during solidification may have resulted in a reduction of eutectic segregation at the end of solidification. 4. Hot isostatic pressing slightly improved strength and ductility by further reduction of the amount of closed porosity. ACKNOWLEDGEMENTS The authors would like to acknowledge the UAB metals casting group for producing the castings evaluated in this study. Work at the Air Force Research Laboratory was supported through the through the Air Force Contract No. FA8650-10-D-5226. REFERENCES 1. J.G. Kaufman and E.L. Rooy, Aluminum Alloy Castings: Properties, Processes, and Applications. ASM International, Materials Park, OH, USA, 2004. 2. D.G. Eskin, Suyitno, L. Katgerman, Prog. Mater. Sci., 49 (2004) 629-711. 3. M. Rappaz, J.M. Drezet, and M. Gremaud: Metall. Mater. Trans. A, 30A (1999) 449 - 455. 4. O.N. Senkov, S.V. Senkova, M.G. Mendiratta, D.B. Miracle, Y.V. Milman, D.V. Lotsko and A.I. Sirko, High Strength Aluminum Alloy Composition, US Patent #7060139, 13 June 2006. 5. O.N. Senkov, Advanced Aluminum Materials for Rocket Turbopump Rotors, SBIR II Final Report, AFRL-PR-ED-TR-2006-0073, Air Force Research Laboratory, Edwards AFB, CA, 2006, 234 p.; SBIR Phase II Special Report, AFRL-PR-ED-TR-2006-0074, Air Force Research Laboratory, Edwards AFB, CA, 2006, 156 p. 6. ASTM E8-04, "Standard Test Methods for Tension Testing of Metallic Materials," Annual Book of ASTM Standards 2005, Vol. 3.01, ASTM International (2005) pp. 62-85. 7. ASTM B557M-10, "Standard Test Methods of Tension Testing Wrought and Cast Aluminum- and Magnesium- Alloy Products [Metric]," ASTM International, West Conshohocken, PA, 2010, DOI: 10.1520/B0557M-10, www.astm.org.

206

Shape Casting: The 4lh International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

RELATIONSHIP BETWEEN STRUCTURE AND PROPERTIES OF AL-CU ALLOYS Alicia E. Ares 1 ' 2 , Liliana M. Gassa ' · 3 , Carlos E. Schvezov1,2, Sergio F. Gueijman2 1 Member of CIC of the National Research Council (CONICET) of Argentina. National University of Misiones; 1552 Azara Street, Posadas, Misiones, 3300 Argentina. Physicochemical Theoretical and Applied Research Institute (INIFTA). National University of La Plata. Diagonal 113 y 64, La Plata. Argentina. 2

3

Keywords: columnar-to-equiaxed transition, Al-Cu alloys, structural and corrosion parameters. Abstract The objective of the present research consists on studying the type of structure (columnar, equiaxed or with columnar to equiaxed transition, CET) using parameters of the solidification process and electrochemical parameters in Al-Cu alloys with different concentrations. In order to obtain columnar, equiaxed and CET structures, the alloys were directionally solidified upwards in an experimental set up with a set of thermocouples in the samples which permit to determine the time dependent profiles of temperature during the process. The electrochemical studies of the samples were realized by using an electrochemical impedance spectroscopy (EIS) technique and potentio-dynamic polarization curves immersed in 3% NaCl solution at room temperature. In general, we observed that the susceptibility to corrosion of the different structures depends on the size of the secondary dendritic spacing and proportion of A^Cu phase and Al-rich phase. Introduction Aluminum and its alloys are characterized by low density (2.7 g/cm3), high electrical and thermal conductivity, and good resistance to corrosion in certain media such as air. Also, the use of shape casting processes allows achieving good mechanical strength at low price. The mechanical strength alloy is achieved. However, these processes generally lower resistance to corrosion. Commercial Aluminum alloys often contain Cu, Mn, Mg, Si, Zn and Li in varying proportions between 0.1 and 10 %. These alloys are widely used in components of transportation due to fuel savings associated with weight reduction benefits similar security without compromising the safety of the occupants. Some of most widely used aluminum alloys are those containing between 4 and 10 wt.% Cu. Metallurgical factors that may affect the corrosion of an alloy include: crystallography, size, shape and heterogeneity of the grain, impurities, inclusions and residual stresses. Alloys (Al-Cu) have a high tensile strength and they are used in aircraft structural parts, car and bus bodies, and fuel tanks. Particularly Duralumin (96% Al 4% Cu) alloy is widely used in doors and windows [1]. Osorio et al. [3] studied the effect of macrosegregation and secondary dendritic arm spacing on the corrosion resistance of alloys Al-4.5%Cu. He found one major factor influencing the corrosion resistance of this alloy: the difference in corrosion potential between the Al-rich phase and the intermetallic particles Al2Cu, pointing that the formation of these particles is affected by the cooling rate imposed during solidification [2].

207

In another study, they analyzed the corrosion resistance in alloys Al-5%Cu and Al-8%Cu, directionally solidified with CET. The same authors found that both columnar and equiaxed morphologies yield similar results in trials of experimental EIS and polarization curves. In addition, they obtained decreasing corrosion resistance for increasing contents of Cu, and they suggest that this behavior may be associated with a decrease in the secondary dendritic arm spacing [3], In this study we characterize the corrosion resistance of Al-Cu alloys directionally solidified in a wide range of concentrations of Al-2wt.%Cu, Al-4wt.%Cu, Al-10wt.%Cu, Al-20wt.%Cu and Al-33.2wt.%Cu (eutectic). Directionally solidified samples present three different grain structures (columnar, columnar-to-equiaxed transition (CET) and equiaxed structure). Experimental Procedure All the alloys (Al-2% Cu, Al-4% Cu, Al-10% Cu, Al-20% Cu and Al-33.2% Cu, wt.%) were directionally solidified in an alumina mould placed inside a furnace equipped with a directional heat extraction system (details of the experimental device can be found elsewhere [4]. After directional solidification, the specimens were cut longitudinally and grounded using SiC papers of different grain sizes, from grade 60 to grade 1200, and then polished with diamond paste of 1 micron. The etching was performed with 15 ml HF, 4.5 ml HNO3, 9.0 ml HC1 and 271.5 ml H 2 0 at room temperature (25 CC), as recommended elsewhere to reveal macrostructure [4]. The position of the CET area was identified by visual observation and optical microscopy, and the distance from the base of the samples was measured with a ruler. These distances were found to vary between 5 mm and 73 mm from the bottom. In Figure 1 (a) shows the macrograph of an experience with Al-10%Cu, showing the three different areas of the specimen, columnar (bottom), CET (middle) and equiaxed (above). To analyze the microstructure, the samples were etched with a solution containing 1 g NaOH in 100 ml of distilled H2O, during a time of 5 to 15 seconds [4]. The macrographs and micrographs of different areas of the specimen of Al-10 wt.% Cu alloy sample can be seen in Figures 1 (b) to 1 (d) and in Figures (e) to (g), respectively. Equiaxed grain size was measured according to ASTM standard El 12 [1]. To determine the size of equiaxed grains used a typical frequency histogram of equiaxed grain size for each of the 10 mm intervals in which the specimen was divided. From these histograms we determined the equiaxed grain size [5]. Similarly, the width and length of the columnar grains was directly measured. Dendritic arm spacing measurements were made using the linear interception technique, preferably in regions close to the positions of the thermocouples during directional solidification in order to make correlations with the parameters of solidification readings of the aforementioned sensors. In order to perform the electrical tests, three samples were obtained for each concentration, one for each of the three basic macrostructures (columnar, transition and equiaxed). Later they were sanded to 1200 SiC particle size, washed in distilled water and dried by natural air and dried by natural flow. All work electrodes were positioned first longitudinally with respect to the counter electrode (Figures 1 (b-g) then transversally to the counter electrode (Figures 1 (h) and (i)).

208

All the electrochemical tests were conducted in a 3% NaCl 300 ml solution at room temperature using an IM6d ZAHNER® electrik potentiostat coupled to a frequency analyzer system, a glass corrosion cell kit with a platinum counter electrode and a sutured calomel reference electrode (SCE), see Figures 1 (j) and (k) show the detail of the experimental arrangement. Polarization curves were obtained using a scanning rate in the range of 0.002 V/s < v < - 0.250 V/s from open circuit potential until to 0.250 V. Impedance spectrums were registered in the frequency range of 10"3 Hz < f < 105 Hz in open circuit.

Figure 1. (a) Macrostructure of Al-10wt.%Cu alloy, (b-d) Macrostructures of the longitudinal work electrodes corresponding to each area, (e-g) Microstructure of the longitudinal work electrodes corresponding to each area, (h-i) Macrostructure and microstructure of the columnar transversal work electrode (j-k) Details of the experimental device. Results and Discussion Grain size (Gs) The values of the grain size measurements for the Al-10wt.%Cu sample are shown in Figure 2, as a function of position in the solidified specimens.

209

As shown, the equiaxed grain size varies from about 1 mm in the CET region and increases to 4.2 mm in the equiaxed region at the end of the specimen, which is also the last part to solidify. In the case of the columnar grain the width varies from 1.2 mm at the base to 4.7 mm at the CET area. Secondary dendritic arm spacing (λ?) Measurements of the secondary dendritic arm spacings (SDAS) include active and inactive branches. In the Figure 3 shows, as expected, a clear increase of local SADS with the distance to the bottom.

Figure 2. Variation of grain size versus the distance from the base of the sample for the Al-1 Owt%Cu alloy. The dashed lines indicated the CET area.

Figure 3. Secondary dendritic arm spacing variation versus the distance from the base of the sample. Al-10%Cu alloy,

Corrosion resistance Figure 4 (a), (b) and (c) shows the current density versus potential (E /1) response of alloys with different amounts of Copper and structures for the work electrode positioned longitudinally respect to the counter electrode. The corrosion reaction being an electrochemical reaction, then, the current value is a direct measure of the rate of dissolution of the alloy. Therefore, only in the case of the equiaxed structure it is observed that as the concentration of Copper increases the rate of dissolution of the alloy decreases. While the shape of the voltammograms is similar for each alloy and structure studied this trend with increasing of Copper concentration is not the same for the other two structures (columnar and CET), where although the alloy which increased flows of solution it is the lowest Copper content (Al-2wt.%Cu), this is followed by the alloy with 20wt.% Copper, indicating that there would be a combination strength / structure differently affects the corrosion resistance of each alloy. Moreover if we compare the current values of the different structures is the columnar which has increased flows of dissolution and equiaxed structure of lower values, regardless of the concentration of the alloy. A similar response was obtained in the transverse specimens in terms of higher current values in the alloy of lower Copper content, but there is no clear correlation between the dissolution process and the Copper content in the alloy (Figure 4 (d)).

210

In all cases, the E / I response of the alloys shows the typical hysteresis indicates the phenomenon of pitting, found that the more susceptible alloys (susceptibility measured as the difference between the potential of pitting, Ep, and repassivation potential, Er) are the lower Copper content in all its structures. The least susceptible are those containing 10wt.% Copper in its composition, see Figure 5. Figure 6 shows the Nyquist diagrams obtained for different Al-Cu alloys tested, we could observe that they are virtually just lines depart from the imaginary axis, which indicate that the alloy forms a thick oxide layer. This type of response is repeated for all structures, as can be seen in Figures 6 (a), (b) and (c). The impedance results correspond to a simple electrical circuit characterized by only one time constant suggesting the existence of the surface oxide upon the surface. The values of these parameters are 3±1 μΡ. cm"2 and 9±1 x 10" Ω cm2, typical for oxides formed on Aluminum and its alloys.

(c) (d) Figure 4. Voltammograms corresponding to Al-Cu alloy work electrodes positioned longitudinally (L) with respect to the counter electrode: (a) columnar, (b) CET and (c) equiaxed. In the case of figure (d) the work electrode is position transversally with respect to the counter electrode.

211

ALLOY

Figure 5. Pitting potential depending on the concentration and structure of the Al-Cu alloy.

(a)

(b)

(c) Figure 6. Nyquist diagrams corresponding to five Al-Cu alloys in different areas, (a) Columnar (b) CET and (c) Equiaxed.

212

According to the Al-Cu phase diagram system (see Figure 7), the microstructure of an hypoeutectic alloy consists on dendrites in Al matrix (a-phase) surrounded by an interdendritic region (eutectic phase). The eutectic is formed by aE - phase and intermetallic Al2Cu with alternating lamellar each. In the alloys under study and following the Al-Cu phase diagram, the interdendritic region is increased from the alloy with lower percentage of Cu up to 100wt.% concentration eutectic, see micrographs in Figure 8.

Figure 7. Al-Cu phase diagram [6].

(a) (b) (c) (d) (e) Figure 8. Microstructures: (a-b) With concentration less than 5.65wt.%Cu: (a) Al-2wt.%Cu and (b) Al-4wt.%Cu, in both it is observed the a-phase with particles of Al2Cu. (c-d) With concentration more than 5.65wt.%Cu: (c) Al-10wt.%Cu and (d) Al-20wt.%Cu, in these samples it is possible to assess the white dendritic phase (a-phase) and the dark interdendritic phase (αε phase + intermetallic Al2Cu). (e) With eutectic concentration: Al-33.2wt.%Cu (ae phase+ intermetallic Al2Cu). Analyzing the variation of the secondary dendritic spacing with the concentration of Copper in the alloy (for concentrations between 2wt.%Cu and 20wt.%Cu) and for the three structures (columnar, equiaxed and CET) it is possible to assess in Figure 9 that λ2 decreases with increasing Cu% until 10wt.% and then increases again to the 20wt.% of Cu. This was verified for the three types of structures and is consistent with what the literature reported for the same alloy system [1]. In Figure 5 it was found that Al-Cu alloys containing 10wt.% Copper are the less susceptible to corrosion. This behaviour is similar to that followed by the λ2 with increasing Cu concentration. For these experiments, there was no correlation of the electrochemical behavior of these alloys with grain size variation as a function of % Cu.

213

70 60

3·40 << 30 20 10

0 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20

WT. % COPPER Figure 9. Effect of Copper content on the secondary dendritic spacing. Conclusions The main conclusions of this investigation are: 1. The corrosion susceptibility depends on the formation of a thick film of oxide as revealed by the EIS analysis. However, the presence of Copper in their composition makes them susceptible to pitting corrosion. 2. Both SDAS and grain size were found to be strongly related with the cooling rate. Besides SDAS was found to have a minimum for a Copper content of about 10 wt.%. 3. The most influencing microstructural variable for corrosion resistance has been found to be the SDAS. It has been found that that resistance decreases with SDAS. Acknowledgments This work was partially supported by CONICET (National Research Council, Argentina). References 1. H. E. Boyer and T.L. Gall, Metals Handbook, (Metals Park, Ohio: American Society for Metals, 1984), 35-18-35-19. 2. W.R. Osório, J.E. Spinelli, I.L. Ferreira and A. Garcia, "The rol of macrosegregation and of dendritic array spacings on the electrochemical behavior of an Al-4.5wt%Cu alloy", Electrochimica Ada, 52 (2007), 3265-3273. 3. W.R. Osório, J.E. Spinelli, I.L. Ferreira and A. Garcia, "Experimental analysis of corrosion resistance on columnar to equiaxed transition region of as cast structures of Al-Cu alloys", Materials Science and Technology, 24 (2008), 1433-1437. 4. G. F. Vander Voort, Metallographic Principles and Practice, (Metals Park, Ohio: ASM International, 2001) 525-661. 5. A.E. Ares, S.F. Gueijman, C.E. Schvezov, "An experimental investigation of the columnar-to-equiaxed grain transition in aluminum hypoeutectic and eutectic alloys", Journal of crystal Growth, 312 (2010), 2154-2170. 6. R. F. Speyer, Thermal Analysis of Materials, (New York, NY: Marcel Dekker Editor, 1994) 30-109.

214

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

MICROSTRUCTURE CHARACTERIZATION OF MAGNESIUM CONTROL ARM CASTINGS Liang Wang1, Ratessiea Lett1'2, Sergio D. Felicelli'·2', John T. Berry2 1. Center for Advanced Vehicular Systems, Mississippi State University Mississippi State, MS 39762 2. Mechanical Engineering Department, Mississippi State University Mississippi State, MS 39762 Keywords: Magnesium, Casting, Microstructure, AZ91, Control arm Abstract Microstructural and mechanical property data were generated from several control arm castings of Mg alloy AZ91 produced for the High Integrity Magnesium Automotive Components (HIMAC) project. The castings were made by four different processes: squeeze cast, low pressure permanent mold, T-Mag, and Ablation. Ten control arms were examined from each of the four casting groups. The microstructure, grain size, pore fraction, and pore size were measured with optical microscopy and image analyzer. Different types of defects are identified to evaluate the four casting processes. In order to explore the presence of oxide films, a series of four-point bend (FPB) tests were performed, and the ultimate bending stress (UBS) was obtained. The mechanical properties of the castings were quantitatively evaluated for reliability using a two-parameter Weilbull distribution function. A detailed metallographic analysis of the fracture surfaces of FPB samples was performed using SEM. This project was sponsored by the United States Automotive Materials Partnership (USAMP). Introduction The magnesium alloy, AZ91, has been used in many automotive applications because of its lightweight, excellent castability and high fluidity [1]. Most of the magnesium automotive parts are produced by high pressure die casting (HPDC) process. However, due to the high filling speeds in HPDC, the melt flow is turbulent and air can be entrapped causing porosity when the melt solidifies [2]. The formation of porosity is one of the primary detrimental factors controlling fatigue lifetime and total elongation in cast light alloy components [3]. Many efforts have been devoted to investigate the mechanisms of porosity formation in the last 20 years. More recently, new mechanisms of pore formation based on entrainment of oxide films during the filling of aluminum alloy castings have been identified and documented [4-9]. Oxide film defects may be contained in most reactive liquid metals such as Al and Mg due to surface turbulence during the melting, pouring and transfer processes in casting. These defects have been observed on the fracture surfaces of tensile test specimens and the oxides have been identified by SEM-EDX analysis [8-10]. In contrast with the efforts devoted to Al-based cast alloys, few studies have been done in Mg alloy castings. Griffiths and Lai [10] investigated the nature of the oxide film defects in pure Mg castings. They found double oxide film defects comprised of folded MgO films on the fracture surface of tensile test bars taken from the castings. The objective of this research is to investigate the existing casting processes and develop new casting technologies for magnesium alloys and explore their capabilities in meeting the current

215

and future engineering challenges required by the automotive industry. We have investigated new casting processes by production of a passenger car magnesium control arm by squeeze cast, low pressure permanent mold (LPPM), and two new emerging casting processes - T-Mag and Ablation. The T-Mag process was invented by the Commonwealth Scientific and Industrial Research Organization (CSIRO), which is the national government body for scientific research in Australia. T-Mag is a permanent mold casting process for magnesium, which can fill the die smoothly from the bottom up. This process can minimize air entrapment and oxide generation, which leads to superior mechanical properties for the Mg castings. Ablation is a process that removes the aggregate mold with water, which is sprayed so as to ablate away the mold, allowing the water to impinge directly on the casting. This technique can easily remove the mold and achieve high cooling rates within the casting [11]. In this project, all control arms, produced by the four casting processes, were provided by various industrial partners. The microstructure characterization and mechanical properties for the samples extracted from the control arms were analyzed. Twelve samples were cut from one control arm of each casting process for the microstructure characterization, including grain size, area percentage of porosity, porosity size distribution, and defect analysis. Four point bend (FPB) tests are useful in determining the flexural properties in simply supported beams. Four point bending is most advantageous in that the maximum bending stress is uniformly distributed between the two top loading points, where as in Three Point Bending, the maximum stress is located directly under the top load [12]. This is important for the samples taken from the casting region, as there may be other defects within the center span aiding fracture. In this study, FPB tests were performed for each casting process. Separate sets of samples were cut from two locations of the control arm representing different thicknesses, with sample locations and dimensions being the same for all control arms. The ultimate bending stress (UBS) was plotted with Weibull statistics method and the casting process with superior mechanical properties was identified. The fracture surfaces were examined with Scanning Electron Microscopy (SEM) to investigate the oxide film and porosity effects on the FPB tests. Experimental Procedures Microstructure Characterization One control arm from each casting group was used to examine the microstructure characterization. Twelve samples were extracted from each control arm. The porosity and grain size distribution were measured for each sample. Significant defects such as shrinkage, cracklikes, gas pores were also identified. All samples were mounted and polished to examine the porosity distribution. The microstructure and grain size were examined after etching. Two images from the typical area of each sample were selected for the porosity analysis. The image analyzer was used to characterize the porosity distribution, including the pore size distribution, average area percentage of porosity, near neighbor distance, largest pore size. Four Point Bend (FPB) test Four Point Bend (FPB) test specimens with dimensions of 70 mm long, 16 mm wide, and 3.5 mm thick, were cut from locations C and D of each control arm (see Figure 1) to characterize the casting mechanical properties. These specimens were tested using an EM Model 5869 Instron machine at a cross-head speed of 3mm/min for the FPB test. Figure 2 depicts the sample arrangement employed in the FPB test.

216

Figure 1. Magnesium control arm illustrating locations of C and D used for FPB test

Figure 2. Sketch of four point bending test geometry.

The initial test apparatus used was a smaller bend test setup. However, as the size of this setup was a limiting factor, a larger test setup was used to test samples from processes that did not fracture before "bottoming out" on the small test setup. This included Ablation samples from locations C and D and T-Mag samples from location D. However, although the larger test setup was used, fracture of the Ablation and T-Mag samples previously referenced still did not occur. A major concern for the samples during the testing in the larger setup was slippage of the ends and contact with the setup in areas other than at the force locations which may cause indentation. The test was manually stopped to prevent further slippage of the sample off of the lower forces and to prevent sample contact with the cornered edges. We could not break any Ablation samples with neither the small nor the large FPB settings, which indicates that the Ablation castings show significantly higher ductility that the other processes. Several Ablation bent samples were broken manually in order to investigate the fracture surfaces. Weibull Analysis The Weibull analysis approach, proposed by Weibull [13], has been frequently used to characterize the dispersion of the mechanical properties. Using the symbol F for the ultimate bending stress (UBS) for the four point bend (FPB) test, the Weibull distribution can be represented using the following form of the Weibull equation [10]

In In

1 i-/%y

(1)

= m\n Fn - min F0

This equation is in the form y = mx + c; therefore, the Weibull modulus m can be determined from the slope of a plot of ln(ln(l/(l-/^. n ))) against lnF„, where Pfn is a measure of the probability of failure. This can be estimated by a commonly applied method: Pf„ =

"-^

(2)

In this equation, n is the number of the result when all results are ordered in an ascending fashion, and N is the total number of results. The scale parameter represents the failure stress below which 63.2 pet of the samples failed. This value is denoted F0 and was obtained from the

217

intercept on the y-axis of the straight line used to determine the Weibull modulus (m). Since, from Eq. (1), y = -m\nF0 a t * = 0,then F0 = exp(.y/-m). The usefulness of this analysis is that the Weibull modulus m is a single value that shows the spread of properties, in this case, the ultimate bending stress (UBS) of the FPB test. Higher Weibull modulus shows a narrower spread of properties, which indicates that the casting process is associated with low numbers of defects in the final casting and a greater reproducibility of properties. In this work the Weibull method was used to analyze the FPB test results for the magnesium control arm samples to evaluate each casting processes. Results and Discussion Micro-porosity was observed in the tested specimens, ranging in size from 5 to 20μιη. Most of pores have the aspect ratio of approximately 0.5. The near neighbor distance of the porosity is in the range of 30 to 70pm depending on the casting groups. Table 1 lists the average porosity analysis results for each casting process. The average area percentage of porosity is in the range of 0.1 - 2.0 wt% for all samples from different casting processes. It can be seen that the gas porosity is not major defect for all castings. The grain sizes for each sample for different casting groups were measured and the results show that larger grains were found in the T-Mag cast (300 ± 50μηι), compared with LPPM (100 ± 20μπι), Ablation (100 ± 30μηι), and squeeze cast (20 ± ΙΟμπι). It would be expected that the Ablation casting shows the smallest grain size due to the high cooling rates. The control arms were heat treated and hence the results do not reflect as cast conditions. Besides micro-porosity, different types of defects were identified including sponge shrinkage, gas pores, and crack-likes. In general, all four casting groups present some sponge shrinkage. Gas pores with large size diameters were also found in some samples for each casting group. The long crack-likes were found only in the squeeze casting samples. Figure 3 shows the typical metallographic images for each casting process. The large gas pore was found in the Squeeze cast (Figure 3(a)), and the sponge shrinkage defects were found in the LPPM (Figure 3(b)) and T-Mag samples (Figure 3(c)). Separated irregular pores were found in the Ablation samples (Figure 3(d)). Casting Groups

No. per mm2

Squeeze LPPM T-Mag Ablation

34.95 76.75 69.16 194.52

Table 1 Porosity analysis results for each casting group Avg. Avg. min Aspect Avg. Near Avg. max size diameter ratio neighbor diameter 2 (μηι ) (min/max) dist. (urn) (μιη) (Mm) 132.94 8.37 15.84 0.53 71.53 84.10 7.06 15.31 0.47 65 69 76.18 7.07 12.43 0.57 50.33 41.65 5.52 9.64 0.58 38.34

Avg. Area percentage

(%)

0.44 0.84 0.49 0.83

The details of fracture surfaces of four point bend (FPB) tests for the specimens from each casting process are investigated through SEM. Figure 4 shows typical oxide film defects from the fracture surfaces of the FPB samples taken from the control arms produced by the squeeze cast and LPPM. They were observed to have a filmlike structure, which a folded and wrinkled appearance was found in the fracture surfaces (Figure 4(a) and (b)). No such features were observed for the T-Mag and ablation processes in this study (Figure 4(c) and (d)).

218

Figure 3 Typical metallographic images for four different casting processes to show différent defects, (a) Squeeze Cast, (b) LPPM, (c) T-Mag, (d) Ablation.

219

Figure 4 Casting defects on the fracture surfaces of FPB test specimens taken from the control arms produced by different casting processes, (a) Squeeze casting, (b) LPPM, (c) T-Mag, (d) Ablation Table 2. UBS for FPB tests in an ascending order for each casting process (MPa) Samples Squeeze Casting LPPM T-Mag 273.82 199.46 313.67 1 274.75 241.64 323.86 2 289.93 252.45 328.77 3 291.00 266.66 332.20 4 296.66 271.70 350.37 5 297.09 273.81 6 353.72 308.31 295.04 356.15 7 309.68 302.16 359.99 8 310.59 305.21 366.92 9 314.30 306.96 377.40 10 315.21 311.43 378.30 11 322.75 314.59 392.74 12 326.50 341.97 13 327.07 343.56 14 328.75 348.23 15 330.02 357.96 16 331.65 17 340.09 18 348.57 19 348.95 20 353.83 21 355.74 22 356.48 23 357.85 24 362.78 25 363.33 26 368.50 27 372.55 28 399.72 29 401.14 30 405.37 31 334.93 295.80 352.84 Mean

220

The ultimate bending stress (UBS) for FPB tests, which were obtained from the measured load and sample dimensions, with the data arranged in an ascending order for each casting process are shown in Table 2, while the Weibull plots obtained from these data are shown in Figure 5. The Weibull parameters have also been summarized in Table 3. Since none of the Ablation samples was broken by the FPB tests, the Ablation process is not included in the Weibull plots. The Weibull moduli for the UBS are 7.93, 11.54, and 17.29 for Squeeze Cast, LPPM, and T-Mag, respectively. Dispersed shrinkage porosity and oxide film defects were observed in the Squeeze Cast samples, which in turn results in more scatter in FPB test results than in LPPM and T-Mag samples. The mean values of the UBS are 295.80MPa, 334.93MPa, and 352.84MPa, for Squeeze Cast, LPPM, and T-Mag, respectively. For the Ablation process, an average UBS level of 536.35MPa was reached before the sample bottomed out. These results indicate that the T-Mag and Ablation castings have better mechanical properties than Squeeze Cast and LPPM.

Figure 5 Two parameter Weibull plot for UBS data of FPB tests for different casting processes. Table 3. Summary of Weibull Analysis of FPB UBS results for each casting process Casting Process Weibull Distribution Weibull Scale Parameter Mean value of UBS Equation Modulus (MPa) (MPa) Squeeze Cast y = 7.93x-45.59 7.93 307 295.80 LPPM y=11.54x-67.60 11.54 348 334.93 T-Mag y = 17.29x-101.94 17.29 358 352.84 Conclusions The microstructure characterization and mechanical properties were evaluated for four different casting processes: Squeeze Cast, Low Pressure Permanent Mold (LPPM), T-Mag, and Ablation. Control arms produced by each casting process were provided by different industrial partners. Samples were extracted from the control arms to examine the microstructure, porosity size distribution, and grain size. Four point bend testing was performed to investigate the mechanical properties and a Weibull analysis was used to quantify specimen failure rate. The four casting processes were evaluated and the following conclusions were drawn: (a) T-Mag castings present better mechanical properties than Squeeze Cast and LPPM according to the Weibull analysis of FPB ultimate bending stress (UBS) results; (b) Ablation castings present significant higher ductility than other processes since no Ablation sample could be broken by FPB tests; (c) The largest average grain size was found in the T-Mag casting; fine dendrites were found in the

221

squeeze casting. Both LPPM and Ablation have similar grain size distribution; (d) Five types of casting defects were identified including micro-porosity, oxide films, sponge shrinkage, gas pores, and crack-likes; (e) Significant shrinkage porosity and oxide film defects were observed in the Squeeze Cast and LPPM castings. Acknowledgements This work was sponsored by the United States Automotive Materials Partnership (USAMP) under the United States Council for Automotive Research (USCAR). The assistance of Mr. Don Penrod of Manufacturing Services and Development Inc. is appreciated. The authors gratefully acknowledge valuable discussions with Professor Elborn Jones and the technical assistance of Mr. Jacob Coleman and Mr. Wilburn Ray Whittington at Mississippi State University. References [I] B.L. Mordike, and T. Ebert, "Magnesium: Properties - Applications - Potential," Materials Science and Engineering A, 302 (1) (2001), 37-45. [2] C. Potzies, K.U. Kainer, "Fatigue of Die-cast Magnesium Alloys," Magnesium Technology, Edited by Alan A. Luo, TMS (The Minerals, Metals & Materials Society), (2004), 275-278. [3] L. Wang, H. Rhee, S.D. Felicelli, A.S. Sabau, and J.T. Berry, "Oxide Film and Porosity Defects in Magnesium Alloy AZ91," Shape Casting: The 3 rd International Symposium, eds. J. Campbell, P.N. Crepeau and M. Tiryakioglu, TMS, Warrendale, PA, (2009), 123-130. [4] John Campbell, Castings Td ed. (Butterworth-Heinemann, London, 2003), 17-69. [5] X. Yang, X. Huang, X. Dai, J. Campbell, and R. J. Grant, "Quantitative Characterization of Correlations between Casting Defects and Mechanical Strength of Al-7Si-Mg Alloy Castings," Materials Science & Technology, 22 (2006), 561-570. [6] J. Campbell, "Entrainment Defects," Materials Science & Technology, 22 (2) (2006), 127145. [7] J. Knott, P.R. Beeley, J.R. Griffiths, N.R. Green, C.J. Newton, and J. Campbell, "Commentaries on 'Entrainment Defects' by J. Campbell," Materials Science & Technology, 22 (2006), 999-1008. [8] R. Raiszadeh, and W.D. Griffiths, "A Method to Study the History of a Double Oxide Film Defect in Liquid Aluminum Alloys," Metallurgical and Materials Transactions. B, 37 (2006), 865-871. [9] J. Mi, R.A. Harding, M. Wickins, and J. Campbell, "Entrained Oxide Films in TiAl Castings," Intermetallics, 11 (2003), 377-385. [10] W.D. Griffiths, and N.W. Lai, "Double Oxide Film Defects in Cast Magnesium Alloy," Metallurgical and materials transactions A, 38 (2007), 190-196. [II] J. Grassi, J. Campbell, M. Hartlieb, and F. Major, "The Ablation Casting Process," Materials Science Forum, 618-619 (2009), 591-594. [12] "Standard Test Method for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials by Four-Point Bending", (ASTM International, v. 08.03, D627202), 513-518 [13] W. Weibull, "A Statistical Distribution of Wide Applicability", Journal of Applied Mechanics, 18(1951)293.

222

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry

Methods and Systems Session Chairs: Alan Druschitz Derya Dispinar

Shape Casting: The 4lh International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

THE EFFECT OF REDUCED MOLECULAR W E I G H T OF THE PATTERN ON THE PROPERTIES OF Al ALLOY CASTINGS MADE BY THE LOST FOAM CASTING PROCESS 'Κ. SIAVASHI, 2C. TOPPING and 'W. D. GRIFFITHS 1. School of Metallurgy and Materials, College of Engineering and Physical Sciences, University of Birmingham, Birmingham, United Kingdom. B15 2TT 2. Isotron, Brunei Close, Daventry, United Kingdom. NN11 8RB Keywords: Aluminium alloys, lost foam casting, mechanical properties, y radiation, molecular weight.

ABSTRACT In the Lost Foam casting of Al alloys, heat from the cast liquid metal causes the foam pattern to degrade and results in the evolution of gas and formation of liquid polymer byproducts. These can cause a reduction in the casting quality if they become entrapped in the liquid metal. However, the liquid polymer byproducts can be absorbed by the permeable pattern coating, once their molecular weight (Mw) is sufficiently reduced by the action of heat. Therefore using a lower molecular weight (Mw) pattern may lead to higher quality castings because less reduction in Mw will be required before absorption of the liquid polymer byproduct into the pattern coating. In this work the Mw of expanded copolymer foam patterns has been be reduced by exposure to γ-radiation. The properties of castings made with these irradiated foam patterns, such as porosity content and fatigue properties, were compared with the properties of castings made from unirradiated foam, to show the advantages of using the former. Introduction In the Lost Foam casting process the pattern is constructed of polystyrene (PS) foam, coated with a permeable refractory layer, typically about 500 μηι in thickness. This is placed into a moulding box and surrounded by loose, unbonded silica sand, which is compacted by vibration. The mould is then cast, with the polystyrene (PS) pattern being degraded by the heat from the liquid metal, to form the desired cast shape. The process confers great freedom of design compared to conventional casting processes. Complex patterns can be made by joining together simpler shapes and cored features can be formed in situ, reducing machining costs. However, the degradation of the polystyrene pattern produces liquid and gaseous byproducts, and if these become entrapped in the liquid metal during mould filling, the quality of the final casting can be considerably reduced. In the case of casting of Al alloys, significant effort has been put into studying the thermal decomposition of the foam pattern and the removal mechanisms of the degradation byproducts, for example, by Zhao et al. [1], Shivkumar et al. [2], Liu et al. [3], Caulk [4], Barone et al. [5], Molibog et al. [6], Sun et al. [7] and Hill et al. [8]. Previous research has suggested that an important feature of the process is the

225

permeable pattern coating, through which the pattern degradation products must pass. At the temperatures associated with casting Al alloys, the pattern tends to degrade to form a mostly liquid residue. It has been proposed that the permeable pattern coating absorbs the liquid polystyrene residue by a wicking action [7], although alternative suggestions have been put forward [1]. Detailed investigation suggested that once the liquid polystyrene residue degraded to reach a critical molecular weight (Mw), (and hence critical viscosity), absorption into the pattern coating occurs. For example, for a coating permeability of 1.2 x 10"12 m 2 , the polystyrene, with initial Mw of 325,000 gmol"1, is wicked into the coating once the liquid degradation byproduct reaches a Mw of about 70,000 gmol"1 [9]. This suggests that improved Lost Foam Al alloy castings could be obtained if the material used for the pattern possessed a lower initial molecular weight. The liquid polystyrene residue might then reach the critical Mw for wicking into the pattern coating more quickly, and be removed more easily. This would presumably reduce the likelihood of it being entrained in the casting, improving casting properties. Apart from polystyrene, foam patterns are also made out of a copolymer of PS and polymethylmethacrylate (PMMA). The Mw of the latter can be reduced by γ-radiation, (which does not have the same effect on polystyrene) [10]. The results presented in this paper show how using irradiated foam patterns can produce improved castings. Experimental Procedure Rectangular plates of a copolymer of 70wt.% PMMA and 30wt.% PS (PROBEAD-70) with a thickness of 10 mm, and length and width 180 x 450 mm, respectively, were irradiated by a cobalt-60 γ-ray source in order to reduce their Mw. The foam plates were exposed to dosages of up to about 190 MRad. In addition to this, the effect of foam type, (the copolymer or pure PMMA), and irradiation source, (γ-rays or an electron beam), were also compared. The resulting foam patterns were cast with an Al-Si-Mg alloy, and the quality of the resulting castings characterised. Another copolymer (PROBEAD-30, COPOL 1.3) (30wt.% PMMA and 70wt.% PS), was also exposed to an electron beam to reduce its Mw with dosages of up to about 180 MRad. The effect of irradiating the foam patterns was established by measuring their Mw and polydispersity using Gel Permeation Chromatography (GPC), carried out by Rapra Technology (Shrewsbury, UK). The results of the GPC measurements were expressed as 'polystyrene equivalent' Mw, rather than absolute values of Mw, but comparisons between results obtained within this work would still be valid. In order to ascertain the effect of γ-irradiation on the mechanical properties of the irradiated foams, a 3 point bending test was carried out on samples of dimensions 80 x 50 x 10 mm, subjected to a 30 kg load applied at 5 mm min"1. Foam strips were then cast of dimensions 10 x 40 x 300 mm, with a 0.3 mm layer of a high permeability coating (1.2 x 10"12 m 2 ) and cast horizontally with 2L99 alloy (Al7wt.%Si-0.3wt.%Mg) at 780°C and 150 mm head height. All castings filled completely.

226

To characterize the quality of the castings obtained, their porosity was measured by image analysis, carried out on polished samples taken from the centre line of the castings. Defects such as internal porosity and surface cavities are considered to be associated with the entrapment of liquid polymer degradation byproducts at the castingcoating interface. The internal porosity was characterised by measurement of the total porosity area, while the surface cavities, which were only found on the bottom casting surface, were characterised by measurement of their total length and frequency. To establish the effect of using reduced Mw foam patterns on the mechanical properties of lost foam Al alloy castings, a fatigue test was used. Test bars, of dimensions 10 x 40 x 300 mm, were taken from the centre line of the strip castings, and given a T6 heat treatment, (solutionised at 535°C for 12 hours, aged at 135°C for 6 hours). These samples were subjected to a high cycle fatigue test using 4 point bending, with maximum and minimum forces of 2.5 and 0.25 kN, frequency of 67 Hz, and loading span on the top and bottom of the specimens of 20 mm and 60 mm, respectively. The samples were placed in the fatigue test with their as-cast surfaces intact, arranged so that the base of the casting faced downwards. This meant that the surface containing the cavities were suspected of being associated with liquid polymer degradation byproducts trapped at the casting-coating interface, experienced the maximum stress. Results In Figure 1 the Mw of the foam pattern is plotted against the irradiation dosage received. The foam pattern (PROBEAD-70) had an original Mw of about 325,000 gmol" ', which was reduced according to the amount of irradiation received, to values as low as about 45,000 gmol"1 with the maximum γ-irradiation, (a dosage of 189 MRad). The other copolymer type, which is usually used for ferrous castings, (PROBEAD-30) also showed a reduction in its Mw due to the effect of the electron beam irradiation. The original Mw of the PROBEAD-30 was 271,000 gmol"1 and it was reduced to about 86,000 gmol"1 by exposure to 160 MRad.

Figure 1. The effect of γ-irradiation on Mw of different types of foam pattern.

227

As the Mw of the foam pattern was reduced by irradiation, the mechanical properties of the foam pattern decreased. The maximum load at fracture plotted against Mw of the foam pattern, has been shown in Figure 2. The maximum reduction of foam strength in the most irradiated foam pattern (189 MRad), was about 60%.

Figure 2. Results of 3 point bending tests on the irradiated foam patterns (PROBEAD70), showing that foam strength was reduced with reduction of its Mw by γ-irradiation. Figure 3 shows how the porosity content of the aluminium alloy castings made with the irradiated foam patterns was reduced as the Mw of the foam pattern was reduced. The porosity content of the casting made with an unirradiated foam pattern (Mw of 325,000 gmol"1) was about 1.6%, and was reduced to about 0.4% in the casting made with the most irradiated foam (Mw of about 45,000 gmol"1). In other words, the porosity content of the castings was decreased to about 25% of the original porosity, by irradiation of the foam patterns. A Fisher test confirmed that the porosity content of the castings made with unirradiated foam and with the most irradiated foam were statistically different at the 99% confidence limit.

Figure 3. Graph showing porosity content of the aluminium castings reduced by decreasing the Mw of the foam pattern (PROBE AD -70).

228

The castings showed defects (cavities) at the base of the cast strips, thought to be associated with globules of liquid polymer degradation byproduct trapped at the castingcoating interface. This is shown in Figure 4, where the surface cavities were associated with porosity immediately above them, indicative of gas released by the liquid polymer degradation product rising up through the liquid metal above and becoming trapped in the solidifying casting. Figure 5 shows the total length of defects found at the base of the horizontal casting strips decreased with decreasing Mw of the foam pattern used, in agreement with the reduction internal casting porosity.

Figure 4. Defects occurring at the base of the castings, (a) Casting made with unirradiated foam pattern, (b) and (c) castings made with irradiated foam patterns, (30 and 75 MRad respectively).

Figure 5. Graph showing the relationship between Mw of the foam patterns (PROBEAD-70) related to the total length of the defects found on the bottom surface of the horizontally cast plates. Figure 6 illustrates how, when the Mw of the initial foam pattern was lower than the critical value, (which was about 70,000 gmol"1 in the case of the high permeability coating used here), the fatigue properties of the castings were much higher than when the castings were made with the unirradiated foam pattern (with an initial Mw of 32500 gmol"1). The castings made with the foam pattern exposed to about 144 MRad of γ-

229

radiation, (Mw of 63,000 gmof1), had a fatigue life increased to nearly twice that of castings made with the conventional, unirradiated patterns. 0»

l

Ü2 υ > "1.5 c o

11 Φ

^0.5 Φ 3

cn « 0 0

U

50000 100000 150000 200000 250000 300000 350000 M w o f the Foam Pattern (gmol 1 )

Figure 6. Results of fatigue tests showing that fatigue properties of the heat treated alloy improved by irradiation of the foam patterns used in the casting process. The SEM results of the fracture surface of the specimens illustrated that the failure of the castings made with an irradiated foam pattern with Mw of 63,000 gmol" , occurred due to a small surface-breaking defect on the bottom surface of the casting (see Figure 7b). However the initiation of the fatigue failure of the casting made with an unirradiated foam pattern was due to a near-surface pore with a diameter 4 times greater than that of the former defect (Figure 7a).

Figure 7. SEM micrographs of fracture surfaces from the fatigue tests, (a) casting made with an unirradiated foam pattern (Mw of 325,000 gmol'1) and (b) casting made with an irradiated foam pattern (144 MRad, Mw of 63,000 gmol'1).

230

Discussion Irradiation of the copolymer foam patterns reduced the Mw significantly, and this was observed in the case of both copolymer types, containing 70% PMMA / 30% PS and 30% PMMA / 30% PS. The reduction in Mw, for any dose of γ-radiation, was greatest in the case of the 70% PMMA copolymer, than in the case of the 30% copolymer, (see Figure 1). Irradiation of pure polystyrene foam (PS) had no measurable effect on molecular weight. These results were obtained by γ-irradiation, but a faster alternative would be to use an E-beam irradiation method, where electrons are driven through the polymer foam structure instead of γ-irradiation. This is substantially faster, about 25 times, (and was used here to prepare the patterns used for cast iron Lost Foam casting). Irradiation of pure PMMA foam reduced the Mw significantly, (by up to 95%), i.e., the effect of irradiation increased in effectiveness with increasing amounts of PMMA in the foam patterns. However, the foam structure was obviously damaged by the radiation dose, with the pattern becoming friable and unusable for moulding. In the case of the copolymers, irradiation reduced mechanical properties, as shown by the 3-point bending test results in Figure 2, but the foam patterns were still readily useable. The effect of the γ-irradiation is to reduce molecular weight by chain scission. In the case of polystyrene, this chain scission is accompanied by cross-linking, which prevents any reduction in Mw; PMMA, on the other hand, has its molecular weight reduced progressively by increasing amounts of chain scission with increasing radiation [10,11]. Measurement of the porosity content of the Lost Foam aluminium castings showed that this was related to defects on the bottom surface of the horizontally cast flat strips, (see Figure 5). Reducing the initial Mw of the foam pattern to, or below, the critical value to cause wicking of the liquid polymer byproduct into the coating reduced casting porosity, (see Figure 3). Since the casting porosity is thought to be associated with the surface defects caused by entrapment of the liquid polymer byproducts at the castingcoating interface, it follows that reducing the initial Mw of the pattern reduces the extent of these surface defects, (for example, as shown in Figure 7), and this is reflected in the improved fatigue properties associated with irradiated foam patterns, Figure 6. Conclusions 1. The Mw of foam patterns containing polymethylmethacrylate (PMMA) was decreased by γ-irradiation and electron beam irradiation. 2. The reduction in Mw due to irradiation increased in effectiveness with increasing amounts of PMMA in the foam patterns. 3. Irradiating the copolymer foam patterns up to values of 189 MRad reduced their flexural modulus, but they could still be used in the casting process. However, when pure PMMA was given a 100 MRad dose, it became too friable to be used. 4. The porosity content of the castings was reduced by reducing the Mw of the foam patterns used in the casting process. 5. The defects at the bottom surface of the castings, due to the entrapment of the liquid polymer degradation byproducts at the casting-coating interface, were

231

shorter than in the case of conventional Lost Foam casting, when reduced Mw foam patterns were used. 6. The fatigue life of the castings was increased by reducing the Mw of the foam patterns used in the casting process, probably because the critical Mw for wicking the liquid pattern degradation byproduct into the permeable coating was more quickly reached. 7. Electron beam irradiation is the preferred method of reducing the pattern Mw, as it was found to be about 25 times quicker in delivering the same dosage as γradiation. Acknowledgements The authors would like to gratefully acknowledge the assistance of Mr. Adrian Caden of the University of Birmingham with the execution of the experiments, and Rapra Technology is thanked for carrying out the Gel Permeation Chromatography analyses. References 1. Zhao Q., Burke J.T. and Gustafson T.W., Foam Removal Mechanism in Aluminium Lost Foam Casting, AFS Trans., 110, 2002, pp. 1399-1414. 2. Shivkumar S. and Gallois B., Physico-Chemical Aspects of the Full Mold Casting of Aluminum Alloys, Part I: The Degradation of Polystyrene, AFS Trans., 95, 1987, pp. 791-800. 3. Liu X.J., Bhavnani S.H. and Overfelt R.A., Simulation of EPS Foam Decomposition in the Lost Foam Casting Process, J. Mat. Proc. Technol., 182, 2007, pp. 333-342. 4. Caulk D.A., A Foam Melting Model for Lost Foam Casting of Aluminum, Int. J. Heat and Mass Trans., 49, 2006, pp. 2124-2136. 5. Barone M. and Caulk D., Analysis of Mold Filling in Lost Foam Casting of Aluminum: Method, Int. J. Metalcasting, 2008, 2, (part 3), 29-43. 6. Molibog T.V. and Littleton H., Degradation of Expanded Polystyrene Patterns, AFS Trans., 110, 2002, pp. 1483-1496. 7. Sun Y., Tsai H.L. and Askeland D.R., Investigation of Wetting and Wicking Properties of Refractory Coating in the EPC Process, AFS Trans., 1992, 100, pp. 297-308. 8. Hill M., Vrieze A,E., Moody TX., Ramsay C.W. and Askeland D;R., Effect of Metal Velocity on Defect Formation in Al LFCs, AFS Trans., 106, 1998, pp. 365-374. 9. Griffiths W.D. and Davies P.J., Wetting and Wicking of Liquid Polymer Degradation Byproducts into the Pattern Coating during Lost Foam Casting of Al Alloys, Int. J. Cast Met. Res., in press. 10. Hill, R., Imperial College, pers. comm., 2009. 11. Wilson, J.E., Radiation Chemistry of Monomers, Polymers and Plastics, Marcel Dekker, New York, (1974).

232

Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

CLASSICAL NONDESTRUCTIVE TESTING TECHNIQUES DO NOT CORRELATE WITH STRENGTH AS DOES PROCESS COMPENSATED RESONANT TESTING Authors: R. H. Nath'.C. C. Grupke2, C. Leonard3, and M. K. Johnson4 1

Magnaflux Quasar, Albuquerque, [email protected] Consultant, [email protected] 'Diversified Machine, Inc., Montague, MI, [email protected] 4 Retired Consultant, Albuquerque, NM, [email protected] 2

Keywords: Process Compensated Resonant Inspection, Classic Nondestructive Testing, Oxides, bifilms Abstract Process Compensated Resonant Inspection (PCRT) outperforms classical nondestructive testing (NDT) on aluminum parts based not just upon the significantly better results obtained from testing millions of automotive parts in production operations with achieved near-zero failures, but also insight gained from destructive testing and analysis. Several different destructive testing efforts were performed. Applicable classical NDT and PCRT were used. The PCRT relies on defining as "acceptable" those parts that are structurally acceptable, as measured by metallurgical, destructive and other nondestructive testing. Castings tested and accepted with classic NDT indications often have structural weakness and are accepted because their major defects are often the result of damaging thermal history or invisible oxide bifilms. Conversely, structurally acceptable castings are often rejected for negative classic NDT indications. Key previously assumed "facts" dealing with the parts and how they are accepted/rejected do not always stand up to more thorough destructive testing and analysis. Introduction The Process compensated Resonant Inspection (PCRT) to be discussed in the following is a technique derived from the Resonant Ultrasound Spectroscopy (RUS) work performed at the Los Alamos National Laboratory by Dr. Albert Migliori. It measures resonant frequency patterns that are directly dependent on the parts' geometry and, more importantly, its elastic properties that are directly related to structural integrity. PCRT is described in detail in an ASTM Standard Practice[l]. From 2004 until present (2010), various studies have been performed using aluminum automotive parts to compare PCRT with classic NDT. The studies started out with an hypothesis derived from both theory and significant testing of PCRT. It was clear to successful users of PCRT that to make it work properly (sort parts which were structurally deficient compared to those which were structurally sound) that it required both types of parts in the database. The classical NDT most often used for light metal parts (x-ray, penetrant, and ultrasound) simply

233

inferred by indications that there might be a crack or inclusion or some other form of defect that might structurally weaken the part. It is important to note that PCRT is the only inspection technique that can find the visually undetectable bifilm inclusions in light metals. During the first set of testing in 2004, aluminum automotive links were tested with both classic NDT and PCRT. Surprisingly, many of the break points and faces had no initiation point indication (as in a crack or a visible oxide inclusion below the surface), nor did the face show anything unusual under standard microscopic examination. It was clear that some seemingly invisible mechanism was at work. (The aluminum chemistry was well within bounds as was standard hardness testing). Had this been but one sample, it could have been called an anomaly and disregarded as such. However, this was not an anomaly. Rather it was pervasive. A number of experiments were undertaken using A356 aluminum parts with lessons learned in one being applied to successive ones. The following addresses results of these experiments. It also shows quite clearly that classic NDT techniques are not nearly as good at finding structurally weak parts as is PCRT. Comparison of Classic NDT to PCRT Links Tests This experiment was carried out on links made with exceedingly bad process control. The variation in the performance of these simple parts reflected very bad casting methods. The fact that the production output consisted of such highly variable parts made an excellent case study subject for comparison of quality assurance methods, however. But from a practical standpoint this link production should have been stopped until it was brought under control. There were two large-scale experiments that looked at the comparison of PCRT to classic NDT. One consisted of a series of experiments using different "links" (suspension links, etc.) and the other used knuckles. The following will concentrate first on the links, then on the knuckles. Initially, the thought was that a set of links may be characterized using un-machined, cast links and then correlate the results with breakage of machined links. Some links were made to be "bad" by adding oxides. The rest were passed by in-place factory, classic NDT. A break cutoff was provided for a pull test of 7000 pounds (31,116 N) because the OEM set that force as a minimum for these links. The sort was based on resonant data from un-machined links, and the classifications were performed on machined parts. PCRT was able to differentiate the low-forcebreaking links (less than 7000 pounds (31,116 N)) from the higher break force links. In doing so, however, a number of "good" links were also rejected- many that had broken at very high levels. In this effort, we conclude that the 7000-pound (31,116 N) break force limit reflects normal automotive industry practice of using a very high design factor (formerly called a safety factor). We also see that the current factory test methods, ultrasound and x-ray, already do falsely reject an unusually large number of links, but some parts that were installed on vehicles were failing. The results from the destructive tests on this original set of parts was considered to be flawed because "the castings cannot be that variable."

234

In 2005, three new sets of parts were procured. One hundred and fifty links were straight from the factory, having passed all the in-place NDT testing (ultrasound and x-ray). Another fifty links were selected because they were rejected by ultrasound NDT, and fifty more were selected because they had been rejected by x-ray NDT. Even though ultrasound and x-ray had passed all of the 150 normal factory shipped links, 18 of these "normal factory shipped" set of 150 failed below the 7000-pounds (31,116 N). Additionally, there were false reject rates of 80% and 86%, respectively, for the other link sets. Link Data and PCRT Analysis Early Link Tests The PCRT sort derived from the early break force analysis was based on the assumption that 7000-pounds (31,116 N) was a reasonable cutoff point for categorizing links as acceptable or unacceptable. Additionally, the data were taken on some links that had intentionally been made with large pieces of oxide in them, some of which broke at very high levels. So, it was to be expected that a number of the links would fail the PCRT test and pass the break test above the 7000 pounds (31,116 N) cutoff limit. Nearly all breaks occurred at the weakest structural area of the links, which is not surprising. Figure 1 illustrates how a machined link could have a large flaw in an area that is otherwise structurally robust (e.g., "b") and would not break at that point below the cutoff break force.

Figure 1. A small flaw in darker areas denoted by "a" will have a much more significant effect than a large flaw at (b) or (c). These (a) areas are also where almost all breakages occurred Many of the break faces showed no obvious signs of flaws. This held true under standard microscopic examination. Yet, the PCRT pattern analysis runs were still able to sort out these low failure level parts. There is a likely explanation that will be discussed later. Further Link Testing - As-Cast Results Resonant data were taken on the additional 250 links described above. A PCRT sort was made from the data analyzed for the earlier link study using as-cast data only. Each link was PCRT sorted in the as-cast condition, based on this early training data. These results were based on the 7000-pound (31,116 N) break force reject level. The links were then machined, and PCRT broadband data were taken again, this time in the machined condition. The links were then tensile force tested to failure. The results were compiled, compared with the early-based PCRT sort, and a new pattern analysis was performed based on the PCRT data taken on the machined links. These data were expected to be more representative of the actual conditions of the links. When the links were machined and destructively tested, the average force break and standard deviation were calculated for each of the separate sets of parts (tested with x-ray, ultrasound, and both, as per above) and for all the parts together. The results are shown in figure 2. Note that the rejected parts did have a slightly lower average break point, but not significantly so compared to

235

the accepted parts. Additionally, recalling that 7000-pound (31,116 N) break force had been specified as a "failure" by the factory, and this would have caused all parts to fail, accepted or otherwise, at significantly less than a 2 σ level, indicating that the 7000-pound (31,116 N) number was almost certainly much too high for the process variation of this "out-of-control" casting operation. Mean Mean less 1 Mean less 2 Mean less 3 Standard Break Standard Deviation Standard Standard Part Type Force (lbs) (lbs) Deviations Deviation Deviations 7607 2089 5518 3430 150 Factory 9696 Accepted 2394 6682 4288 1895 SO Ultrasound 9076 Reject 9278 2421 6857 4436 2014 50 X-ray Reject 7261 5034 9488 2227 2807 All 250 Parts Figure 2. This compilation of break force statistics for the later links shows little difference between classic NDT accepted and rejected parts. Figure 3 shows the binned distributions for these sets of links. (Note break-force values are factory numbers in pounds and must be multiplied by 4.45 to become Newtons.) Simply visually examining these distributions with the averages and standard deviations makes it immediately clear that a 7000-pound (31,116 N) failure force is unrealistic for this casting operation. For the factory-tested parts, this is equal to 1.3 standard deviations below the mean. It is likely that the field failure rate is on the order of several parts to several tens of parts per million, reflecting the very high conservative safety factor in the OEM factory specification.

Figure 3. Break force distributions are very similar for both classic NDT failed and passed links. Note that these distributions show very little substantive differences between the classic NDT rejected parts and the classic NDT accepted parts. The attempts to improve quality with classic NDT were ineffective in this parts casting operation.

236

We did look at the original PCRT sort results (the sort based on the earlier as-cast PCRT sort), and found that PCRT actually did a significantly better job of sorting parts than the classic NDT even using the 7000-pound (31,116 N) break force limit from the factory The problem was that although all the low breaking parts could be rejected most of the rest of the castings were also filled with oxide related defects that were not in highly stressed hoop regions. These structurally defective castings were also rejected. If PCRT were used in this production operation it would have caused the management to stop production and correct the process. The x-ray and ultrasound NDT methods just passed most of these massively defective castings day after day. Knuckle Testing Comparisons A government supported consortium of the "Detroit 3" US automotive manufacturers performed a closely monitored experiment using aluminum rear automotive knuckle in a controlled laboratory environment[2]. Results of most of the link testing were available, and this knuckle experiment took advantage of lessons learned. The objectives were to compare the classic NDT to PCRT using more realistic failure criteria and to understand just what was physically occurring and make comparisons among testing methods. A large number of knuckles were first measured to obtain resonant spectra for PCRT. These were post casting and T6 heat treatment, but not machined. They were then subjected to classical NDT (liquid dye penetrant, and x-ray) testing, which were normally specified for these parts in the factory. About 65 knuckles, whether they passed or failed penetrant or x-ray, were then subjected to destructive multi-axis fatigue testing based on worst-case road conditions. A comparison was made between all the sorting methods (resonance testing was trained on the fatigue testing results). Results comparing the overall effectiveness of the various NDT methods were used to evaluate whether the knuckles would or would not fail above or below a set number of cycles associated with an extremely intensified set of forces far beyond those found in vehicle operation. Prediction results for all NDT methods are presented in Figure 4. (Again, note that there is considerable additional data available regarding this project such as tensile bar results for each knuckle, comparison of digital and film x-ray and the details of the fatigue test equipment and procedures used by Defiance Testing[2]) Figure 4. Knuckle test results show that PCRT had zero false accepts or rejects while x-ray and penetrant testing was essentially ineffective.

Please note in figure 4 that there are a variety of conditions to consider. PCRT is obviously the nondestructive method of choice, and if cost is a driver for a factory, PCRT is also the least expensive on a per part basis. In this sample set, which was intended to be about half good and half bad parts, there were many parts with superficial mould release issues, which caused them to

237

fail penetrant testing, thus, in this sample set, penetrant has a very high (79%) false reject rate but still accepts 4% of all "bad" parts. The worst false acceptance is from x-rays. At 96%, x-rays were basically useless for any oxide issues, and are totally misleading as far as determining whether a part may fail in service from cold laps or bifilm inclusions. Depending on x-ray could be extremely dangerous for improperly accepting structurally unacceptable parts. Only PCRT was without error in this sample set. Indications of Bifilms as Causal for Structural Deficiencies All of the knuckles and links from above were examined after breaking. It was found that even those parts that had massive oxide inclusions and cold shuts sometimes passed conventional NDT but failed the fatigue testing at a low number of cycles when the anomalies were in a high stress region of the part as would be expected. However, another very interesting phenomenon was observed. An unexpected outcome was the added evidence that there was almost certainly another failure method occurring. Some of the parts that broke at low cycles had no visually signs of a weakness or crack initiation flaw, even under detailed laboratory x-ray inspection. There was simply no visible basis for low failure cycles. Here are comments based on a university metallurgical laboratory examination of the break face shown in figure 5. Figure 5. This is a break face, folded open, that failed at a stress

cycle value considerably below the "failure" cutoff. The face shows no visible signs of a flaw that would cause the premature break. Quote: "The SEM [Scanning Electron Microscope] studies confirm no macroscopic defect is present. It follows that a high density of microscopic defects may be the cause of the failure. This assumption appears to be supported by close observation of other figures. 'In particular, [close examination] is strongly indicative of the presence of different concentrations of defects in across the field of view. The large central region lying below the general fracture level is indicative of the crack having been deviated significantly. Since no single large defect is apparent as the cause, and in view of the detailed complexity of the fracture surface, it seems likely that a region containing a high density of smaller defects was responsible. The small facet-like regions in this central sunken area are typical of small bifilms. Although no high oxygen signal was found at any point on the fracture surface, a low level oxygen peak was nearly always present. Thus the presence of very thin, fresh oxide bifilms seems likely. The

238

complete absence of a Mg peak indicates the absence of spinels, corroborating the existence of the oxides as pure alumina, and supporting the view that they are recently formed from recent turbulence (possibly during pouring, but more likely during an inappropriate treatment with a rotary degasser). 'It seems likely that the failure occurred in three stages: "The bifilm-rich areas parted during early cycles of the fatigue process. This is likely to have occurred on the first few cycles, leaving no evidence of fatigue striations. Areas relatively free from bifilms (being a natural feature of the heterogeneous tangled mass of defects in this casting) would subsequently allow the cracked areas to propagate across them as a classical series of benchmarks, denoting the step by step, cycle by cycle, progress of the crack. 'When the total cracked area caused the stress in the residual load-bearing area to exceed the tensile failure stress, then a simple tensile failure would occur, although it seems possible that this may have occurred over a few cycles as opposed to a single final cycle. This area is characterized by classical ductile dimples, denoting a so-called ductile fracture. At the base of each of the dimples is a fractured Si particle, and the walls of the dimple are the shear failure regions of the ductile matrix, culminating in sharp arêtes between dimples. (In my personal view, each of the failed Si particles at the base of each dimple has failed from a bifilm, or the transverse fold of a bifilm. Otherwise it is a problem to explain how cracking could occur in a material like Si that has high intrinsic strength and would not be expected to fail by cracking at the low stresses involved in the failure of an Al alloy.) 'Thus in my view this specimen is showing the complexity of failure during fatigue that might be expected from a casting that is literally crammed with bifilm cracks, almost certainly as a result of poor liquid metal treatment (poor pouring or casting techniques would have introduced significantly larger area bifilms, whereas rotary degassing provides high densities of new, fresh, alumina films, nicely shredded to fine sizes). 'For the kind of density of defects that might be present, the radiograph of unfurled bifilms shown as Figure 2.46 on page 65 of Castings 2 nd Edition by John Campbell[3] is a good illustration. Here the density of defects is clearly sufficiently high that the defects are not free to either sink or float; they are so closely crammed together that they effectively form a continuous crack network through the Al alloy. If this sample had been tested in tension, it would be easy to envision how the crack would have zig-zagged through, deviating significantly from a level fracture surface that would have been typical of a fatigue in a steel, where bifilm density is low as a result of rapid detrainment of entrained defects from steels, giving them high strengths and high ductilities, often in the region of 50% or so." The break face shown in figure 5 is similar to two other knuckles that experienced the same effect of breakage at below the number of acceptance criteria cycles with no visible or obvious metallurgical evidence of a cause. Again, these parts were rejected by PCRT but passed by the classic NDT. With the cast aluminum links, which have similar metallurgical properties as the knuckles, inclusions were scattered throughout. Premature breakages most often occurred only in the

239

bushing mounting area (loops at the ends) at the point of highest stress. Refer to figure 1, a simple model of links tension tested to failure, and note that their most common break points occurred at the bushing ends where the material was the thinnest. This is also the area of highest stress according to the FEM model outputs. Other pictures exist of the break faces and can be seen in Appendix A of the USCAR report[2]. Additionally, the authors have pictures of camber link break faces that show visible oxide inclusions as break initiators and the invisible break face initiations. These are available for review by contacting the principle author. Conclusions PCRT is much more effective than classic NDT at detecting structurally unacceptable parts and at minimizing the rejection of structurally acceptable parts. This is true of aluminum, as has been shown in this paper, and it is also true of other metals. Production testing track records of absolutely minimal field returns (near zero) after testing millions of light metal and ferrous castings demonstrate this practical capability. The reason PCRT is so effective is that it actually measures those parameters that affect the structure as manifested in resonant frequency patterns. This allows it to detect the presence of such things as bifilms that are essentially impossible to detect until such time as there is a failure. They represent discontinuities that change the effective part geometry, which in turn changes the resonant patterns measured and evaluated. PCRT is also capable of detecting that foundry or casting processes that are not working properly by determining part rejection or false rejection in numbers higher than a reasonable very small number. That is, it detects consistency of structural performance. Classic NDT, while only looking for indications, rarely correlates significantly with performance parameters such as tensile or fatigue breaking based on the described tests and in-field usage. Are current casting processes designed to pass today's classic NDT? Avoiding shrink voids, gas bubbles, interrupted pours and other sources of cold laps are good casting practices. If foundries used a quality assurance method that also rejected castings with damaging but invisible oxide bifilm inclusions would they also improve their molten material handling and filling to minimize oxide bifilms? The data generated by studies covered in this paper suggest that the invisible oxides or bifilms are as significant for casting mechanical performance as those anomalies that can be detected with x-ray, ultrasound and dye penetrant. References 1. ASTM Standard E2534-10, Standard Practice for Process Compensated Resonance Testing Via Swept Sine Input for Metallic and Non-Metallic Parts and ASTM E- 200108, Standard Guide for Resonant Ultrasound Spectroscopy for Defect Detection in both Metallic and Non-Metallic Parts. 2. Nath, R.H., et. al., "Quasar International's Final Quasar International's Report on the USCAR-USAMP MPCC A356-T6 Aluminum Knuckle Project," Include Appendices A through J, USCAR, (16 June 2005). 3. Campbell, J„ "Castings" 2 nd edition, Butterworth-Heinemann, (2003)

240

Shape Casting: The 4>h International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

Advanced Methoding Concepts for the Gravity Casting of Steel Alloys Bob Puhakka Alloy Casting Industries, New Hamburg, Ontario Canada Keywords: steel castings, casting defects, casting methoding, bifilms Abstract Wlodawer and Chvorinov provided invaluable early guidelines for the feeding of steel castings, but technology has moved on. This paper describes a new set of concepts that apply to the filling conditions for the entire family of steel casting alloys - from plain carbon to super duplex stainless steels. Application of the concepts is achieving new standards of quality together with reduced costs. The Current State of the Art The casting industry (especially the steel castings industry) has been fixated on fulfilling the known rules for castings derived from the work of Wlodawer and Chvorinov. These rules are, of course, useful since they relate to the feeding of the solidification shrinkage. However, the industry has completely neglected the problems of filling to avoid reoxidation and surface turbulence. As a consequence our current methods cannot produce castings reliably with any sort of reproducibility. This serious problem has unfortunately led to pervasive low expectations for steel gravity sand castings. All previous work on this matter has been relatively ineffectual. A fresh conceptual break from tradition is required. In an effort to better understand the inclusion presence in steel castings a study was carried out inspecting castings from several casting facilities. The nature of the inclusion content was tabulated and presented. [4] The results indicated that in excess of 83% of the defects present in steel castings were reoxidation inclusions. Reoxidation inclusions result when molten steel and air combine. They are solid particles and agglomerates of oxides such as AI2O3, MnO, CaO, T1O2 and FeO which mainly initiate as films on the surface of the liquid metal. Castings that possess a large quantity of reoxidation inclusions will require costly rework to make saleable; and even then they may never be fully reliable. Such castings are prime candidates for leaks, reduced mechanical properties, poor machining and poor weldability. A recently published article on the nature of casting defects states, "Inclusions are generally associated with the flow of liquid metal into the mold during pouring. However, modeling and verification trials in foundries have failed to indicate how gating systems may be universally improved." [5] In fact, the conclusions of almost every research paper on the topic of castings defects in steel have come to nearly identical conclusions. [6] [7] The fill system design logic currently employed in the industry is relatively arbitrary and unscientific. In fact it is common for the filling system to be assembled from pre-formed refractory sleeves as a result of the widespread belief that sand molds will not withstand contact with liquid steel. The parallel channels dictated by the use of this system ensure a maximum of air entrainment, with consequential problems for the casting. This paper will explain in some detail how it is possible to design a fill system that prohibits the formation of reoxidation

241

inclusions. Before we begin, however, it is important to review the current industry fill system design features in detail. The Pouring Cup

F,gure 1

An illustration of the aspiration mechanism of a conical pouring cup

, .

e

. ,

Tnfir.hanwin or a conical nonnno cnn

The traditional conical pouring cup is perhaps the worst single feature of traditional filling systems. The 'design' is for the sole purpose of creating an enlarged, targeted entrance to the downsprue. However, a detailed review of the system reveals that it is in fact an efficient aspiration pump, taking down at least as much air as metal into the running system during the pour[12]. The quantity of entrained air explains the so-called 'Discharge Coefficient' value of only about 0.5. A reasonable input for the system should achieve a value 1.0 otherwise defects are certain to be created. Figure 1 illustrates the aspiration of air that takes places within the conical basin. Usage hints such as "pour to the back of the cup to avoid vortex generation" is irrelevant. Such , . . .,

,■

·

,

, ■

.,

operational details are a diversion, overlooking the major r

damage caused by this feature.

'

σ

J

The Non-Tapered /Reverse Tapered Sprue The taper of the sprue is not only to allow for drawing the sprue pattern from the mold! Parallel refractory runner sleeves or reverse-tapered sprues are widely used despite the well known fact that gravitational acceleration narrows the falling stream predictably, predicting necessarily that parallel and reverse tapered sprues will not constrain the melt, and will therefore allow surface turbulence to create oxide damage. Clearly, the sprue taper must be accurately calculated and moulded for each and every casting. The Sprue-Runner Transition Studies have shown [2] that it is not possible to have an ideal flow behavior when transitioning from a round downsprue into a square or rectangular runner. Thus the continuity of form of the flow channels is critical to success. The Sprue Well

Figure 2- the sprue weil

essentially serves as a meial-air bl

,r

Traditionalists view the well as a 'cushion' for the melt at the sprue exit. Its widespread use appears to have become dogma, somehow supposed to reduce flow velocity and promote laminar flow. Unfortunately however, the well once again removes the constraint on the flowing stream, allowing churning, mixing and combining air with the molten metal. There are many formulae used to calculate sprue well size for a particular gating system but recent research has emphatically confirmed that the ideal volume for a sprue well, contrary to current practice, is actually zero. It is known ^, ^.,

>,

·.

L

c

t

,« , , ι ^

that the well provides a stream of entrained bubbles for some

242

seconds, whereas a simple radiussed turn can redirect the flow without the generation of a single bubble [2]. The Runner Choke

Figure 3 - the choke produces severe localized turbulence resulting in the entrainment of surface films and air bubbles

An industry publication recently sent out a newsletter with the following advice, "Practical Tip: Increase Choke Size When Having Issues with Flow Rates"[8]. The advice overlooks the fact that the provision of any choke accelerates the melt, causing it to jet, resulting in the generation of copious quantities of reoxidation inclusions. Figure 5 shows a screen grab from a fill simulation of a 'textbook' AFS 1:4:4 ratio system with a runner choke. The velocity vector behavior immediately following the choke indicates the location where volumes of air bubbles and reoxidation inclusions will be generated during the pouring of the casting. [10]

The Runner Expansion Methods The practice of reducing the velocity of the advancing fluid front by increasing the cross-sectional area of the runner over the sprue exit area can have disastrous consequences. It has been demonstrated by X-Ray radiographie video for real metal, and many times in computer simulations, that the advancing melt will not expand to fill an enlarged channel. An unfilled runner is subsequently responsible for the entrainment of air, oxide films and mold material - depositing much of this into the casting cavity. There is no 'expansion ratio' capable of producing premium results. The use of traditional expansion ratios must be avoided.

Figure 4 - runner expansion techniques result in unfilled runners producing severe surface turbulence, surface film and air entrainment

The Runner Extension

Figure 5 — the runner extension provides a location of unrestrained flow resulting in the aspiration of air into the casting cavity.

During a gating course early in my career I was told, "Thou shalt not allow the first metal poured to enter the casting cavity." This advice, of course, was an emphasis for the need to include a runner extension into the fill system. At the time this was probably good advice because the traditional filling systems created large quantities of damaged melt. The extension was intended to collect the debris and oxides generated in the running system by the first-

243

arriving metal. In fact, in the naturally pressurised systems described in this report this is no longer the case since the melt arrives at the gates in relatively good condition. What fraction of metal does initially bypass the ingates is nowadays potentially counter-productive. It can often be seen to become part of a rolling-back wave, rolling in air and oxides that will find their way into the casting cavity via the last ingate. [9] [11] The New Methodology To avoid oxides it seemed rational to adopt a system that would assist with the elimination of air from the flowing stream. For this reason, the naturally pressurized filling system design was adopted [2]. This is a system in which the areas of the filling channels are calculated by finding the velocity, V, at each fall distance, h, from the melt level in the pouring basin, assuming no friction. The approach is therefore a simple balance between potential energy, mgh, and kinetic energy, mV2/2. The approach is well known to casting method engineers.

Figure 6-an ideally designed spruedual runner transition. No well, no choke, no runner expansion

The difference in this situation is to accept these areas and provide the filling system with only these calculated areas at every point throughout the downsprue and runners. Only the gates would be increased in size to reduce the velocity of entry to the mold to the critical 0.5 m/s if possible (on occasions this would be raised to 1.0 m/s if necessary, but not beyond this already 'stretched' limit to the Rule). A typical 'sprue exit/runner/gate' ratio for such a system might vary from 1:1:4 to 1:1:20. It must be stated however that the preselection of such a 'ratio' has no part to play in the design of a proper fill system; the ratios simply happen, occurring as a result of the design process.

It is worth repeating that the standard technique of increasing the area of the runner to reduce the velocity of the flow, using for instance a ratio 1:2:4 or other expanding ratios, does not work. It merely provides an unfilled runner in which turbulence can be generated to damage the flow [2], The system requires a specially designed pouring basin of a deltashaped, offset stepped type [2], to reduce so far as possible the ingress of air into the entrance to the filling system. The benefits of such a design were recently excellently illustrated with a video of a water model by workers at CANMET [3] [13]. (Campbell [2] describes a contact pour technique using bottom-teemed ladles to exclude air at the sprue entrance but our lip pour techniques suited to our size of casting did not require this particular solution). The sprue requires a correctly calculated taper. Although Campbell recommends a curved hyperbolic taper [2] it was judged that our castings were not sufficiently tall to benefit greatly from such sophistication. Thus so far we have calculated only the sprue entrance and the sprue exit and connected the two with a straight taper It clearly works well with the limited size of casting we pour (encompassing the range 1 to 2,500 kg).

244

Figure 7 - delta-shaped offset step pouring basin

The other major feature of this approach is that only bottom gating into the mold cavity can be permitted; otherwise the melt falls inside the mold cavity, exceeding the critical fall distance of a few millimeters in which gravity accelerates it to above its critical velocity, so that it starts to jump and splash, creating surface turbulence and entraining defects such as bubbles and bifilms (and sand inclusions, which are an excellent indication of a turbulent filling system - not an indication of poor molding sand). Thus gating at the mold joint (for traditional horizontally parted molds) has become a feature of the past - a luxury that can be no longer risked or afforded. Along with blind risers [14], conical pouring basins and runner chokes; parting line gating serves as a corner-cutting, "cost-saving" practice that produces damaged castings A final discipline is the modeling of every casting to check that the methoding is complete and effective. Thus the action of the filling system is carefully studied to ensure that no pockets of air remain after the first pass of liquid. Such pockets allow turbulence and the entrainment of air and oxides. The system needs to be seen to fill and to pressurize gently against the mold walls. This is the natural pressurization concept, arising from the friction generated by the flow against the channel walls. The channels clearly need to stay full during the filling process. Having filled successfully with only metal (no air in the form of bubbles), the feeding system is then checked to ensure that there is no danger of shrinkage. For some castings this exercise is

vigun

8

- the use of process simulation is absolutely

essential. The unique solidification profile of each and y individual geometry must be evaluated and

ever

not always straightforward. The provision of manipulated. a bottom gate and high level feeders (top if possible) sometimes requires turning the casting through 180 degrees before a workable solution can be found. Occasionally, much use of heavy chills is required to generate favorable temperature gradients. It is at this step that one notices the vast difference in approach compared to the traditional methoding practices. The traditional approach selected the casting orientation based on the 'easiest' way to feed the casting, with the fill system being only an after-thought. The new approach begins with assessing how to fill the mold as perfectly as possible; and then finds a way to meet the feeding requirements. The fill system takes precedence. Only when both filling of the mold cavity is seen to be tranquil, and the action of the feeders is seen to be adequate to all locations of the casting, is the tooling built, the mold made and finally poured. It has become interesting to note that the pouring of the mold is no longer viewed as being in the lap of the gods, but is now viewed as an expected confirmation of the methoding technique. That is not to say that failures have not been experienced. These have occasionally occurred as a result, for instance, of forgetfulness or oversight of some key aspect of the design prior to pouring. Such mistakes illustrate the fallibility of a regime in which single person is totally

245

responsible, but at the same time working under pressure in a production environment. Clearly, like all professional engineering designs, at least one other competent person should be available to check and sign off the design prior to manufacture. The naturally pressurized fill system is now used exclusively at the author's facility. Castings are produced from the entire family of ferrous engineering alloys in sizes ranging from 1 to 2500 kg. The naturally pressurised concept is applied without exception, and found to apply successfully across the complete range of casting sizes. On Filtration The use of reticulated ceramic foam filters is very common in the steel foundry. Almost universally, the designs use the filter as a damage mitigation device even though it is found to be largely unsuccessful if simply introduced into a fill system designed with traditional features. The filter does not right several wrongs! For instance, an unconstrained front-end with a conical pouring basin and parallel sprue will generate reoxidation inclusions that will plug the filter. In addition, an unconstrained, ratio-based runner design will generate reoxidation inclusions postfilter that will now be deposited into the casting cavity[15]. In contrast, the introduction of a ceramic foam filter into a naturally pressurized fill system as a velocity reduction device yields surprising additional benefits. Without the generation of reoxidation inclusions the filtration capacity of the filter now appears to be essentially infinite. This brings significant savings, as systems can be designed with fewer and smaller filters. In fact, recent experience confirms the new fill system designs allows up to ten times more metal through the filter than the manufacturer's recommended filtration capacity; there has been no experience of a filter blocking prematurely. On Fill Rate There continues to be great debate on the determination of target fill rate for the pouring of steel castings. Traditional formulae include considerations such as fluid life and cooling rate etc. There is also the questionable approach of simply taking the square root of the casting weight (in lbs) and using that value as the target pour time (in seconds). All such techniques seem to be largely irrelevant. Furthermore, it is adherence to these largely arbitrary calculation sets that appears to have prolonged the frustrating attempts toward defect elimination for decades. A fill system must not be designed that exceeds the volumetric output of the ladle. If the system is too large to be kept full by the ladle, dramatic and severe aspiration necessarily takes place damaging the casting. A fill system needs to be designed that pours continuously, as quickly as possible, and fills the casting cavity as slowly as possible- all at the same time. 'As quickly as possible' is very simply the maximum volumetric output of your ladle. 'As slowly as possible' is simply maintaining a maximum limit ingate velocity of approximately 0.5 m/s. An arithmetic manipulation of these two competing design targets is the fill system design process. The challenge for the design engineer includes: 1. Know the maximum volumetric output for each and every ladle used in the facility. An intelligent designer will design a fill system customized for the very ladle to be used to pour the casting.

246

2. An empirically determined correction factor for friction and velocity losses. Specific to the particular molding materials, sprue/ runner geometry and number of fill system branches - the design engineer must know this value. A direct comparison of actual fill times vs. ideally calculated (excluding loss compensation) will provide a very specific correction factor to be used. As afirstapproximation africtionalcorrection to thefillrate a value of 1.3 (corresponding to a 30% loss of speed) is a useful starting point. 3. A specifically designed pouring basin for the ladle to be used. The pouring basin is the critical step in the transfer of the molten metalfromthe ladle to the sprue entrance - the basin must be designed for the ladle. It is not unreasonable, then, to imagine that a casting facility will have a basin design for each and every ladle used. Confirmation by Results Steel castings are notorious for their need for excessive dressing, rework and cosmetic repair. The majority of the imperfections visible on the surface are clearly reoxidation inclusions [4] resultingfromthe entrainment of air during thefillingof the mold. In addition, the surface finish is often poor. The carbon steel (ASTM A216 WCB) pump casing shown in Figure 9 is an example of the new approach. It has been shaken out and put through the coarse shot blaster to remove the remaining adhering mold material. This is a casting that is required to be fully pressure tight and therefore subjected to full magnetic particle inspection (MPI), liquid penetrant inspection (LPI) and radiographie testing (RT) to insure this condition. The casting is seen to be clean and devoid of surface inclusions, shrinkage and tearing. In addition, it is perfectly leak-tight.

Figure 9 - a carbon steel pump casing immediately following shakeout and shot blasting. There are no inclusions and the surface finish is excellent.

Conclusions 1. The majority of casting defects (including much gas porosity, perceived microporosity, hot tears, leakers and reoxidation inclusion) are the result of defects (mainly double oxide films and/or air bubbles) entrained by thefillingprocess. The problems arisefromthe use of traditional filling systems (including conical pouring basins, ill-tapered sprue using refractory sleeved running systems, choked runners, parting line gated molds etc.). 2. A naturally pressurizedfillingsystem design can practically eliminate entrained defects, giving essentially perfect castings. Castings are characterized by good surface finish, freedomfromleaks, surface inclusions, porosity, hot tears and cracks. Absence of

247

upgrading work, dressing and welding, has brought significant economies. Additionally, the absence of rework avoids the unintentional masking of initially unseen defects that may contribute to subsequent service failures. 3. The author has validated the success of these techniques on a broad range of alloys in the steel, white/grey/ductile iron, brass, bronze, Ni-base superalloys and high-conductivity (pure Copper) alloy families. It seems reasonable to suppose that the casting manufacturing systems presented here would apply to metals and alloy of all types. Acknowledgements I am grateful to John Campbell, author of 'Castings' for his continued advice and encouragement. John has become a close, personal colleague without whom I would never have been able to accomplish the conversion as quickly and completely. Our near-daily correspondence allows for me to dialogue with another individual in the same headspace regarding the nature of casting defects. Sincere thanks are also due to my co-workers at Alloy Casting Industries in New Hamburg whose dedication and hard work made our revolution possible. References 1. 2. 3. 4. 5.

J Campbell "Castings" 2003 Elsevier. J Campbell "Casting Practice" 2004 Elsevier S Kuyucak; 68th World Foundry Congress, Chennai, India 2008 (February) pp 483-48 J M Svoboda, Monroe R W, Bates C E, Griffin J; Trans Amer Foundry Soc 1987 95 187-202. Predicting the Occurrence and Effects of Defects in Castings; Malcolm Blair, Raymond Monroe, Christoph Beckermann, Richard Hardin, Kent Carlson, and Charles Monroe, JOM 2005 6. J.A. Griffin and CE. Bates, Ladle Treating, Pouring and Gating for the Production of Clean Steel Castings, SFSA Research Report No. 104 (Crystal Lake, I: Steel Founders' Society of America, 1991) 7. P. Scarcer, Jr., CE. Bates, and J.A. Griffin, "Using Gating Design to Minimize and Localize Reoxidation" (Paper presented at the 56th Steel Founders' Society of America National Technical & Operating Conference, Chicago, Illinois, 7-9 November 2002) 8. AFS Education Connections Newsletter, April 20, 2010 9. http://bobpuhakka.blogspot. com/2010/06/not-to-beat-dead-horsebut.html 10. http://bobpuhakka.blogspot.com/2010/04/sacred-cows-of-metal-casting-episode-2.html 11. http://bobpuhakka.blogspot.com/2010/01/left-to-end-of-runner.html 12. http://www.youtube.eom/user/bobpuhakka#p/u/4/i5vWOrQUVLI 13. http://www.youtube.eom/user/bobpuhakka#p/u/57/5Zkh4wrnsAk 14. http://www.youtube.eom/user/bobpuhakka#p/u/46/uwS3-l_rUlg 15. http://www.youtube.eom/user/bobpuhakka#p/u/44/JxnNKiCz8Jk

248

Shape Casting: The 4"1 International Symposium Edited by: Mura! Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

ADVANCED CASTING MOLD DESIGN TECHNOLOGY OF THE LCS WATERJET INLET TUNNEL ENTRY EDGE COMPONENTS LaurentiuNastac1 and John Romanelli2 'Concurrent Technologies Corporation, Pittsburgh, PA 15219, email: [email protected] Concurrent Technologies Corporation, Johnstown, PA 15904, email: [email protected] Approved for public release; distribution is unlimited. This material is submitted with the understanding that right of reproduction for government purposes is reserved for the Office of Naval Research, Arlington, Virginia 22203-1995. Keywords: LCS waterjet entry edge components, casting modeling, mold design, sand mold printing, ASTM A757 C1Q steel, prediction of macro-shrinkage, porosities, hot tears, and cracks Abstract In order to reduce cost, increase performance and ensure quality, this Navy Metalworking Center (NMC) project utilized an advanced casting simulation-based optimization approach to assist in the improvement of the mold design of LCS Waterjet Inlet Tunnel (WjIT) entry edge components. This approach helped to minimize mold filling and solidification-related defects (misruns, coldshuts, shrinkage and porosity and hot tears), as well as post-solidification-related defects (hot and cold cracks, distortion and residual stresses). The results of this optimization were used to more readily achieve first-time quality on the geometrically challenging WjIT components. The melt chemical composition and the thermo-physical and mechanical properties of ASTM A 757 C1Q steel material as a function of temperature were used in the simulation study. To accomplish this work, NMC used Nova Flow&Solid™ and NovaStress™ software. Introduction The water-jet inlet tunnel entry edge (WjIT) for the Lockheed Martin Freedom class Littoral Combat Ship (LCS) was previously manufactured from 13 ASTM A131-DH36 25-mm thick plate segments by welding and grinding. The hydrodynamic profile of the WjIT entry edge is a critical operational factor (see Figure 1). The objective of this Navy Metalworking Center (NMC) manufacturing technology (ManTech) project was to validate that a casting solution is indeed the most promising alternative manufacturing method for WjIT [1], The ManTech project also sought to demonstrate that a cast steel (ASTM A 757 C1Q) provides comparable tensile and impact properties to ASTM A131-DH36 rolled plate steel. There are four WjITs entry edges on the LCS and each is slightly different. The total length of each WjIT entry edge is about 4m and its width is about 1.5 m. The thickness of each WjIT is 25 mm. The casting solid models developed in this project were divided into three segments due to the size, casting, heat treating and handling issues. The segments were designated as Part A, B and C. These segments will be assembled by welding. This paper describes the process modeling approach used in the development of the mold design technology used to produce the WjIT parts.

Figure 1. WjIT geometry (Part B).

249

Objectives and LCS WjIT Casting Requirements The objectives of this work were as follows: (1) apply modern mold design technologies to assist in the manufacturing of the LCS entry edge casting components using: (i) ProMetal Rapid Casting Technology (RCT) [2] and (ii) NovaCast software for mold filling, solidification and stress modeling [3], and (II) achieve first-time quality on the geometrically challenging WjIT components made of ASTM A 757 C1Q steel material with the following requirements: (i) nondestructive testing (NDT) inspection (magnet particle, die penetrant, and radiography inspection); (ii) mechanical properties; (iii) dimensional accuracy; and (iv) surface tolerances. The fabricated WjIT component was previously made from ASTM A131 DH36 steel plate [4], A cast steel alloy that can provide similar tensile and impact properties is ASTM A757 C1Q steel [5]. Chemical composition maximum limits of ASTM A 757 C1Q steel are presented in Table 1. Tensile and impact minimum requirements for ASTM A 757 C IQ and ASTM A 131 DH36 are shown in Table 2. Table 1 Chemical composition of ASTM A 757 C1Q Carbon

Manganese

Phosphorus

Sulfur

Silicon

Nickel

Molybdenum

0.25

1.2

0.025

0.025

0.6

1.5/2.0

0.15/0.30

Table 2 Tensile and impact requirements for ASTM A 757 C IQ and ASTM A 131 DH36 UTS (MPa) 515

YS (MPa) 380

El

(%) 22

RA

CVN (J) 20@-46°C

(%) 35

355 22 49024@-20°C in T direction n/a 620 34@-20°C in L direction T- Transverse direction, L- longitudinal direction

Specification ASTMA757C1Q ASTMA131 DH36

Requirements for NDT were selected in accordance to NAVSEA Technical Publication S9074-ARGIB-010/278 [6] and ABS Guide for Building and Classing Naval Vessels 2004 [7]. Thus, the castings have to be subjected to dimensional, visual, magnet particle (MT), die penetrant (PT) and radiography (RT) inspections. Each casting has to be 100% visually inspected for any surface defect. Criticality level 2 requires 50% minimum RT coverage. Level 4 is required for shrinkage, porosity and inclusion in accordance to ASTM E 446 [8]. Mold Design Technology Development: Optimized Rigging System for WjIT Part B ProMetal RCT designed the rigging system (gating and risering), and NMC performed the simulation work to improve the rigging system design (see Figure 2). The process and material parameters used in the simulation are shown in Table 3. The chemical composition, thermophysical and mechanical properties of C1Q steel material used in the simulations were functions of temperature. Also, the mold material, chills and exothermic neck-down sleeves used in the simulations were also functions of temperature. 10 ppi ceramic foam filters were used on each runner for each component. NMC used Nova Flow & Solid™ and NovaStress™ software [3] for these simulations. These casting simulation software tools were previously validated by CTC [9-11]. For part B, the best casting position was determined to be at 15 degree orientation (e.g., tilting angle). However, due to some casting and mold printing potential issues, the 0 degree casting position was used. Then, the bottom poured gating system was hand calculated and added to the part and a second analysis was performed with mold filling to

250

determine the best location of risers and chills. After 3-5 iterations for each part, an optimized rigging system was developed. The predicted mold filling time was about 10 seconds and the predicted total solidification time of the casting with rig is about 75 minutes.

Figure 2. Optimized casting mold rig design system and mold package for Part B. Table 3. Simulation parameters Simulation parameters

Material type /Value

Mold Mold thickness

Silica Sand/Furan binder Min. 50 mm

Initial Mold and Ambient Temperature

20 °C -25 s

Pouring time (ladle pouring-over lip) Pouring Temperature ASTMA757C1Q

1560°C T L = 1489 °C,Ts= 1437 °C

The casting with rigging weighed 661.5 Kg. No misruns and coldshuts were predicted. To avoid significant turbulence in the in-gates and mold erosion, the sprue was choked and seven 10 ppi ceramic foam filters were used, one for each runner. The predicted flow rate was relatively constant during filling demonstrating that the gating system is properly calculated. To avoid mold erosion, mold wash was applied to the bottom of the sprue well. The predicted (recommended) mold shakedown time was about 24 hours. Figure 3 illustrates the temperature distribution at the end of solidification and the solidification time profile. Figure 4 shows the predictions for CIQ Part B in terms macro- and micro-shrinkage (e.g., shrinkage porosities) profiles. It can be seen from Figures 3 and 4 that the exothermic sleeves and the chills performed very well since the risers are still very hot and all the shrinkage was directed toward the risers. The predicted pressure (e.g., tensile and compressive stresses) and the yield strength of the Part B casting at a maximum temperature of 1400 °C are shown in Figure 5. The predicted pressure and the yield strength of the Part B casting during cooling are shown in Figure 6. The thermal tensile stresses shown in Figure 5 are relatively high especially in the nose area at temperatures near Solidus. The tensile stresses are slightly below the yield stress of the CIQ steel casting at 1400 °C. Therefore, slightly higher hot tearing and cracking tendency would be expected for CIQ component under the current rig design and casting conditions. Mold Design Technology Development: Optimized Rigging System for WjIT Parts A and C A similar methodology was applied to design the molds for parts A and C. Figure 7 shows the predictions for CIQ Parts A and C in terms macro-shrinkage and Figure 8 shows the micro-shrinkage (e.g., shrinkage porosities) profiles for both parts.

251

Figure 3. Predicted temperature distribution at the end of solidification and the solidification time profile for Part B.

Figure 4. Predicted macro-shrinkage and porosity profiles for Part B.

Figure 5. Stress analysis results: Pressure and yield strength distributions at the end of solidification (time ~ 75 min).

Figure 6. Stress analysis results: Pressure and yield strength distributions during cooling (time 200 min).

252

Figure 7. Macro-shrinkage (a) and shrinkage porosities (b) predictions for Part A.

Figure 8. Macro-shrinkage ((a) and (b)) and shrinkage porosities ((c) and (d)) predictions for PartC. Manufacturing of WjIT Entry Edge Parts The assembled mold for Part B of the WjIT is presented in Figure 9. In addition to the WjIT segments, several plate shape samples were cast separately. The plate shape sample was used to develop the appropriate heat treatment parameters.

Figure 9. Assembled cores (mold) for part B of WjIT. Castings were shot blasted to remove the oxide from the surface and were then inspected which consisted of visual and dimensional inspection, magnet particle and radiographie inspection after the risers were removed by flame cut and grinding. Magnetic particle examination indicated that (i) Parts A and C are free of defects and (ii) the presence of three cracks in the nose area of Part B of the WjIT. The cracks were excavated and the crack area was

253

reexamined with the die penetrant. The die penetrant indicated no presence of cracks after excavation. Part B of WjIT was then sent to heat treat. Heat treatment consisted of normalizing at 898°C for 4 hours and then air cooling. The part was again die penetrant examined to make sure there are no cracks before weld repair. Part B was preheated to 186° C; weld repaired using 8018-C3 8 mm diameter electrode and stress relieved at 537° C. After stress relieve, Part B was again magnetic particle inspected and no indication was detected. The shape of the Part B WjIT casting after mold removal in shakeout and after die penetrant and X ray inspection is presented in Figure 10a and 10b, respectively.

Figure 10. WjIT after Mold Shakeout (a) and after Die Penetrant and X-ray Inspection (b). Comparison of Simulation Results with Actual Test Results A comparison between predictions and experimental measurements for porosities are illustrated in Figures 4, 7, 8, and 11-14. The locations for radiographie inspection are also shown in Figures 11, 13, and 14. The predictions compare well with the radiography results in terms of shrinkage amount and location. Figure 12 show the cracks developed after riser removal prior to any weld repairs. The tensile stresses at the end of solidification and during cooling are relatively high in the nose area (see also predictions in Figure 8) and can create hot tears if associated with segregation [12, 13] and porosities that act as nucleation sites for hot tears [14]. Hot tears will act as nucleation sites for cracks that may propagate later on during cooling and further processing (e.g., riser removal and/or welding) [14]. Carbon positive segregation toward the risers can also significantly contribute to the increase of the cracking tendency [14] in steel alloys, especially in the regions beneath the riser necks, where cracks are typically found after plasma riser removal. The casting trials confirm the predictions that Part B has some potential for hot tearing and cracking development and additional steps have been taken to minimize these defects. The steps included mold shakeout after minimum 24 hours, stress relieving immediately after casting and before riser removal, and additional improvements in the mold design minimize the thermal gradients in the casting during solidification and cooling.

Figure 11. Actual shrinkage porosities for Part B (visual inspection and radiography)

254

Figure 12. Part B nose casting showing cracks 1,2 and 3 prior to any weld repair.

Figure 13. Comparison of Shrinkage Porosities (Visual Inspection and Radiography) and Predictions for Part C.

Figure 14. Comparison of Shrinkage Porosities (Visual Inspection and Radiography) and Predictions for Part C. Concluding Remarks A comprehensive simulation was performed to assist in the mold design developments for the LCS WjlT cast steel components to minimize mold filling and solidification-related defects (misruns, coldshuts, shrinkage, porosities and hot tears), as well as post-solidification related defects such as hot and cold cracks, distortion and residual stresses. It was demonstrated that the model predictions compare well with the experiments in terms of shrinkage amount and location.

255

From this simulation study, NMC established that it is, indeed, feasible to produce the LCS WjIT casting components using the specified material and ProMetal RCT advanced molding processing technology [2]. NMC also determined that the main factors that influence the integrity and mechanical properties of these complex castings, and therefore, their quality, are the melt chemistry, the foundry practice and the solidification characteristics [1]. Several additional improvements were made on the mold rig design of all WjIT components under the existing foundry practice to further minimize the amount of shrinkage, porosities, and deformation, as well as the tendency for hot tearing and cracking. Prototype castings were manufactured and successfully passed NDT inspection and met all the mechanical properties and dimensional requirements. The actual results of radiography inspection compared well with the predicted shrinkage and micro-porosity. Acknowledgments This work was conducted by the Navy Metalworking Center, operated by Concurrent Technologies Corporation under Contract No. N00014-06-D-0048 to the Office of Naval Research (ONR) as part of the U.S. Navy Manufacturing Technology Program. Approved for public release; distribution is unlimited. The authors would like to thank ProMetal RCT personnel for successfully manufacturing the cast LCS WjIT entry edge components. References 1. J. Romanelli et al, "LCS Waterjet Inlet Tunnel Manufacturing Improvement S2279: Task 1: Entry Edge Manufacturing Improvement Report" (NMC Final Report, TR No. 10-131,2010). 2. ProMetal RCT (www.prometal-rct.com). 3. Novaflow&Solid™ and NovaStress™ Software (Novacast AB, Sweden, www.novacast.se). 4. ASTM Standard A 131/A 131M-04a "Standard Specification for Structural Steel for Ships" (ASTM West Conshohocken, PA, 2004). 5. ASTM Standard A757/757M-00 "Standard Specification for Steel Castings" (ASTM West Conshohocken, PA, 2000). 6. 9074-AR-GIB-010/278 NAVSEA Technical Publication:"Requirements for Fabrication Welding and Inspection, and Casting Inspection and Repair for Machinery, Piping, and Pressure Vessels (Naval Sea System Command, 1995). 7. ABS Guide for Building and Classing Naval Vessels 2004. 8. ASTM Standard E 466/A 466 "Reference Radiographs for Steel Casting up to 2 inch [51 mm] in Thickness" (ASTM West Conshohocken, PA 2000). 9. L. Nastac et al, "Advances and Challenges in Investment Casting of Ti-6-4 Alloys," International Journal of Cast Metals Research, UK, 19 (2), (2006). 10. L. Nastac, J. Valencia and K. Stefanick, "Stainless Steel Investment Casting Evaluation Rapid Response" (NMC Final Report, TRNo. 05-006 2005). I L L . Kramer et al, "Implementation of Steel Castings to Enhance Reliability and Decrease Cost for the M777 Lightweight Howitzer" (NMC Final Report, TRNo. 07-02,2007). 12. L. Nastac, Modeling and Simulation of Microstructure Evolution in Solidifying Alloys (Springer Verlag, 2004, 305 pages). 13. L. Nastac, A. Patel, and G. Maurer, "Modeling of Macro-segregation in a permanent mold casting" (Proceedings of the International Symposium on Liquid Metal Processing and Casting, Santa Fe NM, September 20-23,2009, TMS, 2009). 14. L. Nastac, "Estimation of Hot Tears in Castings" (CCAT-AFRL meeting, Springfield, OH, November 9-10,2009).

256

Shape Casting: The 4>h International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

E V A L U A T I O N O F T H E D I S T O R T I O N OF A H Y D R O TURBINE BLADE DURING HEAT TREATMENT PROCESS Jinwu Kang, Xiaokun Hao, Gang Nie, Haimin Long, Hailiang Yu and Tianyou Huang Key Laboratory for Advanced Materials Processing Technology, Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China Keywords: Turbine blade casting, Heat treatment, Distortion, Curvature Abstract Hydro turbine blade castings are susceptible to distortion during heat treatment process due to their thin and curved shapes. By the numerical simulation method, the displacement results of the castings can be acquired. However, the displacement consists of distortion and contraction or expansion and depends on the selection of reference points which are hard to be exactly selected. In this paper, a distortion evaluation method is presented, in which the curvature variation of local areas with respect to the original shape is utilized. By adding the displacement results, the finite element model at each step is converted into STL format files. Then, based on the STL files, the curvature around each vertex is calculated and the curvature variation of each step relative the original shape is acquired for the description of distortion. The distortion degree of the whole casting is evaluated by the local variation of curvature, which is independent on references points and also suitable for the evaluation of distortion during heat treating process. Reference points can be selected in the areas with smallest distortion, which can be used for the displacement evaluation. 1. Introduction Distortion is one of often found problems of castings, especially large castings with thin sections such as frame shape or those with curved surfaces. To predict distortion, it is necessary to perform thermal and stress analysis of castings because distortion is the result of non-uniform cooling and closely related to the behavior of casting material. Distortion prediction is usually designated by displacement results directly obtained from stress analysis. And the comparison of the final shape and that of the beginning is used as direct illustration of distortion. [1-6]. However, the displacement results depend on the selection of constraints, as the constraints are changed, the displacement will be also different. However, this has not yet attracted researchers' attention. Castings undergo expansion, contraction and distortion during heat treatment processes; the former two are caused by cooling and phase transformation, the latter by uneven cooling of different positions of the casting. Contraction or expansion is just uniform reduction or enlargement of size, which is different from

257

distortion. Therefore, distortion must be separated from displacement and it should be independent of constraint. The author has proposed three methods, the net distortion (separated from displacement), local surface normal variations and machining allowance [7]. However, the distortion evaluation still depends on the selection of reference points. Chen also proposed to evaluate a qualified impeller casting by the final distribution of uniform machining allowance [8]. In this paper, a curvature method is presented to depict distortion without reference points. And it is used to analyze the distortion of a heavy hydro turbine blade casting. 2. Distortion evaluation by curvature variation Casting process starts from part design which is provided by product design department. Foundry engineers add machining allowance to the part design and enlarge it by contraction rate and design rigging and gating system. For numerical simulation, the casting design has to be enmeshed to finite difference (FD) or finite element (FE) models for thermal and stress analysis, and stress analysis is usually performed using finite element modeling. Numerical simulation provides displacement of casting as the final results. The displacement results of each step are added to the original shape, then, the shape of each step is obtained. The shape in finite element model is converted into STL file which are a group of discretized triangles. Curvature of each vertex of the triangle is calculated. The flowchart of this method is shown in Figure 1. Part design Add machining allowance, contraction rate Casting design (FE model)

Stress analysis of heat treatment process

Displacement results Casting design (Surface STL file)

Simulated casting shape (Surface STL file)

Curvature variation

Figure 1 Flowchart of the distortion calculation methods

258

2.1 Conversion of finite element model to STL format As the stress analysis is finished, the coordinates of each node is modified by adding the displacement result. Then, the finite element model of each step is obtained. Select the surface elements and surface nodes, and then judge the surface of each surface element; if the element type is tetrahedral, each surface is a triangle, which can be directly transformed into the format of STL triangles. As it is hexahedral elements, each quadrilateral shape can be separated into two triangles. 2.2 Calculation of local curvature The curvature of a vertex of a triangle in STL format is calculated as follows. m

1ι,=2π-Σθ„

(1)

Where #,· is the angle ; in the triangle j , m is the number of triangles at node /', as shown in Figure 2. The curvature of each vertex is smoothed by averaging its value and that of its surrounding vertices. The variation of curvature at vertex / is M/=*,'-*,"

(2)

Where M / is the relative variation of curvature at the time t, kt andA:,' are the curvature at the beginning and time t. The variation of curvature is used to depict the distortion.

Figure 2 Discretized surface triangles of casting (STL format) 3. Distortion of a heavy turbine blade during casting process The hydro turbine blade for Three Gorges Project is made of Cr-Ni stainless steel weighing 18 tons after machining. The casting undergoes normalizing, first tempering and second tempering. During normalizing process, the casting is heated to 1050°C

259

and held for a period of time and then cooled by forced air flow. As the casting cools to the martensite transformation start temperature, Ms (276°C), martensitic transformation starts and finishes at Mf (78°C). The casting is simulated by Deform-3D for the normalizing process, with the bottom constrained. Based on the finite element model, the STL file is formed, as shown in Figure 3. Based on the above method, the curvature is calculated, as shown in Figure 3(c). It can be seen along the diagonal line from the bottom right corner to the upper left corner is of big curvature. The negative value of curvature is because of the spatially twisted shape.

(a) finite element model (b) STL format (c) initial curvature distribution Figure 3 The format conversion from finite element model to STL The heat treatment thermal schedule and the heating and cooling curves of the highest and lowest temperature of the blade are shown in Figure 4. There are two small temperature difference peaks during heating and a big peak during fast cooling, representing the uneven temperature distribution which results distortion. The temperature, displacement and the curvature variation results are illustrated in Figure 5. During heat treating process, the two upper corners exhibit relatively significant displacement which depends on the selection of reference points. However, the curvature variation of each step illustrates the actual distortion which is independent of displacement. Displacement is not exactly distortion, while, the curvature variation can directly describe it. That means the area with big displacement may be mainly the result of distortion, or the result of other area which serves as a distortion source. For example, a plate is bent at the center line, the two ends will be greatly displaced, but the two ends remain flat. Under this circumstance, the displacements of the two ends are not caused by the distortion of itself, the real reason is the distortion of the center line, thus, the center line behaves as a distortion source. The bottom region of this blade casting is neglected because of constraints. It can be seen the area in green color is of big curvature variation at 5.5h which serves as a distortion source and caused the distortion at the upper corners. At 22.Ih, temperature distribution is uniform, so there is small and uniform distortion without a significant distortion source. During the beginning of cooling process, the red area serves as the distortion source because of the uneven cooling. As cooling progresses, this region expends and becomes more significant. As the casting is cooled below the start point

260

of martensite transformation Ms (32.8h), the distortion source becomes weaker as shown in Figure 5 (c). Therefore, as big temperature difference exists in casting, there is significant distortion resource. As the casting is in relative uniform cooling or heating conditions, there is no significant distortion resource. From Figure 5, it can also be seen that the displacement results are not consistent with the actual distortion. The distortion is actually the variation of displacement with spatial location, which can be depicted by the curvature variation. 1000 800 V ^600

I

£400 6 ω

H

200 0 0

10

20

30

40

50

60

time (h)

Figure 4 Thermal schedule and heating, cooling curves of the blade

(c) Curvature variation Figure 5 Temperature, displacement and curvature variation of the turbine blade during heat treatment process

261

The final displacement and curvature variation are shown in Figure 6. Finally the maximum displacement mainly exists at the two upper corners. The regions close to the two corners are of little bigger curvature variation, i.e., the distortion finally happens at the two corners. Therefore, the final displacement complies with distortion. The curvature variation of the center region of casting is little. Thus, the reference points for the final distortion comparison in traditional way in production can be selected in the area with no or less distortion, i.e., the dark blue center region of casting shown in Figure 6. The calculated displacement result of point A (as shown in Figure 5 (a)) at the upper right corner of this blade casting is compared with that of the measured of a similar blade casting, as shown in Figure 7. It can be seen that they are basically in agreement with a peak of about 60mm appeared during cooling. However, the direct validation of the curvature variation is hard for real blade castings. Therefore, it is necessary to do experiments with specially designed specimen for further validation.

(a) Final displacement (mm) (b) final curvature variation (radian) Figure 6 Final displacement and curvature variation of the blade

Figure 7 Comparison of calculated and measured displacement results 4. Conclusions Distortion prediction and control is of great significance during production of castings. Based on displacement results and the finite element model obtained from thermal stress analysis of castings, the surface discretization model STL format is acquired. The curvature and curvature variation of local surface area is calculated,

262

which is used to depict the actual distortion, the area with big curvature variation serves as distortion resource during heat treatment process. This method was applied into a heavy hydro turbine blade casting. The results show that the displacement is not consistent with distortion during heat treatment process; however, the final displacement is consistent with distortion and mainly occurred at the area close to the two corners of the blade. Curvature variation is proved to be a useful method for the depiction of distortion independent of reference points. Acknowledgement The project is funded by the project 2007BAF02B02 of the National Science & Technology Program in the Eleventh Five-year Plan Period and Major National Sci-Tech Project of China No 2009ZX04014-082.

REFERENCES 1.

Y. Chen, J. W. Kang and B. C. Liu: Proceedings of the 8th International Conference on 'Modeling of Casting, Welding and Advanced Solidification Processes', June, 1998, San Diego, USA,TMS, 771-778

2.

J.W. Kang, X G. Liu, Y.B. Duan and et. al.: 5th Decennial International Conference on 'Solidification Processing' , July, 2007 , Sheffield, UK .University of Sheffield, 554-557

3.

B.C. Liu, J.W. Kang and S.M. Xiong: Sci. Technol. Adv. Mater., 2001, 2: 157-164.

4.

Y. Song, Y. Yan, R. Zhang and et al : Proc Inst Mech Eng Part B J Eng Manuf., 2002, 216: 1123-1134

5.

Yool-Kwon Oh and Je-Se Choi: Advanced Materials Research Adv. Mater. Res., 2008,47-50: 1043-1046

6.

Sung-Mo Lee, Won-Jae Lee: J Mater Eng Perform., 2005, 14: 388-394

7.

Jinwu Kang, Haimin Long, Tianjiao Wang, Tianyou Huang and Baicheng Liu: Proceedings of The 8th Pacific Rim International Conference on 'Modeling of Casting and Solidification Processes', April 12-15, 2010, Incheon, Korea, 223-228

8.

Lugui Chen, Yong Ling, Xiuhong Kang, Lijun Xia, Dianzhong Li: J Mater Sci Technol,, 2008, 24: 364-368

263

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

The capability enhancement of aluminium casting process by application of the novel CRIMSON method Xiaojun Dai1, Mark Jolly2, Binxu Zeng3 '•2'3School of Mechanical Engineering, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK Key words: aluminium casting process, CRIMSON, melting, oxide film, Abstract The conventional foundry not only frequently use the batch melting where the aluminium alloys are melted and held in a furnace for long time, but also use the gravity sand casting process where the melted aluminium alloys are transferred using a ladle from furnace to pour station and are poured into a mould. During filling a mould the turbulent filling behaviour due to gravity is easily to make the oxide films on the surface of the liquid cracked and trapped into the liquid. Also the long exposing time of the liquid surface with the around air during melting and filling will increase the level of hydrogen absorption. All of the abovementioned factors are often the main reasons for casting defect generation. In this paper, a novel CRIMSON aluminium casting method is introduced which has a number of advantages. Instead of gravity filling method, it uses the single shot up-casting method to realize the rapid melting and rapid counter-gravity-filling mould operations which reduce the contact time between the melt and environment thus reducing the possibility of defect generation. Another advantage is the drastic reduction of energy consumption due to shortened melting and filling time. A simulation software, FLOW-3D, is used to compare the CRIMSON method with the conventional gravity casting process. A tensile bar case is used as a sample to simulate the filling process. Introduction Casting industry has been driven by the requirement of improving product quality and minimisation of the production costs. To satisfy these requirements both casting production process and energy efficiency play vital roles in the foundry. Selection of the proper casting process, facility and minimisation of the energy consumption will be the key factors for the casting enterprise to successfully compete against rivals in the tough market. In conventional foundries, the capacity of a typical aluminium alloy melt furnace usually is between the range of 100 kg and several tonnes. The liquid metal is held at about 700 °C in a holding furnace before it is transferred to a ladle and poured into a casting mould at pour station. It can take several hours or even several days for the liquid metal in a batch to be used up and any leftover metal is poured off to be re-used or scrapped for re-melting or refining in a secondary processing plant [1].

265

Quality issues can arise when the liquid metal reacts with hydrogen, oxygen and water in the atmosphere. An oxide surface layer is created when the melted aluminium alloy is exposed to the air. During filling a casting mould the turbulent filling behaviour of the liquid metal due to gravity is easy to make the oxide films on the surface of the liquid cracked and trapped into the liquid. Also the long exposing time of the liquid surface with the around air during melting, transferring and filling will increase the thickness of the oxide film on the liquid surface and the level of hydrogen absorption. All of these will result in layers of cracked oxide films, porosity and shrinkage which damage the integration of the micro-structure of the alloy, leading to degraded mechanical properties of the end product [2. 3]. In traditional casting industry the energy efficiency of a casting facility depends largely on the efficiency of its melting and heat treating performance. In association with the two performances, over 60 % of the total process energy costs are represented in a typical casting facility [4] where there are huge opportunities for metal casting industry to adopt the best energy practices which will provide the great energy saving potential. To ameliorate the current processes for increasing energy efficiency will have an important effect on reducing the production costs and promoting the competitiveness. For instance, by implementing some state of the art technologies such as the CRIMSON method in aluminium alloy casting will make use of such opportunities.

Figure 1 Schematic plan of the new casting process facility The researchers and engineers from University of Birmingham and a local company, - N-Tec LTD have co-invented a patent CRIMSON (Constrained Rapid Induction Melting Single Shot Up-Casting) method. The objectives to develop this method are to decrease the energy consumption and to meliorate the casting quality within light-metal casting industry. The methodology of the new method is that foundries, using an induction furnace, need only to melt the quantity of metal required to fill a single mould in a closed crucible rather than large batches that use unnecessary energy and create more rejects. As shown in Figure 1. the closed

266

crucible in the induction furnace, then, is transferred to a pour station and the melted metal is pushed up using a computer controlled anti-gravity filling method to fill the mould. Due to a characteristic of rapid melting, transfer and filling in the new method, the holding time of melted metal is minimised, the possibility of hydrogen absorption and formation of surface oxide film are decreased largely and in the mean time a huge amount of energy saving is achieved [5]. In this paper, the tensile test bar was used as a sample and a CFD simulation software (Flow3D) was used to simulate the casting filling process. Sand moulds for the new up-casting method and conventional gravity sand casting were designed to compare their filling behaviours using the numerical method. In addition the traditional melting process from one local company was investigated and it was compared with the new up-casting method. We mainly focus on the issues of quality and energy saving, other issues will be investigated later. The simulation results of filling the two different sand moulds were compared to see how the new method can avoid the turbulent behaviour during filling process. The calculation and analysis of energy consumption were completed to see what the difference between the current melting processes and the novel method. Thus, the quality issue and the potential energy saving for the new method can be found. Runner system design and simulation software Runner system design of CRIMSON method and gravity sand casting method A runner system for CRIMSON method is shown in Figure 2(a) where the system consists of runner, ingate, riser, feeder and six tensile bars. A runner system of gravity sand casting is shown in Figure 2(b) where the system consists of basin, down sprue, runner, ingate, riser, feeder and six tensile bars.

(a) CRIMSON

(b) Gravity sand casting

Figure 2 Structure of the tensile bar with different runner systems. Simulation software and conditions In order to simulate the liquid metal flow in different runner systems, a commercial Flow-3D CFD simulation software was used. A 'velocity magnitude' method of the Flow-3D code was

267

used to predict the flow behaviour of liquid metal during filling a mould. The simulation was implemented using a Workstation with 16.00GB RAM and eight 2.66GHz CPUs. For the simulation of the CRIMSON runner system, Finite Difference Method (FDM) was used to generate the mesh which includes about 170,000 control volumes (cells). The filling flow rate of 0.25 L.s"1 [6] and a pressure of 9 kPa were used. For the runner system of gravity sand casting, the mesh has about 240,000 control volumes (cells). Same main conditions for the simulation of both runner systems are: pouring temperature is 700 °C and the around atmosphere pressure is 1 atm (1.013xl05 kPa). Test facilities and casting sample for energy saving Test facilities CRIMSONfacility: the structure and layout of the novel casting process facility is indicated in Figure 1 .Where its functions and features are: • High power Induction furnace (275 KW): it is used to quickly heat and melt the metal to the required temperature. Each time a billet with required size and weight is put in; • Up-caster: when the crucible with the melted metal inside is ready, it is moved and cramped in the right position in Up-caster and a mould is located on the top of pouring position, a piston in the Up-caster will raise and push the melted metal in the crucible into the mould; • Computer-controlled operation board: the movement of the piston in Up-caster is automatically controlled by the pre-programmed computer program; • Mould transfer stop: after pouring, cooling down and solidification, the mould can be moved to the transfer stop, waiting for lifting and cleaning;

Figure 3 Schematic of the aluminium meltingfurnace in G&W LTD.

268

Melting facility of G&W LTD: Grainger & Worrall (G&W) LTD. is currently using one type of melting furnace (4 tonne, Figure 3) with combining melting process where the primary melting area functioning like a stack melter and gas is used to preheat and melt aluminium ingot, then the melted aluminium alloy flowing along an inclined channel to a refining area where an electric resistance furnace is used. The refined liquid aluminium alloy is held in the electric resistance furnace. The holding time for the furnace is up to 4-5 days. The overheating temperature of A354 aluminium alloy is 760 °C and the pouring temperature is 700 °C. Casting sample and mould selected for comparing the energy consumption Half of sand mould of the tensile bar is shown in Figure 4 which has also been selected to use novel method and conventional casting process from G&W to examine the difference of energy consumption. The mould with a runner system has a profile of 530 mm length x 390 mm width x 100 mm height and has a weight of 4 kg for filling metal [7].

Figure 4 Sand mould of the tensile bar Results and discussion Simulation For the simulation of runner system of CRIMSON method, the filling time is 6.08 seconds. The simulation took about 30 minutes. The velocity magnitude of the liquid metal during filling is depicted in Figure 5 where the filling velocity of liquid metal can be observed using the velocity scale.

Figure 5 Numerical simulation of runner system of CRIMSON method.

269

For the simulation of runner system of gravity sand casting, the filling time is 3.67 seconds. The simulation took about 20 minutes. The velocity magnitude of the liquid metal during filling a mould is described in Figure 6 where the velocity of liquid flow can be judged using the velocity scale.

Figure 6 Numerical simulation of runner system of gravity sand casting. From Figure 5, it is found that the maximum velocity of liquid metal flow during filling is 0.4 m.s"1 which is less than 0.5 m.s"'. The quiet flow behaviour of liquid metal in the runner system during filling was approved proper for avoiding the generation of trapped oxide films, porosity and other casting defects [8]. In addition the filling method using against gravity decreases the exposing time to the air which will reduce the opportunity of generating oxide films. From Figure 6, the maximum velocities of liquid metal flow in downsprue and in runner during filling are more than 1.0 m.s"1, respectively. These phenomena mean that the violent and turbulent follow flow behaviour will easily to crack the oxide films on the liquid and make them trapped into the liquid. Although the filter existed in the runner system can leach the coarse trapped oxide films from the liquid which depends on the size of the holes and effectiveness of the filter, the fine oxide films will still pass through the filter and remain in the liquid. After solidification, these remained fine oxide films will generate the defects such as porosity, shrinkage etc [2, 3]. In the mean time, the turbulent flow behaviour of liquid metal in the downsprue and runner will readily make the air or hydrogen entrapped into the liquid where the porosity or bubble will be formed which will impair the mechanical properties of the casting [9]. In addition the long transferring time from furnace to ladle and the usage of the traditional gravity filling method make the melted liquid metal exposed to the air for long time, which increases the opportunity of generating oxide films on the surface of the liquid metal. Based on the abovementioned discussion, it is found that the up-casting filling process that the new method used can drastically reduce the opportunity of generating oxide film on the surface of the liquid alloy and the potential time for hydrogen absorption. Therefore, the quality of the casting can be secured accordingly. In opposition to the new process, the filling process of the conventional gravity sand casting process has the violent flow behaviour which will easily make the oxide films on the liquid metal and the air entrapped into the liquid

270

which will generate different type of casting defects and damage the mechanical properties of casting. Energy saving The energy consumption of the two types of melting processes was investigated and the results and the calculated energy efficiency were recorded in Table 1. Melting process

Energy consumption

Energy efficiency

G&W (Gas+Ele)

27.86 MJ.kg-1 (Gas: 17.78 MJ.kg-' Eie: 10.08 MJ.kg1) 1.98 MJ.kg'1

Gas: 5.65% Eie: 1.70%

Normal energy efficiency of furnace Gas: 7-19% Eie: 59-76%

57.8%

Eie: 59-76%

CRIMSON Induction furnace

Table I Energy consumption and energy efficiency of the two different meltingfacilities As shown in Table 1, the thermal efficiency of the melt furnace at G&W for gas is 5.65 % and for electricity is 1.70%. The former is near the normal thermal efficiency of crucible furnace using gas (7-19 %). The later is far more less than the normal thermal efficiency (59-76%) of an induction furnace using electricity. This means that there is lot of energy loss for the current melting process at G&W due to the long holding time. Therefore, it is assumed that if the current long melting and holding process at G&W could be replaced by the new single shot melting method, 14 times more energy can be saved. It is estimated that 26 GJ.tonne"1 (7.2 MWh.tonne1) can be saved for producing every tonne of A354 casting alloys when using the new process and thus the melting cost will be significantly decreased. Conclusions and future works From the simulation results, the maximum flow velocity of the new CRIMSON method during filling a mould is less than 0.5 m.s"1. The total flow behaviour is quiet and stable which reduces the possibility of generation of oxide films and other casting defects. In opposition to the new method, the conventional gravity sand casting process has turbulent flow behaviour during filling a casting mould where the maximum velocities of liquid flow in the downsprue and runner are more than 1.0 m.s"1. The violent flow behaviour will not only easily make the oxide films on the liquid metal cracked and entrapped into the liquid but also make the air trapped into the liquid. These entrapped oxide films and air will generate different types of casting defects such as porosity, shrinkage and bubble etc which will damage the mechanical properties of the final casting. The investigation on examining the energy consumption and the melting efficiency of both methods has showed that the CRIMSON process is a novel method for reducing energy consumption. If the traditional foundries could use the new melting method instead of their traditional melting process, the estimated energy savings could be of the order of

271

26 GJ.tonne-1 (7.2 MWh.tonne"1) in this case. This could hugely reduce the production cost by about £546 pounds.tonne"1 (7.6 p.kWh"1) and would tremendously increase the company's competitiveness. To validate the simulation results, the experiment on comparing the mechanical properties of the tensile bar using different filling processes will be implemented in next stage. The other issues of the energy efficiency for the foundry will be considered too where not only the melting process is included, other relevant processes should be considered. Acknowledgement This research project is sponsored by the Engineering and Physical Sciences Research Council (EPSRC) of the UK under the grant of EP/G060096/1. Many thanks to the University of Birmingham, N-Tec LTD. and Grainger & Worrall Ltd for providing the experiment equipment and data. References [1] Casting method saves energy, 25, Mar 2009, Professional Engineering Magazine. http://www.profeng.com/areliive/2009/2206/22060076.htm [2] X. Dai, X. Yang, J. Campbell and J. Wood, 2003, Effects of runner system design on the mechanical strength of Al-7Si-Mg alloy castings, Journal of Materials Science and Engineering. A354. Pg 315-325. [3] X. Dai, X. Yang, J. Campbell & J. Wood, "The Influence of Oxide Film Defects generated in Filling on the Mechanical Strength of Aluminium alloy Castings", Materials Science and Technology, 2004, 20(4), 505-513. [4] R. Eppich, R.D. Naranjo., Implementation of Metal Casting Best Practices, Report of U.S. Department of Energy, 2007. [5] M. Jolly, Energy Saving in the Foundry Industry by Using the "CRIMSON" Single Shot UPCasting Process, 2010 TMS Annual Meeting & Exhibition, Febuary 14-18, 2010, Seattle, WA. [6] J. C. Gebelin, M. Lovis & M.R. Jolly, SIMULATION OF TENSILE TEST BARS: DOES THE FILLING METHOD MATTER?, Symposium on Simulation of Aluminum Shape Casting Processing, TMS2006 March 2006, Warrendale, PA, Eds Q. Wang, M.J.M Krane and P.D Lee. [7] X. Dai, M. Jolly, Potential energy savings by application of the novel CRIMSON aluminium casting process, Applied Energy (Accepted for publication) [8] J. Campbell, (1991) Casting, Butterworth-Heinemann, Oxford. [9] M. Divandari, 2001, PhD Thesis, 'Mechanisms of bubble damage in castings'. University of Birmingham.

272

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

Optimization of the process parameters and tooling improvement for the rheocasting of high quality aluminum components using the SEED process Chang Qing Zheng, Ehab Samuel, Florentin Laplume National Research Council of Canada, Aluminum Technology Centre (ATC-NRC), 501 boulevard de l'Université est, Chicoutimi (QC), G7H 8C3 Keywords: rheocasting, aluminum, SEED process, semi-solid, semi-solid casting, HPDC Abstract The automotive industry has leaned greatly towards the use of aluminum alloys by virtue of their strength and low density. Given this, the potential for aluminum use in the fabrication of vehicle parts has greatly increased. However, there are limited studies devoted to the improvement of the casting process. In the present work, the SEED (Swirled Enthalpy Equilibrium Device) rheocasting method, as developed by Rio Tinto Alcan in collaboration with the Aluminium Technology Center of NRC Canada (ATC-NRC), was analyzed by the authors in an attempt to optimize operating parameters (e.g. proper mold filling, slurry temperature, injection speed, etc.), which affect the final cast part quality. In many of the existing semi-solid casting processes which use billets as feedstock, for example, it is often found that the outer surface of the billets is contaminated. During the injection phase, a billet's external skin comes into contact with air and lubricant, and, as a result, becomes contaminated. The use of such a contaminated billet can often result in an increased rejection rate of cast parts. The SEED process, which uses heat extraction of the liquid aluminum alloy via mechanical agitation (swirling) in a confined cylinder to form the semi-solid billet on site, has already proven successful in producing sound aluminum castings having an excellent combination of strength and ductility. The resulting semi-solid billet, having a microstructure consisting of a-Al globules surrounded by the eutectic phase, is then injected into the cold chamber of an HPDC machine. Introduction Conventional shape casting processes such as pressure die casting, for example, offer both low cost and high productivity. The downside, however, is that pressure die-cast parts are generally prone to fabrication defects and limited mechanical properties [1]. One possible remedy is semisolid pressure die casting, which can yield quality parts comparable to those made with wrought processes; the cost of casting is comparable to using conventional pressure die casting. Indeed, semi-solid pressure die casting can eliminate casting defects and allow for additional improvement, such as the use of heat treatments, to the production process [2, 3]. Within the framework of a collaborative effort between Rio Tinto Alcan and ATC-NRC, studies were carried out to assess the performance of a 357 alloy designed for semi-solid pressure die casting and improve casting tooling design which, in turn, helps to improve mechanical properties; these studies are based on a new design of experiment (DOE) method that builds on experience, i.e. the Expectation-Maximization (EM) Method. The semi-solid billets were prepared using the SEED process [4], as shown in Figure 1.

273

Figure 1. SEED process of Rio Tinto Alcan |3]. This patented method [5] employs the heat extraction of the liquid aluminum alloy via swirling, in a confined cylinder. The resulting semi-solid billet is injected into an HPDC press to yield quality parts. A proper set of working parameters can lead to a uniform and globular microstructure which, in turn, can decrease slurry viscosity and improve die filling; likewise, an improper use of these parameters can result in defects, and reduced mechanical properties [6, 7]. Methodology In the DOE, the EM Method was used to keep the number of tests to a minimum. This consists of: (i) defining parameters (factors and responses) and objectives, (ii) conducting experimental tests and building operational domains, (iii) regressions and building models, (iv) validation and (v) multi-criteria optimization, which consists in maximizing a desirability function that expresses the importance of combined criteria in relation with targeted goals. A direct research method is then applied to find optimal solutions [8, 9]. The operational domain, in a multidimensional space, is modeled as an ellipsoid (Figure 2) and defined with continuous values, which discriminate between accepted tests (inside) and rejected tests (outside). This is not to say that the test points which do not fit inside the operational domain are discarded from the work, rather they are not included in the ellipsoid. However, they are still used in the overall model. Test points have coordinates (a, b, c, d, e...), which change with the number of operational parameters (i.e. factors) being considered.

Figure 2. Simplified view of the multidimensional operational domain, modeled as an ellipsoid.

274

Because this ellipsoid is a crude representation of the operational domain, perfect discrimination between all the accepted and rejected points is not always possible. Usually, limitations and constraints on the factors truncate the ellipsoid, contributing to a more realistic shape of the operational domain. For this work, the ΈΜ Optimization' software by EM Optimization International Inc. was used. Semi-solid 357 aluminum alloy wedge plate parts (Figure 3) were cast using the SEED process highlighted in Figure 1, with a 530-tonne HPDC press. These plates were subsequently subjected to a T6 heat treatment consisting of (i) a solution heat treatment of 540°C for 6 hours, (ii) a 20°C water quench and (iii) an aging step of 170°C for 6 hours. Following this, ASTM B557 round bar tensile samples, with a gage length of 25 mm and a gage diameter of 6.35 mm, were machined at pre-assigned locations along the plate length (Figure 4).

Figure 3. Schematic of the cast wedge plate.

Figure 4. Position of tensile samples and hardness test points on the cast wedge plate. For simplicity, we will only consider the results obtained for section A (Figure 4). In other words, the mechanical properties such as yield strength, ultimate tensile strength, etc. will all be measured from tensile samples taken at position A. Standard liquid metal treatment and control steps were applied to the aluminum in the furnace, e.g. chemical composition check, rotary fluxing, degassing, etc. Results and Discussion In this work, 13 parameters (seven factors and six responses) were considered. These are listed in Tables 1 and 2 for factors and responses, respectively. Figure 5 illustrates the scanned area used for measuring the average a-Al globule size. In this work, 55 castings were produced. Of these, 34 met the pre-established ranges. The distribution of these 55 test points is relatively uniform inside the operational domain (Figure 6). Regression models, based on the tests carried out, were established to quantify the effects of the working parameters on the responses.

275

Table 1. Factors and their typical range.

Table 2. Responses and their typical range.

Figure 5. Scanned section tensile sample for average width value (μιη) of o-AI globules.

Figure 6. Operational domain and test distribution (total of 55 tests).

276

Figure 7 demonstrates the degree of influence of the factors on a given response - in this case, the yield strength at section A. These effects are normalized, and shown in terms of being either positive (moving to the right) or negative (moving to the left). The greater a factor's influence (either positive or negative) on a response, the longer the red bar is in the histogram of Figure 7.

Figure 7. Factors and their relative impact on the yield strength (at section A). According to Figure 7, Sr concentration (factor 'F') has the most influence on the yield strength, while filling speed ('Β') has the least. Die temperature (factor 'D'), on the other hand, has no effect, as it is not shown in the histogram. There is an interaction between filling speed ('Β') and intensification pressure ('C'), as well as pre-filling speed ('Α') and drained mass ('G'). Figure 8 illustrates the behaviour of the yield strength response in more detail. The interactions demonstrate that the effect of factors C and A, for instance, are only significant when they are combined with B and A, respectively. In other words, the relationship is proportional, rather than linear (as in the case of Sr concentration) or squared (as in the case of the metal pouring temperature).

Figure 8. Yield strength at section A relative to filling speed, metal pouring temperature, drained mass, Sr concentration and die temperature. It can be seen from Figure 8 that the yield strength: (i) increases from 277 to 282 MPa as the filling speed increases from 0.2 to 1.4 m/s, (ii) decreases, but then increases, as the metal pouring temperature increases (note that the curve is not linear like the others, owing to the E2 relation shown in Figure 7), (iii) decreases from 292 to 269 MPa as the Sr concentration increases from 30 to 100 ppm, (iv) decreases from 287 to 270 MPa as the drained mass increases from 80 to 268 g, and (v) remains constant at all die temperatures tested. These observations can likewise be

277

made by considering the behaviour of any of the other responses (ultimate tensile strength, elongation, etc.) with respect to the effects of the working parameters. Histograms similar to Figure 7 are presented in Figure 9, for the remaining responses. It can be seen here, for example, that the Sr concentration greatly affects the percent elongation, a-Al globule size and density. On the other hand, the die temperature has little effect on the percent elongation or the hardness.

Figure 9. Relative impact of operational parameters (factors) on responses. As can be seen from Figures 7 to 9, the regression models derived from the multi-dimensional analysis show the degree of influence of the operating parameters (factors) on the alloy's mechanical properties, microstructure (by measurement of primary a-Al phase size) and porosity defects (by density measurement). The overall influence of the operational parameters on the responses studied in this work is fully summarized in Table 3. Table 3. Effects of operating parameters (factors) on responses.

According to Table 3, is can be observed that: (i) the prefilling speed has no significant effect on the majority of the responses, (ii) the filling speed has a significant effect on hardness, (iii) the intensification pressure has a significant effect on ultimate tensile strength and hardness, (iv) the die temperature has a significant effect on percent elongation and a-Al globule size, (v) the metal pouring temperature has a significant effect on yield strength, ultimate tensile strength and density, (vi) the strontium concentration has a significant effect on yield strength, percent elongation, a-Al globule size and density, and (vii) the drained mass has a significant effect on all of the responses, except hardness.

278

Multi-criteria optimization was used to maximize the mechanical properties (yield strength, ultimate tensile strength and elongation) all at once, in order to improve the performance of the parts being cast. One best solution point inside the feasible domain was found using the EM Optimization software, based on the models (Figure 10).

Figure 10. Optimization by EM software and the best solution found. The end point (i.e. best solution) has coordinates (A, B, C, D, E, F, G), which correspond to the optimized values of the working parameters. Subsequent to the maximization of the mechanical properties, 50 wedge plate parts were produced wiüi these optimized parameters, in order to provide the best possible solution (i.e. the best quality parts). In order to ascertain the quality of these optimized plates, once again, ASTM B557 tensile samples were machined and testing resumed. The results obtained for this optimized set of tensile samples are presented in Table 4. Table 4. Optimized parameters and results.

As can be seen from Table 4, the predicted and observed values of the mechanical properties (yield strength, ultimate tensile strength and percent elongation) are in good agreement. Moreover, it is clear from these results that the alloy demonstrates very favourable mechanical property values, in terms of strength and ductility. Conclusions 1. After having tabulated the effects of the working parameters on the responses, it was found that the yield strength is mainly influenced by pouring temperature, Sr concentration and drained mass. Moreover, the ultimate tensile strength is mainly influenced by pouring temperature, intensification pressure and drained mass, and the elongation is mainly influenced by die temperature, Sr concentration and drained mass.

279

2. After having carried out 55 tests, the optimum operating parameters for the 357 alloy in this work were obtained. With this information in hand, a new batch of 50 optimized parts were cast resulting in mechanical (tensile) property values which were in good agreement with the values predicted by the EM Method. 3. The average mechanical (tensile) property values (yield strength = 297±4 MPa, ultimate tensile strength = 350±3 MPa and percent elongation = 8.3±0.8%) obtained in this study demonstrate a very favourable combination of strength and ductility. These properties are attributed to the uniformly globular microstructures obtained using the SEED process, which result in high integrity castings and exceptional values of mechanical properties. Acknowledgements The authors acknowledge National Research Council Canada for the support and permission to publish. The authors wish to express special thanks to Mr. Dany Drolet, Ms. Geneviève Simard and Ms. Marie-Eve Larouche at the Aluminum Technology Centre of NRC Canada. Thanks also to our collaborators at Rio Tinto Alcan R&D in the SEED development Mr. Alain Lemieux. References 1. Major, J.F. and Richman, D.: Aluminum Automotive Castings - An Ever Expanding Role in an Increasingly Competitive Market. Proceedings of the International Symposium on Recent Metallurgical Advances in Light Metals Industries, Canada, Aug. 1995, 25-42. 2. Jorstad, J.L.: Semi-Solid Metal Processing; The High Integrity Die Casting Process. Die Casting Engineer, Jan. 2004, 48(1), 42-48. 3. Yurko, J., Fleming, M., and Martinez, A.: Semi-Solid Rheocasting (SSR™) - Increasing the Capabilities of Die Casting. Die Casting Engineer, Jan. 2004,48(1) 50-52. 4. Doutre, D., Langlais, J. and Roy, N.: The SEED Process for Semi-Solid Forming, in Proceedings of the 8th International Conference on Semi-Solid Processing of Alloys and Composites, Limassol, Cyprus, pp. 397-408 (2004). 5. Doutre, D., Hay, G. and Wales P.: U.S. Patent No. 6,428,636 Aug. 6, 2002. 6. Zheng, C.Q. and Simard, A.: Optimization of Casting Parameters on an Improved AA6061 Aluminum Alloy for Semi-Solid Die Casting, Journal: Advances In Light Weight Materials Aluminum, Casting Materials, and Magnesium Technologies, USA, Apr. 2010, SP-2294. 7. Pineau, F. and Simard, G.: Investigation of the Primary Phase Segregation during the Filling of an Industrial Mould with Semi-solid A357 Aluminium, S2P 2008 International Conference, 2008, 141(43), 635-640. 8. Galopin, M., Dao, T.M., Zheng, C.Q. and Hansquine, S.: A New Approach to Machinability Testing, Seminar and Applications Forum on a Systems Approach to Machining, Institute of Advanced Manufacturing Sciences Inc. (IAMS), Cincinnati, Ohio, USA, May 1993, 4-5. 9. Galopin, M., Dao, T.M., Zheng, C.Q. and Hansquine, S.: A Problem Solving Tool for Optimization of Welding Processes, 5th International Conference on Computerization of Welding Information, Golden, Colorado, Aug. 1994, 9-12.

280

Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SHAPED CASTINGS AND MACHINING John E. Wyatt1 & John T. Berry2 'Mississippi State University Instructional Systems & Workforce Development, Box 9730, MSU, MS, 39762, USA Mississippi State University Mechanical Engineering, Box 9552, MSU, MS, 39762, USA Keywords: Machinability, Machining Sequencing, Residual Stresses Abstract This paper will discuss three aspects of the machining of shaped castings: a) machinability, b) machining sequence, and c) residual stresses, the principal thrust being that of machinability. Currently there is a growing movement to reconsider the whole process of machining from the standpoint of fracture mechanics. This has resulted from the groundbreaking work of Atkins at Reading University (UK), which is relevant to cast alloys which invariably are multiphasic materials. Although secondary and tertiary phases have been blamed for tool wear, they are in fact intrinsically tied into the cutting process. The premise of Atkins is that ductile rupture takes place near the tool tip. These ductile tears manifest as dimples associated with the secondary and tertiary phases. Although some secondary phases will cleave (silicon particles in AISi type alloys) some will decohere from the matrix leaving dimples. The authors will describe some experiments confirming the theories of Atkins. Introduction Ever since F.W. Taylor wrote his treatise "On the Art of Cutting Metals" in 1907 [1], researchers have sought ways of creating chips more efficiently, as well as of modeling the chip formation process. There have been many worthy contributions, including the studies of Ernst, Merchant, and Shaw for example [2, 3, 4]. However, throughout the past century insufficient fundamental thought has been given to the surface that is generated by the machining process. The surface seems to have been regarded as a by-product of the machining process with the chip being the principal product. The reality of this is that the chips are in fact the by-product and the surface generated is the major product of machining. The Current Understanding of the Chip Separation Process With reference to cast and weld-repaired components, the actual mechanism regarding the process of metal removal, especially the act of separation, has not been at the center of metal cutting research in the past. However, recent work has emerged that has shed new light on the surface generation basics. Atkins [5, 6] has suggested an alternative approach than heretofore regarding the way in which metal cutting relates to fracture mechanics. Using Scanning Electron Microscopy (SEM) based approach experimental evidence has been presented by Subbiali and Melkote in the microcutting of 2024-T3 aluminum alloy [7], Melkote et alia[8, 9] have justified this new approach and have established that with small uncut chip thickness values (15 - 105 um) a ductile tear is initiated ahead of the tool nose which then propagates into either the chip or the workpiece, as illustrated in Figure 1. Furthermore, it is likely that this tear could possibly be affected by the tool geometry and other

281

cutting variables. Further discussion with the proposer of this new concept [6], suggests that this tear is present in all forms of machining. This has only recently been confirmed by work at MSU examining larger uncut chip thicknesses (250-500um). Figures 2 (a), (b), (c), and (d) present evidence for the existence of ductile tears, confirmed by the presence of classical "dimpling" phenomenon associated with secondary and/or tertiary phases. Engle and Klingele demonstrated that voids could be formed through ductile tearing that produce dimples associated with constituent phases or inclusions. This is illustrated in Figure 3 [10]. As the nascent chip forms, aided by the ductile tears referred to above, the top half of each tear will be incorporated into the underside of the chip. The lower surface of these tears will pass under the nose radius of the tool, forming part of the tertiary shear zone [7] as illustrated in Figure 4. Assuming that a "dimple" escapes with the chip, its associated nucleation feature (e.g. inclusions or constituent particles) will in turn be incorporated into the newly generated surface. This latter feature is likely to become a nucleation point for fatigue cracking or abnormal component wear in contact with subsequent mating surfaces. It should be noted that the tool tip is the last part of the tool that is in contact with the surface of the workpiece and thus plays the final part in the generation of that surface, very possibly burnishing over the remains of the ductile tear, see Figure 2(d).

Figure 1. A depiction of the presence of a ductile tear ahead of the tool nose. The chip formation process is causing the tear to open. The upper part of the torn surface generated by the cutting operation escapes partly with the chip while the lower portion enters the tertiary shear zone and is then burnished over into the final generated surface.

(a)

(b)

282

(e) (d) Figure 2. SEM Photomicrographs showing dimpling and cracking, (a) Shows cracks appearing on the underside of the chip and at the interface, (b) Shows the dimples inside the crack on the underside of the chip illustrating that these 'dimples' appear throughout the removed material, (c) Shows the interior of the crack at the tool nose interface, (d) Shows the possible remnants of a ductile tear after being burnished into the machined surface. The material is wrought AL6061 in the T6 condition

Figure 3. Engel and Klingele's depiction of ductile tearing in ductile materials by void formation [10].

Figure 4. A depiction of the shear zones in metal cutting This leads to the possibility that the surfaces generated by machining are affected by a combination of the above ductile tear, or possibly a brittle crack, produced by the chip formation process, and the burnishing effects of the tool tip at the tertiary shear zone. These potential effects are poorly understood, especially in as cast materials and weld-repaired surfaces. Properties likely to be affected would be those associated with the superficial state of stress and

283

characteristics of the refurbished layer. Such characteristics include fatigue, corrosion and oxidation resistance. Finally, these effects are especially germane to as cast and weld-repaired surfaces which are often known to contain features which would encourage ductile fracture (intermetallics and non-metallic inclusions, as well as constituent particles in age hardened materials). Thus surfaces so generated are likely to be heavily populated by what were once 'dimple' rupture zones especially in materials containing a high fraction volume of injurious inclusions. Consequently, the subsequent mechanical behavior of the machined surface will be strongly affected by this abundance, along with the stress-state obtained there. As mentioned above, this will be especially relevant to fatigue, corrosion, and oxidation resistance. In view of the changing picture of the manner in which chip formation is initiated and the special role of local ductile failure in the tool nose region of the workpiece, it is highly desirable that parallel research is undertaken in cast light alloys. Since the majority of such alloys are multiphasic, the effects of second phase population and distribution upon both chip formation and subsequent effects of this upon fatigue crack inititiation on the generated surface are ripe for study. These effects may be of special importance where weld-repair is permitted Post Machining States of Stress Residual stress in machined components is critical in determining both the wear and the fatigue characteristics ofthat component [11, 12]. It is not just the levels of stress that are critical but whether the stress is tensile or compressive. A machined surface that has tensile residual stress induced by the machining process would be prone to fatigue cracking, whereas a machined surface with compressive residual stress would be resistant to fatigue cracking [13]. It is widely accepted that the initiation stage of fatigue cracking is facilitated if the residual stresses present at the machined surface are of a tensile nature. Consequently, post-machining operations such as shot peening have been employed to provide a compressive stress field at the surface. The experimental study of residual stress associated with conventional machined surfaces goes back at least 50 years [14]. Much of the early work was concerned with turning operations in an attempt to simplify the subsequent analysis of the mechanical state obtained. Typical milling operations used in finishing castings will generate similar residual stress patterns to those observed when turning, as shown in Figure 5 [14]. The sense (sign) of the superficial residual stresses determined in this study appeared to depend upon the composition and thus hardenability aspects of the various steels examined. Figure 8 shows that the residual stress at the surface is also dependent on tool sharpness which gives rise to the varying residual stress patterns.

Figure 5. The effect of wearland on residual stress in face milling

284

Unfortunately, such stress pattern determination is often expensive and involved. However, a new low cost, readily executed method of estimating superficial residual stress directly has recently been investigated at MSU. The technique involves the change in microhardness indentation size and spacing associated with the removal of residual stresses [15, 16]. Turning to the question of predicting such residual stress patterns by computational techniques, in many previous model experiments some authors have assumed an initial state of zero stress and strain in the elements concerned. The initial state of residual stress in the as-received material or component must also be taken into account in subsequent machining operations. Although the machining operation might well remove layers affecting residual stresses which can cause component distortion, it could later superimpose its own residual stress profile on the surface that the machining process has generated. Consequently, the historical progression of mechanical states must be taken into account if effects of subsequent behavior in monotonie or repeated loading are to be predicted. Much can be learned of the effects of residual stresses upon fatigue life. Work on wrought aluminum alloys by SAAB revealed the effect of high speed machining on the fatigue resistance of an aluminum alloy workpiece under pocket milling operations [17]. Fatigue test data showed that first a decrease in fatigue resistance was observed when the cutting speed increased above the conventional speed level, 100 m/min, but then an increase occurred when the cutting speed was raised towards 3,000 m/min. The minimum resistance appeared to be in the speed range of 500 to 1000 m/min. The fatigue life of the specimens was also dependent upon the cutting mode employed. In reviewing the SAAB investigation, the Swedish Defense Research Agency [18], presented an S-N curve which indicated a serious reduction in notch fatigue life when utilizing high speed milling. Although the curves were of similar shape the high speed machining curve at a cutting speed of 1800 m/min was significantly lower than that of the conventionally machined test piece. The tests were conducted on notched specimens excised from plate (7010-T7451). The exact mechanisms and cause-effect relations that produce residual stresses and surface material properties arising in machining are thus not well documented. Certainly, windows of opportunity exist in which high speed machining can be exploited to provide excellent fatigue lives but these areas have not yet been mapped [17, 18]. Such effects can be expected to be present in weld-repaired surfaces of the type described earlier. Using MSU's low cost residual stress technique, the authors original results as seen in Figure 6 show the corresponding relationship between the fatigue life results for face milling with superficial residual stress measurements on aluminum alloy bar machined over a range of cutting speeds. The residual stress measurements were made using the newly developed method at MSU. The correspondence is indeed remarkable and lays the background for this part of the proposed work [15].

285

Cutting Spc»ei (m/minì

-2000 -1050

8»

1Q0O

Î5Q0

/»DO

2900

2030 i

-V

4000 i :

SQ30 ·

0

506

1000

1500

200O

250C

Cutting Spe«4 (mimili)

(a) Measured Superficial Residual Stress [17]

((,) Fatigue life curve

Figure 6. Correlation between residual stress and fatigue life for face milled [15] Whether these effects involve local reprecipitation or overaging phenomena or features associated with burnished-over remains of ductile tears, is also a moot point, and requires further investigation. In summary, it is usually agreed that compressive surface residual stress improves fatigue life and tensile residual stress decreases it. The literature reveals that residual stress has very much more complex effects on fatigue compared to the general belief. There is thus much to be done before residual stress affects on component life are fully understood. This is especially true for cast components as their initial state of stress needs to be known as well as the microstructure of these layers before a machining strategy can be formulated. In summary, work on the high speed machining of wrought alloys, especially those of aluminum, has established that the superficial residual stress pattern is at least in part responsible for significant variation in their fatigue resistance. The relative importance on tool geometry and cutting conditions, in particular cutting speed should be determined for cast light alloy components where extensive machining is undertaken and where these components are subject to repeated loading. The time is ripe for such investigation. Conclusions Although the primary direction of research interest in the science and engineering of metal cutting over the last half century has been centered upon either the efficiency of removing chips from the workpiece or upon the problems of tool wear improvement, both especially at high cutting speeds, precious little attention has been directed to the nature of the surface so generated. This is important for both wrought and cast materials. One exception has been the investigation of the state of residual stresses at the generated surface, something which is closely linked to fatigue crack initiation. However, much of the literature concerns wrought rather than cast light alloys. Additionally the overall pattern of residual stress, which affects component distortion, of special interest in the assembly of cast components, has not received the attention of investigators in the area of metal cutting. Recent research by Atkins, Melkote and the writers has suggested that the presence of second (and no doubt tertiary) particles affect both the initiation of chip formation as well as the initiation of fatigue cracking in the generated surface. The latter, of course, will also be

286

profoundly affected by the magnitude and sense of the superficial residual stresses implanted or modified by machining. The writers point out that although the machinability of cast alloys of all types is at last commanding appropriate attention, research is badly needed on the specific mechanisms of chip formation and its impact on the nature and properties of the surfaces generated in these materials. Acknowledgements The writers would like to acknowledge the support received from the Office of Research and Development of Mississippi State University, and from the Coleman endowment, for the support of their work. Helpful discussions with Dr. Schreyes Melkote of the Georgia Institute of Technology are also gratefully acknowledged. Finally, the contributions of several senior undergraduates must be recognized. References 1. TAYLOR F.W., 1907, On The Art of Cutting Metals, Transactions of ASME, Vol. 28, USA 2. ERNST H., 1938, Physics of Metal Cutting, Cincinnati Milling Machines L™ 3. MERCHANT M.E., 1945, Mechanics of the Metal Cutting Process, Journal of Applied Physics, Vol. 16, pp.267 - 275, May 4. SHAW M.C., 1984, Metal Cutting Principles, Oxford University Press (U.K.), ISBN 0-19859020-2 5. ATKINS A.G., 1974, Fracture Toughness and Cutting, Int. J. Production Research, Vol. 12 (2), pp. 263-274 6. ATKINS A.G., 2003, Modelling Metal Cutting Using Modern Ductile Fracture Mechanics: Quantitative Explanations for some Longstanding Problems, International Journal of Mechanical Sciences, Vol. 45, pp. 373-396. 7. SUBBIAH S., MELKOTE S.N., 2007, Evidence of Ductile Tearing Ahead of the Cutting Tool and Modeling the Energy Consumed in Material Separation in Micro-Cutting, Journal of Engineering Materials and Technology, Vol. 129, pp. 321-331, April 8. MELKOTE S.N., 2008, Private Communication 9. ALTINTAS Y., 2008, Private Communication 10. ENGEL L„ KLINGELE H., 1981, An Atlas of Metal Damage, Prentice-Hall, Englewood Cliffs, NJ, p. 41 11. LIU C.R., LIN Z.C., BARASH M.N., 1984, Thermal and Mechanical Stresses in the Workpiece During Machining, High Speed Machining, Presented at the Winter Annual Meeting of the ASME, USA

287

12. G. ARNDT, 1971, On the Study of Metal Cutting and Deformation at Ultra-High-Speeds, Proceedings of the Conference of Production Science Industry, Vol. 30, pp. 30-41 13. KALPAKJIAN S., SCHMID S.R. 2001, Manufacturing Engineering and Technology - Third Edition, Addison - Wesley Publishing Company (U.S.A.), ISBN 0 - 201 - 36131 - 0 14. FIELD, M., KAHLES, J.F., 1964, Surface Integrity of Machined and Ground High Strength Steels,OM\C Report 210. 15. WYATT J. E. & BERRY J. T., 2006, A New Technique for the Determination of Superficial Residual Stresses Associated with Machining and other Manufacturing Processes, Journal of Materials Processing Technology, Vol. 171, pp. 132-140 16. BERRY J.T., WYATT J.E., 2005, A Low Cost Method for Determining Superficial Residual Stresses as Applied to Machined Surfaces, US Patent No. 6,934,642, Issued August 23 r i 2005 17. ANSELL H., 1999, Fatigue and Damage Tolerance Aspects of High Speed Machined Airframe Parts, Meeting of the International Committee on Aeronautical Fatigue, July, Bellevue, WA. 18. BLOM A.F., PALMBERG B., 2001, A Review of Aeronautical Fatigue Investigations in Sweden During the Period June 1999 to May 2001, Swedish Defence Research Agency, FOI, The Aeronautics Division, FFA, Sweden, FOI-R-0138-SE.

288

Shape Casting: The 4lh International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

THE ESTIMABLE VALUE OF 'CLEVER' EXPERIMENTS John T. Berry Mississippi State University Mechanical Engineering, Box 9552, MSU, MS, 39762, USA Keywords: Casting Shape, Chilling Power, Experimental Validation Abstract The inexorable growth of computational modeling in materials science and engineering has been associated with a serious decline in the number of 'clever', or at least genuinely useful experiments Experimental evidence, if provided, is often second-hand and merely placed there to validate a model, thus not stimulating further study. Many established concepts in materials science have been overturned as a result of careful experiments. These experiments have often nucleated new concepts and have started a sequence of experiment pacing theory and viceversa. In his long career, the writer has witnessed many examples of this pacing effect, some of which he lists. Several of these resulted in viable industrial processes. He describes unpublished work concerning the heat extractive capacity of molds which is relevant to current processing developments. The experiments concerned spawned further analytical and computational work, providing examples of the said pacing effect. Introduction As in many areas of science and engineering, progress in materials processing research, especially in the solidification area, has been governed by experiment pacing theory and viceversa. Often this progress has been conditioned by what Albert Einstein termed 'the delicacy of observation' (1916). In the past, countless seemingly well accepted theoretical concepts have been questioned and subsequently rejected through the results of well-planned and conducted experiments that were properly interpreted. Certain of these experiments have in turn led to the formation of new theories, which have been proven correct, often modified or even themselves rejected by further experimental work. A few have perhaps resulted in a new process or product. Perhaps a prime example of this 'pacing' effect is one which involves the concept of the dislocation 'pile-up' in low-carbon ferritic steels and its connection to the yield discontinuity. Subsequent research employing the transmission electron microscope (TEM) indicated that dislocation pile-up would only occur in low stacking-fault energy (SFE) materials. Austenitic stainless steels thus exhibit such features, whereas low-carbon ferritic steels, the subject of much previous analysis, do not! Consequently, other theories based on blocked slip-bands and dislocation multiplication became favored as the basis for alternative explanations of yielding and brittle fracture in mild steels.

289

The majority of the major contributions of the last century, and indeed those of the present one, contain examples of experiment and theory pacing each other. During his professional career the writer has witnessed many such examples : • •

The efforts of US and Canadian scholars to enhance understanding and knowledge of the morphology of the freezing front (Winegard 1964) The evolution of the single crystal gas-turbine blade (Kear at alia 1969) The development of controlled-rolling and the rise of the versatile microalloy steels (Microalloying 1975) The confirmation of the damaging effects of entrained oxide bifilms during melting and casting of light alloys (Campbell 2003) The recognition of the contribution of fracture mechanics and the role of second-phase particles in the metal-cutting process (Atkins 2009 together with confirmatory experimental work by Subbiali and Melkote (2008) - see also this conference (Wyatt and Berry 2011))

The writer's initial contact with solidification related research occurred towards the end of a period where continuing experimental work, often performed by practitioners, questioned the pioneer contributions of Nicolai Chvorinov on the solidification times of shaped castings (1938, 1940, 1951, 1953). In this instance that author had performed both analytical and experimental work which was questioned by experiment. At the time of Chvorinov'sl953 publication, much progress in interpreting this work and furthering understanding had taken place in the US (Pellini, 1952,1953) The sequence of the pacing phenomenon which followed is the principal subject of this paper. Certain aspects of this sequence have considerable bearing in certain contemporary developments in the technology of shaped casting production. Clever Experiments And Their Effect On Scientific Progress Although the celebrated Chvorinov rule (1938, etc) relating casting solidification time to the square of the volume to surface area ratio was apparently anticipated by Farquar (1920), as pointed out by Ruddle (1971), it has proved to be of considerable utility to practicing steel foundry personnel, and even of help to modelers in the preliminary design of feeding systems. The volume to surface area ratio or feeding modulus, has also proved useful in other applications, both in the avoidance of solid state embrittling reactions in heavy section steel castings (Monroe and Huff, 2010) and in the control of nodule and cell count in ductile irons (Fras and Lopez, 2010) The original experiments of Chvorinov, first reported in English in 1938 and then later detailed in Czech (1951-1953), wisely covered a wide range of sand-casting sizes in low-carbon steel. They included plates ('Deska') covering sizes from 10x400x400 to 200x1800x2400 cm and included a large very large casting ('Sabota o vaze') of 65 tonnes ! Also included were the results of Briggs for some spherically shaped castings, as well as cylinders ('Valec'). After the original publication of Chvorinov's work, a number of objections, mainly by practitioners, arose saying that the index of two in the proposed rule just did not work in practice. Thus in 1952 the writer was tasked with looking into this proposed relationship, in particular into the extent the mold material and casting shape affected matters. This research took place under the guidance of his doctoral supervisor, Dr. Voya Kondic and a Technical Sub-Committee of the

290

Institute of British Foundrymen (Now the Institute of Cast Metals Engineers, University of Birmingham, UK.

ICME)

at

the

Interestingly, a prominent member of the Sub-Committee (TS46) was Mr. Ronald Ruddle, author of the classic text 'The Solidification of Castings' (1957). The study study was both experimental and to some extent analytical. Several of the findings were reported in TrAFS in 1959 (Berry, Kondic and Martin). Although the experiments performed do not fall into the 'clever' category, they can be regarded as distinctly useful and also of relevance to contemporary developments in certain aspects of mold chilling power, both as far as cooling fins and ablative casting are concerned. In addition to examining mold chilling power, the study looked at casting size and end-effects. The aspects studied, especially end-effects, had probably led to some of the objections raised by practitioners to the original Chvorinov relationship. Much of this work is to be found in Berry, Kondic and Martin, 1959 The interaction of casting size and mold moisture content was of particular interest, both a drying-out effect seen at heavier casting thicknesses and an incredible enhancement of chilling power at small thicknesses. Perhaps a review of some unpublished experiments and just how they led to the development of new ideas, thus illustrating the pacing effect referred to earlier, would be appropriate. Some Useful Experiments And Their Consequences In the abstract to this paper the writer touched upon the importance of validating experiments to justify computation based conclusions and the pacing effect this has on theoretical and computational progress. It is worthwhile to point out that two of the earliest large scale investigations of computer applications to solidification problems - the NSF funded CADCAST program - led by Prof. Robert Pehlke and the writer, together with the previously conducted AFS funded program at the University of Michigan (UM) contained experimental validation at every stage, either from the thermal analysis of castings poured at UM, or by members of the AFS monitoring committee. (See progress reports contained in Trans. AFS) This type of Tnhouse' validation, which represents an ideal situation, is hardly seen today. Returning to the question of the early industry based objections to the Chvorinov rule, it appeared to the writer that many of them clearly arose from end, or corner-effects. It will be recalled that the derivation of the rule was based on the assumption of semi-infinite planar heat flow occurring during the solidification of pure metals or short freezing range alloys. Ruddle had also recognized their importance and had provided a partly experimental, partly analytical method of accounting for effects on solidification time (Ruddle and Skinner, 1951). The writer subsequently used this 'corner-correction' to indicate why the rule was so often refuted, itself an example of experiments pacing theory. Figures 1 to 4 show the results of experiments employing the Ruddle/Skinner correction to the surface area for two eutectic alloys ( 11% Si-Al and 8.5% Al-Cu). This in turn inspired later work of an analytical or computational nature in the nineteen eighties. The PhD dissertation of C.S. Wei (Georgia Tech, 1982) examined how a combination of thermal

291

properties, as well as the included angle of the corner affected the strength of any such correction. One important conclusion was that corner effects were not as strong in low -carbon steel castings as for those in pure aluminum. A possible ramification of this is the relatively small deviation of experiment from prediction in the original Chvorinov plot. A subsequent development, in this case where theory was pacing experiment, led to the design of a test casting to demonstrate the effect of included corner angle. That test casting has since been used in teaching many 'generations' of undergraduates by the writer (See H. Huang, PhD dissertation, University of Alabama, 1982) Yet a further spin-off of this pacing sequence was the interesting concept of replacing the enmeshment of the mold with an interfacial heat-flux map which could be applied to the casting surface. Eisuke Niyama introduced this concept in 1977 applying this approach to a onedimensional semi-infinite mold. In 1981 the writer suggested that this might be applied to more complex geometrical features.(Berry, 1981) Subsequently, Hansen, Berry and Wei, in 1983 described their proposals for the Q-Method. Franklin, and Moosbrugger both provided experimental work elaborating on this method and its potential application (Master theses at Georgia Tech, respectively 1983 and 1985) It was envisaged that this approach of eliminating the thousands of nodes involved in the discretization of the mold would be of great benefit in telescoping computational time. Dantzig and his students at the University of Illinois neatly packaged this concept by designing software (SPIDER) that literally climbed around casting features determining their curvature and ascribing a suitable heat flux expression thereby. They also looked at the question of variation of mold chilling power (Dantzig et alia, 1985) Alas, progress in this general area of interfacial heat flux descriptors was 'Leap-frogged' by the rapid growth in the automated and virtually instantaneous enmeshment software together with the astonishing advances in computational speed. However, the extent of corner and curvature effects, which are driven by the local variation in the interfacial temperature gradient, are fascinating in terms of thermophysical properties, metal superheat, as well as geometry. Figure 5 shows the effect of time on the magnitude of the interfacial temperature gradient at a rightangled three dimensional corner as measured by Franklin for an aluminum alloy sand casting. (1982). The extent of deviations from the Chvorinov planar solution were also shown schematically, Figure 6. Moosbrugger, Berry and Wei, 1986 summarized much of the work in this area. Thus the question of corner and shape effects on the Chvorinov rule are now reasonably well understood. However, a more recent attempt to combine effects of both volume and shape more compactly into one simple equation are especially noteworthy and may well eventually displace the use of the original rule in approximate calculations of solidification time (Tiryakioglu, 1995). Turning to the question of mold chilling capacity and early predictions of solidification times, the sparsity of thermal properties at that time (1950s) did undoubtedly present problems. Chilling capacity was one aspect which the IBF sub-committee cited earlier expressed a particular interest.

292

Three methods were then in use to distinguish between thermal effects of molding media: (a)

(b) (c)

The Chvorinov Curve - Fit method, where the dimensionless temperature in the mold at a specific location is plotted against the quotient of the distance from the mold-metal interface divided by the square root of twice the elapsed time. A set of master curves was then used to assign a value of the mean temperature diffusivity . Figures 7 and 8 present data for a dried coarse-grained clay-bonded sand compacted to two distinct ramming densities. The ordinate of the plot is the dimensionless mold temperature (See Berry, 1954) An entirely analytical method (Russell-Gittus) where the thermal conductivity of the sand mold was calculated from the properties and fineness of its components. (See Berry, 1954) A mold-calorimetric method where measurement of the solidification time of a slab of known geometry was used, along with known thermal properties of the casting medium , to calculate the mean heat diffusivity of the mold. (See Berry, 1954)

Table 1 contains results for a variety of sands, in one case for two different compaction densities. The data are presented as mean values of heat diffusivity. Although the agreement between the results employing each method varied, the data points to the fact that compaction (i.e. ramming density) and grain coarseness are important parameters. These methods might well be reexamined in terms of current practice as well as a means of obtaining accurate temperature dependent of thermal properties for use in modeling. A further aspect of mold chilling and its effect on deviations from the Chvorinov rule which was of interest to the investigating committee was the effects of moisture and/or combustible material present in sand molds. It is worthwhile re-examining these observations in the light of the use of chill-fins, as well as of ablative casting and lost foam casting developments. Perhaps the most dramatic of the experimental results which surfaced at that time was the combined effects of section size and moisture content on chilling power. Figure 9 provides evidence of the power of the presence of moisture in enhancing chilling power of both a coarse grained sand (14-28) and one of a fine grained nature (H). The average fineness levels were 25 and 150 BSS respectively. This is especially true for the case of the small section-size casting (6mm) poured in the moist fine sand. The casting medium was again the 8.5% Al-Cu alloy. (Berry, 1954) This is of special relevance to the use of chill fins and to a lesser extent to ablative casting. Although water and molten metals do not mix, clearly under certain controlled circumstances the results are highly beneficial rather than explosive! At the other end of the scale, where heavy sectioned castings are involved, especially thicknesses greater than 50mm, a distinct 'drying-out' effect was observed. Figure 10 indicates how for castings solidifying at temperatures associated with cast irons and steels this effect becomes manifest. (Berry,1954) Clearly, the permeability of the molding medium would need to be taken into account in attempts to model this phenomenon. In addition to moisture driven effects, the writer was asked to look into the combined effects of moisture and combustible mold additions. At that time coal-dust additions were common in molding practice in the UK. On this side of the Atlantic the literature contains reference to what

293

were termed 'sea-coal' additions. This was a term which had crossed to the US in previous centuries but had virtually been forgotten in the home country where it had originated. The principal reasons were for such additions were related to the improvement of surface quality in iron castings. The exact mechanism involved was much debated at the time. Other combustible additions, such as wood flour were also common at this time. The general observation was that such additions controlled the appearance of the Veining" defect. The results of parallel tests embracing dry sand, green sand and a 'black', sand containing coal-dust each based upon use of the same molding and casting media were somewhat inconclusive. However, there was some indication that the chilling power enhancement brought about by the presence of moisture may have been more effective than convective effects of combustion of the coal-dust. (Berry, 1954) Later work employing an organically bonded silica sand into which aluminum A356 was poured, confirmed this suspicion that combustible materials may, in fact, provide an insulating effect. Figure 11 shows the interfacial temperature gradient (which drives mold chilling power) plotted against the reciprocal of root time, as in figures 6 and 7. Current simulations generally employ standard data bases of thermophysical properties, for example those values quoted by Fras and Lopez (2010). However, it is perhaps appropriate to ask whether the building in of data-shifts reflecting some of the above effects, especially where chill-fins are applied in green sand molding, might be undertaken. Discussion The opening statement of this presentation referring to the decline in the number of truly 'clever' experiments, or even useful ones, may have surprised members of the materials community, especially if one turns to TMS or AFS publications. However, there are now many members of our group as it exists today who were not trained in schools of materials science and engineering where experimental assignments are still commonplace. It is the writer's belief that our colleagues from other disciplines could benefit by emulating this tradition, if only in part. It has been pointed out that several pioneer investigations of the various factors influencing the progress of solidification and of heat flow into the mold were greatly enhanced by the pacing effect of experiment on theory and vice-versa. If we examine the trends in various university curricula we see that many engineering programs have de-emphasized laboratory activity and industry contact. Indeed, it has been alleged that the graduates of certain PhD programs 'don't even know how things are made' (Wyatt, Altan and Berry, 2009). Despite the praiseworthy efforts by Professor Ashby and colleagues in advocating a 'top-down' approach to how materials and shaping processes are intimately connected with the design process (Ashby, 2011), many young PhDs entering the ranks of academia have had little experience of designing and building components, or even of conducting experiments, quite apart from exposure to industry. Graduates of co-operative programs, in addition to those of certain technology curricula, are perhaps exceptions. This is not to deny the significant contributions to modeling of the casting process by computer scientists, mechanicians and graduates of similar disciplines. Indeed, this very series of symposia has been greatly strengthened by their presence and has further enhanced mutual understanding.

294

There is no reason why a similar process of bringing together individuals as undergraduates drawn from the various disciplines should not be encouraged. There is little doubt that the interuniversity teams involved in motor sports and human-power type projects benefit immensely in this aspect - the majority of our students invariably crave the 'Hands-on' approach to learning. This presentation has attempted to show how experiment and analysis, and hence computation, have paced each other and have resulted significant progress in materials processing and even actual product development. Conclusions There are many examples in the broad area of materials processing and eventual product performance, where the mutual stimulation exhibited in both experiment and analysis have proven valuable. A number of specific examples have been cited in the area relating to shape casting production. In particular the influence of casting shape and thermophysical property variation on freezing progress have been exampled. For this process to continue and grow it is essential that engineers and technologists are made aware early in their education and training of the important role played by 'clever', or at least 'useful' experiments. Acknowledgements At the end of a long career it is impossible to cite the very many individuals, mentors, colleagues, industry friends, funding agencies and students who have influenced that career, especially since there is a great likelihood of leaving out and thus disappointing them. 1 would, however, like to acknowledge the individual who not only assisted in putting together my dissertation, but also gave unflinching support and contributed boundless common sense and sound judgment throughout that career, my wife. Nomenclature A

Surface area

V

Volume

t

Time

T

Temperature

x

Distance

R

Volume/Surface area

R'

Volume/Effective surface area

b

Mean heat diffusivity, Vkpc

295

k

Thermal conductivity

p

Density

c

Specific heat

Subscripts s

Solidification

x

Location with respect to mold/metal interface, i

o

Initial, with respect to mold temperature

References Ashby, M. F., "Materials Selection in Mechanical Design," 4th Edn. (2011) Elsevier Atkins, A.,

" The Science of Engineering of Cutting," (2009) Butterworth-Heinemann

Berry, J. T. PhD Thesis, University of Birmingham, UK, 1954 Berry, J. T, Kondic,V. and Martin, G., Trans. Am. Foundry Soc, 67 (1959) 449-476 Berry, J.T., Private communication, NSF CADCAST Project (1981) Campbell, J., Castings 2 nd Edn. (2003) Butterworth-Heinemann Chvorinov, N., Proc. Inst. Brit. Foundrymen, 132 (1938-1939) 29-40, also Giesserei, 27 (1940) 177-186, 201-208,222-225 and Hutnicke Listy, 6(1951) 11-20 Dantzig, J.A. and Lu,S. C , Met. Trans. 16B (1985) 195-202 and Dantzig, J.A. and Wiese, J.W., Met. Trans. 16B (1985) 203-209 Einstein, A., Relativity, The Special and General Theory, (1916) Methuen and Co. Ltd. Farquar, R.B., Trans. Am. Foundry Soc, 29 (1920) 171- 201 Franklin, PH., M.S. Thesis, Georgia Inst. Of Tech. 1982 Fras, E. and Lopez, H., Int. Jnl. Metalcasting, 4 (2010) 35- 58 Huang, H., PhD Dissertation, The University of Alabama, 1994 Manganon, PL. and Heitman, W.E., in 'Microalloying '75', 1977, Union Carbide, NY Monroe.C. and Huff, R., Int. Jnl. Metalcasting, 4 (2010) 27-33

296

Kear, B.H., Leverant, G.R.and Oblak, J.M., Trans. Am. Soc. Metals, 62 (1969) 639 Subbiah, S.and Melkote, S„ Materials Science and Engineering, 474 (2008) 283-300 Moosbrugger, J., M.S. Thesis, Georgia Inst. of Tech., 1985 Moosbrugger, J., Berry, J.T. And Wei, C.-S., ASME Paper 86-WA/HT-95 (1986) Pellini, W.S., Am. Foundryman, 24 (November, 1953 and December, 1953) 58-61 and 62-71 Ruddle, R.W., 'The Solidification of Castings', 2nd. Edn. (1957) Inst. of Metals, London. Ruddle, R.W. and Skinner, B.F., Jnl. Inst. of Metals 79(1951) 35-56 Tiryakioglu, M., Private Communication to J.T. Berry, 1995 Wei, C.-S., PhD Dissertation , Georgia Inst. Tech.., 1982 Wei, C.-S., Hansen, P.N. and Berry, J. T., in ' Numerical Methods in Heat Transfer', Vol. 2, 1983 (Lewis.R., Morgan, K. and Schrefler, B.A., Eds.) 461-471 Wiley Winegard, W.C., 'An Introduction to the Solidification of Metals', (1964) Inst. of Metals, London Wyatt, J.E. and Berry, J. T. (This Symposium) Wyatt, J.E, Altan, T. and Berry, J.T, Paper presented at 2009 SE Regional Meeting of Am. Soc. of Engineering Education

297

Figures

Fig. 1 Analytical Solution (Solid line) With Experimental Results Solidification Time v. Volume/Surface Area,ll% Si-Al Plates in Dried Silica Sand Molds (Berry, 1954)

298

Fig. 2 As Fig. 1 but Solidification Time as a Function of Volume/Effective Surface Area, R'. (Berry, 1954)

299

Fig 3 Analytical Solution (Solid Lines) With Experimental Results Solidification Time v. Volume/Surface Area, 8.5% Al-Cu Plates in Dried Silica Sand Molds of Coarse (14/28), Medium (Bromsgrove, Natural) and Fine (H) Grain Size (Berry,1954)

300

Fig. 4 As Fig. 3 but Solidification Time as a Function of Volume/Effective Surface Area, R'. (Berry, 1954)

301

Fig. 5 Interfacial Temperature Gradient v. Reciprocal of Sq. Root of Elapsed Time for ThreeDimensional Wedge (All Right Angles). Circles are Experimental Values for Pure Aluminum in Dried Silica Sand . Solid Line is Chvorinov Solution For Planar Solidification. (Franklin, 1982)

302

Fig. 6 Schematic of Experimental Relationship of Figure 5, Summarizing Effects of Mold Geometry and Casting Superheat. Linear Relation is Chvorinov Solution For Planar Solidification. (Moosbrugger, Berry and Wei, 1986)

303

Fig. 7 Experimental Determination of Chilling Power Using Chvorinov Curve Fit Method, for Coarse Silica Sand(14/28) Packed at High Density (1.82 gm/cc) (Berry, 1954) Ordinate is Dimensionless Mold Temperature

304

Fig. 8 Similar Data to That of Figure 7, but for Same Sand With Low Density Packing ( 1.34 gm/cc) (Berry, 1954)

305

Fig. 9 Experimental Results Solidification Time v. Volume/Surface Area for 8.5% Al-Cu in Dry and Green Silica Sands of Coarse (14/28) and Fine (H) Grained Nature. Note the Dramatic Chilling Enhancement Present for Light Sectioned Castings. (Berry, 1954)

306

Fig. 10 Experimental Results for Solidification Time v. Volume/Effective Surface Area for Gray Iron Plates (CE 4.2%) in Green and Dry Silica Sand Molds. Note the Drying-Out Effect for Heavy Sectioned Castings. (Berry, 1954)

307

Fig. 11 Interfacial Temperature Gradient Data for a Three-Dimensional Corner (all RightAngles) for Pure Aluminum in Three Different Type Silica Sand Molds, Dry and Green ClayBonded, and Organically Bonded. Note Enhanced Chilling Effect of Moisture Presence and Decreased Chilling Effect of Presence of Organic Bond. (Berry, unpublished work)

308

Tables Table 1. Companson of Mean Heat Difrusivity Values for Various Molding sands as Determined by Three Separate Methods (Berry, 1954) Sand 14/28 (Coarse) 14/28 (Coarse) H (Fine) T (Fine, Natural) Notes:

Compacted Density (gm/cc) 1.82 1.35 1.40 1.19

1 31.0 20.8 18.3 15.8

Mean Heat Difrusivity Method 2 27.4 20.9 13.9 13.0

3 25.2 21.6 17.2 15.8

Figures given as originally converted to CGS units x 1000 Methods of Determination are :

1. 2. 3.

309

Chvorinov Curve-Fit Method Russell-Gittus Method of Calculation Mold-Calorimetric Method

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

AUTHOR INDEX Shape Casting IV A

Antonio Maroto, J Ares, A Au, D

B

Berry, J Bozorgi, S Browne, D Byczynski, G

c

Campbell, J Ceschini, L Cockcroft, S Connolley, T Cuesta, R

D

Dai, X Dispinar, D Druschitz,A Duan,J

E

El-Sayed,M

F

Felicelli, S Flender, E

G

Gassa, L Green, N Griffin, J Griffiths, W Grupke,C Gueijman, S Guo,J

H

Haberl, K Hamilton, R Han,Z Heisser, C Hsu, F Hudak, D

157 207 21

157, 215, 281, 289 113 129 191

J

Johnson, M Jolly, M

K

71, 139, 181 121 21 87 157

Kandeil, A KangJ Kendig, K Khajeh, E Kneissl, C

L

265 173 199 21

Laplume, F Lee, P Leonard, C Lett, R Leung, A Li, J Lin, H Liu,B

149

M

29, 53, 157, 215 3

Mackay, R Maijer, D Mirihanage, W Morri, A

207 13 199 149, 225 233 207 103

N

Nastac, L Nath, R Nguyen, K Nordmark, A Nyahumwa, C

311

113 87 61 3 45 165

233 13, 265

149 257 199 37 113

273 87 233 157, 215 87 61 45 61

191 21, 37 129 121

249 233 21 173 139

o

Onsoien, M.

P

Pabel.T Phillion, A Puhakka, B

R

Reilly, C Rockett, P Romanelli, J

S

Sajja, U Salem, H Samuel, E San José, R Schumacher, P Schvezov, C Scott, S Senkov, 0 Senkova, S Siavashi, K Squatrito, R Stern, J Syvertsen, F

T

Tiryakioglu, M Todaro,! Tomesani, L Topping, C

w

z

.95

Zeng, B Zheng, C Zhu,J

113 87 79,241

13, 21 87 249

53 149 273 157 113 207 103 199 199 225 121 3 173

139, 165 121 121 225

Wang, L Wyatt,J

29, 215 281

Yang, W Yin, H

61 29

3 12

265 273 103

Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011

SUBJECT INDEX Shape Casting IV A

A356 Wheel Al-Cu Al-Cu Alloys Aluminium Aluminium Alloys Aluminium Casting Process Aluminum Aluminum A356 Aluminum Alloy Aluminum Gravity Casting ASTM A757 C1Q Steel Autonomous Optimization AZ91

B

Bifilms Brittle

E

21 37 207 87, 149 225 265 29,273 157 61 45 249 3 215

Electromagnetic Pump Entrainment Eutectic Experimental Validation

F

Facets Fatigue Fatigue Potential Fatigue Specimens Filling Flow-Related Defect Four Point Bend Fractional Step Method Fracture Freckles

139, 173, 233, 241 181

Grain Transport Griffith Crack

Cast Aluminum Alloy 199 Casting 13, 103, 149,215 Casting Defects 139, 241 Casting Methoding 241 Casting Modeling 249 Casting Shape 289 Cellular Automaton 37 Ceramic Foam Filter 45 CET 129 Chilling Power 289 Classic Nondestructive Testing 233 Columnar-to-Equiaxed Transition 207 Confidence Interval 165 Control Arm 215 Cracking Susceptibility Coefficient 113 Cracks 71,249 CRIMSON 265 Critical Gating Velocity 45 Curvature 257

Defects Degassing Dendrite Growth Distortion Double Oxide Film Defects Ductile Ductile Cast Iron

139 181 139 121 3 21 157 53 181 53

G

C

D

157 13 37 289

H

HAZ Heat Treatment High Pressure Die Cast Hot Cracking Index Hot Cracking Susceptibility Hot Tearing Hot Tears HPDC Hydrogen Hypothesis Testing

129 181

71 3, 257 191 113 113 103 249 273 121, 173 165

I

In-Mold Melt Treatment In-Mold Thermal Analysis

L

13, 181 173 29 3, 257 149 181 95

LCS Waterjet Entry Edge Components Lost Foam Casting LPDC

3 13

95 95

249 225 21

M

Machinability Machining Sequencing Macrosegregation Magnesium Mathematical Modeling Mechanical Properties Melting Mesh Adaptation Microporosity Microstructure Modeling Modeling and Simulation Modelling Mold Design Molecular Weight Mould Filling

N

Naturally Pressurized Fill System Ni Superalloys

o

Oxide Bifilms Oxide Film Oxides

P

Permeability Porosities Porosity Precision Sand Prediction of Macro-Shrinkage Process Compensated Resonant Inspection

R

Residual Stresses Rheocasting Runner System

S

Sand Mold Printing SEED Process Semi-Solid Semi-Solid Casting Simulation

Single Crystals Solidification Solidification Under Pressure Squeeze Casting Steel Castings Stress Structural and Corrosion Parameters Super Duplex Stainless Steel

281 281 53 215 21 149, 225 265 53 61, 121 3, 95, 215 3, 29, 103 61 13 249 225 21

T

Tensile Properties Terminal Freezing Range Thermophysical Properties Turbine Blade Casting

u

Ultra-High Strength..

79 71

w

Water Analogue Experiment Weibull Weibull Modulus Weibull Statistics Welds

79 265 13, 233

X

X-Ray Microtomography X-Ray Tomography

37 249 87, 173 191 249 233

281 273 45

249 273 273 273 3

314

71 3, 129 199 61 241 3 207 79

191 113 3 257

199

21 157 165 191 71

37 87