John
Moalli,
Editor
,\."CI"(V (~r PI(/sIICS j,n.\:IfIi?Cr.,
F\kitk,.l:I;.itgnlbmy
Copyright © 2001, Plastics Desig...
553 downloads
2829 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
John
Moalli,
Editor
,\."CI"(V (~r PI(/sIICS j,n.\:IfIi?Cr.,
F\kitk,.l:I;.itgnlbmy
Copyright © 2001, Plastics Design Library. All rights reserved. ISBN 1-884207-92-8 Library of Congress Control Number: 2001091836
Published in the United States of America, Norwich, NY by Plastics Design Library a division of William Andrew Inc. Information in this document is subject to change without notice and does not represent a commitment on the part of Plastics Design Library. No part of this document may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information retrieval and storage system, for any purpose without the written permission of Plastics Design Library. Comments, criticism and suggestions are invited and should be forwarded to Plastics Design Library. Plastics Design Library and its logo are trademarks of William Andrew Inc.
Please Note: Great care is taken in the compilation and production of this volume, but it should be made clear that no warranties, express or implied, are given in connection with the accuracy or completeness of this publication, and no responsibility can be taken for any claims that may arise. In any individual case of application, the respective user must check the correctness by consulting other relevant sources of information. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.
Manufactured in the United States of America.
Plastics Design Library, 13 Eaton Avenue, Norwich, NY 13815 Tel: 607/337-5080 Fax: 607/337-5090
Preface
As the technology of plastics continues to develop, extending from formulation to processing to design, the number of plastic products, especially those replacing metals, increases in parallel. Resin manufacturers create new formulations with improved properties, molds and dies are highly optimized for maximum part performance, and designers have a plethora of new computer aided tools at their disposal. Although it would seem that such technological advances would serve to reduce plastic product failures, the additional complexity involved in modern designs, as well as the constant push for performance at the edge of the material envelope, seems to moderate failure rate reduction. Plastic product failures can be divided into three discrete arenas - improper design, improper manufacturing (including processing and assembly), and improper use of the end product. One way to reduce plastic product failures is to disseminate the knowledge, on a wide variety of topics, which has been accumulated by plastics researchers and practitioners. Such is the purpose of this book, and by sharing the work presented at a number of SPE ANTEC conferences, we hope to answer a wide range of questions that have arisen and anticipate those topics that may present themselves in the future. The first three chapters introduce the reader to a variety of failure mechanisms, starting more generically with ductile failure, fatigue, and oxidation in Chapter 1 and advancing through failures related to processing, assembly and environmental effects in Chapters 2 and 3. All of these are important mechanisms for consideration, as even a well designed part can fail prematurely if it is not processed properly or it is exposed to environments that degrade it's mechanical properties. Chapter 4 discusses fractography and morphology of plastics, which can often be confusing to the uninitiated if they attempt to directly translate techniques established for metals; plastics are known to share some features with metals but are often distinctly different. Nonetheless, examination of fracture surfaces is a valuable tool that can provide insight into product loading history, environmental exposure and even processing conditions. The next 5 chapters in the book were selected to help the reader with a proactive approach in failure prevention. Examination of failure and material models, determination of product life, test methods, design aids and case studies are all presented, and we hope that this information will prove to be useful to product design engineers, mold designers, failure analysts, and general plastics practitioners in all phases of product design and development. John Moalli Menlo Park. California
i
Table of Contents
Preface
vi John Moalli
Chapter 1. Failure Mechanisms 1 Plastics Failure Due to Oxidative Degradation in Processing and Service 1 Myer Ezrin, Amanda Zepke, John Helwig, Gary Lavigne and Mark Dudley Durability Study of Conductive Copper Traces Within Polyimide Based Substrates 9 Elena Martynenko, Wen Zhou, Alexander Chudnovsky Ron Li and Larry Poglitsch Fatigue Behavior of Discontinuous Glass Fiber Reinforced Polypropylene 17 Mustafa Sezer and Ahmet Aran Ductile Failure and Delayed Necking in Polyethylene 25 W. Zhou, D. Chen, Y. Shulkin, A. Chudnovsky, N. Jivraj, K. Sehanobish, and S. Wu Chapter 2. Processing and Assembly The Role of a Heat Affected Zone (HAZ) on Mechanical Properties in Thermally Welded Low Density Polyethylene Blown Film Timothy E. Weston and Ian R. Harrison Effects of Processing Conditions on the Failure Mode of an Aliphatic Polyketone Terpolymer Nicole R. Karttunen and Alan J. Lesser Orientation Effects on the Weldability of Polypropylene Strapping Tape MJ Oliveira and DA Hemsley Joint Performance of Mechanical Fasteners under Dynamic Load Self-Tapping Screws in Comparison with Threaded Inserts in Brass and Plastic Axel Tome, Gottfried W. Ehrenstein, and Frank Dratschmidt Defect Cost Analysis Christoph Roser and David Kazmer
31 31
39 45
53 63
ii
Chapter 3. Environmental Effects Environmental Stress Cracking (ESC) of ABS (II) Takafumi Kawaguchi, Hiroyuki Nishimura, and Fumiaki Miwa, Takashi Kuriyama, and Ikuo Narisawa Residual Stress Development in Marine Coatings Under Simulated Service Conditions Gu Yan & J R White Estimation of Long-term Properties of Epoxies in Body Fluids Steven W. Bradley Mechanical Performance of Polyamides with Influence of Moisture and Temperature – Accurate Evaluation and Better Understanding Nanying Jia and Val A. Kagan Temperature-Moisture-Mechanical Response of Vinyl Ester Resin and Pultruded Vinyl Ester/E-Glass Laminated Composites S. P. Phifer, K. N. E. Verghese, J. J. Lesko, and J. Haramis Freeze-thaw Durability of Composites for Civil Infrastructure J. Haramis, K.N.E. Verghese, and J. J. Lesko
73 73
79 89
95
105 113
Chapter 4. Morphology and Fractography 121 Fractography of ABS 121 Hiromi Kita, Masatoshi Higuchi, Atsushi Miura Fractography of Metals and Plastics 127 Ronald J. Parrington Crack Propagation in Continuous Glass Fiber/Polypropylene Composites: Matrix Microstructure Effect 135 M. N. Bureau and J. Denault, F. Perrin, and J. I. Dickson Fracture Behavior of Polypropylene Modified with Metallocene Catalyzed Polyolefin 143 Laura A. Fasce and Patricia M. Frontini, Shing-Chung Wong, and Yiu-Wing Mai Morphology and Mechanical Behavior of Polypropylene Hot Plate Welds 149 MJ Oliveira, CA Bernardo, and DA Hemsley The Influence of Morphology on the Impact Performance of an Impact Modified PP/PS Alloy 159 S. P. Bistany Morphological Study of Fatigue Induced Damage in Semi-crystalline Polymers 165 Nathan A. Jones and Alan J. Lesser
iii
Chapter 5. Modeling of Failures and Failure processes Failure Analysis Models for Polyacetal Molded Fittings in Plumbing Systems L.J. Broutman, D.B. Edwards, and P.K. So Progressive Failure Analysis of Fiber Composite Structures Matt H. Triplett Calculating Thermally Induced Stresses Using a Nonlinear Viscoelastic Material Model N. Schoeche and E. Schmachtenberg Evaluation of a Yield Criteria and Energy Absorbing Mechanisms of Rubber Modified Epoxies in Multiaxial Stress States Robert S. Kody and Alan J. Lesser
173 173
Chapter 6. Design and Life Prediction Shelf Life Failure Prediction Considerations for Irradiated Polypropylene Medical Devices Michael T. K. Ling, Samuel Y. Ding, Atul Khare, and L. Woo Determining Etch Compensation Factors for Printed Circuit Boards Anthony DeRose, Richard P. Theriault and Tim A. Osswald, and Jose M. Castro Activation Energies of Polymer Degradation Samuel Ding, Michael T. K. Ling, Atul Khare and Lecon Woo Estimation of Time-temperature-collectives at Describing Ageing of Polymer Materials D. Blaese and E. Schmachtenberg
201
Chapter 7. Test Methods Standard Test Procedures for Relevant Material Properties for Structural Analysis Gerald G. Trantina and Joseph T. Woods Factors Affecting Variation in Gardner Impact Testing Mark Lavach Radiation Resistance of Multilayer Films by Instrumented Impact Testing Robert Wojnarowski, Michael T. K. Ling, Atul Khare, and L. Woo Aspects of the Tensile Response of Random Continuous Glass/Epoxy Composites Okenwa I. Okoli, G.F. Smith
179
187
193
201 209
219
227
233 233 241 247
253
iv
Comparing the Long Term Behavior of Tough Polyethylenes by Craze Testing KC Pandya and JG Williams
259
Chapter 8. Failure Prevention Design Aids for Preventing Brittle Failure in Polycarbonate and Polyetherimide Joseph T. Woods And Ronald P. Nimmer 10 Common Pitfalls in Thin-Wall Plastic Part Design Timothy A. Palmer Defect Analysis and High Density Polyethylene Pipe Durability Shaofu Wu, Kalyan Sehanobish, and Noor Jivraj Practical Risk Analysis - As a Tool for Minimizing Plastic Product Failures Subodh Medhekar, John Moalli, Robert Caligiuri Attachment Design Analysis of a Plastic Housing Joined with Snap-fits Dean Q. Lewis and Gary A. Gabriele, and Bob Brown Avoiding the GIGO Syndrome – Combining the Real and Virtual Worlds in Analysis of Polymer Product Failures John Moalli, Steven Kurtz, Robert Sire, Sanjeev Srivastav, Ming Wu
267 267
Chapter 9. Case studies Case Studies of Inadvertent Interactions Between Polymers and Devices in Field Applications Joseph H. Groeger, Jeffrey D. Nicoll, Joyce M. Riley, Peter T. Wronski Case Studies of Plastics Failure Related to Improper Formulation Myer Ezrin and Gary Lavigne Translating Failure Into Success – Lessons Learned From Product Failure Analysis John E. Moalli Index
275 281 289 297
307
313 313 323
329
337
Chapter 1 Failure Mechanisms Plastics Failure Due to Oxidative Degradation in Processing and Service
Myer Ezrin, Amanda Zepke, John Helwig, Gary Lavigne and Mark Dudley University of Connecticut, Institute of Materials Science, Storrs, CT 06269-3136, USA
INTRODUCTION People cannot live without oxygen and water. But these are deadly enemies of polymers, both in processing of plastics formulations and in service. Water is a problem mainly for condensation polymers which degrade by hydrolysis. In this paper the focus is on oxidative degradation. Oxygen degrades polymers to lower molecular weight (MW) by reacting with polymer free radicals to form peroxy free radicals (ROO•) and hydroperoxides (ROOH). Free radicals have an unshared electron and react in any way they can to restore the atom or molecule to a balanced structure. Often that leads to chain scission. As MW goes down most polymer properties suffer. As little as 5-10% reduction in MW may cause failure. Avoiding contact with oxygen and using an antioxidant (AO) as a free radical scavenger are means of preventing degradation. The high temperature required to process plastics is the major cause of degradation in injection molding, extrusion, blow molding, etc. High temperature is needed to fuse polymers and to reduce melt viscosity to a level that the machines can handle. Mechanical shear of the melt and the presence of oxygen, even in small amounts, are major factors in degradation due to processing. The chain carbon atoms attached to a branch, such as methyl group (CH3), tend to split off a hydrogen atom, creating a free radical at a tertiary carbon atom.
CH2
CH(CH3)
CH2
C(CH3)
+ H
2
Plastic Failure Analysis and Prevention
Very little oxygen is needed to react with free radicals during processing. Polymer suppliers usually have very little AO in the resin as sold to processors. Unless additional AO is added, polymer is likely to degrade in process. Polyolefins, which have only carbon-carbon chain bonds (PE, PP, EP and other copolymers) are particularly susceptible to oxidative degradation, in service as well as in processing. Even if additional AO is added, severe processing conditions (high temperature, high shear, long residence time in the barrel), use of regrind, etc. may deplete most of the AO, leaving too little to withstand conditions in service. Even in moderate service conditions, such as a PE eyewash squeeze bottle on a laboratory wall, oxidative degradation can lead to failure in long term applications. Such a PE bottle, which had been on a laboratory wall for 15-20 years, cracked when tested in a safety inspection. Here, too, additional AO is needed to survive many years of service. A complicating factor in processing is formulations containing peroxides to crosslink the polymer. Peroxide causes crosslinking by decomposing to free radicals ( ROOR → 2RO· ). The high content of peroxy free radicals formed abruptly reacts with the polymer to cause crosslinking. These free radicals may react with the AO, leaving the system without enough AO for the polymer to survive processing and service. The AO system must be chosen accordingly. Commonly used methods of analysis to determine if failure is due to oxidative degradation are differential scanning calorimetry (DSC) for oxidative induction time (OIT), (ASTM D3895) or oxidative induction temperature (ASTM D3350). Infrared spectroscopy (IR) may detect bound oxygen as carbonyl (C=O), which forms increasingly as AO becomes exhausted. A third method is change in MW measured as an increase in melt flow rate (MFR), (ASTM D1238). This is a very practical method because it relates directly to MW, i.e., a small reduction in MW gives a large increase in MFR. The applicable relationship is n=KM3.4. Gel permeation chromatography (GPC) is also useful for monitoring MW changes in processing or service. The DSC methods require about an hour or less, after establishing test conditions, and are most useful for comparing materials, e.g., before and after processing, or after service. They are a practical method of determining the relative amount of AO remaining. When a sample’s AO content is zero, oxidation exotherm starts very soon after oxygen is admitted into the DSC cell. Additional information on the DSC methods is given in the next section. IR is useful mainly to detect bound oxygen, which occurs when most or all of the AO has been depleted. Examples are given below of failure due to oxidative degradation for (1) HDPE power cable jacket; (2) PE low voltage cable in a power plant control room; (3) PP rotors in a hot
Plastics Failure Due to Oxidative Degradation
3
water system; (4) EPDM hot water check valve; and (5) EVA (ethylene vinylacetate) hot melt adhesive degraded in a heated reservoir. A recent case of PP failure in hot water heaters, most likely due to oxidative degradation, was reported in Consumer Reports, July 1999, p. 8. PP that replaced copper dip tubes brings cold water to the bottom of the heater. The PP has been disintegrating into small pieces and clogging pipes and other water delivery systems, and preventing normal operation of the hot water heater. Class action law suits have been filed in some states. Sixteen million heaters were made between 1993 and 1996 with PP dip tubes that may be defective.
EXPERIMENTAL METHODS DSC - OXIDATIVE INDUCTION TIME AND OXIDATIVE INDUCTION TEMPERATURE The OI time method requires selecting an isothermal temperature, first equilibrated in nitrogen, then changing to oxygen for the test. For PE and PP a common temperature is 200°C. The OI temperature method is much simpler, because a routine programmed temperature run is made using oxygen from the beginning. At some temperature an exotherm will be observed. OI time testing is critically dependent on selecting an appropriate isothermal temperature. Switching from nitrogen to oxygen may cause an upset in the baseline which could complicate interpretation of the result, i.e., did the exotherm start right away (zero minutes OIT) or is it at a higher value, which may be difficult to detect with some polymers. The instrument used was a TA Instruments model 2920. The OI time method generally followed the guidelines of ASTM D3895. The isothermal temperature depends some-what on the material being tested, selected to give a time to exotherm of about 60 minutes for the most highly stabilized samples. Surface to volume ratio and sample weight affect the response to oxygen, affecting how readily and reproducibly the OI time is determined. The ASTM method calls for extrapolation of the initial part of the exotherm to the baseline (see DSC figures below). Extrapolation introduces some variability because not all samples in a group have similar exotherm slopes. A more realistic measure of OI time or OI temperature is the initial onset time or temperature (see DSC figures below). That is when the reaction with oxygen starts. Since the deflection off the baseline is often slight and gradual, a common or standard method needs to be used to detect the initial onset point because the temperature or time is not important by itself, only in comparison with similar samples with different process or service history. Uniform contact with the sample pan bottom is ideal. If OI time is required, an OI temperature run is helpful in selecting an appropriate isothermal temperature.
4
Plastic Failure Analysis and Prevention
MOLECULAR WEIGHT-RELATED METHODS Melt Flow Rate - ASTM D1238 If a sample no longer has AO, some degradation may occur in the MFR test at elevated temperature. That will give a value which is the sum of the change due to the sample’s pretest history and its MFR test. This effect can be judged by running the test at various heating times in the barrel before extrusion. If MFR goes up as test time increases, degradation in the test is indicated. Gel Permeation Chromatography This method gives molecular weight distribution by passing a solution through columns of controlled pore size. The instrument used is a Waters 150C with THF solvent at 35°C and 1 ml/min. flow rate. INFRARED SPECTROSCOPY A Spectratech micro IR in the reflectance mode with a germanium crystal was used with a Nicolet Magna 560 FT/IR. While special attention is paid to the carbonyl (C=O) region at 1700-1750 cm-1, the rest of the spectrum may indicate other changes due to processing or service.
EXPERIMENTAL RESULTS HDPE POWER CABLE JACKET FRACTURE Black HDPE jacket of a medium voltage power distribution cable experienced cracking in a certain pattern (Figure 1). The jacket was in contact with a copper foil wrap directly underneath it. Fracture occurred where the cable was wet in a manhole. Fracture lines were at the points where two layers of copper tape overlapped, putting pressure on the jacket along these lines. Fracture did not occur where jacket was dry. OI time extrapolated value at 199°C in oxygen was six minutes in areas where the copper and jacket had been wet. Using the initial onset temperature instead of the extrapolated value, OIT was zero. That is, all AO was depleted. The role of copper is important. In the wet state some copper is converted to the ionic form. Copper ion is a catalyst for reactions which result in oxidative degradation of PE. This failure was affected strongly by Figure 1. HDPE power cable jacket fractured by oxidative degradation due to water and copper in contact the contact of PE with copper in the ionic form due with jacket. Copper below jacket shows through to water immersion. It also illustrates that an OIT cracks.
Plastics Failure Due to Oxidative Degradation
Figure 2. Unused PP rotor and closeup of a degraded fin of a used rotor.
5
Figure 3. DSC-01 time at 200°C of #1 unmolded PP pellets (>50 min.); #2 unused molded rotor (4 min.); #3 rotor degraded in service (1 min.).
value of six minutes by extrapolation is really a value of zero at onset of the exotherm, for all practical purposes. The curvature of the cable meant that bending stress was also a factor in the jacket failure. 3.2 PE INSULATION IN A POWER PLANT CONTROL ROOM Sections of 1/8" diameter PE insulation cracked, threatening the safety system of which the fractured PE was a part. Failure occurred in about 10 years particularly where wiring was near fluorescent lighting. At such locations OIT values were very low or zero minutes. In the worst cases of embrittlement, IR revealed carbonyl bond oxygen (C=O). Fluorescent lighting appears to have enough ultraviolet radiation to accelerate oxidative degradation. Areas well away from fluorescent lighting were relatively undegraded. PP ROTORS IN A HOT WATER SYSTEM In this case pellets, as-molded rotors, and degraded rotors were available to track the OIT values from “cradle to grave.” Degraded rotors in service for about a year experienced substantial degradation under service conditions of hot water, steam and air. Figure 2 shows a complete unused rotor, which measures about 2" in diameter. The photo includes a closeup of the degraded end of a fin of a used rotor. Figure 3 is the DSC-OIT thermogram of pellets, unused rotor as molded and a rotor degraded in service. Isothermal DSC temperature was 200°C. The OI time for pellets was >50 minutes; for as-molded rotors, time was 4 minutes; and degraded rotor was 1 minute. Time to purge the DSC cell with oxygen following equilibration in nitrogen was about 1 minute, so that the OI time of degraded rotor was practically zero. The pellets were well stabilized (OIT >50 min.), so that an OIT value of 4 minutes for molded rotor indicates that most of the AO was consumed in processing. The type of service
6
Plastic Failure Analysis and Prevention
Figure 4. EPDM hot water check valve molded on metal support. Surface is degraded and uneven; metal spring in center has broken through the degraded EPDM. Overall diameter is approx. 5/8 inch (20 mm).
Figure 5. DSC-OI time at 210°C of EPDM valve - #1 failed in service (0 min.); #2 unused valve (0 min.); #3 a different EPDM that did not fail in service (>50 min.).
(hot water, air) readily consumed the remaining AO leaving the rotors very susceptible to degradation in service. It is possible that a different AO system might have provided better protection in processing and service. EPDM EMBRITTLED IN HOT WATER CHECK VALVE The part resembles a small mushroom with the EPDM covering a metal support (Figure 4). In the figure, the EPDM over the metal part in the center has fractured and been lost, leaving the metal exposed. A spring-loaded metal part in the center controls the valve. Failure in service was due to development of roughness of the EPDM surface and to the metal part breaking through the EPDM. OI time was performed at 210°C for valve material failed in service (#1), prior to service (#2), and for a different unused EPDM valve that had not failed in service (#3). Figure 5 shows that sample #3 had OIT of >50 minutes. The other two, made with a different EPDM formulation than #3, had OIT of zero. OI temperature runs were made to see how the exotherm onset temperatures would compare to the OI times at 210°C. Figure 6 gives OI temperature runs for samples 1, 2 and 3. Representative initial onset and extrapolated values are indicated. Each test was done in duplicate, with remarkably good reproducibility. The greatest uncertainty is in the initial onset temperature for the good EPDM that did not fail (#3). The curve slopes up much more gradually than for #1 and #2. The OIT values are given in Table 1 for the initial onset and extrapolated values. They differentiate between #1 and #2, while OI time at 210°C gave values of zero for both. OI time tests at lower isothermal temperatures also gave very low values for #1 and #2. Thus OI temperature provided useful information more readily than OI time did, or would have required considerable effort to select the optimum temperature.
Plastics Failure Due to Oxidative Degradation
7
Table 1. Oxidative induction temperature of EPDM OIT, oC Sample Initial onset
Figure 6. DSC-OI temperature of same samples as in Figure 5 - see Table 1 for initial onset and extrapolated OI temperatures. Values are shown for #1. 123° initial onset and 147° extrapolated.
Extrapolated
#1 - failed in service
122, 124 avg. 123
146, 147 147
#2 - same as #1 - unused
160, 159 avg. 160
175, 174 175
#3 - different EPDM unused
203, 167, 185 avg. 185
270, 270, 272 271
Table 1 indicates that the AO level as molded, before service, was enough to give an OI temperature initial onset value 37°C higher than after failure in service and 28°C higher by extrapolation. By comparison to #3, an earlier formulation that never failed in service, the newer material as made is much lower, 25° by initial onset and 96° by extrapolation. Clearly, the service condition requires an AO capability like that of #3. In the manufacture of the new valve, some AO is depleted in processing, and the amount remaining is not enough to protect Figure 7. Micro IR reflectance spectra of EPDM valve - the material in service (failure was in three #1 degraded outer surface - bound oxygen at 1500-1800 months). Another possible contributor to the cm-1 not present below surface; #2 inner bulk below problem was that EPDM was crosslinked with degraded outer surface. dicumyl peroxide. As indicated above, free radicals from peroxide would have reacted with some of the AO. Infrared spectroscopy (Figure 7) shows considerable bound oxygen in the surface of the failed material (upper curve), from 1500-1800 cm-1, that is absent in the inner bulk of the same failed EPDM (lower curve). These spectra indicate that the oxidative degradation is limited to a very thin surface layer. However, embrittlement at the surface causes fracture
8
Plastic Failure Analysis and Prevention
to occur through the full thickness, following fracture initiation at the surface. DEGRADED HOT MELT EVA ADHESIVE The color after being in the hot reservoir was dark brown, compared to light tan pellets before heating. EVA (ethylene vinylacetate) can degrade by deacetylation, i.e., loss of acetic acid, with formation of a C=C double bond in the chain. This is independent of oxidative degradation, which occurs because of the ethylene units in the copolymer. OI temperature was performed to see to what Figure 8. Micro IR reflectance spectra of EVA hot extent oxidation is a factor also. For pellets initial melt adhesive - #1 unfused pellets; #2 dark brown onset temperature was 164° and by extrapolation adhesive from heated reservoir. 216°C. Dark material gave corresponding values of 148° and 206°C. These differences seem small relative to the very great difference in color. GPC for molecular weight distribution gave some reduction in MW of EVA, but also a considerably smaller GPC peak for degraded material. That corresponded to material which did not dissolve in hot toluene, whereas toluene did dissolve the pellets well. Toluene solutions were injected into the THF GPC solvent. IR spectra (Figure 8) of pellets and degraded material are different in ways that are not clear. The VA content has not changed much. The large peak near 1400 cm-1 is gone, as well as the small one near 900 cm-1. These correspond to CaCO3 filler, which appears to have reacted, possibly with acetic acid from vinylacetate. In this case oxidative degradation may have been partly responsible for the severe darkening, but other factors not readily understood may have been even more responsible.
SUMMARY OI time and temperature are useful methods for deter-mining if oxidative degradation is the cause of polymer property loss due to processing and to service conditions. The methods are particularly attractive for polyolefins, which degrade to lower molecular weight oxidatively, and are very dependent on AO to minimize degradation. The OI temperature method is much simpler than OI time and in this study revealed differences between materials even better than OI time, in some cases. Together with other methods, such as GPC, melt flow rate and IR spectroscopy, failures due to oxidative and other causes of degradation can be determined. The result is a better understanding of the extent to which failure can be ascribed to processing, to service and to reactions other than oxidation.
Durability Study of Conductive Copper Traces Within Polyimide Based Substrates
Elena Martynenko, Wen Zhou and Alexander Chudnovsky Fracture Mechanics and Materials Durability Laboratory, Civil and Materials Engineering Department, The University of Illinois at Chicago, 842 W. Taylor Street (M/C 246), Chicago, IL 60607, USA Ron Li and Larry Poglitsch Motorola Inc., Automotive and Industrial Electronics Group, 4000 Commercial Ave, Northbrook, IL 60062, USA
INTRODUCTION Flexible circuits are widely used in various electronic packages. As the complexities of electronic packages grow, high reliability of assembled components is critical to maintain final product quality, especially in light of trends toward miniaturization and higher levels of integration. Electronic packages with FPC are used in every conceivable application from heart pacemakers, to automotive instrument clusters and to missile guidance systems. FPC failures may lead to serious consequences. A detailed understanding of why and how electronic packages fail greatly aids in the development of high-performance packaging with enhanced reliability. A variety of factors essential in the electrical, mechanical, and thermal design can contribute to the packaging failures. Properties of materials such as interconnection alloys, metal plating, laminates, adhesives, etc. can be a source of catastrophic failures if not properly understood and selected. Modern electronic systems in many applications experience severe vibrations and shocks. The failure appears to be due to submicroscopic cracks that grow into visible cracks and lead to a complete rupture without warning under repeated loading. Therefore, fatigue resistance is of major importance in reliability assessment of various electronic packaging. The primary objective of this study is to determine high cycle fatigue resistance of certain flexible circuits. Various failure modes and mechanisms in electronic packages have been addressed in literature.1-3 However, limited information has been presented on high cycle fatigue resistance of FPC. To the authors knowledge there was no reported study for a thermal fatigue testing of the material systems under consideration. The objective of this
10
Plastics Failure Analysis and Prevention
study is to address the fatigue resistance as the function of temperature, displacement and frequency.
EXPERIMENTAL SETUP AND APPARATUS To perform reliable high cycle fatigue testing, precise equipment is required. Specially designed experimental setup includes a sine servo controller, electrodynamic shaker, power amplifier, continuity monitor, temperature chamber with temperature control panel and sample fixture. The sine servo controller is designed specifically for use in controlling wide band electrodynamic vibration shakers in sinusoidal testing applications. A wide operating frequency range makes it adaptable to almost any test situation from research and calibration to production testing. A 40 (75 peak) pound force electrodynamic shaker is designed for generalpurpose vibration testing of small components and stress screening of electronic sub-assemblies. It provides a force output proportional to the input drive current from a power amplifier and consistently reproduces the waveform within the specified level and frequency bandwidth limits. The continuity monitor is a high frequency event detector that was used to determine the number of cycles to failure in each individual trace of the FPC samples.
MATERIALS AND EXPERIMENTAL PROCEDURES FPC samples have been provided by Motorola Inc. All samples represent the single-sided conductor layer, double-access covered FPC (manufactured in accordance with IPC-FC-240 requirements) with a composite structure consisting of polyimide dielectric laminate and copper circuit traces, i.e., three layers – two dielectric layers and a single conductor.4-7 Flexible circuitry is typically a composite, of metal foil conductors, and a flexible dielectric substrate. The substrate insulates the conductors from each other and provides much of the circuit’s mechanical strength. Plastic films, synthetic papers, and resin-impregnated fabrics have been used as dielectrics in flexible circuits, but polyimide and polyester films satisfy the widest spectrum of requirements.8,9 The conductor material in FPC must survive processing and provide adequate electrical performance in the service environment. Conductor properties influence the flexural fatigue life of a flexible circuitry assembly. In many “static” applications, bending is limited to installation and servicing. In “dynamic” applications, the assembly is flexed or folded repeatedly during normal use. For dynamic applications, conductors should be of the minimum acceptable thickness and have high fatigue ductility. Conductors made of copper foil provide the best balance between conductivity, ease of processing, and cost.10 Three material systems (A, B and C) have been selected for our study. All of them consist of polymer matrix (polyimide) with embedded copper circuit traces. Dog-bone shape
Durability Study of Conductive Copper Traces
11
samples have been prepared for fatigue studies. There are two asymmetric holes introduced for identification purpose to ensure consistent orientation. Sample thickness is 0.15 mm. Overall 16 samples of material system A have been tested. For each test condition at least 2 but in most cases 3 samples have been tested. This results in 16 to 24 data points for each loading condition. To assure proper sample alignment during installation and testing, required bending as well as precise bonding at the selected domains special sample placement fixture has been designed. The 3M Com. adhesive film is used for bonding purposes. In our study flexes are bonded to the sample fixture via this pressure sensitive adhesive (PSA) film. Symmetrical bending is assured by bonding of the flexes to the fixture in specified domains. Specimen installation on the electrodynamic shaker is a complex procedure requiring special alignment steps and proper connection to continuity monitor. Displacement is transmitted through a vertical rod connected to the electrodynamic shaker. Displacement range is continuously detected by a sine servo controller and automatically adjusted via power amplifier according to the feedback reading of the accelerometer attached to the electrodynamic shaker. This experimental setup guarantees precise reading of displacement and acceleration ranges selected for testing conditions. General schematic of load-transmitting principle is shown in Figure 1. A temperature chamber has been employed to perform thermal fatigue testing with precisely Figure 1. General schematic of load-transmitting support. controlled temperature gradient. The following setup has been used on sine servo controller: output - manual; sweep – manual, continuous. Temperature in the range of 100±5°C has been selected for thermal fatigue studies. Frequencies of 60 Hz and 100 Hz have been selected for our study and displacement range from 1.27 mm to 3.81 mm has been utilized. Initial and final count settings of the continuity monitor have been recorded for all eight traces in each sample and fatigue lifetime has been calculated based on those records. Each sample failure has been followed by a microscopical analysis via an optical microscope attached to a computerized image analyzing system. Digitized images of fracture surfaces of FPC circuit traces are provided in this study. Scanning Electron Microscopy (SEM) has been used in the analysis of failure modes of copper traces. Detailed fracture analysis
12
Plastics Failure Analysis and Prevention
has been performed and failure modes have been established for each material system under investigation.
RESULTS AND DISCUSSIONS Generally, the fatigue life is determined as the number of fatigue cycles required to produce a failure at a given stress level or under a given test conditions. In this study fatigue performance of certain material systems has been analyzed based on fatigue lifetime comparative analysis in addition to fracture analysis. High cycle fatigue testing has been performed under room (23 °C) and elevated (100 °C) temperatures. Two frequency “windows” have been selected: 60 Hz and 100 Hz. The following acceleration values have been selected Figure 2. S-N (D-N) diagram for material system A. 15 g, 23 g and 30 g. They are related to the following displacement values: 1.27, 1.78, 2.79 and 3.81 mm depending on frequency. Fatigue resistance of chosen material systems has been analyzed as a function of temperature and frequency. The fatigue data for this system are reflected in Figure 2 where the number of cycles is given versus displacement, which can be directly related to stresses. Three data points are given for each displacement/loading condition: minimum, maximum and average. These data represent the average values of fatigue performance of particular traces of various samples tested under the same conditions. This representation provides the scatter of fatigue resistance of different traces subjected to the same testing conditions. As can be seen, higher displacements (stresses) significantly reduce fatigue lifetime of selected material system under room temperature. Similar trend is observed for elevated temperature (100°C). Frequency effect can not be explicitly formulated in this study. However it is implicit in displacement effect, which corresponds to a particular acceleration under certain frequency. To convert obtained data into conventional S-N curves, the Finite Element Analysis (FEA) has been performed. Dynamic stresses have been calculated based on FEA model. Shell elements with plasticity capability are used to simulate the flexible substrate. The bending profile of the substrate is measured from actual bending results. The flexible substrate is bonded to a rigid plate through pressure sensitive adhesive (PSA). The two rigid plates are fixed. The thin substrate is subject to displacement load along its symmetric plane. The finite element model is shown in Figure 3.
Durability Study of Conductive Copper Traces
Figure 3. The Finite Element Analysis (FEA) model.
Figure 5. The FEA data: tensile stress versus displacement.
13
Figure 4. The FEA data: tensile strain versus displacement.
Figure 6. S-N diagram for material system A constructed based on experimental and FEA data.
The actual construction of the flexible substrate consists of several components, and the FEA model is simplified. The substrate is represented by a thin layer of equivalent thickness and effective material property. The mechanical behavior of the flexible substrates is studied in a separate publication [11]. The stress-strain curves measured at various temperatures can be found in that reference. Shown in Figure 4 and Figure 5 are the numerical results from FEA. They present tensile strain and tensile stress as a function of displacement load, respectively. Both graphs indicate nonlinearity as displacement increases. These results have been used to generate the conventional S-N fatigue curves. Figure 6 provides the typical S-N curve for a material system A, on the basis of experimental data and FEA results. As described above, three data
14
Figure 7. Digital image of fracture surface of material system A under ambient temperature (23°C).
Plastics Failure Analysis and Prevention
Figure 8. Digital image of fracture surface of material system A under elevated temperature (100°C).
points (minimum, maximum and average fatigue lifetime) are given for each loading condition to reflect the scatter of fatigue performance of various traces subjected to the same testing conditions. Clear decline in the fatigue performance of this material system is observed with the increase of stresses. It is worthy noting that the fatigue data obtained in this work results from a non-zero mean process. The non-zero stresses are induced by the substrate forming. Analysis of the effect of mean stress is to be reported elsewhere. In contrast with to our expectations, the higher temperature increases fatigue lifetime of material system A. This is clearly seen in Figure 2. It may be related to the softening mechanisms within the polymer matrix caused by elevated temperatures, which increase adhesiveness of circuit traces, slows down brittle failure process and increases fatigue lifetime. Final comparative analysis of all three material systems is unfinished at the present. Finite Element Analysis showed that within the applied loading range, the temperature effect on stress is significant while it is negligible for strain. For a displacement controlled failure process, rising temperature is likely to delay the failure time. Fracture analysis reveals the potential causes of the effect via direct observation of the failure modes. Figure 7 is the image of fracture surface in material system A under room temperature. Well-developed and aligned across the width cracks can be seen on the bonding line of the sample setup. Optical microscopy of various samples confirmed the fact that all the cracks were initiated within the copper circuit traces. However, it had been found that in few cases crack occurred along the metal/polymer interface. Simultaneous single cracks could be observed in various traces without interconnection at the initial stages. However, on the final stages of failure process they are all connected with each other through the polymer matrix. Temperature affects the failure mechanism in this particular material structure. Figure 8 presents the fracture surface of copper circuit traces (system A) under elevated
Durability Study of Conductive Copper Traces
15
temperature (100°C). An array of multiple cracks within the same copper trace could be observed. It can be attributed to softer polymer matrixes or adhesives due to higher temperature, which reduce “rigid” movements of the flex and diminish to some extent brittle crack initiation within the copper traces. Therefore, there is an apparent variation in the failure mechanism with temperature.
CONCLUSIONS AND FUTURE WORK High cycle fatigue resistance of copper circuit traces in FPC (three material systems) as a function of frequency, displacement and temperature has been analyzed. Novel testing procedure has been designed and new experimental setup has been developed. Comparative analysis of selected materials based on fatigue lifetime evaluation is in progress. Failure analysis has been performed and failure mechanisms have been identified for material system A. Typical S-N curves for the same material system are constructed. Wider frequency and temperature range are being analyzed at present. Reliability assessment is also being performed via statistical analysis of the data.
ACKNOWLEDGEMENTS The financial support of Automotive and Industrial Electronics Group, Motorola Inc. is gratefully acknowledged. Assistance in manufacturing and assembling of the continuity monitor from Mr. Ted Lester and his invaluable inputs to this project are greatly appreciated.
REFERENCES 1 2 3 4 5 6 7 8 9 10 11
P. Viswanadham and P. Singh, “Failure Modes and Mechanisms in Electronic Packages”, Chapman and Hall, New York, 1998. Proceedings: Eighth Electronic Materials and Processing, Ed. S. Rao, ASM International, 1994. J. C. Cluley, “Electronic Equipment Reliability”, John Wiley and Sons, New York, 1974. S. W. Hinch, “Handbook of Surface Mount Technology”, Longman Scientific & Technical, New York, 1988. W. Sikonowiz, “Designing and Creating Printed Circuits”, Hayden Book Company, Inc., New Jersey, 1981. G. L. Ginsberg, “Printed Circuits Design”, McGraw-Hill, Inc., 1991. C. F. Coombs, Jr., “Printed Circuits Handbook”, 3 rd edition, McGraw-Hill, Inc., 1988. D. Steinberg, “Vibration Analysis for Electronic Equipment”, Second Edition, John Wiley and Sons, New York, 1988. “Handbook of Flexible Circuits”, Ed. K. Gilleo, Van Norstrand Reinhold, 1992. S. Gurley, “Flexible Circuits: Design and Applications”, Marcel Dekker, Inc., New York and Basel, 1984. M. Lu, Z. Qian, S. Liu, R. Li and L. Poglitsch, “Thermo-Mechanical Behaviors of Flexible Substrates”, J. Electronic Packaging, in press, 1999.
Fatigue Behavior of Discontinuous Glass Fiber Reinforced Polypropylene
Mustafa Sezer Arcelik A.S., Istanbul, Turkey Ahmet Aran Istanbul Technical University, Istanbul, Turkey
INTRODUCTION Fatigue is known as the failure of materials under the cyclic loads below their yield strengths. Although there is a tendency about the increasing use of plastics in mechanical components, the designers can’t find enough data to predict the fatigue performance of plastics as it is for metals. Since the polymers are not rigid in their neat polymer state, they can be reinforced with long and short glass, carbon, aramid, polyester, or other fibers to increase the stiffness, stability, heat conductivity and fatigue resistance. Injection molding process is the most convenient to add fillers and reinforcements to the polymers. Wohler (S-N) curves are the conventional method to investigate the fatigue behavior of polymers. S-N curves have been used to investigate the tensile and flexural fatigue behaviors of long and short glass fiber reinforced polymers.1-4 Fatigue crack propagation (FCP) method is also very popular method to determine the fatigue behavior of glass fiber reinforced polypropylenes.5-8 In this study, fatigue behavior and failure of 30 wt% short glass fiber reinforced chemically coupled and uncoupled polypropylenes were investigated. One aim of this study is to create design data for the above mentioned materials in the design of dynamic components. To increase the knowledge capacity about the failure and fatigue mechanism of the materials is the second aim of the study.
EXPERIMENTAL In this study 30 wt% chemically coupled (CCPP) and uncoupled (UCPP) glass fiber reinforced polypropylene's supplied by TARGOR were used (GC30H251 and GF30H152). Maleic anhydride grafting were used in CCPP to increase the bonding property between the fibers and the matrix polymer. The polypropylenes used as matrices were homopolymer and isotactic. The average diameter of glass fibers were 10 microns. A separate study was per-
18
Plastics Failure Analysis and Prevention
formed with 500 fibers before and after the injection to evaluate the fiber breakage during the injection molding. According to the results, the glass fiber lengths have lowered to 15% during the injection.9,10 Materials properties used in the studies are given in Table 1. Table 1 Physical and mechanical properties of materials used in the tests CCPP (GC30H251)
UCPP (GF30H152)
67 MPa
41 MPa
6645 MPa
5773 MPa
Elongation at break (23oC, 5 mm/min)
2.2%
1.3%
Izod impact strength (23oC, notched)
10 kJ/m2
4 kJ/m2
151oC
95oC
Tensile strength (23oC, 5 mm/min) Young’s modulus (23oC, 1 mm/min)
Heat deflection temperature /A (1.8 MPa)
ISO 527 Type I tensile test specimens were used for static and dynamic tests. The width of parallel portion and measuring length were 10 and 50 mm respectively. A semiautomatic injection molding machine, MANUMOLD 77/30, was used to produce test specimens. Injection molding parameters were determined according to ISO 294, ISO 1873-2 and the proposals of material manufacturer. Detailed injection molding parameters were in reference.11 Test specimens were conditioned under 23οC and 50% humidity according to ISO 291 in HERAUS HC 4030 circulated climatic cabin before the static and cyclic tests. An Instron electromechanical test machine was used for the static tensile tests. For the fatigue tests, an MTS servo-controlled hydraulic test machine was used. The load ratio (minimum load/maximum load) was chosen as 0.1 and sinusoidal wave form was used in the fatigue tests. To check the reliability of taking the stroke as specimen elongation, some dummy cyclic tests were done with marked tensile test specimens. During these tests, stroke and elongation data were saved by both MTS and “Kodak Ektapro Motion Analysis (KEMA)” systems. Then elongation of marking traces were measured on the monitor of KEMA system by help of frozen and magnified images. Very satisfied results were obtained. Wohler curves were constructed with tensile test specimens which were subjected to fatigue tests at different stress coefficients. Fatigue tests conditions were same with above and each point on the curve is the average of 5 experiments. Frequency is an important parameter to occur the hysteretic failure or mixed mode (hysteretic heating and mechanical failure) failure on the test specimens. Pilot tests which were performed at 1 Hz, 5 Hz and 10 Hz, to determine the effect of test frequency on heat-
Fatigue Behavior of Discontinuous Glass Fiber
Figure 1. The variation of maximum fatigue stress of CCPP and UCPP with the number of fatigue cycles (Wohler curves).
19
ing. Temperatures were observed first by the contactless infrared thermometer and then by surface type thermocouples. Temperature was nearly constant for 1 Hz and early failures were observed for 5 Hz and 10 Hz.11 After tensile tests were applied to fatigued and unfatigued test specimens, scanning electron microscope (SEM) investigations were performed.
RESULTS AND DISCUSSION WOHLER (S-N) DIAGRAMS Fatigue test results are shown in Figure 1 and Figure 2 as Wohler curves. The fatigue strength of CCPP is much more than UCPP (Figure 1) as expected. But normalized fatigue strength (ratio of upper fatigue stress level to tensile strength) of UCPP is higher Figure 2. The variation of normalized of maximum fatigue stress of CCPP and UCPP than the CCPP (Figure 2). with the number of fatigue cycles (normalized Wohler curves). The deterioration trends of normalized fatigue strengths are similar. No endurance limits were determined for both CCPP and UCPP as seen in diagrams (Figure 1 and Figure 2). Fluctuations of fatigue test results for both materials are within acceptable limits and results can be used in engineering calculation. According to Wohler curves, improving fiber-matrix bonding does not only improve the tensile
20
Plastics Failure Analysis and Prevention
Figure 3. Fracture surfaces in FA (left) and in FFA (right) of 10 cycles fatigued of CCPP at 500 magnifications.
Figure 4. Fiber surfaces in FA (left) and in FFA (right) of naturally fatigued of CCPP at 1000 magnifications.
strength but the fatigue strength also (Figure 1). In despite of improved bonding results, normalized fatigue strength of CCPP is lower than UCPP (Figure 2) because of the debonding process. MICROSCOPIC INVESTIGATIONS Tensile tests are performed on unfatigued and fatigued specimens and their fracture surfaces were investigated. Although only brittle fracture was observed on the fracture surface of unfatigued CCPP, the crack surface of 10 cycles fatigued specimen shows 2 different regions (Figure 3). In the first region (FA), holes were observed around the fibers in the matrix and since fiber-matrix bonding does not exist anymore, the matrix has started to bear the load (Figure 3). Ductile type fracture was observed in this region (FA). The other region called as FFA has brittle fracture surface (Figure 3). Here, matrix and fibers have still good bonding. FFA type fracture was observed at the fiber ends, where the bond of matrix and fibers exists (Figure 4) and some matrix particles were seen on the surface of pulled-out fibers.
Fatigue Behavior of Discontinuous Glass Fiber
21
Figure 5. Fracture surfaces in elastic deformation portion of unfatigued UCPP at 200 magnifications (left) and fiber surface in FFA of naturally fatigued of UCPP at 1000 magnifications (right).
On the fracture surfaces of unfatigued UCPP specimens, big deterioration were observed at the fiber-matrix bondings and ductile fracture was observed on the matrix (Figure 5). On the contrary, we have not seen matrix particles on the surface of pulled-out fibers in FFA for UCPP (Figure 5). FATIGUE FAILURE MECHANISM On the crack surfaces of fatigued CCPP specimens, brittle fracture observed only in one region. Totally brittle fracture (no ductile fracture) was observed on the fracture surface of unfatigued specimens. When the brittle fracture occurred, the matrix particles were seen on surfaces of the pulled-out fibers. As a result, if some ductile fracture on the matrix and no adhering matrix particles on surfaces of the pulled-out fibers were observed, we can conclude that some fatigue loading have been applied to the specimen. At the fiber ends local stresses are high and poor bonding may occur. During fatigue loading, failure starts as debonding at the fiber ends and develops along the interface. Since failure propagates along the interfaces no adhering matrix particles were seen on surfaces of the pulled-out fibers. This type of failure results as a ductile failure on the matrix. In tensile test, matrix cracks in a brittle manner since there is no time to occur a failure at the interfaces. Failure occurs at the matrix near the fibers instead of interfaces and these matrix particles will be seen on the pulled-out fibers. In UCPP, ductile fracture was also observed on the fracture surface of unfatigued specimens. Besides, no adhering matrix particles on surfaces of the pulled-out fibers in FFA were observed because of the poor bonding in UCPP. During the microscopic study, we have also noticed that the debonding effect of the alternating stresses (the region where ductile fracture is observed) propagates as “failure front”.
22
Plastics Failure Analysis and Prevention
According to the above observations, the occurrence of a fatigue failure can be summarized as follows: 1. Failure starts as voids at the fiber ends. The fiber ends are the potential stress concentration areas because both they are discontinuous regions for stress lines and they don’t have enough sizing. 2. Fiber-matrix debonding propagates along the fibers with cyclic movements. 3. Because of the debonding process, matrix starts to bear the load and ductile deformation is observed. 4. Composite replies with more strain to the applied load for the reason of the debonding fibers. This case results with the new fiber-matrix debondings. 5. The voids continues to grow and merge into cracked areas. The cracked areas develops to form a big main crack. 6. For the reason of the debonding fibers, load bearing capacity of composite to become lower. After the FA grows to the extend that loads cannot be carried by the remaining cross-section and material fails under tensile stresses in a brittle manner.3
CONCLUSIONS 1. Fatigue failure mechanisms of CCPP and UCPP are nearly the same, except that UCPP has weaker fiber-matrix interface and crack propagates faster than CCPP. 2. During the microscopic study we notify that fatigue crack propagates as a “failure front”. 3. Fluctuations in the fatigue test results for both materials are within acceptable limits and the results can be used it in engineering calculation confidently. Fatigue data can be shown as S-N curves and both materials do not show any endurance limits.
ACKNOWLEDGMENTS The authors wish to thank for the support of Arcelik A.S. The authors also wish to thank to Research and Development Center for giving a permission to use the test capability and thank to Mr. Turgay GONUL to help us in SEM studies.
REFERENCES 1
2 3
Grove, D., Kim, H., Cooper, D. and Ellis C., Longitudinal Fatigue Behavior of Long and Short Glass Reinforced, Injection Moldable, Polypropylene Composites, 26th International SAMPE Conference, pp. 281-295, October 17-20, (1994). Grove, D. and Kim, H., Fatigue Behavior of Long and Short Glass Reinforced Thermoplastics, Advances in Automotive Plastic Components and Technology SP-1099, SAE Inc., pp. 77-83, February, (1995). Grove, D. A. and Kim, H. C., Effect of Constituents on the Fatigue Behavior of Long Fiber Reinforced Thermoplastics, ANTEC’95, SPE, (1995).
Fatigue Behavior of Discontinuous Glass Fiber
4 5 6 7 8 9 10 11
23
Dally, J. W. and Carrillo, D. H., Fatigue Behavior of Glass-Fiber Fortified Thermoplastics, Polym. Eng. Sci., Vol. 9, No. 6, November, (1969). Carling, M. J., Manson, J. A., Hertzberg, R. W. and Attalla, G., Effects of Fiber Orientation and Interfacial Adhesion on Fatigue Crack Propagation in Short-Glass-Fiber Reinforced Polypropylene Composites, ANTEC’85, SPE, (1985). Karger-Kocsis, J., Freidrich, K. and Bailey R. S., Fatigue Crack Propagation in Short and Long Glass Fiber Reinforced Injection-Molded Polypropylene Composites, Adv. Composite Mater., Vol. 1, No. 2, pp. 103-121, (1991). Karger-Kocsis, J., Freidrich, K. and Bailey R. S., Fatigue and Failure Behavior of Short and Long Glass Fiber Reinforced Injection-Molded Polypropylene, Sci. Eng. Composite Mater., Vol. 2, No. 1, pp. 50-67, (1991). Harmia, T., Fatigue Behavior of Neat and Long Glass Fiber (LGF) Reinforced Blends of Nylon 66 and Isotactic PP, Polym. Composites, Vol. 17, No. 6, pp. 926-936, (1996). Sezer, M., ABN-082 “Reinforced Polypropylene”, Research and Development Center of Arcelik A.S., Internal Research Report, (1996). Sezer, M., ABR-056 “Design and Manufacturing Properties of Reinforced Polypropylene”, Research and Development Center of Arcelik A.S., Internal Research Report, (1997). Sezer, M., Fatigue Behavior of Discontinuous Glass Fiber Reinforced Polypropylene, Ph. D. Thesis, Istanbul Technical University, (1999).
Ductile Failure and Delayed Necking in Polyethylene
W. Zhou, D. Chen, Y. Shulkin, A. Chudnovsky Department of Civil & Materials Engineering, University of Illinois at Chicago, Chicago, Illinois 60607, USA N. Jivraj, K. Sehanobish, S. Wu The Dow Chemical Company, Freeport, Texas 77541, USA
INTRODUCTION The time dependent micromechanisms of deformation and fracture has been observed in many engineering materials. Extensive studies of time dependent strain localization in form of shear bands and crazes of semicrystalline and amorphous polymers have been published.1,2 Creep yielding in polymers is well known to occur at high strain.3-5 The same phenomenon is also recognized in nonpolymeric materials, for example, a change in mechanism from ductile failure of ligaments between voids to transgranular and intergranular microcracking in chromium steel under creep has been reported.6 The neck formation and propagation constitutes the yielding process in a PE tensile drawing experiment. The engineering yield stress, σ y, is taken as the first maximum of the engineering stress. The engineering draw stress, σ dr, is defined as an essentially constant engineering stress under which the neck propagates. With the increasing application of polyethylene as a durable material, its fracture behavior has received considerable attention. It is generally accepted that all the modes of fracture in polyethylene are intimately associated with the development of cavitation, drawing and crazing or material as a precursor to fracture. In the present work, the time dependent (delayed) necking is investigated under displacement (ramp) and load (creep) control conditions. The results obtained from these experiments are employed to characterize a similar phenomenon, the ductile failure of PE pipes. Ductile failure of a PE pipe manifests itself in appearance of a bulge on the pipe wall (ballooning). The bulge is extended in the longitudinal direction of the pipe and accompanied by significant thinning of the wall. The study of PE ductile failure under sustained hydrostatic pressure tests (e.g., ASTM 2837 and ISO/TR 9080) is a very involving and time consuming process. In addition, the experiments with pipes can often require large amount of material that may not be available at an early stage of material development. Thus a
26
Plastics Failure Analysis and Prevention
methodology of material durability characterization based on tensile specimens testing is of practical and economic importance.
EXPERIMENTAL DETAILS The ASTM D-638 Type-IV tensile specimens were prepared from a European PE 100 HDPE grade 32 mm diameter pipes. The specimens, with the gage length and the cross section of 30 mm and 2 mm x 7 mm. respectively, and grip lengths of 40 mm, were cut in machine direction. All tests were conducted at ambient temperature (23°C). Tensile ramp tests were performed on an Instron Testing Machine at various strain rates. Long-term creep tests were conducted on a tensile creep station equipped with a LVDT. For the tensile creep tests, several stress levels were chosen above and below the draw stress σ dr observed in the ramp test with strain rate 3.3 × 103 s-1. Initial loading rate in the load application on the creep station is about 3 × 10–2 MPa s-1. The rate of creep strain was taken as a creep rate at the steady stage of the process.
THE RESULTS OF TENSION TESTS The engineering yield stress, σ y, is associated with significant shear band formation.7-9 A twostage process of onset of yielding has been reported.10 The first stage is the formation of thin microshear band packets spreading over a narrow region prior to the second stage of a large scale shearing of that narrow region resulting in a macroshear band formation. The second stage of this process produces most of the shear strain in the shear band. On further strain, the shear bands coalesce to form a well-defined neck whose boundaries propagate under essentially constant draw stress, σ dr. Figure 1 shows the engineering stress–engiFigure 1. Stress-strain diagrams of PE 100 HDPE ramp · · -1 test at strain rates. 1 - ε = 0.33 s , 2 - ε = 0.033 neering strain curves at various strain rates under · · s-1, 3 - ε = 0.0033 s-1, 4 - ε = 0.00033 s-1. displacement control conditions (ramp tests). The strain rate dependence of yielding stress σ y and drawing stress σ dr are presented in Figure 2. Decrease in strain rate from 3.3 × 10–1 s-1 to –3 3.3 × 10 s-1, i.e., by two orders of magnitude, resulting in decrease of σ y by about 30%. The draw stress σ dr is less dependent on the strain rate, i.e., as a change in the strain rate from 3.3 × 10–1 s-1 to 3.3 × 10– 3 s-1, resulting in only 10% decrease in σ dr. Thus, the differ-
Ductile Failure and Delayed Necking
Figure 2. Ramp test. Dependence of yielding and drawing on strain rate.
27
Figure 3. Time-to-necking and -to-ductile failure vs. applied stress (HDPE: BG10050).
ence between the yield stress and draw stress, σ y - σ dr, is reduced with the decrease of strain rate. If this tendency is extrapolated further, one can identify a certain strain rate, at which the value of yield stress would coincide with that of draw stress. This common value of σ y and σ dr is called the characteristic stress and denoted by σ 0. According to the data shown on Figure 2, the characteristic stress σ 0 of HDPE is about 10 MPa and corresponds to the ramp test with strain rate ε· 0 = 10–7 s-1. It suggests that the yielding at this strain rate occurs with no overshooting. In the creep tests, at loads above the characteristic stress σ 0, progression of a homogeneous stretching was observed until a sudden neck formation at time tn called the time to (delayed) necking. The relation between time to necking and applied stress are shown in Figure 3 by filled points. It is instructive to compare these data with stress vs. time in ductile failure of HDPE pipes at high stresses.12
INTERNALLY PRESSURIZED PIPE VS. TENSILE BAR In a thin-walled pipe of large length under internal pressure p, the hoop, axial and radial stresses (Figure 4a), σθ , σ s, and σ r are related to each other as σ s = νσ θ , σ r = –p (<0), and | σ r | << σ θ with ν being Poisson’s ratio. If one disregards the small component σ r, the state of the pipe becomes plane strain, ε s = 0 . As seen from Figure 4b, the pipe experiences yielding (ductile failure) in the hoop direction. Unlike the pipe, the prismatic bar cut out from the pipe in its axial direction and subjected to axial tension (Figure 5a) has only one non-zero stress component σ s. In this connection two questions arise. The first is about the difference in failure behavior of the pipe material with respect to hoop θ and axial s directions. In other words, would failure tests in axial direction be valid to characterize the failure in hoop direction? Comparison of the ductile failure behavior in s-direction (tests of the
28
Figure 4. Internally pressurized pipe: (a) hoop ϑ −. axial s- and normal n-directions; (b) Ductile failure (PE 100 HDPE).
Plastics Failure Analysis and Prevention
Figure 5. Isotropy of pipe material with respect to ductile failure: (a) test in axial s-direction; (b) test in hoop ϑ -direction.
tensile specimen, see Figure 5a) and that in θ -direction (tests of the pipe, Figure 5b) shows that within an experimental error the pipe material is isotropic with respect to necking. The second question is related to the fact that the tensile bar is in an uniaxial stress state, whereas the pipe is in biaxial or even triaxial stress state. Is it correct to judge the pipe failure based on experiments with the tensile bars? Let the yielding point of the pipe material be determined by the Tresca or von Mises criterion. For the pipe, one has σ 1 = σ 0 , σ 2 = νσ0 and σ 3 = -p with | σ 3| << σ 1. If the first criterion is accepted, σ y = σ 1 - σ 3 = σ 0 where σ y is the yield stress at uniaxial tension, i.e., σ s and σ r do not influence the yielding point. For the second criterion, σ y ≈ σ 0 1 – ν + ν 2 , and since the material is almost incompressible ( ν ≈ 0.45 ), σ y ≈ 0.87σ 0 . Thus, the bar is supposed to be equivalent to the pipe if the
Ductile Failure and Delayed Necking
29
stress of the bar is 13% higher than that of the pipe. However, the correspondence of the bar and pipe needs to be further investigated. The correlation between time to ductile failure and time to necking is seen from Figure 2. The open circles represent the ISO TR 9080 Hoop Stress Testing datapoints of ductile failure for a European PE 100 pipe grade HDPE. The above data also allow the estimation of the stress σ c of ductile-brittle transition at 20oC. The approximate value of σ c is 10 MPa, i.e., about the same as the characteristic stress σ 0 (see Figure 2). The equality of σ c and σ 0 is not a mere coincidence. Indeed, the yielding in PE at stress σ y is directly associated with ductile failure, and results from breaking up of the lamellae crystal stacks as well as individual crystal, resulting in largely non-recoverable morphological rearrangements.13 The brittle fracture is associated with a craze formation triggered by cavitation within the amorphous phase at stress level of σ dr and below.9,11 Thus, when σy ≅ σ dr ≅ σ 0 both ductile and brittle mechanisms of failure may occur at the same time and which one of them actually takes place is controlled by chance. It results in a large scatter of time to failure at stress about σ 0 and corresponds to the transition in fracture mechanisms from ductile above σ 0 to brittle below that value.
CONCLUSIONS The estimation of the time to ductile failure on the basis of the test for delayed necking is in a good agreement with the direct observations of PE pipes ductile failure in the sustained hydrostatic pressure test performed in accordance with ISO/TR 9080. A recent advances in an understanding of the delayed necking phenomenon, specifically σ y and σ dr dependence on strain rate in the ramp tests, led to identification of the characteristic stress σ 0 that corresponds to the ductile-brittle transition in PE fracture mechanisms. This stress, when placed on the stress-lifetime diagram, connects the ductile failure line with the brittle one. Thus, this paper, in combination with previously reported studies on brittle fracture, as well work being presented at this conference,14 offer a basis for an accelerated testing for PE pipes failure for the entire range of stresses that encompass both ductile and brittle modes of PE failure.
REFERENCES 1 2 3 4 5 6
Crist, B. Structure and Properties of Polymers, Vol 12 of “Materials Science and Technology”, Thomas, E.L., VCH Publishing 1993. Andrews E. H. Fracture in polymers, American Elsevier, New York, 1968. Turner, S. in 'The Physics of Glassy Polymers' (Ed. R. N. Haward), Wiley & Sons, New York, 1973, p. 250. Ender, D. H. and Andrews, R. D., J. Appl. Phys., 1955, 36, 3057. Brady, T. E. and Yeh, G. S. Y., J. Appl. Phys., 1971, 42, 4622. Gooch, D., J. Metals Science, 1982, 16, 79.
30
7 8 9 10 11
12 13 14
Plastics Failure Analysis and Prevention
Ma, M., Vijayan, K., Hiltner, A., Baer, E. and Im, J.: Shear yielding modes of polycarbonate. J. Mater. Sci., 24: 2687 -2696, 1989. Kim, A., Garrett, L. V., Bosnyak, C. P. and Chudnovsky, A., Modeling the process zone kinetics of polycarbonate, J. Applied Polymer Sci., 49: 877-883, 1993. Stokes, V. K. and Bushko, W. C.: Use of plastics and plastic composites: Materials and Mechanics Issues. V. K. Stokes, ed.: 1- 21, MD-vol 46. American Society of Mechanical Engineers, New York, 1993. Matsuoka, S.: Relaxation phenomena in polymers. Hanser Publishers, New York, Ch. 3, 1992. Zhou, W., Chudnovsky, A., Sehanobish, K. and Bosnyak, C.P.: Understanding of the time-dependent fracture phenomena of polycarbonate. Proc. of the 3 rd International Symposium on Risk, Economy and Safety, Failure Minimization and Analysis, Pilanesberg, South Africa, July 6-10, 1998, page 139-150. Williams, J. G, Fracture Mechanics of Polymers, Ellis Horwood Ltd., 1984. Zhou, W., Chudnovsky, A., Sehanobish, K. and Bosnyak, C.P. SPE’99, 1999. Fan, J., Chen, D., Shulkin, Y., Chudnovsky, A., Jivraj, N., Sehanobish, K., “ Application of the Crack Growth Layer Model for Understanding of the Correlation between Lifetime and Creep Behavior in Polyethylene”, SPE ANTEC 2000.
Chapter 2 Processing and Assembly The Role of a Heat Affected Zone (HAZ) on Mechanical Properties in Thermally Welded Low Density Polyethylene Blown Film
Timothy E. Weston Pennsylvania College of Technology, USA Ian R. Harrison Pennsylvania State University, USA
INTRODUCTION Low density polyethylene (LDPE) thin films are produced by blown film extrusion. In practice, the lay-flat produced during this process is slit to thin film or used in tubular fashion to produce products such as bags. In either case, it is common to thermally weld these films during the manufacturing process. The properties of the products produced depend on the strength of the film as extruded as well as the strength of the thermal weld used to join pieces of film together. One important property in LDPE film applications is cold service temperature. As the service temperature for LDPE is lowered, the material and film undergo a ductile to brittle transition. Film properties on the brittle side of this transition exhibit significantly lower deformation to failure than is exhibited in ductile film. This lower observed ultimate deformation lowers impact properties and other properties related to ultimate deformation and the drawing process such as toughness. Thermal welding affects the low temperature utility of these thin films by shifting the ductile/brittle transition temperature for the film and weld system to higher temperatures than the film without welds. Therefore, as the cold service temperature increases, the ductile to brittle transition temperature is encountered closer to ambient temperature, thus reducing the low temperature performance of welded film. It has been suggested in the literature that this shift in low temperature utility arises from a geometric concentration of stress at the weld - film interface.1
32
Plastics Failure Analysis and Prevention
Other literature sources report changed mechanical properties in a well-characterized area termed the heat affected zone (HAZ). The HAZ is defined as unmelted but changed material directly adjacent to a thermal weld. The reports of a HAZ are limited to the joining of larger parts. These larger parts were welded by other welding techniques such as butt or hot plate welding. Several key differences exist between the large parts studied and the results reported here. Clearly, in previous works parts are more massive and the weld times observed are significantly longer than weld times in thin film welds. Larger parts cool more slowly and retain the weld energy longer than a thin film. Therefore, the development of a HAZ in large parts is a function of thermal transport over much longer heating times. Thin film welding times are much shorter, typically 2-3 seconds versus minutes for large parts, and therefore the presence of a HAZ may not be a significant factor in the performance of thin film welds. A HAZ analogous to those observed in thick parts has never been reported in the case of thin film welds. The problem is then: (1) to determine if there exists a HAZ present in thermally welded LDPE thin film. (2) to determine whether this HAZ has different properties than the original blown film. (3) to determine if the shift in low temperature ductile/brittle failures are caused by geometric stress concentration or the presence of a HAZ weaker than either the original blown film or the thermal weld.
EXPERIMENTAL EXPERIMENTAL OVERVIEW Three experiments were performed in order to confirm the role of a Heat Affected Zone (HAZ) in the low temperature performance of LDPE thin films. Low temperature mechanical testing was used to determine the correlation between weld temperature and change in ductile/brittle service temperature. Optical microscopy was used to detect the presence of a HAZ and also to determine any correlation between size of the HAZ and welding temperature. Finally, Small Angle X-Ray Scattering (SAXS) was used to determine if the optical observation of a HAZ was related to a measurable structural or morphological difference between the HAZ, weld, and parent film. Welded samples were prepared using a Doboy . Model HS-C continuous thermal welder (Doboy Packaging Machinery, Inc. New Richmond WI) traveling at 20 feet/min. Figure 1 is an illustration of this welder. Two film materials, Winzen Stratofilm and Raven Astrofilm, were provided in 6 foot wide rolls of lay flat. The first material was produced by Winzen International, Figure 1. Doboy thermal welder.
The Role of a Heat Affected Zone
33
Inc. San Antonio, Texas designated Stratofilm 372, roll # 40949120 with nominal 0.8 mil film thickness. This material was selected because of failed welds during high altitude flights in the NASA scientific balFigure 2. Position of loon program. This film the HAZ in thermal has been a balloon industry welds. benchmark as early as Samples were Figure 3. Film welding showing the 8-ply arrangement. 1974. welded longitudinally in eight layers simulating high altitude balloon cap production. Figures 2 and 3 illustrate of this arrangement. Welding temperature was varied to produce five test groups representing weld temperatures of 100, 125, 150, 175, and 200οC. Welding pressure was controlled using adjustment screws on the Doboy welder in series with leaf style springs. MECHANICAL TESTING Mechanical testing was performed using an Instron 4200 tensile testing system equipped with a Series IX Automated Materials Testing System v6.02 data analysis package. A low temperature chamber was available for the low temperature experiments described below. Samples were divided into one inch wide specimens three inches in length using a sharp razor blade. Specimens that did not cut cleanly on the first try were discarded as potentially problematic due to the presence of notches. All samples were trimmed to provide an overall length of three inches. Both room temperature and low temperature mechanical tests were performed in accordance with ASTM Method D882 “Standard Test Methods for Tensile Properties of Thin Plastic Sheeting”. Strain rate for room temperature experiments was 508 mm/min (20 in/min). For both film and weld samples, an initial distance between the grip of 50.8 mm (2 inches) was employed. All samples were tested in replicates of five samples per condition. Low temperature specimens were allowed to equilibrate for a minimum of 45 minutes at test temperature. This sample conditioning is similar to those found in other works.1
34
Plastics Failure Analysis and Prevention
OPTICAL MICROSCOPY Different regions in welded LDPE thin film can be seen with the naked eye. In contrast, LDPE film is commonly translucent and very uniform in appearance. In welded film, the weld area is much more transparent than the parent film and contains minor visual inconsistencies. While not immediately obvious, there exists a region between the film and weld that is clearly not like either the parent film or the weld area. This region is small, on the order of several (2-3) mm in width, but is easily identified with the naked eye. The purpose of this experiment was to use optical microscopy to help understand the size and nature of this zone found between the weld and the parent film. For this experiment two optical microscopes were used. An Olympus SZH-ILLB stereoscope was used for low power observations, and an Olympus BH-2 microscope was used for higher power observations. The Olympus BH-2 microscope was equipped with crossed polarizers used to increase contrast in stressed samples. Both microscopes were compatible with an Olympus Model PM 10 AD Photomicrographic system. Both systems were also compatible with a Javelin MTV3 video camera attachment. SMALL ANGLE X-RAY SCATTERING (SAXS) The equipment used for the SAXS experiment was a Siemens FK-60 Cu X-ray tube operating on a Siemans Kristaloflex K710 generator. X-rays were pinhole collimated using three successive pinholes to reduce parasitic scatter. The X-ray detector used was a STAR-1 computer integrated system, manufactured by Photometrics Ltd. This system uses a CCD (charge coupled device) for X-ray detection. The STAR-1 data acquisition system is comprised of a control, interface, and instruments units. The STAR-1 was connected, using a IEEE-488 interface, to a Macintosh IIci computer. Dark current (noise) measurements were collected and computer subtracted from the scattering data. A commercial software package, IP Lab-PMI, was used to manipulate the scattering data to intensity plots or color enhanced images.
RESULTS AND DISCUSSION MECHANICAL TESTING Samples from the Winzen Stratofilm 372 were welded at 100, 125, 150, 175, and 200οC. The parent film was tested as a control. The results are also graphically illustrated in Figure 4. Welding at temperatures of 125οC or greater produced welds that varied by less than 2% in both stress at yield and 6% in % strain at break. Samples with welds produced at 100οC failed at much lower yield stress and % strain at break. The 100οC welded samples were observed to undergo failure in the weld region. The failure mode is commonly known as a
The Role of a Heat Affected Zone
35
“peel seal”. All other welded samples failed in a ductile fashion with deformation occurring throughout the length of the samples. The actual location of failure was Figure 5. 7.5x Microphotograph of W200 welded LDPE film w/rule. random throughout the sample and did not occur at the weld. The parent film also exhibited ductile failure with significant strain prior to break. The stress at break value for the film was very much in line with the values observed with the welded samples. This increase can be attributed to the weld not participating in the strain of the sample because Figure 4. Room temperature stress and strain. of a thickness roughly twice the film section thickness. All film and all welded samples failed at approximately 300% strain during ambient temperature testing. In contrast, all welded samples failed at approximately 20% strain or less at -80οC. The brittle failures at -80οC all occurred in the parent film parallel with and directly adjacent to the weld. The parent film failed in both ductile and brittle modes at -80οC. Stress at break values were comparable Figure 6. 7.5x Microphotograph of W200 welded LDPE film. for both ductile and brittle failures at all three test temperatures. As expected, the stress at break values did increase with decreasing temperature. The test condition of most interest is -60οC. At this temperature, there is a clear correlation between the number of samples exhibiting ductile failures and the welding temperature. OPTICAL MICROSCOPY Optical microscopy was used to observe and document the existence of a well-defined HAZ in samples welded from Winzen Stratofilm 372. Figure 5 is a microphotograph of the W200 welded film sample under combination front lighting and back lighting. The rule shown at the bottom of this microphotograph suggests the size of the HAZ as observed by low power microscopy to be on the order of 1.5 -3 mm. Figure 6 is a microphotograph of sample W200
36
Plastics Failure Analysis and Prevention
under 7.5x magnification. The weld is the light streak running vertically across the microphotograph. This microphotograph was lit to emphasize the film side of the weld. The film side of the weld is seen in the center to upper portion of the microphotograph. The terminal side of the weld is seen in the lower portion of the microphotograph. The parent film is seen in the top of the microphotograph. From this Figure 7. 15x Microphotograph of W200 figure it is possible to observe the darker, more opaque welded LDPE film. nature of the film, the lighter, more translucent nature of the welded material, and the brown region in between the weld and the film. This brown region corresponds to the predicted position of a HAZ. It should be noted as well that while we struggled to document this obvious transition area using microphotography, it is easily visible to the naked eye, and the lighting necessary to produce a suitable photograph detracts from the ease of identifying this region. Figure 7 is a 15x magnification microphotograph of this same region. In this figure, the weld runs vertically across the microphotograph. The parent film is located to the left of the microphotograph. The HAZ can be observed in the center of the microphotograph between the film and weld. This microphotograph suggests there may be more than simple structure to the HAZ. One pattern common to much of the HAZ region is the finger like pattern caused by alternation of light and dark areas on the microphotographs. These “fingers” seem to form in the HAZ at the weld/HAZ interface. The finger like structures are often large enough to be easily seen using the naked eye. While this structure is very obvious, their cause is not. Possible causes include microscopic thickness variations, variations in crystallinity, or melt dynamics associated with the welding process occurring at 20 feet/min. It would be necessary to design a study to control these variables to determine the reproducibility and source of the “finger” structures. SMALL ANGLE X-RAY SCATTERING (SAXS) A series of SAXS experiments were designed in an effort to confirm structural differences between the weld, HAZ, and parent LDPE film. Because of the weak scattering nature of LDPE and the film thickness the data was collected from stacked samples. Data from several different stacking experiments did show approximately a 10Å difference between the weld region and HAZ. The corresponding difference in Plot Meridian FWHM is 0.2–0.3 mrad. This data reinforces the hypothesis that there are structural differences underlying the optical observations discussed previously.
The Role of a Heat Affected Zone
37
CONCLUSIONS The optical microphotographs presented provide strong visual evidence for the existence of a HAZ in these welded LDPE thin films. The literature revealed that private sector companies had visually observed a HAZ in these LDPE films but did not realize the relevance of their observations. Our measurements of the observed HAZ scaled directly to the increase in heat input to the weld as measured by welding temperature. This observation is consistent with known modFigure 8. Brittle Failure Rate (-60ºC) and HAZ width els for thermal conductivity and literature vs. weld temperature reports of HAZ behavior. Mechanical data presented provides strong evidence of affected mechanical properties in the welded LDPE thin film. In particular, the low temperature performance of the welded film system is compromised by a shift in the temperature where ductile to brittle transition occurs in the mechanical properties. The data presented clearly demonstrates the weld region would be expected to fail long before the surrounding parent film. Finally, a correlation has been established between the increase in brittle failures and the width of the HAZ as observed by optical microscopy. Figure 8 clearly indicates this relationship. The graph in Figure 8 also demonstrates the dependence of the HAZ width and the rate of brittle failures with welding temperature. If geometry of the samples caused the failures, as observed by Simpson and Bowman, there would be no change in brittle failures with weld temperature. The increase in both HAZ width and brittle failure rate with increasing temperature suggests the increasing size of the HAZ is the key factor in the increased brittle failures seen in the welded LDPE thin film system studied here. Future work on this research will seek to more accurately establish the nature of the structure, morphology, and extent of this HAZ and then tie these structural observations to possible causes in shifts in cold brittleness temperature. Ultimately, a model for predicting the generation of a HAZ and its affect on properties such as cold brittleness will be developed.
REFERENCES 1 2 3 4
Simpson, M., Bowman, J, Polymer Engineering and Science, 31:7, 487 (1991) Simpson, D.M., Harrison, I.R., ANTEC ‘93, 1206, (1993) Winzen International, Winzen International, Inc. (1985) Winzen International, Winzen International, Inc. (1994)
The Role of a Heat Affected Zone
5 6 7 8 9 10
Weisman, D. and Alexander, H., Intern. J. Polymeric Mater., 3:33 (1974) Nieh, J.Y., and Lee, L.J., ANTEC ‘ 93, 388 (1993) Stevens, S.M., ANTEC ‘96, 1275, (1996) Stevens, S.M., ANTEC ‘94, 1258, (1994) Jang, B.Z., Uhlmann, D.R., Vander Sande, J.B., Journal of Applied Polymer Science, 29, 3409 (1984) Gupta, A., Simpson, D.M., and Harrison I.R., ANTEC ‘93, 1201, (1993)
38
Effects of Processing Conditions on the Failure Mode of an Aliphatic Polyketone Terpolymer
Nicole R. Karttunen and Alan J. Lesser Polymer Science and Engineering Department, University of Massachusetts, Amherst, Massachusetts 01003, USA
INTRODUCTION It is well known that processing conditions can play a significant role in the physical and mechanical behavior of polymeric materials. Conditions such as process temperature, cooling rate and shear rate can have significant impact on the properties of the final product. The levels of orientation, residual stress, crystallinity, etc. are a few examples of characteristics that may be altered by varying such conditions during processing. These, in turn, affect the behavior of the material. This investigation reports the effects of shear rate and cooling rate on the failure mode of extruded tubes. The material studied is a semi-crystalline aliphatic polyketone terpolymer.
EXPERIMENTAL MATERIAL AND SPECIMEN PREPARATION Material was supplied by Shell Chemical Company. The aliphatic polyketone consists of perfectly alternating units of carbon monoxide and ethylene, with a small portion (approximately 6 mol percent) of ethylene units substituted with propylene units: -(CO- CH2CH2)m-(CO-CHCH3)nProperties of this material have been examined in many studies.1-4 Material was provided in the form of extruded pipe. Five samples were produced by processing at different extrusion conditions. The chosen conditions represent the range of practical processing conditions, and therefore a range of morphologies that is most likely to be created in actual products. As a reference, an “original morphology” (OM) was produced at standard extrusion conditions. The thermal history was altered for two additional samples. These were either cooled slowly (AM-“annealed morphology”) or rapidly (QM“quenched morphology”) from the melt. The extrusion rate was altered for two additional samples, representing the highest and lowest practical shear rates. These samples are referred to as “rate 1” (R1) and “rate 2” (R2), however, the corresponding extrusion rates
40
Figure 1. Schematic of biaxial testing specimen.
Plastics Failure Analysis and Prevention
are not available from the supplier. As a note, the samples are labeled as different “morphologies”, however, a priori, the actual morphologies of the samples were not known to be different. Rather, it is the possible similarities and differences in morphology that were to be determined and used as a “link” between the processing conditions and failure response. The diameter of the specimens was approximately 22 mm with a wall thickness of 2 mm. The length was 15 cm. At each end, the cylinders were sealed with compression fittings and secured to a pipe fitting, which was threaded into steel cylinders. These cylinders were gripped in an Instron tension-torsion machine. A tap in the upper steel cylinder allowed internal pressurization with nitrogen gas. A schematic is shown in Figure 1. MULTIAXIAL TESTING PROCEDURE
The testing procedure has been described in detail elsewhere,5 and will only be discussed briefly here. The hollow cylindrical specimens were subjected to uniaxial and biaxial states of stress at 20°C at a nominal octahedral shear strain rate, of 0.05 min-1. Tests were performed in an Instron 1321 tension-torsion machine modified with a Tescom ER3000 pressure regulator. The ratio of axial load to internal pressure was held constant for each test. Different stress states were applied to specimens by changing the value of this ratio. Each of the stress states within a given yield locus was applied at the same nominal octahedral shear strain rate. This was chosen as the method of control in order to be consistent with viscoelastic theory. CRYSTALLINE PHASE Crystallinity measurements were made by differential scanning calorimetry as well as by density gradient column. Calorimetry was performed on a Du Pont DSC 2910 at a heating rate of 10oC/min. The values for the crystalline and amorphous phase densities used in crystallinity calculations were taken from.6 Crystalline orientation was determined by wide-angle x-ray diffraction, using pin-hole collimated, monochromated CuΚ α radiation. Patterns were collected on a GADDS detection system (Brucker).
Effects of Processing Conditions
41
Figure 2. Plot of octahedral shear stress as a function of volumetric strain for the original morphology sample. Solid and dashed lines represent ductile and brittle-like behavior, respectively.
RESIDUAL STRESS MEASUREMENT
Figure 3. Representation of the ductile-brittle/transition for the original morphology sample. Dashed lines represent the stress state for the corresponding images. The dotted line indicates the case of equibiaxial tension.
Residual stresses may be induced during processing due to the inability of the material to contract upon cooling. Such stresses have been shown to affect the material behavior.7-8 To determine if residual stresses were a factor in this behavior of this material, measurements were made using a procedure described by Clutton and Williams.9 In this procedure, sections of tube are cut to various lengths and “strips” of material are removed from each section. By measuring the amount of ring closure or opening, the residual stress may be determined.
DISCUSSION OF RESULTS FAILURE BEHAVIOR The original morphology sample was used as a “reference” for comparison of the other samples. From a plot of the octahedral shear stress as a function of the volumetric strain for this sample, Figure 2, it is observed that several stress states result in a ductile-type response (the criteria used to determine ductile behavior is achievement of a zero-slope condition in this type of stress-strain curve). However, while it would be expected that the largest volumetrically straining stress states would be most likely to produce brittle response, this is not observed for this material. Some of the macroscopic failures for the original morphology are displayed in Figure 3.
42
Plastics Failure Analysis and Prevention
Table 1. Crystallinity and melt temperature values determined by DSC and density gradient column Crystallinity by DSC, %
Tm, oC
Crystallinity by column, %
Original
34
225
34
Annealed
33
223
34
Quenched
33
223
34
Rate 1
33
224
35
Rate 2
35
224
36
Figure 4. Representation of the ductile-brittle/transition for the annealed morphology sample. Dashed lines represent the stress state for the corresponding images. The dotted line indicates the case of equibiaxial tension.
The cooling rate had significant effect on the material behavior. Macroscopically, the failures of the quenched morphology sample appeared more brittle in nature than the original morphology. However, by the criteria described previously, in most stress states, ductile yield was achieved. In contrast, the annealed morphology was quite a bit more brittle in failure than the original morphology or the quenched morphology. In fact, the annealed morphology was the only sample for which a brittle type of response was observed for the case of equibiaxial tension, Figure 4. From this figure, it can be seen that the brittle-like regime for this sample was considerably larger than the other samples. The extrusion rate seemed to be inconsequential to the failure behavior. The sizes of the ductile and brittle-like regimes for these two extrusion rates were similar to the original morphology. Again, ductile-types of failures were observed for the case of equibiaxial tension. EFFECTS ON CRYSTALLINE COMPONENT It appears as though the crystalline component was relatively unaffected by the different processing conditions. Table 1 presents the crystallinity as determined by DSC and by density gradient column, with no considerable differences between the samples. It in known that the crystallization kinetics of this polymer are very rapid,10 and therefore significant changes in the crystallinity were not expected. Additionally, none of the samples exhibited
Effects of Processing Conditions
Figure 5. Sample WAXD patterns for polyketone terpolymer hollow cylinders: (a) original morphology, (b) annealed morphology.
43
orientation in the crystalline phase Table 2. Residual stress values calcu- (e.g., Figure 5). lated for each sample. Similar patterns were observed for Residual hoop Residual axial all samples. As a stress, MPa stress, MPa note, is it acknowledged that Original 15.7 1.7 other differences Annealed 16.8 0.6 may exist in the crystalline phase, Quenched 18.6 0.8 for example, crysRate 1 16.9 0.2 tallite size. HowRate 2 16.5 0.5 ever, as a “rough” comparison of the apparent crystallite size, half-peak widths of the WAXD patterns were determined and found to be similar in all cases. While there were no observed differences between the crystalline components of the samples, it is noted that this material is approximately 65% amorphous. Therefore, any effects on the amorphous component may have considerable effect on the behavior of this material. This is the topic of future work. RESIDUAL STRESS
The amount of residual stress calculated for each sample is presented in Table 2. The stress in the hoop direction is considerably greater than that in the axial direction. Due to the presence of these stresses, the stress state that was applied to the specimens during testing was not equal to the actual stress state on the specimen. Rather, the applied stress was superimposed upon the existing residual stress. The original yield loci for the samples are presented in Figure 6 as axial stress vs. hoop stress. However, the presence of residual stress effectively shifts the yield loci in Figure 6 to that shown in Figure 7. In this figure, it can be seen that the brittle failures now fall in line with the case of equibiaxial tension. Thus, by considering the level of residual stress pre-existing in the specimens, we can account for the observation of ductile response in the stress state for which a brittle failure was expected.
44
Plastics Failure Analysis and Prevention
Figure 6. Yield loci for the various samples. Dashed line indicates equibiaxial loading. Solid and hollow symbols represent ductile and brittle-type failures, respectively.
Figure 7. Yield loci shifted to account for residual stress. Dashed line indicates equibiaxial loading. Solid and hollow symbols represent ductile and brittle-type failures, respectively.
CONCLUSIONS It has been observed that processing conditions are an important factor in the failure mode of aliphatic polyketone. While processing variations yield negligible changes in percent crystallinity or crystalline orientation of the material, it has been determined to affect the ductile/brittle transition through the creation of residual hoop stresses within the samples. While the residual stress accounts for the observed ductile-like behavior under equibiaxial tension, it does not account for the larger range of the brittle-like regime of the annealed sample. Further work is necessary in order to explain this observation.
ACKNOWLEDGEMENTS The authors gratefully acknowledge Shell Chemical Company for the material and financial support. In particular, the authors would like to thank Dr. C. C. (James) Kau and Dr. Piero Puccini for helpful discussions.
REFERENCES 1 2 3 4 5 6 7 8 9 10
Ash, C., International Journal of Polymeric Materials, 30:1 1-13 (1994). Ash, C. E. and Flood, J. E., Polymeric Materials: Science and Engineering, 76 110-111 (1997). Garbassi, F. and Sommazzi, A., Polymer News, 20:7 201-205 (1994). Kelley, J. W., Roane, D. R. and Le, D. M., 53rd ANTEC 3819 (1995). Karttunen, N. R. and Lesser, A. J., Journal of Materials Science, submitted. Lommerts, B. J., Klop, E. A. and Aerts, J., Journal of Polymer Science: Part B: Polymer Physics, 31 1319-1330 (1993). Williams, J. G. and Hodgkinson, J. M., Polymer Engineering and Science, 16:12, 785-791 (1976). So, P. and Broutman, L. J., Polymer Engineering and Science, 21:113, 822-828 (1981). Clutton, E. Q. and Williams, J. G., Polymer Engineering and Science, 35 1381 (1995). Holt, G. A. and Spruiell, J. E., Polymeric Materials: Science and Engineering, 76 112 (1997).
Orientation Effects on the Weldability of Polypropylene Strapping Tape
MJ Oliveira Dept Eng Polímeros, Universidade do Minho, 4800 Guimarães, Portugal DA Hemsley Polymer Microscopy Services, 52 Springfield Close, Loughborough LE 12 5AN, UK
INTRODUCTION Polypropylene tape with high tensile strength is commonly used for strapping many products ranging from light cardboard packs to heavy loads such as pallets of bricks or bundles of pipes. It is produced by extrusion followed by drawing at moderate temperatures to achieve high molecular orientation. For reducing fibrillation and to improve the weldability, the tape is embossed after the drawing stage by means of textured hot rolls. The strapping cycle comprises feeding, tenFigure1. Welding sequence. sioning and sealing of the tape around the pack. Thermal welding is the more common sealing process of polypropylene tape, replacing the traditional steel stapling. The welding process involves four steps (Figure 1): i the heater blade moves in to between the tapes ii the sealing block moves upwards slightly pressing both surfaces against the blade iii after a fixed heating time the heater blade retracts iv the sealing block moves to squeeze the melted surfaces together and cut the unused tape. After a cooling period the welded strap is released. The production of a strong weld is critical for a good performance of the strap. In this study polypropylene tapes, produced with different draw ratios or having different types of embossing patterns, were welded with a strapping machine. The effect of the welding tem-
46
Plastics Failure Analysis and Prevention
perature and the surface profile of the sealing block on the morphology and failure behavior will be presented.
EXPERIMENTAL This study was carried out on polypropylene strapping tape, of cross-dimensions of 12 x 0.6 mm, supplied by Gerrard Industries (U.K.). All the tapes were produced from polypropylene homopolymer of MFI about 4 g/600 s (230ºC, 21.6 N). The identification of the tapes is shown in Table 1. It was known that the tapes T5 to T9 were drawn in a oven at 95ºC at draw ratios of 5:1 to 9:1, as is indicated by the subscript. A sample of the extrudate from which the tapes were drawn was included in the testing program. Tape TS, of unknown draw ratio, was included in this study due to the interest in studying a tape with a different type of embossing. Table 1. Properties of the tapes Shrinkage, % Sample indentity
• •
• • •
Density, kg m-3
Birefringence x103 130oC
150oC
T5
902
28
7.4±0.5
32±1.5
T6
902
30
11.0±1.0
34±2.5
T7
901
36
12.6±0.5
38±3
T8
901
34
15.8±0.5
39±1
T9
889
36
16.3±0.5
41±1
TS
905
28
9.4±0.5
28±1
The following tests and equipments were used to characterize the tapes: density measurement by the column gradient method; microscopical observation and birefringence measurement, in cross-sections cut along the drawing direction, with a Zeiss Universal polarizing microscope, equipped with a Ehringhaus quartz compensator; scanning electron microscopy to observe the embossed surface, using a Leica S 360; determination of the shrinkage on annealing at 130ºC and 150ºC in an air circulating temperature controlled oven; determination of the tensile strength and elongation at break using a JJ type T 5002 tensile testing machine at a rate of 200 mm/min. The effective cross- sectional area of the
Orientation Effects on the Weldability
47
tape was evaluated taking into account the weight and the density of a precisely measured length of tape The tapes were welded using a semi-automatic Gerrard SA 600–1Z strapping machine at temperatures between 340ºC and 490ºC at 30ºC intervals. The complete welding cycle was 1.4 s. These welds were made using the sealing block with a serrated profile provided with the machine. For investigating the effect of the block profile and welding pressure some welds were produced with tape TS at 400ºC with a flat block and a block of higher height, respectively. The effect of misalignment on the strength of the seal was analyzed on tapes welded at 400ºC by imposing to the tape ends the maximum misalignment allowed by the machine (~2 mm). The microstructure and the mechanical behavior of the welds were studied by means of microscopy and by shear and peel testing.
RESULTS PROPERTIES OF THE TAPES The Tables 1 and 2 include the results of the measurements performed on the tapes. The density and the birefringence of the undrawn extrudate was 905 kg m-3 and 3x10-3, respectively. Table 2 – Mechanical properties of the tapes Sample
Break load, N
Break strength, MPa
Elongation at break, %
T5
1509±35
241±6
45±2
T6
1974±38
315±6
40±1
T7
1761±48
368±10
34±1
T8
2395±81
398±13
32±2
T9
2367±49
406±8
31±1
TS
1447±47
349±11
40±2
The effect of the draw ratio on the birefringence is shown in Figure 2. The birefringence of the extrudate is very low, compared to that of the tapes, evidencing the strong effect of the drawing operation on the molecular orientation. The increase in birefringence with the draw ratio is sharper for the lower draw ratios. Above the draw ratio of 8:1 the bire-
48
Plastics Failure Analysis and Prevention
Figure 3. Effect of the draw ratio on the mechanical properties of the tapes. Figure 2. Effect of the draw ratio on the birefringence of the tapes.
fringence tends to level off. Similar behavior was observed by Fransen et al.1 in polypropylene tapes and by Pezzuti et al.2 in polyethylene films. It can be observed in the Figure 3 that the tensile strength of the tapes increases steadily with the draw ratio up to the value of 8:1 and then begins to level off similarly the birefringence behavior. The strain at break decreases with the draw ratio displaying a pattern that is nearly a mirror image of the tensile strength vs. draw ratio curve. An identical behavior was observed by Ram et al.3 in polypropylene tapes. The accentuated decrease in denFigure 4. Variation of shrinkage with draw sity shown by tape T was certainly caused by voiding and 9 ratio. splitting within the structure, suggesting that the improvement in strength by increasing the draw ratio is reaching a limit. As was shown by Mahajan4 for HDPE tapes drawn at 95ºC and 120ºC those defects increase with draw ratio and with the decrease of the drawing temperature. As it is shown in Figure 4, the shrinkage increases with draw ratio. The values obtained at 150ºC are particularly relevant, from the user’s point of view, as they indicate the tendency of the tape to contract near the weld zone. The high shrinkage shown by the tapes at this temperature, between 32 and 41%, suggests that the depth of the heated zone must be kept to a minimum and the tape ends tightly gripped to avoid spoiling the seal by retraction at the welding stage.
Orientation Effects on the Weldability
Figure 5.Types of embossing: (a) - T5-T9, (b) - TS.
49
Figure 6. (a) -Example of a bad weld (TS-340ºC). (b) Example of a good weld (TS-370ºC).
The Figure 5 shows the two types of embossing patterns of the tapes. The embossing pattern of tape TS is deeper and has sharper corners than in the other tapes. The embossing induced splitting of the tapes, this being more severe in the tapes drawn at higher ratios. WELDING BEHAVIOR OF THE TAPES The welding temperature has a marked effect on the morphology, strength and failure behavior of the welds. When the heating tool temperature is too low (below 400ºC for most of the tapes) the welds show voids and splits at the interface with the unmelted material. As is illustrated in Figure 6-a, the low temperature and scarcity of the melt prevented the complete filling of the gaps between the matting surfaces resulting in a poor weld. The welds made on tape TS showed higher splitting than the others for identical welding temperatures. This behavior is certainly caused by the deeper embossing of this tape that demands a higher amount of melt to fill the gaps at the weld zone. The joints welded at too low temperatures showed low strength (Figure 8) and elongation at break. Depending on the type and amount of defects, the welds broke either at the
50
Plastics Failure Analysis and Prevention
Figure 8. Effect of the welding temperature on the breaking load under shear testing.
seam or at the interface with the unmelted material. The increase in welding temperature up to 430ºC or 460ºC in the case of tape T7 resulted in a reduction of the flaws and increase in the shear Figure 7. Typical fracture paths of a good weld (TS strength and ductility of the welds. Flow lines – 370ºC). (a) shear test; (b) peel test. and swirls were observed at the widest regions, while at the thinner zones the material oriented in the axial direction of the tape. Figure 6-b shows a typical cross section. Most the welds showed the maximum strength and ductility for a welding temperature of 430ºC (Figure 8). A typical failure of these welds is shown in Figure 7-a. The fracture path runs alternately on both sides of the boundaries of the weld zone, evidencing that the adhesion at the mating surfaces was stronger than at interface with the basic material. The excessive heating of the tapes, by using tool temperatures of 460ºC or higher changed the morphology and failure behavior of the weld and reduced its strength (Figure 8). The microstructure, that was too fine to be resolved by optical microscopy in the previously referred welds became spherulitic for some tapes and occasionally showed cracks at the mating layer. These cracks probably resulted from a combined effect of degradation of the polymer by excessive heating and increasing contraction upon cooling of the spherulitic structure. The failure generally started at one of the weld ends and moved soon to the basic material.
Orientation Effects on the Weldability
51
The peel tests were less effective in assessing the quality of the welds than the shear tests. Above an optimum temperature the peel strength is almost unaffected by the welding temperature. These results correlate well with the failure behavior of the welds under this test. Except in the case of the samples welded at the lowest temperature that fractured through the joint interface, all the others failed through the original material (Figure 7-b). This may be explained by the reduced interfibrillar strength arising from the high molecular orientation of the tape, which made it more susceptible to crack propagation than the unoriented material of the weld. The modifications made on the surface pattern of the pressing block, namely the flattening of the surface and the increase in the height to increase the welding pressure caused some modifications on the weld morphology. However it only had a marginal effect on the shear strength. The use of the flat block in the tape with deeper embossing caused the welds to be more uniform and this seems to reduce the dispersion of the results. As expected the misalignment of the tape ends reduced the shear strength of the joint. In the case of the machine used, the maximum misalignment caused a reduction in strength of 4%. The draw ratio of the tapes appears to influence its welding behavior. The tapes with higher orientation (T7 - T9) produced thinner welds and with more splits and voids than the less oriented samples. The increase in orientation increases the melting temperature.5 However for the range of draw ratios used here the difference in melting temperature that could be expected is not enough to explain the reduction in thickness observed. The increase in stiffness with the draw ratio certainly favored the squeezing of the melt out of Figure 9. Effect of the draw ratio on the maximum effi- the weld region, and is probably the main cause ciency of the tapes. of the reduction in thickness observed. The high shrinkage coupled with the high stiffness of the more oriented tapes is also probably the cause of the higher incidence of voids at the weld zone. The joint efficiency, defined as the ratio between the forces to break the welded and unwelded tape, decreases with the draw ratio up to the draw ratio of 8:1 (Figure 9). Thus, the improvement in strength achieved by drawing the tapes at higher ratios is lost on welding.
52
Plastics Failure Analysis and Prevention
CONCLUSIONS The study carried out in polypropylene tapes with different draw ratios to investigate its welding behavior allowed to draw the following conclusions: 1. The welding temperature has a strong influence on properties of the welds. An optimum welding tool temperature around 430ºC could be defined for most of the tapes. 2. The microstructure of the welds is in general very fine and shows a much lower orientation than the tapes. Welding temperatures of 460ºC or higher produce coarser textures and favor the occurrence of voids and splits at the joints. 3. The welding temperature influences the fracture path of the welds. Below the optimum temperature it runs along one of the interfaces, at the optimum range it alternates between the two interfaces and above the optimum temperature generally moves away from the weld zone. 4. The orientation of the tapes influences the morphology of the welds. The increase in orientation reduces the thickness of the weld zone and favors the formation of voids. 5. The welding efficiency decreases with increasing the orientation of the tapes.
ACKNOWLEDGEMENTS The authors express their appreciation to Gerrard Industries for supplying the samples and lending the equipment.
REFERENCES 1 2 3 4 5
P. J. Franssen, J. M. A. Jansen and B. C. Roest, Polypropylene Fibres and Textiles, 2nd Int. Conf. Plast. and Rubber Inst., London, 26-28, Sept., 1979. J. L. Pezzuti and R. S. Porter, J. App. Polym. Sci., 30, 4251-4259, 1985. A. Ram, J. Soker and J. Adorian, Plast. Rubb. Proc. Appl., 1, 363-368, 1981. S. J. Mahajan, B. L. Deopura and Y. Wang, J. Appl. Polym. Sci., 60, 1539-1549, 1996. A. O. Ibhadon, J. App. Polym. Sci., 43, 567-571, 1991.
Joint Performance of Mechanical Fasteners under Dynamic Load - Self-Tapping Screws in Comparison with Threaded Inserts in Brass and Plastic
Axel Tome; Gottfried W. Ehrenstein Institute of Polymer Technology, Univ. Erlangen, Am Weichselgarten 9, 91058 Erlangen, Germany Frank Dratschmidt EJOT GmbH&Co KG, Bad Berleburg, Germany
INTRODUCTION The use of thermoplastic components subjected to dynamic loads and at elevated temperatures becomes more common. The application of glass fiber reinforced plastics for highly loaded parts, e.g. automotive injection molded under-the-hood applications, require an overall knowledge of the joining properties.1 For low loads, snap-fit joints are especially economic. In case of higher loaded plastic parts, threaded inserts or self-tapping screws can be used for transferring high forces. This paper compares experimental results of the joint performance between self-tapping screws and both brass and plastic inserts subjected to static and dynamic loading conditions. In particular, the static and dynamic load limits will be discussed.
JOINTS WITH THREADED INSERTS For joints involving threaded inserts, an insert with an internal thread is placed within one of the parts to be joined, thereby enabling the other part to be screwed onto it with standard screws. Insert joints are used for parts that are subjected to special requirements or for purposes of facilitating assembly and maintenance. Beside standard threaded inserts in metal, which are generally assembled into the component by the hot-embedding or ultrasonicembedding methods, threaded inserts made of glass-fibre reinforced plastics, which matrix matched to the component, have been on the market for some time for use with the ultrasonic-embedding process. These have considerably shorter acoustic irradiation times and to not need to be removed from the component for recycling.
54
Plastics Failure Analysis and Prevention
JOINTS WITH SELF-TAPPING SCREWS For screw joints with self-tapping screws, a boss with a molded or drilled pilot hole is needed. The screw itself forms the thread during assembly. Self-tapping screws reduce molding and assembly costs by elimination the need for molded threads or secondary tapping operations as necessary for thread inserts. If unlimited repeat assembly is not required, special screws designed for plastic allow for a limited number of disassembly/reassembly cycles. Ten or more reassembly cycles with no deterioration in the joining properties are often possible. The clamp force/prestress force and back out torque are time dependent at a rate that is dependent on the stress relaxation characteristics of the thermoplastic boss material.2
JOINT DESIGN The design methods and dimensioning guidelines of such joints are based on static load tests usually at ambient temperatures.3-6 Besides dynamic loading,7-9 the joint is additionally exposed to temperature effects. Temperatures above 80°C up to 150°C in applications under-the-hood are common. Often the temperature is not only static, but superposed by temperature cycles. The prestress force of joints with self threading joints, which determines the quality the joint, are affected by elevated temperatures and temperature cycles.2 Earlier investigation show that the boss design can be calculated approximately.3-6 In practice, the theoretical analysis provides a starting point for design. For fine-tuning of design parameters such as the optimum pilot hole diameter, screw/insert engagement length and boss wall thickness according to the screw/insert-plastic combination require the necessity of experimental investigations.
EXPERIMENTAL MATERIALS AND SPECIMEN PREPARATION Glass fiber reinforced polyamide/nylon 6 with 30 wt% short glass fibers (PA6-GF30, Durethan® BKV 30H, BAYER AG) are used for injection molded boss test specimens. The boss geometry was kept constant (wall thickness: 4.5 mm; boss length: 26 mm; pilot hole diameter 4.0 mm) except for the pilot hole diameter for the threaded inserts which was reamed to 5.8 mm respectively up to 8.5 mm according to the insert, Figure 1a. All boss specimens for self-tapping screws had a lead-in section with a diameter of the normal screw (5.0 mm) to a depth of 2.0 mm. The moisture content of the polyamide specimen was about 1.5±0.2 wt%. Figure 1b shows the design parameter of the boss specimen and the metal screws, which are specially designed for the use with plastics. Both screws have a single lead thread without a cutting slot, a wide thread spacing (P ≈ 0.44 ⋅ d), small core diameters (dcore ≈
Joint Performance of Mechanical Fasteners
55
Figure 1. Configuration of thermoplastic boss metallic and plastic inserts (a, right) and self-threading screws (b, left)
0.54⋅d, d = normal screw diameter) and small thread angles (α = 30°) (manufactured by EJOT/Germany and RIBE/ Germany). Beside self-tapping screws, standard commercial hot-embedded threaded brass-inserts and ultrasonic-embedded plastic-inserts in sizes M4 and M5 (manufactured by Böllhoff/ Germany) were examined. The plastic insert were made of semi-aromatic copolymer based on PA6-GF60 (Grivory® GV 6H, EMS Chemie). TEST PROCEDURES During insertion of the screw, interesting aspects include the torque profile and the prestress force when tightening the screw. In order to measure these values, a screw rig is used that enables the screw assembly to be carried out at continuously adjustable speeds up to 2000 min-1. During assembly, the number of revolutions are constant. The characteristic measurement values (torque, prestress force, screw engagement length, turn angle) are measured continuously. Evaluation is carried out by a linked computer system. The metallic inserts were hot-embedded the plastic inserts ultrasonic-embedded with the parameters listed in Table 1. Table 1. Embedding parameters for the insert joints studied Ultrasonic-embedding (plastic-inserts)
Hot-embedding (metallic-inserts)
Frequency
20 kHz
Temperature
240 to 260 oC
Insertion time
0.2 to 0.5 s
Insertion time
1 to 2 s
Holding time
0.3 to 0.5 s
Holding time
2 to 7 s
Pressure
cylinder diameter 50: 1 to 2 bar
Pressure
6 bar
56
Plastics Failure Analysis and Prevention
The axial/pull-out strength under quasi-static loading was measured in the short-time pull-out test. To assess the anchoring strength under dynamic load, it is possible to make use of the highly sensitive and precise hysteresis measuring method which has been successfully used for material testing with plastic.7 Based on static load tensile tests, force controlled dynamic tensile swell tests were conducted using a Schenk servo-hydraulic testing machine in a load controlled manner. The applied waveform was sinusoidal at a constant minimum-to-maximum load ratio (R = Fu/ Fo) of 0.1. The test frequency was 5 Hz. Tests were performed at 23°C and 50% relative humidity. The characteristic values were measured according to the hysteresis measurement method. This method is described in several publications (e.g. 6,7). Hysteresis measurements allow for the simultaneous determination of four different properties during the dynamic test: forces (stresses), displacement (extensions), stiffness and mechanical energy. In addition, the damping factor is calculated from the stored energy and the energy loss, for both linear and non-linear viscoelastic behavior. Two different loading principles were used in the fatigue test. Stepwise load increased experiments reveal significant value changes and allow the determination of limits for cyclic loading. Then single load level tests prove the correctness of these limits.
RESULTS AND DISCUSSION STATIC LOAD LIMITS The determining factor for the strength behavior of insert joints are the shear stress that develops at the interface between the insert and the plastic, or between the screw flank and the insert (also for self-tapping screws) as well as the strength of the joint to the component (failure mode boss fracture). The axial pull-out force of brass-insert, plastic-insert and selftapping screw joints are shown in Figure 2. For insert joints size M5 the axial pull-out force is comparable to self-tapping screw joints with a screw diameter dl = 5 mm, however only for a reduced screw engagement length of lE = 8 mm (standard design: lE = 2dl = 10 mm). DYNAMIC LOAD LIMITS Figure 3 shows the dynamic characteristic values of brass-insert, plastic-insert and self-tapping screw joint in the stepwise load increased dynamic loading test. These are measured results for individual component joints which are subjected to statistical fluctuations with respect to the maximum upper load (see error bar in Figure 4). The loading force is shown at the top left (controlled value), displaying the curves for upper, Fo, middle, Fm, and lower, Fu, load plotted over the number of cycles. Starting at 0.4 kN the upper load was increased by 0.2 kN after every 5,000 cycles. Prior to each increase, the joint was unloaded and
Joint Performance of Mechanical Fasteners
57
Figure 2. Axial pull-out force of self-tapping screws ( ∅ 5 mm) in comparison with hot-embedded metallic inserts (size M4/ M5) and ultrasonic-embedded plastic inserts (size M4/M5) in PA6-GF30 (23oC/50% RH).4
Figure 3. Stepwise load increased dynamic loading of plastic and brass inserts (with force unloading at the end of each level) as well as self-tapping screw (type I) (without force unloading at the end of each level) in PA6-GF30 (23oC/50% RH)
returned to the damage-free basic load level that had prevailed at the start of the test. This
58
Plastics Failure Analysis and Prevention
Figure 4. Dynamic load limits of self-tapping screws ( ∅ 5 mm) in comparison to hot-embedded metallic inserts (M4/M5) and ultrasonic-embedded plastic inserts (M4/M5) in PA6-GF30 (23oC/50% RH).
unloading was only done with insert joints. The displacement (top right graph) of the joint undergoes a sudden jump when the load is increased, with the plastic-insert and self-tapping screw displaying a slightly higher spontaneous displacement at the start of the test. During the initial load steps, the system displays a linear viscoelastic behavior, since the displacement increases in proportion to the load. After approximately 50,000 cycles, i.e., as of an upper load of 2.0 kN, a more pronounced increase in displacement is observed for both the insert joint as well as the self-tapping screw joint. This would indicate a greater level of creep and hence an increase in the non-linear viscoelastic deformation, which is irreversible. In the current example, failure occurs in the brass-insert joint at approximately 75,000 cycles, i.e., with the load increase to 3.0 kN, through the boss tearing off. This is shown by the sharp increase in the displacement. The plastic insert and self-tapping screw joint failed after approximately 95,000 cycles (3.4 kN), i.e., through thread pull-out/shear-off or rather screw pull-out (self-tapping screw). The highly unproportional increase in damping and displacement that occurs with each load increase, together with the unproportional decrease in the stiffness above 30,000 cycles (brass-insert), i.e. 1.6 kN respectively above 50,000 cycles (plastic-insert/self-tapping screw), i.e. 2.0 kN, points to increasing non-elastic deformation or to potential damage development. The increasing rise in the damping, with a continuous reduction in the stiff-
Joint Performance of Mechanical Fasteners
59
Figure 5. Single load level experiment (Fo = 2.8 kN) with joints in PA6-GF30.
ness of the brass-insert at the same time, shows that initial damage to the joint and probably cracking occurs after 30,000 cycles already, which ultimately leads to the boss fracture after 75,000 cycles (3.0 kN). The dynamic load limits of the brass-insert, plastic-insert and selftapping screw are summarized in Figure 4. Under dynamic load, the pull-out strength decreases for PA6-GF30 to approximately 60% of the pull-out strength of plastic-inserts and self-tapping screw for quasi-static loading and to approximately 70% of this value for brassinserts. Figure 5 shows results from the single-step fatigue test with an upper load of 2.8 kN. The upper loads are in the range of the failure loads established in the load-increase test. While the brass-insert only displays a unproportional increase in the dynamic characteristics values above 480,000 cycles, and fails through boss fracture at approximately 500,000 cycles, the plastic insert fails above 250,000 cycles through thread pull-out/ shear-off. The dynamic behavior of the self-tapping screw is limited by screw fracture. Therefore, the interphase screw flank/plastic does not determine the dynamic load limit, but the endurancelimit of the metallic screw depends on the small core diameter. With an reduction of the upper load, the dynamic long-time behavior increases to larger cycles. Figure 6 shows joints after 106 cycles without rupture by an upper load of 2.0 kN (single load level experiment). For self-tapping screws and plastic-inserts no cracks are evident. At the runout of the brass-insert (here: size M4) notch effects cause cracks to develop, which was evident by with an increase in damping after 700,000 cycles. The decisive factor is that, even on fatigue specimen that has not ruptured, it is possible to observe cracks. This would suggest that the fatigue strength had not been attained, with this upper load at least, considering the characteristic values for stiffness and damping and microscopic studies.
60
Plastics Failure Analysis and Prevention
Figure 6. Joints after 10” cycles without rupture with an upper load of 2.0 kN. For self-tapping screw and plastic inserts in opposite to brass-insert, no cracks are evident. The cracking at the runout of brass-insert was evident with an increase in damping after 700,000 cycles.
CONCLUSIONS The static pull-out limit of joints with brass-inserts (size M5) is comparable to joints with plastic-inserts (size M5) as well as self-tapping screws (∅ 5 mm), e.g. in PA6-GF30. The joint fail via pull-out of the insert or self-tapping screw. The dynamic load limit measured in the stepwise dynamic load increase test of selftapping screw and plastic insert is about 3.4 kN, and brass-inserts about 3.0 kN. The joint failures via boss fracture (brass-insert), thread pull-out/shear-off (plastic-insert) and screw pull-out (stepwise load increase test) or screw fracture (single load test). In opposite to joints with self-tapping screws or plastic inserts, cracks can be observed in microscopic studies at joints with brass-inserts, e.g. even on fatigue specimen that have not ruptured after 106 cycles. An estimate of the upper load for durable load transformation in the plastic component is not possible at present on the basis of the measured results that have been obtained. It is, in fact, doubtful whether a fatigue strength range can be given at all in view of the notch effects. The advantage of joints with brass-inserts is the limited relaxation of the clamp/prestress force, e.g. by fixing the metallic component on the brass-insert. In every case in which the plastic component is fixed, the clamp/prestress force will be reduced by the time
Joint Performance of Mechanical Fasteners
61
dependent relaxation. In this paper we have not discussed this effects for joints with selftapping screws and plastic-inserts.
ACKNOWLEDGEMENTS The authors would like to thank the screw producing companies Ribe and EJOT, Germany, which supported these investigations. BAYER AG, Germany, is gratefully acknowledged for material support.
REFERENCES 1 2 3 4 5 6 7 8 9
Smock, D. (1994), „Plastics move under the hood“, Design News, 10 pp. 148-152. Tome, A. and Ehrenstein, G.W. (1999), „Time dependent prestress force of threaded joints. Antec 99, pp. 1327-1331. Onasch, J. (1982), "Zum Verschrauben von Bauteilen aus Polymerwerkstoffen mit gewindeformenden Metallschrauben”; Ph. D. Thesis, University of Kassel. Ehrenstein, G.W. (1995), "Mit Kunststoffen konstruieren”, Munich, Carl Hanser Verlag. Onasch, J. and Ehrenstein, G.W. (1982), "Calculation methods for the joining of plastic parts by thread-forming screws”, Translated from Kunststoffe, 11, pp. 720-724. Großberndt, H. and co-authors (1988), „Die automatische Schraubenmontage“, Ehningen, Expert Verlag. Dratschmidt, F. and Ehrenstein, G.W. (1997), "Threaded Joints in Glass Fiber Reinforced Polyamide”, J. Polym. Eng. Sci., Vol. 37/4 pp. 744-755. Dratschmidt, F. (1999), "Zur Verbindungstechnik von glasfaserverstärktem Polyamid”, Ph. D. Thesis, Institut for Polymer Techmology, Uni Erlangen/Germany. Tome, A., Dratschmidt, F. and Ehrenstein, G.W. (1999), „Threaded Inserts in Plastic“, Plast Europe Vol. 89 (5), pp. 16-18.
Defect Cost Analysis
Christoph Roser and David Kazmer University of Massachusetts Amherst, Amherst, Mass., USA
INTRODUCTION Engineering design is a complex and sophisticated task in order to create a successful product. One major element of a successful product is to reduce the cost of the product while satisfying the specifications. Specifications are the requirements towards the design responses to satisfy the user requirements. This is done by constraining the design responses. Here the development team has to achieve a trade-off between the satisfaction of the constraints and the cost minimization. However, the constraint satisfaction and the total cost of the design are coupled, because increased quality usually creates increased cost, but also reduces the likelihood of defects. The demonstrated methodology will aid the development team in understanding this coupling in order to minimize the total cost of the product by decomposing the total cost and assigning the cost towards the different cost drivers. Then the development team can analyze what causes the cost of the product and thus improve the design towards a more economic design. This enables the development team to improve the design focused on the actual cost drivers, avoiding the improvement of design responses which do not contribute towards the total cost. However, it has to be taken under consideration that a design performance is driven by other reasons than cost. This value analysis is not discussed in this paper.
SYSTEM OVERVIEW
Figure 1. System overview.
The engineering design model consists of input parameters, which are determined by the product development team, and output responses, which are a function of the input parameters. Furthermore a design is subject to noise, distorting the input parameters and the resulting output responses. Figure 1 visualizes the system as utilized in this methodology.
64
Plastics Failure Analysis and Prevention
The input parameters xi are applied to the design and the processing of the design, which in turn will determine the output responses yj. However, as the input parameters of the design and process are subject to noise the output responses will exhibit variation. Therefore, both the input parameters and the output responses are no crisp values but rather a probabilistic distribution. In the following, this distribution is denominated as φ . An example for a normal distributed input parameter is given in,1 where the additional information of the mean and the standard deviation is required. 2
1 φ x = -------------- e i σ 2π
–( xi – µ ) - ----------------------2 2σ
[1]
The distribution f of the input factors xi will determine the distribution f of the responses yj. It is assumed that the distributions of the constrained responses yj are independent of each other. φ y = f ( φ x , φx , φ x , …φ x ) [2] j 1 2 3 0 In this case, the output responses are divided into three groups, which are part of every design. One response is the utilized material cost CM, including all materials and standard components needed to create a product, including waste material like sheet metal cut offs or injection molding runner material. It does not include secondary supplies like maintenance material for the production process. These costs are included in the second group of processing cost CP, which also includes machine cost, man-hour cost, amortized tooling, etc. The third and last group contains the cost due to inadequate quality, in which the resulting part properties are compared with the correlated constraints. Examples for this group are the weight, the strength, chemical resistance, and so on. These responses can be constrained in three different ways such that the response has to be below, above or in between a given limit, where LSL is the lower specification limit and USL is the upper specification limit.
LSLj < yj yj < USLj LSLj < yj < USLj
[3]
MARGINAL PART COST The marginal part cost CMP is simply the sum of the material cost and the process cost. It is the cost required to produce one additional quantity of the design.
CMP = CM + CP
[4]
Defect Cost Analysis
65
YIELD In this methodology, a good part is defined as a part satisfying all specifications, and a bad part is defined as violating one or more specifications. Due to the stochastic nature of the input variables and the processing, the same input parameter values will not always result in identical output response values but Figure 2. Distributed response. rather a distribution. Therefore, there is a certain possibility of violating a constraint. The probability density function (pdf) of a normal distributed example is given in Figure 2, where the response is bounded by two constraints, with the likelihood of defect parts equal to the shaded areas of the curve outside of the limits and the likelihood of good parts equal to the non shaded area of the curve between the limits. The probability P of satisfying one constraint if the response yj is calculated by integrating the response distribution between the correlated LSL and the USL. In case the constraint is one sided then the other side is set to - ∞ for the LSL or to + ∞ for the USL.1 The probability can be evaluated from the evaluation of constraint satisfaction. USL i
Pi =
∫
φ y dy i
[5]
i
LSL i
This integration will always yield a value between zero and one inclusively if the distribution φ of yi is a valid probability density function. The joint probability of success for meeting multiple quality requirements, PJoint, is calculated by multiplying the probabilities P of success of each response y for all n constrained responses, assuming the independence of the responses y.2 PJoint represents the percentage yield of acceptable parts. n USL i
n
P Joint =
∏ Pi i=1
=
∏ ∫
φ y dy i i
[6]
i = 1 LSL i
TOTAL COST After calculating the marginal part cost CMP and the total yield PJoint of the product, the total cost CT can be calculated. The marginal part cost occurs for every produced part, no matter if the part is defect or acceptable. However, every defect part is rejected, therefore the total cost of the production of one good part also has to include the appropriate cost for the production of the defect parts. Therefore, the total average part cost is calculated as the marginal part cost divided by the yield.
66
Plastics Failure Analysis and Prevention
CM + CP C MP C T = ------------- = --------------------------------USL i n P Joint
∏ ∫
[7]
φ y dy i i
i = 1 LSL i
COST DRIVER ANALYSIS So far, the demonstrated procedure is - although not standard in the industry - rather well known in the research community. However, this paper will now reverse analyze the total cost CT of the product in order to determine the cost driver in the product. One possible approach would be to determine the cost of all defects and compare it with the cost of all defects excluding the investigated defect. This is similar to the cost change if the probability of constraint satisfaction for the investigated response would be perfect with no change to the other responses. Two cost drivers are already known, the material cost CM and the process cost CP. The defect cost is then the remainder towards the total part cost. This cost due to the defects CD can be calculated by subtracting the marginal part cost CMP from the total cost CT. [8] CD = CT - CMP Now it is possible to compare the defect cost of all defects with the defect cost excluding the analyzed constraint j. The defect cost excluding constraint j, C Deφj , is calculated similar to the defect cost by ignoring the performance requirement yj.
1 C Deφj = C MP ------------------------------------------USL n
∏
i
∫
[9]
φ y dy i i
i = 1, i ≠ j LSL i
As CDeφj ≤ C D the decrease in cost CDej for a optimal constraint satisfaction Pj = 1 would be the difference between the defect cost CD and the defect cost excluding j C Deφj . C Dej = C D – C Deφj [10] While this approach is mathematically justified, it has one flaw reducing its understandability and easy to use. As the joint probability of success, PJoint and the resulting total cost depend on all responses, there exists interaction between CD and CDej, i.e., a part may be rejected due to multiple defects. Therefore, the sum of all C Dej may exceed the defect cost CD.
Defect Cost Analysis
67
n
CD ≤
∑ CDej
[11]
j=1
Although the methodology is mathematically correct, it may lead to confusion if it is expected that the sum of all single defect costs equal the total defect cost. Similarly, a defect cost can be measured by calculating the expected defect cost if all other qualities are assumed perfect, Pi = 1, i ≠ j , and only the defect probability of response j is taken under consideration. 1 C Doj = CD – C MP -------------------------USL
∫LSL
i
[12]
φ y dy j j
i
For the same reasons mentioned above, the sum of all single defect costs, CDoj, may be less than the total defect cost CD. n
CD ≥
∑ CDoj
[13]
j=1
Therefore, the presented methodology of defect cost analysis will evaluate the ratio of a single defect probability with the sum of all defect probabilities. This is done by dividing the probability of failure of one response i by the sum of all probabilities of failure, which yields the effect E of the response i on the total cost, where Ej may range from 0 to 1. 1 – Pj E j = ---------------------------n
[14]
∑ ( 1 – Pi ) i=1
This percentage effect Ei of each constraint violation on the cost due to the yield is then simply multiplied with the cost due to the yield CY to get the cost due to the violation of a single constraint Ci. 1 – Pj C j = C Y E j = C Y ---------------------------n
[15]
∑ ( 1 – Pi ) i=1
The cost effect of each defect and quality specification can be measured monetary. In addition, the sum of all single defect costs equals the total defect cost.
68
Plastics Failure Analysis and Prevention
n
CD =
∑ Ci
[16]
i=1
If it is assumed that all quality responses are optimized towards a Pi of one as shown in [15] by calculating Cj, then the CDej will overestimate the defect cost, whereas the CDoj will underestimate the defect cost.
COST DRIVER EVALUATION As stated in [2], the input variables and the output responses are related. Therefore if the effect of the output responses y are known, it is possible to determine which response is the least satisfying, i.e. the most expensive response. Then it is possible to reverse the relation between the input variables and the response to evaluate the input variables, which can improve the selected response. The goal is to adjust those input variables to minimize the total cost of the product.
EXAMPLE The methodology will be demonstrated on an injection molded part as shown in Figure 3. This part consists of a flat plane with slots and holes, a small protrusion and a runner system. To simplify the relations between the input parameters and the output responses it is assumed that the part shows the behavior of a flat plate of similar dimensions. The investigated input parameters listed below include a material parameter, a design parameter and two process parameters: • Molecular Weight Figure 3. Example part. • Melt Temperature • Injection Time • Thickness In order to estimate the quality distribution, a standard deviation was estimated for each input from the knowledge of the process characteristics and applied to the input parameter. The underlying relations between the inputs and the responses are known from different models and simulations, including structural analysis, material prediction, and process simulation. These models also included noise and uncertainties as normal distributions were applied to all model parameters and other input parameters not listed above. To achieve dis-
Defect Cost Analysis
69
tributed output responses these models and simulations where run repeatedly utilizing design of experiments, with the input variables and model parameters distributed according to the probabilistic density functions of the input parameters. The distribution of the responses was estimated as normal distribution based on the sample data, where m is the number of runs per experiment, l is the numerator for the different runs and i is the numerator for the responses. m
∑ yikl =1 µ il = k-----------------∀( i, l ) m
[17]
m
∑ ( yikl – µil )
2
k=1 - ∀ ( i, l ) σ il = -------------------------------------m–1
[18]
The performance specifications included moldability, cost and design responses. Deflection Material Cost Process Cost Cycle Time Shear Rate Maximum Injection Pressure Due to the computation time required to calculate the quality responses, a second order response surface was fitted through the data points based on a central composite design of experiments, predicting the mean and the deviation of the output responses.3 Using this response distribution, the probability of success for one response Pi was determined for each response and subsequent the probability of satisfying all responses Pjoint was calculated according to [5] and [6]. Figure 4. Yield. Figure 4 shows the relations between the input parameters and the yield. It can be seen that the injection time and the thickness have the most significant effect on the yield, as the total yield varies between 0 and 1, whereas the melt temperature and the molecular weight have very little effect on the yield, which is almost constant. Note that the figure shows the rela• • • • • •
70
Figure 5. Marginal cost.
Plastics Failure Analysis and Prevention
Figure 6. Total cost.
tions at the midpoint of the design space. The relations change if the graphs are plotted at another point in the design space due to interactions. Figure 5 shows the relations between the input parameters and the marginal cost, where the centerline is the expected mean and the dashed lines on both sides represent three standard deviations from the mean. The marginal cost was created based on the volume of the part, its production time and the required Figure 7. Cost driver. machine size based on the maximum pressure. As there is large variation in the marginal cost, the mean values are flat and only the thickness of the part shows an effect due to the increase in material consumption and cooling time. Using [7] the total cost is derived based on the marginal part cost and the yield as shown in Figure 6. It can be seen clearly that the total cost rises if the yield goes down. This graph can be derived from Figure 4 and Figure 5. Note that the total cost increases dramatically if the yield becomes very low. Using the information from above it is possible to estimate the effect of the various cost drivers towards the total cost using [8] and [15]. This effect is shown in a bar graph in Figure 7 for a sub optimal point in the design space with the following values: Melt Temperature: 294°C Molecular Weight: 100,000 mers Injection Time: 0.6 s
Defect Cost Analysis
71
Thickness: 1.7 mm The total cost of producing one part for the given design space point is $0.85. The cost drivers for this given design point are Material Cost: $0.22 Process Cost: $0.03 Fill Pressure: $0.15 Deflection: $0.45 Shear Rate: $0.00 Melt Front Temperature: $0.00 It can be seen that in this case the main cost driver are deflection defects, followed by the material cost and the fill pressure defects. Therefore, significant cost is generated because the product is not stiff enough and is likely to bend or break under the specified load. Applying the relations between the responses and the input parameter, it was determined that the thickness is the main driving input variable for those responses. By increasing the thickness to 1.95 mm, it was possible to reduce the defect cost due to the stiffness and fill pressure to $0.00, with only a slight increase in material cost to $0.25 and process cost to $0.04, minimizing the total cost of the product to $0.29 and creating a robust designed part.
CONCLUSIONS The analysis of the cost drivers enables the development team to improve the design by focusing on the most significant contributions toward the cost. This analysis is measured in monetary units, making the results easy to understand and to estimate the impact of the different cost drivers. Future research includes an improved estimation and visualization of the significance of the input parameters towards the cost drivers, enhancing the understanding of the relation between the input parameter and the total cost. It is also intended to implement the shown methodology in a commercially available CAD software package.
NOMENCLATURE µ σ φx ϕy
Mean Standard Deviation Distributed Input Parameter
CD
Distributed Response Cost due to Defects
C Deφj
Cost due to Defects excluding j
CDej CDoj
Cost reduction due to Defects excluding j Cost due to Defects of Response j only
72
Ci CM CMP CP CT Ei i,j k l LSL m n o Pi PJoint USL xi yi
Plastics Failure Analysis and Prevention
Cost due to a Single Response Material Cost Marginal Part Cost Process Cost Total Cost Percentage Effect of Response I Counters Sample Counter Experiment Run Counter Lower Specification Limit Sample Size Number of Constrained Responses y Number of Input Parameters x Probability of Success for one Response Joint Probability of Success Upper Specification Limit Input Parameter Output Response
ACKNOWLEDGMENTS The author would like to thank Stephen Shuler and Don Richwine from GE Plastics. Portions of this work was supported by the National Science Foundation (Grant No. DMI-9702797).
REFERENCES 1 2 3
Devore, J.L. (1995) Probability and Statistics for Engineering and the Sciences. Wadsworth. Papoulis, A. (1991) Probability, Random Variables, and Stochastic Processes. McGraw-Hill. Schmidt, S.R., et al. (1994) Understanding Industrial Designed Experiments. Air Academy Press, Colorado Springs, Colorado.
Chapter 3 Environmental Effects Environmental Stress Cracking (ESC) of ABS (II) Takafumi Kawaguchi, Hiroyuki Nishimura, and Fumiaki Miwa Osaka Gas Co.,Ltd., Japan Takashi Kuriyama, and Ikuo Narisawa Yamagata University, Japan
INTRODUCTION It has been known that plastics fail even under very low stress when they are in contact with particular chemical agents. This phenomenon is called environmental stress cracking (ESC). ESC is of great importance in materials selection, because it sometimes becomes the cause of failure in actual use of plastic parts. Woshinis and Wright investigated the cases of many failure cases of plastic parts and they concluded that about one-third of the plastic parts failures were caused by ESC.1 In particular, outer parts more frequently come into contact with many kinds of agents. The evaluation of the resistance of plastics to ESC is, therefore, important in material selection of outer parts. For the material of outer parts, acrylonitrile-butadiene-styrene (ABS) co-polymer is widely used because of its favorable cost/performance ratio, luster, and resistance to impact. There are some reports on the ESC of ABS. They have shown that some kinds of chemical agents such as organic solvents and surfactants cause ESC of ABS.2,3 The authors investigated the mechanism of the crack propagation of ABS in non-ionic surfactants by ECT tests and a transmission electron microscope (TEM), and reported that the level of local stress at the crack tip was the dominant factor which determined the mechanism of crack propagation.4,5 It was found in the study that when the local stress at the crack tip was low, a small massive crazed zone originated from the penetration of a nonionic surfactant and a crack propagated by ESC. As the local stress ahead of the pre-crack tip was relatively high, the toughening mechanism due to the deformation of the rubber particles and the crazing acted ahead of the crack tip and resulted in the arrest of crack propagation.
74
Plastics Failure Analysis and Prevention
In this study, the dependence of the ESC of ABS on temperature and on the kind of surfactant was investigated by ECT tests. The fracture surfaces of the specimens in these tests were investigated by a scanning electron microscope (SEM).
EXPERIMENTAL MATERIALS AND SURFACTANTS The material used in this study was obtained from Ube Saikon Co., Ltd. Two types of non-ionic surFigure 1. The shape of the specimen for ECT tests. factant (surfactants A and B) were used in this study. One was a kind of poly-oxyethylenealkylphenyllether and the other was a kind of polyoxyethylenealkylether. Their molecular structures are shown below. Surfactant A: 4-C8H7C6H4(OCH2CH2)12OH Surfactant B: C12H25(OCH2CH2)4OH ECT TESTS AND SEM ECT tests were performed under constant loading conditions. Crack length was measured by a CCD camera. The ECT tests were performed in non-ionic surfactants, and the temperatures of the specimens and the surfactants were kept constant (23°C or 50°C) during the tests. Figure 1 shows the shape of the specimens used in the ECT tests. They were cut out from compression-molded sheets. The fracture surfaces of the specimens in ECT tests were observed by a Hitachi SEM S-530 at an acceleration voltage of 20 kV.
RESULTS AND DISCUSSION OBSERVATION OF CRACK PROPAGATION BEHAVIOR IN ECT TESTS Figures 2 and 3 show the result of an ECT test performed in surfactant A for σ = 8.2x105 Pa at 23°C and 50 °C, respectively. The x-axis denotes time after the initiation of crack propagation and the y-axis denotes crack length. In Figure 2, it is seen that the curve of crack length can be divided into three regions (regions A, B, and C). In region A, crack propagation began after an incubation time. The whitening zone ahead of the crack tip could not be recognized by a CCD camera. In region B, crack propagation stopped. In this region, the whitening zone ahead of the crack tip was rather large and it could be recognized by a CCD camera. Following this, the crack propagated with a repetition of region A and region B, and in the end, ultimate failure occurred (region C).
Environmental Stress Cracking of ABS
Figure 2. The ECT test result in surfactant A for σ = 8.2 x 105 Pa at 23oC.
Figure 4. The ECT test result in surfactant B for σ = 8.2 x 105 Pa at 23oC.
75
Figure 3. The ECT test result in surfactant A for σ = 8.2 x 105 Pa at 50oC.
Figure 5. The ECT test result in surfactant B for σ = 8.2 x 105 Pa at 50oC.
The behavior of crack propagation shown in Figure 3 was quite different from that in Figure 2. In the initial stage of the test, the crack propagation rate was nearly constant and the arrest of crack propagation was not observed. At a crack length of about 25 mm, the crack propagation stopped and the whitening zone ahead of the crack tip was observed. The total time to failure was rather short compared with that in Figure 2. Figures 4 and 5 show the results of an ECT test performed in surfactant B for σ = 8.2 x 105 Pa at 23oC and 50 °C, respectively. Although the total time to failure was also much shorter when tested at a higher temperature (Figure 5), the crack propagation behavior was similar between the results tested at different temperatures. It was also found that surfactant B gave shorter total time to failure than surfactant A. From these results, it was expected that the penetration of surfactant A into the materials was not so active as that of surfactant B. Figure 6 shows the SEM image of the fracture surface of the ECT test shown in Figure 2. The crack propagation direction is shown by arrows. In Figure 6, the fracture surface near the pre-crack is flat and it indicates that the crack propagated by ESC. This flat area corresponds to region A in Figure 2. In region B, the local stress at the crack tip was higher than that in region A, and there were several areas which had rougher structures than that in
76
Plastics Failure Analysis and Prevention
Figure 6. The SEM image of the fracture surface of the ECT test shown in Figure 2.
Figure 7. The SEM image of the fracture surface of the ECT test shown in Figure 3.
region A. The rough areas seemed to be the trace of the whitening zone observed in the ECT test, and these areas corresponded to the point of the arrest of crack propagation in the ECT test. The fracture surface of region C was rather rough and it indicated that the specimen fractured in that region mainly by stress.
Environmental Stress Cracking of ABS
77
It was found from these results that the change of morphology of the fracture surface of the ECT specimen corresponded to the behavior of the crack propagation in the ECT test. Figure 7 shows the SEM image of the fracture surface of the ECT test shown in Figure 3. In Figure 7, the flat area is relatively wide and it corresponds to the fact that in the ECT test, the arrest of the crack propagation was not observed until the crack length was long enough. DEPENDENCE OF CRACK PROPAGATION BEHAVIOR ON TEMPERATURE AND ON THE KIND OF SURFACTANT From these results, it was found that the rise of temperature had different effects on different surfactants. For the case of surfactant B, the rise of temperature had the effect not only of shortening the total time to failure but also of changing the mode of crack propagation. The rise of temperature caused the specimen to be more liable to rupture by ESC without the arrest of crack propagation caused by the change of morphology at the crack tip. On the other hand, for the case of surfactant A, which gave longer total time to failure than surfactant B, the rise of temperature had the effect of shortening the total time to failure but the crack propagation behavior did not change very much. As described above, the authors reported the mechanism of the ESC of ABS, which was investigated by ECT tests and a TEM. The crack propagation mechanism is shown schematically in Figure 8. These studies revealed that when the local stress at the crack tip was low in the initial step of the ESC of ABS, a small massive crazed zone originated by the penetration of the non-ionic surfactant. As the local stress ahead of the crack tip was relatively high because of the crack growth, the toughening due to the deformation of the rubber particles and the crazing occurred ahead of the crack tip and resulted in the arrest Figure 8. Schematics of crack propagation mechanism in ECT tests. (a) shows the craze of crack propagation. The structure of the damaged zone in small area at the crack tip and (b) shows the ahead of the crack tip was similar to that of PA/PPO craze and the deformation of the rubber partialloy, which indicated that shear banding and crazing cles in large area at the crack tip. coexited.6 From the results of this study, it is suggested that the change of morphology at the crack tip was also affected by the penetration of the surfactant and the crack propagation rate. When the crack propagation rate was high and the penetration of the surfactant was not very active, the morphology change shown in Figure 8(a) tended to appear. On the other hand, when the crack propagation rate was low and the penetration of the surfactant was active, the morphology change shown in Figure 8(b) tended to appear.
78
Plastics Failure Analysis and Prevention
The results of this study can be understood from the mechanism of crack propagation explained above. When the temperature was low, the crack propagation rate was low and the specimens were more likely to have the morphology shown in Figure 8(b), and it resulted in the arrest of crack propagation shown in Figures 2 and 4. When the temperature was high, the rise of temperature had different effects on surfactants A and B. For the case of surfactant A, the penetration of the surfactant was not very active, and the crack propagated mainly with the change of morphology shown in Figure 8(a). For the case of surfactant B, the penetration of the surfactant was more active, and the crack propagated with the change of morphology shown in Figures 8(a) and (b).
CONCLUSIONS Environmental stress cracking (ESC) of acrylonitrile-butadiene-styrene (ABS) copolymer caused by two kinds of non-ionic surfactants was studied by edge crack tension (ECT) tests. The dependence of the ESC on temperature and on the kind of surfactant was investigated. The fracture surfaces were investigated by a scanning electron microscope (SEM). It was found that when the temperature was low, the crack propagated by ESC and successive arrest of crack propagation which was caused by the change of morphology in large area at the crack tip. It was also found that when the temperature was high, the rise of temperature had a different effect on each surfactant. In that case, the penetration of the surfactant into the specimen was active, and the crack propagation behavior was almost the same as that at low temperature. On the other hand, if the penetration of the surfactant into the specimen was not so active, the crack propagated mainly by ESC.
REFERENCES 1 2 3 4 5 6
W. A. Woshinis and D. C. Wright, Adv. Mat. Proc., 1994, vol 12, p 39. D. L. Faulkner, Polymer Eng. Sci., 1984, vol 24, p 1174. R. P. Kambour and A. F. Yee, Polymer Eng. Sci., 1981, vol 21, p 218. T. Kawaguchi et al., ANTEC '98, p3175. T. Kawaguchi et al., J. Polym. Eng. & Sci. (in press). H-J. Sue and A. F. Yee, J. Mat. Sci., 1989, vol 24, 1447.
Residual Stress Development in Marine Coatings Under Simulated Service Conditions
Gu Yan & J R White University of Newcastle upon Tyne, UK
INTRODUCTION Polymer coatings are used extensively for corrosion protection of metals in marine environments. Solvent loss and, in the case of thermosets, the curing process, causes shrinkage of the coating. When it is applied to a stiff substrate the shrinkage in the plane of the coating is resisted and bi-axial tensile residual stresses form. If application of the coating is made at a temperature different from the subsequent service temperature then there will be further residual stresses that result from differential thermal expansion of the coating and substrate. The coating will always have a greater thermal expansion coefficient than the substrate so if the service temperature is less than the application temperature there will be a further increment of tensile residual stress from this source. The stresses may lead to failure of the coating by causing it to crack or to detach from the substrate (flaking, de-lamination, blistering etc.). It is therefore important to have methods to measure the level of residual stress so that its contribution to failure may be assessed. If the substrate is thin and if the coating is applied to one side only then the tensile stress in the coating causes the coating-substrate combination to bend to restore moment equilibrium, with the coating side becoming concave. The severity of the curvature depends on the level of stress. The most common experimental method of measuring the residual stress in coatings uses a measurement of the curvature, from which the stress can be computed provided that the thicknesses and Young's moduli of the two components are known. In the research described here this approach was used and the behavior of a thermoplastic coating and a thermoset coating were compared during solvent evaporation/curing. The curvature was measured using a strain gauge, permitting continuous monitoring. The method was further extended to examine the changes in residual stress when the coating was submerged in water and when it was removed again to dry out. The coatings formed the basis of an experimental undercoat-top coat system of the kind commonly used in marine applications and measurements were also made on bi-layers.
80
Plastics Failure Analysis and Prevention
EXPERIMENTAL SAMPLE PREPARATION The coatings used here were experimental materials designed for marine applications. The thermoplastic, coded "Anticorrosive A" (or "AA" from now on) was a single mixture containing a vinyl copolymer, solvent (xylene), tar pitch, pigments (57% solids by weight, 38% solids by volume). The thermoset was a two-component system, prepared just before application as with any common commercial two-pack epoxy, and was coded "Anticorrosive B" (or "AB"). AB also contained hydrocarbon resin and pigment and the solvent was Shellsol A. The coating thickness was computed from the mass of the coating after the solvent had disappeared and the density of the solid residue. The coatings were applied to thin steel shim substrates, chosen because the coatings were designed for marine applications on steel structures. Substrates 100, 150, 200 and 250 µ m thick were used both to determine the best value and as a check on the reproducibility of the residual stress measurements. The shim was cut into coupons 150 mm x 25 mm. Surfaces were prepared for coating and for strain gauge attachment using emery papers #120, #500 and #800 and finally cleaned using acetone or xylene. A strain gauge was attached to the side of the substrate that was to become the uncoated side using a cyanoacrylate adhesive. A microcrystalline wax coating (M-Coat W1) was applied over the strain gauge and lead connections to water proof them and permit operation when submerged in water. The substrates were held flat on a magnetic table and the coating applied by hand brush. The coatings were suitable for spraying but spray equipment was not available that could be used in or near to the laboratory in which the subsequent measurements were conducted. The coatings were allowed to dry or cure in a room held at 30±1oC and the strain gauge signal was monitored continuously. Solvent evaporation was normally monitored for 14 days after which time the changes recorded in curvature or mass were minimal in AA. Thus samples for investigation of the effect of water immersion or for overcoating with AB were dried for 14 days before the next phase of the experiment. WET/DRY CYCLING Coated substrates with strain gauges attached to the uncoated surface were placed in an empty tank in a room at 30±1oC then submerged in distilled water at 30oC, taking care not to disturb the strain gauge reading during filling. The strain gauge signal was monitored for 24 hours or 48 hours then the tank was emptied carefully. The strain gauge reading was monitored for a drying out period equal to the initial immersion period then the cycle was repeated.
Residual Stress Development in Marine Coatings
81
TEMPERATURE CYCLING Tests were conducted with the samples immersed in water at 5oC then at 30oC using a 48 hour dwell time. EVAPORATION KINETICS The solvent evaporation kinetics of the coatings were investigated by measuring the weight changes on specially prepared substrates without strain gauges attached. The coating was applied and the first weighing made as rapidly as possible in a Mettler AT analytical balance measuring to 100 µ g. Readings were then taken every 10 seconds for the first 10 minutes then at increasingly long intervals. Measurements were continued for 14 days. ABSORPTION/DESORPTION KINETICS The samples used for the study of evaporation kinetics were then used to investigate the absorption and desorption of water. During absorption, the samples were immersed in water and removed periodically for weighing. Each time they were removed the surface water was removed with blotting paper, they were weighed, then returned to the immersion tank as rapidly as possible. At the end of 48 hours they were removed from the tank, the surface water removed and they were allowed to dry out in room air, taking weighings periodically. After 48 hours of drying out the samples were re-immersed and the cycle repeated. YOUNG'S MODULUS OF THE COATINGS To calculate the residual stress from the curvature of the film plus substrate requires knowledge of the Young's modulus of the coatings. This was measured using tensile tests conducted on dog-bone shaped samples cut from free films of AA and AB. Measurements were made on samples as follows: (a) as-prepared; (b) after immersion in water for 6 days at 30oC; (c) after immersion in water for 5 days then dried out for 1 day at 30oC; and (d) after immersion in water for 3 days then dried out for 3 days at 30oC. The Young's modulus was calculated from the small strain part of the load-deformation curve, which was fairly linear. For the combinations of coating and substrate thicknesses used in this work the Young's modulus is not very critical in the measurement of residual stress. CALCULATION OF RESIDUAL STRESS The residual stress in the coating was calculated using an elastic analysis that assumed that the curvature was spherical, that is that the curvature transverse to the coupon axis was equal to that measured along the coupon axis. The analysis was basically that described by Corcoran.1,2
82
Plastics Failure Analysis and Prevention
Figure 1. Development of residual stresses during solvent evaporation in AA coatings of different thickness.
Figure 2. Development of residual stresses during solvent evaporation in AB coatings of different thickness.
RESULTS RESIDUAL STRESSES DURING SOLVENT EVAPORATION Figures 1 and 2 show measurements of the residual stresses in coatings of different thickness for periods of 14 days or more. The stress in AA coatings of thicknesses ranging between 113 µ m to 200 µ m converges to a common value of approximately 1.2 MPa after about 12 days (Figure 1). The coatings with thicknesses below 150 µ m display a stress maximum (of nearly 1.8 MPa for the thinnest) at short times (<10 hours) before decaying to the common value. For coatings thicker than 150 µ m the rate of approach to the final stress value is progressively slower as the thickness is increased. With AB the stress built up most rapidly in the thickest coatings but appeared to be approaching a constant value after 14 days whereas the stress in the thinnest coating (212 µ m) was still climbing after 21 days (Figure 2). The mass loss measurements showed a three stage process. The first stage is free surface evaporation, followed by a mixed kinetics stage, and finally diffusion controlled evaporation.2 Since solidification does not proceed uniformly across the coating some stress build up occurred due to solidification near the edges while the mass loss characteristic was still in the first stage.2 This effect was greater in AA than in AB.2 RESIDUAL STRESSES DURING WET/DRY CYCLING Figure 3 shows the variation in residual stress in AA during wet/dry cycling with a period of 48 hours (that is 24 hours water immersion followed by 24 hours drying out in a room at 30oC). The stress increased rapidly to about 0.4 MPa during the first 2 hours of water immersion then increased much more slowly during the remainder of the first immersion
Residual Stress Development in Marine Coatings
Figure 3. Development of residual stress in an AA coating 165 µ m thick during wet/dry cycling at 30oC (48 h period).
83
Figure 4. Development of residual stress in an AB coating 212 µ m thick during wet/dry cycling at 30oC (48 h period).
period. Some reduction of stress was observed during the following drying out period. For subsequent cycles the changes were more modest, but for each complete cycle the increase obtained during immersion was greater than the drop obtained during drying out. It is notable that the stresses recorded are all tensile: if the major effect were swelling of the coating by the uptake of water then the stress would have been compressive and the curvatures would have.been in the opposite sense. The stress observed in AB was also tensile Figure 5. Development of residual stress in a bi-layer coating consisting of 256 µ m AB on top of 179 µ m immediately after immersion (Figure 4) but it o AA during wet/dry cycling at 30 C (48 h period). quickly reversed to become compressive within half an hour (see reference (2) for a presentation of the results with an expanded time scale). After a compressive minimum of about 1 MPa the stress magnitude in this coating (212 µ m thick) reduced for the remainder of the period of water immersion. On removing the water from the tank a further increment of compressive stress was observed but this was quickly reversed and at the end of the first dry period the stress was only slightly compressive. Further wet/dry cycles gave a compressive increment during the wet period followed by a larger tensile increment on drying out so that the net effect was a drift towards tensile
84
Figure 6. Development of residual stress in an AA coating 252 µ m thick during wet (5oC)/dry (room temperature) cycling (96 h period).
Plastics Failure Analysis and Prevention
Figure 7. Development of residual stress in an AB coating 235 µ m thick during wet (5oC)/dry (room temperature) cycling (96 h period).
stresses (Figure 4). The amplitude of change observed during a cycle increased progressively. Results similar to those for AB are shown in Figure 5 for a bi-layer of 256 µ m AB over 179 µ m AA. The main difference between Figures 4 and 5 is the stress scale, which is expanded for Figure 5. Note that for the bi-layers it is assumed that the application of the AB top coat does not change the characteristics of the AA and that the change in curvature of the AB+AA+substrate combination is caused by stress changes in the top coat only. Changes in stress in AA caused by absorption of solvent from AB are ignored. Detailed differences occurred in the stresses observed for different coating thicknesses and, for bi-layers, different combinations of coating thicknesses.2 Tensile stresses of nearly 2 MPa were observed during the drying out phase of the second and third cycle of an AB coating 293 µ m thick. In bi-layers the stress after several cycles depended on the relative thickness of the two components and could be either tensile (generally when AA thickness was greater) or compressive (generally when AB thickness was greater).2 TEMPERATURE CYCLING Cooling samples to 5oC produced large tensile stresses which relaxed significantly during the cold dwell (Figures 6-8). In AB the stress reversed on returning to 30oC and the stress changes were repeated each temperature cycle (Figure 7). In the AA coating there was a progressive drift to higher (tensile) stresses (Figure 6). Bi-layers showed behavior closer to AB than to AA (Figure 8).
Residual Stress Development in Marine Coatings
85
DISCUSSION The residual stress development in AA coatings was similar to that observed by Croll3,4 who also found that the residual stress in thermoplastic coatings reached an equilibrium value that was independent of the thickness and that the thickest coatings took the longest time to reach equilibrium. Tensile stresses form as the result of the volumetric shrinkage that accompanies the loss of solvent. During the early part of this process the coating is still fluid and stresses begin to form Figure 8. Development of residual stress in a bi-layer only when sufficient solvent has been lost for the coating consisting of 264 µ m of AB on top of 142 µ m o of AA during wet (5 C)/dry (room temperature) cycling coating to develop some energy elastic resistance (96 h period). to deformation. The time dependence of stress build up is determined by the diffusion of solvent through the coating and by the relaxation processes in the coating. The concentration profile will be dependent on the coating thickness and the relaxation rate will depend on the concentration. It is thus curious that the final stress level should be independent of coating thickness. The residual stresses in AB thermoset coatings were also tensile but showed greater scatter in magnitude and did not always approach a steady value even after 22 days. Croll5 also investigated thermoset coatings but used a solventless amine-cured epoxy. In his studies the coatings developed compressive stresses when thin (<55 µ m, thinner than any of the coatings investigated in the current work) and tensile stresses when in the range of thicknesses used here. Croll could not use solvent evaporation to explain stress development and he attributed the tensile stress to structural changes during the curing process. He surmised that compressive stresses were caused by swelling due to water absorption (from the atmosphere). No attempt was made to control the humidity in the experiments reported here and the small lack of consistency between different runs with AB coatings may have been caused by different contributions from this source. In the case of thermoset coatings the diffusion of solvent becomes progressively more difficult as the polymer network develops and release of solvent may proceed for an extended period of time. When AB was overcoated on top of a dry AA coating, solvent release from AB was not only into the air at the free surface but also into AA at the interface between the two coatings. Solvent entering AA will cause swelling giving an increment of compressive stress so that the overall build up of stress was much slower than for a similar AB coating applied
86
Plastics Failure Analysis and Prevention
direct to the substrate and the increment of stress due to the AB coating was much less than that obtained with an AB coating alone.2 The behavior of the coatings when immersed in water and on subsequent drying out requires careful consideration. The initial tensile stress observed in AA coatings has not been explained with certainty. It is speculated that water may plasticize the coating, assisting the escape of residual solvent (or some other minor component). Subsequent changes in stress on dry/wet cycling are small but the sense of the changes are opposite to those which would be caused by water swelling during immersion and reversal of this effect during drying out. It is as if water has occupied the free volume and provided attractive forces to draw the molecules closer together. After water immersion the measured Young's modulus of AA was higher than after solvent evaporation and it increased still further if allowed to dry out partially. This could be explained if water acted both to plasticize the polymer and to provide stronger intermolecular bonds and if the water participating in plasticization was less tightly bound (and more easily lost on drying out) than that providing intermolecular bonding. An initial increment of tensile stress was also observed in AB coatings on water immersion, possibly caused by a similar mechanism to that in AA. After about half an hour this effect reversed and subsequently for all phases of the wet/dry cycling the changes in stress were consistent with swelling by water (giving compression) with reversal during desorption of water. The overall drift in stress in the tensile direction could be due to further solvent evaporation (assisted by water plasticization of the coating). Broadly similar results were obtained by Negele and Funke6 using a simpler epoxy coating. Of perhaps greatest interest here are the results obtained with AB coatings on top of AA coatings. The results are explainable qualitatively in terms of water diffusing through the AB coating and on into the AA coating during immersion and then this process reversing during drying out. The concentration gradients will be complex and will cause significant inertia in the time signature of the changes. As a result of the different stress responses of AA and AB coatings to water the sense of stress in the bi-layer coatings depended on the relative thickness of the two layers, with smallest stresses occurring when their thicknesses were approximately equal. The largest stresses were obtained during the temperature cycling experiments. Differential thermal contraction is believed to be responsible for the generation of tensile stresses of the order of 4MPa in AB coatings on immersion into water at 5oC. Partial relaxation of this stress then occurred and this caused the formation of compressive stress when the sample was restored to a higher temperature. The behavior of AA was basically similar but with a drift towards a permanent tensile stress. AB on top of AA showed behavior similar to that of AB.
Residual Stress Development in Marine Coatings
87
CONCLUSIONS The highest residual stresses observed in this study were caused by differential thermal contraction between coating and substrate. A temperature change similar to that between a dry dock in a warm climate and the open sea gave stresses of 4 MPa and more, a significant fraction of the failure strength. Other sources of residual stress are complex and are probably highly specific to the coating composition. When using bi-layered coatings the changes in stresses were moderated somewhat and it appears that a significant and beneficial reduction in the stress magnitude can be achieved by appropriate combination of thicknesses of the two layers.
ACKNOWLEDGEMENTS The authors acknowledge Courtaulds Coatings for providing the materials used in this study and for the provision of a strain gauge signal conditioning unit. We are grateful to M Buhaenko for advice and for stimulating discussions throughout the project.
REFERENCES 1 2 3 4 5
E M Corcoran, J.Paint Technol., 41 (1969) 635 Yan Gu, MPhil thesis, University of Newcastle upon Tyne (1997) S G Croll, J.Coatings Technol., 50 (638) (1978) 33 S G Croll, J.Appl.Polym.Sci., 23 (1979) 847 S G Croll, J.Coatings Technol., 51 (659) (1979) 49
Estimation of Long-term Properties of Epoxies in Body Fluids
Steven W. Bradley Materials Performance, Inc., College Station, TX, USA
INTRODUCTION The selection of materials for use in the human body often requires unique evaluation techniques. Beyond the normal toxicological and tissue rejection considerations, the saline body fluid environment, elevated temperature (37°C) and complex stress loadings must be considered. In this work, two candidate epoxies were evaluated for the attachment of two polysulfone lumens. Both candidates displayed adequate strength under dry room temperature loading. The preferred candidate material was selected for ease of use as a room temperature cure epoxy. However, the material had a potentially low glass transition temperature, Tg. A course of investigation was as follows: (1) perform an extensive literature review to determine what the potential effects of moisture were on the adhesive strength, cohesive strength and Tg of the candidate epoxies, (2) compare glass transition temperatures of moisture saturated versus dry candidate materials and (3) establish a protocol using master curves for the accelerated performance evaluation of candidate epoxies.
THEORY OF MOISTURE EFFECTS Absorbed moisture may degrade an epoxy adhesive in at least four ways: (1) reduction in interfacial adhesion; (2) reduction of the cohesive strength of the adhesive due to plasticization; (3) reduction in the glass transition temperature; and (4) swelling which produces local residual stresses which can be significant.
REDUCTION IN INTERFACIAL ADHESION Previous work1-3 addressed the effect of moisture on epoxy adhesives used in conjunction with metals which found that soaking and subsequent testing generally produced reductions in strength of less than 50%. Continuous loading or stress in a moist environment generally produced even greater reductions in creep rupture strength at times of one year or longer, with values of 80-90% reported in some cases.1 Previous work4,5 alluded to good interfacial bonding between a DGEBA/DDS epoxy system and polysulfone (PSF), where PSF was a
90
Plastics Failure Analysis and Prevention
second toughener in the system. However, the effect of moisture and a more quantitative measure of the interfacial strength between common epoxies and polysulfone were not found in the literature. This appears to be an important area for experimental work to confirm the behavior of epoxy adhesives for PSF, both dry and wet.
REDUCTION IN COHESIVE STRENGTH Generally the degradation of an epoxy is found to be a reduction in the cohesive strength, though the failure is more often adhesive at higher temperatures. The reason for this behavior can be understood by examining Figure 1,2 which shows load-displacement behavior for an epoxy which is cured at 121°C and has a Tg = 99°C. Note as the test temperature approaches Tg, the materials yield strength drops dramatically. Furthermore, the load supported at Tg would not be possible in anything but a shorttem test. For longer term testing or service loading, the viscoelastic creep of this material would produce a time dependent failure at a very small load. Generally, a lower cure temperature gives a lower Tg and moisture shifts the Tg to a value Figure 1. Illustration of loss in strength for an epoxy 10-40°C lower than the dry Tg. Thus, an epoxy approaches Tg.2 adhesive with a moderate cure temperature, saturated in water, can have a Tg that is very near room temperature, giving essentially zero yield strength at room temperature and about the same creep rupture strength for longer term applications. Sharon et. al.6 provide a summary of both the variation of Tg with moisture content and the effect of service/test temperature relative to Tg in determining mechanical properties. It is important to keep in mind that the mechanical properties given6 are short term tensile properties. Establishing an operating temperature for longer term service relative to Tg would be different (and much lower relative to Tg) than what one would infer from these results.
REDUCTION IN GLASS TRANSITION TEMPERATURE Previous authors6-10 give examples of moisture absorption by epoxy resins and the consequent effect on the glass transition temperature, Tg. Moisture absorption ranges from 2.4% to greater than 5% by weight, with greater moisture absorption at higher temperatures and
Estimation of Long-term Properties of Epoxies
91
for crosslinking agents that are more hydrophilic. The effect of saline body fluids (0.9% salt) appears to be similar to seawater where the salt water has been found to be absorbed at a slightly lower rate, with presumably the salt not being readily absorbed. A lowering of Tg by 10-40°C had been noted due to absorbed moisture.6 Thus, the initial Tg (dry) needs to be sufficiently high so that the wet Tg exceeds the service temperature by >50°C for long term, continuous load applications. If a Tg shift of 40°C due to absorbed moisture is possible and a service temperature of 37°C is envisioned, then a dry Tg of 37+40+50=127°C would be prudent.
SWELLING INDUCED STRESSES Moisture induced swelling can produce significant residual stresses in a part. The actual magnitude depends on both the amount of moisture absorbed, its distribution, and the degree of constraint provided by the system. If, for example, the horizontal bond line were to experience swelling, the constraint of a vertical bond line would result in the development of residual stresses along the vertical portion of the bond line loaded in shear. These stresses might be significant relative to those produced by the service loads. Again, measured swelling strains in combination with analysis can answer such questions.
EXPERIMENTAL & RESULTS DETERMINATION OF GLASS TRANSITION TEMPERATURE Samples of both epoxies were placed in 0.9% saline solution that simulated body fluid. The Table 1. Average Tg for two fluid saturation of the sample was measured by epoxies at dry and saturated con- comparing the initial weight of the epoxy with the weight after soaking in solution. On average, the ditions water uptake was 1.5%. The samples were still gaining a small amount of water weight when the Material Average Tg, oC tests were performed due to time constraint. The Epoxy A (dry) 59.5 samples were analyzed on a Thermal Analysis 9900 DSC. The materials were heated at a rate of Epoxy A (sat.) 48.9 5, 10, 20oC/min to about 180oC. All results were Epoxy B (dry) 89.8 within 2-3oC of each other. Software was required to determine the Tg as the moisture Epoxy B (sat.) 82.3 absorption smoothed the break in the DSC output curve that indicates a glass transition. Several runs were performed for each material to provide for minor differences in curing, saturation or sample size. The results for each epoxy,
92
Plastics Failure Analysis and Prevention
both with and without water, are shown in Table 1. Epoxy A had a saturated Tg 10.6°C lower than the dry samples. Epoxy B had a saturated Tg 7.5°C lower than the dry samples. Additional moisture uptake might shift the Tg as much as 15°C. USING DYNAMIC MECHANICAL ANALYSIS TO CREATE MASTER CURVES Dynamic mechanical testing provides information for the storage modulus, loss modulus and phase lag during testing. For a given temperature/frequency sweep, variations in these parameters provide information about molecular relaxation processes occurring in the material. Several peaks may form during a sweep which provide valuable information about the material. The first peak below the melting point is generally known as the “alpha” peak, or the glass transition temperature, Tg. The next peak that may appear is commonly known as the “beta” peak that quantifies smaller changes in molecular conformation when a certain activation energy is reached. In the tensile mode, the oscillatory strain is described by: ε = ε o sin ( ωt ) [1] where ε o is the strain amplitude and ω is the angular frequency. The corresponding stress is given by: σ = σ o sin ( ωt + δ ) [2] where σ o is the stress amplitude, ω is the angular frequency and δ is the phase lag between the sinusoidal stress and strain curves. The greater the viscosity the greater the phase lag and the higher the elasticity, the smaller the phase lag. For the DMA, the relation between stress and strain is described as: σ = σ o E' sin ( ωt ) + E'' cos ( ωt ) [3] where E' and E'' are the storage and loss modulus, respectively. The time dependent nature of viscoelastic materials has allowed the development of “master curves” that predict the behavior of polymers far beyond practical testing times. Stress-relaxation curves of a polymer made at different temperatures are superimposed by horizontal shifts along a logarithmic time scale to give a master curve. At temperatures below Tg, the free volume is small and aT is best modeled by the Arrhenius equation: Ea 1 1 η- = ----- --- – ----ln a T ≅ ln ---R T T r ηr
[4]
Master curve data was plotted for three different temperatures. All measurements were made at 5 K intervals and thus are at 35°C, 50°C, and 90°C rather than at 37°C for example. The plot at 35°C may be.interpreted as the behavior of the two tested epoxies if kept in a nominally dry condition. The plot at 50°C may be interpreted as the behavior of the epoxies at 37°C with moisture saturation. The 13-degree difference was a result of the plasticization effect of moisture on the epoxy which lowers the Tg as shown in Table 1. The time equiva-
Estimation of Long-term Properties of Epoxies
93
Table 2: DMA shear modulus values Material
15 min modulus, MPa
1 year modulus, MPa
Epoxy A (dry)
700
100
Epoxy A (wet)
300
7
Epoxy B (dry)
900
600
Epoxy B (wet)
700
100
Figure 2. Master curve for Epoxy B at 50ºC.
lent to a one year test is found in Figure 2 by multiplying the time in seconds by four as the time to load is only ¼ of a cycle. The inverse of this number is the frequency on the plot (8E-9 Hz). A comparison of shear modulus for wet versus dry epoxy as taken from DMA plots at 50°C (wet equivalent) and 35°C (dry) are shown in Table 2. Note that the moisture saturated Epoxy A is completely in the rubbery plateau after one year and would not be useful for bonding purposes. Epoxy B saw a significant modulus reduction due to the proximity to Tg after one year. Figure 3. Arrhenius plot for Epoxy B activation energy determination. A simple calculation can be made using the Arrhenius relationship that would provide the required temperature elevation to represent long term behavior of the material. The Arrhenius relationship requires an activation energy which can be found from the slope of a lnaT vs 1/T plot as shown in Figure 3 for Epoxy B. Calculations for Epoxy B using equation 4 indicate that testing dry specimens for 15 minutes at 72°C will give similar behavior to testing specimens for 1 year at 37°C wet, as far as the modulus is concerned. If significant reductions in the modulus are noted, a significant reduction in strength can also be anticipated, in as much as the same molecular motions that give small elastic strains must be observed to allow the deformation processes that ultimately lead to failure. Similar calculations for Epoxy A would require testing at 100°C to indicate the long-term cohesive strength at 37°C, wet. This was far above the glass transi-
94
Plastics Failure Analysis and Prevention
tion temperature of 59°C for Epoxy A indicating that the long term behavior prediction by elevated temperature testing for Epoxy A is not possible. This was also indicated further by the nonlinearity of the Arrehenius plot for Epoxy A.
CONCLUSIONS Moisture and temperature both act to reduce to strength of epoxy in bodily fluids. Two epoxies were evaluated for adhering polysulfone lumens together. Each material was moisture saturated and then place in a DSC fo Tg determination. The drop in Tg for the two materials was between 7 and 10°C. Further moisture saturation might have produced Tg drops as high as 15°C according to literature. DMA temperature/frequency sweeps were performed at temperature shifts equivalent to the moisture effect on Tg. Epoxy A lost more than 97% of its modulus after one year and was in the rubbery region of the modulus plateau. Epoxy B lost 85% of its modulus after the first year. Finally, a calculation using the Arrhenius relationship found that mechanical testing at 72°C for 15 minutes would give some indication of long term cohesive strength at 3°C, wet.
REFERENCES 1 2 3 4 5
6 7 8 9 10
C.O. Arah, et. al., “The Correlation between Adhesive Stress-relaxation and Joint Performance” SAMPE Journal, Vol. 25, No. 4, pp. 11-13 (Jul 1989). R.A. Jurf, J. R. Vinson, “Effect of Moisture on the Static and Viscoelastic Shear Properties of Epoxy Adhesives,” Journal of Materials Science, Vol 20, pp. 2979-2989 (1985). M.R. Bowditch, D. Hiscok, D. Moth, “Relationship Between Hydrolytic Stability of Adhesive Joints and Equilibrium Water Content,” Int. Journal of Adhesion & Adhesives – Adhesion 90, V. 11, no. 3 (1991). B.G. Min, Z.H. Stachurski, J.H. Hodgkin, “Microstructural Effects and the Toughening of Thermoplastic Modified Epoxy Resins,” Journal of Applied Polymer Science, Vol. 50, pp. 1511-1518 (1993). J.L. Hedrick, M.J. Jurek, I. Yilgor, J. McGrath, “Chemical Modification of Matrix Resin Networks with Engineering Thermoplastics. I. Synthesis and Properties of Epoxy Networks Modified with Amine Terminated Poly(Aryl Ether Sulfone) Oligomers,” Polymer, V. 32, no. 11, pp. 2020-2032 (1991). G. Sharon, H. Dodiuk, S. Kenig, “Hygrothermal Properties of Epoxy Film Adhesives,” Journal of Adhesion, pp. 87-104 (1989). K. Takahashi, “Mositure instusion in SiO2/Epoxy Interfaces,” Mat. Res. Soc. Symp. Proc., Vol. 153, pp. 187-192 (1989). J. Gorbatkina, N. Shaidurova, “The Effect of Aging in Water on the Strength of Fiber-Polymer Systems,” J. of Adhesion, Vol. 35, pp. 203-215 (1991). J. Comyn, et. al., “Effect of Water on Adhesives & Adhesive Joints,” Plastic Rubber Process Appl, V3, n3, pp. 201-205 (1983). C. Ombra, et. al, “Long Term Aging Behaviour of Epoxy Matrices in Aggressive Environments,” Interrelations Between Processing Structure and Properties of Polymeric Materials, pp. 615-626 (1984).
Mechanical Performance of Polyamides with Influence of Moisture and Temperature – Accurate Evaluation and Better Understanding
Nanying Jia and Val A. Kagan Honeywell, Morristown, NJ 07962-1021, USA
INTRODUCTION The plastic industry has no doubt witnessed in recent years an increase in interests and demands in using thermoplastics, such as polyamides, to replace certain metals and thermosets in manufacturing automotive air induction and power train systems, lawn/garden and other power tools. Technologies have also advanced to accommodate these demands by developing materials and products with higher performance, less weight, more time/cost savings, optimized welded joints, and better resistance to fatigue and environmental changes.1-3 The important roles today’s thermoplastic structures are designed to play have made it increasingly critical for the materials to perform, especially under adversary working and environmental conditions such as cyclic stress/strain, high and low temperatures, and changing humidity. The short-term and long-term mechanical properties (tensile and fatigue) of polyamide (PA), or nylon, based plastics under dry-as-molded (DAM) conditions were analyzed previously.1-2 Fatigue properties of unfilled nylon 6 and nylon 66 at room temperature conditions with the influence of absorbed moisture were also discussed.4 The absorbed moisture (≤2.5 wt%) decreased fatigue crack growth rates, which might reflect the ability of tightly bounded water to enhance chain mobility. Results on combined moisture-temperature effects on short-term and long-term properties of polyamides, on the other hand, were rarely found in the literature. The current investigation has been focused on the combined effects of moisture-temperature on the short-term (tensile) properties of reinforced nylon 6. Specifics in nylon conditioning using ISO and ASTM procedures are discussed elsewhere.5 The purpose of this investigation is to help designers of plastic parts and assembled product, material developers, and the database users to correctly interpret the data on tensile properties.
96
Plastics Failure Analysis and Prevention
THE EFFECT OF MOISTURE Despite its high performance and easy processing, Nylon’s tendency and ability of absorbing moisture from the surrounding environment has made it a constant challenge for processing and design engineers alike. The moisture is known to affect a range of polymer properties, which in turn impact processability, dimensional stability, mechanical, acoustic, electrical, optical, and chemical properties, and ultimately the performance of products.6-7 The moisture in nylon acts as a plasticizer that reduces the entanglement and bonding between molecules, therefore increases their volume and mobility.7 The moisturized material exhibits lower glass transition temperature (Tg), which makes it easier for further crystallization. The increase in moisture may cause profound changes in a material’s behavior under load; it reduces strength, stiffness, and natural frequency, while increasing energy absorption and ductility in the material. Practically, the best way to minimize the moisture uptake is to select plastics with low absorption rate or design products in ways to prevent excessive absorption. Under dry-as-molded (DAM) conditions, polyamide, or nylon, usually contains 0.10.3% water. At room temperature and 50% relative humidity (RH), type 6 polyamide could eventually absorb 2.75% water. Every 1% moisture increase in nylon may result in 0.2 to 0.3% increase in its dimension.6 This change in dimension would have to be accommodated by preconditioning parts prior to service. SAMPLE CONDITIONING – METHOD AND ANALYSIS Realizing the fact that the properties of dry-as-molded materials often do not reflect their true behavior in service due to the changes in properties caused by the subsequent moisture uptake, sample conditioning is therefore applied to adjust the moisture content in the materials to a desired level so that their properties and performance can be properly tested and analyzed. In this regard, ISO-291 defines the following two standard atmospheres for conditioning: • “Atmosphere 23”: 23/50 (temperature in °C/relative humidity in %) as recommended for most applications; • “Atmosphere 27”: 27/65, as recommended for tropical regions. For practical purposes it is very important to analyze the properties of thermoplastics conditioned under “Atmosphere 23” (23°C/50%RH), which is recommended for most industrial applications such as automotive, lawn & garden, power tools, appliances, and so on. The rate of moisture absorption, however, is very low under “Atmosphere 23”. In this environment, it would take more than a year for the moisture in an ISO-3167 multipurpose test specimen (4 mm thick) of PA 66 to reach equilibrium. To accelerate this process, one
Mechanical Performance of Polyamides
97
must increase the conditioning temperature, and/or the relative humidity. Several such conditioning procedures that may be applied to thermoplastics are shown in Table 1. Table 1. Methods of Moisture Conditioning for Thermoplastics Standard
Procedure
Use/Comment
ISO-291; ASTM D618
Six conditions varying by atmosphere, temperature, water, duration, relative humidity
Standard procedure prior to testing; not a good choice for DAM nylons due to their sensitivity to moisture
ISO-62; ASTM D 570
Water absorption at 23±1°C, 50±1°C, 105-110°C between 0.5 and 24 h.
See Table 2 for data. The moisture at equilibrium varies between 1 to 14% depending on temperature & RH; the microstructure of polymers may be affected by high temperature and high RH (e.g. boiling water).
ISO 483; ASTM E 104
Conditioning done in saturated salt solutions with different RH and temperature.
See Table 3 for results.
ISO-1110
Conditioning performed at 70°C and 62% RH.
Materials tend to “over-saturate” under using this procedure, comparing to the “Atmosphere 23”
Table 2. Water absorption values for selected thermoplastics after 24 h. (ISO-62, ASTM D570)6 Material
Table 3. Influence of relative humidity on water absorption in non-filled nylons (at 23°C in air)7 Relative humidity, %
Water absorption Type of PA
PP
< 0.01%
PC
0.15%
Nylon 11
30
50
62
100
PA 46
1.4
3.8
5.0
15
0.25%
PA 6
1.1
2.75
3.85
9.5
Nylon 6
1.3%
PA 66
1.0
2.5
3.6
8.5
Cellulose acetate
1.7%
98
Plastics Failure Analysis and Prevention
Table 2 lists the water absorption values for several plastics as determined by ASTM D-570 after 24 h immersion at 23°C. Equilibrium values for water absorption will be significantly higher for these materials. The water absorption will also be higher at elevated temperatures. For unfilled nylon 6, the moisture at equilibrium at 23°C and 50% RH is 2.75% (see Table 3). Once conditioned, it becomes important for the moisture level to be determined accurately. Moisture analysis for pellets and parts is critical for manufacturing and other processes such as molding, testing, and end-use. Table 3 shows the equilibrium moisture levels at different RH for several commercial nylons, and Table 4 lists several methods of determining moisture for nylon based thermoplastics. Table 4. Standard methods of moisture analysis for nylon (PA) thermoplastics Standard
Use/Comment
ASTM D789 (Karl-Fischer)
Analysis is based on titration with a Karl Fischer reagent. It is sensitive to moisture from 0.1% to 0.2% with typical sample weight 20 ~ 30 g. Smaller sample size is preferred for higher moisture content
ASTM D 4029
Analysis is based on release of water vapor, which is carried away by an inert gas into an electrolytic cell. It can determine moisture in nylon at a level < 0.1% from a sample 2 ~ 4 g in weight. In a “dry” state the moisture content in nylon based thermoplastics is between 0.05% (nylon 46) and 0.3% (nylon 612) (ASTM D 4066). For nylon 6 the number is ~ 0.2%.
ISO-1110, ASTM D 570
Analysis is based on weight gain. A precision of 0.001 g is required for the weight determining device.
MATERIALS AND EXPERIMENTAL PROCEDURES MATERIALS The material used in this investigation was heat-stabilized nylon 6 with 33 wt% glass fiber reinforcement. The material was injection molded into several configurations: • 4 mm thick ISO multi-purpose tensile bars (ISO-3167); • 3.2 mm thick ASTM Type I tensile bars (ASTM D638) and ASTM flex bars (ASTM D 790); • 4 and 6.4 mm thick plaques. The molded specimens were tightly sealed prior to conditioning and testing in order to preserve their dry-as-molded state while the moisture content remains at ~ 0.2%.
Mechanical Performance of Polyamides
99
PROCEDURE FOR MOISTURE CONDITIONING The procedure for sample conditioning was based on ISO-1110. The molded specimens were loaded into an environmental chamber where the temperature and relative humidity were maintained at 70°C and 62%, respectively. The moisture uptake in the sample was periodically calculated by recording the weight gains using a Mettler balance. The water absorption was also determined on a small group of samples using a Karl-Fischer unit. The moisture content in samples with different thickness was plotted against the conditioning time, as shown in Figure 1. Although the method in ISO-1110 can greatly accelerate the moisture absorption in nylon compared to the standard method such as Figure 1. Moisture absorption vs. time for 33% glass fiber reinforced nylon 6. The number in mm by each “Atmosphere 23”, prolonged exposure of samples curve indicates the thickness of the samples.5 under conditions specified in ISO-1110 (70°C and 62%RH) has been found to cause “over conditioning” in materials by injecting more moisture than what one can ever obtain when conditioned under “Atmosphere 23”. For this reason, the conditioning was terminated once the moisture in the material was found to have reached the equilibrium level under “Atmosphere 23”. Samples were then sealed tightly in moisture proof bags until tested or analyzed. PROCEDURE FOR TENSILE PROPERTY TEST The tensile properties of reinforced nylon 6 used in this study was obtained by following ISO-527. Tests were conducted on ISO multipurpose specimens using an Instron 4505 universal testing system. Tests at high and low temperatures were conducted in an environmental chamber attached to the Instron frame. During the test, the temperature at the center of the chamber was maintained at ±2°C within the set point. The detailed description of test setup can be found elsewhere.1-2
RESULTS AND DISCUSSIONS Figures 2a ~ 2c compare the tensile behavior (stress-strain curves) of DAM and moisturized nylon 6 at –40°C, 23°C, and 120°C, temperatures among those typically found in the enduse conditions. Changes in tensile strength, Young’s modulus, and strain at yield versus moisture are shown in Figures 3 and 4.
100
Plastics Failure Analysis and Prevention
Figure 2a. Effect of moisture on tensile behavior of reinforced nylon 6 (33% glass fiber) at low temperature.
Figure 2b. Effect of moisture on tensile behavior of reinforced nylon 6 (33% glass fiber) at room temperature.
Figure 2c. Effect of moisture on tensile behavior of reinforced nylon 6 (33% glass fiber) at elevated temperature.
Figure 3. Effect of moisture on tensile properties of reinforced nylon 6 (33% glass fiber); tensile strength and strain at yield vs. moisture at room temperature.
The effect of temperature on tensile behavior can be found in Figures 5a and 5b for DAM and conditioned materials, respectively. The decrease in strength and modulus due to the rising temperature is shown in Figure 6.
Mechanical Performance of Polyamides
101
Figure 4. Effect of moisture on Young’s modulus: 33% glass fiber reinforced nylon 6.
Figure 5a. Effect of temperature on tensile behavior of dry-as-molded nylon 6 (33% glass fiber).
Figure 5b. Effect of temperature on tensile behavior of conditioned nylon 6 (33% glass fiber).
Figure 6. Effect of temperature on tensile strength and Young’s modulus of dry-as-molded nylon 6 (33% glass fiber).
Although it has been well known that the moisture in thermoplastics will in general serve as a plasticizer that reduces a material’s strength and increases its ductility, the current study indicates clearly that the net impact of moisture on tensile properties of nylon depends also on temperature. At –40°C, change in tensile behavior was insignificant within the elastic limit of the material (Figure 2a). About 10% decrease was found in the ultimate tensile
102
Figure 7. Impact of temperature and moisture on tensile strength of reinforced nylon 6 (33% glass fiber).
Plastics Failure Analysis and Prevention
Figure 8. Impact of temperature and moisture on tensile strain @ tensile strength of reinforced nylon 6 (33% glass fiber).
strength after conditioning, but the total elongation (or strain) at failure remained virtually unchanged. The most significant change in tensile behavior due to moisture can be found at room temperature where the tensile strength was reduced by about 40% after conditioning, while the elongation at failure increased by 150% (Figure 2b). At elevated temperatures (80°C and above) where considerable plastic deformation in nylon has been resulted from heat, it was found that the conditioned samples exhibited not only a further reduction in strength, but lower overall elongaFigure 9. Impact of temperature and moisture on Young’s tion as well (Figure 2c). modulus of reinforced nylon 6 (33% glass fiber). Figures 2a to 5b seem to suggest that an increase in temperature or moisture can each achieve similar results in nylon, which is to reduce the material’s strength and increase its ductility. However, the combination of temperature and moisture did not result in more ductility in the material that would allow it to deform further before failure. Instead, the material failed sooner with lower strength, indi-
Mechanical Performance of Polyamides
103
cating that the high temperature and moisture together are able to cause deterioration in the reinforced nylon, and possibly other thermoplastic materials as well. The tensile properties of nylon 6 at various temperature and moisture combinations are shown in Figures 7, 8, and 9. The values of these properties were obtained by measuring the moisture content in individual tensile bars immediately after each test using a Karl- Fischer analyzer described in Table 4. Each point on Figures 7 ~ 9 represents a single test. At each temperature (other than 23°C), the first test was conduced 1 hour after specimens were placed in the preheated chamber, and the subsequent tests were conduced approximately one every 20 minutes. One can see that, at each temperature, the tensile properties, especially the tensile strength and Young’s modulus, correlate fairly well with the moisture measured individually from the test bars. The results in Figures 7 ~ 9 also indicate that, in a hot and relatively dry environment, nylon can quickly lose its absorbed moisture, making the initial moisture level reported at the end of conditioning meaningless as a reference parameter to reflect the material’s moisture state. The rise or fall in the material’s properties (e.g., tensile parameters) following the change in temperature and relative humidity is something any design engineer must consider if he or she wishes to design and model structures or products made of thermoplastics.
SUMMARY AND CONCLUSIONS The individual and combined effects of temperature and moisture on the tensile properties of glass fiber reinforced nylon 6 were investigated and characterized. The sample conditioning was performed in an environmental chamber at 70°C and 62% RH as specified in ISO1110. The conditioning was conducted until the moisture in the samples reached the equilibrium level under the standard conditions, i.e. 23°C/50%RH. The tensile property tests were conducted between –40°C and 150°C on ISO multipurpose test specimens with moisture levels from 0.2% (DAM) to 2.6% (conditioned). The following conclusions can be made as a result of the current investigation: 1. For nylon thermoplastics used in this investigation, both temperature and moisture can cause a decrease in strength and an increase in ductility. Among the selected temperatures, the moisture was found to cause the greatest change in tensile properties at 23°C. After conditioning, the material has lost 40% of its original tensile strength, and at the same time, the total elongation has increased by 150%. 2. At –40°C, nylon lost 10% of its tensile strength, but other characteristics such as tensile strain and Young’s modulus remained largely unchanged, especially within the elastic limit. 3. At 80°C and above, the tensile strength and Young’s modulus decreased further, while the elongation or strain to failure increased.
104
Plastics Failure Analysis and Prevention
4. At elevated temperatures, the material properties have further deteriorated by adding moisture into the structure. Comparing to the dry-as-molded materials, the conditioned materials have lower tensile strength, and they also fail sooner. 5. Due to the rapid loss in moisture in a high temperature and low humidity environment, the initial moisture value obtained from a given nylon thermoplastic can quickly become meaningless once the material is exposed to such an environment. The rise or fall in the materials properties (e.g., tensile parameters) following the change in temperature and relative humidity is something one must consider in design, modeling, and manufacture of parts and products using moisture sensitive thermoplastics.
ACKNOWLEDGMENT The authors wish to express their sincere appreciation to Mr. Howard Fraenkel for the moisture conditioning and weight change measurement, and to Dr. Rich Williams for the moisture measurement using a Karl-Fischer unit.
REFERENCES 1 2 3 4 5
6 7 8 9
N. Jia and V. A. Kagan, “Compatibility Analysis of Tensile Properties of Polyamides Using ASTM and ISO Testing Procedures”, SPE Annual Conference Proceedings / Antec’98, Vol.2, Materials, pp. 1706-1712, (1998). N. Jia and V. A. Kagan, “Effects of Time and Temperature conditions on the Tensile-Tensile Behavior of Short Fiber Reinforced Polyamides”, SPE Annual Conference Proceedings / Antec’97, Vol. 2, pp. 1844-1848, (1997). V. Kagan, “Good Vibrations Join Thermoplastics Fast”, Machine Design, August 5, 1999. P. Bretz, R. Hertzberg, and J. Manson, “Influence of Absorbed Moisture on Fatigue Crack Propagation Behavior in Polyamides”, Journal of Materials Science, vol. 16, Part 1: Macroscopic Response, pp. 2061-2069, (1981). N. Jia and V. A. Kagan, “Tensile Properties of Semi-Crystalline Thermoplastics – Performance Comparison under Alternative Testing Standards”, SAE Technical Paper Series, SAE International Annual Congress and Exposition, submitted to SAE’2000, Detroit (2000). Engineering Plastics, Engineering Materials Handbook, Vol.2, ASM International, 883 pages, (1988). M. Kohan, Nylon Plastics Handbook, Hanser / Gardner Publications, Inc., New York, 631 pages, (1995). N. Rao and K. O’Brien, Design Data for Plastics Engineers, Hanser / Gardner Publications, Inc., Cincinnati, 208 pages, (1998). C. MacDermott and A. Shenoy., Selecting Thermoplastics for Engineering Applications, Marcel Dekker, Inc., New York, 305 pages, (1997).
Temperature-Moisture-Mechanical Response of Vinyl Ester Resin and Pultruded Vinyl Ester/E-Glass Laminated Composites
S. P. Phifer, K. N. E. Verghese, J. J. Lesko Materials Response Group, Department of Engineering Science and Mechanics, Virginia Tech J. Haramis Department of Civil Engineering, Virginia Tech, Blacksburg, VA 24061-0219, USA
INTRODUCTION High strength composite laminates are finding applications in infrastructure and construction industries. The high strength to weight ratio, corrosion resistance and durability are appealing, but the higher initial fabrication cost, in the past, has prevented composites from competing with products such as steel beams and wooden studs. With the advent of continuous pultrusion of constant cross-section laminated composites using low cost resin and fiber systems such as vinyl ester and E-glass, composites are successfully competing with these traditional materials. However, uncertainties in strength and durability of pultruded composites have hampered their wide-range use. Pultruded composites are similar to aerospace grade composites but differ in that the resins are more brittle, have higher shrinkage and less toughness. The fiber architecture is different due to the pultrusion requirement that all offaxis plies to the pultrusion direction must be attached to the on-pultrusion-axis plies. Only fiber aligned with the pultrusion axis can be pulled through the pultrusion die. Tricot stitching is generally used to attach off-axis plies to the on-axis plies which creates fiber bundles and gaps between the bundles. When the stitched plies are stacked to form a laminate fiber undulation and resin rich pockets are formed. With the high shrinkage resin, microcracking and delamination are plentiful in the as-processed condition. With fiber undulation, resin rich pockets, microcracking and delamination, pultruded laminates will potentially have lower strength, stiffness, moisture, temperature, and fatigue resistance than comparable aerospace grade composites. E-glass fiber is known to degrade in moist environments due to an ionic exchange of the H+ from the water with the metals such as sodium (Na+),1-4 which are loosely bound as
106
Plastics Failure Analysis and Prevention
oxides in the silica. The resin acts to reduce the rate of degradation but matrix cracking, delamination, and exposed edges are sights for accelerating degradation. The purpose of this research will be to experimentally obtain cross-ply, angle-ply, and resin strength and stiffness as a function of test temperature and time at elevated temperature immersion in tap water. Tensile and shear properties can then be determined to be used in specific application structural durability analysis.
MATERIALS UNREINFORCED RESIN MATERIAL The resin used to make samples for mechanical testing of the unreinforced polymer was DerakaneTM 441-400 supplied by the Dow Chemical Company. The resin consists of a 690 g/mole, number average molecular weight oligomer and is diluted with 28% styrene in order to reduce its room temperature viscosity. The Derakane resin was cured with benzoyl peroxide (BPO), obtained from Aldrich. The material was 97% pure (lot # ES 03918CS). A BPO concentration of 1.1 wt% in Derakane was utilized (or a 0.011:1 ratio of BPO to Derakane). The resin was added to a suitable glass container. An Arrow 850 high torque stirrer was placed into the blend and stirring was set at the medium speed. The BPO was slowly added to the stirring mixture. After the BPO had dissolved (approximately 1 hour), the blend was degassed utilizing approximately 24 in Hg vacuum. The blend was allowed to degas for approximately 30 minutes. The catalyzed Derakane resin was cured utilizing the following procedure. An 20.34 cm x 15.24 cm x 0.635 cm vertical mold was utilized for the curing process to prevent the formation of air voids. The mold was treated with mold release and then assembled. The blend was added very slowly to the mold. After the top had been secured to the mold, it was then placed in a Fisher Isotemp forced convection oven. The material was cured utilizing the following cure cycle: 1 hour hold at 65°C heating at 10°C/min to 150°C 20 minute hold at 150°C. The resin undergoes a rapid, free radically initiated addition copolymerization reaction to form the crosslinked network. After the mold had cooled to room temperature, the cured blend was removed. The samples were then cut on a water cooled circular table saw and then milled and ground to the required dimensions of 17.8 cm x 2 cm x 0.32 cm. In order to get the dog-bone geometry, the resin specimens were then placed in a fixture and hand fed through a four-fluted carbide end-mill attached to a router. The width of the gage section of the dog-bone is 1.27 cm in accordance with the ASTM D638-95 test standard.
Æ
Æ
GLASS FIBER/VINYL ESTER COMPOSITE The glass fiber reinforced composites were made using the DerakaneTM 411-350 vinyl ester resin that was supplied by the Dow Chemical Company. The E glass fiber was obtained
Temperature-Moisture-Mechanical Response
107
from Owens Corning and the lay up is [One, Nexus NS veil cloth/One, CDM “1810” C, 0/ 90°/continuous strand mat complex/Two, CD 185, 0/90° stitched mats] s. The composites were made using pultrusion in the Dow facility at Freeport, Texas. The width of the die used was 18 cm. The pultrudable grade of DerakaneTM 411-350 consists of fillers such as calcium carbonate, UV stabilizer, air release agent and mold release agent. A proprietary initiator package was used to cure the vinyl ester using a free radical mechanism similar to the one explained above. During pultrusion both the pull force and die temperature were monitored at all times. Die temperatures were measured at the entrance, midpoint, and exit using thermocouples. For the vinyl ester laminates, die temperatures were set at 240°F, 280°F, and 280°F, respectively. A pull speed of around 45 cm/minute was used and pieces 54 cm long were cut periodically using a tabletop circular saw. The final test specimen dimensions of 17.8 cm x 2.54 cm x 0.4 cm were achieved by using a water-cooled circular saw. The edges of the specimens were not sealed but were polished using 100, then 400, followed with 600 grit wet sandpaper.
EXPERIMENTAL PROCEDURE Tensile tests were performed in accordance with ASTM D3039-95a, and D3518-95 for the cross-ply and angle-ply (±45°) composites respectively and ASTM D638-95 for the resin in dog-bone form. These tests were performed using an Instron test frame operated in displacement control at a rate of 1.25 mm/min. Both extensometers and strain gages were used to measure strain. In the case of the resin samples, the aluminum extensometer tabs and strain gages were found to induce failure at those sites when tested at room temperature and below. Use of silicone extensometer tabs alleviated this problem.
RESULTS AND DISCUSSION UN-AGED (AS RECEIVED) MATERIAL Mechanical Properties Unaged properties of the resin, cross-ply and angle-ply laminates are shown in Table 1. Error bars in the figures represent ± 1 standard deviation. Typical stress-strain plots, normalized by ultimate strength, are also indicated in Figure 1 to present the plasticity of the resin and angle-ply specimens at room temperature. Test Temperature Effects In order to evaluate the effects of temperature on mechanical properties of these materials, tensile tests were performed at temperatures ranging from –50°C to 90°C. A comparison of the tensile strength and stiffness was made for the resin, cross-ply and angle-ply laminates. The room temperature tensile strength data, normalized to room temperature ultimate
108
Plastics Failure Analysis and Prevention
Figure 1. Tensile stress/strain curves normalized to peak tensile strength.
Figure 2. Tensile strength normalized to room temperature unaged tensile strength.
strength, showed that the resin and angle-ply follow the same curve from 0°C to 95°C, Figure 2. However, a peak near -25°C is seen to exist for the resin. This is speculated to be due to the increasingly brittle nature of the resin that makes it extremely notch sensitive at subambient temperatures. On the other hand, the angle-ply laminates, withstand substantial microcracking, permitted higher strength at sub-ambient conditions. At 95°C the strengths of both angle-ply and cross-ply laminates were seen to monotonically decrease to 60% and 30% respectively of their room temperature, unaged strengths. The tensile modulus of the resin and shear modulus of the laminate reduced monotonically with increased temperature. The tensile modulus of the laminate on the other hand was seen to be constant from -50°C to 25°C after which it decreased monotonically, Figure 3. Table1. Unaged 25°C mechanical tensile properties with standard deviation Tensile modulus, GPa (Msi)
Ultimate strength, MPa (Ksi)
Strain at ultimate strength (mm/mm)
Resin
3.32±0.05(0.48±0.01)
82.6±1.0(12.0±0.1)
4.85±0.59
Angle-ply
3.71±0.23(0.54±0.03)
55.8±1.3(8.1±0.2)
4.41±0.26
Cross-ply
23.9±3.4(3.47±0.50)
431±12(62.5±1.8)
2.58±0.35
Specimen
ENVIRONMENTALLY AGED MATERIAL Moisture Absorption Samples were immersed in water at 45°C and 80°C. Although from previous studies performed by the authors the resin was observed to follow a Fickian uptake behavior when
Temperature-Moisture-Mechanical Response
Figure 3. Tensile modulus normalized to room temperature unaged modulus
Figure 5. Cross-ply tensile strength normalized to unaged 25°C strength verses time of immersion.
109
Figure 4. Pultruded vinyl ester E-glass laminate % moisture content (MC) verses time of immersion.
Figure 6. Angle-ply shear strength normalized to unaged 25°C shear strength verses time of immersion.
immersed in water at 80°C with an equilibrium moisture content of 1.1%, large deviations were noticed for the composite. Therefore the experiment was arbitrarily terminated after 80 days at which point the moisture contents were 0.5% and 1.2% by weight respectively. Figure 4 shows the uptake plots at both 45°C and 80°C for the cross ply laminate. Others5-7 have shown that the equilibrium moisture content in similar vinyl ester/E-glass composites should be reached within approximately 10-12 days in 93 and 80°C temperature water immersion. In the present case from a simple rule of mixtures approach, it is clear that this is higher than the moisture content for the resin by itself when normalized to percent resin content of 37% in the composite. The E-glass fiber, interface, matrix cracks, and delaminations are speculated to be influencing the moisture absorption rate and equilibrium value. The degradation is therefore both a function of temperature and time.
110
Plastics Failure Analysis and Prevention
Mechanical Properties (Immersion Aging Time Effect): In order to assess the effects of time of exposure on the mechanical property retention, cross-ply and angle-ply laminate specimens were aged for 4, 16, and 80 days in 45°C and 80°C temperature water. Tensile tests were then performed at room temperature. This nonfickian moisture absorption behavior mirrors the trend of cross-ply tensile strength reduction with immersion time, as shown in Figure 5. For the cross-ply laminate, immersion for 4 days in 45°C and 80°C water, the strength was reduced to 85% and 65% respectively when compared to the room temperature, as-fabricated strength. Prolonged immersion for 80 days further reduces the strength to 65% and 35%. The modulus of the cross-ply laminate reduces to 95% and 85% with immersion of 80 days in 45°C and 85°C water respectively. Experimental observation of the failure mechanism for the cross-ply laminate revealed changes from the unaged to aged material. A sudden event marked the failure of the unaged and 80 day aged cross-ply specimens. The failure of the 4 and 16 day aged cross-ply specimens began with a few loud pops followed by a continuous popping over 5 to 15 seconds. The stress strain curves for the 4 and 16 day aged specimens therefore became nonlinear near end of life due to this slow failure sequence. The shear response measured from the angle-ply laminate is less affected by immersion in 45°C and 85°C water (Figure 6). Immersion of 4 days reduces the shear strength to 95% and 93% respectively, and immersion of 80 days reduces the strength to 95% and 70% respectively. Shear modulus, not shown in the figures, is little effected by immersion aging. Mechanical Properties (Test Temperature Effects) Cross-ply and angle-ply laminate specimens, immersed in 45°C and 80°C temperature water for 80 days, were tensile tested at –50°C, 25°C and 50°C. The aged, cross-ply specimens, Figure 7, had nearly parallel strength/temperature curves to the unaged but were reduced by 35% and 65% for 45°C and 80°C/80 days aging respectively. The shear strength of the angle-ply, Figure 8, is not reduced at 45°C/80 day immersion but at 80°C/80 days immersion the shear Figure 7. Comparison cross-ply tensile strength relative strength/temperature curve is parallel to the to test temperature for specimens unaged and immersed unaged composite but is offset by 25-30%. The for 80 days in 45°C and 80°C water. tensile modulus of the cross-ply laminate aged at 45°C for 80 days showed little difference when compared to the unaged material (Figure 9). When this was compared to the changes in shear modulus of the aged angle-ply laminates, despite the overall test temperature related trends, little or no change was noticed as a func-
Temperature-Moisture-Mechanical Response
Figure 8. Comparison angle-ply shear strength relative to test temperature for specimens unaged and immersed for 80 days in 45°C and 80°C water.
111
Figure 9. Comparison cross-ply tensile modulus relative to test temperature for specimens unaged and immersed for 80 days in 45°C and 80°C water.
tion of aging conditions. All of the above mentioned trends indicate that the properties of the matrix polymer is not affected by hygrothermal aging and is only affected by test temperature. The E-glass reinforcement however is extremely sensitive to both temperature and moisture.
CONCLUSIONS The tensile and shear strength of the cross-ply and angle-ply monotonically decreased with increasing temperature over the range of –50 to 95°C. In contrast, the resin tensile strength peaked near -25°C. The Derakene 411-400 resin had a Fickian moisture uptake when immersed in 80°C water. The vinyl ester/E-glass composites exceeded the resin equilibrium moisture content when normalized with resin content of the composite and did not reach equilibrium within the 80-day immersion period in 45, 65, or 80°C temperature water. The fiber dominated tensile strength of the cross-ply laminate after immersion, decreased rapidly initially. In 14 days the cross-ply tensile strength was reduced to 78% and 48% for 45°C and 80°C immersion temperatures respectively; while only an additional 12% and 15% reduction was noted from 14 to 80 days of immersion respectively. From the data obtained at different test temperatures on hygrothermally aged specimens it was noticed that the resin was affected primarily by temperature and not aging condition while the glass fiber was sensitive to both water as well as temperature. Visible discoloration was noticed on the samples that were immersed in water. The rate of discoloration was seen to be a function of both aging temperature as well as aging time. Failure surfaces of samples tested after saturation indicated that this discoloration was limited to the outer layers of the composite even after 80 days in 80°C water. It was speculated that damage (debonding) to the surface veil cloth as well as outer
112
Plastics Failure Analysis and Prevention
ply fibers was responsible for this discoloration. Preliminary scanning electron microscopy (SEM) work indicated the formation of cracks around the several fibers on samples exposed to 80°C water. The progression of these cracks both in terms of magnitude and position relative to the surface plies seems to be dependent on the length of exposure.
FUTURE WORK Tensile testing of the unreinforced resin after saturation in water at 45°C and 65°C will be necessary, in order to better understand the hygrothermal response of the polymer. To complete the study on the effects of aging time on property retention, 4, 16 and 80-day aging tests at 65°C for the cross-ply laminate will be performed. Similar tests will also be carried out at additional temperatures of 20°C and 35°C to ascertain the thermal dependence of the rate of strength reduction. The goal will be to devise a.scheme to calculate a characteristic time for degradation and the activation energy for the process, similar to ideas presented by Phani and Bose.8
ACKNOWLEDGEMENTS The authors wish to thank Dr. P. M. Puckett, Herbert Englen of The Dow Chemical Company, Freeport, Texas and Timothy MacNeil of Owens Corning for supplying the materials and pultrusion facilities. The authors also express their gratitude to Dr. D. A. Dillard at Virginia Tech for time on the Instron.
REFERENCES 1 2 3 4 5 6 7 8
Charles R. J., Journal of Applied Physics, 29:11 1549-1560 (1958). Metcalfe A. G. and G. K.Schmitz, Glass Technology 13:1 5-16 (1972). Bank L. C., T. R. Gentry, and A. Barkatt, Journal of Reinforced Plastics and Composites, 14 559-587 (1995). Schultheisz, C. R. et al. ASTM STP 1242, 257-286 (1997). Springer G. S., Environmental Effects on Composite Materials, Volume 2 59-78 (1984). Springer G. S., Journal of Composite Materials, 14 213-232 (1980). Sridharan S, A. H. Zureick, and J. D. Muzzy, ANTEC’98, 2255-2259 (1998). Phani K. K, N. R. Bose, Composites Science and Technology, 29, 79-87, (1987).
Freeze-thaw Durability of Composites for Civil Infrastructure
J. Haramis Department of Civil Engineering, Virginia Tech K.N.E. Verghese, J. J. Lesko Materials Response Group, Department of Engineering Science and Mechanics, Virginia Tech, Blacksburg, VA 24061, USA
INTRODUCTION Fiber reinforced polymer (FRP) composite materials are under consideration for use in civil infrastructure within the U.S. primarily due to their high strength and stiffness to weight ratios and their design flexibility for specific structural characteristics. In addition, the serviceability and functional service life of a composite structure such as a bridge may be greater than those built using conventional structural materials. One uncertainty that hampers attempts to routinely implement FRP in highway structures is proof of environmental durability. As of today a comprehensive database of information does not exist nor does a fundamental understanding of the physics of the problem at hand. Freeze-thaw durability is one such environmental condition that is not well understood. Limited research has been conducted in this area but it is typically targeted to aerospace applications.1 Work targeted specifically to civil infrastructure applications by Gomez and Casto2 has reported mechanical data on freeze-thaw tests conducted on isophthalic polyester and vinylester pultruded glass reinforced composites. Specimens were aged in accordance with ASTM C666 (namely, 40°F to 0°F followed by a hold at 0°F and a ramp up to 40°F followed by a hold) while submerged in 2% sodium chloride and water. Specimens were removed after every 50 cycles and tested in flexure mode. The results clearly indicated a reduction in flexure strength and modulus after 300 cycles. Lord and Dutta3 were, according to the authors, one of the first to highlight the importance of cracks in the matrix and fiber-matrix interface as being the cause of the damage in composite materials. When these cracks form beyond a certain critical size and density they coalesce to form macroscopic matrix cracks which tends to increase the diffusion of water into the system. Water can then condense within these cracks resulting in crack propagation
114
Plastics Failure Analysis and Prevention
as well the formation of micro and macro level ply delamination during the expansion of water undergoing a liquid-solid phase transition. In Kevlar fabric laminates used by Allred4 subjected to two hour temperature cycles from –20°F to 125°F, ultimate tensile strength of the laminate was found to decrease by 23% after 360 cycles and by 63% after 1170 cycles. In work by Verghese, et al.5 differential scanning calorimetry (DSC) was used to identify the nature and presence of freezable water for each constituent material within an Eglass/vinylester composite, i.e., matrix, and interphase (via an assembled composite). Thawing heat flow measurements taken for a single cycle (-150°C to +50°C, 5°C/min) on saturated, neat, unreinforced vinylester resin samples indicated no thawing endotherm and thus the absence of freezable water. This was attributed to the fact that water would reside in the free volume of the resin. Since this free volume size is on the order of about 6-20 Å, Thompson’s equation indicates that these voids will be thermodynamically too small for water to freeze. Heat flow measurements taken for an E-glass/ vinylester composite with the same cycle parameters clearly indicated a melt endotherm at –6.8ºC thus indicating the presence of small voids at the interphase region within the composite and potential susceptibility to freeze-thaw degradation. Cyclic DSC cycling (–18°C to +4°C, 5°C/min) of an E-glass/ vinylester composite displayed a shift up in the thaw endotherm as cycling progressed, indicative of freeze-thaw damage via increased void size. It seems unlikely that water can freeze in limited void system of a neat polymer resin, but the crack dimensions in a composite system appear are large enough to facilitate the freezability of water (see Figure 1). Work is underway to understand the effects of moisture and freeze-thaw on residual mechanical properties of [0°/90°]3s, glass/ vinylester and glass/epoxy composites. This work will help further our understanding of the effects of different environmental conditions, in particular, freeze-thaw. Figure 1. Typical crack in toughened vinylester composite (0.24mm by 0.08mm).
EXPERIMENTAL PROCEDURE OVERVIEW
Both saturated and dry fiber reinforced polymeric composite samples will be placed in an accelerated freeze-thaw environment and tested for mechanical property degradation, changes in crack density, and moisture uptake at specified intervals to assess damage. A group of saturated controls will be held at a constant temperature above freezing and will be
Freeze-thaw Durability
115
tested in the same manner as the freeze-thaw samples. All samples are pultruded glass reinforced cross-ply (0/90) laminates with polymeric matrix materials. MATERIALS All materials were pultruded using an open resin impregnation bath. During pultrusion both pull force and die temperature were monitored at all times and allowed to reach steady state before any material was considered usable. Three different matrix resins were used in this study: a toughened vinyl ester, an untoughened vinyl ester, and an epoxy. Two different fiber lay-ups were used designated by the letters “L” and “P”. The vinylester laminates were pultruded with the “L” lay-up while the epoxy laminate employed the “P” lay-up in order to prevent scaling problems in the epoxy resin system. Tables 1 and 2 detail the two lay-ups.
Table 1. Fiber Lay-up “L”
Table 2. Fiber Lay-up “P”
Reinforcement #1
2 x XCDM 1810 “C” Complex
Reinforcement #1
2 x XCDM 1810 “C” Complex
Reinforcement #2
4 x CD 185 Complex
Reinforcement #2
4 x CD 185 Complex
Surface veil
2 x Nexus 039
Reinforcement #3
1 x 300 g/m2 (1.0 ox) M8643
Surface veil
2 x Nexus 039
From the batch of pultruded material, 510 samples were cut to 25.4 mm by 177.8 mm (1 inch by 7 inches) from the larger as-received panels using an abrasive wet saw. Special care was taken to align all saw cuts with the principal material directions with the long direction corresponding to pull direction. All laminates were nominally 4 mm (0.160 inches) thick. After the cutting operation, the edges of each sample were wet sanded smooth with 400-grit abrasive paper and blown dry with compressed air. All samples destined for saturation were edge coated with an oven cured two-part epoxy to prevent moisture infusion through cut edges. AS-RECEIVED TESTING AND ANALYSIS Ultimate tensile strength, stiffness, and strain-to-failure were determined quasi-statically for each class of as-received material in accordance with ASTM D 3039 “Standard Test Method
116
Plastics Failure Analysis and Prevention
for Tensile Properties of Polymer Matrix Composite Materials” using a deflection rate of 2.5 mm/min (0.10 inches/min). A total of thirty samples were tested, ten of each material type. In addition, two samples of each material were set aside for crack density analysis using x-ray and optical microscopy techniques. SATURATION AND MOISTURE UPTAKE ANALYSIS 324 samples were fully saturated in a 65°C (149°F) water bath with moisture uptake measured throughout the saturation process. Weight measurements were taken hourly on the first day the samples were placed in the saturation tank, every three hours on the second day, every four hours the third day, every six hours the fourth day, and once everyday thereafter. The samples reached saturation within 45 days. FREEZE-THAW CONDITIONING
Figure 2. Freeze-thaw conditioning chamber.
Figure 3. Unloaded sample freeze-thaw tray.
The freeze-thaw conditioning parameters chosen for this research study were based on ASTM C 666 “Standard Test Method for Resistance of Concrete to Rapid Freezing and Thawing”. This test protocol calls for a ramp down from 4.4°C (40°F) to -17.8°C (0°F) followed by a hold at 17.8°C (0°F), a ramp up to 4.4°C (40°F) and a hold at -17.8°C (0°F). There may be a minimum of 4.8 and a maximum of 12 conditioning cycles per day with 75% of the cycle time set aside for freezing and 25% for thawing. Two high performance cascading refrigeration freeze-thaw conditioning chambers (see Figure 2) will be used to achieve a ten cycle per day rate. A series of trays was fabricated to hold each sample in accordance with ASTM C 666, namely to surround the samples with between 0.8 mm (1/32 inch) to 3.2mm (1/8 inch) of water. Additional design goals for the trays were to minimize the volume of water and maximize convective heat transfer with the air inside chamber. A finished tray is shown in Figure 3. Each tray can hold eight samples and every other slot was machined all the way through to allow airflow vertically through the trays.
Freeze-thaw Durability
117
A second type of tray was also fabricated to allow the application of four point bending loads to each sample capable of causing 0.55% strain at the centerline on the surface of the tension face, see Figure 4. This strain level was chosen because it is beyond the “knee” in the stress-strain curve for each material in this research study and it likely opens up cracks that might be large enough to allow additional freezing to take place. This tray can also hold up to eight samples. Three levels of freeze-thaw exposure will be Figure 4. Loaded sample freeze-thaw tray. included in this study: 100, 300, and 500 cycles. A set of control samples will also be placed in a constant 4.4°C (40°F) bath for the duration of each freeze-thaw exposure level. POST-CONDITIONING TESTING AND ANALYSIS Similar to the as-received testing regime, ultimate tensile strength, stiffness, and strain-tofailure were determined quasi-statically in accordance with ASTM D 3039 for each class of material after saturation. A total of thirty samples were tested, ten of each material type. In addition, two samples of each material were set aside for crack density analysis using x-ray and optical microscopy techniques. Similar mechanical testing and crack density analyses will be performed at each of the three specified freeze-thaw conditioning levels for both unloaded and loaded samples in each of the general conditioning categories, i.e., saturated freeze-thaw (144 samples), saturated constant temperature (144 samples), and dry freeze-thaw (144 samples). CRACK DENSITY ANALYSIS Crack density was/will be assessed using non-destructive acousto-ultrasonic (AU) techniques with confirmation by optical microscopy. AU is an ultrasonic NDE technique useful for quantifying small, distributed changes not easily detected with traditional ultrasonic techniques. It returns results related to the ability of the interrogated material to transfer mechanical energy. In this setup, the transducers are spaced in a manner (typically, far apart) so that there is no direct or minimally reflected energy transmission between them and the primary transmission is by plate wave propagation. AU results are calculated from the energy spectral distribution of the AU signal by moment analysis. The most useful AU parameters tend to be the area under the curve (zeroth order moment) and the centroidal frequency (ratio of the first and zeroth moments). The area is directly related to the amount of energy transfer along the specimen. Shifts in the
118
Figure 5. As-received and post-saturation strength data.
Plastics Failure Analysis and Prevention
Figure 6. As-received and post-saturation stiffness data.
centroidal frequency also indicate changes in the materials ability to propagate particular modes. Typically, the spectral content of the AU signal has several relatively discrete frequency peaks with a few dominant ones. Each peak corresponds to a different mode or order of plate wave propagation. The modes are typically categorized as symmetric (extensional) and antisymeFigure 7. As-received and post-saturation strain-to-failure metric (flexural). The order of a particular mode data. indicates the complexity of that type of deformation. Since each peak represents energy propagating with a different type of deformation, it possibly can be sensitive to different types of degradation. AU samples will be taken from the general sample population and have already been baselined for as-received and post-saturation conditions. For optical microscopy, representative sections of material will be cut, potted, and polished for each post-conditioning group and examined using an inverted optical microscope.
RESULTS Though this research is still in its early stages, some data can be reported. Tensile strength, stiffness, and strain-to-failure are shown in Figures 5 through 7. Strength for the toughened vinylester was 389 MPa in the as-received condition versus 237 MPa for the post-saturation condition. Likewise, for the untoughened vinylester, strengths were 432 MPa versus 240 MPa. For the epoxy, strengths were 424 MPa versus 237 MPa. Stiffness for the toughened vinylester was 19.9 GPa in the as-received state versus 22.1 GPa for the post-saturation condition. Stiffness for the untoughened vinylester, was 23.9
Freeze-thaw Durability
119
GPa for both conditions. For the epoxy, stiffness was 26.2 GPa versus 25.6 GPa. Strain-to-failure for the toughened vinylester was 2.49% in the as-received condition versus 1.26% for the post-saturation condition. For the untoughened vinylester, strain-to-failure was 2.58% versus 1.36% and for the epoxy, 2.29% versus 1.20%. In summary, strength and strain-to-failure Figure 8. Moisture uptake curves for saturation at 65ºC. were approximately 50% lower after saturation for all three materials. Stiffness effectively remained unchanged. The moisture content at saturation was 0.70% for the toughened vinylester samples, 0.74% for the untoughened vinylester samples, and 0.84% for the epoxy samples. A moisture uptake curve is presented in Figure 8. It is clear from the curves that the composite does not obey Fick’s law. Because of this, the moisture aging experiment had to be terminated after 45 days at which point the heaters were turned off and the tank was allowed to equilibrate to room temperature.
CONCLUSIONS AND FUTURE WORK It is virtually impossible to freeze water in a highly crosslinked amorphous polymer. This is in part due to geometric space constraints in addition to hydrogen bonding that further impedes the process. In the composite system however, the crack dimensions are large enough to facilitate the freezability of water. The authors believe that this is this mechanism of freezing and the associated volume increase during the transition leads to the propagation of cracks and the accumulation of damage. Work is underway to begin freeze-thaw cycling and damage assessment as discussed earlier in this paper. It is hoped that information regarding crack density will be used to better understand the trends in strength and stiffness of these composites and the effect on freeze-thaw durability.
ACKNOWLEDGMENTS The authors would like to thank Dow Chemical Corporation, Dr. P. M. Puckett, and Herbert Englen for the use of their facilities and personnel in the pultrusion activities related to this research as well as their donation of the various resin systems. The authors would also like to thank Tim MacNeil and Owens-Corning for their donation of the glass reinforcement complexes used in this research.
120
Plastics Failure Analysis and Prevention
REFERENCES 1 2 3 4
5
Mitra, Dutta and Hansen, “Thermal Cycling Studies of a Cross-plied P100 Graphite Fiber Reinforced 6061 Aluminum Composite Laminate”, Journal of Materials Science, 26, November 1991. Gomez, J. P. and Casto, B., “Freeze-Thaw Durability of Composite Materials”, Virginia Transportation Research Council, 1996. Lord, H. W., and Dutta, P. K, “On the Design of Polymeric Composite Structures for Cold Regions Applications”, Journal of Reinforced Plastics and Composites, 7, 1988. Allred, R. E., “The Effects of Temperature and Moisture Content on the Flexural Response of Kevlar/Epoxy Laminates: Part II [45/0/90] Filament Orientation”, Environmental Effects on Composite Materials, Volume 2, George Springer, editor, Lancaster, PA: Technomic Publishing Company, Inc., 1984 Verghese, N., Haramis, J. Morrell, M.R., Horne, M.R., and Lesko, J.J. “Freeze-Thaw Durability of Polymer Matrix Composites in Infrastructure”, To be published in the proceedings of Duracosys’99.
Chapter 4 Morphology and Fractography Fractography of ABS Hiromi Kita, Masatoshi Higuchi, Atsushi Miura National Institute of Technology and Evaluation, Takashi Kuriyama, Ikuo Narisawa, Yamagata University, Japan
INTRODUCTION Failure analysis of plastic products is significantly difficult due to the insufficient systematic data of plastics. In order to solve this essential problem and to develop the technology to determine the cause of product failure, the systematic data for the fracture surface information and material strength of polymers have been required. In this paper, the application of fractography has been examined aiming on ABS polymers, which are broadly used as general consumer products and structural materials. Through the comparison of the fracture surface information obtained as a result of the experiments carried out under static and cyclic loading, the influences of repeated loading, notch, grades, loading level and ambient temperature on the fracture surfaces have been investigated.
EXPERIMENTAL Table 1. Experimental materials A(Heat Resistant)
B (High Rigidity)
C (Flame Resistant)
120,000
140,000
120,000
Flexural strength, MPa
73.5
78.4
65.7
Flexural modulus, MPa
2450
2548
2254
Impact strength, kJ/m2
14.7
9.8
12.7
Materials : ABS (Grades) Molecular weight
Three grades of ABS (Acrylonitrile-butadiene-styrene copolymer), heat resistant (material A), high rigid (material B) and flame resistant (material C) grades were used in this study as listed in Table 1.
122
Plastics Failure Analysis and Prevention
Figure 1. A schematic and photographs of fracture surface and the side of Compact Tension specimen (Material A). The magnification of fracture surface near the notch is also shown for (a) and (d).
Two types of specimens, Compact Tension (CT) specimen (Figure 1) and un-notched rectangular bar specimen (Figure 2) were used. The former is defined in ASTM D5045-93 (Standard test methods for plane-strain fracture toughness and strain energy release rate of plastic materials) except for the dimension of notch, and the latter is the preferred test specimen defined in ISO 178:1993 (Plastics-Determination of flexural properties). Static loading tests were carried out with both the CT and un-notched specimens at a testing speed of 10, 100 and 1000 mm/min. Fatigue tests were conducted at a test frequency of 5 Hz under load controlled condition. The maximum load level was set at 20, 40, 60, 80% of the tensile strength at yield or the flexural strength of the static tests at a testing speed of 10 mm/min. Note that only CT specimens are tested for the 20% load level. All the tests were performed at 0, 23, 50°C. The fracture surfaces were observed by using optical microscope (OM) and scanning electron
Fractography of ABS
123
Figure 2. A schematic and photographs of fracture surface of un-notched rectangular bar specimen (Material A).
microscope (SEM). In the CT specimens, the cross section observations were also carried out.
RESULTS The fracture aspects of CT specimens are shown in Figure 1. Striations were observed in all of the fatigue fracture surfaces from the OM observations (Figure 1 (a)-(c)), but never in the static ones (Figure 1 (d)). Striations tended to appear closer to the notch and increase its maximum spacing width with the increase of the applied load level (Figure 1 (b), (c)). Tearlines appeared on the condition at high ambient temperature and low loading level (Figure 1 (b)), or at high ambient temperature and low testing speed. Serration appeared only in the fatigue fracture surfaces on the condition of low load level (Figure 1 (a)), while only voids and fibrillation was observed in the static fracture surfaces (Figure 1 (d)) from SEM observations. And fibrillation caused by the delamination of flame retardant could be also observed in material C. Stress whitening appeared on the cross-section of the fractured specimens (Figure 1 (a)-(d)) and the behavior was dependent on the grade of the materials as follows, i.e., the half-ellipse shaped whitening area in material A, the dendrite shaped area with a slight whitening in material C, both half-ellipse and dendrite shaped whitening area in material B. But, the distinction between them became difficult with the decrease of the ambient temperature. All the stress whitening area tended to appear closer to the notch and increase in depth with the increase of the applied load level. The fracture aspects of un-notched rectangular bar specimens are shown in Figure 2. Un-notched rectangular bar specimens did not fail in all of the static loading conditions,
124
Plastics Failure Analysis and Prevention
though fractured in the cyclic loading conditions. Striations observed in all of the fatigue fracture surfaces although little could be determined for its appearance. The number of the fracture origin increased with the increase of the applied load level under the cyclic loading conditions (Figure 2 (a), (b)). The fractographic analysis of ABS polymers is summarized in Tables 2 and 3. With the systematic arrangement of the information obtained from the experiment the determinable properties became obvious as follows. The existence of repeated loading can be easily recognized for the appearance of the striations on the fracture surfaces. Notch effect is determined by the existence of the fracture origins. The grade of ABS is distinguished by the observation of additives and the stress whitening phenomenon on the cross sections near the fracture surfaces. The magnitude of stress affects the appearing position of the striations, while the ambient temperature influences the depth of the stress whitening with the notched specimens. Consequently, the estimation of the fatigue life still remains as a clue, hence further study is probably needed. Table 2. Fractography of ABS Compact Tension
Fatigue
OM
SEM
Static
OM
Striation Tear Line
High temperature, low load level
Whitening
Increased with the increase of temperature Closer to notch with the increase of load level
Void, Fibril
High temperature, high load level
Serration
Low load level
Tear Line
High temperature, low speed
Whitening SEM
Void, Fibril Un-notched Rectangular Bar
Fatigue
OM
Striation Origin
Static
Un-failed
Increase with the increase of load level
Fractography of ABS
125
Table 2. Fractography of ABS Grades of ABS Whitening
A: half-ellipse B: half-ellipse and dendrite C: dendrite
Additives
C: delamination of flame retardant
Table 3. Application of fractography to ABS Properties
Determination
Repeated Loading
Easily determined
Notch Effect
Determinable
Grades
Determinable
Stress Dependence
Determinable
Temperature Dependence
Determined on condition
Fatigue Life Estimation
Further study necessary
Notes
Determinable when notched
CONCLUSIONS This study has explored the application of fractography in ABS polymers. And the results presented in this paper, indicates that many causes of the failure can be determined through the observation of the fragments. In conclusion, we are eager to contribute to the expedition of failure analysis and to the prevention of repetition, through continuous acquisition of fracture surface information and material strength data of polymers to form a database.
ACKNOWLEDGEMENT The authors would like to express their sincere gratitude to Dr. K. Mizutani of Technology Research Institute of Osaka Prefecture for his advice and assistance during this work.
REFERENCES 1
Narisawa and T. Kuriyama, Seikei-Kakou, 1:5 529 (1999).
Fractography of Metals and Plastics
Ronald J. Parrington IMR Test Labs, 131 Woodsedge Drive, Lansing, NY 14882, USA
HISTORICAL PERSPECTIVE Plastics have been in existence for approximately 130 years. John Hyatt patented nitrocellulose, the first commercial plastic, in 1869. However, full-scale development and use of plastics is only about 50 years old. In contrast, metals have been in use for many hundreds of years. The application of engineering materials is unavoidably accompanied by the occurrence of failures, many of which have been catastrophic. The consequences of material failures; including deaths, financial losses and legal ramifications; have encouraged the development of effective failure analysis methods. Although the cost of failure analysis may exceed the value of the part, the cost of service failures usually far exceeds the cost of failure analysis. Many of the techniques utilized over the years for the evaluation of metals have been successfully applied to plastics with only minor modifications. Fractography is arguably the most valuable tool available to the failure analyst. Fractography, a term coined in 1944 to describe the science of examining fracture surfaces, has actually been utilized for centuries as part of the field of metallurgy. Even before that, however, Stone Age man possessed a working knowledge of fracture. Archeological findings of lithic implements, weapons and tools shaped from stone by controlled fracture, indicate that prehistoric man knew how to: (1) select rocks with favorable fracture behavior; (2) use thermal spalling to detach bedrock from the working core; and (3) shape stone by pressure flaking. Fractography as we know it today, developed in the 16th century as a quality control practice employed for ferrous and nonferrous metal working. “De La Pirotechnia” published by Vannoccio Biringuccio1 in 1540 is one of the first documents to detail fractographic techniques. Invention of the optical microscope in 1600 provided a significant new tool for fractography. Yet it was not utilized extensively by metallurgists until the 18th century. In 1722, R.A. de Réaumur2 published a book with engravings depicting macroscopic and microscopic fracture surfaces of iron and steel. Interestingly, the categories of macroscopic features developed by de Réaumur have remained essentially unchanged through the centuries.
128
Plastics Failure Analysis and Prevention
Partly due to the development of metallographic techniques for examining cross sections of metals, interest in microfractography waned during the 19th century. Metal workers continued to utilize fractographic techniques for quality assurance purposes but, for the most part, researchers and publications ignored fractography. Several technological developments in the 20th century revitalized interest in fractography. Carl A. Zapffe3 developed and extensively utilized fractographic techniques to study the hydrogen embrittlement of steels. His work lead to the discovery of techniques for photographing fracture surfaces at high magnifications. The first fractographs were published by Zapffe in 1943. An even more revolutionary development was the invention of the scanning electron microscope (SEM). The first SEM appeared in 1943. Unlike the transmission electron microscope, developed a few years earlier, it could be used for fracture surface examination. An SEM with a guaranteed resolution of ~500 angstroms became commercially available in 1965. Compared to the optical microscope, the SEM expands resolution by more than one order of magnitude and increases the depth of focus by more than two orders of magnitude. The tools for modern fractography were essentially in place before plastics achieved widespread use.
FAILURE ANALYSIS OVERVIEW The general procedure for conducting a sound failure analysis is similar for metallic and nonmetallic materials. The steps include: (1) information gathering; (2) preliminary, visual examination; (3) nondestructive testing; (4) characterization of material properties through mechanical, chemical and thermal testing; (5) selection, preservation and cleaning of fracture surfaces; (6) macroscopic examination of fracture surfaces, secondary cracking and surface condition; (7) microscopic examination; (8) selection, preparation and examination of cross sections; (9) identification of failure mechanisms; (10) stress/fracture mechanics analysis; (11) testing to simulate failure; and (12) data review, formulation of conclusions and reporting. Although the basic steps of failure analysis are nearly identical, some differences exist between metals and plastics. Nondestructive testing of metals includes magnetic particle, eddy current and radiographic inspection methods that are not applicable to plastics for obvious reasons. However, ultrasonic and acoustic emission techniques find applications for both materials. Likewise, different chemical test methods are necessary. Typical test methods for metals are optical emission spectrometry (OES), inductively coupled plasma (ICP) and combustion. Fourier transform infrared (FTIR) spectroscopy is extensively used to identify plastics by molecular bonding and thermal testing, differential scanning calorimetry (DSC)
Fractography of Metals and Plastics
129
and thermogravimetric analysis (TGA), is also very important for polymer characterization. Energy dispersive X-ray spectroscopy (EDS), used in conjunction with the SEM, is a very practical tool for elemental chemical analysis of metals and plastics. Also noteworthy, different chemical solutions are required for metals and plastics to clean fracture surfaces and to etch cross-sections to reveal microstructure. Figure 1. Fracture of a glass-filled nylon threaded part due to stress concentration. (8.4X)
CAUSES OF FAILURE
Of course, the primary objective of a materials failure analysis is to determine the root cause of failure. Whether dealing with metallic or nonmetallic materials, the root cause can normally be assigned to one of four categories: design, manufacturing, service or material. Often times, several adverse conditions contribute to the part failure. Many of the potential root causes of failure are common to metallic and nonmetallic materials. Improper material selection, overly high Figure 2. Cross section showing fracture along the knit line of a Teflon PFA lined impeller. (3.8X) stresses, and stress concentrations are examples of design-related problems that can lead to premature failure. Material selection must take into account environmental sensitivities as well as requisite mechanical properties. Stress raisers are frequently a preferred site for fracture origin, particularly in fatigue. These include thread roots (Figure 1), sharp radii of curvature, through holes, and surface discontinuities (e.g., gate marks in molded plastic parts). Likewise, many manufacturing and material problems found in metals are also observed Figure 3. Cross section of a Delrin hinge that fractured or have a corollary in plastics. Weldments are a (arrows) through an area of porosity. (8.4X) trouble prone area for metals, as are weld lines or knit lines in molded plastics (Figure 2). High residual stresses can result from metal
130
Plastics Failure Analysis and Prevention
forming, heat treatment, welding and machining. Similarly, high frozen-in stresses in injection molded plastic parts often contribute to failure. Porosity and voids are common to metal castings and plastic molded parts (Figure 3). These serve as stress raisers and reduce load carrying capability. Other manufacturing- and material-related problems that may lead to failure include adverse thermo-mechanical history, poor microstructure, material defects and contamination. Environmental degradation is one of the most important service-related causes of failure for metals and plastics. Others include excessive wear, impact, overloading, and electrical discharge.
FAILURE MECHANISMS Another key objective of failure analysis is to identify the failure mechanism(s). Once again, some failure modes are identical for metals and plastics. These include ductile overload, brittle fracture, impact, fatigue, wear and erosion. Analogies can also be drawn between metals and plastics with regards to environmental degradation. Whereas metals corrode by an electrochemical process, plastics are vulnerable to chemical changes from aging or weathering. Stress corrosion cracking, a specific form of metallic corrosion, is similar in many ways to stress cracking of plastics. Both result in brittle fracture due to the combined effects of tensile stress and a material specific aggressive environment. Likewise, dealloying or selective leaching in metals, the preferential removal of one element from an alloy by corrosion, is somewhat similar to scission of polymers, a form of aging which can cause chemical changes by selectively cutting molecular bonds.
FRACTOGRAPHY When material failure involves actual breakage, fractography can be employed to identify the fracture origin, direction of crack propagation, failure mechanism, presence of material defects, environmental interaction, and the nature of stresses. Some of the macroscopic and microscopic features employed by the failure analyst to evaluate fracture surfaces of metals and plastics are described below. Note, however, that many of the fractographic features described for plastics are not observable for reinforced plastics and plastics containing high filler content.
MACROSCOPICALLY VISIBLE FRACTOGRAPHIC FEATURES On a macroscopic scale, all fractures (metals and plastics) fall into one of two categories: ductile and brittle. Ductile fractures are characterized by material tearing and exhibit gross
Fractography of Metals and Plastics
Figure 4. Beach and radial marks emanate from the origin (“O”) of this torsional fatigue fracture. (0.5X).
131
Figure 5. Brittle fracture of an epoxy layer displays a mirror zone, rib marks and hackles. (5.6X)
plastic deformation. Brittle fractures display little or no macroscopically visible plastic deformation and require less energy to form. Ductile fractures occur as the result of applied stresses exceeding the material yield or flow stress. Brittle fractures generally occur well below the material yield stress. In practice, ductile fractures occur due to overloading or under-designing. They are rarely the subject of a failure analysis. Fracture analysis usually involves the unexpected brittle failure of normally ductile materials. Many macroscopically visible fractographic features serve to identify the fracture origin(s) and direction of crack propagation. Fractographic features common to metals and plastics are radial marks and chevron patterns. Radial marks (Figure 4) are lines on a fracture surface that radiate outward from the origin and are formed by the intersection of brittle fractures propagating at different levels. Chevron patterns or herringbone patterns are actually radial marks resembling nested letters “V” and pointing towards the origin. Fatigue failures in metals display beach marks and ratchet marks that serve to identify the origin and the failure mode. Beach marks (Figure 4) are macroscopically visible semielliptical lines running perpendicular to the overall direction of fatigue crack propagation and marking successive positions of the advancing crack front. Ratchet marks are macroscopically visible lines running parallel to the overall direction of crack propagation and formed by the intersection of fatigue cracks propagating from multiple origins. Brittle fractures in plastics exhibit characteristic features, several of which are macroscopically visible (Figure 5). These may include a mirror zone at the origin, mist region, and rib marks. The mirror zone is a flat, featureless region surrounding the origin and associated with the slow crack growth phase of fracture. The mist region is located immediately adjacent to the mirror zone and displays a misty appearance. This is a transition zone from slow to fast crack growth. Rib marks are semi-elliptical lines resembling beach marks in metallic fatigue fractures.
132
Plastics Failure Analysis and Prevention
MICROSCOPICALLY VISIBLE FRACTOGRAPHIC FEATURES
Figure 6. Dimpled appearance typical of ductile fracture of metallic materials. (375X)
Figure 7. Fracture of a polyethylene tensile test specimen exhibits material stretching. (375X)
Figure 8. Brittle fracture of a FC-0205 powder metal control rod displays cleavage facets. (375X)
On a microscopic scale, ductile fracture in metals (Figure 6) displays a dimpled surface appearance created by microvoid coalescence. Ductile fracture in plastics (Figure 7) is characterized by material stretching related to the fibrillar nature of the polymers response to stress. Although a part may fail in a brittle manner, ductile fracture morphology is frequently observed away from the origin, if the final fast fracture occurred by ductile overload (e.g., the “shear lip” in metal failures). The extent of this overload region is an indication of the stress level. Brittle fracture of metallic materials may result from numerous failure mechanisms, but there are only a few basic microfractographic features that clearly indicate the failure mechanism: (1) cleavage facets (Figure 8); (2) intergranular facets (Figure 9); and (3) striations (Figure 10). Cleavage facets form in body-centered cubic (BCC) and hexagonal close-packed (HCP) metals when the crack path follows a well defined transgranular crystallographic plane (e.g., the {100} planes in BCC metals). Cleavage is characteristic of transgranular brittle fracture. Intergranular fracture, recognizable by its “rock candy” appearance, occurs when the crack path follows grain boundaries. Intergranular fracture is typical of many forms of SCC, hydrogen embrittlement and temper-embrittled steel. Fatigue failures of many metals exhibit striations at high magnifications (normally magnifications of 500 to 2,500X are required). Striations are semi-elliptical lines on a fatigue fracture surface that emanate outward from the origin and mark the crack front position with each successive
Fractography of Metals and Plastics
Figure 9. Intergranular fracture of an embrittled cast steel pneumatic wrench. (375X)
133
Figure 10. Fatigue striations are visible on this Type 302 stainless steel spring fracture. (1500X)
stress cycle. The spacing of fatigue striations is usually very uniform and can be used to calculate the crack growth rate if the cyclic stress frequency is known. Striations are discriminated from striation like artifacts on the fracture surface in that true fatigue striations never cross or intersect one another. Plastics do not display cleavage and intergranular fracture. However, like metals, striations are found on fatigue fracture surfaces. Often, striations in plastics are observable at much lower magnifications (<100X). In addition to mirror zones, mist regions and rib marks, brittle fracture of plastics may display hackles, Wallner lines and conic marks. Hackles (Figure 5) are divergent lines radiating outward from the fracture origin. They resemble river patterns observed on the cleavage facets of transgranular brittle fractures of metals, but run in the opposite direction. Wallner lines are faint striation-like markings formed by the interaction of stress waves reflected from physical boundaries with the advancing crack front. Conic marks are parabolic-shaped lines pointing back towards the origin. Hackles and Wallner lines may or may not be visible without the aid of a microscope.
CLOSING REMARKS Fractographic techniques developed and applied to metal failures for centuries have been readily adapted to the fracture analysis of plastics since their emergence as a key engineering material over the last 50 years. However, more work remains to be done to advance fractography of plastics. One notable area for research is fracture analysis of composites, reinforced plastics, and plastics containing high filler content. Fractures of these materials are currently dismissed as inherently lacking meaningful fractographic features. Finally, there is a definite need for an authoritative publication on fracture in plastics.
134
Plastics Failure Analysis and Prevention
ACKNOWLEDGEMENTS The author gratefully acknowledges the contributions of Dave Christie and Steve Ruoff of IMR Test Labs.
REFERENCES 1 2 3 4 5 6
Brostow, W. and Corneliussen, R.D., Failure of Plastics, Hanser publishers, Munich, 1986. Davies, T.J. and Brough, I., “General Practice in Failure Analysis”, Metals Handbook, 9 th edition, Volume 11, ASM, 1986. Ezrin, M., Plastics Failure Guide: Cause and Prevention, Hanser publishers, New York, 1996. Fractography and Atlas of Fractographs, Metals Handbook, 8 th edition, Volume 9, ASM, 1974. Larson, F.R. and Carr, F.L., “How Failures Occur … Topography of Fracture Surfaces”, Source Book in Failure Analysis, ASM, 1974. Portugall, U. and Steinlein, K., Practical Metallography, 36:8, 446-462 (1999).
BIBLIOGRAPHY 1 2
3
Biringuccio, V., “De La Pirotechnia”, Venice, 1540; see translation by C.S. Smith and M.T. Gnudi, “The ‘Pirotechnia’ of Vannoccio Biringuccio”, AIME, New York, 1942. Réaumur de, R.A., “L’Art de Convertir le Fer Forgé en Acier, et L’Art d’Adocir le Fer Fondu”, (“The Art of Converting Wrought Iron to Steel and the Art of Softening Cast Iron”), Michel Brunet, Paris, 1722; see translation by A.G. Sisco, “Réaumur’s Memoirs on Steel and Iron”, University of Chicago Press, 1956. Zapffe, C.A. and Moore, G.A., Trans AIME, 154, 335-359 (1943).
Crack Propagation in Continuous Glass Fiber/ Polypropylene Composites: Matrix Microstructure Effect
M. N. Bureau and J. Denault Industrial Materials Institute, National Research Council Canada, Canada F. Perrin and J. I. Dickson École Polytechnique, Montreal, Canada
INTRODUCTION Continuous fiber composites with commodity thermoplastic matrix systems of relatively low glass transition temperature, such as polypropylene/glass fiber (PP/GF), are increasingly used in various applications, namely in the automotive sector. These relatively new composites combine the general advantages of semi-crystalline thermoplastic matrix composites with respect to molding processes, damage tolerance, chemical and environmental resistance and recycling possibilities, with the need for lower costs of production and higher production rates in large volume markets such as the construction, transport and automotive industries.1 Impact studies on discontinuous GF composites with a PP matrix have suggested1 that the use of a tough thermoplastic matrix, instead of a thermoset matrix with glass transition well above room temperature, in composites may result in significant toughness improvement, especially at low temperatures (e.g., -40°C). It has been established that the fracture behavior of GF composites with a PP matrix or a polyethylene terephthalate (PET) matrix is influenced by the matrix crystalline structure2-3 as well as the laminate configuration.4 Among several processing parameters affecting the mechanical behavior of thermoplastic GF composites, the cooling rate employed during the molding process, by modification of the matrix morphology, appears to be critical.2-4 In continuous GF composites with a PP or a PET matrix, mode I quasi-static fracture studies showed that the fracture behavior of is dominated by fiber pullout, whereas mode II quasi-static fracture studies showed that their fracture behavior is matrix dominated.3,5 Moreover, fatigue crack propagation studies6 have shown that interlaminar fracture, mostly delamination under mode II conditions, plays a key role in the fatigue fracture of continuous GF composites. Fatigue crack propagation is also generally very sensitive to the morphol-
136
Plastics Failure Analysis and Prevention
ogy and the microstructure of the materials tested. Mode I fracture testing is thus appropriate to study the influence of GF/matrix interactions, and mode II fracture testing appropriate to study the influence of the matrix itself on the fracture process in these composites. The objective of this paper is to study the effect of the matrix morphology, by using two different cooling rates in the molding process, on the mechanical behavior of a PP/GF composite.
EXPERIMENTAL MATERIAL AND SPECIMEN PREPARATION Pre-impregnated unidirectional 0.25 mm thick tapes, supplied by Baycomp Canada, made from a blend of chemically modified PP with pure PP reinforced with 60 wt% of continuous E-glass fibers coated with a specific thermoplastic sizing, were employed. The nominal thickness of these tapes was 0.25 mm. Unidirectional PP/GF plates of 4 mm in nominal thickness were prepared from 16 layers of these tapes. These specimens were compression molded at 200°C for 5 minutes in a Wabash press at 0.69 MPa and cooled to room temperature at two different rates Figure 1. Cumulative distribution of the spherulite diam- of 1°C/min or 10°C/min. A permanganate chemeter obtained by image analysis of SEM observations in ical etching technique of pre-polished surfaces the composites studied. was used to reveal the crystalline structure of the PP/GF composite. Image analysis of scanning electron microscope (SEM) observations was used to characterize the spherulitic matrix morphology (Figure 1). As shown previously,2,7 the cooling rate employed during the molding process resulted mainly in differences in the matrix morphology of the PP/GF composite employed. At higher cooling rates (10°C/min), a relatively fine spherulitic matrix morphology, 26 µm in average diameter, with an amorphous phase surrounding the fibers and between the spherulites, was observed. At low cooling rates (1°C/min), coarser spherulitic matrix morphology, 34 µm in average diameter, with fibers free of amorphous phase, was observed. THREE-POINT BENDING TESTS The three-point bending tests were performed following the ASTM D-790M standard test method at 23°C using a computer-controlled mechanical tester. These tests were done at a
Crack Propagation
137
crosshead speed of 1.3 mm/min with a ratio of span to specimen thickness of 16. The flexural2 modulus, the flexural strength and the strain at failure on the tensile stress side of the three-point bending specimen (lower fiber) were respectively calculated according to ASTM D-790M. A minimum of five tests was performed for each reported value and the standard deviation was less or equal to 5%. DOUBLE-CANTILEVER BEAM TESTS The mode I quasi-static fracture tests were performed using the DCB specimen configuration according to ASTM D-5528 standard test method. The nominal specimen dimensions were 20 mm in width, w, 150 in length, L, and 4 mm in thickness, 2h. An initial crack length, ao, of 35 mm was obtained using a 25 µm thick film of PET inserted at mid-thickness of the plates prior to consolidation. The DCB specimens were tested at a crosshead speed of 2 mm/min using an Instron computer-controlled Figure 2. Comparison of the experimental compliance plot- mechanical tester. As shown in Figure 2, good ted against the normalized crack length with the compliance agreement for both composites was obtained calculated from the beam theory for the DCB (Eq. 1) and between the compliance measurements on the ENF (Eq. 3) specimens. load-displacement curve at a given crack length, a, and the compliance expression derived from the beam theory, given by the following equation: 3
8a C = ------------3 Ewh
[1]
where E is the Young’s modulus. Using the compliance calibration method, the critical strain energy release rate, Gk, can be obtained using its definition: 2
P ∂C G Ic = ------c- ------2w ∂a
[2]
where Pc is the critical load determined from the non-linearity on the load-displacement curve and ∂C / ∂a is given by the compliance calibration. This critical strain energy release rate corresponds to an energetic expression of the fracture toughness. Two critical strain energy release rates are reported for each composite tested, one at the onset of crack propagation, GIc (onset), and one at the crack propagation plateau, GIc (propagation). These two GIc s correspond to the fundamental values from a typical R-curve of a tough material.
138
Plastics Failure Analysis and Prevention
END-NOTCHED FLEXURE TESTS The mode II fatigue crack propagation tests were performed using the ENF specimen configuration.3-5,8-9 The ENF specimen is a three-point bending specimen with an imbedded through-width delamination located at the laminate mid-thickness. The nominal total span, 2L, the nominal thickness, 2h and the nominal width, w, were respectively 110 mm, 4 mm and 24 mm. An initial delamination length, ao, of 24 mm was obtained using a 25 µm thick film of PET inserted at mid-thickness of the plates prior to consolidation. These tests were performed at 23°C with a sinusoidal waveform, a cyclic frequency of 5 Hz and a load ratio of 0.1. To avoid compressive forces resulting in artificially high crack propagation resistance and to minimize the influence of sliding friction on the strain energy release rate, a minimum of 26 mm and a maximum of 51 mm were respectively employed for the crack length.8 The crack length was monitored during the fatigue tests using the compliance method.8-9 Each fatigue test was verified by a second test. As shown in Figure 2, good agreement was obtained for both composites (cooled at a rate of 1°C/min and 10°C/min) between the compliance measurements and the compliance expression derived for the ENF specimen9 from the beam theory given by the following equation: 3
3
2L + 3a )C = (---------------------------3 8Ewh
[3]
where E is the Young’s modulus. Using the compliance calibration method, the crack length and the fatigue crack growth rate, da/dN, where N is the number of cycles, can thus be obtained. Each fatigue crack growth rate was obtained over a minimum propagation distance of 0.2 mm. With the crack length obtained, the mode II strain energy release rate, ∆ GII, expressed as a function of the constant load amplitude, ∆ P = Pmax - Pmin, can be calculated using a modified version of its definition: 2
∆P - ∂C ------∆GII = --------2w ∂a
[4]
where ∂ C/ ∂ a is given by the compliance calibration. The fatigue crack propagation by delamination is thus expressed in terms of da/dN versus ∆ GII curves on a log-log scale.
RESULTS AND DISCUSSION THREE-POINT BENDING The typical curves of the three-point bending tests performed on the PP/GF composite are presented in Figure 3. These curves show that the flexural modulus of the unidirectional specimens is not significantly affected by the cooling rates in the range tested, whereas the flexural strength and the strain at failure varied with the cooling rates employed. The specimens molded with a cooling rate of 1°C/min showed lower flexural strengths and the strains
Crack Propagation
139
Table 1. Results of the three-point bending tests in the PP/GF composites Cooling rate Mechanical property
Figure 3 Typical flexural stress – deflection curves of the three-point bending tests performed on the PP/GF composite. The open and closed symbols represent specimens molded with a cooling rate of 1°C/min and 10°C/min, respectively.
1oC/min
10oC/min
Flexural modulus, GPa
24
25
Flexural strength, MPa
405
470
2
6
Strain at failure, %
at failure than those molded with a cooling rate of 10°C/min. These results are presented in Table 2. Results of the mode I Table 1. quasi-static fracture tests These tests show that the fracture behavior of these composites is affected by the microCritical strain energy release 2 structure of the PP matrix. It should be kept in rate, GIc, J/m Cooling mind that an amorphous phase surrounding the rate onset propagation fibers is observed2 at relatively high cooling rates (10°C/min), whereas the fibers are free of 1oC/min 200 770 amorphous phase at low cooling rates (1°C/ min). In addition, short beam shear tests showed 1270 2010 10oC/min that considerably lower apparent interlaminar shear strengths were obtained at low cooling rates. It thus appears that the fracture behavior of the PP/GF composite is affected by the efficiency of stress transfer between the matrix and the fibers and also between the spherulites in the matrix itself. At lower cooling rates, the highly crystalline spherulitical structure led to a reduced presence of the amorphous phase between spherulites and between the fibers and the spherulites, causing premature fracture. MODE I QUASI-STATIC FRACTURE The results of the mode I quasi-static fracture tests are shown at Table 2 in terms of critical strain energy release rate GIc at the onset of crack propagation and during propagation
140
Figure 4. Mode I quasi-static fracture surfaces of the PP/ GF composite molded with a cooling rate of: a) 10°C/min and 1°C/min.
Plastics Failure Analysis and Prevention
(steady crack growth), using DCB specimens of PP/GF composite molded with cooling rates of 1°C/min and 10°C/min. According to the threepoint bending results, the DCB test results in Table 2 reveal that the cooling rate employed during the molding process of the PP/GF composites critically affects their resistance to steady crack growth (fracture toughness). The critical strain energy release rates at the onset of propagation and for propagation respectively drop by a factor of 6 and 3 when a cooling rate of 1°C/min is employed instead of 10°C/min. Mode I quasi-static fracture surface observations (Figure 4) show that a different mode of fracture is observed for each cooling rate. At 10°C/min, a transpherulitic fracture characterized by fiber/matrix delamination resulting from the presence of the ductile amorphous PP phase at the fiber surface2 is observed (Figure 4a). In this case, the PP spherulitic matrix presents extensive stretching resulting from crack bridging. At 1°C/min, an interspherulitic fracture is observed (Figure 4b). In this case, lower cooling rates result in the reduced presence of the amorphous phase between the fibers and the spherulites and also between spherulites in the matrix itself, leading to weaker interspherulitic regions (2) and lower fracture toughness (Table 2). MODE II FATIGUE CRACK PROPAGATION
The log-log curves of the fatigue crack growth rates da/dN plotted against the strain energy release rate ∆ GII are shown in Figure 5. Despite the non-negligible amount of experimental variation observed in the crack propagation data, the fatigue crack propagation curves in Figure 5 show that crack propagation occurs at growth rates 6 to 10 times higher, at a given ∆ GII, when a cooling rate of 1°C/min instead of 10°C/min is employed. The results in Figure 5 also indicate that the fatigue failure occurs at a ∆ GII of approximately 730 J/m2 for a
Crack Propagation
141
Figure 5. Fatigue crack growth rates da/dN plotted against the strain energy release rate ∆ GII in the PP/GF composites.
cooling rate of 1°C/min, whereas it occurs at a ∆ GII of approximately 1250 J/m2 for a cooling rate of 10°C/min. The mode II fatigue fracture surfaces are shown in Figure 6. These mode II fatigue fracture surfaces are in agreement with the mode I quasi-static fracture surfaces (Figure Figure 6. Mode II fatigue fracture surfaces of the PP/GF com4). At 10°C/min (Figure 6a), a transpheru- posite molded with a cooling rate of: a) 10°C/min and 1°C/ min. The macroscopic fatigue delamination propagation direclitic fracture with the formation of cusps in tion is from left to right. the PP matrix and some matrix/fiber debonding, is observed. These cusps, formed by shear loading of the matrix in the plastic zone, have been reported in other thermoplastic composites.5,8 However, such cusps have not, to the knowledge of the authors, been reported in PP/GF composites. At 1°C/min (Figure 6b), an interspherulitic fracture is observed. As mentioned previously, the low cooling rates used in the latter resulted in weak interspherulitic regions, providing an easy interspherulitic crack propagation path and resulting in a more brittle behavior (higher growth rates at given ∆ GII and lower ∆ GII at fatigue failure). This is in agreement with the mode I quasi-static fracture tests results showed previously.
142
Plastics Failure Analysis and Prevention
CONCLUSIONS Lower cooling rates during the molding process of PP/GF composites lead to reduction in the flexural strength and the strain at failure. The compliance of DCB and ENF specimens of PP/GF composites follows the prevision of the beam theory. The critical strain energy release rates obtained from mode I quasistatic loading is considerably affected by the cooling rate employed during the molding process of the PP/GF composite. Lower cooling rates result in reduced resistance to steady crack growth, both at the onset of propagation and for propagation itself. The fatigue crack propagation behavior in mode II cyclic loading is also strongly affected by the cooling rates employed during the molding process. The fatigue crack growth rates at given levels of strain energy release rate and the range of strain energy release rates obtained reveal a considerably poorer fatigue crack propagation behavior in the lower cooling rate PP/GF composite. In mode I quasi-static fracture and in mode II fatigue fracture testing, an interspherulitic mode of fracture was observed in the lower cooling rate composite, whereas a transpherulitic mode of fracture was observed in the higher cooling rate composite. This effect is related to the reduced presence of a ductile amorphous phase at the fiber-matrix interface and between the spherulites in the matrix as the cooling rate decreases, resulting in poorer quasi-static and fatigue crack propagation behavior.
ACKNOWLEDGEMENTS Financial support from Natural Sciences and Engineering Research Council of Canada (NSERC) and National Research Council of Canada (NRC) are gratefully acknowledged.
REFERENCES 1 2 3 4 5 6 7 8 9
Karger-Kocsis, J., Polypropylene: Structure, Blends and Composites, Vol. 3, Chapman & Hall, London, (1995). Youssef, Y. and J. Denault, Polym. Compos., 19 301-9 (1998). Ye, L. and K. Friedrich, J. Mater. Sci., 28 773-80 (1993). Ye, L., A. Beehag and K. Friedrich, Compos. Sci. Technol., 53 167-73 (1995). Gilchrist, M., N. Svensson and R. Shishoo, J. Mater. Sci., 33 4049-58 (1998). Reifsnider, K.L., Compos. Mater. Ser., 4 11-77 (1991). Bureau, M.N. and J. Denault, Polymer Composites ’99, SPE, October 6-8, 1999, p. 155-166. Trethewey Jr, R., J. W. Gillespie Jr and L. A. Carlsson, J. Compos. Mater., 22 459-83 (1988). Russell, J. and K. N. Street, ASTM STP 876, 349-70 (1985).
Fracture Behavior of Polypropylene Modified with Metallocene Catalyzed Polyolefin
Laura A. Fasce and Patricia M. Frontini Division Polimeros, Dpto. De Ingenieria en Materiales, Instituto de Investigaciones en Ciencia y Tecnologia de Materiales, 7600 Mar del Plata, Argentina Shing-Chung Wong Division of Materials Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798 Yiu-Wing Mai Centre for Advanced Materials Technology, Department of Mechanical and Mechatronic Engineering , The University of Sydney, NSW 2006 Australia.
INTRODUCTION Toughening of brittle homopolymers continues to generate great scientific and commercial interest. Recent developments of metallocene catalyzed thermoplastic elastomers have created new frontiers for research in characterizing the toughening behavior of novel polymer blends containing physically miscible phases. In this paper, we present a variety of fracture assessment schemes for polypropylene homopolymers containing an elastomeric phase (ENGAGE POes 8100®). Puncture test at 3.5 m/s was conducted to assess the failure behavior under biaxial loadings. The J-integral fracture toughness was also measured to describe the resistance against crack initiation at a quasi-static loading rate. Fracture and deformation mechanisms were examined using petrographic thin sectioning technique1 and tensile dilatometry,2,3 respectively.
EXPERIMENTAL WORK Materials were compression molded and cut to standard specimens for fracture-mechanical tests. Homopolypropylene was mechanically mixed with 10, 20 and 30 wt% metallocene catalyzed polyethylene (ENGAGE POes 8100®) Crack tip plastic zone was investigated using single-edge- double-notch four-pointbend (SEDN-4PB) specimens coupled with petrographic thin sectioning technique.4,5 The polished thin sections were observed under an optical microscope.
144
Plastics Failure Analysis and Prevention
To investigate the deformation mechanism of each individual blend composition, dilatational strains were measured using a computer-controlled Instron 5567 system at a constant crosshead speed of 2.5 mm/min. Highly senstive clip-on extensometers were used to monitor strains in the longitudinal and transverse directions. The thickness strain and the width strain were assumed to be equal. The dilatational strains of the deformed samples were then calculated from2,3 2 ∆V ------- = ( 1 + ε x ) ( 1 + ε z ) – 1 V
[1]
where ε x denotes the strain in the loading direction and ε z denotes the strain in the lateral direction. The true stress, σ , of each sample was obtained from P σ = ---------------------------------2 Wo To ( 1 – ε z )
[2]
where P is the applied load, Wo and To are the original width and thickness, respectively. It is generally known that the technique of tensile dilatometry provides a volume dilatation slope that is indicative of dilatational plasticity. This slope is that of the volumetric strain versus axial strain plot. Puncture tests were conducted on clamped circular disks of 3.2 mm in thickness and 9 mm in diameter according to the ASTM standard 3029 at 3.5 m/s. All studied samples displayed a certain degree of stable crack growth and hence J-integral analysis was applied. J-R curves were determined by the normalization method6,7 using only a single pre-cracked specimen load-displacement record. Pre-cracks were introduced by a razor blade sliding into the machine notch. The specimens were then loaded in an Instron model 4467 testing machine at a rate of 1 mm/min at ambient temperature. After unloading, the specimens were fast fractured. The amount of crack extension, ∆a, for stable crack propagation was determined post mortem from the fracture surface. ∆a was monitored by marking with an alcoholic iodine solution before unloading. The data for the JR curve were best fitted to a power law relationship, J = C 1 ∆a
C2
[3] where C1 and C2 are curve-fitting parameters. The fracture initiation toughness was determined by the intersection of the JR-∆a curve with a vertical line at 0.2 mm offset and by comparison with a blunted precracked sample, Jspb.8
RESULTS AND DISCUSSION Puncture energies for different blend compositions are given in Table 1. Clearly, introduction of the rubbery phase greatly enhances the impact toughness of the homopolymer. The puncture energy increases 10-fold from PP homopolymer to a 10%-rubber blend. For 20
Fracture Behavior of Polypropylene
145
Figure 1. Load – time curves for punture tests. Figure 2. Visual examination of puncture specimens.
Table 1. Puncture energy Material
Energy
Standard deviation
PP homopolymer
2.1 J
±0.17
PP+10% mePE
26.4 J
±13.24
PP+20% mePE
45.3 J
PP+30% mePE
42.0 J
Table 2. J-integral fracture toughness Material
J0.2
Jspb
PP homopolymer
4.87 N/mm
4.99 N/mm
PP+10% mePE
3.52 N/mm
3.19 N/mm
±1.18
PP+20% mePE
4.02 N/mm
4.19 N/mm
±0.09
PP+30% mePE
3.16 N/mm
3.51N/mm
and 30% rubber blends, the fracture energy is about 20 times that of homopolymer. Figure 1 illustrates the typical load-time curves for the specimens tested. A visual examination of the tested samples greatly facilitates the understanding of the failure mode (see Figure 2). PP homopolymer fails in a brittle manner characterized by circumferential branching cracks (Figure 2a). A large scatter of observations was found, however, in 10% rubber blend. The material fails from a fully brittle mode to a semi-ductile mode for different specimens tested (Figure 1). Note that the blend exhibits large radial cracks from the contact point of the plunger. Conversely, the 20 and 30% blends displayed a clear ductile failure mode as confirmed by optical observations of the broken pieces, which show stress-whitening and permanent deformation around the plunger contact point. The slightly reduced puncture energy for additional rubber content is attributed to the softening effect induced by the elastomeric phase.
146
Plastics Failure Analysis and Prevention
Figure 3. J-R curves using normalized method.
J-R curves for PP modified with metallocene poly-ethylene are shown in Figure 3. The homopolymer displayed a small amount of stable crack growth after which it became unstable. At ambient temperature and under quasi-static conditions the rubber softening effect promoted stable crack growth in modified materials but fracture initiation values diminished (Table 2). Toughness enhancement under impact conditions and at lower temperatures would be expected for the rubber-modified blends. Crack fronts were not straight despite the presence of side grooves. Using petrographic thin sectioning technique, the arrested crack-tip can be studied using light microscopy. Figure 4 displays the crack- tip Figure 4. Arrested crack in SEDN-4PB specimens. deformation zones of 0, 10, 20 and 30% rubber blends. Note that there is little deformation surrounding the crack in the PP homopolymer (Figure 4a). However, an increase in rubber content dramatically introduces characteristic flow lines radiating from the crack tip. Extensive plastic deformation occurs around the crack tip for 20% and 30% rubber blends. It is not clear whether the plastic flow was asscciated with volume-conserving (shear yielding) or dilatational (crazing, rubber cavitation) mechanisms. To examine the tensile dilatometric properties, Figure 5 plots the magnitudes of axial, transverse and volume strains versus load-line displacement for 30% rubber blend.
Fracture Behavior of Polypropylene
147
Table 3. Volume dilatation slope Material
Volume dilatation slopes
PP homopolymer
1.166x10-1
PP+10% mePE
-2.426x10-2
PP+20% mePE
-5.456x10-3
PP+30% mePE
-8.502x10-4
Figure 5. Strains versus load-line displacement.
While both axial and tranverse strains increase with load-line extension, the volume strain as obtained from Eq. [1] decreases with axial strain. Table 3 compares the volume dilatation slopes of different blend compositions. From the negative nature of the dilatation slopes, it is conjectured that PP homopolymers modified with metallocene catalyzed PE exhibit strain-induced molecular restructuring arising from, for example, enhanced crystallinity as commonly found in semi-crystalline polymers like nylons.3,9 The results also suggest that there is negligble volume dilatation and the mechanism responsible for deformation could possibly arise from volume-conserving shear. While this is consistent with the softening effect of the matrix upon introduction of the elastomeric phase whereby shear resistance is greatly reduced at a lower loading rate, the extensive stress whitening around the crack tip awaits a more definitive examination like TEM to identify the fracture mechanism under a triaxial state of stress.
CONCLUSIONS Puncture energy and J-integral fracture toughness were measured for a series of PP homopolymers modified by metallocene catalyzed PE. A transition from fully brittle to ductile fracture was observed as the rubber content increased under impact conditions. For the modified materials, toughness was optimal for the 20% rubber blend because additional rubber caused softening effect. Measurements of the dilatational strains suggested the toughening mechanism under uniaxial loads for the blends could be derived from a volumeconserving deformation mechanism. A more definitive approach, however, is required to verify the dilatometric implications.
148
Plastics Failure Analysis and Prevention
References 1 2 3 4 5 6 7 8 9
Takemori, M. T. and Yee, A. F., Fractography and damage, in Impact Fracture of Polymers – Materials Science and Testing Techniques, Takahashi, K. and Yee, A. F. (editors), (Kyushu University, Fukuoka-shi, Japan, 1992) pp. 331-392. Bucknall, C. B., Toughened Plastics, (Applied Science Publishers, London, 1977) p. 197. Nair, S. V., Wong, S.-C. and Goettler, L. A., J. Mater. Sci. 32 5335-5346 (1997). Sue, H.-J. and Yee, A. F., J. Mater. Sci. 24 1447-1457 (1989). Sue, H.-J. and Yee, A. F., J. Mater. Sci. 28 2975-2980 (1993). Bernal, C. R., Cassanelli, A. N. and Frontini, P. M., Polymer Testing 14 p. 85 (1995). Bernal, C. R., Montemartini, P. E. and Frontini, P. M., J. Polym. Sci. Part B: Phys. Ed., 34 p. 1869 (1996). European Structural Integrity Society (ESIS), Technical Committee 4, Polymers and Composites, A Testing Protocol for Conducting J-Crack Growth Resistance Curve Tests on Plastics, 1992. Bucknall, C. B., Heather, P. and Lazzeri, A. J. Mater. Sci. 24 p. 1489 (1989).
Morphology and Mechanical Behavior of Polypropylene Hot Plate Welds
MJ Oliveira, CA Bernardo Dept Eng Polímeros, Universidade do Minho, 4800 Guimarães, Portugal DA Hemsley Polymer Microscopy Services, Loughborough, U.K.
INTRODUCTION Hot tool welding is a widespread and reliable technique for thermoplastics.1 It is applied to join components of variable complexity and size produced by extrusion, injection molding or others methods. In cases where the surfaces to be joined are flat, the hot tool is a plate with the temperature controlled on both sides; the process is called hot plate welding. In this process the surfaces to be joined are maintained under pressure against the hot plate, then the plate is withdrawn, and finally the matching surfaces are pressed together for joining. To avoid excessive lateral flow of melt out of the joint, some machines are equipped with rigid stops that bring the pressure automatically to zero once the part length is reduced to a fixed amount. This type of machines presents some advantages over the pressure controlled ones.1,2 and are particularly suitable for parts with small cross-sectional area, as the ISO tensile bars used in this work. Although significant research has been done to simulate3 and optimize4 the process, the relationship between the weld strength and the process parameters is not yet fully understood. Various types of tests have been used to evaluate the strength of the welds, including tensile,5 impact,6 and long term creep rupture tests.6,7 Microscopical examination has also been used to characterize hot plate welds.5,8-10 However, the evaluation of the weld quality is still uncertain. Polypropylene, a very versatile material, is used in applications ranging from extruded pipes to injection molded engineering parts that are welded by the hot plate process. The microstructure of this polymer is very sensitive to the crystallization conditions and to the thermo-mechanical process history. This makes the material particularly interesting for exploring processing-microstructure-property relationships. With this purpose, an injection molding grade of polypropylene was used to produce bars that were hot plate welded under
150
Plastics Failure Analysis and Prevention
a wide range of conditions. The morphology and failure behavior of the welds were assessed by microscopy and by short and long term mechanical testing.
EXPERIMENTAL The welding experiments were carried out with parts cut from ISO type tensile test bars. The bars were injection molded from polypropylene homopolymer (ICI GWM 22) of MFI 4.4g/600s, and cut in half. The welds were prepared with a purposely built semi-automatic hot plate welding machine. The hot plate temperature, the heating time and the joining displacement were varied using the conditions in Table 1. To guarantee a good contact between the hot plate and the surfaces to be welded, the latter were smoothed using a microtome. A fixed heating displacement of 0.25 mm was used in all the welds. Table 1. Welding conditions Weld
T/t/d, ºC/s/mm
RMD
Lo-d, mm
Seam temp., ºC
W1
190/50/0.35
0.78
0.01
170
W2
190/75/0.35
0.64
0.20
172
W3
190/100/0.35
0.60
0.23
174
W4
190/100/0.70
1.00
~0
167
W5
215/20/0.35
0.84
0.07
174
W6
215/50/0.35
0.52
0.32
188
W7
215/75/0.35
0.43
0.47
192
W8
215/50/0.70
1.00
~0
167
W9
215/75/0.70
0.86
0.12
174
W10
240/20/0.35
0.60
0.23
193
W11
240/50/0.35
0.38
0.57
208
W12
240/75/0.35
0.31
0.78
213
W13
240/20/0.70
1.00
~0
167
W14
240/50/0.70
0.76
0.22
182
Morphology and Mechanical Behavior of Polypropylene
151
Table 1. Welding conditions Weld
T/t/d, ºC/s/mm
RMD
Lo-d, mm
Seam temp., ºC
W15
240/75/0.70
0.62
0.43
191
W16
265/20/0.35
0.50
0.36
210
W17
265/50/0.35
0.31
0.77
228
W18
265/75/0.35
0.26
1.02
234
W19
265/20/0.70
0.98
0.01
170
W20
265/20/1.00
1.00
~0
167
W21
265/75/0.70
0.51
0.67
209
W22
290/20/0.35
0.43
0.46
229
W23
290/50/0.35
0.28
0.92
250
W24
290/20/0.70
0.87
0.10
180
T/t/d -hot plate temp./heating time/welding displacement RMD- ratio of melt displacement Lo-d – length of melt left at weld zone (from each part)
The melted zone depth, Lo, was measured in half bars that were heated in the welding machine, cooled and analyzed by polarized light microscopy. The edge of the melted zone was identified by analysis of the microstructure modifications caused by the heating process.11 The temperature profile of the melted zone was estimated using the heat conduction equation for a semi-infinite solid. For that, the thermal properties of the material were estimated from the experimental values of Lo.11 The microscopical examination of the welds was carried out with a polarizing microscope in 10-15 µm thick cross-sections. It included the analysis of the crystalline texture, size and shape of the beads, and the measurement of the insertion angles of the bead notches. The dimensions of the cross-section of the welded bars at the seam plane were also evaluated by microscopy. The welds and the original bars were subjected to several mechanical tests: • tensile testing at 500 mm/min of intact specimens and also of specimens notched with a hole of 3 mm diameter (Figure 1a).
152
Plastics Failure Analysis and Prevention
•
Figure1. Specimen geometry for testing. a – hole notch, b – curved notch.
tensile impact testing at room temperature using a digital pendulum machine. To increase the stress at the weld region, the specimens were notched with a wide radius notch (Figure 1b). • tensile creep-rupture testing at 70ºC in a water/5% Teepol bath at a nominal stress of 8 MPa. For the evaluation of the tensile strength and the resilience, the weld cross-sections were considered rectangular and the dimensions taken at the seam plane. The surfaces of the specimens fractured by mechanical testing were analyzed by scanning electron microscopy. The fracture cross-sections were observed by polarizing microscopy to identify the fracture path.
RESULTS AND DISCUSSION MELTED ZONE DEPTH The melted zone depth, Lo, was determined by measuring the distance between the outer surface of the bar and the row of small spherulites that limit the recrystallized region.12 The temperature at the weld seam, worked out from the heat conduction equation for a semi-infinite solid, was considered to be the one at the plane of the joining stops. The ratio of melt displacement, RMD = d/Lo, was determined from Lo and d, the welding displacement, defined by the position of the stops. The values of RMD and the weld seam temperatures are shown in Table 1. MORPHOLOGY OF THE WELDS
Figure 2. Weld of Type P1 (W13). 1 – bar skin; 2 – oriented texture.
The ratio of melt displacement is the parameter with greater influence on the weld morphology, particularly at the beads. Values of RMD below 0.5 originate a sharp and deep central V-notch in the beads, whereas above 0.5 this notch is smoothed and the severity of the contact notch increased. As Figure 2 illustrates, large welding displacements cause the folding of the bead over the bar and, in most cases, the sticking to the unmelted bar. This link is established through a transcrystallization process, caused by the nucleating action of the solid
Morphology and Mechanical Behavior of Polypropylene
153
bar onto the molten bead contacting it. The folding of the bead over the bar raises sharp contact notches. The sticking affects the position of its root. All the beads have a zone of oriented texture related to the original skin of the bar (Figures 2-5). The occurrence of this texture was referred to previously.11 It shows that in those zones the heating of the melt was not enough to erase the memory of the previous texture and to prevent its nucleating action during recrystallization. This memory and also the temperature Figure 3. Weld of Type P2 (W10). a – low bir. spherulites; b – high bir. spherulites; c- boundary layer affected the deformability of the melt as the shape of the beads suggests. In the welds made at 190ºC the deformability of the melt was low and consequently the beads acquired a lobular contour. In contrast, the beads of welds made at higher temperatures have a smoother aspect. Another interesting feature that may be of value in quality control is the occurrence of Type III spherulites at the outer surface of the welds made at 190ºC. It is known that melt shearing and an adequate rate of cooling favor this type of structure. These conditions appear to be present only for that particular hot plate temperature. Changes in the RMD change the depth of melt remaining at the central zone (Lo-d) and the seam temperature, as is shown in Table 1. As the values of those parameters increase, the crystalline texture tends to be more spherulitic, mostly of the low birefringent type. The Type III high birefringence spherulites only occur in regions where a high shear is combined with a relatively low temperature (~180ºC), as is the case of the weld shown in Figure 3. It was observed that the PTFE coating contaminated the surfaces contacting the hot plate. This is evidenced by the row-nucleated spherulites formed at the weld seam. The morphology of the welds may be summarized in four major types: Type P1 - this type of welds (as illustrated in Figure 2) has a narrow weld zone with a highly oriented texture. The beads are large and generally sticking to the unmelted material. This causes a great increase of the cross-section at the seam plane and severe contact notches, often located away from the weld region. Type P2 - in these welds the dominant microstructure is spherulitic (Figure 3) although some effect of orientation is still present in the crystallized texture. This is evidenced by the presence of Type III spherulites and by the transversely oriented texture of the boundary layers. The beads stick to the bar. Type P3 - in clear contrast to the previous types this group of welds display detached beads and a wide and uniform spherulitic central zone (Figure 4).
154
Plastics Failure Analysis and Prevention
Figure 4. Weld of Type P3 (W11) – interface with transcrystalline texture.
Figure 5. Weld of Type P4 (W23).
Type P4- the most characteristic features of these welds, illustrated in Figure 5, are the occurrence of voids and the very deep and sharp central notch of the beads. These features result from the use of a too small welding displacement in the joining of a large and hot amount of melt. Consequently, the shrinkage ± was very high, favoring the formation of voids and reducing the cross-sectional area at the seam plane. MECHANICAL PROPERTIES AND FAILURE BEHAVIOR The mechanical behavior of the welds varied with the welding conditions and the type of test used. Table 2 shows the results obtained. The relation between failure and weld morphology is referred below. Table 2. Mechanical properties of the welds Tensile strength, MPa
Resilience, kJ/m2
Creep rupt. time, h
W4/P1
11.3±0.7
54.23±4.8
64±10
W5/P1*
14.9±1.2
46.8±4.3
47±8
W18/P1
11.2±1.2
36.5±6.7
51±6
W13/P1*
7.0±2.4
35.1±6.2
39±3
W19/P1*
12.6±0.7
37.5±5.6
51±3
W20/P1*
8.5±1.6
33.2±3.0
41±4
W1/P1-P2
14.6±0.8
61.7±11.3
54±4
Weld/type
Morphology and Mechanical Behavior of Polypropylene
155
Table 2. Mechanical properties of the welds Tensile strength, MPa
Resilience, kJ/m2
Creep rupt. time, h
W24/P2
15.2±1.5
64.6±2.3
50±10
W2/P2
13.7±0.2
63.1±7.6
61±12
W3/P2
14.8±0.4
61.1±7.2
63±10
W6/P2
16.5±0.8
60.1±4.7
58±7
W9/P2
14.0±0.7
50.1±10.9
63±12
W10/P2
18.9±1.4
65.9±4.6
49±9
W14/P2
17.4±1.1
66.0±2.6
48±5
W7/P3
18.6±0.8
64.0±6.5
79±15
W11/P3
19.8±0.8
56.5±13.4
58±7
W15/P3
18.4±0.8
62.2±7.0
56±17
W16/P3
21.2±1.0
70.1±4.6
77±11
W22/P3
18.9±1.1
64.9±6.4
41±4
W25/P4
11.4±6.0
23.0±6.4
45±4
W12/P4*
4.0±0.5
23.7±6.7
11±5
W17/P4*
5.7±2.0
25.6±7.5
0
W21/P4
16.8±5.0
41.0±16.7
22±13
W23/P4*
5.2±1.1
28.3±3.4
58±10
orig. bar
37.8±0.4
95.8±5.8
11±8#
Weld/type
* - failed at the seam plane when tested with beads on # - time to yield.
TENSILE TESTING Only a small number of welds failed at the weld zone in this test (indicated in Table 2). Failure was caused either by a high level of transversal orientation (in welds of type P1) or by
156
Plastics Failure Analysis and Prevention
voids and deep V-notches at the beads (in welds of type P4). However, other welds with identical features broke away from the weld zone, which confirms the low ability of this test to discriminate between good and bad welds.5,8 For the specimens notched with a hole at the center of the weld, the test showed to be more discriminative. Due to the stress concentration caused by the hole, the beads had a diminished role in the failure process and the rupture Figure 6. Fracture surface of weld W14 broke in the ten- was initiated and controlled by the weakened sile test. zones of the central part of the weld. In the welds of type P4, that showed the lowest strength, the voids or cracks caused the failure. In the welds with transverse molecular orientation the strength increases and the failure mode changes as the orientation decreases. Failure occurred either at the seam plane (P1) or at the boundary layer (P2), depending on the occurrence of the peak of molecular orientation. For the type P3 welds, of higher strength and elongation at break, failure occurred at the seam plane. The fracture showed ductile patches on the surface (Figure 6) and dense cracking underneath. In contrast, the weakest welds failed without signs of ductility. IMPACT TESTING The smooth and curved notch machined on the specimens only removed a small portion of the beads. This allowed the failure to be controlled by the morphology both at the central zone, as in the previous test, and at the beads. Thus, as more weakening features are influencing the failure, this impact test is considered to be more discriminative of the weld quality. In the welds of type P1 the dominant feature was the high molecular orientation causing the initiation of the fracture at the central zone. As the contact notches of the beads were frequently located away from the weld region, although very severe, did not contribute to the failure. In contrast, it caused the initiation of the fracture in welds of type P2. Depending on the notch location and the level of molecular orientation at the boundary layers, the fracture progressed either through one of those layers or through the unwelded zone (Figure 7). In type P3 welds the fracture was frequently located at the interface zone. The coarser and more brittle microstructure of this zone, the transcrystalline texture caused by the PTFE, and the stress concentration effect of the central V-notch, all contributed for the failure. As was
Morphology and Mechanical Behavior of Polypropylene
Figure 7. Cross-section of weld W14 broke in the impact test.
157
Figure 8. Cross-section of weld W3 broke in the creep rupture test.
expected from their microstructure, the P4 welds failed at the weld interface zone as in the previous tests. CREEP-RUPTURE TESTING The data in Table 2 shows that only a reduced number of welds failed at the weld zone. In the majority of the welds the fracture was initiated by the contact notch of the beads and, due to its location away from the weld zone, the fracture propagated through the base material. Figure 8 shows the fracture path for this type of failure. This helps to explain how welds produced with different processing conditions and having distinct microstructures showed identical failure times. Thus, in contrast to what has been referred to by other authors,6,7 this test showed not to be discriminative in assessing the weld quality. This behavioral difference is certainly related to the occurrence of the sticking of the bead to the bar in a great number of welds, which showed to be determinant of the failure behavior in this test. PROCESSING PARAMETERS AND WELD QUALITY It was shown that the weld performance is influenced by a number of morphological features or defects. Their occurrence depends on the welding parameters, sometimes in a contradictory manner. Thus, for example, a large melt displacement favors the setting of molecular orientation but attenuates the severity of the central V-notch and increases the load bearing area. Conversely, small welding displacements reduce the severity of the contact notch but worsen the central notch and the adhesion at the interface. In general terms, the properties of the welds reached their maximum for intermediate values of RMD, in the range from 0.38 to 0.75. The resilience, determined with the tensile impact test showed to
158
Plastics Failure Analysis and Prevention
correlate particularly well with the RMD and the morphology of the welds, as shown in Figure 9 However, the quality is affected not only by the RMD but also by the interface temperature. For the set of welds studied in this work, better performance was obtained for the conditions 240ºC/ 50 s/0.35 mm and 265ºC/20 s/0.35 mm that lead to an identical interface temperature (~210ºC).
CONCLUSIONS
Figure 9. Effect of the ratio of melt displacement on the resilience of the welds.
The following main conclusions can be drawn from this work: 1. The weld morphology depends on the welding parameters and on the previous microstructure
of the parts. 2. The influence of the morphological features of the welds is enhanced differently by the various tests. The tensile test on hole-notched specimens enhances the central zone while the creep rupture test enhances the beads. In the tensile-impact test the fracture behavior is affected by the overall morphology. 3. The optimum welds are void and orientation free at the central zone, and have detached beads with the central V-notches outside the bar plane. 4. The characterization of the weld morphology, in complement with mechanical testing, is of great value on predicting the quality of welded joints.
REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12
Stokes, V.J, Polym. Eng. Sci., 29, 1310 (1989). Watson, M.N. and Murch, M.G., Polym. Eng. Sci., 29, 1382 (1989). Potente, H., Kunststoffe, 67, 98 (1977). Potente, H. and Natrop, J., M.N. and Murch, M.G., Polym. Eng. Sci., 29, 1649 (1989). Atkinson, J.R. and DeCourcy, G.W., Plastics and Rubber Process and Applications, 1, 287, (1981) Zeeuw, K. and Potente, H., SPE 35th ANTEC Proceedings, 55 (1977). Hertforth, H., Weld. in the World, 11, 1456 (1974). Barber, P. and Atkinson, J.R., J. Mat. Sci., 9, 1456 (1977). Potente, H. and Reinke, M., Plastics and Rubber Process and Applications, 1, 149, (1981). Stevens, S. M., SPE ANTEC Proceedings, 1275 (1996). Oliveira, M.J., “Study of the Microstructure of Polyethylene and Polypropylene Welds”, PhD thesis, University of Minho, (1989). Oliveira, M.J. and Hemsley, D.A., British Polymer Journal, 17, 269 (1985).
The Influence of Morphology on the Impact Performance of an Impact Modified PP/PS Alloy
S. P. Bistany Montell Polyolefins, 912 Appleton Road, Elkton, MD 21921, USA
BACKGROUND Catastrophic failures were experienced in blowmolded parts during low temperature impact testing. The resin used to make the parts is normally ductile for the same test temperature and impact speed. The two main differences noted between past performance of the resin and this failure event was the process producing the part and thickness of the part wall. Blowmolding is a process that produces hollow parts.1 While several forms of blowmolding exist, extrusion blowmolding was used to make this particular part. In this process, a tube of resin (or parison) is first extruded vertically down between two mold halves. Once the parison is in position, the mold closes on it and a gas, typically air, is blown into the parison forcing it onto the mold surface which gives the part its final shape. This process exposes only one side of the parison to a cool mold surface allowing the internal surface to cool freely. With respect to wall thickness, impact data is typically generated on 3.2 mm injection molded panels. The average wall thickness on the blowmolded parts was on the order of 5.0 to 7.6 mm. Analysis of the parts ruled out contamination with dissimilar resin. Also, since the MFR of samples taken from the part were similar to the virgin pellets, degradation during processing was ruled out. Testing to investigate thickness effects on impact were performed along with SEM analysis of the failed parts.
FAILURE INVESTIGATION THICKNESS EFFECTS ON IMPACT To determine if wall thickness played a role in the impact failures, test panels were injection molded at thicknesses similar to those found in the part and compared to the standard test thickness of 3.2 mm as well as test samples cut from the failed part. Instrumented impact testing at –30oC and 2.2 m/s was performed. The results in Table 1 below show that thickness does not appear to be the cause of the failures since the thicker injection molded panels
160
Plastics Failure Analysis and Prevention
experienced ductile failure. However, all blow molded test samples experienced brittle failures. Table 1. Instrumented Impact Performance Molding Process
Thickness, mm
Failure Mode
Injection
3.2
Ductile
Injection
6.4
Ductile
Blow
5.0
Brittle
Blow
6.7
Brittle
MICROSCOPY INVESTIGATIONS
Figure 1. Blow molded part outside surface.
Figure 2. Blow molded part inside surface.
Figures 1 and 2 are SEM micrographs comparing the outside and inside surfaces of the blow molded part respectively. The outside surface cooled much more rapidly being in contact with the mold surface while the inside surface cooled at a much slower rate. Figure 2 details micro-voids and spherulite boundaries on the inside surface that resulted from the slow cooling. The outside surface shows a morphology free of micro-voids and spherulite boundaries.
The Influence of Morphology
Figure 3. Blow molded part outside surface cross-section.
161
Figure 4. Blow molded part inside surface cross-section.
Over the past several years, there have been many studies performed on this type of morphology. These studies have found that during crystallization, non-crystallizable polymer chains act as impurities and are forced to the spherulite boundary region.2 As a result, these polymer chains form weak links between spherulites. The micro-voids that form as a result of the contraction of material at the boundaries act as stress concentrators and are often crack initiation sites.3 Additional research has shown that once cracking has been initiated, the weak boundaries provide a “path of least resistance” and hence make good crack propagation sites.4-7 Attempts have been made to strengthen the spherulite boundaries by addition of copolymers which migrate to the boundaries during crystallization.8 The copolymers function to cocrystallize between spherulites that in turn form links between spherulites and thus strengthen the boundaries. To further illustrate the radical difference in morphology between the outside and inside surfaces, cross-sections of these surfaces were taken. The impact modifier was etched out for each sample prior to SEM analysis. Figure 3 shows the outside surface cross-section. Here the impact modifier particles are relatively small and striated. Figure 4 shows the inside surface cross-section. Here the impact modifier appears to coalesce and is not striated. Coalescence of impact modifier particles is known to lead to a reduction in impact strength.9 Additionally, the asymmetry of the striated particles is more effective at increasing toughness than spherical particles.
162
Figure 5. Injection molded panel surface cross-section.
Plastics Failure Analysis and Prevention
Figure 6. Injection molded panel core cross-section.
To compare morphologies resulting from the blow molding and injection molding processes, cross-sections of an injection molded panel were prepared in the same manner as that for the blow molded part. Figure 5 shows the cross-sectional view at the outside surface of the panel. The morphology is similar to that of the outside skin of the blow molded part where we see striated impact modifier. Figure 6 shows the cross-sectional view of the injection molded panel core. Here we see relatively small well disperse impact modifier particles. This type morphology is more effective at increasing toughness than that observed in Figure 4.
TEST METHOD DEVELOPMENT An attempt was made to reproduce the morphology observed on the inside surface of the blow molded part. Injection molded panels with a wall thickness of 6.4 mm were prepared. Modifications were made to a thermoforming unit that allow a solid metal sheet to be shuttled in and out of the thermoforming oven. Panels were placed onto the metal sheet and shuttled into the oven exposing the top of the panels to the IR heaters. The panels were exposed over a range of time intervals. At each interval, the impact performance was measured and the surface morphology studied. A transition from ductile to brittle failures occurred at an oven residence time of 80 seconds. Figure 7 shows the surface of a control panel that was not exposed to the heat source. These samples failed ductility. Figure 8 shows the surface of the panel after exposure to the heat source for 80 seconds. Here we see that while the sizes of the micro-voids are smaller than those in Figure 2, they are still large enough to cause brittle failure. This methodology became a screening tool for product development efforts aiming to increase the impact performance of blow molded parts.
The Influence of Morphology
Figure 7. Control panel surface.
163
Figure 8. Exposed panel surface.
SUMMARY The outside surface of the blow molded part is impacted during the impact event causing the inside part surface to experience high tensile forces. As a result of slow cooling of the inside surface, large spherulites develop along with micro-voids formed by the contraction of resin at the spherulite boundaries. Past research has found the spherulite boundaries to be weak sites prone to crack initiation and propagation. In addition, the slow cooling rate at the inside surface promotes impact modifier coalescence and leads to a further reduction in impact strength. These morphological features have been compared to those for an injection molded panel of same material and are believed to be the cause of the brittle part failures.
REFERENCES 1 2 3 4 5 6 7 8 9
Plastic Blow Molding Handbook, Edited by Norman Lee (1990). Keith, H. D., Padden, F. J., Spherulitic Crystallization from the Melt I. Fractionation and Impurity Segregation and Their Influence on Crystalline Morphology, J. Appl. Phys., 35, 1270 (1964). Varga, J., Supermolecular Structure of Isotactic Polypropylene, J. Mat. Sci., 27, 2557 (1992). Friedrich, K., Analysis of Crack Propagation in Isotactic Polypropylene with Different Morphology, Colloid & Polymer Sci., 64, 103 (1978). Greco, R., Ragosta, G., Isotactic Polypropylenes of Different Molecular Characteristics: Influence of Crystallization Conditions and Annealing on the Fracture Behavior, J. Mat. Sci., 23, 4171 (1988). Schultz, J., Microstructural Aspects of Failure in Semicrystalline Polymers, Poly. Eng. Sci., 24, 770 (1984). Botsis, J., Oerter, G., Friedrich, K., Fatigue Fracture in Polypropylene with Different Spherulitic Sizes, SPE ANTEC, 3294 (1996). Lustiger, A., Marzinsky, C., Mueller, R., Spherulite Boundary Strengthening: A New Concept in Polymer Blends, SPE ANTEC, 1506 (1998). Wu, S., A Generalized Criterion for Rubber Toughening: The Critical Matrix Ligament Thickness, J. Appl. Polym. Sci., 35, 549 (1988).
Morphological Study of Fatigue Induced Damage in Semi-crystalline Polymers
Nathan A. Jones and Alan J. Lesser Department of Polymer Science and Engineering, University of Massachusetts, MA 01003 USA
INTRODUCTION It is well established that the fatigue response of polymeric materials has two distinct regimes.1 At high frequencies and/or stress levels, the fatigue life of the polymer is dominated by hysteretic heating. As the stress level and/or test frequency is reduced, another failure mode is observed. This failure mode is more brittle and post-mortem fractographic analyses have shown the failure mechanism is associated with the nucleation and growth of flaws in the material to a critical size. This regime is referred to as the mechanically dominated regime or the high-cycle regime. The total lifetime in the mechanically dominated regime is usually attributed to the nucleation and growth of defects to a critical size. Crack growth kinetics have been measured on a wide range of polymers2 and have generally shown to follow a Paris-type law once the flaw grows beyond the threshold flaw size. For many polymers, this threshold size for fracture initiation is on the order of a millimeter.3 One the other end of the spectrum, the nucleation and growth of incipient cracks have been reported by Zhurkov, Kuksenko4-6 and Kausch.7 Zhurkov and Kuksenko reported results for the kinetics of the initiation of incipient submicrocracks in a number of polymers. They found that the initiation of damage occurred primarily by the nucleation of incipient submicrocracks. For the case of polypropylene, they reported the characteristic diameter to be on the order of 32-35 nm and a corresponding saturation density of 7x1014 cm-3. Moreover, the increase in number density to the saturation conditions followed a thermally activated process. Hence, they concluded that the initiation of damage in polymers is governed by a thermally activated process which describes the increase in number density of these submicrocracks. Consequently, the kinetics of damage evolution are described on the nanometer scale by a thermally activated process and on the millimeter scale by a Paris-type law. However, little quantitative work has been conducted describing the kinetics of damage evolution on
166
Plastics Failure Analysis and Prevention
the micrometer scale. This paper summarizes results from the first phase of an investigation to qualitatively and quantitatively describe fatigue damage and its evolution in a commonly used semi-crystalline polymer. For this study, we selected isotactic polypropylene (iPP) as a model polymer characteristic of many semi-crystalline thermoplastics in use today. Herein, we report results from an experimental investigation whereby in-situ measurements of dynamic visco-elastic behavior and energy density evolution are complimented by a comprehensive microscopic investigation to describe the kinetics and energetics of damage evolution on the micrometer scale.
EXPERIMENTAL Samples of isotactic polypropylene (iPP, Mn = 108,344, Mw = 330,871, MFR=0.7, supplied by Advanced Elastomers) were compression molded into plaques 2.8 mm thick using a hot press. The polymer was melted at 230°C, pressed to 3.5 MPa and cooled by flowing cold water through pipes in the hot press whilst maintaining the pressure at 3.5 MPa. The resulting plaque of iPP was cut into ASTM D638 Type II tensile bars using a specially designed cutting die. Fatigue tests were conducted at room temperature on an Instron Model 1321 servohydraulic machine in a load-controlled mode.1 The fatigue loads were applied with a sinusoidal waveform and a frequency of 2 Hz. The maximum and minimum stress levels were set at 25.3 MPa 2.2 MPa respectively and were maintained at a constant level for each fatigue test. Using the appropriate software and hardware, hysteresis loops of the crosshead load, axial strain, and transverse strain were stored digitally for pre-specified cycles through the duration of the test. Afterward, the hysteretic data was processed to calculate inelastic strain, dynamic viscoelastic parameters (E', E", and tan δ ), and energy densities for each stored cycle.1 Separate samples were fatigue damaged to different levels by systematically removing specimens after exposure between 105 and 106 cycles. These samples were subsequently used for microscopic examination of the fatigue damage. In preparation for microscopic examination, samples were faced off in a cryogenic microtome at a temperature of -100°C. Faces were cut both parallel to the stress direction. These blocks were etched in four consecutive etches consisting of: 0.35 g potassium permanganate, 10 ml sulfuric acid and 10 ml phosphoric acid.8,9 After each etch the samples were washed in 10 ml of a 2:7 mixture of concentrated sulfuric acid/distilled water, 10 ml hydrogen peroxide, 10 ml distilled water and 10 ml acetone. Faced and etched samples were coated in gold and viewed in a Jeol JSM-CF35 SEM operating at 20kV.
Morphological Study of Fatigue Induced Damage
167
These iPP samples were reasonably transparent, allowing TOM through the thickness of the whole tensile bars to be collected using an Olympus BH-2 optical microscope and a 35 mm film camera. Thermal characteristics were determined by DSC using a TA instruments DSC 2910 under nitrogen with a heating rate of 20°C/min. X-ray diffraction patterns, both WAXD and SAXD were obtained using various generators and a Siemens GADDS two-dimensional detector to record the data.
DISCUSSION OF RESULTS UNDAMAGED MORPHOLOGY
Figure 1. SEM micrograph of undamaged iPP.
An SEM micrograph of an undamaged sample that had been etched is shown in Figure 1. It shows lamellar crystals radially distributed in spherulites which have a diameter of about 70 µ m. WAXD showed the polymer to be 100% α -phase iPP10 with no discernible orientation. The SAXD showed an isotropic ring corresponding to a lamellar repeat distance of 14.2 nm. DSC measurements gave a melting point of 171.1±0.5°C and an melting endotherm of 97.1±3.6 J/g. Taking the enthalpy of fusion of iPP to be 209 J/g11 the degree of crystallinity is thus 46.4±1.7%. IN-SITU MEASUREMENTS OF i PP FATIGUE
Figure 2 shows the dynamic viscoelastic response of the iPP resulting from exposure to fatigue. In Figure 2, the loss modulus is represented by squares and the storage modulus is represented by the circles. The observation that can be made with respect to the data are that both the storage and loss moduli are decreasing throughout the fatigue life of the polymer. Another key point to notice is that two distinct regimes appear in rate of decay in the elastic (storFigure 2. Changes in loss modulus (squares) age) modulus. Below 103 cycles the rate of decay in and storage modulus (circles) during fatigue the elastic modulus is much greater that beyond this of iPP. point. This indicates that different processes may be operative during the fatigue process and are responsible for the fatigue softening. Another interesting feature of the data is that an apparent stiffening of the polymer is observed between 105 and 106 cycles.
168
Plastics Failure Analysis and Prevention
In Figure 3 the strain energy density is plotted with circles and the irreversible work data are plotted with squares. The strain energy continually increases as the fatigue process continues which is consistent with the fact that the material is becoming more compliant. Also, the irreversible work density of the material continually decreases over coincident with increasing cycles. This observation also supports the dynamic viscoelastic response of the material. Also, both of these energy densities show the similar change in their rates at approxiFigure 3. Changes in strain energy (circles) and irreversible work (squares) during fatigue. mately 103 cycles. Similar behavior are also noticed in the potential energy density evolution and the irreversible deformation. In these types of experiments the potential energy density curves usually follow closely to the strain energy density curves since they together reflect changes in internal energy density under the controlled load conditions. Although Figures 3 and 4 contain data for a single specimen, the reproducibility among different samples was very good. DAMAGE CHARACTERISTICS
Figure 4. SEM micrographs of damaged samples.
Visible Appearance of Fatigue Damage The procession of defects through the tensile bars caused the polymer to whiten and white lines with the appearance of shear bands where observed on some samples at an angle of 60° to 70° to the stress direction. The whitening was first observed between 1,000 and 10,000 cycles and the degree of whitening was observed to increase as the samples were further fatigued. It is well known that this stress whitening of polymeric samples is commonly caused by defects scattering visible light.4
Scanning Electron Microscopy Figure 4 shows SEM micrographs of microtomed and etched samples of a heavily fatigued iPP sample, which should be compared to Figure 1. On the undamaged sample no defects were seen, the small holes appearing in the sample are attributed to being etch artifacts. However the heavily fatigued sample (106 cycles, see Figure 4A) showed many defects of
Morphological Study of Fatigue Induced Damage
169
different sizes around the specimen ranging in diameter from 10 to 100 µ m and up to about 1 µ m thick. Many of the mature crazes appear to evolve from smaller crazes by coalescence. Note that the drawn fibrils between the defect surfaces (see Figure 4B) are clearly visible which confirms that fatigue damage evolves through craze formation and growth. It can be seen in Figure 4A that the crazes are initiated in and grow through both the bulk of the spherulites and the spherulite boundaries. Also there is no evidence to suggest that the craze growth is affected by the orientation of the crazes to the lamellae. Except for some kinks along the craze length (which are discussed below) the crazes grow perpendicular to the stress direction. Transmission Optical Microscopy TOM images showing the progression of fatigue damage are shown in Figure 5 for 105 to 106 cycles. Optical micrographs for 103 cycles or less show no crazes, whereas those of 105 (Figure 5A) cycles or greater show many crazes with dimensions of a few µ m or more. It can be seen qualitatively that as the number of fatigue cycles increases the average length of the crazes increases and that the number of individual crazes decreases. There are also a number of distinctive features about the arrangement of these crazes. Close inspection of Figure 5D shows a process zone of many small crazes surrounding the larger craze. This indicates that the growth of the larger crazes initiates additional crazes in the vicinity of the main craze tip. As the tip of the main craze grows past smaller craze the main craze shields the smaller crazes reducing the stress intensity around them. This slows the growth rate of the smaller crazes. This has the result that the main craze will be surrounded by many smaller crazes (commonly referred to as a process zone) all of about the same size. Crazes were often found to stack themselves in cascades, see Figure 5C. These cascades could be attributed to the appearance of the macroscopic white lines 60° to 70° to the stress direction on the sample. The arrangement of these crazes and their shapes are observed to change during the fatigue lifetime of the sample. Initially the mean craze length is much less than the mean crazeFigure 5. Optical micrographs craze nearest neighbor distance, see Figure 5A. Thus there are of iPP samples fatigued for different numbers of cycles. few craze-craze interactions, the distribution of crazes is random Scale bars represent 100 µ m.
170
Plastics Failure Analysis and Prevention
and most crazes have a circular disc shape. As the average craze length becomes comparable to the craze-craze nearest neighbor distance craze-craze interactions become important, see Figure 5C. Two or more crazes may merge to form one larger craze giving rise to large crazes with a characteristic zigzag appearance, see Figure 5D. Crazes formed from the merging of many crazes will have irregular shapes. Once the diameter of any one craze is large compared to the mean craze-craze distance it is unlikely that any one craze will exist without interacting with another craze. During fatigue craze initiation, propagation and coalescence occur simultaneously. As the number of cycles increases craze propagation and coalescence becomes the primary mechanisms of fatigue damage and the crazes become larger, acquire the kinks of characteristic of coalesced crazes and also become fewer. As the mass fraction of crazed sample was low, techniques such as WAXD and DSC were unable to determine any change in physical properties due to high cycle fatigue. The mature crazes detailed in this paper are often as large as 1 µ m by 100 µ m, well beyond the 200- 300 nm limit of SAXD. DAMAGE MECHANICS The previous section showed that fatigue damage in iPP manifests itself in the form of a regularly spaced ensemble of crazes. We start by approximating the stress intensity of a single craze in an infinite body as that due to a small crack in an infinite body subjected to a remotely applied stress superposed with a closure traction acting across the face of the crack. Hence the stress intensity for a single craze becomes K= σ ( πl ) 0.5 where σ = σ ∞ – σ c . The change in potential energy of the body due to a single craze can be calculated by: l
l
2 2
2 2 l∆Π -------------------- = 2 ∫ G dl = ------ ∫ K tot dl = πσ Eo Eo b 0
[1]
0
where b is the thickness of the specimen, ∆Π denotes the total change in potential energy of the body, G is the energy release rate, and Eo is the elastic modulus of the undamaged iPP. In order to calculate the change in potential energy due to an ensemble of crazes we must first consider the effect of their interaction on the modulus of the material. We start with the definition of a damage parameter (scalar damage parameter) given by Bristow12 for a 2-D case: 2 1 2 ρ = --- Σl i = nl A
[2]
where n is the number density of crazes, l is the craze radius, and a A is the size area of a representative volume of material (i.e., representative volume = b*A).
Morphological Study of Fatigue Induced Damage
171
In 1993, Mark Kachanov,13 showed that the following nonlinear approximation actually remains highly accurate at high craze densities provided the cracks are homogeneously distributed. This is given by: 1 - = ------------------1 E- = ------------------2 1 + πρ Eo 1 + πnl
[3]
In the above equation E denotes the effective modulus and Eo denotes the initial modulus of the iPP. Using a similar approach as above to calculate the change in potential energy per unit of representative volume (i.e., potential energy density) yields the following expression for an ensemble of crazes: lK 2 2 2 4 σ π 2 tot ∆Π -------- = n ⋅ 2 ∫ ------------ dl = --------- ( nl + 0.5πnl ) E Eo V 0
[4]
Finally expressing equation 4 in terms of the Griffiths energy for a single craze (i.e. G=2 γ ), the following expression can be arrived at: 2 3 ∆Π -------- = γ ( 2πl + πn l ) V
[5]
Hence, the above equation describes the change in potential energy density of an ensemble of crazes in terms of the specific energy of craze formation, the number density of crazes, and the craze radius. QUANTITATIVE CHARACTERIZATION OF FATIGUE DAMAGE In the previous section we used the definition of a damage parameter ρ defined by Bristow. Image analysis was conducted on the images obtained for the TOM’s, using Zeiss Image v.1.0, to measure an effective number density (# of crazes/m2) and the average craze radius. In Figure 6 the craze radius is represented by squares and the craze density by circles. This Figure quantitatively shows what was qualitatively claimed above. With increasing numFigure 6. Evolution of fatigue damage between bers of cycles the average craze radius increased 105 and 106 cycles. Circles represent craze however the number of crazes decreased. As we number density and squares represent craze have already discussed above this is due to craze radius. coalescence. The combined effect of the changes to craze radius and number density in an increase in the amount of fatigue damage. From the measured potential energy density evolution, the damage evolution is given by equation 6
172
Plastics Failure Analysis and Prevention
and was calculated to average approximately 13 J/m2 and to be constant over 900,000 cycles. ∆Π γ = -------------------------------------3 V ( 2nl + πn2l )
[6]
REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13
Lesser A.J. (1995) J. Polym. Sci., Part B, Polym. Phys. Ed., 58, 869-879. Hertzberg R.W. (1980) Fatigue of Engineering Plastics, Academic Press, New York pp 61-67. Williams J.G. (1984) Fracture Mechanics of Polymers, John Wiley and Sons, pp 124-130. Zhurkov S.N., Kuksenko V.S., Slutsker A.I. (1969) Soviet Physics - Solid State, 11, 2, 238. Zhurkov S.N., Marikhin V.A., Slutsker A.I. (1959) Soviet Physics - Solid State, 1, 1060. Kuksenko V.S., Slutsher A.I. and Yasttrebinskii A.A. (1968) Soviet Physics - Solid State, 9, 8, 1869. Kausch (1978) Polymer Fracture, Polymers/ Properties and Applications, Springer Verlag. Patrick M. Bennet V., Hill M.J. (1996) Polymer, 37, 24, 5335-5341. Naylon K.L., Phillips P.J. (1983) J. Polym. Sci., Part B, Polym. Phys. Ed., 21, 2001-2026. Turner-Jones A., Aizlewood J. M., Beckett D.R. (1964) Makromol. Chem., 75, 134. Quirk R. P., Alsamarraie M.A.A. (1989) Polymer Handbook Third Ed., Wiley Interscience, New York, Ed. Brandrup J., Immergut E.H. V/29. Bristow J.R. (1960) J. Appl. Phys., 11, 81-85. Kachanov M. (1993) Advances In Applied Mechanics, Vol. 30, Hutchinson J. Wu T., eds., Academic Press, pp 259-445.
Chapter 5 Modelling of Failures and Failure Processes Failure Analysis Models for Polyacetal Molded Fittings in Plumbing Systems
L.J. Broutman, D.B. Edwards, and P.K. So L.J Broutman & Associates, Ltd., Chicago, IL, USA
INTRODUCTION Residential hot/cold water plumbing systems made completely of plastic were introduced in the mid 1970’s. These systems consisted of polybutylene piping and polyacetal (Celcon and Delrin) fittings. The manner of joining was to crimp either an aluminum or copper crimp ring onto the tubing with a specially designed crimp tool so as to effect a seal between the barbs on the fitting and the polybutylene pipe. These systems, as well as the constituents, eventually became covered by ASTM standards such as D3309 and F845. In addition, polybutylene was able to achieve a rating of 0.7 MPa (100 psi) water pressure at 180ºF through the Plastic Pipe Institute (PPI). Celcon was able to achieve a pressure rating only at room temperature. This plumbing system was installed throughout the U.S. from the mid-70’s until 1986 when Celcon was withdrawn from site-built homes and replaced with copper or brass fittings. Celcon was still utilized in factory-built construction until the early 1990’s. The components were manufactured by at least 5 different companies and sold under their individual trade names. The pipe was extruded to common ASTM size specifications and the fittings and crimp ring dimensions were also ASTM specified. In a matter of only 3 or 4 years, numerous leaks occurred, partly due to poor installation practices. However, this did not account for all of the observed leaks. It became clear that material degradation by oxidation, particularly of the polyacetal fittings, was also occurring and in the early 1980’s, both the polyacetal and polybutylene were no longer recommended for hot water recirculating plumbing systems. In these systems, the materials on the hot water side are exposed 24 hours a day to elevated temperatures.
174
Plastics Failure Analysis and Prevention
Unfortunately, with time, the leak rate substantially increased and significant litigation between homeowners, developers, contractors, manufacturers and material suppliers ensued. Individual lawsuits and class-action suits became common in the 1980’s and Figure 1. An axisymmetric model showing the FEA mesh and boundary conditions. 1990’s. Eventually a plumbing claims organization was established to replumb homes at no cost to the homeowners and this is still in existence. This failure of a plastic plumbing system has probably been the most costly litigation involving polymeric materials, and ranks with other costly litigation such as the silicone breast implant litigation.
POLYACETAL FITTINGS The polyacetal fitting is represented in Figure 1, which is a finite element model used to calculate strains in the fitting due to crimping.1 The crimping process reduces the diameter of the crimp ring, which causes bending of the fitting. The bending strain in the axial direction is dependent upon the degree of crimping, and the axial position of the crimp ring. In addition to the axial bending strain, the fitting is subjected to a circumferential compressive strain. The common sizes of this system are 3 mm (½ inch) and 6 mm (¾ inch). The fitting strains are somewhat higher in the 3 mm (½ inch) fitting than in the 6 mm (¾ inch) fitting. The fitting strains are principally caused by the crimping process. The strains or stresses due to water pressure are not substantial in the area of the crimp ring because of the reinforcement provided by the crimp ring. The maximum tensile strains created by the crimping process occur on the tensile wall of the fitting and range from approximately 1.5% to 2.5% for allowable crimp ring dimensions and placements.1 Since the dimensions of the crimped system remains approximately constant, the stresses in the fitting relax with time and temperature. Relaxed stress values calculated from Celcon stress relaxation data are approximately 10 MPa to 14 Mpa (1500 psi to 2000 psi) after 10 years of service and exposure to a temperature of 40ºC.
SURFACE EMBRITTLEMENT OF POLYACETALS When polyacetals such as Celcon are exposed to chlorinated water (hypochlorite ions), the surface is chemically degraded, causing substantial reductions in molecular weight. Eventually the degraded layer whitens due to microvoid formation. This degraded surface becomes embrittled and allows cracks to initiate even at extremely low applied strains or stresses. The tensile creep or stress rupture characteristics of Celcon M25-04 injection molded bars
Failure Analysis Models
175
(thickness = 1 mm) are shown in Table 1. The failure time is a function of stress, chlorine concentration and temperature. The surface embrittlement process allows cracks to initiate in short time periods, thus reducing lifetimes when compared to testing in a non-oxidizing environment.
Table 1. Tensile creep rupture strengths of Celcon M25-04 in chlorinated water Stress, MPa (psi)
Chlorine concentration, ppm
Temp., ºC
Failure Time*, h
10 (1500)
0.5
70
1125
14 (2000)
0.5
70
167
14 (2000)
2
80
36
14 (2000)
2
70
122
14 (2000)
2
60
367
* Averages of 7 specimens
FAILURE MODEL FOR CELCON FITTINGS A failure model has been developed for Celcon fittings based on inspection of hundreds of actual fittings removed from service from various locations around the U.S. In addition to visual and microscopic analysis, other factors such as the fitting strain analysis and Celcon creep rupture characteristics were considered in the development of the model. Figure 2 shows the five stages of the failure model. Stage 1 involves degradation and embrittlement of the exposed inner surface of the fitting. The rate of degradation is so great that the process is insensitive to strain. Thus, the depth of the degraded or whitened layer is relatively uniform and independent of the strain level. Stage 2 involves continued embrittlement such that microcracks form at 90º to the principal stress direction and propagate from the inside surface through the whitened layer. A critical stage is Stage 3, where microcracks in the whitened layer begin to extend beyond the uniform degraded layer. This process occurs as a result of the hypochlorite ion/ water solution flowing into the microcracks and diffusing into the Celcon around the crack tip. Degradation of the Celcon in advance of the crack tip embrittles the material, allowing the crack to propagate further into the now brittle zone at the crack tip. The transition from Stage 2 to Stage 3 can occur even when the applied strain is near zero. This is confirmed by the observation of Stage 3 cracks even in areas of the fitting where the strain was less than 0.1% as calculated by finite element analysis.
176
Plastics Failure Analysis and Prevention
Figure 2. Five stages of failure for celcon fittings.
Continued crack growth occurs in Stage 4. Final rupture occurs in Stage 5 when the cracks reach the outer surface of the fitting. The cracks (Figure 2) all have a whitened degraded layer on all sides and triangular whitened areas on cut cross-sections, which contain a crack at their center. Crack growth in Stage 4 is a strong function of the applied stress, chlorine concentration and temperature.2
Failure Analysis Models
177
The anticipated lifetimes of Celcon fittings can be estimated from tensile creep rupture data as shown in Table 1. In addition, one needs to know the relaxed stress, exposure temperature and chlorine levels in the water. For example, based on a relaxed stress of 10 MPa (1500 psi), T=70ºC and 0-5 ppm Cl, a creep rupture time of 1125 hours was determined. If one assumes exposure to 70ºC for only one hour per day, the approximate lifetime would be 20,000 hours, or 2½ years. At lower, more realistic temperatures experienced by residential plumbing systems, lifetimes have been predicted between 5-15 years. These are in close agreement with actual field experience.
CONCLUSIONS A failure model for polyacetal (Celcon) fittings has been developed and applied with good success to predict lifetimes of fittings in service. Of importance is the significant effect of chlorinated water on reducing the lifetime of Celcon fittings.
REFERENCES 1 2
Y. Chen, B.D. Agarwal and L.J. Broutman, “Finite Element Analysis of Plastic Plumbing Assemblies,” SPE ANTEC ’99. In Seok Oh, Paul K. So and L.J. Broutman, “Crack Growth Studies for Polyacetal Resins in Chlorinated Water,” SPE ANTEC ’99.
Progressive Failure Analysis of Fiber Composite Structures
Matt H. Triplett U. S. Army Aviation and Missile Command, Redstone Arsenal, AL, USA
INTRODUCTION Fiber composite structures are used extensively in aerospace, automotive and sports equipment applications. Understanding the stress-strain behavior of fiber composite laminates to ultimate failure and the ability to predict ultimate strength is critical in the design of safe and lightweight structures. Each laminate will behave differently when loaded from first ply failure to ultimate failure. In order to assess the total stress-strain behavior of a given laminate, a failure criteria to determine when a lamina, i.e. ply, has failed is required. The failure of the first ply is commonly referred to as first ply failure. Following the failure of a ply, an appropriate material degradation criteria is required. The analysis of successive ply failures and material degradation, i.e., stiffness reduction beyond first ply failure is referred to as post-failure analysis or progressive failure analysis. Progressive failure analysis is performed until the laminate reaches ultimate failure. Many criteria exist for the prediction of lamina failure, for example: Maximum Stress, Maximum Strain, Tsai-Hill,1 Tsai-Wu,2 Hashin-Rotem,3 and Hashin.4 There are other failure criteria, many of which have been reviewed and compared to each other and experiments.5-7 Numerous studies address material degradation and progressive failure analysis of fiber reinforced composites.6-15 Most of these studies7-15 utilize combinations or modifications of the Maximum Stress, Hashin, or Tsai-Wu lamina failure criteria. Sun, et al.,6 utilized the six failure criteria listed above and two different stiffness reduction methods and compared the predicted ultimate strength to experiments of various laminates. The predictions were obtained using an iterative computer program to simulate the tensile experiments. These various laminates were tested in tension at varying off axis load angles. The results show that the Maximum Stress, Maximum Strain, and HashinRotem failure criteria predict the ultimate strength better than the Tsai-Hill, Tsai-Wu, or Hashin criteria. In addition, the ultimate strength predicted using the Tsai-Hill, Tsai-Wu, and Hashin criteria do not follow the general trend of the experimental data over the range of load angles tested for three of the four laminates. The Hashin-Rotem criteria performed
180
Plastics Failure Analysis and Prevention
slightly better than the Maximum Stress or Strain criteria and has not been implemented into a finite element analysis, therefore the Hashin-Rotem criterion is chosen for this study. Progressive failure studies6-16 utilize numerous methods to degrade lamina properties. Some8,11,12,15 degrade the properties as a function of some experimental or estimated coefficient. Most6,7,9,10,13,14 simply reduce the stiffness to zero based on the mode of failure. The latter method will be employed in this study. Progressive failure analysis studies have typically been performed using either a special purpose code, or a complex, specially developed element or material model subroutine for commercial finite element codes. The study presented here utilizes a simple solution dependent variable subroutine and the ABAQUS finite element code to perform progressive failure analysis.
LAMINA FAILURE CRITERIA The Hashin-Rotem lamina failure criterion is summarized below. For fiber failure: σ 11 -------- = 1 X
[1]
where σ 11 is the fiber direction stress and X is the fiber direction strength, either tensile or compressive. For matrix failure: 2 σ 12 2 22 σ ------------ Y + S = 1
[2]
where σ22, σ 12 are the transverse direction stress and in plane shear stress respectively, Y is the transverse direction strength, either tensile or compressive, and S is the in plane shear strength.
STIFFNESS REDUCTION The material properties required to define the unidirectional behavior of a lamina including transverse shear are as follows; Fiber direction modulus E11 Transverse direction modulus E22 In plane modulus G12 Transverse shear moduli G13, G23 In plane Poisson Ratio ν 12 As previously stated, many studies simply reduce the material properties to zero based on the failure mode but there are numerous variations of this method. For example, matrix stiffness reduction can be separated such that E22 and G12 are reduced separately based on the assumption that transverse matrix failure does not necessarily inhibit the ability to carry
Progressive Failure Analysis
181
shear load.6 The stiffness reduction method employed here is similar to that of Ganapathy, et al.13 and is shown below. For fiber failure based on equation [1] all properties are reduced. E 11 = E 22 = G 12 = G13 = G 23 = ν 12 = 0 [3] For matrix failure, only the matrix dependent properties are reduced. E 22 = G 12 = G13 = G 23 = ν 12 = 0 [4] Actually, for numerical stability, ABAQUS requires that the value of the moduli be greater than zero. Therefore that value is reduced to a very small number, in this case one.
INCREMENTAL ANALYSIS PROCEDURE In order to perform a progressive failure analysis an incremental load analysis is required. The laminated shell element properties available in ABAQUS allow for multiple layers and multiple integration points in each layer. For each load increment the stresses at each integration point are passed to a user subroutine that checks for the failure modes in equations 1 and 2. Based on the failure modes, the properties are reduced according to the criteria in equations 3 and 4. This procedure is continued until ultimate failure of the laminate is reached.
COMPARISON TO EXPERIMENTS 6
Sun, et al. performed tensile experiments at various off axis load angles on numerous laminates fabricated with AS4/3501-6 carbon epoxy unidirectional plies. Table 1 shows the laminates and off axis angle tested. Table 2 shows the material properties of the AS4/3501-6 plies. A one element finite element model has been used previously to study tensile loading of composites.7 This method will also be used for this analysis. Table 1. Laminate and off Axis Loading Angles6 Laminate Off
Axis Angle
[0/45/-45/90]s
0°-22.5° every 7.5°
[90/30/-30]s
0°-22.5° every 7.5°
[0/45/-45]s
0°-30° every 7.5°,26°,45°
[90/0/90/0]s
0°-7.5° every 1.5°,15°,22.5°
182
Plastics Failure Analysis and Prevention
Table 2. Properties of AS4/3501-6 Plies6 E11
153.7 GPa
Xcompression
-2013.0 MPa
E22
11.0 GPa
Ytension
67.0 MPa
G12=G13*=G23*
6.9 GPa
Xcompression
-206.8 MPa
ν 12
0.32
S
110.3 MPa
Xtension
2171.0 MPa
Ply Thickness
0.13 mm
*Not included in reference 6, assumed for this study
Table 3. Comparison of Experiments and Analysis Ultimate Strength, MPa Laminate and Off Axis Angle
Experiment6
Analysis
[0/45/-45/90]s 0°
765
721
7.5°
752
731
15°
774
788
22.5°
832
885
[90/30/-30]s 0°
966
1064
7.5°
908
885
15°
837
782
22.5°
807
731
[0/45/-45]s 0°
883
756
7.5°
843
769
Progressive Failure Analysis
183
Table 3. Comparison of Experiments and Analysis Ultimate Strength, MPa Laminate and Off Axis Angle
Experiment6
Analysis
15°
929
859
22.5°
1028
1026
26°
1129
1167
30°
1074
1026
45°
818
744
[90/0/90/0]s 0°
1126
1073
1.5°
1140
1038
3°
1074
808
4.5°
1018
600
6°
861
473
7.5°
713
452
15°
394
346
22.5°
288
288
Figure 1. Comparison of ultimate strength for [0/45/-45/90]s laminate
Figure 2. Comparison of ultimate strength for [90/ 30/-30]s laminate
184
Plastics Failure Analysis and Prevention
Figure 3. Comparison of ultimate strength for [0/45/-45]s laminate
Figure 4. Comparison of ultimate strength for [90/0/90/0]s laminate
The comparison of the finite element progressive failure analysis to the experiments are shown in Table 3 and Figures 1 through 4. The agreement is very good for the laminates shown in Figures 1 through 3. For the [90/0/90/0]s laminate in Figure 4, the agreement is not as good between 3° and 7.5° load angles. This is true for all of the failure criteria studied by Sun, et al.6
CONCLUSIONS The results presented here provide a simple method to perform progressive failure analysis of complex composite structures utilizing finite element analysis. This technique can be especially useful for the analysis of impact damage and residual strength studies of fiber composite structures. The method can also be used to reduce the time and cost of designing fiber composite structures by reducing the number of fabricate and test iterations required to obtain an optimum weight and strength design. Future studies should focus on comparison to biaxially loaded laminates and laminates loaded in compression.
REFERENCES 1. 2. 3. 4. 5. 6.
Tsai, S. W., “Strength Theories of Filamentary Structures,” Fundamental Aspects of Fiber Reinforced Plastic Composites, R. T. Schwartz and H. S. Schwartz, Eds., Wiley Interscience, New York, 1968. Tsai, S. W., and Wu, E. M., “A General Theory of Strength for Anisotropic Materials,” J. Composite Mater., Vol. 5, 1971. Hashin, Z. and Rotem A., “A Fatigue Failure Criteria for Fiber Reinforced Materials,” J. Composite Mater., Vol. 7, 1973. Hashin, Z., “Failure Criteria for Unidirectional Fiber Composites” J. Appl. Mechanics, Vol. 47, 1980. Nahas, M. N., “Survey of Failure and Post-Failure Theories of Laminated Fiber-Reinforced Composites,” J. Composites Technol. Res., Vol. 8, 1986. Sun, C. T., Quinn, B. J., Tao, J., and Oplinger, D. W., “Comparative Evaluation of Failure Analysis Methods for Composite Laminates” DOT/FAA/AR-95/109, 1996.
Progressive Failure Analysis
7. 8. 9. 10. 11. 12. 13. 14. 15.
185
Ochoa, O. O. and Engblom, J. J., “Analysis of Progressive Failure in Composites,” Composites Sci. Technol., Vol. 28, 1987. Chang, F. K. and Chang, K. Y., "A Progressive Damage Model for Laminated Composites Containing Stress Concentrations," J. Composite Mater., Vol. 21, 1987. Yener, M. and Wolcott, E., "Damage Assessment Analysis of Composite Pressure Vessels Subjected to Random Impact Loading," J. Pressure Vessel Technol., Vol. 111, 1989. Chang, F. K. and Lessard, L. B., "Damage Tolerance of Laminated Composites Containing an Open Hole and Subjected to Compressive Loadings: Part I-Analysis," J. Composite Mater., Vol. 25, 1991. Reddy, Y. S. and Reddy, J. N., "Three-Dimensional Finite Element Progressive Failure Analysis of Composite Laminates Under Axial Extension”, J. Composites Technol. Res., Vol. 15, 1993. Shahid, I., and Chang, F. K., "An Accumulative Damage Model for Tensile and Shear Failures of Laminated Composite Plates," J. Composite Mater., Vol. 29, 1995. Ganapathy, S., Tripathy, B., and Rao, K. P., "Damage and its Growth in Laminated Composite Circular/Rectangular Plates Undergoing Large Deformation," Composite Structures, Vol. 32, 1995. Eason, T. G. and Ochoa, O. O., "Modeling Progressive Damage in Composites: A Shear Deformable Element for ABAQUS®," Composite Structures, Vol. 34, 1996. Yen, C. F., Cassin, T., Patterson, J., and Triplett, M., "Progressive Failure Analysis of Thin Walled Composite Tubers Under Low Velocity Impact, 39th AIAA SDM Conf., Long Beach, CA, 1998.
Calculating Thermally Induced Stresses Using a Nonlinear Viscoelastic Material Model
N. Schoeche and E. Schmachtenberg Institute for Plastics in Mechanical Engineering, Altendorfer Str. 39, Essen, Germany
INTRODUCTION Plastics have a high thermal expansion coefficient. Thus, if the dimensional change is obstructed stresses are induced. During a cooling process these are tensile stresses. Because of the temperature and time depending behavior of plastics the course and the value of the tensile stress depends on the temperature history, the load history and the material itself. On the one hand thermal induced tensile stresses represent an additional load in the material which can be the cause for unexpected failure. On the other hand stresses relax upon time and temperature. The higher the temperature the higher the relaxation potential. Therefore, the analytical calculation of thermally induced stresses is extremely difficult. Numerical models open the possibility to predict thermally induced stresses in order to know if the thermal history of plastic part is relevant for possible unexpected failure.
MODELING OF THE MATERIAL BEHAVIOR To be able to consider the temperature influence as well as the time depending behavior an extended spring-dashpot model has been applied as it is described by Schmachtenberg, et al.1-5 In difference, some modifications concerning the temperature related description of the material behavior have been made. This includes the viscous behavior and the thermal expansion coefficient which is also depending on the temperature. Figure 1. Extended spring-dashpot model. The extended spring-dashpot model is shown in Figure 1. The following model's properties are to be given: • number of spring-dashpot elements, • Young's modulus for each element,
188
Plastics Failure Analysis and Prevention
•
description of the temperature depending viscous behavior (“flow lines”) for each element. The viscous behavior is represented by the dashpot. The nonlinear viscous behavior is described by “flow lines” which are the “flow stress” of each element over the temperature for a given flow rate. The model's parameters are derived from simple tensile tests which are performed for difFigure 2. Flow stress per element vs. temperature (PEferent temperatures. With this “fingerprint” of HD). the material the calibration is possible. First of all, the number of elements n is determined. The wider the range of temperature to be calculated the more elements are necessary. Then, the Young's modulus of each element is determined. The sum of all moduli must be equal to the Young's modulus of the material for the lowest temperature E0,max. The distribution of the moduli over the elements has an influence on the simulation's accuracy. In the simplest distribution all Young's moduli Ei are equal and can be determined by E0, max E i = ---------------n
[1]
Using the stress-strain curves it is possible to calculate the “flow lines” now. If the model can be adjusted appropriately to the measured curves the information about time and temperature dependency is available in the model. A link between time and temperature is necessary which is done by using the Arrhenius function k --- – -------- · T T ref εref = ------10 · ε 1
1
[2]
The flow lines in Figure 2 are calculated in a way that the simulated stress-strain curves fit with the measured ones. A detailed description is given elsewhere.6,7 Increasing the external strain by ∆ε with an actual strain rate, ε· ref leads to an increase in the element's stress [3] ∆σ i = Ei ∆ε In the first step it is assumed that all elements behave elastic. Knowing the element's stress the viscous behavior can be calculated using the “flow lines”. For the actual stress in the element the temperature where the element starts to flow is taken from the “flow lines”. As this temperature is not necessarily equal to the actual temperature ϑref in the element equation 2 is used to calculate the actual strain rate of the dashpot. Because the numerical
Calculating Thermally Induced Stresses
189
methods enables to calculate in time steps the elongation ∆ε p of the dashpot can be calculated from this strain rate. This means that the change in the element's stress is not as assumed in equation 3 but ∆σ i = Ei ( ∆ε – ∆ε p ) [4] The stress at the actual time t of each element is σ i ( t ) = σ i ( t – 1 ) + ∆σ i [5] The sum of all element's stresses σ i equals the total stress in the material σ =
∑ σi
[6]
i
•
The thermal and mechanical behavior of the material itself is described as follows: the measured stress-strain curves are approximated by 1 – D1 ε σ = E0 ε ------------------1 + D2 ε
[7]
Therefore, the material's parameters E0, D1 and D2 are temperature dependent. The thermal expansion coefficient α th is given as a function of temperature. For the time-temperature shift the value k is set constant. The influence of the temperature dependence on the k value is small enough to be neglected. With this information and after calibrating the model with stress-strain curves it is possible to predict stress-time curves for any possible temperature and strain history. As a restriction only temperatures in the temperature range measured are eligible.
• •
MEASUREMENT OF THERMAL INDUCED STRESSES IN A GEOMEMBRANE For the evaluation of the model it is necessary to compare simulation results with measurements. Until now, uniaxial thermal stresses have been measured. Thus, the model is limited to uniaxial loads. An universal testing machine is used where strain rate, strain or force can be hold constant during the measurements. To obtain data about the thermal expansion the force has been held constant at zero Newton while the specimen is heated. For thermal cycling tests the strain has been held constant in order to measure the time and temperature depending stress. Specimens of the size 4 mm x 15 mm x 120 mm are clamped inside the temperature chamber of the testing machine as shown in Figure 3. The upper clamp can be moved by the traverse. The force is measured with a load cell. The strain of the specimen is measured by an mechanical extensometer near the clamps. This is necessary because the thermal expansion of the extension pieces will influence the expansion of the specimen otherwise. By measuring the strain on the specimen it is assured that the extension pieces have no influ-
190
Plastics Failure Analysis and Prevention
Figure 4. “Fingerprint” of the PE-HD material.
ence on the specimen's mechanical behavior. Due to the thermal properties of plastics and the dimensions of the specimens the temperature in the chamber is different to the temperature in the specimen during heating or cooling. ThereFigure 3. Measurement device (tensile testing machine). fore, the temperature in the material is measured in a second, shorter specimen which is clamped into one clamp next to the other specimen. As an example the response of a PE-HD geomembrane for two load cases is investigated. Load case 1: The geomembrane has a strain of 2% (e.g., due to installation) at 30°C. Then, it cools down within 30 minutes to 0°C (night). This temperature is held for 30 minutes to give time for relaxation. Within 60 minutes the temperature is increased to 60°C (sunny day) and held for 30 minutes. The cycle continues with cooling down to 0°C within 60 minutes and repeating the cycle. Load case 2: The temperature profile of load case 1 is applied but without a prestrain. Thus, compression occurs and due to high relaxation of the compression stress it is assumed that relative high tensile stresses are induced. Both load cases have been observed for 23 h which simulates a period of 7 days.
COMPUTER SIMULATIONS Figure 4 shows the stress-strain curves („fingerprint") of the PE-HD material used. The thermal coefficient has been determined to be linear with the temperature α ( ϑ ) = ( 84 + 2ϑ ) × 10
–6
[8] 1/K The time-temperature shift coefficient k is set to 10000 K constantly. The model has 20 spring-dashpot elements with a Young's modulus of 80 MPa in each element. It has been calibrated between 0°C and 80°C as shown in Figure 4. The “flow lines” resulting from these parameters are presented in Figure 2.
Calculating Thermally Induced Stresses
Figure 5. Stress due to thermal cycling in a geomembrane with 2% strain.
191
Figure 6. Stress due to thermal cycling in a geomembrane without strain.
The load cases are input by a list of time steps with the temperature and strain applied at this time. The software divides each time step in smaller steps using linear interpolation and iterates to a convergent solution.
INTERPRETATION The result for load case 1 is shown in Figure 5. The dotted stress line represents the measured curve. It is almost identical to the simulated line. Due to the strain of 2% the stress starts at a value of 11 MPa and relaxes. The cooling process shrinks the material at the same time. The simulation predicts the point when the increasing stress grows faster than the relaxation very well. Also, the peaks of high and low temperature are predicted very well. For long term behavior the maximum stress is near 7 MPa and the minimum stress near 2 MPa. The results for load case 2 are presented in Figure 6. The simulation is capable to show that the stress level increases significantly. Compression stress which is induced by high temperatures relaxes fast while low temperatures induce tensile stress increase up to 4 MPa. The maximum stress levels tend to increase with time. It can be assumed that this load case leads to a reduced life time because of an increasing tensile stress in the geomembrane. Deviations between measured and simulated curves are caused by environmental conditions (e.g., support of the specimen to prevent buckling in load case 2) which cannot be simulated by the model.
CONCLUSIONS For uniaxial load and moderate changes in the thermal boundary conditions the model and the model parameters presented are capable of a very good prediction of the stress curves. With this method it is possible to determine critical strain and thermal load cases. Also, this tool can be helpful to estimate the life time of plastic parts which are used under changing
192
Plastics Failure Analysis and Prevention
thermal conditions. The results reveal that even without prestrain a significant tensile stress can be induced in a geomembrane. Improvement will be done to be able to calculate three-dimensional stress conditions and to deal with fast temperature changes.
ACKNOWLEDGMENT This project is funded by the German Research Foundation (DFG). We are also thankful to Borealis GmbH, Germany, for supporting us with material.
REFERENCES 1 2 3 4 5 6 7
Schmachtenberg, E. (1985), „Mechanical Properties of Nonlinear Viscoelastic Materials", PhD-thesis at RWTH Aachen Michaeli, W., Fölster, Th., Lewen, B. (1989), „Beschreibung des nichtlinear-viskoelastischen Verhaltens mit dem Deformationsmodell", Kunststoffe 79 (1989) 12, pp. 1356-1358 Michaeli, W, Mohr-Matuschek, U., Lewen, B, Fölster, Th. (1990), „Kunststoffgerechtes Konstruieren, Kunststoffe 80 (1990) 3, pp. 352-355 Menges, G., Wenig, M., Fölster, Th. (1990), „Deformation Behavior of Thermoplastics for Non-Uniform Stress Distributions", Kunststoffe German Plastics 80 (1990) 9, pp. 39-40 Partom, Y., Schanin, I. (1983), „Modeling Nonlinear Viscoelastic Response", Polym. Eng. Sci., October 1983 Vol. 22 Schmachtenberg, E., Schöche, N. (1996), „Modeling of non-linear viscoelastic material behavior”, conference proceedings of MATEH '96 in Opatija, Croatia Schmachtenberg, E. (1996), „Entwicklung eines Werkstoffmodells zur Beschreibung thermisch induzierter Eigenspannungen für viskoelastische Werkstoffe", DFG-research report, University of Essen
Evaluation of a Yield Criteria and Energy Absorbing Mechanisms of Rubber Modified Epoxies in Multiaxial Stress States
Robert S. Kody and Alan J. Lesser Polymer Science and Engineering Department, University of Massachusetts, Amherst, Massachusetts 01003, USA
INTRODUCTION It is well established that in many rubber-modified systems, a primary toughening mechanism is the relief of hydrostatic stress through rubber particle cavitation or disbonding, followed by inelastic void growth in the matrix material.1,2 This process has been reported in rubber toughened epoxies,3,4 showing that both rubber particle cavitation and inelastic void growth are primary toughening mechanisms. Of these two mechanisms, rubber particle cavitation has received the most attention in the literature. Two parameters reported to control the effectiveness of cavitation are particle size and modifier surface energy.2,5,6 Earlier studies have shown, that rubber particles exhibit an increased resistance to cavitation as particle size is decreased. Recently, this sizeeffect on the cavitation resistance has been modeled by Lazzeri and Bucknall,5 and Dompas and Groeninckx.2 Both criteria are based on energy balance principles and consider that the energy available to produce cavitation is the volumetric strain energy stored in the rubber particle, Uo, given by: 2 2 3 U 0 = --- πR K r ε v 3
[1]
where R is the initial particle radius, Kr is the bulk modulus of the rubber, and ε v = εii is volumetric strain. Uc is the energy barrier that must be overcome by the available strain energy, and for Bucknall’s model is given by eqn. [2], and for Groeninckx’s model by eqn. [3]. 2
3 3 2 r - + 4πr 2 Γ + 2πr 3 G ρF ( λ ) U c = --- πKr R εv – ---- r f 3 3 R 2
U c = 4πr Γ
[2] [3]
194
Plastics Failure Analysis and Prevention
where r is the radius of the cavitated void, Γ is the surface energy of the rubber, Gr is the shear modulus of the rubber, ρ represents the density ratio of the rubber before and after cavitation taken equal to 1, and F( λ r) is a numerical integration of the shear strain function of the rubber after cavitation, taken to be equal to 1. In both models, the necessary condition for particle cavitation occurs when: Uc ≤ U0 [4] Each models yields a “scale effect” for rubber particle cavitation where the larger particles cavitate at lower volume strains than the smaller particles. The primary difference between the two models being that Bucknall includes the shear energy and residual volumetric energy stored in the rubber particle after cavitation. For rubber particles with Gr=0.4 MPa, Kr=2 GPa, and Γ =0.03 Jm-2, both models are plotted in Figure 1, showing the smallest cavitated rubber particle for a given ε v. Again, both models predict the observed trend; The largest rubber particles cavitate at the lowest volumetric strains.2,6 Figure 1. Minimum particle radius that will cavitate versus volumetric strain, for rubber particles with Kr=2 GPa, Additional studies have shown that toughGr=0.4 MPa, and Γ =0.03Jm-2. ness increases with decreasing particle size down to a minimum critical size, below which the particles do not cavitate and little toughening is realized. However, if particle cavitation does occur, the higher toughness measured for the smaller particles is attributed to the smaller inter-particle distance, which describes the size of a ligament between voids. The effect of inter-particle distance has been discussed by others7,8 and is closely related to inelastic deformation of the matrix material. After particle cavitation occurs, it is well accepted that the majority of the energy dissipation comes from irreversible deformation of the matrix material. This deformation usually takes the form of shear bands and inelastic void growth. Unfortunately, models that predict their contribution have been scarce in the literature. Until recently,5,9 the literature has been absent of any models that allow for a prediction of the full yield response of rubber-modified polymers in arbitrary stress states, whereby cavitation and inelastic void growth are considered. In 1993 Lazzeri and Bucknall5 made a first attempt at introducing a yield criterion for rubber-modified polymers. Their model considers that all particles cavitate and act as pores in the matrix material. Next, they employed a theory developed by Gurson10 for a perfectly
Evaluation of a Yield Criteria
195
plastic media containing pores. Lazzeri and Bucknall modified Gurson’s model, which is based on a von Mises criterion, to include the effects of hydrostatic stress by through a coefficient of internal friction, µ .11 Bucknall’s yield function, Φ , was introduced as: 3σ m σe µe σm µ e σ m - + 2f cosh ---------- Φ = ------ + ------------- 2 – ----------- σ 0 2σ 0 σ0 σ0
- f2 – 1 = 0
[5]
where: σ e is the von Mises equivalent stress, σ 0 is the yield stress in the absence of hydrostatic stress, σ m is hydrostatic stress, µ e is the tensile equivalent coefficient of internal friction, where µ e = ( 3µ ) ⁄ ( 2 ) , and f is the volume fraction of pores in the matrix. By assuming that all of the rubber particles cavitate before yield, one can use the volume fraction of rubber particles for f in eqn. [5] to predict the yield envelope of rubber toughened polymers. This paper reports the results from an experimental investigation that evaluates the utility of eqn. [5] for predicting the multiaxial yield behavior of rubber-modified polymers. Herein, we present results describing the macroscopic yield/failure envelopes of rubbermodified epoxy networks. The failure envelopes encompass stress states ranging from uniaxial compression to biaxial tension. Additional studies and calculations are conducted to more quantitatively determine the energy dissipation through particle cavitation and irreversible matrix deformation. These studies involve loading and unloading samples in biaxial tension to various stress levels and studying their morphologies.
EXPERIMENTAL The rubber modified epoxy used in this study was made from a diglycidal ether of bisphenol A (Shell Chemical Co.’s EPON 828), reacted with 0%, 10%, and 20% carboxyl terminated butadiene acrylonitrile (CTBN 1300X8) rubber by weight. The epoxy was cured with 1,3phenylenediamine and aniline, to a molecular weight between crosslinks, Mc=900 grams/ mol. The properties of CTBN rubber12 and details on the calculation of Mc can be found elsewhere.13 In this paper, the materials will simply be referred to as samples A, B, C, and D, where: A is the unmodified epoxy, B is the 10% CTBN-modified epoxy with particle radius, R=1 µ m, C is the 21% CTBN modified epoxy with R=1.4 µ m, and D is a 21% CTBN modified epoxy with R=5.3 µ m. To examine morphology, the samples were cryofractured in liquid nitrogen, coated with a 100 Å thick layer of gold palladium and examined in a JOEL 35F scanning electron microscope, SEM, in secondary electron imaging (SEI) mode. The resin was spun-cast into thin-walled hollow cylinders as shown in Figure 2. The mechanical tests were conducted on a biaxial tension-torsion machine, modified to supply internal pressure to the hollow cylinders. All samples were tested in stress states ranging from uniaxial compression to biaxial tension, at 21°C and a constant octahedral shear strain
196
Plastics Failure Analysis and Prevention
rate, γ· =0.0028 min-1. Details of the fabrication and testing procedure are described elsewhere.14 oct
RESULTS AND DISCUSSION MACROSCOPIC YIELD ENVELOPE A Plot of equivalent stress, σ e versus hydrostatic stress, σ m, as measured from hollow cylinder tests, are shown in Figure 3. The solid and hollow symbols represent ductile yield and brittle fracture, and the solid and dashed lines represent the linear regression fit yield and fracture envelopes. First, we see that when the Figure 2. Thin walled hollow cylunmodified epoxy (A) is subjected to stress states with an increasinder geometry. ing hydrostatic component, the yield strength decreases and the failure mode changes from ductile yield to brittle fracture. Also note that in the rubber-modified systems (B, C, and D), macroscopic yield is not always realized before the specimens fail. However, their strengths do follow that of the yield envelope. This suggests that significant yielding is occurring and there is no change in failure mode. In this paper, any failures that follow the macroscopic yield envelope of the material are considered to have yielded. Using this definition, samples B, C and D show that the addition of rubber particles Figure 3. Equivalent stress versus hydrostatic stress, for CTBN-modified epoxy. The solid symbols, hollow sym- reduces the yield strength of epoxy, but supbols, and lines represent yield, fracture, and the yield/ presses brittle fracture in the highly confined fracture envelopes, respectively. stress states. The only difference between samples C and D is the size of the rubber particles. Figure 3 shows that increasing rubber particle size may affect the ability of the samples to reach yield, as the samples with large particles often failed by a brittle mode prior to yield. This could be attributed to samples D having larger inter-particle distances. Finally note that µ e decreases with the addition of rubber particles: µ e=0.55 for the unmodified epoxy (A), µ e=0.51 for 10% CTBN samples (B), and µ e=0.46 for 21% CTBN samples (C and D). COMPARING RESULTS TO YIELD MODEL PREDICTIONS The experimentally measured yield strengths for samples A, B, C and D are presented in Figure 4, with the predictions made by eqn. [5]. The solid symbols and dashed lines repre-
Evaluation of a Yield Criteria
197
sent the measured yield strengths and yield envelopes. The solid lines are the predictions from eqn. [5] for epoxy with µ e=0.55 and 0, 10 and 21% CTBN rubber by volume. From Figure 4, we first see that eqn. [5] predicts the yield envelope of the pure epoxy. This is expected because when f=0, eqn. [5] reduces to a modified von Mises yield function, which has been shown to describe the yield behavior of epoxies.6,14 For the rubber modified samples, the actual yield strengths are lower than predicted by the model. This suggests that the partiFigure 4. Comparison of the yield strength and eqn. (5) predictions for rubber-modified epoxy. The symbols and cles might act as stress concentrators which is not dashed lines represent the yield strengths and envelopes, accounted for in eqn. [5]. Finally we notice that solid lines are model predictions. in contrast to eqn. [5], the actual yield envelopes of the rubber-modified systems remain linear with increasing σ m, and µ e decreases with rubber content. Also, it should be mentioned that eqn. [5] does not predict the brittle failure mode observed in these systems. Possible reasons for the discrepancy between the data and eqn. [5] predictions can be attributed to eqn. [5] being based on a model for a perfectly plastic porous media which approximates the exact solution with a hyperbolic cosine function. This solution was further modified in an approximate fashion to accommodate for the pressure sensitivity that polymers exhibit. Finally, the modifier is introduced in this model through the modifier concentration only and does not incorporate the effect of particle size or properties of the modifier. Unfortunately, these simplifications appear to be too limiting for accurate prediction of the yield response of these types of systems. This result also underscores the need for more refined theoretical models for these materials. ENERGY ABSORBING MECHANISMS Selected tests were conducted to assess the energy absorbing mechanisms that are activated before gross yielding. Efforts were made to isolate the effects of cavitation and plastic matrix flow. The tests involved loading and unloading a hollow cylinders in equi-biaxial tension to a range of stress levels prior to failure, and monitoring changes in stiffness, irreversible work, and morphology. A plot of σ a, versus ε a, for samples C loaded in biaxial tension is shown in Figure 5. Up to a critical stress level of σ a ≅ 19 MPa, the deformation is elastic. Just beyond this, the loading stiffness changes, and deformation becomes increasingly irreversible as noted in the
198
Plastics Failure Analysis and Prevention
Figure 6: SEM micrographs of samples C, loaded and unloaded in biaxial tension to σ a= 0, and 16 MPa.
Figure 5: Axial stress versus axial strain for samples C; hollow cylinders loaded and unloaded in biaxial tension to stress levels of: σ a=0, 16, 31, and 41MPa.
hysteresis curves. The hysteresis curves indicate that ~19 MPa marks the threshold for significant increases in irreversibility. However, the SEM micrographs in Figure 6 indicate that significant cavitation occurs well before this threshold is reached. This suggests that the primary mechanism for the irreversibility is not associated with Figure 7: Irreversible work versus maximum axial stress, cavitation, and hence this threshold must be for samples C and D loaded and unloaded in biaxial tension. Energy absorbed through particle cavitation, UC is associated with the onset of plastic matrix flow. calculated using equations (2) and (3). A second series of tests were conducted on samples D. Again the hysteresis curves showed that ~19 MPa marks the threshold of significant irreversibility. This further supports the idea that particle cavitation occurs well before this threshold value, since the cavitation stress is particle size dependent. QUANTIFICATION OF ENERGY DISSIPATION The irreversible work done to the rubber-modified samples loaded and unloaded in biaxial tension, wi, was calculated from the hysteresis of Figure 5. In Figure 7, wi is plotted versus the maximum σ a for both samples C and D, showing that only above the critical stress of σ a ≅ 19 MPa is significant irreversible work done to samples C and D. Note that the irreversible work plotted in Figure 7 are measures of the total energy dissipated. To estimate the amount of energy dissipated by particle cavitation, we start by considering that all particles
Evaluation of a Yield Criteria
199
cavitate and use equations [2] or [3] to estimate the contribution of each particle, Uc, to the total irreversibility using eqn. [6]. U c u c = ------------- ( f ) 4--- 3 3 πR
[6]
In Figure 7, Uc as predicted by equation [6], using both Bucknall’s and Groeninckx’s models are plotted with wi versus maximum applied σ a. Both plots shows that the irreversible work done to the samples is primarily due to irreversible deformation of the matrix material. In fact, when extrapolated to the failure stress of samples C and D, Bucknall’s model predicts that less than 5% of the total energy is stored in the cavitated rubber particle, and Groeninckx’s model predict a value of less than 2%. Therefore, we conclude that the primary energy dissipating mechanism in rubber-modified epoxy, tested in biaxial tension, is irreversible flow of the matrix material.
CONCLUSIONS Biaxial testing of hollow cylinders showed that adding rubber particles to epoxy: decreases the yield strength, reduces the pressure sensitivity of yield, and suppresses brittle fracture in the more confined stress states. Comparison of the measured yield envelopes with a recently published model based on a plastic porous media showed deviations. Additionally two energy absorbing mechanisms were investigated: rubber particle cavitation and plastic matrix flow. Biaxial tests and morphological studies revealed a critical stress above which deformation becomes increasingly irreversible. This threshold stress was found to be independent of particle size, which supports the contention that this threshold is associated with the onset of matrix flow and not particle cavitation. Finally, cavitation models showed that cavitation accounts for less than 5% of the total energy dissipated in deforming a rubbermodified sample to macroscopic yield in biaxial tension.
ACKNOWLEDGEMENTS The authors wish to acknowledge the financial support from the NSF, Materials Research Science and Engineering Center (MRSEC) and Shell Chemical Co., Elastomers Department. The authors also thank Mike Modic and Mike Masse at Shell Chemical Co. for their helpful comments and suggestions.
REFERENCES 1 2 3
D. S. Parker, H.-J. Huang, and A. F. Yee, Polymer, 31, 2267 (1990). Dompas and Groeninckx, Polymer, 35, 4743 (1994). R. A. Pearson and A. F. Yee, J. Mater. Sci., 26, 3828 (1991).
200
4 5 6 7 8 9 10 11 12 13 14
Plastics Failure Analysis and Prevention
H.-J. Sue, J. Mater. Sci., 27, 3098 (1992). A. Lazzeri and C. B. Bucknall, J. Mater. Sci., 28, 6799 (1993). J. N. Sultan and F. J. McGarry, Polym. Eng. Sci., 13, 29 (1973). S. Wu, Polymer, 26, 1855-1863 (1985). R. J. M. Borggreve, et.al., Polymer, 28, 1489 (1987). Huang and Kinloch, J. Mater. Sci., 27, 2763 (1992). A. L. Gurson, J. Eng. Mater. Technol., Trans. ASME, 99, 2 (1977). S. S. Sternstein and L. Ongchin, A.C.S. Pol. Prep., 10, 1117 (1969). A. F. Yee and R. A. Pearson, J. Mater. Sci., 21, 2462 (1986). E. D. Crawford and A. J. Lesser, J. Appl. Polym. Sci., in press (1997). R. S. Kody and A. J. Lesser, J. Mater. Sci., (1997).
Chapter 6 Design and Life Prediction
Shelf Life Failure Prediction Considerations for Irradiated Polypropylene Medical Devices
Michael T. K. Ling, Samuel Y. Ding, Atul Khare, and L. Woo Baxter International, Round Lake, IL 60073, USA
INTRODUCTION Ionizing sterilization is gaining popularity in medical device and packaging industry because of its convenience and lower cost. The mode of sterilization is a consequence of the high energy electrons released from the interaction of the gamma ray photons or electron beam particles with the materials being sterilized. These high energy electrons in turn interact with the DNA sequences in the microbiological burdens, through permanently altering their chemical structures to render them innocuous. The high energy electrons, however, can also initiate ionization events in the material being sterilized. It can create peroxy and hydroperoxy free radicals in the presence of oxygen, and start the degradation cascade. This could result in an unacceptable color formation, excessive pH shifts and high extractable. Furthermore, the degradation could also lead polypropylene (PP) to well-publicized catastrophic failure during post radiation shelf life storage. From product quality and application viewpoints, it is thus highly desirable to develop a simple and rapid method to characterize the impact or most importantly the shelf life of radiation sterilized packaging materials or devices. The purpose of this paper is to examine the various important parameters that need to be taken into consideration when a specific method is applied to predict the shelf life of a product made of polypropylene material.
202
Plastics Failure Analysis and Prevention
EXPERIMENTAL Technique used in this study includes ASTM D3895-92 isothermal oxidative induction time (OIT) from Dupont 1090 thermal analyzer with 910 differential scanning calorimetry (DSC) cell. OIT was conducted under air flow condition of 100 CC/min. The thickness is usually 5 mil or thinner. Failure morphology was examined by JEOL 35CF-SEM after sputter coating with palladium for surface conductivity.
RESULTS AND DISCUSSIONS THE EFFECT OF LONG LIVED FREE RADIALS Both the sterilizing action and the degradation caused by ionizing radiation are believed to result from Compton secondary electrons from the primary interaction event. The high energy gamma photons or accelerated electrons (from the e-Beam source) interact with the atoms in the material, creating a secondary high energy electron and a recoiling photons or electrons. These electrons can lead to a series of secondary ionization events in localized spurs. The cascade is propagated until all the excess energy above the ionization threshold is dissipated. Thus from a single incoming photon or electron, a binary tree configuration of secondary electrons are generated, and they are responsible for the bioburden kill and material degradation. Catastrophic failures have been reported during the PP shelf life storage period. Intense investigation has come to the following hypothesis: Long lived free radicals trapped in the crystalline domains slowly migrate towards the crystalline /amorphous interface where they react with available oxygen to form peroxy and hydroperoxy radicals and initiate degradation near the interface. When sufficient number of the tie molecules between crystallites were cut through this chain scission process, PP elongation is reduced dramatically and catastrophic failures followed. At the same time, since the outer surface is more exposed to atmospheric oxygen, the extent of degradation near the surface is much greater than that of the interior. A brittle layer is then formed and has the same effect as forming sharp notches on the sample, creating stress concentrations. Once the notch reaches a critical size, failure occurs. To confirm that long lived free radicals do play a significant role in the post irradiation PP degradation, a PP film sample was examined by electron paramagnetic resonance (EPR) spectroscopy at room temperature, Figure 1; the PP film was about 6 months old after an irradiation dose of 25 KGy dose at about 10 Figure 1. EPR signal of PP, 6 KGy/hr rate. It is seen that the free radical mediated oxidative degmo. post irradiation.
Shelf Life Failure Prediction
Figure 2. PP OIT vs. antioxidant concentration.
203
Figure 3. Combined Arrhenius plot of OIT data (DSC) and oven data (brittleness).
radation continues in polypropylenes long after the irradiation event. Incidentally, the strong EPR signal was completely eliminated when the sample was annealed in a vacuum oven at 90°C, a temperature much above the glass transition for the amorphous phase for PP, and well into the alpha relaxation for the crystalline phase of PP. SHELF LIFE PREDICTION BY OIT (OXIDATIVE INDUCTION TIME) FOR PACKAGING FILMS The remaining antioxidant in the PP material post irradiation is very critical to protect against further degradation. The action of the phenolic antioxidant is mainly that of a hydrogen donor in eliminating organic free radicals and becomes sacrificially consumed in the process; once it is completely consumed, catastrophic failure of PP will occur. Therefore, the ability to measure the remaining antioxidant is very useful for shelf life prediction. Confirming what have been widely reported in the literature,1 we also found that, OIT at various temperatures was an excellent linear function of active antioxidant content, Figure 2. The exceptionally linear response of OIT at multiple temperatures strongly indicates the potential of using this method for simple and very rapid, although non-specific assay for active antioxidant determination. OITs of a radiation copolymer PP film of about 5 mil in thickness was determined before and after 20 KGy of gamma exposure at about 6 KGy/hr dose rate. The data was plotted in the Arrhenius form in Figure 3. It is clearly seen that the gamma exposure has significantly reduced the OIT throughout the temperature range studied. To access stability at
204
Plastics Failure Analysis and Prevention
still lower temperatures, where the OIT detection becomes difficult, the gamma exposed films were subjected to oven aging at 90°C and 60°C. Their failure times were noted when film samples became brittle. Interestingly, when the oven failure data were plotted onto the Arrhenius plot with the gamma samples, a continuous curve with diminishing slope toward lower temperatures emerges. This kind of continuity of functional behavior of OIT data at higher temperatures and oven stability data closer to ambient could, at least in principle, proFigure 4. Strain rate effect on PP. duce long term property prediction based on OIT data, provided that the rate of slope change can be determined separately. It was observed that the slope of Log OIT vs. 1/T at any 1/T is nearly constant for many of the polypropylene tested regardless homopolymer or copolymer in nature or radiation history. For example, the 20KGy curve of Figure 3 can be shifted vertically by a single factor to coincide with the non-irradiated sample. This implies that if a “master curve” could be constructed, the shelf life of the PP can be predicted once the shift factor is determined. The determination of the shift factor should only involve a single, or at the most, a few carefully chosen calibration points. A detailed evaluation of this technique including the applicability of the master curve is currently underway and results will be reported in the future. SHELF LIFE PREDICTION CONSIDERATION FOR THICK DEVICE For a typical packaging film which is relatively thin compared to device, shelf-life prediction using OIT method as mentioned above is quite satisfactory according to our opinion. However, for a device many times thicker, the ductility of the entire device greatly depends on the ductility of the surface even though the bulk of the core material is still very ductile, the geometry, and the strain rate involved during in use. If the surfaces were brittle and their thickness exceeded the critical thickness, failure is likely to occur. A few methods have been reported for PP shelf life predictions: 1. Molecular weight equivalent method3 2. High surface area, high orientation acceleration test3,4 3. Variable Q10 and D&A techniques5 based on bending angles Typically, the three points bending flexural test under ASTM D-790 is specified for polypropylene device post radiation testing. Under ASTM D-790, a strain rate of 0.01 min-1 (or 0.00016 S-1) is specified for the outer fiber. From Figure 4, one can see that all polypro-
Shelf Life Failure Prediction
Figure 5. 15 year PP surface embrittlement.
205
Figure 6. PP core, 15 yr. crazing, large strain.
pylenes regardless of molecular weight are quite tough. However, in an actual device application, significantly higher strain rates are encountered. For example, a suddenly loaded syringe flange is estimated to be subjected to a strain rate of about 0.5 sec-1 at the root. This is a factor of 3000 times higher than the ASTM bending test. In addition, strain rate at stress concentrated areas also shifted toward much higher levels. Based on the above discussion, strain rate, stress concentration factor (or geometry factor), molecular weight are all important factors that determine the ductility of the PP device. Knowing the brittleness dependency of strain rate, it is critical to chose the appropriate testing strain rate for an accurate shelf life prediction; too low the testing rate would lead to exceptionally high shelf lives and too high the testing rate would make all samples brittle. It is therefore extremely critical to determine the actual worst case strain rates to predict the shelf life of the device. PP with an amorphous (Tg) of about 0oC is not a tough material. Especially when subjected to high rates of strain, the glass transition is shifted to a higher temperature. The socalled ductile-brittle transition for PP is estimated to lie in the strain rate range of between 0.1 and 10 sec-1 at room temperature.2 When the oxidative degradation progresses from the external surface, the fraction of the brittle layer steadily increases with time. The brittle layer has the same effect as forming surface notches, and the notches can shift the deformation to much higher strain rates. Hence, when a sufficient brittle layer has built up on the surface, even a slow deformation (strain) rate can have the same effect as a high rate event, catastrophic brittle failures becomes the inevitable outcome. The growth rate of the brittle layer depends on the copolymer type and content of the PP, stabilizers package and morphology. Oxidation rate of semi-crystalline PP also depends on the degree of crystallinity. A gamma irradiated (50 KGy) PP copolymer (about 2 wt% ethylene) tensile bar was examined after 15 years of storage at ambient. A layer of surface embrittled polymer degradation product appears to cover the entire surface, while the core
206
Plastics Failure Analysis and Prevention
of the sample remained relatively ductile. The scanning electron microscopy (SEM) morphology of a high speed fracture plane cross section, Figure 5, indicated about 100 micron thickness for the totally degraded layer; this is about 6.7 µ m/year. When the brittle surface layer was removed, and the remaining interior sample subjected to a sharp bend of 180 degrees, very ductile behavior with massive crazing and stress induced whitening were observed, Figure 6. The formation of massive crazing in the stress whitening zone indicates that the mechanical behavior of the core material is near indistinguishable from that of unirradiated polypropylene. From the core of the sample, the residual antioxidant content as assayed by the OIT, lies along the same trend line of un-aged samples. This indicated that the interior of the sample, where ambient oxygen could not diffuse in through the surface in sufficient quantities, 15 years post irradiation storage has little effect on the material property. This is a strong evidence that during post irradiation storage, the oxygen diffusion is the limiting factor for degradation, and the brittle behavior is mainly due to the brittle skin that form a notch and magnified the local stress and therefore the strain rate to cause a brittle failure. Only a slight deformation is needed to cause the brittle surface to fracture to form cracks. These cracks act as sharp notches and the stress concentration factor depends on the radius of the crack tip. The stresses become infinite as the radius of the crack tip approaches zero; in this case a stress intensity function must be used to express the stress in the vicinity of the crack. For the propose of discussion, the notch is assumed to be blunt with a radius of r and a crack length of a, then the stress in the direction of the applied stress can be approximated by an elliptical solution and is expressed as6 σ = σ0 1 + 2 ( a ⁄ r )
1 --2
= σ0 KI
[1]
Where KI is the stress concentration factor and reduces to KI=2(a/r)1/2 for a relatively sharp notch(a>>r). σ 0 is the applied stress at far field. Notches not only introduce tri-axiality into the stress field, but also alter the strain rate in the highly stressed region around the root of the notch. The strain rates increase linearly with KI or (a/r)1/2 since strain is directly proportional to the stress for a linearly elastic material. The significance of strain rate on PP’s ductility is seen in Figure 4. From equation [1], one can see that with a steady increase in the notch depth, the stress concentration factor also increases. Hence, for a typical notch radius of 2 microns, these stress concentration factors are obtained. Brittle Layer Thickness, (microns)
Stress Concentration Factor, KI
1 10 100
2.4 5.5 15.1
Shelf Life Failure Prediction
207
It can be seen then, stress concentration factor of about 2 to 15 are encountered. This is also the factor based on linear elasticity which allow one to scale the local strain rate based on the brittle layer thickness. Using data from Figure 4, one can estimate the ductility for the sample or device. For example, for a 10 micron brittle layer thickness, or an acceleration of about 5, and a nominal strain rate of 0.5 sec-1, both the low and high molecular weight samples are becoming brittle, while the medium molecular weight sample remains relatively unaffected. However, at a brittle layer thickness of 100 microns, or an acceleration factor of about 15, the medium molecular weight sample also lost its ductility completely. With these parameters quantified, one can begin to construct a model for property prediction. In the case of near complete depletion of antioxidants, immediately after radiation one can factor the brittle layer advancing rate into the component thickness and the expected strain rate.during use, to arrive at a reasonable conclusion on storage shelf life at any temperature where ductile behavior can be maintained. As usual, all predictions based on theoretical considerations need to be validated with parallel real time aging studies.
SUMMARY Factors governing polypropylene’s post radiation shelf life to maintain the mechanical ductility were identified. The oxygen diffusion limited degradation model first proposed by Gillen and Clough7 was experimentally confirmed. From a sample with 15 years of real time ambient aging, the advancing velocity of the surface embrittled zone was determined. It is proposed to incorporate the surface brittle layer thickness into an overall model based on stress concentration factor and the strain rate dependence for device and component shelf-life prediction. An OIT method was found useful in determining the stability of PP at high temperatures, and the data may be extrapolated to ambient temperature to predict the shelf life of PP by following the curvature, non-Arrhenius behavior, of Figure 3.
REFERENCES 1 2 3 4 5 6 7
G. N. Foster, Oxidation Inhibition in Organic Materials, Vol.2, J. Pospisil, P. Klemchuk eds., CRC Press, Boca Raton (1989). R.J. Rolando, W. Krueger and H. Morris, SPE Proceeding, ANTEC, 657, 1986. L. Woo at al, Shelf Life Prediction Methods and Applicabilities, ANTEC ‘91, P1854. L. Woo, C. Sandford, and R. Walters, in Advances in Biomaterials, P.52, Edited by S.M. Lee, Technomic Publishing, Lancaster, 1987. Donohue, and S. Apostolou, Predicting Post-Rad Shelf Life from Accelerated Data: The D&A Process, ”Technical papers, Brookfield, CT, SPE, 42:2819-2822, 1996. J.G. Williams, Stress Analysis of Polymers, Halsted Press, John Wiley& Sons, NY, 1973. K. Gillen and R. Clough, Irradiation Effects on Polymers, D. W. Clegg and A. A Collyer Eds. Elsevier Applied Science, New York, 1991.
Determining Etch Compensation Factors for Printed Circuit Boards
Anthony DeRose, Richard P. Theriault and Tim A. Osswald Polymer Processing Research Group, University of Wisconsin-Madison, Madison, WI, USA Jose M. Castro Ohio State University, Columbus, Ohio, USA
STANDARD PROCESSING METHOD Metal-clad, multi-layered, fiber mat reinforced, thermoset resins are typically processed in either an autoclave or within a press. In either case, the individual plies are first sequentially layered, as depicted in Figure 1 and pressed as a stack, or book of laminates as shown in Figure 2. The book is then subjected to a mold temperature and pressure schedule to advance the degree of cure and adhere the plies to form the laminate composite. A typical mold temperature and pressure schedule for the pressing of laminate comFigure 1. Schematic of metal-clad, multi-layered, fiber matrix reinforced thermoset laminate. posites is illustrated in Figure 3. In stage I of the processing schedule shown in Figure 3, the resin viscosity decreases due to the influence of the heat transfer from the molds and flow occurs upon the application of pressure. Stage II is characterized by the steady increase in the degree of cure. Since the mold temperature during stage II is higher than the glass transition temperature, Tg, of the resin, the residual stress developed due to curing strain and flow are in competition with the stress relaxation process within the polymer network. In the ideal case, it has been found that residual stress is completely relaxed by the time the material is cooled below Tg.1 The pressure profile during stage II is maintained at a high level to impede the growth of voids which may exist within the resin. The mold temperature profile associated with stage III is typically a crash, or sudden cooling, of the laminate through the Tg of the resin. The cycle terminates when the laminate can be eas-
210
Plastics Failure Analysis and Prevention
Figure 3. Platen temperature and pressure scheduling.
ily handled and removed from the press. Optimization of this time period will lead to the minimization of the processing induced stress. Figure 2. Schematic representation of ply Once the laminates are removed from the press the sequence within a six laminate book. circuit patterns are developed by chemically removing the unwanted metal from either side of the laminate. During this operation much of the residual stresses that have developed within the metal layers are transferred to the thermoset layers. The cured and etched laminates can then be sandwiched between prepregs and restacked in the mold. Repeating this cycle leads to the production of multi-layered circuit boards.
MODEL DEVELOPMENT CURE KINETICS The conversion of a thermoset to a glassy solid involves two transitions: gelation and vitrification. Gelation is the time at which covalent bonds form to generate a three-dimensional network that gives rise to long range elastic behavior. Vitrification is when Tg of the system rises to the cure temperature and when further reaction is prohibited or is dramatically reduced. The degree of cure can be assumed to be proportional to the number of bonds formed during crosslinking such that each bond releases the same amount of heat. Thus, the degree of cure can be defined as Q c = ------QT
[1]
where QT is the total heat of reaction and Q is the amount of heat released up to the current time. The rate of heat generation can be expressed as
Determining Etch Compensation Factors
dc Q· = Q T -----dt
211
[2]
where the cure rate term can be described using an empirical model. One such model is the Kamal-Sourour2 reaction model written as m n dc ------ = ( k 1 + k 2 c ) ( 1 – c ) dt
[3]
where k1 and k2 are kinetic rate constants, and both m and n describe the reaction order. The kinetic rate constants have been modified by the Rabinowitch model3 as suggested by Havlicek and Dusek,4 to account for the diffusion of the chain segments as 1 1 1 ---- = -------- + ----ki k i, c k d
[4]
where ki,c is the Arrhenius kinetic rate constants for both 1 and 2 and kd is the Arrhenius kinetic rate constant for the diffusive effects described by –Ed –b k d = a d exp --------- exp ------ RT f
[5]
where a and b terms are rate constants, Ed are activation energies, f accounts for the equilibrium fractional free volume, R is the gas constant and T is the temperature. When the Tg of the material rises beyond the cure temperature, i.e., after vitrification, diffusion becomes the dominating mode of transport. The ability to predict the point of gelation and Tg is important in being able to predict the formation of residual stress. At the point of gelation, a network of crosslinked chains propagates throughout the resin, thus enabling the semi-solid resin to withstand stress.5 A statistical approach to predict the degree of cure in which the resin can withstand residual stress is proposed by Flory’s theory of gelation.3 For example, a stoichiometric mixture of DGEBA and a tetra-functional aromatic amine would yield a theoretical conversion at gelation of 0.57.7 Predicting the progression of Tg is important in determining the point where residual stress development during cooling is initiated. For a system exhibiting a one-to-one relationship between Tg and conversion, DiBenedetto’s equation8 is an approach for stoichiometric ratios to express this relationship using only a single parameter model. In order to solve the kinetic models aforementioned, we must first solve the transient heat transfer from the mold to the laminate. The governing equation for heat transfer is the energy balance for heat conduction expressed as 2 ∂T · ρC p ------ = k ∇ T + ρQ ∂t
[6]
This equation can be reduced to one-dimension assuming that the planar heat transfer through the laminate can be neglected compared to the through thickness temperature transients. Laminates are typically thin enough such that the exothermic heat generated due to
212
Plastics Failure Analysis and Prevention
conversion can be conducted out of the resin relatively quickly, and therefore, the source term, Q· , can be neglected. The Crank-Nicolson Method was used to approximate Eq. [6] which results in a system of equations that can be solved using the Thomas Algorithm.9 MECHANICAL PROPERTIES As a thermoset cures, the mechanical properties of the material behave similar to a viscous liquid in its uncured state and similar to an elastic solid in its fully cured state. It is convenient to divide the curing into three stages, as shown in Figure 4. In Stage I the material is considered a viscous fluid, with negligible stiffness. In Stage II, conversion has advanced to form a network gel and the mechanical properties, such as the stiffness, Figure 4. Conceptual model for the progression of cure develop. The progression of the elastic modulus of thermoset resins. during this stage has been reported to have a linear relationship with the degree of cure.10 Stage III indicates that the degree of cure of the material has reached a terminally high value. For epoxy resin systems it has been found that the curing strain due to conversion can be neglected in the residual stress analysis.5 For laminate composites, the CTE (coefficient of thermal expansion) of the plies, non-uniform cooling, and the removal of the copper during the etching stage of the process significantly influences the residual stress profile leading to dimensional movement.11 Thus, the factors that dominate the residual stress profile are the mechanical properties of the individual plies at the point of cool down. To evaluate the dimensional movement of the laminate, the generalized Classical Lamination Theory was implemented.12 This theory allows for the prediction of the thermal residual stress as a function of the complex coupling effects associated with planar movement and warpage. The generalized theory for stress takes the form 0
{ σ } k = [ Q ] k ( { ε } + z { κ } – { β } k ∆T k )
[7] where {σ} is the stress, [Q] is the reduced stiffness matrix, {β} is the CTE and ∆Τ is the temperature difference at the kth laminate. The dimensional movement of the laminate is associated with mid-plane strain, {ε0}, and the mid-plane curvature, {κ}. The additional movement that occurs during the etching process is a result of the transfer of residual stress from the metal layers to the reinforced thermoset as the metal is chemically removed. Using the predicted residual stress profile in the first half of the simulation, we can determine how much of this stress is removed as a function of the metal removed during etching. The subsequent strains that result from removing this portion of the residual
Determining Etch Compensation Factors
213
stresses can then be calculated. This method allows us to specify the amount of metal that is removed in either of the planer directions for each side of a laminate independently. Etch compensation factors are then developed for each laminate based on the total predicted shrinkage. These etch compensation factors can then be used as a factor by which the artwork should be magnified to take into account the final amount of shrinkage the laminate will undergo during processing. This will insure that each layer of circuits will align as expected within the final board. The etch compensation factors in the warp and fill direction for the kth laminate can be expressed symbolically as W
W
F
= 1 + εk
fk = 1 + ε
[8]
k
F
[9] where and are the predicted total strains in the warp and fill directions of the kth laminate, respectively. fk
W εk
F εk
NUMERICAL PREDICTION The simulation was used to analyze an unsymmetric laminate similar in construction to the diagram shown in Figure 1. The general laminate construction is shown in Table 1, and the more specific mechanical properties of the individual components of this laminate are shown in Table 2. Table 1. Prepreg laminate construction and properties. Prepreg A
Prepreg B
Units
Resin
Epoxy
Epoxy
-
Resin mass percent
63
52
%
Init. degree of cure
28
28
%
Mat Weave
Plain
Plain
-
Warp fibers
1800
450
m/kg
Fill fibers
1800
900
m/kg
Thickness
0.043
0.086
mm
The laminate was first examined at the stage in the process after removal from the press cycle, prior to the etching process. The residual stress profiles in both the warp and fill directions are shown in Figure 5. The total movement in the warp direction was 0.00232
214
Plastics Failure Analysis and Prevention
[mm/mm] with a radius of curvature of 1550 [mm] towards Prepreg A. In the fill direction, the total movement was 0.00257 [mm/mm] with a radius of curvature of 2245 [mm] toward Prepreg B. Since the radius of curvature was smaller in the warp direction than in the fill direction the dominant mode of warpage was in the direction toward Prepreg A. Table 2. Properties of individual plies. Material Metal
Prepreg A
Property
Value
Units
Thickness
0.036
mm
Density
8950.0
kg/m3
Thermal cond.
386.0
J/(s m K)
Specific heat
385.0
J/(kg K)
Thermal exp. coef.
16.5e-6
1/K
Elastic modulus
110.0
GPa
Shear modulus
40.0
GPa
Poisson’s ratio
0.355
-
Thickness
0.042
mm
Resin mass ratio
0.63
%
Glass (warp/fill)
1.0
-
Density
1538.8
kg/m3
Thermal cond.
0.17
J/(s m K)
Specific heat
116.0
J/(kg K)
Exp. coef. (warp)
18.2e-6
1/K
Exp. coef. (fill)
18.2e-6
1/K
E-Modulus (warp)
14.03
GPa
E-Modulus (fill)
14.03
GPa
Shear modulus
4.64
GPa
Determining Etch Compensation Factors
215
Table 2. Properties of individual plies. Material
Property
Value
Units
Prepreg A
Poisson’s ratio
0.32
-
Prepreg B
Thickness
0.084
mm
Resin mass ratio
0.52
%
Glass (warp/fill)
2.0
-
Density
1669.9
kg/m3
Thermal cond.
0.20
J/(s m K)
Specific heat
150.2
J/(kg K)
Exp. coef. (warp)
11.3e-6
1/K
Exp. coef. (fill)
24.0e-6
1/K
E-Modulus (warp)
21.90
GPa
E-Modulus (fill)
13.54
GPa
Shear modulus
5.41
GPa
Poisson’s ratio
0.31
-
Because we want to be able to calculate the movement of etched laminates in order to compensate the artwork, the analysis was repeated with the simulation of the etching stage incorporated. A typical circuit pattern is shown in Figure 6. From this photograph the distribution of the remaining metal on the laminate was determined, which was then used in the simulation. Figure 5. Stress profile of the unetched laminate. This distribution is shown in Figure 7, and shows that approximately 3.2 percent of the circuitry runs in the warp direction (180 degrees), while approximately 2.4 percent runs in the fill direction (90 degrees) on either side of the laminate. Figures 8 and 9 show the additional
216
Plastics Failure Analysis and Prevention
Figure 7. Circuit orientation distribution on laminate (180 is warp direction, 90 is fill direction).
Figure 6. Photograph of the circuit pattern on a single laminate (both sides shown).
Figure 9. Predicted movement in the fill direction as metal is progressively removed (96.8 percent of metal was removed from the warp direction initially).
movement resulting from including the etching process in the simulation as a function of the amount of metal removed in the warp and fill Figure 8. Predicted movement in the warp direction as directions, respectively. These additional strains metal is progressively removed (97.6 percent of metal are a result of the changing laminate equilibrium was removed from fill direction initially). as the constraint of the metal is progressively removed. It is these values of strain that we then use to compute the etch compensation factors by which the artwork can be compensated. For example, Figure 8 shows that depending on the amount of metal removed in the warp direction during etching, the resulting movement in the warp direction can be as large as 0.8 mils per 10 cm of laminate. Subsequently, Figure 9 shows that etching metal in the fill direction can result in a movement of approximately 2.0 mils per 10 cm of laminate, or about twice the shrinkage for similar conditions as in the warp direction. As circuit patterns continue to decrease in size quantifying these movements will become important in reducing the number of circuit board failures.
Determining Etch Compensation Factors
217
CONCLUSIONS Based on the nature of the process and the materials involved in the manufacturing of thermoset laminates it is difficult to avoid dimensional movement. This movement can result in the failure of circuit boards due to the misalignment of the circuitry on the various layers of the board. Simulation is perhaps the least expensive method for determining what this movement will be before the circuits are etched such that compensation factors can be applied in order to avoid these failures. To increase the accuracy of these compensation factors, the model should be extended to capture the influence of the resin flow during the early stages of the process.
ACKNOWLEDGMENTS The authors would like to gratefully acknowledge Allied Signal Laminate Systems, Inc, and the State of Wisconsin, Office of University-Industry Research for their support.
REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13
Ochi, M., K. Yamashita and M. Shimbo, J. Appl. Polym. Sci., 43, p.2013, (1991). Kamal, M. R. and S. Sourour, Polym. Eng. Sci., 13 (1), p.59, (1973). Rabinowitch, E. Trans. Faraday Soc., 33, p.1225, (1937). Havlicek, I. and K. Dusek, in Crosslinked Epoxies, B. Sedlacek and J. Hahovec, eds., Walter de Gruyter, New York, (1987). Wang, H.-B., Y.-G. Yang, H.-Y. Yu and W.-M. Sun, Polym. Eng. Sci., 35, p.23, (1995). Flory, P. J. in Principles of Polymer Chemistry, Cornell University Press, Ithaca, New York, (1953). Riccardi, C. C., H. E. Adabbo and R. J. J. Williams, J. Appl. Polym. Sci., 29, (1984). DiBenedetto, A. T. J. Polym. Sci. Polym. Phys., 25, p.1949, (1987). Stikwerda, J. C., in Finite Difference Schemes and Partial Differential Equations, Wadsworth & Brooks/Cole, California, (1989). Kim, K. S. and H. T. Hahn, Comp. Sci. Tech., 36, p.121, (1989). Brahatheeswaran, C and V. B. Gupta, Polymer, 34 (2), p.289, (1993). Gibson, R. F. in Principles of Composite Material Mechanics, McGraw-Hill, New York, (1994). Karkanas, P. I., I. K. Partridge and D. Atwood, Polym. Int., 41, p.183, (1996).
Activation Energies of Polymer Degradation
Samuel Ding, Michael T. K. Ling, Atul Khare and Lecon Woo Baxter Healthcare, Round Lake, IL 60073, USA
INTRODUCTION In the study of polymer degradation and durability, there is little reliable, predictive methodology that is universally valid over wide spans of temperature and time. Many of the high temperature “accelerated” oven tests have been deemed unrealistic for different mechanisms were prevalent. In the mean time, for practical reasons, experimental time spans of much longer than a year are extremely difficult to conduct. In the medical plastics industry, products are frequently sterilized by ionizing radiation, which severely depletes the antioxidant package. Yet to conform to many regulatory requirements, a scientifically based estimate of post irradiation shelf life must be provided. Thus a better understanding on the time and temperature influence on the material's performance is a necessity for product introduction. In this study we have examined the Arrhenius activation energy as a function of temperature for many polymer systems important in the medical industry. Data from oxidative induction time (OIT), accelerated oven aging, and real time ambient storage to up to 23 years will be used to determine the functional behavior and quantitative significance of the activation energy.
EXPERIMENTAL AND MATERIALS Technique used in this study includes ASTM D3895-92 isothermal OIT from Dupont 1090 thermal analyzer with 910 differential scanning calorimetry (DSC) cell. Forced convection air circulating ovens were used at various temperatures to assess long-term oven age shelf life with sample embrittlement as end points. Morphological studies were done using a Reichert FC4E cryo-ultramicrotome to prepare undistorted material blocks for SEM analysis. SEM analysis was done with the JEOL 35CF-SEM after sputter coating with palladium for surface conductivity. In addition, other available characterization data were incorporated into this study. The materials studied consist of polypropylene (PP), high density polyethylene (HDPE), low density polyethylene (LDPE), EPDM and polyester thermoplastic elastomers.
220
Plastics Failure Analysis and Prevention
Gamma exposure at various doses was conducted in a laboratory gamma cell at dose rates of approximately 6-10 KGy/hr.
RESULTS AND DISCUSSION Both the OIT and chemo-luminescence data support the general mechanism of degradation where the primary alkyl free radicals are propagated through atmospheric oxygen diffusing into the polymer via the formation of peroxy and hydroperoxy free radicals (Figure 1). The rate limiting steps in this complex chain reaction scheme determine the overall degradation rate. Figure 1. Oxidative kinetic chain reaction. In this regard, the action of stabilizer, such as phenolic antioxidants, chocks a section of the degradation loop by eliminating organic free radicals, or decomposing the hydro-peroxides, and becomes.sacrificially consumed in the process. The activation energy of the degradation rate, as expressed in the Arrhenius form, will be affected by factors such as polymer composition, stabilizer package, and polymer morphology. GAMMA RADIATED POLYPROPYLENE Catastrophic failures have been reported during the PP shelf life storage period. Intense investigation has come to the following hypothesis, that, long lived free radicals trapped in the crystalline domains slowly migrate towards the crystalline /amorphous interface where they react with available oxygen to form peroxy and hydroperoxy radicals and initiate degradation near the interface.1,2 When sufficient number of the tie molecules between crystallites were cut through this chain scission process, PP’s elongation is reduced dramatically and catastrophic failures followed. To establish that long lived free radicals do play a significant role in the post irradiation PP degradation, a PP film sample was examined by electron paramagnetic resonance (EPR) spectroscopy. A distinct free radical spectrum was detected at room temperature about 6 months after irradiation. Incidentally, the strong EPR signal was completely eliminated when the sample was annealed in a vacuum oven at 90oC, a temperature much above the glass transition for the amorphous phase for PP, and well into the alpha relaxation for the crystalline phase of PP.3 In a separate study, a radiation grade PP's OIT under air flow conditions of 100 ml/min was determined and the result compared with the same film sample (about 130 micron in thickness) after 20 KGy of gamma exposure at about 6 KGy/hr dose rate. To access lower
Activation Energies of Polymer Degradation
Figure 2. Combined Arrhenius plot of OIT and oven times for gamma irradiated PP.
221
Figure 3. Bell Lab OIT and oven aging data on LDPE cable compound.
temperatures thermal stability, where the OIT detection becomes difficult, the gamma-exposed films were subjected to oven aging at 90 and 60oC and their failure times noted. When the OIT and oven failure times were plotted onto the Arrhenius plot for the gamma irradiated samples, a continuous curve with diminishing slope toward lower temperatures emerges (Figure 2). This kind of continuity of functional behavior of OIT data at higher temperatures and oven stability data closer to ambient could, at least in principle, produce long term property prediction Figure 4. LDPE cable activation energies. based on OIT data, provided that the rate of slope change can be determined separately. Further experiments along this line of reasoning are being carried out currently to explore the boundary of validity for several polymer systems. LOW DENSITY POLYETHYLENE This continuous curve behavior was very reminiscent of the data on crosslinked low-density polyethylene cable compounds studied with OIT, oxygen uptake, and oven aging experiments at the former Bell Telephone Laboratories,4 (Figure 3). When the high temperature results were extrapolated by the Arrhenius equation to lower temperatures, grossly and physically impossible optimistic results were obtained. An examination of the activation energies indicated a more than four-fold difference between the high temperatures and near ambient (Figure 4). This observation prompted the Bell lab researchers cautioning against using the high temperature OIT for low temperature durability predictions. Nevertheless, by
222
Figure 5. EPDM rubber OIT and oven aging.
Plastics Failure Analysis and Prevention
Figure 6. EPDM activation energies.
recognizing the curved nature of the durability function, one can indeed achieve realistic predictions. EPDM RUBBER A new class of thermoplastic elastomers was created when olefinic polymers (polyethylene, polypropylene) are dynamically vulcanized with a crosslinkable elastomer such as ethylene propylene diene rubber (EPDM). These so-called thermoplastic vulcanizes are quite resistant to oxidation and studies have been available on Figure 7. OIT of polyester thermoplastic elastomer. their stability over long period of time.5 A general-purpose thermoplastic sample of 50 Shore D hardness was chosen for the OIT study. Published data from a long term oven aging study for 50% strength reduction was plotted on the same graph for comparison (Figure 5). The activation energies at several temperatures are plotted in Figure 6. POLYESTER ELASTOMER Polyester thermoplastic elastomers (TPE) based on polybutylene terephthalate (PBT) hard segment and tetramethylene ether (PTMO) soft segments constitute an important class of medical elastomers because of their wide property range, solvent bonding capability, oxidative stability and processing ease. OIT measurements conducted in air at higher temperatures again coincided with the oven aging data reported in the literature (Figure 7). The activation energies at several temperatures are also calculated.
Activation Energies of Polymer Degradation
Figure 8. PP surface ”fibrils” at 60X.
223
Figure 9. PP crack depth in microns.
PP SURFACE EMBRITTLEMENT For an electron beam irradiated PP film sample undergone oven aging at 90oC, a curious phenomenon was observed. About 3 weeks into oven aging, surface fibrils orthogonal to the exposed film edge surface became visually observable. Under optical and electron microscopic observation, these fibrils are showed to be shallow, surface cracks (Figure 8). These cracks appeared to grow in number and their depth (measured in cross section by SEM) increases linearly as a function of time. The linear crack depth growth significant accelerated at approximately 25% of the film thickness (or about 50% of the film volume) (Figure 9). The activation energies determined from the rate of surface embrittlement are: 30 70 90 Temperature, oC 16 41 82 Ea, kJ/mole PP ACTIVATION ENERGY FROM OIT, OVEN AGING AND REAL TIME AGING Recently, in the authors’ laboratories, several prototype and production polypropylene bottles, which have been stored under ambient conditions for up to 23 years, were discovered. This “find” could allow the calibration of our long-term durability prediction methods. When the OIT of these products were determined, an excellent linear relationship with storage time, pointing to the zero OIT time of about 30 years. Hence, we can state, with reasonable assurance, that the durability of this particular grade of PP in the thin film form, under ambient storage, is about 30 years. When this data was combined with newly generated OIT and oven life data, plotted in the Arrhenius form, a continuous curve covering nearly 8 decades (100 million folds) of time was obtained (Figure 11). And when the local slopes
224
Plastics Failure Analysis and Prevention
Figure 10. Rate of degraded layer formation.
Figure 11. Combined OIT, oven aging and real time aging data of PP.
were measured and converted to the activation energy at various temperatures, again, a concave curve resulted. Yet another significant observation on PP is that the activation energy from thermal aging processes when compared with the activation energies of the rate of brittle layer formation, near identical results are obtained. This apparent “ self-similarity “, or near identical activation energies at the same temperature exhibited for different degradation measurement parameters and different grades of polypropylene, could lead to much simplified modeling and understanding of the degradation and durability process. Efforts are currently underway to gather more supporting data on this self-similarity and the utility it could bring.
SUMMARY A broad based study on the kinetics of polymer degradation was conducted. The Arrhenius activation energy was used as the parameter to follow the rate dependence with temperature. For most systems, a monotonic increasing trend with temperature was evident. This finding explains the frequent observation that kinetic parameters obtained at high temperatures often lead to grossly optimistic results at ambient. For a polypropylene copolymer system, combined data from OIT, oven aging and real time storage of up to 23 years, yielded one of the most complete data sets covering over 8 decades of time. When the activation energies from thermal processes were compared with the rate of surface embrittlement, a striking self similarity, or near identical activation energies at the same temperature were evident. This observation could lead to broader applications and further understandings on the polymer degradation.
REFERENCES 1.
G. N. Foster, in Oxidation Inhibition in Organic Materials, Vol.2, J. Pospisil, P. Klemchuk eds., CRC Press, Boca Raton (1989).
Activation Energies of Polymer Degradation
2. 3. 4. 5.
225
L. Matisova-Rychla and J. Rychly, in Polymer Durability, R. Clough, N. Billingham, and K. Gillen Eds, Am. Chem. Soc., Washington, DC (1996). L. Woo, J. Palomo, T.K. Ling, E. Chan, C. Sandford, Medical Plastics and Biomaterials, 3, (2), 36 (1996). H. E. Bair, Thermal Characterization of Polymeric Materials, p. 869, E. Turi Ed., Academic Press, New York, (1981). N. R. Legge, G. Holden, and H. E. Schroeder, eds, Thermoplastic Elastomers, Hanser MacMillan, New York, (1987).
Estimation of Time-temperature-collectives at Describing Ageing of Polymer Materials
D. Blaese and E. Schmachtenberg University of Essen, Institute for Plastics in Mechanical Engineering (IKM), Altendorfer Str. 3, 45127 Essen, Germany
INTRODUCTION The development of innovative products by using plastic materials represents a great challenge in respect to the material selection for the new product. The selection of a suitable material for a given application is not only influenced by the construction of the product and the processing of the material but also essentially influenced by the operating conditions of the product. Figure 1. Classification of thermoplastic materials. One problem at the material selection is the large number of different obtainable plastic materials. Only for thermoplastic materials more than 10,000 trading products are available. The prices for these materials differ between 1 $/kg (e.g. PE) and 100 $/kg (e.g. PAEK). Closely associated with these costs for the raw material the properties of these plastic materials are very different. One example for this connection between price and property of a material is illustrated in Figure 1 for the property of heat resistance. This connection shows that the production costs of technical products of plastic materials are the main reasons for a systematic material selection. An evaluation of the production costs for an exemplary product which is produced by the injection molding process the costs of the material amounts to 55% of the production costs (Figure 2). This percentage of the material cost depends on the applied material and also on the output of the product.
228
Plastics Failure Analysis and Prevention
Figure 2. Production costs (example: injection molding, polyamide, weight 190 g, 180,000 pieces a year).
This points out that a systematic material selection is essentially important for the eco- Figure 3. Decrease of strength vs. time for polyethylenepipe. nomic production of plastic products as well as for the suitability for the application. Furthermore plastic materials age under operating conditions. The loading of a product for example is given by the influence of a medium, a temperature and also a mechanical loading. Figure 3 shows the decrease of strength of polyethylene pipes. In this figure three different sections with three different gradients can be distinguished. The reason for these sections are the three different damaging processes of polyethylene by loading under temperature and a medium. This causes that the ageing of a plastic material is essentially influencing the lifetime of a product. Because of this it can be shown that the ageing and lifetime of a product is an essential aspect in the course of the material selection which has to be taken into consideration. Especially the influence of a medium can cause different interactions with the material and combined with this the influence of a medium effects different failure mechanisms of the product. It has to be taken into consideration that the different mechanisms cannot be analyzed separately. On the contrary to this a superposition of these effects is the result and they are influencing each other. But how to analyze the lifetime of a product and the suitability of the material for the given application after preselection of a plastic material?
TIME-TEMPERATURE-EXTRAPOLATION WITH ARRHENIUS Generally such an estimation of lifetime can be conducted by suitable and praxis relevant product tests. Hereby it has to be guaranteed that the complexity of the application also can be reproduced in the product test. It is very important for the concept of these tests that a characteristic product value can be measured. This characteristic product value represents a
Estimation of Time-temperature-collectives
229
scale for the damaging of the product and depends on the loading of the product. In practice a method is established which estimates the limits of application by time-temperature-extrapolation of the measured damaging processes. So a lifetime prediction is possible by using the time-temperature-shifting-principle (Figure 4). This time-temperature-shifting-principle says that the influence of a high temperature for Figure 4. Time-temperature-shifting-principle. a short time causes a comparable damaging as a low temperature for a long time. One possibility to describe the time-temperature-shifting is given by the statement of Arrhenius. This is shown in equation 1: t - = 10 -----tref
1 1 k --- – -------- T T ref
[1]
In this statement t stands for the required lifetime at a temperature T and tref stands for the testing time at the testing temperature, Tref. Thereby the shifting factor k is the only unknown. This factor contains the influences of the used material, the given operating conditions like the medium and the loading as well as the influence of the construction. The determination of the shifting factor k can be determined by the described product tests. For at least two different temperatures at different testing times products has to be taken from the test stand and the characteristic product value has to be determined. For two different temperatures two products are inspected which show an identical damaging according to the measured characteristic product value. This is equivalent for an identical damaging of the product in account of the ageing of the material. With these two pairs of values with the following equation 2 the shifting factor k can be determined. t1 log ---- t2 k = ----------------------1 – ----1 ----T T 1 2
[2]
With the determined shifting factor k the time-temperature-shifting can be calibrated for a given application. This shifting factor k is only valid for the given application, the used material and the examined product. A transfer to other products is only limited possible because of these reasons.
230
Plastics Failure Analysis and Prevention
By using the statement of Arrhenius for time-temperature-shifting it has to be noticed that this statement is only valid in certain limits. These limits are given by the equilibrium temperatures of the plastic materials as the melting temperature or the glass transition temperature. Beyond it an extrapolation should not be conducted over a too large temperature range. The insecurity of the results increases with increasing difference between testing temperature and operating temperature.
ESTIMATION OF TIME-TEMPERATURE-COLLECTIVES In practice a loading under a constant temperature for the whole lifetime is not realistic. On the contrary a complex time-temperature-profile is given (Figure 5). In this case it is possible to suppose the highest temperature for the whole lifetime as the critical temperature for the strength of the product. This leads to a high demand on the properties of the used material. So this procedure leads to a large over-dimensioning of the product and combined with this to a very uneconomic solution of the product idea. Figure 5. Time-temperature-profile. A second way to consider such time-temperature-profiles is given by an estimation of the separate time-temperature-collectives with regard to their contribution of ageing. For this the following procedure is imaginable: 1. An "influence" is defined as a specified time-interval during which a specified medium with a constant temperature and under a constant loading is influencing the product. 2. This influence effects an "ageing" of the product. Different influences can be taken into consideration for the determination of ageing of the product by addition of the different parts. So an "ageing progress" is given by an influence. 3. If the ageing is advanced to a critical state the product changes its properties (e.g. decrease of strength, change of the molecular structure of the product or other). 4. The estimation of the different parts of ageing caused by an influence is conducted with the help of the time-temperature-shifting-principle. For each influence a reference-time would be calculated which causes a comparable ageing progress at reference temperature. By addition of the separate reference times a total testing time for the product can be evaluated. This testing time at reference temperature
Estimation of Time-temperature-collectives
231
Figure 6. Percentages of time and ageing.
effects a comparable damaging of the product Figure 7. Internal pressure strength for describing the damage. as it will be caused by the influence of the timetemperature-profile for the demanded lifetime. The example described in the following explains this principle and shows at the same time the measurement of a suitable indicator-property for products loaded with an internal pressure. For this product which is predominated loaded with a temperature, a medium and an internal pressure the time-temperature-profile is shown in Figure 5. With equation 1 it can be determined that with assumption of a maximum temperature of 130°C for the total lifetime, a testing time of about 300,000 hours at a testing temperature of 120°C results. By taking into consideration a maximum temperature of 82°C the result is 720 hours which is still a long testing time. With the principle of estimation of time-temperature-collectives as described above the testing time can be reduced to 107 hours at a testing temperature of 120°C. Thereby the separate time-temperature-intervals are separately inspected. For each interval a testing time at testing temperature can be estimated with Arrhenius. With the described procedure the ageing progress can be estimated. The addition of these ageing progresses is the real damaging caused by the time-temperature-profile. The percentages of time and ageing evaluated with this principle are shown in Figure 6. By this it can be illustrated that especially the high temperatures are very critical and that they cause a high percentage of ageing although they effect only for a short time. Because of this it is obvious that a dimensioning with this maximum temperature for the whole lifetime leads to a large over-dimensioning of the product. The characteristic product value to describe the damaging of the product caused by the ageing progress of the material is measured with a so-called "bursting test". This means that the parts are tested with a raising internal pressure up to a failure. The bursting strength is an indicator for damage. A cor-
232
Plastics Failure Analysis and Prevention
responding pressure-time-curve for this internal pressure strength is given in Figure 7 for the examined product. For examination of the above described theory such as pressure-timecurve was measured for a special product. A few parts of this product are operating under real conditions for a determined time. The internal pressure Figure 8. Comparison between calculated and measured damaging. strength of these parts can be compared with the measured values of testing (Figure 8). The operating time can be converted with the statement of Arrhenius and so these results can be compared with the results of the product tests. The results of this comparison illustrate that the parts under the real operating conditions show a higher internal pressure strength than evaluated with the described theory. This shows that the estimation of time-temperature-collectives is an estimation to the secure side. On the other hand with this theory a large over-dimensioning can be avoided.
Chapter 7 Test Methods
Standard Test Procedures for Relevant Material Properties for Structural Analysis
Gerald G. Trantina and Joseph T. Woods General Electric Corporate Research and Development, Schenectady, NY 12301, USA
INTRODUCTION Engineering thermoplastics exhibit complex behavior when subjected to constant, increasing, or cyclic mechanical loads. As these materials begin to be used more in load-bearing designs, engineers must be able to predict the structural performance of actual molded parts. However, the necessary material properties to do this are usually not available. While standard data sheet properties can be useful for initial material selection, they are inadequate to predict the structural performance of a part. And even when the necessary engineering data exist, they are usually not measured at the same time, strain rate, temperature, or stress as those of a particular application. The structural analyst’s task is to predict the performance of a design at end-use conditions in terms of both operating temperature and loading (constant, increasing, or cyclic). To do this, two types of information are required: data to perform structural analysis calculations and data to assess performance. The material properties required for structural analysis of thermoplastic components will be presented. The shortcomings of data sheet data will be overcome with standard test procedures that provides relevant material properties for structural analysis. There is a significant amount of activity focused on developing standards for plastic materials test data. Certainly, inconsistent test procedures and material data reduce the credibility of the plastics industry. Adoption of uniform international standards will address this problem. However, to the design engineer, the usefulness of the data is also extremely important. If the data is dependent on the test specimen geometry or the type of loading, it cannot be used to predict part performance. Issues that will be addressed in this paper
234
Plastics Failure Analysis and Prevention
include the stiffness and strength of fiber-filled materials, ductile-brittle behavior of unfilled materials and time-dependent material behavior. This material behavior will be described in the context of how the data is used to predict structural performance of parts. For the design engineer, relevant material properties are the key to successful thermoplastic applications with short product development times.
STIFFNESS AND STRENGTH An accurate characterization of the strength and stiffness of glass-filled thermoplastics is necessary to predict the strength and stiffness of components that are injection molded with these materials. The mechanical properties of glass-reinforced thermoplastics are generally measured in tension using end-gated, injection-molded ASTM Type I (dog-bone) specimens. However, the gating and the direction of loading of these molded specimens yields nonconservative stiffness and strength results due to the high axial orientation of glass fibers that occurs in the direction of flow (and loading) during molding. Previous studies1,2 have shown that injection-molded, glass-reinforced thermoplastics are anisotropic values of stiffness and strength in the cross-flow direction are substantially lower than in the flow direction. The tensile stiffness and strength were measured by using dog-bone specimens that were cut in both the flow and cross-flow direction from edge-gated plaques of various thicknesses. The cross-flow tensile modulus and strength of 30% glassfilled materials is approximately 60% of the flow properties. Table 1 illustrates this for tensile and flexural strength. However, before the issues of predicting part strength need to be addressed, the part must be designed for stiffness. Table 1. Average tensile and flexural strength of 152 x 381 x 3 mm fan-gated plaques Material
Testing direction
Tensile strength, MPa
Flexural strength, MPa
Flexural/tensile strength ratio
PBT + 30% glass fiber
Flow Cross-flow
125 71
119 111
1.59 1.56
PC + 30% glass fiber
Flow Cross-flow
122 78
194 114
1.59 1.46
PPO/PA + 30% glass fiber
Flow Cross-flow
143 92
209 108
1.46 1.17
PA + 33% glass fiber
Flow Cross-flow
169 96
236 120
1.40 1.25
Standard Test Procedures
235
Table 1. Average tensile and flexural strength of 152 x 381 x 3 mm fan-gated plaques Material
Testing direction
Tensile strength, MPa
Flexural strength, MPa
Flexural/tensile strength ratio
PC + 15% glass fiber
Flow Cross-flow
87 68
130 102
1.49 1.50
PPO/PA + 10% glass fiber
Flow Cross-flow
84 80
110 101
1.31 1.26
A simple example of part stiffness for fiberfilled materials is shown in Figure 1. Structural stiffness is dependent on material properties, part geometry and for fiber-filled materials, the gate location - where the polymer enters the mold. In Figure 1 the deflection of a 50 cm (side A) by 25 cm (side B) rectangular plate is shown as a function Figure 1. 30% glass-filled PBT plate (50 cm x 25 cm) of plate thickness. The plate is loaded with a presdeflection versus thickness for edge gates with orthotropic analysis and isotropic analysis with flow modu- sure of 0.5 kPa and is fixed on all four sides. The lus (data sheet). material is a 30% glass-filled PBT. The upper curve is for a plate with an edge gate on side B and the lower curve for a plate with and edge gate on side A. The linear structural analysis is performed with a standard orthotropic material model.1 It should be noted that the plate is stiffer with the gate on the long side (side A) since the short side (25 cm) controls the plate deflection and the flow direction and the stiffest material behavior is in the short side direction. Also, an isotropic analysis would produce a nonconservative, erroneous result. For example, for a 3 mm thick plate, the plate deflection for a side B gated plate is 0.52 mm and for a side A gated plate is 0.32 mm (Figure 1) using an orthotropic analysis while the deflection using an isotropic analysis with the flow direction modulus (data sheet) would be 0.26 mm (a 19% error compared to the optimum gate A configuration). Next, the strength of parts molded with glass-filled materials must be considered. While this is a complex failure analysis problem that is beyond the scope of this paper, an important observation about the relationship of the tensile strength to the flexural strength should be noted. For unfilled materials, the tensile strength is 2/3 of the flexural strength. This relationship holds because the procedure for computing the flexural strength is elastic and does not account for the fully plastic behavior of the bending test. A simple accounting for this plastic behavior produces a factor of 2/3 to be multiplied times the simple elastic
236
Plastics Failure Analysis and Prevention
beam bending equation. Reference 3 provides the mathematical and mechanical justification of this 2/3 factor. This is validated by noting that for over 800 unfilled resins the flexural strength is 1.56 times the tensile strength.3 Thus, the design engineer should use the tensile strength or multiply the flexural strength by 2/3. For fiber-filled materials there is a significant reduction in the ductility with strains-tofailure of 3-4% versus roughly 100% for unfilled materials. However, the matrix resin can achieve very large strains and significant plasticity even though the overall strains may be only 3-4%. Also, the stress-strain curves become nearly flat prior to failure indicating gross plastic behavior. Therefore the same mechanics concepts applied to unfilled resins3 might be applied to fiber-filled resins. Table 1 summarizes the ratio of the flexural to the tensile strength of six resin/fiber systems. The average ratio is 1.42. Thus, there is substantial evidence that the tensile strength rather than the flexural strength should be used for failure predictions of parts molded from fiber-filled materials or the flexural strength should be multiplied by 2/3. However, as mentioned earlier, the details of the failure prediction method is complex1 and is beyond the scope of this paper.
DUCTILE-BRITTLE IMPACT The design engineer is continually challenged to predict part performance under high loading rates and sometimes low temperatures. Of particular concern is the identification of combinations of stress state, strain-rate and temperature that lead to brittle part failure. It is well-known that triaxial stress states created by notches, holes, fillets and thick sections increase the potential of brittle failure. Unfortunately there are no simple methods to predict brittle failure of thermoplastic parts. Standardizing on the Charpy or Izod impact test is not going to improve this situation.3 However, there are new approaches useful to the design engineer.4 Most unfilled engineering thermoplastics exhibit ductile behavior in tensile tests with increasing strength as strain rate increases (Figure 2) and/or temperature decreases. However, parts typically have complex geometry with local areas of triaxial stress states. The controlled testing of notched beams with various notch radii and beam thicknesses, where the load-displacement response is measured for various loading rates and temperatures, provides useful information for 4 Figure 2. Yield stress at room temperature versus a failure criteria for plastic parts. These tests also strain rate for PC, PC/ABS and ABS. illustrate the extremely limited usefulness of Izod or Charpy impact tests.
Standard Test Procedures
237
Geometries producing tensile, triaxial stress states having a large maximum principal stress components are more susceptible to brittle failure. If brittle failure is a possibility for the material, then the actual stress state Figure 4. Strength ratio versus strain rate for PBT at o produced within the -30 C. part and the material’s failure criterion must be considered. One way of characterizing the susceptibility of a geometry to brittle failure is to numerically calculate the geometry’s peak maximum principal stress level and divide it by the yield stress of the material. This ratio is Figure 3. Geometric severity ratios referred to as the “geometric severity ratio.” Since the geometric for typical part features. severity ratio normalizes the peak, maximum principal stress level in the geometry by the yield stress of the material used in the analysis, it represents a purely geometric factor. Note that the geometric severity ratio corresponds to a worst case loading condition, i.e., a load level that results in the maximum possible level of maximum principal stress in the part assuming elastic-perfectly plastic material behavior. Geometries having larger geometric severity ratios are more likely to fail brittlely. Figure 3 provides a list of geometric severity ratios for common, generic geometries. In many geometries, local stress levels will be affected by geometric parameters such as radius, thickness, etc. For these geometries a geometric severity ratio range is provided. Using Figure 3 geometric severity ratios can be chosen which represent features of the part being designed. These geometric severity ratios can then be compared to a strength ratio, the material’s critical maximum principal stress to yield stress ratio, i.e., its failure criterion, at the appropriate rate and temperature (Figure 4) to determine if brittle failure is possible for an arbitrarily high structural load. If brittle failure is a possibility, then maximum principal stress levels must be kept below those required to initiate failure by adjusting local geometric parameters or by limiting the loading on the structure.
TIME AND TEMPERATURE DEPENDENT DEFORMATION For the design engineer, the purpose of measuring time-dependent deformation is to obtain useful information that can then be used to predict time-dependent part performance.5 The
238
Plastics Failure Analysis and Prevention
important consideration here is the type of loading-tension versus flexure and constant load versus constant displacement. In the creep test, a specimen is rapidly loaded and then the load is held constant at a constant temperature. For a tensile specimen, the strain should be measured with an extensometer in the gauge section where the stress is uniform. For the stress relaxation test, the specimen is rapidly loaded and then the displacement or strain is held constant. The creep test is more common and probably simpler since dead weight loading can be used with multiple creep stations. The uniaxial tensile test is considered most useful for producing accurate, consistent results that can be easily interpreted. An extensometer is used to measure uniform strain accurately, and the stress is uniform and equal to the load divided by the specimen’s net-section area. Flexural (bending) creep tests have been widely used with polymers. However, in the flexural test, linear-elastic, time-independent beam equations are used to calculate the bending stress using the applied load. Unfortunately, because of the time-dependent, nonlinear, stress-strain response of thermoplastics, the simple bending equations are often inadequate. Constant stress is not maintained because of stress redistribution - the stress distribution is not linear through the thickness of the beam. The heat deflection temperature (HDT) is a common measure of heat resistance in the plastics industry. Such a test, which involves variable temperature and arbitrary stress and deflection, is of no use in predicting the structural performance of a thermoplastic at any temperature, stress, or time. In addition, it can be misleading when comparing materials. A material with a higher HDT than another material could exhibit more creep than the other material at a lower temperature. For purposes of predicting part performance and for material selection, tensile creep data is the desired measurement. Since load-bearing applications are designed for stiffness and strength, time-strain data are not directly useful. However, a curve-fitting interpretation of these data that retains the important features common to many of the creep models - log-time representation and Arrhenius relations for temperature - is useful to the design engineer. In this section, the steps required to translate the initial strain-time data to an isochronous stress-strain curve for any time or temperature will be described. Several methods for fitting and extrapolating time-strain data could be used. The objective should be to obtain the most accurate fit while achieving reasonable extrapolation predictions with minimum complexity. It has been determined based on visual inspection, that a second-order polynomial function in log time can be used where: 2
ε t = A ( log t ) + B ( log t ) + C
The constants are determined by a least-squares curve-fitting method. An example for a PC/ABS blend tested at 10.3 MPa and 50oC is shown in Figure 5. Engineering judgment must be used concerning the appropriate extrapolation in time. Caution should be exercised
Standard Test Procedures
Figure 5. Strain versus time for PC/ABS with a constant stress of 10.3 MPa and temperature of 50oC.
239
Figure 6. Isochronous stress-strain curve for PC/ABS at 40oC and 100 hours.
when more than one order of magnitude of time extrapolation is used. Based on fundamental principles for thermally activated processes, use of the Arrhenius relation is reasonable to interpolate and extrapolate with temperature. Plotting the strains versus temperatures for a particular time and stress on a natural logarithm (ln) of strain versus the inverse of the absolute temperature graph results in a linear interpolation or extrapolation based on the Arrhenius relation. Finally, the isochronous stress-strain curve is produced by choosing the appropriate temperaFigure 7. Effective modulus versus stress for PC/ABS at ture and plotting the stress-strain points taken 40oC and 100 hours. from the Arrhenius plots at that temperature and the previously chosen time. For example, for the PC/ABS blend for 100 h and a temperature of 40oC, the isochronous stress-strain curve is produced (Figure 6). For constant stress applications, the isochronous stress-strain curve can be used with standard equations by choosing the appropriate “effective modulus” considering the range of stresses in the application. This requires engineering judgment where higher stressed parts would typically be analyzed with a lower “effective modulus.” For example, Figure 7 shows the effective modulus calculated from the isochronous stress-strain curve (Figure 6)
240
Plastics Failure Analysis and Prevention
as the secant modulus. The use of this modulus based on the maximum stress in the part should provide a conservative estimate of the time and temperature dependent deflection of the part. When the isochronous stress-strain curve is highly nonlinear or the part geometry is complex, finite-element structural analysis techniques can be used. Then, the complete nonlinear, isochronous stress-strain curve can be used in a nonlinear finite-element analysis or a linear effective modulus can be used in a linear analysis.
CONCLUSIONS Uniform standards for measuring mechanical properties of plastics will lead to consistent test data. However, for the design engineer, the usefulness of the data in predicting part performance is also very important. If the data is dependent on test specimen geometry or the type of loading, it cannot be used to predict component behavior. Data reported for flexural strength of unfilled and fiber-filled thermoplastics is about 50% greater than the tensile strength. This is simply a miscalculation of the flexural strength since elastic beam equations are used for the nearly fully plastic behavior of thermoplastics. The proper flexural strength calculation would be 2/3 times the elastic beam equation. The flexural strength would then be about the same as the tensile strength. Also, gate location and its effect on part deflection for fiber-filled materials can be treated with an orthotropic stress analysis. Izod and Charpy impact tests are simply two different beam bending loads (cantilever and 3-point bending) yielding single-point information. To predict ductile/brittle behavior of parts, geometry can be captured with a geometric severity factor and compared to the strength ratio - the material’s critical maximum principal stress at the appropriate strain rate and temperature divided by the yield stress. Creep data displayed as “effective modulus” graphs provide useful geometry-independent design information for situations where time and temperature are important. HDT is simply a geometry and loading dependent temperature that is not useful for design purposes. The plastics industry must strive to develop standard mechanical tests that are independent of specimen geometry and thus useful to the design engineer who is responsible for part performance.
REFERENCES 1 2 3 4 5
Ambur, G. and Trantina, G., “Structural Failure Prediction with Short-Fiber Filled, Injection Molded Thermoplastics,” Society of Plastics Engineers, 1988 ANTEC Conference Proceedings, pp. 1507-1511. Stokes, V.K., Inzinna, L.P., Trantina, G.G., Liang, E.W., and Woods, J.T., “Mechanical Properties of Long-Fiber Filled Injection-Molded Thermoplastic Composites,” 1994 ANTEC Conference Proceedings. Trantina, G.G. and Oehler, P.R., “Standardization - Is It Leading to More Relevant Data for Design Engineers,” 1994 ANTEC Conference Proceedings. Woods, J.T. and Nimmer, R.P., “Design Aids for Preventing Brittle Failure in Polycarbonate and Polyetherimide,” 1996 ANTEC Conference Proceedings. Trantina, G.G. and Nimmer, R.P., Structural Analysis of Thermoplastic Components, McGraw-Hill, New York, 1994.
Factors Affecting Variation in Gardner Impact Testing
Mark Lavach Elf Atochem North America
INTRODUCTION 1
Previous work concluded that the Gardner Impact Test is useful to find the MFE for brittle thermoplastics such as acrylic and HIPS with standard deviations between 8% and 10%. Standard deviations of 15% were found for more ductile materials such as ABS or PC. Major sources of variation included the mounting of the apparatus (floor vs. bench), material quality, test temperature, and the operators’ definition of failure. In an independent study,2 Paxon Polymer found that testers mounted on standard laboratory benchtops yielded higher MFE's than those bolted directly to the floor. We have also observed that standard laboratory bench tops can crack after repeated use of the tester. This could seriously compromise the accuracy and precision of the test. More recently, the Vinyl Siding Institute (VSI) in their exhaustive round robin testing using PVC siding found residual standard deviations on the order of 15%.3 In a series of recent tests, we examined the effect of supporting table mass on the MFE of embossed and non-embossed PVC siding, along with several other typical thermoplastics. We hypothesized that the heavier supporting tables would absorb less impact energy due to less flexural or compressive deflection of the supporting table. The net result would be a decrease in the samples’ MFE. Further, we believed that the precision of the test could be improved. Additionally, we looked into the effect of reducing the drop height increments to increase the test's precision.
EXPERIMENTAL In many installations, the Gardner Tester is bolted to a table facilitating operation of the equipment by placing the equipment at an ergonomically friendly height. In our labs, we studied the effect of bolting the tester to tables of varying weight, ranging from 57 lbs to 440 lbs. The 57 lb table had a wooden top with steel frame supports. Heavier tables were constructed entirely of wood and could best be described as "butcher block" tables. These tables were not bolted to the floor. The testers were also fixed to the floor in an attempt to represent a supporting table of infinite mass. Initial studies were performed by a single operator
242
Plastics Failure Analysis and Prevention
using non-embossed PVC siding sourced from a single box. Siding was cut into 2"x2" squares, and always taken from the same siding panel location. This material would also be used in all subsequent testing and as a control in tests involving other materials. MFE's were calculated using the Bruceton Staircase or "Up and Down" method.4 Twenty samples were used for each calculation. Tests were run in pass/fail mode, and no attempt was made to find the brittle/ductile transition point for each of the plastics. Additionally, 20 samples were run for each test configuration to allow us good approximation for the test's starting point. Total variability studies incorporated 3 different operator teams who ran duplicate tests (20 samples per test) using two different testers on the three different tables.
RESULTS AND DISCUSSION Figure 1 shows results from a single operator test of non-embossed PVC siding. These results a fall off in the siding's MFE with increasing table mass. Statistical analysis of the data suggests that MFEs obtained using the 57 lb table were significantly different from those obtained with the 440 lb table. Both the floor supported and 440 lb table supported testers gave similar results. Twenty tests (400 samples) were then replicated on the 440 lb table to determine test variation. Using this test design, we found variaFigure 1. MFE vs. base weight (single operator). tion in the MFE of the tested siding to be about 11%. Variation in the MFE was not reduced during the testing cycle indicating that prior knowledge about the test’s starting point had no influence on reducing overall test variation. Differences in failure type classification by test operators could be an important source of test variation. Using the same non-embossed siding, we selected 3 teams of operators who ran duplicate tests on each of the three tables using two different testers. For this study, sample thickness variation was less than 2% of the mean thickness which should have minimal impact on test results. Using an ANOVA (Analysis of Variance) based experimental design, we could not only determine differences in supporting table results, but also look at differences between the operator teams and test equipment. Interactions, significant differences, and test variation could also be evaluated. Mean Failure Energy results are shown in Figure 2, where once again our initial hypothesis was supported: that is, increasing supporting table mass decreases a sample’s MFE. Differences in MFEs between the lightest and heaviest tables were again considered statistically significant. Differences between the 440 lb table and the floor were considered statistically insignificant.
Factors Affecting Variation in Gardner Impact Testing
Figure 2. MFE vs. base weight (multiple operator).
243
Figure 3. Gardner test equipment bias.
Use of ANOVA based analysis also allowed us to investigate other contributions to variations Table 1. Tup and weight effects in MFE determinations. While base mass was found to have the greatest effect on MFE, statisTable wt, Weight MFE, Tup # tically significant differences were found lbs # in-lbs between the two testers and the three different 215 1 1 120.9 operator teams. There was significant equipment bias between the two testers (Figure 3). Mass 215 1 2 118.5 measurements of both the tup and falling weight 215 2 1 97.5 showed only small differences between the items. Falling weight guides were of identical 215 2 2 96.3 radius, and there appeared to be no interference 440 1 1 114.9 in the guides. Using the two heaviest tables (215, 440 lbs), tups and weights were exchanged 440 1 2 112.9 between the two testers. This testing revealed 440 2 1 98.0 that use of one tup repeatedly resulted in lower MFEs regardless of other equipment changes. 440 2 2 96.5 This indicated that the equipment bias was correlated to differences in the tup (Table 1). Both tups had smooth, round surfaces of similar radii. Higher MFEs were found with the older of the two tups, which could be a result of strain hardening in the tup. Differences in MFEs obtained by the various teams were also found. One team consistently observed lower MFEs than the two other teams. These results confirm the SPI findings that failure identification differences between operators can influence test results.
244
Plastics Failure Analysis and Prevention
We also looked at the effect of the various test variables (table, team, and tester) on the residual standard deviation of the test. Residual standard deviation (RSD), or coefficient of variation (COV), is defined as the mean divided by the standard deviation,5 and is useful for comparing means of different magnitudes, units, or test conditions. RSD normalizes the standard deviation allowing for quick comparisons of variation over a wide range of test conditions and impact strengths. Using ANOVA based analysis we found the residual standard deviation to be independent of the tests’ controlled variables (table, team, tester). Residual standard deviations from these tests were well in line with the estimated test variation observed in the single operator control studies. Following these tests we evaluated the effects of supporting table mass on the MFE of embossed PVC siding. Embossed siding, which helps simulate a woodgrain appearance in PVC, also introduces an impact “flaw” into the siding surface. This flaw acts as a stress concentrator similar to a notch in a pendulum impact test. In general, the net result is a reduction in the part’s impact resistance. For the samples, we quantified the depth of the embossing pattern in an attempt to measure its’ effect on the test. In limited testing, we could not directly quantify the effects of embossing on impact resistance. More samples would be needed for a quantifiable determination. However, increasing base mass again reduced all of the sample MFEs. For one sample the effect was not only significant, but resulted in the sample's MFE falling below the 60 in-lb minimum standard for PVC siding.6 Differences were not only significant, but also important. Too often we talk of test differences as being statistically significant, but offer no opinion on their relative importance. Table 2. Effect of Base Weight on Thermoplastic MFE Supporting Base Weights, MFE (in-lbs) Plastic
Supplier
Tup type 57 lbs
215 lbs
440 lbs
ABS
C.125
273
256
246
HDPE
C.125
190
185
184
HIPS
C.125
179
164
170
PP (0.077”)
A
H.250
121
120
110
PP(0.121”)
B
C.125
196
185
184
H.250
112
102
101
PVC
Factors Affecting Variation in Gardner Impact Testing
245
We also tested a variety of other plastics. For these materials, 2 inch disks of each polymer were used in testing. Samples were prepared via injection molding. For stronger samples a conical tup was used. For all polymers similar trends regarding sample MFE vs. base mass were observed (Table 2). Once again, differences between the floor supported and the 440 lb tester were found to be insignificant. While increasing base mass reduced a sample’s MFE, it had no effect on the residual standard deviation found in the test. Test variation was independent of the supporting table mass. Variation also appeared to be independent of a samples MFE. Tough materials exhibited similar residual standard deviations to brittle ones (Table 3). In general, an operator’s familiarity with the test specimen had little impact on test variation. PVC, which we test regularly in our labs, appeared in the middle of total test variation. To reduce operator variation in failure determinations, careful inspection of some samples with either water or a stainable dye is recommended. Finally, we evaluated the influence of drop height increments on test variation with the 440 lb table. Our Gardner test tower is equipped with 8 in-lb (1") increments. For samples of known MFE, reduction of the drop height increment from 8 in-lb to 4 in-lb (0.5") increments significantly reduced a tests total variation (Table 4). Further reductions in the test increments are not likely to be as beneficial as one could begin to over-control the test. The Bruceton Staircase calculation requires somewhat robust changes in energies to be effective. Minimization of these changes could impact the calculation’s ability to predict a sample’s MFE. Ability to accurately place the falling weight at 2 in-lb (0.25") increments with an 8 lb. falling weight is also debatable. Table 3. Variation in Gardner Test Results Plastic
Test size
Ave. RSD
Std. Dev., in-lbs
ABS
21
5.2
2.8
HDPE
21
4.1
2.0
HIPS
21
9.4
6.4
PP (0.077”)
21
10.7
6.6
PP (0.121”)
21
7.9
6.0
PVC
21
9.1
3.9
246
Plastics Failure Analysis and Prevention
Table 4. Effect of Drop Height Increments on PVC MFE and RSD Increment size, in-lbs
Cases
MFE, in-lbs
RSD, %
4
21
99.2
3.4
8
24
95.5
4.8
CONCLUSIONS We have shown that providing a solid, rigid support with increasing mass will lower a sample’s MFE. Heavy, rigid bases have less deflection and compression than lighter ones. resulting in more energy transferal to the sample during the impact event. Based on statistical analysis, this table should weigh more than 400 lbs and be constructed in a fashion which eliminates possible deflection and compression of the support. The test can also be subject to significant equipment bias, with different testers giving different results. For pass/ fail determinations, operator interpretation of failure could also result in differences in sample MFEs. We would expect that the magnitude of variation would increase for determinations of the brittle/ductile transition point. Increasing base mass had no impact on test variation. However reducing drop height increments reduced the overall test variation. Overall test variation appeared independent of impact energy.
ACKNOWLEDGMENTS The author wishes to thank Elf Atochem North America for its support while conducting this project. Special thanks are given to personnel who participated in the drop testing of the samples as well as to BASF, Dow Chemical, Paxon Polymer Company, Exxon Chemicals and the Vinyl Siding Institute (VSI) who supplied samples for testing.
REFERENCES 1
2 3 4 5 6
SPI/SPD Technical Committee Report, Sheet Producers Division, “Recommended Guidelines for the Standardized Use of the Test Method Described in ASTM D-3029-82 for Determining the Relative Impact Resistance of Custom Extruded Sheet.”, April 1994. Paxon Polymer Company, Private Communication. Vinyl Siding Institute Final Report, “Gardner Impact Round Robin Test Program”, June 1997. Brownlee, K.A., Hodgest, J.L.,Jr., and Rosenblatt, Murray, “The Up and Down Method with Small Samples,” American Statistical Association Journal, Vol 48, 1953, pp 262-277. Havlicek, L.L., Crain, R.D., “Practical Statistics for the Physical Sciences”, American Chemical Society, Washington, DC., 1988, pp.77-78. ASTM D-3679, “Standard Specification for Rigid Poly(Vinyl Chloride)(PVC) Siding.” Annual Book of ASTM Standards, Vol 8.02.94.
Radiation Resistance of Multilayer Films by Instrumented Impact Testing
Robert Wojnarowski, Michael T. K. Ling, Atul Khare, and L. Woo Baxter International, Round Lake, IL 60073, USA
INTRODUCTION In the medical packaging industry, multilayer films are frequently used for a multitude of applications ranging from primary sterile fluids containers to secondary and tertiary protective packaging. Also, terminal sterilization by ionizing radiation has steadily gained popularity for its effectiveness and simplicity. However, frequently material degradations also accompany the sterilization process. This is due to the depletion of antioxidants during and after the irradiation step. In addition, many of the material degradation processes continue long after the radiation treatment. This is due to actions of long lasting free radicals, peroxy and hydroperoxides created during irradiation, as they continue to react with atmospheric oxygen diffusing in from the external of the packaging to propagate the oxidative chain reaction. In this article we will present the development of a radiation resistant multi-layer film with the aid of an instrumented impact tester where material degradation can be quantified. RADIATION DEGRADATION OF POLYMER FILMS Both the sterilizing action and the degradation caused by ionizing radiation are believed to result from Compton secondary electrons from the primary interaction event. The high energy gamma photon or accelerated electrons (from the e-Beam source) first interact with the atom in the material being irradiated, creating a secondary high energy Figure 1. Compton scattering with matter. electron and a recoiling photon or electron, with reduced energy compared with the incoming beam. The cascade is propagated until all the excess energy above the ionization threshold is dissipated. Thus, from a single incoming photon or electron, a shower (Figure 1) of secondary electrons is generated and they are responsible for the bio-burden kill and material degradation.1
248
Plastics Failure Analysis and Prevention
Both thermal analysis and chemiluminescence data support the general mechanism of degradation where the primary alkyl free radicals are propagated through atmospheric oxygen diffusing into the polymer via the formation of peroxy and hydroperoxy free radicals (Figure 2). In this regard, the action of the phenolic antioxidant is mainly that of a hydrogen donor in elimiFigure 2. Oxidative kinetic chain reaction. nating organic free radicals, hence being sacrificially consumed in the process. Catastrophic failures have been reported during the PP shelf life storage period. Intense investigation has come to the following hypothesis that long lived free radicals trapped in the crystalline domains slowly migrate towards the crystalline/amorphous interface where they react with available oxygen to form peroxy and hydroperoxy radicals and initiate degradation near the interface.2-4 When sufficient number of the tie molecules between crystallites are cut through this chain scission process, the elongation of the PP is reduced dramatically and catastrophic failures follow. To confirm that long lived free radicals do play a significant role in the post irradiation PP degradation, a PP film sample was examined by electron paramagnetic resonance (EPR) spectroscopy. A distinct free radical spectra, indicative of the “living” free radical reaction was detected 6 months after an irradiation dose of 25 KGy at a rate of about 10 KGy/hr. This finding confirms the long held view that the free radical mediated oxidative degradation continues in polypropylenes long after the irradiation event. Incidentally, the strong EPR signal was completely eliminated when the sample was annealed in a vacuum oven at 90°C, a temperature much about the glass transition for the amorphous phase for PP, and well into the alpha relaxation (Tm) for the crystalline phase of PP. It is therefore imperative to develop a rapid testing method where physical degradation of film properties after radiation sterilization can be assessed. Since many of the shipping protocols involve winter shipping simulation, where temperatures approaching freezing are frequently encountered, an ASTM low temperature condition would be required. In order to avoid large numbers of samples of actual packaging to be tested, it was decided a film testing methodology be employed. By quantitatively monitoring the real time high speed stress strain behavior under various temperatures, a better understanding on the material behavior can be obtained and a predictive correlation with actual packaging achieved.
Radiation Resistance of Multilayer Films
249
RESULTS AND DISCUSSION A custom designed instrumented impact tester (Figure 3) was constructed based on the load frame of a Dynatup drop weight impact tester. HARDWARE DESCRIPTION An environmental chamber was fitted to the bottom of the impact tower with an opening for the instrumented tup to pass through. An insulating cover is placed on the opening and removed just Figure 3. Instrumented impact tester. prior to the release of the weighted tup. A 15 cm circular film holder modeled after an embroidery hoop was constructed with aluminum with an “O-ring” groove containing a Viton O-ring to elevate the taut film above the inner diameter edge of the holder. In this way, the film is prevented from contacting the relatively sharp radius of the holder and more reproducible data obtained. A multi-sample rack was also constructed so several films can be conditioned in the environmental chamber simultaneously. Temperature control was implemented with a proportional digital temperature controller with a type J thermocouple directly touching the film being tested. Subambient cooling was achieved with liquid nitrogen feeding from a 100 liter dewar through a throttling valve directly into the environmental chamber. In conjunction with the digital temperature controller, temperatures as low as -150°C can be obtained. The mass of the traveling carriage with the impact tup was adjusted and, depending on the drop height, a minimum available kinetic energy of about 30 J was maintained. An optical flag 1.0 cm in width was attached to the drop weight carriage and the tup velocity just prior to impact measured with an optical gate. Typically, impact velocity of about 3 meters per second was used. SOFTWARE IMPLEMENTATION All aspects of data acquisition, user dialog, impact energy calculations, temperature control, and data presentation and archiving were handled by a Lab-View virtual instrument program by National Instruments on a personal computer. The software was constructed in a modular form to facilitate documentation and validation. Briefly, each one of the modules are described below: (1) Temperature control: a digital temperature controller using an Iron Constantan (Type J) thermocouple with an RS-232 interface was used to control temperature and communicate with the main program.
250
Plastics Failure Analysis and Prevention
(2) Velocity sensing: the transit time for the 1.0 cm aluminum optical flag through an optical gate is measured by a timer function on the National Instruments digital I/O board on the personal computer. (3) High speed force sensing: the strain gage output from the impact head is first conditioned by an instrumentation amplifier before conversion by a high speed analog/digital converter on the National Instruments interface board at up to 20,000 conversions/second 50µ second / datapoint). (4) Energy integration: the impact force is integrated against displacement (converted from tup velocity) to obtain the impact energy as a function of displacement, and plotted on a dual vertical axes on the graphical presentation. A typical graphical display is shown in Figure 4. EXPERIMENTAL DESIGN
Figure 4: Sample film impact data at two temperatures.
Three films were selected for low-temperature impact testing: a PP alloy film, a multi-layer film containing the same PP alloy, but with a thin PP layer (about 10% of overall thickness), and a multi-layer film consisting of the same PP alloy and thickness ratio, but with a modified PP layer. The modification of the film construction was the incorporation of a small amount of elastomeric impact modifiers5 into the PP layer, resulting in a “toughened” film. The three films were sterilized at three radiation dosages: 20, 50 and 90 KGys (± 10 KGys). Samples from each dose (including control) were impact tested to determine the effect of the modification. The samples were tested at 5°C with a falling weight of 5.77 Kg at a velocity of roughly 3 m/s. The resulting morphologies and impact energies were compared to determine any differences between the films. IMPACT DATA
Figure 5. PP films - impact energies vs. dose.
Integrated impact energy data for each film is plotted as a function of radiation dose in Figure
Radiation Resistance of Multilayer Films
251
Figure 6. Film impact morphologies.
5. The PP alloy film maintains almost 90% of its impact energy over the range of radiation dosages with no morphological changes. The multilayer film with the unmodified polypropylene layer suffered a dramatic loss in ductility above 2 MRad (or 20 Kgy). By examining the impact energy and morphology of the PP alloy, it is evident that the addition of the PP layer has significantly reduced the impact energy. In addition, the failure morphology of the PP alloy/unmodified PP layer film had transformed from ductile to brittle. Although the bulk of the film layers are quite ductile, the film as a whole is brittle. This is another one of the classic cases where in a multilayer construction, the overall material ductility is governed by the brittle layer. Evidently, when the brittle layer fails, sharp cracks are created in a very short time. At the tip of these cracks very high stress concentrations are localized which lead to very high effective strain rates of loading. Under these high strain rates and highly concentrated local stresses, brittle failures are propagated through the ductile layers. As the data in Figure 5 indicates, the impact energy for the multilayer film with the modified PP layer remained high up to 80 KGy, the upper limit of our experiment. More importantly, the failure morphology remained ductile throughout all dose conditions. Figure 6 shows the different morphologies observed between the PP alloy, the unmodified multilayer film and the modified multilayer film.
SUMMARY AND CONCLUSIONS Three films were evaluated by impact testing at 5°C to assess the physical degradation after radiation sterilization. The films were selected to examine the relationship of a PP layer when modified with the addition of elastomeric impact modifiers. The morphologies and impact energies were examined. It was found that the addition of a PP layer, contributing as little as 10% to the overall thickness of a film, reduces the impact energy significantly and
252
Plastics Failure Analysis and Prevention
changes the morphology from a ductile fracture to a brittle fracture. However, with the addition of the impact modifying elastomer, the film can maintain close to 90% of its original impact strength and remain ductile.
REFERENCES 1 2 3 4 5
K. Gillen and R. Clough, in Irradiation Effects on Polymers. D. W. Clegg and A. A Collyer, Eds., Elsevier Applied Science, New York. R. J. Rolando, "Radiation Resistant Polypropylene: New Development," J. Plastic Film & Sheeting, 9, (4), 326. 1993. L. Woo, J. Palomo, T.K. Ling, E. Chan and C. Sandford, "Shelf-Life Prediction Methods and Applications," Medical Plastics and Biomaterials, 3, (2), 36. 1996. G. Herbert, “The Effect of Molecular Orientation on the Radiation Stability of Polypropylene”, Medical Plastics and Biomaterials, 3, #3, 40-43. 1998. N. R. Legge, G. Holden, and H. E. Schroeder, eds. Thermoplastic Elastomers. Hanser Macmillan, New York. 1987.
Aspects of the Tensile Response of Random Continuous Glass/Epoxy Composites
Okenwa I. Okoli, G.F. Smith Warwick Manufacturing Group, University of Warwick, Coventry CV4 7AL
INTRODUCTION The automobile sector continues to provide significant growth opportunities for polymer composites. Although it is recognized that the growth of plastics consumption in cars in the nineties will be evolutionary rather than revolutionary, there is no doubt that the well proven advantages of polymer composites such as weight savings, corrosion resistance and functional integration are more important than ever for the industry.1 Nevertheless, fibre reinforced composites (FRC) are still regarded as relatively new materials within the mechanical engineering field and often lack the detailed material property data associated with metals. In particular, the use of composites in safety critical applications, leads to uneasiness since the mechanical response in crash applications is not well understood.2 The need for a full characterization of the behavior of fibre reinforced polymer composites under dynamic loading conditions has prompted numerous investigations in recent years.2-12 However, when compared to metals, relatively few studies have been conducted to investigate polymer mechanical properties at high strain rate.13 In addition, the increasing use of fibre reinforced composites has prompted the need to ascertain the fibre contents necessary to provide the essential mechanical properties. In safety critical applications, it is therefor necessary to investigate the effect of increasing fibre content on the energy released or absorbed by the structure. This work set out to investigate the effect of strain rate on failure mechanisms and that of fibre content on energy absorption in random continuous glass/epoxy composites.
EXPERIMENTAL WORK Two materials were tested. The first was a 3 mm thick random continuous glass/epoxy Von Roll Isola composite laminate. The composite had a fibre weight fraction of 65%. The second material were locally manufactured (W.M.G.) random continuous laminates with different fibre volume fractions (15.5, 20.7, 26.9, 38.0 and 41.2%).
254
Plastics Failure Analysis and Prevention
The apparatus and procedure used to obtain the tensile properties have been described elsewhere.2
RESULTS AND DISCUSSION EFFECT OF STRAIN RATE ON FAILURE MECHANISMS The failure mechanisms will be discussed with the aid of photomicrographs. Figure 1 shows a magnified (x 678) fracture surface of a 68.660 laminate tested in tension at a cross-head rate of 1.7x10-2 mm s-1. Fracture is in the direction parallel to fibre reinforcement and as such, distinct river marks can be observed, indicating the direction of crack propagation. These river marks correspond to fracture ridges formed by minutely displaced failure planes.14 There is little evidence of fibre-matrix adhesion, which suggests a poor interfacial Figure 1. Von Roll Isola Grade 68.660 Laminate showing bond. This results in the gaps observed at the fibre-matrix debonding (x678). attached ends of the fibres, and the long fibre pull-outs and indicates composite toughness. However, the visible fibre ends indicate brittle failure in the fibres. Figure 2 shows a magnified (x 439) fracture surface of a 68.660 laminate tested in tension at a cross-head rate of 83x10-2 mm s-1. Catastrophic failure in the matrix with signs of delamination and fibre pull-out can be observed. Catastrophic failure, as reported by Broutman15 is typical in cases of low levels of adhesion in the fibreFigure 2. Von Roll Isola Grade 68.660 Laminate showing catastrophic failure in matrix with signs of delamina- matrix interface. This kind of failure results in tion and fibre pull-out (x439). high energy absorption, with the occurrence of multiple delaminations and does not allow for significant fibre failure. However, the increase in tensile strength at this loading rate (see Table 1), is due to the increased strength of the glass fibres with strain rate. It has been demonstrated, that the tensile modulus of elasticity,16 and tensile strength,17 of glass fibres increases with strain rate. It then follows that the observed rate dependence of the failure strength follows from the increased strength of the glass fibres. In consequence,3 the energy
Aspects of the Tensile Response
255
involved in the failure of the FRC specimens as determined from the area under the stressstrain curve, increases with strain rate. River marks are also visible, showing the direction of crack propagation in the matrix. Table 1. Tensile property data of 3 mm thick Von Roll Isola grade 68.660 glass/ epoxy laminates at low strain rates Cross-head, mm s-1 (x10-2)
Tensile, MPa
Strain rate, s-1 (x10-3)
Log of strain rate, s-1
1.7
280
14.8
-3.8300
8.3
301
73.2
-3.1357
17.0
304
100.0
-3.0000
83.0
318
700.0
-2.1549
Table 2 The energy to failure of the W.M.G. manufactured random continuous Glass/Epoxy composite laminates obtained at different % fibre volume fractions Volume Fraction (%)
15.5
20.7
26.9
38.0
41.2
Energy (J)
1.641
6.306
7.438
5.254
5.251
EFFECTS OF FIBRE VOLUME FRACTION ON ENERGY ABSORPTION The energy to failure of the random continuous (W.M.G.) laminates obtained at different fibre volume fractions is presented in Table 2. Figure 3 shows the variation of expended energy with fibre volume fraction. The relationship is non-linear, with considerable scatter in the data. Expended energy was found to increase to a peak value, then decrease as fibre volume fraction was increased. This behavior, can be attributed to the failure modes of the composite laminates. It was reported,18 that impact energy increases with fibre volume fraction. However, increasing the fibre content decreases the volume of matrix between fibres, and reduces the inter-laminar strength of the composite.19 Figure 4 shows a magnified (x750) fracture surface of an W.M.G. laminate with 41.2% fibre volume fraction tested in tension at a cross-head velocity of 83x10-2 mm s-1. It was reported20 that composites with fibre volume fractions (40-50%) commonly exhibit brittle failure with fibre pull-out. This trend can be observed in the Figure 4. The broken fibre ends are flat indicating brittle failure, and signs of fibre pull-out can be found in the matrix. Traces of matrix adhesion can be observed on the fibres. Agarwal and Broutman20 reported
256
Plastics Failure Analysis and Prevention
Figure 3. Expended Energy variation of the W.M.G. manufactured random continuous glass/epoxy composite laminates with % fibre volume fraction.
that initiation energy increases with interface strength. This may be applied to the present situation where interface strength has fallen due to increased fibre volume fraction with a resultant fall in energy.
CONCLUSIONS The random continuous (Von Roll Isola 68.660) laminates showed low levels of fibre-matrix interfacial bonding bringing about the long fibre Figure 4. W.M.G. Random continuous glass/epoxy lamipull-outs observed. This resulted in catastrophic nates with 41.2% fibre volume fraction showing matrix exhibiting signs of fibre pull-out and fibres showing little failure with increase in test speed. The foregoing matrix adhesion (x750). suggests that although the fibres fail in a brittle mode, the matrix failure mode is dominant. In addition, increasing the test speed results in catastrophic failure due to enhanced crack propagation rate and an increase in fibre tensile strength. The effects of fibre volume fraction on expended energy were studied for random continuous (W.M.G.) laminates. Energy was found to increase to a peak value with increasing
Aspects of the Tensile Response
257
fibre volume fraction to an optimum value (26.9%) after which further increase in the volume fraction brought about a decrease in energy. The point above which increasing the fibre volume fraction becomes detrimental to energy absorption is considered to be where the flowability (measure of the extent to which the movement of the resin is allowed to fill all parts of the mould) of resin is restricted by the glass fibres resulting in poor wetting and consequently, poor fibre-matrix interfacial bonding.
REFERENCES 1 2
3 4 5 6 7 8
9 10
11
12
13 14 15 16 17 18
M. Sönmez, Plastics Consumption in Automotive Applications, in Automotive Manufacturing International, Sterling Pub. Grp., 1993, (pp. 210-214). O.I. Okoli, G.F. Smith, Overcoming Inertial Problems in the High Strain Rate Testing of a Glass/Epoxy Composite. Proceedings of Society of Plastics Engineers Annual Technical Conference, Advanced Polymer Composites Div., Vol. 2, ANTEC, May 1995, (pp. 2998-3002). J. Harding, L.M. Welsh, A Tensile Testing Technique for Fibre-Reinforced Composites at Impact Rates of Strain. J. Materials Science, Vol. 18, 1983, (pp. 1810- 1826). A.M. El-Habak, Effect of Impact Perforation Load on GFRP Composites. Composites, Vol. 24, No. 4, 1993, (pp.341-345). D. Delfosse, G. Pageau, R. Bennett, A. Poursartip, Instrumented Impact Testing at High Velocities. Journal of Composites Technology and Research, JCTRER, Vol. 15, No.1, 1993, (pp. 38-45). A.M.A. El-Habak, Compressive Resistance of Unidirectional GFRP Under High Rate of Loading, Journal of Composites Technology and Research, JCTRER, Vol. 15, No.4, 1993, (pp. 311-317). O.I. Okoli, High Speed Performance of Composite Materials, in Engineering Polymers Integrated Capability (EPIC) Conference, work area 2d, University of Warwick, UK, March 1996. O.I. Okoli, A. Abdul-Latif, G.F. Smith, The Impact Response of Glass Fibre Reinforced Composites: A Comparison Between Finite Element Results and Experimental Data. Proceedings of Society of Plastics Engineers Annual Technical Conference, Advanced Polymer Composites Division, Vol. 2, ANTEC, May 1996, (pp. 2504- 2509). O.I. Okoli, G.F. Smith, The Effects of Strain Rate on the Failure Energy of Fibre Reinforced Composites. Proceedings of the First International Conference on Composite Science and Technology, Durban, South Africa, June 1996, (p. 359). G. Zhou, Characteristics of Impact Energy Absorption During Damage Development in Laminated Composites. Proceedings of the 4th International Conference on Deformation and Fracture of Composites, Manchester, Institute of Materials, March 1997, (pp. 55-68). O.I. Okoli, G.F. Smith, The Effects of Strain Rate and Fibre Volume Fraction on the Failure Modes of Fibre Reinforced Composites. Proceedings of the 4th International Conference on Deformation and Fracture of Composites, Manchester, Institute of Materials, March 1997, (pp. 77-88). O.I. Okoli, G.F. Smith, Semi-Empirical Relation for the Determination of Dynamic Young's Modulus in Woven Glass/ Epoxy Reinforced Composites. Proceedings of Society of Plastics Engineers Annual Technical Conference, Advanced Polymer Composites Division, Vol. 2, ANTEC, April 1997, (pp. 2373-2376). S.M. Walley, J.E. Field, P.H. Pope, N.A. Safford, The Rapid Deformation behavior of Various Polymers. J. Physics III France, 1, 1991, (pp. 1889-1925). B.W. Smith, Fractography for Continuous Fibre Composites, in Engineered Materials Handbook: Composites, Vol. 1, 1989, (pp. 786-797). P. Yeung, L.J. Broutman, The Effect of Glass-Resin Interface Strength of Fibre Reinforced Plastics. Polymer Engineering and Science, 1978, Vol. 18, PT. 2, (pp. 62- 72). A.E. Armenàkas, C.A. Sciammarella, Response of Glass-Fibre-Reinforced Epoxy Specimens to High Rates of Tensile Loading. Experimental Mechanics, October 1973, Vol 13, (pp. 433-440). L. M. Welsh, J. Harding, Dynamic Tensile Response of Unidirectionally- Reinforced Carbon Epoxy and Glass Epoxy Composites. Proc. 5th Int. Conf. On Composite Materials ICCM V, TMS-AIME,1985, (pp. 1517- 1531). P.K. Mallick, Fibre-Reinforced Composites: Materials, Manufacturing, and Design, Marcel Dekker Inc., New York and Basel.
258
19 20
Plastics Failure Analysis and Prevention
G.L. Farley, R.K. Bird, J.T. Modlin, The Role of Fibre and Matrix in Crash Energy Absorption of Composite Materials. Proc. American Helicopter Society, National Specialists' Meeting, Crashworthiness Design of Rotor Craft, 1986. B.D. Agarwal, L.J. Broutman, Analysis and Performance of Fibre Composites. 2nd ed., 1990, John Wiley and Sons, Inc., New York.
Comparing the Long Term Behavior of Tough Polyethylenes by Craze Testing
KC Pandya and JG Williams Department of Mechanical Engineering, Imperial College, London, UK
INTRODUCTION The time dependent nature of the mechanical properties of polyethylene has been the cause of a number of field failures in commercial pipelines and has been investigated by a number of workers.1-3 Such failures normally occur through the development of long term slow crack growth mechanisms in pipes subject to some form of constant loading during service. Estimating the lifetime of existing pipes and increasing the slow crack growth resistance of new pipelines requires a proper understanding of the structure–property relations that govern the initiation of slow crack growth. This can then be used to improve cost effectiveness by maximizing durability and minimizing the need for replacement. Near the crack tip, under the effect of a high local triaxial stress state, small microvoids may open up and subsequently grow in size and coalesce. A fully formed craze consists of a network of large coalesced voids interspersed between fibrils which are highly orientated in the stress direction. It is now well established that crazing is the precursor to slow crack growth in polyethylene2,4 and that the nature of the separation processes that lead to the breakdown of the craze governs the resistance of the material to slow crack growth. The appropriateness of the choice of experimental method and means of analysis depends on the inherent properties of the material under investigation. In low and medium toughness grades of polyethylene the craze zone may be contained within a K or J dominant stress field allowing conventional fracture mechanics to be used. With increasing optimization of structural properties such as molecular weight and distribution of short chain branches in the toughest grades of polyethylene, a substantial craze zone forms ahead of the crack tip invalidating the use of a single parameter fracture criterion. A new method of craze analysis is presented here using high constraint circumferentially deep notched tensile (CDNT) specimens.5,7 The notion behind the test is that when a specimen is loaded in tension, the deep symmetrical notches develop a highly constrained region within the confined ligament which cavitates and fails thus replicating the worst case damage mechanisms seen in polyethylene pipe in the field. The method involves measuring
260
Plastics Failure Analysis and Prevention
the local separation properties of a craze under both constant load and constant speed conditions and quantifying the long term behavior of different grades of tough polyethylene on this basis. Focussing in this way specifically on the local decohesion within the fracture process zone represents an important departure from mainstream fracture methods that may seek instead to accommodate crazing within a continuum analysis. In recent times a lot of research has been carried out on the analysis of interfacial decohesion through the development of cohesive zone modelling techniques.6 The basis of this method lies in characterizing crack growth in a material through the specification of a local fracture criterion which relates cohesive stresses to separation at an interface. Experimentally measured rate dependent traction – separation curves presented in this paper represent an example of such a criterion. The curves contain all the necessary information pertaining to load transfer and energy consumption mechanisms within the craze and in principle may be applied to the general problem of interfacial decohesion in polyethylene in any geometry under a variety of loading conditions.
TEST PROCEDURE AND SPECIFICATIONS Rectangular specimen blanks were cut from compression moulded plaques supplied by BP Chemicals, placed in a four jaw self centring chuck and notched on a lathe using a single point notching tool. The circumferential notch was then sharpened with a new razor blade. The geometrical specifications of the notched specimens are as follows: Figure 2. Test Procedure under constant load condiL x B x W: 120 x 16 x 16 mm tions. Ratio of ligament to bulk cross sectional area: 1/10 Figure 1. Test ProceAngle of notching tool: 19ο dure under constant speed conditions. The experimental apparatus for tests under constant speed conditions is shown in Figure 1. The specimens were tested on an Instron machine with an extensometer mounted on the specimen to measure the craze extension as the test proceeded. Load was measured by a load cell and the load time trace recorded. Tests were run over a speed range of 50.0 mm/min to 0.005 mm/min. The experimental apparatus for
Comparing the Long Term Behavior
261
tests under constant load conditions is shown in Figure 2. Table 1. Selected structural properties of the test The total extension was meamaterials sured using a linear voltage displacement transducer from Material Density, kg/m3 Mw, g/mol Comonomer/1000C which the craze separation was obtained by correcting for the PEI 940 185,000 4.5 bulk extension of the speciPEII 947 310,000 2.5 men. A data logging system was used to record the extenPEIII 947 290,000 1.5 sion as a function of time as PEIV 954 355,000 0.0 the test proceeded. The grades of polyethylene under investigation are shown in Table 1. PEI and PEII are members of the commonly used generic group of polyethylenes of PE80 and PE100 respectively. PEIII and PEIV are an experimental copolymer and homopolymer respectively.
TEST RESULTS At constant speed, the interfacial holding traction, which is defined using the original ligament area, is seen to vary as the test proceeds, rising from zero along the so called cohesion branch of the traction – separation curve to a maximum value and then falling to zero again along the decohesion branch as the craze surfaces separate. The curve is governed by two properties, the craze (maximum) stress and the break separation which together with the shape of the curve define the work of separation ( γ ). Variations in the structural properties of the test Figure 3. Comparison of traction - separation curves. materials and the nature of their rate dependence are reflected in the differences in the values of these properties and in the general shapes of the curves.7 Traction – separation curves at a speed of 50.0 mm/min and 0.005 mm/min for the four grades of polyethylene shown in Table 1 are compared in Figure 3. For PEI, PEIII and PEIV differences in the shape of the traction – separation curve at high and low speeds indicate a transition from ductile behavior at high speeds to macroscopically brittle behavior at low
262
Figure 4. Comparison of energy - time behavior.
Plastics Failure Analysis and Prevention
Figure 5. Stress - time behavior under constant load and constant speed conditions.
speeds, shown by the much lower value of γ . However the behavior of PEII differs from the other grades at low speeds. This is more clearly seen if γ (the area under the traction - separation curve) is plotted as a function of test speed as shown in Figure 4 and indicates the behavior of PEII at low speeds of a rising energy of separation with falling test speed. The long term behavior of the grades can be analysed under constant load conditions by plotting the net section stress as a function of time to failure. Each grade of polyethylene is seen to have a specific stress – time dependence, with variations in the trends related directly to differences in the structural properties. Figure 5 combines the craze stress obtained under constant speed conditions with the net section stress obtained under constant load conditions as a function of time to failure for each grade. Doing so allows the behavior of each grade to be mapped over failure times ranging from seconds to weeks and for comparisons to be made between the performance of each grade under different loading conditions. There is a clear trend of falling stress with falling test speed under constant speed conditions and with falling load under constant load conditions.
DISCUSSION OF RESULTS In Figure 3 at 50.0 mm/min the break separation for PEI, PEII and PEIII are comparable while the value is much lower for PEIV. In general, for the same test speed craze stress falls with falling density. This may be explained in the following way. Crazing is a voiding and cavitation process. Given that we take the material in each grade to be homogenous prior to craze development the question arises as to where the microvoids nucleate. A possible answer lies in the degree of free volume that is available within a material. A low density
Comparing the Long Term Behavior
263
material is likely to have a higher percentage of free volume than a high density material due to the difference in the degree of packing of the molecular chains. Consequently the stress required to produce a craze is much lower in PEI than PEIV. The curves at 0.005 mm/min in Figure 3 demonstrate the effectiveness of the test method in being able to differentiate between the behavior of each grade in the quasi-static speed range. The craze stress again follows the changes in density. However, in contrast to the high speed tests, the decohesion behavior of each grade at low speeds is dependent more on the structural properties of the molecular chain rather than the density such that differences in the values of break separation between the grades are.greater than at high speeds. PEII has a much higher break extension than the other grades at 0.005 mm/min. Consequently the energy of separation at this speed for PEII is twice that for the other copolymers. It seems that the optimization of the high molecular weight and appropriate chain branching in PEII results in it having the highest break separation. A large amount of energy is thus needed for chains to overcome secondary forces and slip past each other. The disentanglement processes that precede craze breakdown are thus greatly inhibited. In contrast, in the homopolymer PEIV, the extent of stable craze growth is extremely limited indicating the ease of disentanglement of a smooth molecule. Given the high molecular weight of PEIV this also indicates the greater importance of short chain branching over molecular entanglements in inhibiting craze breakdown. The trends for energy of separation as a function of test speed are compared for each grade in Figure 4. PEII occupies an upper energy band across the intermediate and low speed range. All four grades seem to show a high speed (50.0 mm/min to 1.0 mm/min) plateau region. Such high energy ductile failure is accompanied by the formation of a small number of large voids, the growth of which results in a decrease in stress transfer and high energy of separation. A transition occurs in the behavior for all grades at around 0.1 mm/ min. At speeds lower than this PEI, PEIII and PEIV show a decreasing energy of separation with falling test speed. However, for PEII at low test speeds, the energy value rises with falling test speed, governed in the main by rising break separation as was seen earlier in Figure 3. Such behavior is quite contrary to the more generally expected low speed, macroscopically brittle behavior shown by the others. Understanding this behavior has important implications for the development of materials with long term resistance to slow crack growth mechanisms. Stress - time data obtained under constant speed and constant load conditions are shown in Figure 5. The two loading conditions do not predict the same ranking of the grades. For example, choosing time to failure as the criterion for assessment, at a stress of 40 MPa using the constant speed data the grades are ranked in order of descending toughness as PEIV, PEIII, PEII and PEI. However, at 20 MPa using the constant load data the
264
Plastics Failure Analysis and Prevention
Figure 6. Application of measured craze properties to a cohesive zone model using a three point bend geometry.
ranking in order of descending toughness is PEII, PEI, PEIII and PEIV. The short term constant speed data reflect the differences in densities of the grades while the long term constant load data are affected more by differences in molecular weight and branch chain density. It is also clear, bearing in mind the results of the energy speed behavior in Figure 4, that the mapping of stress time behavior alone does not provide sufficient information for the assessment of toughness.
APPLICATION OF RESULTS TO COHESIVE ZONE MODEL A predictive 'cohesive zone' type fracture model is introduced based on the 'Finite Volume' method which incorporates the measured traction - separation curve as a local fracture criterion.7,8 A three point bend specimen is taken as an illustrative example as shown in Figure 6, where using geometrical symmetry only half the specimen is modelled. A measured traction - separation curve is specified along the plane of symmetry and governs the opening of the craze surfaces. Bulk stress-strain properties obtained experimentally from standard plane stress dog bone specimens are also specified. The central premise of this approach is that the measured traction curves are a material property and can be used to model decohesion in any physical situation for a given rate and temperature. Experiments were run at slow rates on three point bend specimens for PEI and PEII. Model predictions of load - time behavior and crack growth from numerical simulations of the three point bend specimens have been found to agree very well with experimental results as shown in Figure 7 for tests run at 0.1 mm/min. This is potentially a powerful method given that the fracture model is based on the mechanics of craze separation, the physical basis of which has been experimentally verified. The prediction of Figure 7. Comparison of experimental and predicted load failure through the examination of interfacial - time curves in a three point bend geometry. decohesion would mean that it should be applicable to many different physical situations. There thus appears to be considerable scope for extending this method to other geometries, including pipe geometries, and other loading conditions in order to obtain quantitative predictions of the initiation of slow crack growth in tough polyethylenes.
Comparing the Long Term Behavior
265
CONCLUSIONS Circumferentially deep notched tensile specimens were used successfully to assess the behavior of tough polyethylenes on the basis of their craze mechanics under constant speed and constant load conditions. Experimentally measured traction - separation curves under constant speed conditions were shown to provide good discrimination between the grades and highlighted the superior toughness of a particular grade of PE100. Results under constant load conditions provided a useful stress - time characterization of the grades and also highlighted in general the inferior performance of homopolymers compared to copolymers. A means of further analysis was introduced by applying measured traction - separation curves to a cohesive zone model. Experimental and numerical results have been found to be in excellent agreement. This indicates the possibility of applying the model to a range of fracture problems including the prediction of the initiation and growth of slow cracks in pipes.
ACKNOWLEDGEMENT The authors thank BP Chemicals for supplying the test materials and for their financial support of this research project. Thanks is also due to Dr. Ivankovic of Imperial College for his advice on the numerical model.
REFERENCES 1 2 3 4 5 6 7 8
Chan M.K.V., Williams J.G., "Slow Stable Crack Growth in High Density Polyethylenes", Polymer, 24, (1983) p. 234 – 244 Bhattacharya S.K., Brown N., "The Initiation of Crack Growth in Linear Polyethylene", J. Mater. Sci., 20, (1985) p. 2767 – 2775 Egan B.J., Delatycki O., "The Morphology, Chain Structure and Fracture Behavior of High-Density Polyethylene", J. Mater. Sci., 30, (1995) p. 3307 – 3318 Friedrich K., "Crazing and Shear Bands in Semi-Crystalline Thermoplastics", Adv. Polym. Sci., 52/53, (1983) p. 225–274 Duan D., Williams J.G., "Craze Testing for Tough Polyethylenes", J. Mater. Sci., 33, (1998) p. 625 - 638 Needleman A., "An Analysis of Decohesion along an Imperfect Interface", Int. J. Fract., 42, (1990) p. 21 - 40 Pandya K.C., Williams J.G., "Measurement of Cohesive Zone Parameters in Tough Polyethylene", Polym. Eng. Sci., submitted for publication. Pandya K.C., Ivankovic A., Williams J.G., "Predicting Crack Growth in Tough Polyethylene from Measured Cohesive Zone Traction - Separation Curves", 11th International Conference on Deformation, Yield and Fracture of Polymers, (2000) abstract submitted.
Chapter 8 Failure Prevention Design Aids for Preventing Brittle Failure in Polycarbonate and Polyetherimide
Joseph T. Woods And Ronald P. Nimmer GE Corporate Research And Development, Schenectady, NY 12301, USA
INTRODUCTION When assessing structural part performance, failure is an important concern. This is particularly true in impact events where impacted parts are required to absorb a certain impact energy without failing. Automotive bumper impacts and drops of electronic enclosures are good examples. Added to this concern is the possibility of two different failure modes, ductile and brittle. In a ductile failure the part fails in a slow, noncatastrophic manner in which additional energy is required to further spread the damage zone. In contrast, a brittle failure is characterized by a sudden and complete failure which, once initiated, requires no further energy to propagate. The failure criteria for the two failure modes differ. Generally, effective stress (von Mises stress) is used to assess when plastic (permanent) deformation has initiated. If some permanent deformation is acceptable, then a strain to failure criterion may be used as the ductile failure criterion indicating when tearing is expected to occur. Brittle failure criterion for many polymers have not been firmly established and can vary from one material to the next. However, maximum principal stress has been identified and validated as a reasonable brittle failure criterion for polycarbonate and polyetherimide. Since brittle failures are more catastrophic, they will be the focus of the remainder of this paper. A brief review will be given describing the brittle failure mechanism for polycarbonate and polyetherimide. Next, the work leading up to the development and validation of brittle failure criterion for polycarbonate and polyetherimide will be reviewed. Building upon this understanding of the brittle failure behavior of polycarbonate and polyetherimide, design aids to prevent brittle failure will be presented and described in the final section.
268
Plastics Failure Analysis and Prevention
BACKGROUND The first step in being able to predict and prevent brittle failure is identifying and understanding the brittle failure mechanism. In polycarbonate and polyetherimide, craze formation and growth has been identified as the brittle failure mechanism. Crazes are similar to internal voids within a material and once initiated act as an internal crack or defect. With continued loading these voids grow and often lead to brittle fracture. In polycarbonate and polyetherimide brittle failures are preceded by the formation of one or more crazes, which develop under large triaxial stress fields. These large triaxial stress fields typically develop subsurface to a local stress concentrator at the boundary of an elastic-plastic zone.1-3 In these areas very large hydrostatic and maximum principal stress levels develop producing the craze.3-4 Having identified the brittle failure mechanism, the next step in predicting and preventing brittle failure is to develop a brittle failure criterion. Since crazing precedes brittle failure in polycarbonate and polyetherimide, a craze initiation criterion can be used as a conservative brittle failure criterion. Several different crazing criteria were considered, including maximum hydrostatic stress, maximum principal stress, and three dimensional extensions of criteria proposed by Sternstein and Ongchin5 and Oxborough and Bowden.6 In the end, a rate dependent, critical, maximum principal stress criterion was chosen based upon its ease of generation, its ability to distinguish between ductile and brittle failure, and its accuracy.7 This criterion was validated in ribbed plate geometries tested over a range of strain rates and in all cases was able to distinguish between ductile and brittle behavior.7 Moreover, brittle failure loads were predicted to within 20% of actual brittle failure loads using this failure criterion. As mentioned above, craze initiation criterion have been established for polycarbonate and polyetherimide in the form of a rate dependent, critical, maximum principal stress criterion. In other words, craze initiation occurs when the maximum principal stress level in a part reaches a critical, rate dependent value, Equation [1]. If maximum principal stress levels within a part are kept below those required for craze initiation, then brittle failure will not occur. c · σ1 = σ1 ( ε )
[1] The critical values of maximum principal stress required for craze initiation as a function of rate are shown in Figures 1 and 2 for polycarbonate and polyetherimide, respectively. The same data is shown in Figure 3, where the critical maximum principal stress as a function of rate has been normalized by the rate dependent, yield stress of the material. As will be seen later, this representation of the failure criterion simplifies the application of the design aids which have been developed.
Design Aids for Preventing Brittle Failure
Figure 1. Critical value of maximum principal stress for craze initiation for PC.
269
Figure 2. Critical value of maximum principal stress for craze initiation for PEI.
Unfortunately, even though brittle failure criterion exist for these two materials, simple hand calculations or traditional finite element analyses can not be used to predict failures. In most cases the large, localized, maximum principal stress levels required to initiate crazing in these materials can only develop at the elasticplastic boundary of a constrained, localized plastic zone. Such stress states develop near three dimensional, stress concentrators such as corners. In order to accurately predict the three dimensional stress fields in these areas, very Figure 3. Ratio of critical maximum principal stress for refined, three dimensional, elastic-plastic, finite craze initiation to yield stress. element analyses are required. The time required to mesh and analyze actual parts in this manner, which may have many potential failure initiation sites, is impossible at this point in time. Since maximum principal stress levels can not be accurately and readily calculated for complex part geometries, design aids to assist the designer in preventing brittle failure have been developed.
DESIGN AIDS FRACTURE MAPS Most unfilled thermoplastics can behave either ductilely or brittlely depending upon the stress state, strain rate and temperature of the application. In polycarbonate and polyether-
270
Plastics Failure Analysis and Prevention
imide and other thermoplastic materials as well, severe triaxial stress states are more likely to cause brittle failures than are simple uniaxial or biaxial stress states. In some situations the combination of material, strain rate and temperature may not be severe enough to cause brittle failure even under the most severe stress state. An initial conservative method to determine if brittle failure is a concern is to look at a material’s performance under a very severe state of stress at the strain rate and temperature of interest. If the material behaves ductilely under this “worst case” condition, brittle failure of the component is unlikely to occur. A method of describing a material’s ductility under a severe stress state is to calculate its “ductility ratio”. A materials ductility ratio, Equation [2], is defined as the ratio of its actual failure load in a notched beam geometry to its maximum, ductile, load carrying capability in an unnotched beam geometry where the height of the unnotched beam is equal to the net section height of the notched beam geometry. The ductile load limit can be determined Design Aidsly or with a plastic hinge calculation assuming fully developed plasticity over the entire cross section and perfectly plastic material behavior, i.e. no strain hardening.3 A ductility ratio of 1.0 corresponds to a ductile failure. Ductility ratios below 1.0 correspond to varying levels of brittle failure. P Pductile
failure Ductility Ratio = ------------------
[2]
2
σ y bh P ductile = -------------L
[3]
where: yield stress at appropriate rate and temperature σy b beam thickness h beam height L span Ductility ratios can be plotted as a function of strain rate at different temperatures to create “fracture maps”, which map out regions of ductile versus brittle behavior. Fracture maps can be used in three ways: • To compare the relative ductility of materials at different rates and temperature • To choose a material which will behave ductilely at a certain rate and temperature even in a severe stress state • To determine if brittle failure is a concern, if a material has already been selected for an application Fracture maps for polycarbonate and polyetherimide at room temperature are shown in Figures 4 and 5, respectively.
Design Aids for Preventing Brittle Failure
Figure 4. Fracture map for PC (Lexan 141) at room temperature.
271
Figure 5. Fracture map for PEI (Ultem 1000) at room temperature.
GEOMETRIC SEVERITY RATIOS As we have seen, geometries producing tensile, triaxial stress states having a large maximum principal stress component are more susceptible to brittle failure. Fracture maps can be used to determine if a material may fail brittlely under a “worst case” stress state at a given rate and temperature. If brittle failure is a possibility for the material, i.e. if its ductility ratio is less than one, then the actual stress state produced within the part and the material’s failure criterion must be considered. One way of characterizing the susceptibility of a geometry to brittle failure is to numerically calculate the geometry’s peak maximum principal stress level and divide it by the yield stress of the material. This ratio is referred to as the “geometric severity Figure 6. Geometric severity ratio”. Since the geometric severity ratio normalizes the peak, ratios for common features. maximum principal stress level in the geometry by the yield stress of the material used in the analysis, it represents a purely geometric factor. Note that the geometric severity ratio corresponds to a worst case loading condition, i.e. a load level that results in the maximum possible level of maximum principal stress in the part assuming elastic-perfectly plastic material behavior. Geometries having larger geometric severity ratios are more likely to fail brittlely. Figure 6 provides a list of geometric severity ratios for common, generic geometries. In
272
Plastics Failure Analysis and Prevention
Figure 7. Rib positioning vs. bending direction.
many geometries, local stress levels will be affected by geometric parameters such as radius, thickness, etc. For these geometries a geometric Figure 8. Geometric severity ratio as a function of fillet severity ratio range is provided. Using Figure 6 radius and plate thickness for rib thickness for rib thickness/plate thickness=0.5. geometric severity ratios can be chosen which represent features of the part being designed. These geometric severity ratios can then be compared to a material’s critical maximum principal stress to yield stress ratio, i.e. its failure criterion, at the appropriate rate and temperature (Figure 3) to determine if brittle failure is possible for an arbitrarily high structural load. If brittle failure is a possibility, then maximum principal stress levels must be kept below those required to initiate failure by adjusting local geometric parameters or by limiting the loading on the structure. DETAILED RIB DESIGN - DESIGN CURVES The ribbed plate geometry has been analyzed in detail to quantify the effects of local geometric parameters upon its geometric severity ratio. The rib plate geometry deserves special consideration since it is a very common geometry with a high geometric severity ratio. Ribs are often added to thermoplastic parts to increase part stiffness, but may reduce part strength by promoting brittle failure. When designing with ribs many options are available to reduce the likelihood of brittle failure. The simplest option is to position the ribs so that the fillet radii between the ribs and the plate are in compression rather than tension (Figure 7). This will create compressive, triaxial stress states near the fillet rather than tensile triaxial stress states. If this first option is not practical, then stresses near the fillet radii should be kept below those required to cause brittle failure. In some cases this can be accomplished by adjusting local geometric parame-
Design Aids for Preventing Brittle Failure
273
Figure 9. Geometric severity ratio as a function of fillet radius and plate thickness for rib thickness for rib thickness/plate thickness=0.75
Figure 10. Geometric severity ratio as a function of fillet radius and plate thickness for rib thickness for rib thickness/plate thickness=1.0
ters such as fillet radii and plate thickness. Geometric parameter studies have been performed quantifying the local stress fields near the fillet radius in a ribbed plate geometry (Figures 8-10). In Figures 8-10, ratios of maximum principal stress to yield stress are plotted as a function of fillet radius and plate thickness for rib thickness to plate thickness ratios of 0.5, 0.75, and 1.0, respectively. If possible, the combination of plate thickness, rib thickness, and fillet radius should be chosen to keep the geometric severity ratio (Figures 8-10) below that required to initiate failure (Figure 3). If this can be accomplished, brittle failure will not occur even for an arbitrarily high load level. If local geometric parameters can not be chosen to keep the geometric severity ratio below the failure criterion, then brittle failure is possible and can be prevented only by limiting the severity of the loading.
SUMMARY A comprehensive approach, from identifying a failure mechanism to developing design aids, has been presented to assist designers in preventing brittle failure in polycarbonate and polyetherimide. Subsurface crazing has been shown to precede brittle failure in these two materials. Since crazing precedes brittle failure, a craze initiation criterion can be used as a conservative brittle failure criterion. A rate dependent, maximum principal stress criterion has been chosen for the craze initiation criterion. Using this knowledge of the brittle failure behavior of these two materials, a three step approach has been presented to assist designers
274
Plastics Failure Analysis and Prevention
in preventing brittle failure. The first step is to look at a material’s ductility ratio at the strain rate and temperature of interest. If a material has a ductility ratio of 1.0, then brittle failure is not a concern since the ductility ratio represents the relative ductility of the material in a “worst case” stress state. If the ductility ratio is less than one, then the generic, geometric severity ratios representative of geometric features of the part should be compared to the failure criterion of the material. These geometric severity ratios will often be functions of local geometric parameters. In these instances a range of values is listed for a given geometry. If the failure criterion is less than the entire range of geometric severity ratios, brittle failure will not occur, regardless of the magnitude of the applied load. If this is not the case, the final step is to examine the effect of local geometric parameters upon the geometric severity ratio. In some cases, local parameters can be adjusted to keep a geometry’s geometric severity ratio below the failure criterion. If local parameters can not be adjusted to “guarantee” ductile behavior, then the load level on the structure will determine whether or not brittle failure will occur. It should be noted that many of the concepts presented in this paper are specific to two materials, polycarbonate and polyetherimide, and to two specific grades, Lexan 141 (PC), and Ultem 1000 (PEI). These concepts may not apply to materials with different failure mechanisms.
REFERENCES 1 2 3 4 5 6 7
M. Ishikawa and I. Narisawa, J. Mater. Sci., 18, 1947 (1983). N.J. Mills, J. Mater. Sci., 11, 363 (1976) R.P. Nimmer and J.T. Woods, J. Polym. Eng. Sci., 32, No. 16., 1126 (1992). J.T. Woods and H.G. deLorenzi, J. Polym. Eng. Sci., 33, No. 21, 1431 (1993). S.S. Sternstein and L.Ongchin, Am. Chem. Soc., Div. Polym. Chem., Polym. Prepr., 10, 1117 (1969). R.J. Oxborough and P.B. Bowden, Philos. Mag., 28, 547 (1973). J.T. Woods , R.P. Nimmer, and K.F. Ryan, 1995 Antec Proceedings, Boston, 3923 (1995).
10 Common Pitfalls in Thin-Wall Plastic Part Design
Timothy A. Palmer Bayer Corporation, 100 Bayer Road, Pittsburgh, PA 15205, USA
DEFINITION OF THIN-WALL For the purposes of this paper, a thin-wall part is defined as one injection molded in an engineering thermoplastic resin (e.g. PC, PC/ABS, PA6), having projected area greater than 8 square inches and nominal wall thickness less than 0.060" (1.5 mm). Today, many thin-wall applications push beyond this defined limit and use nominal wall thicknesses less than 0.040" (1.0 mm).
PITFALL #1: DESIGNING WITH TOO MUCH VARIATION FROM THE NOMINAL WALL THICKNESS After the molten resin is injected into the mold cavity, different areas of the plastic part experience different levels of volumetric shrinkage proportional to wall thickness. In conventional moldings packing pressure is applied to force more molten material into the thicker areas, minimizing the effects of differential shrinkage. Unlike conventional parts, molten resin in thin-wall parts solidifies only a few seconds after the end of fill, giving packing pressure little time to act. The thinnest walls solidify before significant volumetric shrinkage can occur. Thicker areas take longer to freeze, experiencing very high volumetric shrinkage. In the worst case, material around the gate can solidify before any area of the part can be adequately packed-out. The notion that molten plastic follows the path of least resistance is especially true in thin-wall housings. Often, advancing flow will simply not fill the thinnest areas of a part, creating either non-fill or gas entrapment. Because of these difficulties, thin-wall parts should be designed with uniform wall thickness as much as possible. This allows molded parts with low differential volumetric shrinkage, improved dimensional quality and reduced chance of cosmetic problems caused by non-fill or gas entrapment. However, the decision to use nominal wall design must be made early in the design cycle due to the restrictions it may impose. Often, additional wall thickness must be added to the inside of a housing opposite areas such as label recesses to
276
Plastics Failure Analysis and Prevention
maintain the nominal wall thickness. Note that as with conventional parts, sharp edges in the flow path and at internal corners should be avoided.
PITFALL #2: USING IMPROPER RIB TO WALL THICKNESS RATIO The thick section formed by the intersection of a rib and the nominal wall tends to experience greater volumetric shrinkage than the rest of the part, causing sink opposite the rib. In conventional housings, rib base thickness is based on a percentage of the attached nominal wall, varying from 50 to 66% depending on the degree of cosmetic perfection desired. This design practice acts to reduce the thick section and make it easier to pack-out, largely eliminating visible sink. When standard rib design rules are applied to thin-wall parts, the resulting rib designs are usually too thin to fill properly, especially after draft is added. If the ribs can be filled, freeze-off usually occurs well before the rest of the part, with shrinkage much different than in the attaching nominal wall. To allow the ribs to fill properly, a 1:1 rib to wall thickness ratio can be used in walls less than about 0.050" thick. Any resulting sink marks tend to be much less noticeable than with conventional parts, especially if the opposing surface is textured. In a thin-wall part, there is much less material at the rib/wall intersection to shrink and cause sink than in conventional molded parts.
PITFALL #3: CONSIDERING ONLY EASY-FLOW RESINS FOR THIN-WALL APPLICATIONS Thermoplastic resins are often available in a range of molecular weights. Grades with lower molecular weight typically have lower melt viscosity and flow farther under the same pressure than their higher molecular weight counterparts. Unfortunately, easier flow usually comes at the expense of physical properties such as yield strength and impact strength. In addition, a material's resistance to UV light and chemical attack are reduced with decreasing molecular weight. Because thin-wall applications can be difficult to fill, the expected flow properties of low molecular weight resins seem desirable. Figure 1 shows the difference in predicted filling pressure between high and low molecular weight grades of polycarbonate for a sample housing. Mold-filling analysis results for the 0.040" (1.0 mm) nominal wall show that regardless of molecular weight, high-performance injection molding equipment is probably required. In this case, using a lower molecular weight resin may sacrifice material properties without significantly reducing production costs.
10 Common Pitfalls
277
Figure 1.
PITFALL #4: RELYING ON FIBER-REINFORCED RESINS TO PROVIDE RIGIDITY The structural rigidity of a thin-wall housing is greatly reduced versus its thick-wall counterpart due to the reduction in section modulus. From the standard engineering beam bending formula (w/both ends simply supported), the maximum deflection is inversely proportional to the thickness cubed, so under identical loads, a beam 0.040" thick has deflection 8 times a wall 0.080" thick. A potential solution for thin-wall housings is to use a fiber-filled resin, which typically increases the material modulus by about 50% (10% glass fiber-filled). However, maximum deflection is only inversely proportional to the material modulus, so the unfilled beam only deflects 1.5 times more than the fiber-filled one. Because the wall thickness effect dominates over the effect of fiber reinforcement, the rigidity of thin-wall housings cannot be expected to compare to thick-wall, conventional housings. Rigidity of thin-wall applications will still depend on assembly with the product's other internal components, regardless of the resin used. Impact properties are also important for thin-wall housings given their widespread use in hand-held products prone to being dropped. Fortunately, thinner walls may perform slightly better in a drop impact because more flexible walls have better energy absorption. However, the addition of fillers can sharply reduce these properties. For example, the notched izod impact strength of 0.125" thick polycarbonate is reduced from 17 ft lb/in to 2 ft lb/in when 10% glass is added. These examples suggest that the liabilities of fiber-filled materials may outweigh their benefits in most thin-wall parts.
278
Plastics Failure Analysis and Prevention
PITFALL #5: IMPROPERLY LOCATING GATES Thin-wall applications push thermoplastic resins and standard injection molding equipment to their respective limits, but properly locating gates is often overlooked as a way to widen the available processing window. Unfortunately, gate locations are often chosen after part designs are finalized, leaving only a few locations where gate vestige is allowed. A better approach is to pick gate locations early in the design cycle to optimize filling, and then position label areas or other styling to conceal any remnant of vestige. In conventional as well as thin-wall parts, filling pressure is minimized when all of the last areas to fill do so simultaneously. This phenomenon is known as balanced filling and promotes uniform solidification and packing of the part. When wall thickness is uniform in a thin-wall part, gate locations should be chosen so that the longest flow paths from all gates are equal in length. However, if a thin-wall part has non-uniform wall thickness, truly balanced filling is difficult to achieve. In fact, some degree of filling imbalance may actually improve the moldability of a non-uniform wall part. Mold-filling analysis is required to optimize such cases. When analyzing a thin-wall part, the mold-filling analyst should always consider the part and the delivery system (e.g., three-plate runner, hot manifold), because pressure consumed in these components can have a much greater effect on flow balance in thin-wall parts than in conventional designs.
PITFALL #6: USING SLOW INJECTION RATES While high injection pressures are required to fill thin-walled parts, delivering the molten resin at a sufficient injection rate is also an important parameter. To prevent early freeze-off, the molding machine must inject material at a rate high enough to produce shear heating at the flow-front. Once the flow-front temperature begins to drop, the pressure required to advance it can quickly exceed press capabilities, resulting in non-fill. Today's closed-loop, electronic controls allow nearly any injection rate to be set at the press, but close examination of the actual ram velocity vs. position trace may show that the desired injection rate can only be achieved over a small portion of the injection cycle, if at all. In this case, a "high-performance" injection molding press designed specifically for high injection rates will be required. Such machines have the ability to deliver high pressure at very high injection rates through the use of accumulators or other methods.
10 Common Pitfalls
279
PITFALL #7: USING MORE GATES THAN NECESSARY In many thin-wall applications, numerous gates are used when fewer would be suitable because the material is not expected to flow more than a few inches beyond the gate. However, as mentioned in #6, significant flow in thin walls is possible when flow front velocity is high enough. The rapid freeze-off expected in thin walls typically occurs because the flow front velocity is too low to generate shear heating. While the ability to maintain high Figure 2. flow-front velocity is largely dependent on the capabilities of the injection molding press, the number of gates used also plays an important role. Assuming radial flow from a pin-point style gate, the flow front velocity is inversely proportional to the distance flowed. If a square housing is fed through a centrally located gate (Figure 2), flow front velocity at the end of fill is Q/2 π Rt, where Q is injection rate, R is the radial distance flowed and t is part thickness. When multiple gates are used to fill the part, flow distance is reduced, but the input flow rate must be divided among the gates. In this example, the four gate system has half the flow front velocity of the single gate system at the end of fill. The part with a single, center gate has higher flow front velocity at the end of fill, no major knitlines and avoids gas entrapment at the center of the part.
PITFALL #8: UNDERSIZING GATES Because higher injection rates are used in thin-wall molding, larger gates are required to prevent cosmetic damage caused by excessive gate shear. Externally heated hot drops or valve-gated drops allow large gate diameters with clean degating. The following formula can be used to estimate the required pin-point or hot-tip gate orifice diameter. D =
3
32Q ---------nπγ
Here, the diameter D is a function of Q, the volumetric flow rate from the nozzle, n, the number of gates and γ the shear rate limit. For engineering thermoplastics the shear rate limit is usually 20,000-40,000 1/s, depending on the shear-sensitivity of the resin. Use a limit of 20,000 1/s for shear-sensitive resins. Note that this formula assumes equal flow
280
Plastics Failure Analysis and Prevention
Figure 3. Suggested pin-point gate detail for thin-wall parts requiring large gates.
passes through each gate. It can also be used to size tunnel gates, which should have at least a 20° included angle and be at a 45° angle to the parting line. If a three-plate runner is used, large gates may cause damage the thin nominal wall during degating. This can be avoided if a reinforcing dome is used opposite the gate as shown in Figure 3. Keep in mind that pressure imbalance between multiple drops in cold, three-plate runners may be more than with hot runner systems.
PITFALL #9: UNDERESTIMATING CLAMP TONNAGE REQUIREMENTS In thin-wall molding, it is not uncommon for the process window to be limited by the mold blowing open due to high cavity pressures. With conventional parts, clamp tonnage estimates of 3 tons per square inch are often adequate. Thin-wall applications must typically allow for more than 5 tons of clamp per square inch of the mold cavity projected area. If the part to be filled is large, the mold and backup plates should be about twice as thick as conventional parts to prevent flexing during high-pressure injection.
PITFALL #10: INADEQUATE VENTING IN THE TOOL The fast injection rates used in thin-wall molding require larger parting line vents, primarily to prevent flow hesitation as air is pushed from the cavity at the end of fill. However, the higher injection pressures and better flowing resins used increase the risk of parting line flash. A mold designed with a generous number of thinner vents may be the best compromise. Proper venting in the areas where air is chased at the end of fill is especially critical. Air trapped ahead of a quickly converging flow front can significantly increase filling pressure requirements.
Defect Analysis and High Density Polyethylene Pipe Durability
Shaofu Wu, Kalyan Sehanobish, and Noor Jivraj Texas Polymer Center, The Dow Chemical Company, Freeport, Texas 77541, USA
INTRODUCTION Thermoplastics, in particular polyolefins, are gaining considerable market share in pipe applications such as gas and water supply systems. To ensure proper performance of such pipes over the required lifetime, durability analyses are needed to adequately account for the effects of loading, time, temperature, environmental conditions, as well as the role of pipe defect and imperfections on relevant polymer properties and pipe performance. Durability analysis involves defect characterization, crack initiation and propagation mechanisms, and long term performance prediction. Internal pressure testing of pipes is highly dependent on the defect properties and population, as well as the toughness of material. Thus both defect population and toughness should be used as a measure to differentiate pipe materials for durability analysis. Characterization of defects inside the extruded pipes also provides valuable information about the origin of defects. Suggestions can be made to eliminate these defects by adjusting or changing the processing conditions and environmental conditions. Thus, characterization of defects is critical for durability analysis of pipe materials. Defect analysis of high density polyethylene pipe material is discussed in this paper. Various analytical techniques, such as fractography, hot stage microscopy, energy dispersive X-ray (EDX), micro-transmittance infrared spectroscopy, scanning electron microscopy (SEM), film stretching, and transmission electron microscopy (TEM) are used to characterize the defect properties and size distribution. Some of these analytical techniques, such as optical and FTIR microscopy, have been well documented to identify defects in polyethylene.1,2 In this paper, we will focus on defect characterization, the correlation of the defect properties and long term performance of pipe.
DEFECT CHARACTERIZATION Several PE 100 HDPE pipes which had been subjected to long-term pressure testing at 80οC with internal pressure of 5 MPa (in accordance with ISO TR 9080) have been examined by
282
Plastics Failure Analysis and Prevention
several analytical techniques in order to determine the nature of the existing defect particles, which appear to contribute to the failure of these pipes. Scanning electron microscopy (SEM), energy X-ray Figure 1. (a) SEM image of fracture surface in a pipe. The primary failure initiation site is dispersive located at bottom-center in the figure and is near the inner wall of the pipe; (EDX), microtransmit(b). Large (~75 µ m diameter) particle located at center of main failure initiation site tance infrared spectrosshown in (a). copy, and hot-stage microscopy were used to determine the size distribution and compositions of defects present in these pipes. The fracture surface shown in Figure 1a was produced by freeze fracturing an Figure 2. Example of defect particle on the fracture surface of different pipes. arc-shaped section of the pipe which contained the site at which a leak was first observed when the pipe failed the long-term test. A particle approximately 75 µm in diameter is located at the center of the round, crater-like, penny-shaped domain located at bottom-center in the SEM montage (see Figure 1b). This particle appears to have been the main initiator of failure in this pipe. Depending on the defect properties different interfacial adhesion with the matrix are observed and is shown in Figure 2. In general, material with a better adhesion between the defect and matrix has a longer lifetime with the hydrostatic test. Four tested pipes made from HDPE made using INSITETM Technology have also been evaluated. These pipes have much longer lifetime under the test conditions compared with the pipes made from HDPE made using the traditional Ziegler catalysts. Sections from the main crack sites of each pipe were cut and brittle fractured. Fracture surfaces were analyzed by optical microscopy and SEM. It was observed that fractures were initiated at the inner surface of pipes because of higher residual stress level at the inner surface. Failures initiated
Defect Analysis
283
both from small defect particles (15 to 20 µ m) and at sites where no defects were observed. SEM micrographs of the fracture surface of two pipes are shown Figure 3. Example of defect particles on the fracture surface of pipe made by INSITE Technology. in Figures 3. The defect particles observed on the fracture surfaces also show good interfacial adhesion with the matrix. We did not observe any other pre-initiated crack/craze on these fractured surfaces. We also randomly selected several sections from these pipes and fractured these sections. Again, no clear crack initiation sites were found on these fracture surfaces. Thus the defect particles observed in these pipes can not be related to any fracture event. It is speculated that these gel particles may have similar rigidity to the matrix and good interfacial adhesion and are less responsible for fracture. DEFECT SIZE DISTRIBUTION Figure 4 shows the particle size distributions obtained from fracture surfaces in two tested pipes. These two pipes have a quite different lifetime under the hydrostatic test (70 hours vs. 500 hours with 5.5 MPa hoop stress at 80oC). Particles were observed on fracture surfaces which were produced by freeze fracturing sections of pipe containing visible cracks at the
Figure 4. Particle size distribution of particles observed at fracture surfaces in one pipe.
284
Plastics Failure Analysis and Prevention
outer surface of the pipe. Most of the particles below 10 µ m were located on a fracture surface produced by notching and fracturing pipe in liquid nitrogen, where no externally visible cracks were located. While the relative populations of defects in these two pipes cannot be determined accurately from this information, it is clear that a pipe which has a longer lifetime also contains a significant number of defects, often as large or larger than pipes which exhibited a shorter lifetime. CRACK SIZES VS. DEFECT SIZE IN TESTED PIPES For the tested pipes, there are many natural cracks which initiated during testing, in addition to the major crack that caused failure of the pipe. In most cases, these natural cracks were initiated from defects distributed inside the pipe. Natural cracks in two tested pipes have been analyzed by cryogenic fracture of tested pipes. The results and their correlation with defect sizes are shown in Figure 5. These results suggest that the crack size Figure 5. The defect sizes vs. the crack sizes in two tested pipes. is not proportional to the defect size in these pipes. The crack size depends not only on the size of defect, but also the geometry, rigidity, connectivity with surrounding materials, and location in the pipe. A defect having a poor attachment with the surrounding material will form a natural void, therefore becoming a site for crack initiation; while a defect with a good interfacial adhesion with the surrounding material will generate a lower stress concentration, therefore delays the crack initiation. Examples of these defects are shown in Figure 7a and b. Failure analyses of tested pipes show that cracks usually initiate from a singular defect particle located near the inner surface of pipes. Crack initiations in these sites are primarily caused by a relative higher stress state in the inner wall of the pipe. However, pipes with a similar defect population have shown a quite different lifetime under internal pressure testing. These results suggest that the long-term performance depends not only on defect size, but also on defect properties and residual stress induced by the process. DEFECT STIFFNESS Since the defect particle stiffness and interfacial adhesion is critical to the long term performance of pipe materials, film stretching is used in this study to characterize these properties. Development of this method is based on the analysis of stress state around a sphere inclusion.3 For the case of plane stress, the hoop stress near the interface can be expressed as:
Defect Analysis
285
A – 3B ------- cos 2θ σ 00 = 2E1 ---4 2 r r
[1]
2ν 1 2ν 2 1 – ------------- E 1 – 1 – ------------- E 1 – 1 – ν 1 2 ν 2 p - ----------------------------------------------------------------------------A- = ----------2 2ν 2 4E 1 1 – ------------a - E 1 + E2 1 – ν 2
[2]
where
E 1 – E2 B p --------------------------------------------------- = -------4 4ν 1 4E 1 a - E 2 E1 + 3 – ------------1 – ν1
[3]
and E1, v1 and E2, v2 are the modulus and Poisson’s ratio of matrix and inclusion, separately. For a special case of cavity (E2=0), the hoop stress will be –p for θ = 0 and 2p for θ = 90o. For the perfectly rigid inclusion (E2=) and assume the Poisson’s ratio of the matrix is 0.4, the hoop stress will be 1.67p for θ = 0 and –1.33p for θ = 90o. Thus by examining the deformation behaviors near the defect particle during stretching, the stiffness and interface adhesion of defect can be characterized. A sketch of test setup is shown in Figure 6. Micrographs of the deformation behavior of gel defect particles at three stages of stretching are shown in Figures 7. All of pictures were taken under optical microscopy and with a magnification of 20x. Each figure illustrates the behavior of gel particle before stretching, after necking passing the gel particle, and after film fracture. The non-uniform deformation around the gel particle shown in Figure 7a, for example, indicates that there was stress concentration at upper and lower points of interface after necking passing the gel particle. Based on the stress analysis that soft gel particles will cause high stress concentration at upper and low points, it is concluded that the
Figure 6 (a). Sketch of an inclusion inside polymer and (b). Sketch of setup of gel particle property evaluation by film stretching method.
286
Plastics Failure Analysis and Prevention
Figure 7. Micrographs (x20) of gel particle deformation behaviors in films; a). relatively soft particle; and b). relatively rigid particle.
gel particle shown in Figure 7a is relatively softer comparing to the matrix (HDPE). Figure 7a also showed that the gel particle shape starts to change when we continue stretching the film that is also the evidence of a soft gel particle. Figure 7b shows an example of a relatively hard gel particle in HDPE matrix. The gel particle shape still remained the same or became sharper after necking passing the gel particle and non-uniform deformation around the gel particle was observed. This suggests the gel particle had a higher elastic modulus comparing with the matrix (hard gel particle).
CONCLUSIONS Fractographic analysis of tested pipes shows that cracks are mostly initiated from large (>50 µ m) defect particles. The results show that pipes with smaller defect particle size (less than 20 µ m) have a better long-term property. However, pipes with a similar defect particle size do not necessarily have a similar long-term performance. Besides the size, defects with high rigidity, sharp edges, and low connectivity with the matrix are more susceptible to crack ini-
Defect Analysis
287
tiate during testing or service. Residual stress generated during processing can also affect the long term performance.
REFERENCES 1 2 3
A. Chudnovsky, K. Sehanobish, and S. Wu, Methodology for durability analysis of HDPE pipe, ASME 1999 PVP Conference, August, Boston, 1999. K. Sehanobish et al., Fractographic analysis of field failure in polyethylene pipe, J. Mater. Sci. Lett., 4, 890-894, 1985. J.N. Goodier, Concentration of stress around spherical and cylindrical inclusions and flaws, J. Appl. Mechanics, Trans. of ASME, Vol. 55, No. 39, pp 39- 44, 1933.
Practical Risk Analysis - As a Tool for Minimizing Plastic Product Failures
Subodh Medhekar, John Moalli, Robert Caligiuri Exponent Failure Analysis Associates, Menlo Park, CA, USA
INTRODUCTION In 1988, the International Organization for Standardization issued the ISO 9000 series of business management standards which required organizations to develop formalized Quality Management Systems that ideally are focused on the needs, wants, and expectations of customers. For example, in accordance with QS 9000 (the automotive analogy to ISO 9000) standards, compliant automotive suppliers must utilize Failure Mode and Effects Analysis (FMEA) in the Advanced Quality Planning process and in the development of their Control Plans. Other tools such as Fault Tree Analysis (FTA), Preliminary Hazards Analysis (PHA) and Mean Time Between Failures (MTBF) Analysis are increasing being used by product designers and manufactures to evaluate and minimize the product risk. A majority of risk analysis techniques in use today have originated from industries with increased perceived risk (or higher public exposure/scrutiny). These industries include Nuclear, Aerospace, Chemical and Petrochemical industries, where risk evaluation techniques such as Event Tree Analysis (ETA), Fault Tree Analysis (FTA), Hazard and Operability Studies (HAZOP) were developed and utilized to manage the industry specific risks. Each of the risk analysis technique has its own advantages and shortcomings. The quantitative analysis techniques such as Fault Tree Analysis and Event Tree analysis can yield definitive results but usually require enormous effort in model development, data collection and quantification of uncertainties. The qualitative (or semi-quantitative) risk analysis tools such as PHA or FMEA yield quick results that are not quantitative, but can be costeffectively used to prioritize the relative risks (in relation to other risks present or posed by the product/system). From a practical standpoint, we believe the FMEA to be one of the most efficient tools and have developed some methods to simplify it. In this paper, we describe some of these practical and simplified approaches, and how they can be effectively used to design and minimize the risk of plastic product failures.
290
Plastics Failure Analysis and Prevention
FMEA METHODOLOGIES FMEA’s are performed to identify failure modes with the intention of reducing the likelihood of the failure and/or the severity of the consequences and can be applied to plastic products in any stage of their life cycle. Although FMEA’s are often performed during the early stages of design and development (or modification), they sometimes are executed after some feedback is obtained from product use (i.e., after a product failure). MIL-STD 1629A1 describes the approach developed by the military in which scenario development can be either inductive or deductive. FMEA’s performed on concept designs or leading edge products generally use a partially deductive approach in which the first step is definition of a set of undesirable end conditions.2 Each condition defines a specific loss of functionality or performance, and failure scenarios leading up to these conditions are developed. On the other hand process FMEA’s generally develop failure scenarios inductively by examining undesirable variations in process parameters and examining the resultant effects. We believe that the deductive, or “top-down” approach, is the most practical form of the FMEA. It is generally more simple to perform than an inductive, or “bottom-up” analysis, and often prevents the FMEA team from getting hung up in details that usually do not improve the resolution of the analysis. This deductive approach is cost effective as it only “digs” as deep as required. Time and resources are not wasted on pursuing or detailing out those scenarios that, in the end, are not significant contributors to the risk. In addition, resources are not wasted attempting to quantify risks that are not very significant or not realistic.
A SIMPLIFIED FMEA APPROACH The objective of the FMEA is to create a living document that becomes a basis for making strategic engineering decisions. In a similar fashion to others, we characterize the relative risk contribution of potential failure scenarios associated with the process or product in terms of a risk priority number (RPN). This RPN is obtained as a product of three indices representing, respectively, the severity of the failure consequences, it’s likelihood of occurrence, and it’s detectability. The process we have developed to simplify the FMEA process employs a three-phase approach. In the first phase we develop a common and consistent framework for the analysis. We assemble an FMEA team and use a combination of brainstorming sessions (which is the traditional FMEA method) and focused evaluations of the functionality of each component under both normal and abnormal conditions. We then use the deductive approach and develop a handful of “generic” failure modes, each with an associated relative severity
Practical Risk Analysis
291
index representing its potential safety or business impact, and numerous potential failure scenarios involving these modes. Depending on the nature of the information in this phase, we also develop a relative scale for the likelihood of occurrence and detectability indices, from which values are objectively assigned to individual scenarios. These indices are constructed to be broad brush, rather than detailed, and are usually qualitative rather than quantitative. For example, instead of trying to precisely evaluate specific frequencies of occurrence, we will simply construct a scale based on “occasionally”, “more than a few times”, “observed once or twice in a product’s lifetime”, and “never observed”. In almost all cases, we find that this simplification provides adequate resolution yet shortens the analysis time substantially. A typical example of simplified occurrence indices is displayed n Table 1. Table 1. Sample simplified frequency of occurrence index Evidence About the Failure Scenario
Occurrence Index
Documented “frequent” occurrence in this or similar application.
10
Known to have occurred “a few times” with documented evidence.
8
Known to have occurred once with documented or reported evidence in this or similar application.
6
Anecdotal evidence of previous occurrence of this or related failure scenario.
4
No previous history, but greater potential to occur.
2
No previous history, but potential to occur.
1
In the second phase, we compile information relevant to each individual failure scenario in “evidence sheets”. This can range from qualitative or anecdotal information to formally documented, quantitative information. Typical information sources are field service data from previous generation or similar products, supplier quality control data, process control and production quality assurance data, descriptions of previous failure experiences, and published failure data. The information collected and tabulated in this phase has applicability beyond the FMEA, and forms the basis of the “living” document. It is critically important that the document remain “living”, that is, it must be continually updated. An FMEA which is per-
292
Plastics Failure Analysis and Prevention
formed and never acted upon or updated can represent a potential liability to a company in the event of subsequent litigation concerning the component being analyzed. In the third phase, we calculate the RPNs and graphically analyze their distribution. This provides guidance as to the risk contributors that may require action. We also develop appropriate action strategies in this phase and evaluate their potential for risk mitigation.
CASE STUDY – COMPRESSED NATURAL GAS FUEL SYSTEM As part of an initiative on reducing vehicle hazards, the National Highway Traffic Safety Administration (NHTSA) funded research on alternatively fueled vehicles. We performed a FMEA on fuel system components of compressed natural gas (CNG) vehicles, including the composite high pressure fuel cylinders, under this research program. The fuel system examined was quite complex, with 17 subsystems composed from hundreds of components supplied by several vendors. The number of potential failure scenarios was enormous, and a standard type of risk analysis embodying a root cause, inductive type approach with a traditional detail driven FMEA would have been a prohibitively complex and expensive process. The deductive approach allowed each supplier to think of their subsystem or component not only in the context of immediate functional specifications, but ultimately as a part of the final product delivered to the consumer. Failures of the individual components were viewed as causes leading to more generic failure modes whose impact on the performance or safety of the vehicle could be quantified, rather than developing specific detailed failure scenarios of each individual component. The first phase of the FMEA was identification of the appropriate team. In this case, instead of formulating the group from design and production staff of a vehicle manufacturer, the team was expanded to include all of the potential suppliers as well. By bringing together all of the participants in the product design, it was possible to evaluate the functionality of each of the subsystems both under normal and abnormal conditions, and to consider the interactions between subsystems and the vehicle platform itself. The team then mutually developed scales for the severity, occurrence, and detectability indices by defining them in a relative manner to reflect the full range of conditions encountered in the scenarios without overt specification or quantification. This allowed the key issues to be identified without becoming lost in the details. Specifically, the consequences of a particular failure scenario on the performance of the product were identified by the severity of failure index shown in Table 2. The severity (SV) was assigned on a scale of 1 to 10, with 10 representing the most severe mode. The probability that a particular sequence of events leading to a failure scenario will occur was identified by the likelihood of occurrence index displayed in Table 1. The occurrence index
Practical Risk Analysis
293
(OC) was also assigned on a scale of 1 to 10, with 10 representing the highest frequency. The probability that a particular sequence of events leading to a failure will be controlled by detection or mitigation before the consequences occur is normally quantified in the third index, sometimes called the detection index. Since this FMEA covered all aspects of the CNG system, design as well as usage, this index was used instead to quantify the effect of a variety of risk minimization measures. Accordingly, the risk minimization index (MN), shown in Table 3, was also assigned on a scale of 1 to 10, with 10 representing the highest risk or minimal possibility of control. Table 2. Generic failure modes and associated severity indices Generic Failure Mode
Description of Potential Effects
Severity Index
Customer dissatisfaction
warranty claim, customer only uses gasoline and resulting environmental impact, loss of repeat sales, inconvenience, customer anxiety
1
Leakage (does not involve injury)
smell of gas, customer discomfort, warranty claims, reduced operating range, inconvenience
2
Driveability and performance
collision, collision/injury, warranty claims, recall
5
Loss of compliance
recall, warranty claim, customer inconvenience
7
Vehicle inoperative
walk-home, warranty claim, inconvenience
8
Loss of crashworthiness
collision resulting in: explosion and fire, explosion without fire, injury, property damage; reduced range, render vehicle inoperative, damage to vehicle, recall
9
Large gas release (customer may have advanced warning)
fire, explosion, asphyxiation of operator, vehicle becomes inoperative, property damage, reduced range, damage to vehicle, smell of gas, noise and resulting anxiety, injury, warranty claim, recall, cryogenic burn
9
Catastrophic high pressure failure (unexpected event)
explosion and fire, explosion without fire, injury, property damage, reduced range, vehicle becomes inoperative, damage to vehicle, warranty claim, recall
10
294
Plastics Failure Analysis and Prevention
Table 3. Sample simplified control measures indices. Range of Control Measures
Probability Index
High degree of control measures implemented or planned, assigned a LOW detection index range of 1-3
2
Moderate degree of control measures implemented or planned, assigned a MEDIUM detection index range of 4-6
5
Low degree of control measures implemented or planned, assigned a HIGH detection index range of 7-10
8
It is important to note from Tables 1 through 3 the generic, non-specific nature of the definitions. This prevented endless debates over assignment of values and allowed the FMEA team to focus on the key issues, and not degenerate into needless debate over insignificant issues, a common problem with FMEAs on complex systems. In the second phase of the FMEA, each of the main subsystems and components in the CNG system was reviewed to identify sequences of events or scenarios which could lead to any of the generic failure conditions. These scenarios were defined as potential failures of each major component in the design which could lead to performance problems or safety issues. The next step in Phase 2 involved assigning the frequency of occurrence to individual scenarios. This index assignment was based on reviewing available evidence about failure scenarios, including preliminary testing data, published failure data, and non-quantitative historical and anecdotal information. The final step in Phase 2 was identifying the probability that a particular sequence of events leading to a failure could be detected or mitigated through manufacturing process controls, design changes, or validation testing. This index, designated MN, was lower for design changes or mitigation measures which were reasonably expected to lower the risk of failure. In Phase 3 of the FMEA, the product of the three indices, SV, OC, and MN, was calculated. This number is known as the Risk Prioritization Number, or RPN. The RPN values ranged between 0 and 725, as shown in Figure 1. Examination of the specific data showed that fuel system leaks external to the tank potentially created greater risk than the tank itself; a non-intuitive answer that could not have been arrived at by following a standard detail oriented approach towards risk assessment. Since this analysis was performed on a generic fuel
Practical Risk Analysis
295
Figure 1. Histogram of RPN’s from simplified FMEA.
system, no further actions were taken. A more typical case would entail mitigation of appropriate high RPN items through design changes or warnings.
CONCLUSIONS FMEA can be a practical tool for risk assessment in plastic products. If a deductive or “top down” approach is used, the FMEA process can be completed within a reasonable time frame. Qualitative indices can also be used to reduce the burden of being precise or quantitative while maintaining the intended resolution of the FMEA.
REFERENCES 1. 2.
Department of Defense, “Procedures for Performing a Failure Mode, Effects, and Criticality Analysis”; MIL-STD1629A, Global Engineering Documents, 1980. D.H. Stamais, “Failure Modes and Effects Analyses, FMEA From Theory to Execution”, ASQC Quality Press, 1994.
Attachment Design Analysis of a Plastic Housing Joined with Snap-fits
Dean Q. Lewis and Gary A. Gabriele Rensselaer Polytechnic Institute, Troy, NY, USA Bob Brown Lucent Technologies, Holmdel NJ, USA
INTRODUCTION Integral attachment features (snap-fits) are commonly used today in the manufacturing of plastic parts and a highly recommended means to reduce overall product costs.1,2 Integral attachment features can provide benefits to part design by removing tool requirements, reducing the part count, reduced installation time, and providing feedback upon attachment (i.e., the "snap").3 Many products that once used traditional means of attachment (i.e., screws, adhesives) are now being redesigned for attachment with snap-fits. While snap-fits may provide many benefits, their inclusion into a product must be Figure 1. Photo of product housing. carefully designed to prevent a reduction in product quality. A redesigned desktop telephone housing was used as a case study on the implementation of snap-fit attachment to achieve reduction in assembly costs and its effect on robustness, specifically drop testing. The housing was changed to replace the original attachment using screws with snap-fits. Five screws were eliminated from the product and replaced with five cantilever-hole features located around the perimeter of the upper housing with slotted catches on the lower housing. To allow for disassembly for servicing during product life, access holes were Figure 2. Disengaged cantilever-hole snap fit. included for four of the snaps. Figure 1 is a photo of the assembled housing and Figure 2 shows a photograph of a snap-fit used in this product.
298
Plastics Failure Analysis and Prevention
One of the requirements the product has to pass is standard impact testing for telephone products. The product must withstand drops from a typical desk height of 76 cm (thirty inches). Failure in these tests would occur if the attachment of the housing failed, if the product no longer operated properly, or if the interior components became dislodged or exposed. PROBLEM With the redesign of the housing for attachment using snap-fits, the product would no longer consistently pass the drop testing. Some of the snaps, anywhere from one to all five of them, would become disengaged during these drops. The original screw bosses had not been removed from the design, so the screws could still be used to assist the snap-fits in attachment. It was found that by including one or more of the original screws, the product attachment was more secure and likely to pass the drop tests. OBJECTIVE AND CONSTRAINTS The desired design of this housing was one that did not require the use of separate mechanical fasteners, such as screws. The attachment design needed to be modified to achieve this goal. The product was required to pass the impact tests stated earlier. The exterior of the product had to appear identical to the previous design with the exception of the size and location of access holes for the snaps within certain aesthetic restrictions. Additionally, the location or geometry of other components in the product assembly could not be changed.
RESEARCH APPROACH With the goal of improving the housing attachment strategy, a process of evaluating the existing integral attachment design, physical drop testing of sample housings, and attachment design concept generation was proposed to provide suggestions as to how the product could be improved. The reasons for failure were investigated and used during the product redesign. Prototypes implementing some of the changes were then tested. Some general guidelines were then stated to provide improved product robustness during impact. FAILURE ANALYSIS The first step in analyzing the problem of product failure during impact testing was to determine the possible cause or causes of this failure. Different established techniques for analyzing product failure exist. It was determined that one of these, the fault tree analysis (FTA), could provide a clear explanation of test failure in diagram form. Additionally, drop tests of the assembly were proposed to gather more details on attachment failure. A drop test apparatus was constructed for consistency of drops. It was
Attachment Design Analysis
299
comprised of a support frame from which the housing was hung by string from a small ball supported from the top of the apparatus. The housing position could then be adjusted until it was at a height of 76 cm and parallel to the ground. The ball was released, and the housing could be examined after it hit the ground. The drops were recorded on videotape. The results of the drop testing were then used in determining which of the failure modes found in the FTA were most likely occurring. Failure Tree Analysis Failure tree analysis (FTA) is a means to clarify any and all causes of failure by graphically representing a hierarchy of these causes in a tree structure.4 The main failure is stated at the top with a few more general causes stemming down from that point, and repeated until the root causes are all listed at the bottom of each branch. Any type of failure was included in the FTA, but specific areas that dealt with integral fastening and attachment design were noted for further investigation during this study. Failure to pass impact-testing requirements was the primary failure for this investigation. The reasons for which the product would not pass the test included housing failure or mechanical or electronic failure of com-ponents. Component failures will not be discussed here because the focus of this study is on the housing failure. Failure of the housing was broken down into two further categories: either the housing was actually damaged or the housing disengaged. Damage to the assembly would cause failure if any part of the housing was damaged or if the method of attachment physically failed (broke). Since breakage of the snaps was not a problem, this area will not be considered in the study. Considering housing disengagement, the attachment method could have failed because the degrees of freedom of the two parts were not fully constrained, one or more attachment features disengaged during the test, or one or more attachment features were not fully engaged prior to the drop test. Assuming that the engagement of the features was confirmed prior to the tests, causes of problems prior to drop tests will also not be investigated. To ensure secure attachment, all degrees of freedom between the parts in the assembly need to be constrained. If this is not done completely, the housing could disengage during drop testing. Either improper location or number of features or a problem with tolerances or manufacturing could have caused this. One possible cause of feature disengagement during the test is that the maximum retention force for the design of the feature was exceeded. The retention force could be exceeded due to unexpected loading conditions or due to errors in feature design or manufacturing. Flexibility of the housing could also cause feature disengagement. The top and bottom parts of the thin-walled housing, without sufficient reinforcement or material strength, could deform enough to cause the snaps to come apart. Under proper loading conditions, the fea-
300
Plastics Failure Analysis and Prevention
ture itself might cause the disengagement due to its flexibility. Improper orientation, type, or sizing of the feature, flexibility in the basepart near the snap, insufficient material strength, or manufacturing errors could all contribute to this cause. Summary of Failure Tree After generating the FTA chart, the specific causes of failure that were involved in the attachment of the two housings and those that could possibly result in the snaps disengaging during the impact test found. A limited number of possible disengagement failure modes were found and include the following: • all of the degrees of freedom were not fully constrained in the design due to either an insufficient number of snaps, an improper location of the snaps in the design, or some limited motion allowed in specific directions due to tolerancing that could result in relative part motion and disengagement (Figure 3 modes 1 and 2); • the maximum retention force given for the snap was exceeded causing disengagement Figure 3. Failure modes. (Figure 3 mode 3); • the housing was too flexible due to either insufficient reinforcement in key locations or the product's thin-wall design (Figure 3 mode 4); • the snaps were too flexible due to either the basepart near the snap being too thin, the wrong type of snap used in the design, or the dimensions of the snap were sized improperly (Figure 3 mode 5). A further detailed examination of these failure causes revealed more information on how the snaps could become disengaged during impact testing. If the upper housing is too flexible, it could bow outward at impact (Figure 3 mode 6). This could cause the side edges of the upper housing to pivot in such a way that the cantilever-hole snap pulls away from the catch on the bottom housing. Also, by having four of the five snaps on the two side surfaces (Figure 3 mode 2), bowing of the top housing was not properly constrained. Finally, it was found during inspection of the assembly that when the bottom housing is loaded near the
Attachment Design Analysis
301
location of the snap-fit, the wall actually pushes the snap away from the catch (Figure 3 mode 4). The snap could almost be disengaged just by pushing on the side wall by hand. The true reason for failure could be one or a combination of these causes. Drop Testing While the failure tree analysis was extremely useful if determining different possible failure modes, it was not known whether one or more of these causes could be attributed to the true reason for product failure. Physical testing and evaluation was necessary to determine if the suspected failure modes were feasible and occurring in the drop tests. Also, by altering the assembly, housing, or attachment features, the effects of some of these causes could be investigated. In order for the current product design to pass impact testing, two screws were included in the design. In order to find the effect of these screws, tests were conducted using two, one, and no screws with all five snaps engaged. These tests can also be used as a comparison to the other tests. Figure 4 shows the location of the snap-fits used in the attachment and the drop orientation. Having all the attachment features along a specific loading direction could cause failure Figure 4. Location of snap-fit feature on housing. especially when the loading direction is the same as the disengagement direction. On the housing, four out of the five attachment features are oriented in the same direction. To study the effect of this possible cause, drop tests were conducted with the cantilever-hole feature on the top edge removed (#5 in Figure 4). No screws were used in these tests in order to see the results of the flexibility of the housing and the orientation of the snaps. Flexibility of the housings and of the snaps could also be investigated by adding additional reinforcement to the housing. To examine the effects of the top housing flexibility, a layer of composite material was bonded across the surface of one of the upper housings. This layer added some stiffness to the surface that would make it more difficult for it to deform. On a different housing, a layer of epoxy was applied to the back of the cantilever hook features to test against the rigidity of the snaps. Drops were conducted with these modifications and no screws were used for additional attachment. Most of the drop test failures were conducted from a height of 76 cm and done as a right-side drop as shown in Figure 4. As a comparison to the side drop, additional drops on the bottom side of the housing were conducted without screws.
302
Plastics Failure Analysis and Prevention
Summary of Drop Test Results The drop test data is organized in Table 1 to represent the progression of the design along with some additional solution theories. The modified design with four snaps resulted in the most disengagement. Adding the fifth snap decreased the number of snap failures. Adding one screw reduced the failures even more, and the second screw eliminated the attachment disengagement problem entirely. Taking the screws out and reinforcing the snaps did not improve the design, while the reinforced face reduced the frequency of disengagement. Dropping the housing on the bottom side, also, did not result in as many failures as the right-side drops. Table 1. Snap disengagement data
Drop conditions
# of drops
# of Disengagements of Snap #X #1
#2
#3
#4
#5
Total # Diseng.
1
Right side, four snaps, no screws
5
4
4
4
4
na
16
2
Right side, five snaps, no screws
6
3
4
0
0
0
7
3
Right side, five snaps, one middle screw
10
3
0
0
0
0
3
4
Right side, five snaps, two screws
4
0
0
0
0
0
0
5
Right side, five snaps, reinforced snaps
4
3
3
0
0
0
6
6
Right side, five snaps, reinforced face
4
1
1
0
0
0
2
7
Bottom side, five snaps, no screws
10
1
3
0
0
0
4
DISCUSSION POSSIBLE FAILURE MODES DURING IMPACT LOADING Comparing drop condition 5 to condition 2, it can be seen that reinforcing the snaps did not improve the performance of the design. Considering this, it is very likely that the failure mode of snap flexibility (mode 5 in Figure 3) is not contributing to failure during drop testing. The other listed failure modes all deal with some type of housing motion or deformation. Since the amount of impact loading was found to be higher than expected, it is likely that the parts are undergoing larger deformations than expected. This could cause the snaps
Attachment Design Analysis
303
to become disengaged in one of the proposed failure modes when the parts are moving relative to each other. Noting the difference in frequency of disengagement when dropped on different impacting surfaces (condition 7 vs. condition 2), and also when the fifth snap is included in its different orientation (condition 1 vs. condition 2), the cause of failure could be due to the number and orientation of the fasteners. This is because the disassembly motion is in the same direction as the loading force during right-side drops. Changing the loading direction during the bottom-side drops decreased the total number of disengagements. Also, adding the snap in a different orientation helped keep the parts together when the side snaps were being loaded. Considering condition 6, adding stiffness with the composite material reduced the ability for the upper housing to flex. This did help reduce the number of failures that occurred during testing. Unfortunately, instead of the housing deforming, the impact energy resulted in a different type of motion that caused the screw bosses to fail. They appeared to have been bent over due to the relative motion between the two parts.
CONCLUSIONS The attachment design for the housing using five cantilever-hole snap-fits was a very well implemented design under most conditions. These attachment features are very robust and can withstand high loads. It should be pointed out that during the impact loading, the snaps themselves never fail due to breakage. Also, stiffening the snaps during the testing did not improve the performance of the design. This shows that the sizing done for the features was appropriate for the application, even under the higher than expected impact loading. The problem was not snap-fit failure, but disengagement due to part flexibility. Overall, it appears that during impact loading, the parts undergo some relative motion that causes the snaps to disengage. Reducing the flexibility of the parts by making the top part stiffer was not a solution to the problem either. While it reduced the frequency of disengagement, it resulted in damage to the housing when the screw bosses broke. The impact energy was just transmitted to the next point of joining which could not handle the large loads. The solution to the problem lies in allowing flexibility in the parts while keeping the snaps engaged. The parts need to deflect together to prevent the snaps from coming disengaged during impact, or different types of snaps should be implemented. Reducing relative part motion can be accomplished with additional locating and locking features. By placing more cantilever hole features along the top and bottom edges, the housing will have a more robust attachment for impact loading. With features oriented in different directions, the parts will not be as likely to separate when the loading conditions would cause one or two of the snaps to disengage. This can be witnessed with the testing
304
Plastics Failure Analysis and Prevention
conducted with and without the fifth snap-fit along the top edge. Including this snap reduced the frequency of disengagement of the two closest snaps considerably. An attachment in the center of the housing could also be used to limit relative part motion. The screw used in the center provided the needed attachment but could be replaced by a snap-fit attachment. An annular snap (similar to a pen cap) or a dual cantilever hook feature which protrudes out of the inside surface of the top housFigure 5. Bayonet-and-finger snap-fit. ing and penetrates the bottom housing could be used to replace the middle screw boss. These features might not be the best selection in this application, though, noting the problems encountered with screw boss breakage. The bayonet-and-finger feature (Figure 5) would be a better selection for this application because it can provide some flexibility and has the ability to pivot, allowing part deflection, but limiting the amount of relative motion. More robust attachment can also be achieved by adding additional locating features. Walls and supports can be designed into the parts to hold the housing together more securely. Raised walls that result in overlap between the upper and lower housing can be included along the right and left sides of the bottom housing. By designing the walls to provide a tight fit between the parts, side impact forces will be transmitted to these walls instead of the snaps helping prevent disengagement. The tighter fit will also prevent any relative pivoting of the edges that could cause the snaps to disengage. Also, some smaller reinforcement could be built up around the snaps to provide a similar effect of reducing relative part motions locally around the snap to prevent disengagement. Finally, snap disengagement can be reduced by limiting the deflection of the snaps or by using a different type of snap. Additional features can be added to common snap-fits to help prevent them from becoming disengaged. Walls can be included around the snaps to prevent them from disengaging under side loading conditions. Also, different types of snaps with disengagement motions that are not in the direction of the impact loading would be useful for this type of application. The fingergrip snap shown in Figure 6 has a vertical disengagement motion, not encountered on side impact loading, but could during top and bottom drops. The torsional snap Figure 6. Alternate snap-fit concepts.
Attachment Design Analysis
305
shown in Figure 6 has a twisting disengagement motion that most likely would not be encountered under any condition. This type of snap could be disengaged with a screwdriver through the bottom access hole. Finally, using a simple cantilever hook snap-fit in a different orientation (see Figure 6) would help prevent test failure by changing the disengagement orientation such that failure modes 4 and 6 do not result in disengagement.
ACKNOWLEDGMENTS The authors wish to acknowledge John Kowalik of Lucent Technologies and Dr. Suat Genc of Plug Power for their support and assistance in this study.
REFERENCES 1. 2. 3.
4.
Boothroyd, G., Dewhurst, P., and Knight, W., 1994, Product Design for Manufacture and Assembly, M. Dekker, New York, NY. Chow, W. W-L., 1978, Cost Reduction in Product Design, Von Nostrand Reinhold, New York, NY. Messler, R. W., Jr., Genc, S., and Gabriele, G. A., 1997, “Integral attachment using snap-fits: a key to assembly automation, part 1 - introduction to integral attachment using snap-fit features”, J. Assembly Automation, Vol. 17, No. 2, pp. 140-152. Reliability Toolkit: Commercial Practices Edition, Rome Laboratory and the Reliability Analysis Center.
Avoiding the GIGO Syndrome – Combining the Real and Virtual Worlds in Analysis of Polymer Product Failures
John Moalli, Steven Kurtz, Robert Sire, Sanjeev Srivastav, Ming Wu Exponent Failure Analysis Associates, Menlo Park, CA, USA
INTRODUCTION Consumer and industrial product designs using polymeric materials continue their advance as alternatives to designs using more traditional materials, particularly metals. While considerable research and development work has gone into understanding physical and mechanical behavior of polymers, the amount of available practical information is small when compared with metallic materials. The rapid expansion of polymer applications sometimes suffers from potential failure risk associated with limited application experience and limited robustness of pertinent codes and design rules. We have discussed in the past the need for proper end-use testing prior to commercialization when application experience is minimal.1 Recent advancements in computational power and desktop analytical capability have provided a means of predicting polymer component behavior for design and failure analysis. As an example, desktop computational power now allows virtually any design engineer to numerically simulate complex mechanical and thermal behavior via a finite element analysis (FEA). Unfortunately, the predictive accuracy, and sometimes even relevance, of the outputs from such analyses are far from reasonable due to inaccuracies in inputs, including boundary conditions and material behavior. The objective of this paper is to describe some common mistakes made in FEA studies and offer guidelines for effective use of FEA as a tool in the analysis of polymer products. The reader should note that, in the interest of privacy, some of the case studies have been made generic.
FINITE ELEMENT ANALYSIS FEA is a method that subdivides a continuum into subregions of discretized elements, which then enables conversion of a problem with an infinite number of degrees of freedom
308
Plastics Failure Analysis and Prevention
to one with a finite number, thereby vastly simplifying the solution process. With the help of FEA, many of the complex structural and thermal problems faced in product design and failure analysis, which might otherwise be considered impossible to solve using traditional methods, can now be tackled with ease. It is, however, this ease, coupled with a lack of knowledge of FEA and its limitations, that can produce erroneous evaluations or poor designs, and in many cases, a false sense of security. In order to effectively model material behavior in response to load, most FEA codes require input of at least a few material properties. This first phase of model development is often the source of errors that lead to erroneous output. For example, many modelers use property data provided from resin supplier specification sheets that is likely derived from optimum processing conditions and does not always reflect properties measured from asmanufactured components. The modulus of elasticity and strength of a polymer can easily vary by 30 percent or more under different processing conditions. Specification sheet data may be useful for preliminary designs, but usually is not valid for failure or precise predictive analyses. A better way to determine mechanical properties for model input is to excise coupons from the part of interest and measure the actual behavior under controlled conditions. The latter can be accomplished through the use of mini tensile coupons or small punch testing.2 The small punch test uses a miniature disc-shaped specimen, about 0.5 mm thick, through which a punch is forced. The applied load and resultant deformations are measured to create a stress-strain curve of the tiny piece of material. Care must be taken when machining test specimens removed from the component; heat buildup from cutting or milling can change the part morphology and resultant elastic properties.
Figure 1. Scanning Electron Micrograph of UHMWPE small punch sample. The sample deformed readily and displayed good ductility.
Figure 2. Scanning Electron Micrograph of UHMWPE small punch sample that was believed to be similar to the sample shown in Figure 1. The sample displayed much less ductility even though the composition was identical.
Avoiding the GIGO Syndrome
309
A dramatic example of the necessity of mechanical property measurement for model input is shown in Figures 1 and 2 which display small punch specimens of what was thought to be identical UHMWPE samples; the specimen in Figure 1 displays much more ductility than its counterpart in Figure 2. If we had assumed identical properties for both materials, our models would have clearly been in error. Figure 3. Illustration of stress-strain behavior that will Another problem that can be encountered be assumed by FEA codes if minimum mechanical property data (modulus of elasticity and poisons ratio) when using specification sheet data is the limited are input. The actual behavior is typical of ductile engiamount of information available. If one restricts neering polymers. the model input to the modulus of elasticity and poisons ratio, the FEA can only simulate linear-elastic behavior until failure. Figure 3 shows, schematically, how model assumptions can differ markedly from actual, measured behavior. Many FEA codes will allow the modeler to input points from an actual stressstrain curve for precise constitutive behavior prediction. In some cases, it may be necessary to measure material response in both tension and compression as mechanical properties are known to differ under these conditions in some polymer systems. It is quite possible that elastic properties will have to be determined at multiple locations on a given component. If the part is known to have a skin-core varying morphology, coupons should be excised at and below the surface and properties measured at each location. If the geometry of the part results in substantial mold flow variation, multiple coupons should again be evaluated to capture spatial variations in properties. The appropriate parameters can then be assigned to particular elements in the model. None of the above mentioned methods consider time dependant behavior. For example, if a polymer being examined is known to creep, then input of viscoelastic parameters may be necessary. The code being used must also be capable of performing viscoelastic analyses and the modeler must be knowledgeable of the different ways the code can handle such behavior. Environmental conditions can also lead to time variant mechanical properties. A model that employs initial elastic constants for a material that becomes less stiff, for example, after exposure to an aggressive environment will not accurately predict component response. In some cases, an iterative approach which uses different mechanical properties at different time steps in the analysis is useful. Proper input of loading, both thermal and mechanical, and associated boundary conditions is also imperative for assuring output representative of actual behavior. If, for exam-
310
Plastics Failure Analysis and Prevention
ple, a support condition is assumed to be fixed when some displacement or rotation can actually occur, erroneous output may result. When modeling thermal response, it is critical that the temperature distribution across the part is accurately input; if one inputs a uniform temperature when actual conditions vary across the component surface, effects such as warping will not be properly considered. Finally, after the modeler is satisfied that the input is correct, the analyses should be run and examined for reasonable output. Gross errors are usually simple to detect; output stresses or deformations that are excessive or unreasonable are readily visible during post-processing. Subtle Figure 4. Finite element model of a hip implant. The input properties for the UHMWPE liner were detererrors can be more difficult to find, and can be mined from laboratory experiments. minimized by running known load cases and comparing to known part performance. A convenient, and often overlooked, means for validating a complex non-linear material model is to perform an FEA of the test specimen to predict the test results.
CASE STUDIES HIP IMPLANT An accurate FEA model of the ultra high molecular weight polyethylene (UHMWPE) used as acetabular liners in hip implants was desired to perform a detailed evaluation for the assembled component after implantation, when it would be subjected to cyclic loading during the patients’ during normal daily activities. When implanted, the UHMWPE liner is attached to a metal shell that is typically screwed into the patient’s pelvis. The required accuracy of the model demanded precise input of as-formed UHMWPE mechanical properties and a detailed understanding of the locking mechanisms attaching the polyethylene liner to the metal shell. The objective of this analysis was to quantify relative motion of the cup within the shell during the expected loads, and ultimately to quantify wear (material removal) from both the frontside and backside surfaces of the implant. The analysis proceeded in two stages. During the first phase, laboratory experiments were conducted during which a UHMWPE cup was inserted into metal shell, and an LVDT was attached to the UHMWPE to measure the relative motion of the cup in the metal shell. Material properties were obtained from measurements of the elastic and large deformation mechanical behavior of the UHMWPE.
Avoiding the GIGO Syndrome
311
During the second phase of the investigation, a fully parameterized, three dimensional finite element model was created of the femoral head, the polyethylene liner, and the metal shell, as shown in Figure 2. The loading, boundary conditions, and material properties based on laboratory experiments were incorporated into the model. The relative motion between the liner and the shell was found to compare favorably with the laboratory measurements, supporting the validity of the implementation. After the validity of the model was established, parametric analyses were performed to examine the effect of modifying the design of the metal shell, such as by adding multiple screw holes. THERMAL STRUCTURAL MODELING OF POLYCARBONATE OVERMOLDED WITH POLYURETHANE Crazing was observed in a rigid polycarbonate cover overmolded with flexible polyurethane. A preliminary finite element analysis based on an assumed thermal gradient of the overmolded material and invariant material behavior was performed to evaluate residual stresses. The analysis over-predicted residual stresses in the polycarbonate due to the thermal cycles in the overmolding process; the assumed temperature inputs were a likely source of the erroneous results. A program was initiated in which the mold was instrumented and actual in-process temperatures were measured. The actual peak temperatures were found to be much lower than those previously assumed based on the nominal nozzle injection temperature. The end result was that a more realistic set of input thermal boundary conditions produced lower predicted residual stresses that were more consistent with other observations. Most importantly, the stress magnitudes were shown to be comparable to those in the preovermolded substrate, thereby resulting in an approach to eliminating the stress and resultant crazing.
SUMMARY FEA is a useful tool for product design and failure analysis; it is widely available and can provide useful insight into complex problems. However, results of an FEA are only as good as the input material properties boundary conditions. Because the mechanical properties of polymers are highly dependant on processing conditions, we recommend the use of actual, measured mechanical behavior as the best input for constitutive modeling of polymers. Great care should also be taken to ensure proper application of input thermal and mechanical loads. Finally, FEA models should be validated by comparison of measured and predicted results from known load cases.
312
Plastics Failure Analysis and Prevention
REFERENCES 1. 2.
JE Moalli, “Translating Failure Into Success – Lessons Learned From Product Failure Analysis”; SPE ANTEC 1999, New York, NY. Kurtz SM, Foulds JR, Jewett CW, Srivastav S, Edidin AA: “Validation of a Small Punch Testing Technique to Characterize the Mechanical Behavior of Ultra-high Molecular Weight Polyethylene”; Biomaterials 18: 1659-1663, 1997.
Chapter 9 Case Studies
Case Studies of Inadvertent Interactions Between Polymers and Devices in Field Applications
Joseph H. Groeger, Jeffrey D. Nicoll, Joyce M. Riley, Peter T. Wronski Altran Materials Engineering, a Division of Altran Corporation, Cambridge, MA, USA
INTRODUCTION Polymeric compounds are selected for a wide range of applications by technical persons with a variety of backgrounds. Initial choices may be moderated by other specialists who are often unaware of the potential pitfalls and adverse interactions associated with the use of cost-effective or inappropriate alternate materials. Manufacturers who provide subcomponents may not be included in the design reviews of finished products into which their components are being used. Additionally, suppliers of commercial polymeric materials may be unaware of how their materials are being applied. As a result of these and other considerations, materials selections may be made based on a review limited to basic engineering properties. Considerations of long-term performance and response to specific operating conditions requires a degree of attention and insight that may be overlooked. Several case histories are cited in which some aspect of materials selection and design were deficient in the application. A thermally activated electrical switch formerly made with a phenol formaldehyde thermoset resin was redesigned to include a thermoplastic resin. Localized heat associated with the arcing activity of the switch contacts caused thermal erosion of the housing, releasing reactive sulfur compounds which then reacted with the electrical contact faces, causing irregular performance and eventual contact welding. A pressure relief device in a consumer product was found to have highly variable performance as a result of extensive processing aid additions to the base polymer, selection of poor quality raw materials, and no attention to a root cause analysis with a review of the compound. Plasticizers released from PVC wire insulation at elevated operating temperatures wicked
314
Plastics Failure Analysis and Prevention
along the conductor strands and onto relay contacts, resulting in a power plant shutdown. Components from a pharmaceutical product container were found to be exuding phthalate compounds which were not expected based on an initial review of the raw materials. These cases are presented as constructive examples for those seeking to maximize the performance and useful life of devices making use of polymeric components through an integrated materials selection and design approach.
DECOMPOSITION OF THERMAL SWITCH A thermal limit switch used in a number of domestic and commercial appliances was historically manufactured with either a ceramic or a crosslinked phenol formaldehyde, providing many years of reliable service. A change in materials had been implemented to facilitate processing, resulting in a housing made with thermoplastic polyphenylene sulfide (PPS). The housing contained silver-laminated bronze electrical contacts, one of which was mounted on a bimetallic arm to provide thermally-controlled switching action. Failures of this switch were encountered wherein the contacts were found to weld together, resulting in a thermal runaway condition caused by a failure to interrupt current to the heater that the switch was intended to control. A forensic review of representative failed switches was undertaken. Figure 1 presents a scanning electron micrograph of the surface of a contact removed from a failed switch. On the surface, many melted areas are clearly visible. Some of these are flat, showing the previously molten condition of the metal contacts. Metallographic cross-sections through such a contact showed severe localized melting. Elemental analysis of the contact surface indicated that silver sulfide was present. This compound produced an insulating layer on the surface of the contact, resulting in erratic current flow and localized heating due to limitation of the available contact surface area. Switches in various stages of degradation were operated with thermocouples placed on the contacts and housing. Measurements indicated significant resistive heating, merely due to flow of the rated current. Chemical analysis of the polymer heated to the as-found level, using gas chromatography and mass spectrometry (GC/MS), confirmed formation of hydrogen sulfide, carbonyl sulfide, sulfur dioxide, hydrogen, and methane. Examination of the switch housing interior surfaces surrounding the contacts revealed significant erosion of the polymer as shown by the light Figure 1. Surface of contact showing raised areas colored oval region in Figure 2. Closer examination where welding occurred, 150x.
Case Studies of Inadvertent Interactions
Figure 2. Interior surface of switch housing showing polymer erosion, 9x.
315
Figure 3. Top view of pressure relief device.
revealed the glass and mineral reinforcement particles within the PPS compound standing in relief, due to polymer pyrolysis. This damage was due to the intense localized heating produced by arcing as electrical contact between the switch contacts was established then broken during normal operation. The combined evidence of contact melting and PPS pyrolysis suggested short-term temperatures in excess of 600°C.
INCONSISTENT PRESSURE RELIEF MEMBRANE A pressure relief membrane used in a consumer product was found to exhibit erratic performance both in quality assurance testing and in the consumer market. The pressure relief device was a critical component and played an integral role in product function and safety. The device was manufactured using a compounded thermoplastic polypropylene which was injection molded into the necessary form. As can be seen in Figure 3 the molded part is quite complex in design; consisting of numerous ribs, radial formations and most importantly, the thin membrane which acts as a pressure rupture diaphragm. The latter is coined in the injection molding process. Investigation of the device revealed many areas of misapplied designs and a general focus on processing performance instead of functionality. The thermoplastic compound which was used to fabricate the units made use of a fairly complicated formulation. The original base resin was dropped from the supplier’s product line and alternates were substituted. In conjunction with these changes, increased device anomalies and difficulties controlling the burst pressure range were experienced. After a preliminary materials investigation of the disclosed formulation, interactions of the materials being used were
316
Plastics Failure Analysis and Prevention
identified as being inordinately complex and in some cases inappropriate for this application. Organic chemical analyses of representative devices were conducted using GC/MS. This method was selected to confirm the identity of the organic ingredients and processing aids in the questionable formulation. GC/MS analysis of the seals revealed significant formulation variations between different lots of material. It was determined that the use of additives such as the antioxidants, antiblocking agents, internal lubricants, and other processing aids was inconsistent. The most significant variations were among materials not specified in the formulation. Processing aids such as silicones (used as internal lubricants to modify flow behavior), plasticizers (typically used for increasing impact resistance and adding flexibility), and waxes (used as lubricants and flow modifiers) were noted to be present in many of the device lots. These components appeared at random and were not used consistently. It was suspected that they were added as on-line processing aids to assist with mixing by the compounding operators and/or to achieve a target melt flow index. The formulation suffered from years of incremental modification for performance and processing issues which often suppressed the symptoms but never addressed the root causes. For example, there were three agents listed in the formulation which served as antioxidants. Due to the nature of their chemical functionality, these materials did not enjoy a positive synergy. Instead they competed in the formulation causing none of these materials to offer as much protection to the resin and other organic components in combination as they would when used individually. The antioxidant package was further complicated when a review of their functional characteristics was completed. Originally, the molded pressure relief device suffered from a reaction with copper within the contacting unit surfaces. A metal deactivating antioxidant was added to the formulation to correct this problem. A review of the formulation clearly indicated that the original antioxidant was an amine (nitrogen-hydrogen) compound. This antioxidant sustained limited thermal decomposition during processing, leading to the production of amine compounds. These reacted with copper, leading to the formation of blue-colored copper compounds. While the addition of the metaldeactivator was successful in reducing this occurrence, the original antioxidant was left in place. The replacement and original antioxidants were not chemically compatible, nor was the amine antioxidant stable with respect to the antioxidant supplied in the base polypropylene resin. A third antioxidant was then added to improve oxidative stability. A different problem was noted when a scanning electron microscope (SEM) was used to examine selected areas of representative seals. The high magnification of the SEM provided a view of the relative size of the individual filler particles and their alignment in key areas such as the diaphragm. Examination revealed that the filler materials had a tendency to agglomerate in this region and that the overall filler concentration in the diaphragm area
Case Studies of Inadvertent Interactions
Figure 4. Micrograph of diaphragm cross-section, 605x.
317
Figure 5. Micrograph of diaphragm corner, 226x.
was inconsistent throughout many devices. As shown in Figure 4, the talc particles were quite large when compared with the overall thickness of the diaphragm. As illustrated in this micrograph, the particles aligned in the plane of the membrane and created a stacking effect. In this case, the shape of the particles was inappropriate due to the flow mechanics in the mold cavity. Figure 5 shows the corner at the edge of a representative diaphragm. The filler particles in this area were also dramatically aligned along the curvature of the diaphragm. This suggested that the resin flow in this area during molding was restricted by the presence of the talc particles. This caused the residual stresses in the diaphragm area to be quite high and the particle size of the talc to vary depending on the level of flow restriction during injection. The effect of the talc particle size variation on the inconsistent performance of the seals was significant. This characteristic mainly affected the flow rheology of the compound under high shear conditions during injection molding. The talc particle size, in comparison to the diaphragm thickness, also lead to an erratic influence on the tear characteristics during product performance. Talc agglomeration and absence of bonding with the base polymer further contributed to poor performance. Inconsistent diaphragm burst performance was caused by a combination of chemical, physical and rheological phenomena. The lots of devices which exhibited a particularly high burst pressure were the result of a very fine particle size talc in conjunction with a low concentration of processing aids. The increased strength of the base resin and lack of large talc particles for burst initiation necessitated high burst pressures. The devices which exhibited lower diaphragm burst pressures suffered from a combination of large talc particles and an absence of lower molecular weight polymer to assist with the flow and wetting of the filler. This resulted in high orientation effects which led to very high residual stresses causing premature failure. These anomalies illustrate the combined effects of the uncontrolled chemical
318
Plastics Failure Analysis and Prevention
additives and random talc particle size on the consistent performance of the compound. In this formulation, even if extreme care were taken in manufacturing, the number of materials involved and the inherent variability and performance of the talc made it virtually impossible to produce a consistent product.
PLASTICIZER BLOOM FROM PVC CABLE JACKETS During inspections at a nuclear power plant, green liquid deposits were found concentrated on the surface of selected low voltage cables, at their terminations as well as in the instrument panel in which these cables ended at connections. The cables were rated at 600 volts and incorporated a cross-linked polyethylene (XLPE) insulation with a polyvinyl chloride (PVC) jacket. The estimated age of the cables was 20 years. The green liquid deposits were determined to be non-drying, with a high viscosity, and good lubricity. Analysis of this liquid by Fourier Transform Infra-Red Spectroscopy (FTIR) confirmed that it was mostly adipic acid diethyl ester. This compound is a common plasticizer for PVC and is typically yellowish in color. An FTIR absorption peak unaccounted for by adipic acid diethyl ester was assigned to a silicone fluid (diphenylsilane). This may be attributed to a second plasticizer used in these cable jackets. Samples of the liquid were pyrolized and the residue was analyzed with energy dispersive X-Ray analysis (EDX). This revealed the presence of copper with traces of aluminum, silicon, calcium, iron, and lead. The presence of copper salts in the fluid was responsible for the noted green color. The presence of these green fluid deposits closely followed a record ‘heat wave’ in this particular region. It was deduced that this elevated regional temperature caused the sudden appearance of these exuding plasticizer compounds from the PVC cable jackets. These compounds can cause severe consequences in electrical systems due to their insulating properties. If these compounds were allowed to migrate into electrical switches, relays, or meters they would inhibit proper performance. In this particular case, the plasticizer impinged on the jackets of adjacent cables, causing them to swell then split. In another identical occurrence, a plant shutdown resulted when plasticizer crept onto the surface of electrical contacts used for a pump motor relay.
EXTRACTS FOUND IN PHARMACEUTICALS The presence of two plasticizers, dioctyl phthalate (DOP) and diisooctyl phthalate (DIOP), in a drug formulation caused significant concern to the pharmaceutical companies since aromatics of this type are under regulatory scrutiny. Investigation into the origins of these contaminants led to analytical review of elastomeric components of the product container. Extensive GC/MS analysis isolated the source of the DIOP as being the elastomer raw mate-
Case Studies of Inadvertent Interactions
319
rial. Further research indicated that the supplier of this elastomer was adding DIOP during manufacture to act as a melt-flow modifier to control the Mooney Index of the final product. The DOP, however, was traced to contamination from the polymer compounding equipment. Frequently, oils used to lubricate mixing equipment exude into the compound being produced through dust seals, for example. Knowing this, manufacturers will often utilize lubrication products which are compatible with the polymer products that are produced. In this case, however, the oil utilized contained DOP which would be acceptable in many thermoplastic compounding applications, not slated for medicinal use. The resultant extraction of DOP from components of the product container, however, was not acceptable.
DISCUSSION Development of thermoplastic and thermoset polymer compounds is a mature science that continues to grow with the development of new types of additives, changing regulatory requirements, and proprietary considerations. The selection of all materials that are incorporated into a compound may follow lines that are not always clear. Some ingredients may be outdated. Others may have been added for a customer-specific end use and the compound later became available for the general market. A very wide range of off-the-shelf compounds are available for engineering applications. Many will fit into the existing requirements or designs and/or processes may be altered to accommodate the compound that best fits the needs. These choices, though, are often limited to the general engineering/technical properties without sufficient detailed consideration of the materials in context of the application. An ideal situation is one in which the end-use manufacturer has available the equipment necessary to develop a polymer compound specifically suited to an individual application. In this clean sheet approach, each ingredient may be carefully considered in context of the application, aging characteristics, processing effects, and synergy with other formulation components. Compounding facilities need not be directly available; contract organizations are available and many of the commercial polymer compound suppliers offer custom compounding services. Analyses of plastics failures and contamination issues often indicate that it is necessary to return to the basics and re-examine the material in context of the application. With this approach, a polymer would be formulated using a minimum number of ingredients, each of which would be the most appropriate and efficient for the end use. By reducing the number of ingredients, the controls necessary for each supplier are greatly simplified.and the potential for adverse interactions reduced. Many raw materials are more complex than may be apparent and, in some cases, the ‘hidden’ ingredients may be detrimental to an application. Virtually all commercial elastomers are supplied with an antioxidant already included and
320
Plastics Failure Analysis and Prevention
the type may change periodically. Master-batching agents and processing aids, such as calcium stearate, may be used when adding antioxidants to a raw polymer. Crosslinking additives and their synergists are another source of antioxidants and other compounds. Crosslinking is a chemically challenging process in which thermal decomposition of a reactive peroxide is typically used to provide free radicals. This requires an additional antioxidant to protect the polymer, while reaction products including acetophenone, cumyl alcohol, and acetic acid become available to interact with the other raw materials or additives. In the case of the thermal limit switch, the choice of materials for the housing inherently led to a reduction in the useful life of the device. The stability and useful life of the switch could be readily enhanced through the use of a polymeric housing that does not produce reactive gaseous products. Many thermoset materials are available, as are ceramics. While the near-term economy of using a thermoplastic material may have appeared attractive, the long-term effect on performance may not have been readily apparent when a material substitution was made. In the second situation corrective measures were implemented so that predictable and consistent performance of the pressure relief device could be attained. Compound reformulation took place which included the careful selection of a clean homopolymer base resin, a specially designed and compounded antioxidant and a low aspect ratio, small particle size reinforcement. The compound simplification, in combination with highly functional components, allowed for exceptional performance and reliability. In the example of plasticizer bloom from a set of cables, it is interesting that the simple loss of a compounding ingredient could lead to such indirect, but major consequences. In this case, exposures to long-term conditions of elevated temperature could be surmised, based on the application and service environment. Grafted plasticizers are available. Alternatively, though, a complete reconsideration of the material in this environment would have been beneficial. A polymer compound that is inherently flexible would not involve a plasticizer and the potential adverse effects of its loss. Finally for the pharmaceutical container component example, reformulation of the raw polymer compound, as well as substitution of machine lubricant with a food-grade aliphatic mineral oil was necessary, followed by substitution of increased purity raw materials, before use of this material could be continued.
CONCLUSIONS It is important that the total life cycle of polymeric compounds be considered in context of the end-use application. Some basic guidelines can be developed from a review of situations in which the process was not optimized. The application should be well understood in terms of stresses (thermal, chemical, physical, radiation, etc.). Near- and long-term exposures must be considered.
Case Studies of Inadvertent Interactions
321
Review the candidate or existing material with a fresh perspective and careful attention to all raw materials, their quality, and roles. Simplicity of design facilitates processing, cost control, longevity, and quality assurance. This requires that raw materials be inherently suited for the compound, rather than placing a strong reliance on complex additive packages. The performance of a compound cannot be limited by processing or post-processing handling. Attention to detail is necessary to assure that the intended and realized formulations are identical. While some of these suggestions may seem tedious, it is often the case where shortterm economy and lack of application-specific insight may lead to significant losses when a poorly selected material fails in service.
REFERENCES 1 2 3 4 5 6 7 8 9
Hoffman, Werner (1989). Rubber Technology Handbook. New York, NY. Hanser Publishers. Bhowmick, Anil and Howard Stephens, Eds (1988). Handbook of Elastomers. New York, NY. Marcel Dekker, Inc. Schnaebel, Wolfram (1981). Polymer Degradation: Principles and Practical Applications. New York, NY Hanser Publishers. Sekutowski, Dennis (1992). "Inorganic Additives". in Plastics Additives and Modifiers Handbook, Jesse Edenbaum, Ed. New York, NY. Van Nostrand Reinhold. Gachter, R. and H. Muller (1993) Plastics Additives Handbook. Cincinnati, OH. Hanser Gardner Publications. Charrier, JM (1990). Polymeric Materials and Processing. New York, NY. Hanser Publishers. Barth, H. and Mays, J. (1991). Modern Methods of Polymer Characterization. New York, NY. John Wiley Publishers. Engineering Plastics and Composites (1990). Metals Park, OH, ASM International. Rauwendaal, C. (1991). Mixing In Polymer Processing, New York, NY. Marcel Dekker.
Case Studies of Plastics Failure Related to Improper Formulation
Myer Ezrin and Gary Lavigne University of Connecticut - Institute of Materials Science, Storrs, CT 06269-3136
INTRODUCTION Plastics failures can usually be ascribed to the material, design or processing, or combinations thereof.l Previous ANTEC papers by the authors have dealt with all of these factors. In this paper we focus on the formulation or composition aspect of the material in failure. Three of five cases cited involve adhesive failure. In all cases chemical analysis was required to determine if the formulation was a major contributing factor to failure.
CASE STUDIES 1. ADHESIVE FAILURE OF A MULTILAYER FILM The formulation called for a certain order of films and surfaces for good bonding. Failure to bond well at certain layers was found to be due to improper order of surfaces, so that two in contact would not bond. Simple reversal of one layer solved the problem. The error was found by infrared spectroscopy of surface composition. 2. FAILURE OF EPOXY RESIN TO HARDEN An adhesive using a two-part epoxy resin failed to harden in the field. An analysis showed that the hardener had been omitted, thus absolving the supplier of the adhesive from responsibility. The method of analysis was thermal desorption gas chromatography/mass spectroscopy (TD/GC/MS)2,3 using the Direct Dynamic Thermal Desorption Device developed at the University of Connecticut by Gary Lavigne. Sample contained in a glass tube, held in place by glass wool, is devolatilized in the heated GC injection port. Typical time and temperature are two minutes at 300°C, after which the sample tube is withdrawn using gas pressure. The tube, held upside down, is in the closest possible direct contact with the GC column, thus eliminating transfer lines or paths. The dynamic feature is that continuous helium gas flow carries off volatiles as they are produced, transferring them to the head of the column at low temperature prior to the usual temperature programming of the GC. The
324
Figure 1. Thermal desorption GC/MS chromatogram of epoxy resin that did not harden. Y axis - ion count; X axis - retention time in minutes. Peak no. (1) - high content of unreacted bisphenol A - diglycidyl ether.
Plastics Failure Analysis and Prevention
Figure 2. Thermal desorption GC/MS chromatogram of normal two-part epoxy resin that hardened. Y axis - ion count; X axis - retention time in minutes. Peak no. (1) very low content of bisphenol A - diglycidyl ether.
GC/MS used is the Hewlett/Packard 6980, the GC column was SGE PBX5 0.22 mm ID x 25 meters 0.1 micron film. Temperature program was 150°C/minute to 300°C. Figure 1 is the chromatogram of adhesive that did not harden. It contains a high content of unreacted bisphenol A diglycidyl ether. Figure 2 shows normal two part resin. The diglycidyl ether (unreacted epoxy resin) is gone and other peaks from the hardener are seen. This analysis provided unequivocal proof that the hardener had been omitted by the customer in the field. 3. DELAMINATION OF TIN COATING FROM COPPER SHIELD OF AN ELECTRICAL POWER CABLE Tin coated copper shield over a power cable was held in place by an impregnated fabric tape wrap. Removal of the tape during cable installation caused tin to be removed from the copper shield. Infrared analysis detected calcium carbonate in quantity on both surfaces of the tape, i.e., throughout the tape. Normal tape that did not cause tin to be removed did not contain calcium carbonate. While it is not clear why calFigure 3. Subtracted IR spectrum of problem tape minus cium carbonate would cause the problem, there normal tape. Y axis - absorbance; X axis - wavenumbers was no question that it was associated with in cm-1. Peaks (1), (2)-calcium carbonate. improper formulation of the tape. Figure 3 was obtained by subtracting the spectrum of normal tape from that of the problem tape. The peaks near 1400 and 900 cm-1 are definitive for the carbonate.
Case Studies of Plastics Failure
Figure 4. IR spectrum of unused grease (4a) and used degraded grease (4b). Y axis - absorbance; X axis wavenumbers in cm-1. Peak no. (1) - new peak in used degraded grease.
325
Figure 5. Thermal desorption GC/MS chromatogram of unused grease. Y axis - ion count; X axis - retention time in minutes. Peak no. (1) - triphenyl phosphate (overload).
4. STIFFENING OF GREASE LUBRICATING A MOVEABLE SCREW IN A SERVO MOTOR This failure occurred during proof testing of a servo motor in development, evaluating the ability of the grease on a moveable screw in the part to provide smooth movement of the screw back and forth. The grease lost its lubricating quality, causing the motor to bind. The grease contained a fluoropolymer, a polyol ester fluid and triphenylphosphate. Infrared analysis and TD/GC/MS revealed that the phosphate content was greatly reduced in degraded grease and carboxylic acids had been hydrolyzed from the polyol ester. The reaction product of hydrolysis of triphenylphosphate is phosphoric acid, the likely cause of hydrolysis of polyol ester. It also contributes to corrosive attack of the metal screw. Elemental analysis of grease showed metal content from the screw. Figure 4a is the IR spectrum of unused grease, and Figure 4b is for degraded grease. In 4b, the ester band at 1740 cm-1 is reduced relative to the carbon-hydrogen bands near 3000 cm-1. A new band formed at approximately 1650 cm-1 due to COOH of carboxylic acid. TD/ GC/MS provided further insight into chemical changes that had occurred. Figure 5 is for unused grease. Major peaks are for triphenylphosphate (overload) and a mixture of polyol esters. Figure 6 is for failed grease, showing a much lower content of phosphate and new peaks for heptanoic, octanoic and decanoic acids. The latter formed from partial hydrolysis of the polyol ester. The same GC column and thermal desorption conditions were used as in case 2 above. Figure 7 shows TGA thermograms of unused and degraded grease. The weight loss at 500-600°C in 7a is the fluoropolymer. Inorganic ash content was 2% for unused, and 15% for used grease. The higher ash content was shown to contain elements from the screw.
326
Plastics Failure Analysis and Prevention
Figure 6. Thermal desorption GC/MS chromatogram of used degraded grease. Y axis - ion count; X axis - retention time in minutes. Peaks (1), (2), (3) - heptanoic, octanoic, decanoic acids; peak (4) - greatly reduced content of triphenylphosphate.
Figure 7. Thermogravimetric analysis of unused grease (---) and used degraded grease (__).
This case illustrates how an additive (triphenylphosphate) can change in service giving a new material (phosphoric acid) which can drastically affect performance. IR and GC/MS were essential in revealing the chemical changes that are responsible for this failure. 5. UNINTENTIONAL COLOR IN A PORTION OF AN ELECTRICAL CABLE PE coaxial cable had unintentional color which faded out after several feet. Concern about possible effect on electrical performance led to attempts to determine the source or cause of the color. Contamination with regrind PE from another product was suspected. Infrared showed no difference between colored and normal PE. Thermal desorption GC/MS at 300°C also showed no difference. These analyses indicated that the material responsible for color was probably present at very low level. While that provided some assurance that the effect on electrical properties would probably be low, a more positive identification of the contaminant was needed to satisfy the customer and the manufacturer. Supercritical fluid extraction was performed to help isolate from the PE whatever was causing the color. A simple device made here was pressurized to 6000 psi with carbon dioxide at room temperature. 77 mg of PE, first ground to small particle size, was extracted. The CO2 extract was deposited directly in thermal desorption glass tubes. TD/GC/MS at 350°C gave the chromatoFigure 8. Thermal desorption GC/MS chromatogram grams in Figure 8. Figure 8a (colored PE) shows of supercritical fluid extract of normal PE (8a) and colored PE (8b). Y axis - ion count; X axis - retention material at 20-30 minutes GC retention time which time in minutes.
Case Studies of Plastics Failure
327
is not present in normal PE. Mass spectral analysis found diglyceride of palmitic acid, and other fatty esters, in the colored PE extract. The weight percent of material from the contaminated colored PE was so low that only by concentrating it by supercritical extraction was it possible to detect it by GC/MS. Subsequent analysis of candidate recycled PE containing the fatty esters found in the cable in question provided a means of locating the source. If the contaminant PE had not been colored, its presence would not have been known. While the chemical nature of the contaminant proved to be no problem to electrical properties, the manufacturer had no choice but to try to locate the source of the contaminant, be it serious or not.
CONCLUSIONS Infrared spectroscopy has been used for many years as a key method of analysis to aid in determining causes of failure and variable performance. This paper gives examples of the utility of IR. It gives spectra or fingerprints of the whole material. Without separation of components, so that each can be identified, IR is limited in the information it can furnish. TD/GC/MS has become as important, if not more important, than IR. By separating a mixture into components, and providing mass spectra of each, details of composition are learned that IR alone cannot provide.
REFERENCES 1 2 3
M. Ezrin, Plastics Failure Guide - Cause and Prevention, Hanser, 1996. M. Ezrin and G. Lavigne, Failure Analysis Using Gas Chromatography/Mass Spectroscopy,l/ SPE-ANTEC, 2230-2233, 1991. M. Ezrin and G. Lavigne, Application of Direct Dynamic Headspace GC/MS to Plastics Compositional and Failure Analysis, SPE-ANTEC, 1717-719, 1992.
Translating Failure Into Success – Lessons Learned From Product Failure Analysis
John E. Moalli Exponent Failure Analysis Associates, Los Angeles, CA, USA
INTRODUCTION Given the large number of plastic product failures that have occurred over the years, one would expect that, at least among products of similar complexity, the number of failures should be reduced with time simply from gaining experience with them. Although this may be true for a single manufacturer, it does not seem to translate across the industry as a whole; certain failure modes tend to repeat themselves with undesirable frequencies. From the academic standpoint, many researchers have developed tools in the areas of fracture mechanics, stress relaxation, physical aging, fatigue, and friction. Unfortunately, the average plastics designer does not have access to these sophisticated and sometimes cumbersome techniques. The objective of this paper is to describe some common failure modes and present them in a simple fashion that should enable the reader to learn from others’ mistakes. The reader should note that, in the interest of privacy, some of the case studies have been made generic.
THE DESIGN PROCESS
Figure 1. Schematic of a typical process used in design of plastic components. The product is conceptualized, analyzed, sized and then fashioned into a prototype. The prototype is tested and evaluated and any anomalies revealed are corrected such that the process continues until the final commercial product is realized. Redesign and improvement may continue through the product cycle as appropriate.
The design process is sometimes illustrated as a circular one (Figure 1) which involves conceptualization of a product, construction of a prototype, evaluation and testing of the prototype, and return to the conceptualization/design phase to correct any deficiencies discovered during testing. Clearly, this is a great simplification, with each one of these phases representing multiple steps.
330
Plastics Failure Analysis and Prevention
For example, in order to design and build a prototype, one must come up with dimensions that satisfy the relevant design criteria. A common basis for sizing a part involves its capability to support stress. When considering the stresses on a part, one must evaluate the residual or built-in stresses from processing, as well as external of applied stresses. The sum of these stresses (the total stress) should be kept less than the working stress, a value often supplied by a resin supplier which is less than the strength of the material and considers the operating environment and type of load. By way of example, a certain resin may have a strength of 70 MPa under ambient conditions. Under elevated temperatures and continuous application of load, the supplier may recommend limiting the working stress to 17.5 MPa. A common error in this early design phase is to size the part based on external loads only and limit the applied stresses to the allowable working stress; ignoring the contribution of residual stress can potentially have catastrophic effects. The designer is left little recourse in the event of a product failure, as the material is now being used in a manner inconsistent with the supplier’s recommendations; allowable working stresses have been exceeded. Evaluation of stress in and of itself can also lead to problems in finished products. With the advent of fast personal computers and inexpensive finite element analysis (FEA) codes, designers can be presented with tools that can, in the absence of accurate inputs, give misleading results. A common mistake is to used published mechanical property data as input to the FEA. The published data may be fine for a comparative analysis between different resins or grades, or perhaps for examining the relative effects of a geometry change. However, accurate results are more readily attained by measuring the mechanical properties of coupons excised from the part itself; it is well known that processing can affect material behavior. A proper FEA model will also consider the anisotropy in mechanical properties which can develop in certain types of processing. All too often, this variation is ignored and unexpected failures result. The testing phase of the design process is also one that must be carefully executed. The type of loads which a component may experience (i.e., impact, fatigue, thermal, etc.) must be considered. Although it is impossible to envision every loading scenario, those which are reasonably expected in the service environment should be evaluated. This type of testing is sometimes called end-use testing, and involves subjecting a component to the loads, load cycles, temperatures, and environments (i.e., chemical exposure or UV) that it may experience during its life. Often one or more of these variables is changed to accelerate the time of which the end-use test is conducted. We have examined many failures which could have been identified and eliminated through proper end-use testing. Accurate definition of the environment in which the component is end-use tested is also important. It is simple to say that a part will be used outside and therefore should be
Translating Failure into Success
331
tested with UV exposure, but what about more abstract issues such as the material with which the part will be cleaned? Could it be exposed to adverse environments during shipping or storage? If the part is available in different colors, could color be a factor in the temperature the part experiences in service? Additionally, what type of testing should be done to determine the useful life of a product? One method employed in the plastic pipe industry, and Figure 2. Schematic of curve used to evaluate strength of a product later in its service life. If product is only tested to 1,000 hours and then extrapolated to described in standard test meth100,000 hours as shown, the knee in the curve is missed and strength is overesods, is to test the part at differtimated. ent stress levels and measure the time to failure. Higher stress levels fail at lower times and, after application of a safety factor, an extrapolated time is used to define a limiting stress level. More than one pipe has failed because the tester did not consider the change in material behavior, and has extrapolated data erroneously as shown in Figure 2. The standard test methods attempt to limit this kind of problem by forcing the experimenter to have sufficient data for a statistically valid extrapolation. One approach to defining a testing program is to evaluate the relative risk of different failure scenarios using a failure modes and effects analysis (FMEA). The product of the FMEA is a table which quantifies the relative risk of different scenarios and aids the designer in choosing which modes to examine.
LEARNING FROM FAILURES Even the most thoroughly designed and evaluated products can experience failure. Proper analysis and examination of field failures can allow the designer to determine if the failure was from a one-of-a-kind unexpected event, or perhaps a load case that was not considered and should have been. If a program is initiated which tracks incoming field returns, repetitive failures can be discovered and remediated. In products that are related to consumer safety, such programs are a necessity; if the data are not acquired gradually over time, gov-
332
Plastics Failure Analysis and Prevention
ernmental agencies will likely require compilation in a rapid, crisis type mode in the event of a recall. Tracking programs can vary from simple to complex, but at a minimum should attempt to capture demographics, the user’s description of the failure, actual versus intended use (in terms of environment and loading), and the manufacturer’s assessment of the cause of failure. If these failure modes are ones captured in an FMEA, the results of that analysis can be modified to reflect new failure frequency data, and any remedial measures can be imposed to aid in overall risk reduction. Although we do not formally track all of the plastics failures we have examined, we have seen some repetitive themes. In terms of load cases that are not properly considered by the designer, one of the most common theme that appears in our failure analysis activities is thermal expansion. When selecting plastics for replacement of metals, the fact that plastics typically move an order of magnitude more than metals must be considered. This also becomes an issue when plastics are rigidly attached to metals; if one restricts the contraction that plastics undergo with cooling, tensile stresses and cracks may develop. Restriction of movement during heating often causes plastic components to buckle. These general rules can be reversed, however, with composite materials, which can have small or even negative thermal expansion coefficients. The issue of flame retardants in plastics is also one that seems to be prevalent today. Flame retardants are essential in certain applications, but are not universally appropriate in all plastics. For example, polyolefins are known for their propensity to burn once ignited -a property that can be greatly reduced by the addition of a flame retardant package. In some cases, though, polyolefins are best used without flame retardants even in applications with known risk of ignition. These include conditions where the retardants can compromise key properties, or where the product is protected from burning by monitoring or fire-suppression equipment. Flame retardants also do not prevent combustion in all cases. Flame retardant polymers can burn readily if the geometry of the part is suitable for combustion; end-use evaluation on the final component rather than on test coupons can be especially relevant in burning characteristic studies.
CASE STUDIES COMPONENT HOUSING An ABS component housing was found to be cracking with a failure rate of about ten percent. Fracture surface analysis showed that the failures initiated on the inside of the part, thus reducing the concern for external environmental effects. When the interior components were examined, it was found that some of them were metallic and rigidly attached to the
Translating Failure into Success
333
Figure 3. Graphic illustrating failure in a component housing. The metal box, shown in the right side of 3A, was rigidly attached to the housing with 4 screws as seen in 5B. When the part was cooled, restricted contraction of the ABS housing caused tensile stresses to develop that lead to subsequent cracking.
housing. Subsequent experiments revealed that rapid cooling of the housing could reproduce the observed cracking; the metal was restraining thermal contraction of the plastic, causing elevated tensile stresses and cracking (Figure 3). Proper consideration of thermal stresses and end-use testing would have identified and eliminated this failure mode. BRAKE CUP A rubber cup used as the primary seal in a truck master cylinder was found to be torn after inspection following a severe accident. It was contended that the cup was defective and contained manufacturing defects. Fracture surface analysis revealed a rapid fracture and severe component wear. A key tool in successfully the cup’s performance was an FEA that showed the stresses in the cup under use were inconsistent with failure scenarios proffered by others. Even though the cup was quite small, coupons were excised from a similar one and tested for mechanical properties that were input into the FEA. The actual, measured properties were different than the published ones and thus were critical to developing an accurate and representative model. COMPOSITE PRESSURE VESSEL A filament wound fiberglass self contained breathing apparatus (SCBA), with a service pressure of about 31 MPa, ruptured during storage in a fire truck. Inspection of the vessel revealed a regular cracking pattern consistent with stress corrosion behavior (Figure 4). Chemical analysis of the surface showed the presence of acidic (low pH) components found
334
Plastics Failure Analysis and Prevention
Figure 4. Photograph of stress corrosion cracking in filament wound fiberglass pressure vessel with an aluminum liner. The fiberglass overwrap failed from stress corrosion cracking after exposure to an acidic substance. Many designers are unaware of the susceptibility of fiberglass to acidic environments.
in aluminum rim cleaners. These compounds can severely stress corrode glass fibers under stress. Many designers are not aware of such stress corrosion phenomenon; we have seen this type of failure several times in the last few years. SWAMP COOLER A plastic swamp cooler was designed and constructed by a manufacturer who sought market differentiation from metallic coolers; the plastic device would not rust. The cooler was made from polypropylene and contained an internal pump that was known to fail with regular frequency. In one case, this pump failed and caused the cooler to ignite and initiate a subsequent house fire (Figure 5). This cooler was unmonitored on the roof of a house and contained a known ignition source. It thus represented a perfect candidate for addition of flame retardants; end-use testing would have likely revealed this to the manufacturer.
Translating Failure into Success
335
Figure 5. Polypropylene swamp cooler 1 minute and 44 seconds after ignition from an external flame source (A). The lack of flame retardant in this product lead to the house fire shown in B.
SUMMARY Examination of many different plastic component failures has revealed some recurring themes. As a simple tool, we present the pneumonic REDUCE to remind designers of some actions that may help reduce failures in their components:. Re-Evaluate – If failures come in from the field, determine the cause, track the frequency, and make changes if necessary. Learn from your mistakes and those of others. End-Use Test – Prior to commercialization, evaluate the product under expected service conditions. Do not use the consumer as the test bed. Design – Invoke a cyclic design process that allows for product improvement before commercialization. Establish proper design parameters when sizing a part. For example, don’t size the wall of a component solely on flame resistance if stress is the driving force. Understand – Know how your finished product will be used in service. Is it a component of complete unit? Will it be attached to something that may restrict its movement? Will it be subjected to cyclic (fatigue) loads? Calculate – What are the stresses on your part? Consider residual as well as applied stresses. If you don’t have the tools to calculate stresses, try to measure them. Immersion of molded parts into appropriate solvents can reveal areas of high residual stress. End-use testing can address applied stress concerns. Environment – What kind of conditions will the part be exposed to? Don’t forget to consider thermal as well as chemical effects.
Index
A ABS 73, 121, 241 activation energy 219 adhesives 9, 89 aerospace 113, 179, 289 ageing 228 amorphous 25 anchoring strength 56 antioxidants 316, 206, 247 appliances 96, 314 arcing 313 Arrhenius 219, 229 Arrhenius equation 92 ASTM 173 automotive 9, 53, 95, 135, 179 B ballooning 25 blends 143 blistering 79 blowmolding 159 body fluid 89 bonding 21, 96, 323 fiber-matrix 20 boundaries 161 brittle 21, 143 C cables 318, 324, 326 cars 253 cavitation 25, 146, 193
chain mobility 95 chlorine 174 circuit boards 210 coatings 79 coefficient of thermal expansion 212 cohesive strength 89, 94 composite composites 105, 113, 135, 179, 253 Compton secondary electrons 202 construction 105 container 313 continuum 307 cooling 39, 212 copper 11, 212 corrosion 79, 105 corrosion resistance 253 cost 63 crack 17, 95, 143 front 131 propagation 73 tip 73, 146 cracking 161 cracks 112 crash 253 craze 261 radius 171 formation 267 crazes 169 crazing 25, 73, 146, 206, 259
338
creep 25, 58, 89, 90 crystallinity 167 crystallites 220 crystallization 149, 161 cup 333 cure 210 cyclic loading 310 D damping 58 decohesion 261 deformation 143, 146, 168 degradation 201 degree of cure 212 delamination 79, 105 density 262 design 63, 329 diaphragm 317 diffusion 113, 206 DIOP 318 disbonding 193 displacement 58 DOP 318 drawing 25 DSC 2, 114, 128, 167 dynamic load 56 E EDS 129 EDX 318 electronics 9 electrons 201 elongation 202, 220 energy 262 entanglement 96 environmental stress cracking 73 EPDM 219
Index
epoxy 89, 195 EPR 220 erosion 313, 314 Event Tree Analysis 289 extrusion 31, 159 F failure analysis 307 criterion 179 mode 290, 293 modes 300 progressive analysis 179 scenario 291 Failure Mode and Effects Analysis 289 failure tree analysis 299 fatigue 9, 17, 60, 165 Fault Tree Analysis 289 fibers 20, 179 fibrillation 45, 123 fibrils 169 filler 316 films 31, 48 finite element analysis 307 fitting 174 flaking 79 flaws 165 FMEA 331 force 55 fractography 121, 127 fracture brittle 21 surface 20 free radicals 220 free volume 114 frequency 165
Index
FTIR 128, 281, 318 G gamma 202 gas fuel system 292 gelation 210 Gerdner impact test 241 glass fiber 135 glass transition temperature 89 GPC 2 grease 325 growth 165 H HDPE 281 heat resistance 227 hip implant 310 HIPS 241 housing 297, 313, 332 hydroperoxides 247 hysteresis 56 I ICP 128 impact 159, 300 inelastic void 193 initiation 143 injection molding 227 insulation 313 interphase 114 ionization 201 IR 327 irradiation 247 ISO 289 J joints 95 L lamellae 29, 169
339
lamina 179 laminates 9, 105, 179, 209 Lamination Theory 212 lifetime 165, 228 ligaments 25 load 58 cyclic 17 dynamic 60 loads 233 local stress 73 lubricants 316 M marine 79 material cost 227 Mean Time between Failures 289 medical 201 metallurgy 127 microcracking 25, 105 microscope 127 microscopy 281 missile 9 mobility 96 moisture 89, 92 molding 135 molecular weight 174 morphology 158 motor 325 N necking 29 nitrocellulose 127 nucleation 152, 165 O OIT 203 optical emission spectroscopy 128 orientation 45
340
oven aging 219 overmolding 311 oxidative induction time 219 oxygen 201, 206, 247 P pacemakers 9 packages 9 packaging 201, 247 parison 159 PBT 235 PC 241 PE 326 petrochemical 289 pharmaceutical 313 photons 202 piping 173 plastic deformation 267 plasticization 89 plasticizer 318 plasticizers 316 plumbing 173 polyacetal 173 polyamides 95 polybutylene 173 polycarbonate 267, 311 polyester 219 polyetherimide 267 polyethylene 25, 31, 48, 143, 219, 259, 281 polyimide 10 polyketone 39 polypropylene 45, 135, 143, 149, 165, 201, 206, 219, 316 polysulfone 89 polyurethane 311
Index
PPS 314 Preliminary Hazard Analysis 289 prepregs 210 prestress 55 processing 39 pultrusion 105 PVC 313, 318 pyrolysis 314 R radiation treatment 247 radicals 201, 247 recycling 135 relay 313 residual stress 79, 210, 212, 311 risk 289 analysis 289 prioritization number 294 priority number 290 rubber 333 rubber particles 73, 193 S SAXD 167 SAXS 32, 34 screw 55 SEM 11, 73, 112, 123, 128, 159, 161, 168, 195, 219, 282, 316 semicrystalline 25 serration 123 severity 293 shear 146 shifting factor 229 shrinkage 79, 105 siding 244 size
Index
critical 165 sizing 330 snap 297 softening 167 solvents 73 spherulites 161 sports equipment 179 steel 25 sterilization 201, 247 stiffness 105, 179 strain 144 strain energy 122 strapping 45 stress 174, 244, 330 field 259 intensity 170 striations 123, 132 submicrocracks 165 surface embrittlement 174 surfactants 73 swelling 89 switch 313, 314 T talc 317 tape 45, 324 TD 323 TEM 73, 281 temperature 31, 89, 159, 219 tensile stress 187 tensioning 45 TGA 129 thermal expansion coefficient 187 thermal gradient 311
341
thermoplastics 233 tie molecule 220 tissue rejection 89 tools 95 toughening 143 toughness 122 traces 11 triphenylphosphate 325 U UV 331 V viscoelastic deformation 58 vitrification 210 voids 154 volume 96 von Mises stress 267 W wall thickness 159 warping 310 water 113 WAXD 167 weldability 45 welding 149 whitening 206 Y yield stress 25 yielding 146