No title

The current issue and full text archive of this journal is available at http://www.emerald-library.com IJCST 12,1 12 R...

Author: Emerald

288 downloads 923 Views 10MB 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

DOWNLOAD PDF

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,1

12 Received March 1999 Accepted July 1999

A preliminary study of the low-load lateral impact compression of fabrics P.M. Taylor and D.M. Pollet

Department of Mechanical, Materials and Manufacturing Engineering, The University of Newcastle, Newcastle upon Tyne, UK Keywords Fabric, Compression, Garments, Automation Abstract A swinging pendulum is arranged to impact and bounce from a fabric sample. The energy losses during impact compression are calculated from the changes in potential energy of the pendulum after allowing for other losses. It is found that the compression energy loss, calculated from the static characteristics, gives only a rough guide to the impact compression losses, so other factors must be considered. Further experimental work is proposed to determine more precisely the contribution of the fabric properties to these differences in the characteristics.

1. Introduction The compressibility of fabric is one of the basic mechanical properties, closely related to fabric handle in terms of softness and fullness (Elder et al., 1984) and of major importance in garment automation where it can be related to friction (Ajayi, 1988). Any surface treatment of the fabric such as singeing or pressing generally applied to enhance the hand will have a significant influence on the compressibility. Therefore, compression measurements can be used to assess the surface related properties such as dry abrasion of fabrics (Ukponmwan, 1994) or resilience and wear in carpets (Dunlop and Sun, 1989). Most of the compressibility tests to date have been carried out under static conditions which means that the compressive load is slowly increased up to a maximum pressure while at the same time the thickness reduction is registered. However, in several garment processes the fabrics are compressed dynamically, for example the fabric feeding mechanisms in sewing machines, gripping of single or multiple plies of fabric and certain fabric ply separation devices. This paper proposes a new technique to measure low force impact compression on fabrics and compares the obtained results with those using the static compression characteristics. In this way, we can see if the static characteristics can be used to predict dynamic behaviour.

International Journal of Clothing Science and Technology, Vol. 12 No. 1, 2000, pp. 12-25. # MCB University Press, 0955-6222

2. Background 2.1 Pressure-thickness curve Compression is defined as the decrease in intrinsic thickness when applying a lateral pressure and displays a typical curve as in Figure 1. The intrinsic thickness or maximum fabric thickness is determined as the thickness of the fabric subjected to a barely perceptible pressure which is generally about 1 percent of the maximum applied pressure. Furthermore, from Figure 1, a The authors wish to thank the EPSRC for a studentship award to support this work.

A preliminary study

Lateral pressure (cN/cm2)

25 (3)

20 15 compression

13

10 (2)

5 (1)

0

0

Figure 1. Theoretical compression-release curve

release

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Thickness reduction (mm)

hysteresis effect is obvious during the compression-release cycle resulting in a lower volume during the release, which will reduce continually during the first few successive compression cycles. In order to compare compression curves from different fabrics Kawabata, (1980) introduced four parameters as part of the KES-F (i.e. Kawabata Evaluation System for Fabrics). The four parameters express the work of compression, WC (WC' is the area under the release curve); the linearity, LC; the resilience of the fabric, RC and the relative compressibility, EMC, as follows: ZTm PdT

WC T0

LC WC=0:5Pm T0 ÿ Tm RC WC 0 =WC

1

EMC 1 ÿ Tm =T0 with T0 the thickness at minimum pressure of 0.5 gf/cm2 and Tm the thickness at maximum pressure (Pm) of 50 gf/cm2. In general, a woven fabric can be approximated by a three-layer structure consisting of the inner core and two outer layers at each face as sketched in Figure 2 (de Jong et al., 1986). The core is a dense assembly of fibres/yarns with air spaces in between, whereas the outer layers consist mainly of air with some projecting fibres (2 percent of total fabric mass per unit area for wool (de Jong et al., 1986)). In the process of fabric compression, three stages can now be distinguished depending on the pressure (see Figure 1) (Matsudaira and Qin, 1995; Postle, 1971). First, at low pressure the compression plate comes into contact with the hairs and/or protruding fibres from the outer face of the fabric and the compression force varies almost linearly with the thickness and so this first region of the compression characteristic is presumed to be elastic. Increasing the pressure overcomes the interyarn and interfibre friction, any buckles are flattened and fibre slippage takes place reducing the air spaces

IJCST 12,1

14

between the fibres. The fabric thickness decreases non-linearly with increasing pressure in this second region. Increasing the pressure even further compresses the fibres laterally hence the pressure rises greatly for a small reduction in thickness. This third region can be considered as the initial elastic region of the fibre bulk. For filament spun fabrics only two stages can be considered because of the absence of any protruding surface fibres. Recently, Hu and Newton (1997) proposed a five-layer structure by subdividing the outer layers of de Jong's model (1986) into two separate layers. The first layer in their model still containing hairy fibres now also includes crowns above the average geometrical thickness (i.e. calculated from the crimp height and the thread diameter). The second layer is ``another'' compressible layer forming the firm structure of the fabric whereas the third layer is the incompressible layer as in de Jong's model. Compression has been the subject of many studies (Bogaty et al., 1953; Dunlop, 1974; Dunlop and Sun, 1989; Lee and Carnaby, 1992; Matsudaira, 1995; van Wyk, 1946 a,b) though the most versatile approach is still dominated by the van Wyk equation which relates the pressure increase to the inverse cube of the volume. The equation further includes the Young's modulus, the mass of the fibres, and the bulk density at low pressure. Thus the compressibility of a fabric is independent of the fibre diameter or the elasticity and recently Dupuis et al. (1995) confirmed that compression of fabrics is more dependent on the structure then on the material of the yarns. However in spite of its usefulness, van Wyk's model (1946 a,b) has a limitation in that it includes also a constant (i.e. K-constant) of which the real physical meaning is undetermined. 3. Analysis of the energy losses Let be the angle of inclination of pendulum in Figure 3 with mass, m, at a distance lm from the centre of gravity to the pivoting point. The potential energy, Wp, with reference to point, p, the lowest point of the swing, can be written as follows: P

v

Figure 2. A three-layer model of woven fabric under lateral compression proposed by de Jong et al. (1986)

v’

P

A preliminary study

lm θ

l

mg

15 u0

p

Figure 3. Force diagram of a simple pendulum

T

Wp mglm 1 ÿ cos 2mglm sin2 2

2

After impact, the pendulum arm rebounds only to an angle l since part of the potential energy is lost during the compression of the fabric, W, and part is lost in the natural damping of the pendulum, Z. Hence, the difference in potential energy between the first two impacts can be written as: 2 2 1 W Z 2mglm sin ÿ sin 3 2 2 Since is measured indirectly from the distance, u0 , between the impact table and the pendulum, equation (3) becomes: u0 2 u1 2 W Z 2mglm ÿ 2l 2l 4 mglm ÿ 2 2 2 u0 ÿ u1 2l So, the amplitude of the pendulum arm gradually decays and the successive energy differences for the corresponding fabric compressions can each be calculated according to equation (4). However, the pendulum loss, Z, will need to be subtracted from each impact. 4. Experimental procedure The impact compression is measured on a pendulum, which is based on a design originating from Henning (1934), known as the ``Pendultex'', and used by van Wyk (1946a) in his study on the compression of wool fibres in bulk. The principle is based on measuring the compressional resistance by the consequent damping of the pendulum.

IJCST 12,1

16

Figure 4. Impact compression tester

Fabric compression is studied here instead of fibre compression so the pendulum in this study and pictured in Figure 4 is simpler in construction than the original ``Pendultex'', which comprised a lever and piston mechanism to compress the enclosed fibres in a compartment. The instrument (Pollet, 1998) here comprises a pendulum with mirror finished pressure foot (10 cm2), which impacts on the fabric sample attached by its topmost edge to a rigid vertical plate (i.e. the ``impact table''). The pendulum arm is suspended on precision bearings and housed at the top of a large aluminium framework. A solenoid mechanism connected to the suspension rod releases the pendulum arm automatically from a pre-determined variable angle. In addition, a 1,550g mass on the pendulum arm giving an overall weight of 4.5kg can be balanced to give variable impact forces on the fabric sample. The impact table is adjustable relative to the pendulum as sketched in Figure 5 and houses a laser sensor (LD1605-4 by "1), which measures the varying impact displacement by pointing the laser beam at a metal indicator placed at the side of the pressure foot. Several small holes were drilled in the impact table at the fabric area to investigate the effects of possible air damping. The full process is co-ordinated by a PC via a PC-30AT card (12 bit sampling), which triggers the measurement

A preliminary study

Pendulum

Laser Sensor

17

Fabric Clamp

Impact Table

Pressure Foot

Figure 5. Detailed sketch of the impact compression tester

Fabric Specimen

by pulling in the plunger in the solenoid and then logging the captured displacement data from the laser sensor. To perform a compression test, the impact table is first adjusted so that the pendulum arm, hanging perfectly vertical at rest, barely touches the table. At the same time, the laser sensor is set to measure a minimum distance ( zero). Second, an oblong fabric strip (5.0 10cm) of which further details are listed in Table I is clamped to the impact table at the top edge of the fabric. The strip is allowed to hang down freely and then lightly clamped at the bottom to prevent the fabric from flapping. Care has to be taken that the fabric is not stretched when attached Fabric Fibre Code Content C1 W1 KC1 KC11 KA11

Cotton Wool Cotton Cotton Acrylic

Fabric Structure

Sett (thd/cm) P T

Twill Twill Fleece knit Double knit Double knit

19 12 16 14 15

Notes: P: (warp/wale); T: (weft/course)

17 13 11 15 10

Mass (g/m2)

Thickness at 0.5 cN/cm2 (mm)

517 352 257 183 264

1.03 3.09 1.43 1.03 1.23

Bending length (mm) ASTM 1388 P T 121.7 39.7 37.5 30.1 29.7

87.2 40.5 24.0 16.8 23.5

Table I. Fabric descriptions

IJCST 12,1

18

to the table as this might alter the damping properties of the material. The pendulum arm is locked at an angle (i.e. 28 from the vertical in this study). Data logging is started and simultaneously the solenoid is retracted to release the pendulum, which then swings downwards to impact against the fabric sample. 5. Results Before tests can be carried out, it is necessary to measure the natural losses of the equipment, which need subtracting from the actual measurement. The impact table is therefore removed to allow free swinging of the pendulum arm and the laser sensor is attached near the top of the rig pointing towards the arm. The pendulum is then released from the same angle as in the fabric tests (i.e. 28) and damped solely by bearing friction and air resistance. A nearly linear pendulum damping is displayed in Figure 6, except at extremely small angles. Considering now the free damping of the pendulum, the successive energy loss, Z, for each inclination angle can be derived by setting W in equation (3) equal to zero. Yet, because an angle of zero degrees can never be obtained since the material thickness offsets the pendulum arm, the loss of the pendulum has been calculated according to the angular distance. For example, if the pendulum compresses a material to a thickness of 2.67mm then the arm is still at an angle of 0.28. So, if the pendulum is at an angle, = 0.38, and rebounds to an angle, 1 = 0.258, then a total angle of 0.158 has been covered which from the linear approximation in Figure 7 corresponds to an energy loss of 4 10±06 Nm. Note that the energy loss for a half swing still needs to be divided by two as the pendulum arm moves forwards and backwards during that swing. The expected damping output from the fabric compression tests is displayed in Figure 8. The pendulum arm starts from a rest position at a maximum distance (e.g. 13.0mm), hits the fabric sample and consequently compresses the Pendulum displacement (mm)

0.6 0.4 0.2 0 –0.2 –0.4 –0.6 –0.8

Figure 6. Free oscillation of the impact pendulum

0

0.5

1

1.5 2 Time (ms)

2.5

3 x 104

Note: Release angle: 28, sampling frequency: 250 Hz, negative displacement for pendulum coming towards the laser sensor

Pendulum energy loss Z (Nm)

8

A preliminary study

x 10–5 experimental

7 6 full swing

5 4 3

19

half swing

2 1 0

0

0.1 0.3 0.4 0.5 0.6 Inclination angle of pendulum (degrees)

Figure 7. Energy loss of the pendulum in free oscillation

Pendulum displacement (mm)

14 12

potential energy changes of the pendulum

10 8 6 variation in fabric compression

4 2 0

0

1

2

3 Time (s)

4

5

6

fabric to a certain thickness. The fabric's stored compressive energy is then released back to the pendulum giving it kinetic energy, which causes the pendulum arm to bounce back to a certain distance. This distance is, however, less than the starting distance because of energy losses in the cycle. The above process of impact, compression and rebound continues until the initial potential energy of the pendulum is totally consumed by the energy losses. Thus, each compression graph is characterised by two envelopes (see Figure 8). The envelope of maximum distances gives the differences in potential energy between the successive impacts whereas the minima correspond to the compressed thickness of the fabric. Figures 9-11 display the raw impact response (i.e. pendulum loss included) for fabric KC11, C1 and W1 respectively as examples of a knitted structure and woven structures with variable compressibility. Similar graphs are found for fabrics KC1 and KA11. Owing to a limited sensing range of 4mm, the first impacts of each test are clipped and therefore omitted from the figures and calculations. Further, note the thickness offset for the respective fabrics. This thickness does not correspond to the zero pressure thickness because the

Figure 8. An illustration of the dynamic pendulum compression of fabric

20

Pendulum displacement (mm)

IJCST 12,1

3 2.5 2 1.5 1 0.5 1.5

2

2.5

4

3

2

1

3 3.5 Time (s)

4

4.5

Pendulum displacement (mm)

Figure 9. Note: Release angle: 28, sampling frequency: 1,000Hz, ``non-compressed' fabric Impact response of KC11 thickness: 1.04mm

Figure 10. Impact response of C1

3 2.5 2 1.5 1 0.5

1

1.5

2

4

3

2

1

2.5 3 3.5 Time (s)

4

4.5

5

Note: Release angle: 28, sampling frequency: 1,000Hz, ``non-compressed'' fabric thickness: 1.15mm

Pendulum displacement (mm)

4 3.8 3.6 3.4 3.2 3 2.8 1

2.4 2.2

Figure 11. Impact response of W1

3 2

2.6

2

2.5

3

3.5 Time (s)

4

4.5

5

Note: Release angle: 28, sampling frequency: 1,000Hz, ``non-compressed'' fabric thickness: 2.85mm

pressure foot and pendulum arm rest against the fabric surface at a slight angle to the vertical. For the static compression, this thickness is equivalent to compression forces in the range of 0.2-0.4 cN/cm2. 6. Analysis of the results The dynamic compression energy lost during the successive impacts is calculated for each fabric according to Equation (4) and displayed for example for KC11 in Figure 12. Next, the static hysteresis is calculated from a representative static deflection characteristic as displayed for example for KC11 in Figure 13 (Pollet, 1998), and compared in Figure 14 to the dynamic energy lost when compressing the fabrics to an identical thickness. When a sample is compressed to a minimum thickness, say 909m, the compression characteristic is assumed to follow the curve containing the line segment A shown in Figure 13. In general, this will give an overestimate of the static hysteresis loss. A similar procedure is followed for the other fabrics, which are given in Figures 15-18. The ratios between the dynamic and the static hysteresis together with some KES-F compression parameters are summarised in Table II. From Figures 14-18, it is seen that the impact losses for all fabrics except W1 are larger than the static hysteresis at high impact and decrease towards the static value when the impacts reduce. On the one hand, the compression losses are overestimated because of the assumptions made in the shape of the hysteresis curve. If this were corrected then the static curves of Figures 14-18 would lower. This could lead to the expected result that the two characteristics should become closer together as the impact velocities tend to zero. On the other hand, the impact compression losses in the fabric may also be overestimated because of possible vibratory impact losses within the pendulum itself. This is likely to be manifested most strongly at the initial high impact. A combination of these two effects should bring the corresponding static and 4

A preliminary study

21

x10–5 1

Energy loss (Nm)

3.5 3 2.5 2 1.5

2

1 0.5

3 4

0 0.85 0.9 0.95 1 Thickness of sample at max impact compression (mm)

Note: The numbers in the above figure correspond to the numbeed impacts in Figure 9

Figure 12. Energy loss during impact for KC11

22 Figure 13. Static compression of KC11

Pressure (cN/cm2)

IJCST 12,1

10 9 8 7 6 5 4 3 2 1 0 700

9

A

800 900 1000 1100 1200 1300 1400 Thickness of the sample (µm)

x10–5

Figure 14. Comparison between the static and dynamic energy losses for KC11

Energy losses (Nm)

8 7 6 5 4 3 2 1

0 0.85 0.9 0.95 1 Thickness at maximum compression (mm)

9 Energy losses (Nm)

8

Figure 15. Comparison between the static and dynamic energy losses for KA11

static impact

7

x10–5 static impact

6 5 4 3 2 1 0 0.9 0.95 1 1.05 1.1 1.15 Thickness at maximum compression (mm)

impact curves closer together. However, note that the static characteristics do not include viscous effects in the fabrics, which will be highest during the initial impact.

9

x10–5

Energy losses (Nm)

8

static impact

7 6 5

23

4 3 2 1 0 1.1 1.15 1.2 1.25 1.3 1.35 1.4 Thickness at maximum compression (mm)

9 Energy losses (Nm)

Figure 16. Comparison between the static and dynamic energy losses for KC1

x10–5

8

static impact

7 6 5 4 3 2 1

0 0.98 1 1.02 1.04 1.06 1.08 1.1 1.12 Thickness at maximum compression (mm)

9 8 Energy losses (Nm)

A preliminary study

7

Figure 17. Comparison between the static and dynamic energy losses for C1

x10–5 static impact

6 5 4 3 2 1 0 2.45 2.5 2.55 2.6 2.65 2.7 2.75 2.8 2.85 Thickness at maximum compression (mm)

Fabric W1 shows a quasi-linear relationship between the higher static and the lower dynamic hysteresis. This linearity is also reflected in the high linearity parameter (see LC coefficient in Table II derived from the static compressions

Figure 18. Comparison between the static and dynamic energy losses for W1

IJCST 12,1

24 Table II. Ratio of the dynamic compression energy to the static energy for successive impacts

Dynamic compression energy/static energy Impact

C1

W1

KC1

KC11

KA11

1

6.96

0.79

1.70

1.92

1.35

2

2.49

0.55

1.09

1.09

0.80

3

1.29

0.39

0.95

0.62

0.53

4

0.57

0.71

0.38

0.34

Average

2.83

1.11

1.00

0.75

LC(±) RC (%) EMC (±)

0.144 56.6 0.408

0.57 0.471 59.9 0.372

0.337 54.2 0.408

0.266 50.2 0.406

0.290 57.5 0.431

(Pollet, 1998). Furthermore, note in Table II that the average ratio between the dynamic and static losses of all knitted samples has the same order of magnitude. 7. Conclusions A new experimental method has been established to measure the impact compression on fabric. A pendulum impacts the fabric and gradually reduces its potential energy during a number of swings, because of energy losses during the impaction. Clearly, the static compression losses just give a rough guide to the impact compression losses and so other factors must be brought into the equation. First, the static characteristics should be repeated for each relevant maximum compression in order to get the true static hysteresis loss. Secondly, further instrumentation is to be added to determine if there are vibratory losses within the pendulum. A high bandwidth force sensor is to be added to the impact table so that the full impact characteristics and losses may be measured directly. With this work, it should then be possible to get more precise results and hence determine the contribution of viscous effects to the overall impact characteristics. References 1. Ajayi, J.O. (1988), ``Some studies of frictional properties of fabrics'', Doctoral Thesis, University of Strathclyde, Glasgow. 2. Bogaty, H., Hollies, N.R.S., Hintermaier, J.C. and Harris, M. (1953), ``The nature of a fabric surface: thickness-pressure relationships'', Textile Research Journal, Vol. 23 No. 2, pp. 10814. 3. de Jong, S., Snaith, J.W. and Michie, N.A. (1986), ``A mechanical model for the lateral compression of woven fabrics'', Textile Research Journal, Vol. 56 No. 12, pp. 759-67. 4. Dunlop, J.I. (1974), ``Characterizing the compression properties of fibre masses'', Journal of the Textile Institute, Vol. 65, pp. 532-6.

5. Dunlop, J.I. and Sun, J. (1989), ``The dynamic mechanical response of carpets'', Journal of the Textile Institute, Vol. 80 No. 4, pp. 569-78. 6. Dupuis, D., Popov, G., and Viallier, P. (1995), ``Compression of greystate fabrics as a function of yarn structure'', Textile Research Journal, Vol. 65 No. 6, pp. 309-16. 7. Elder, H.M., Fisher, S., Armstrong, K. and Hutchison, G. (1984), ``Fabric softness, handle, and compression'', Journal of the Textile Institute, Vol. 75 No. 1, pp. 37-46. 8. Henning, H.J. (1934), ``EÂber die Messung der Bauschelastizitt von Textilien'', Melliand Textilber, Vol. 15 No. 12, pp. 537-40 (in German). 9. Hu, J. and Newton, A. (1997), ``Low-load lateral-compression behaviour of woven fabrics'', Journal of the Textile Institute, Vol. 88, Part 1 No. 3, pp. 242-54. 10. Kawabata, S. (1980), The Standardization and Analysis of Hand Evaluation, 2nd ed., Textile Machinery Society of Japan, Osaka. 11. Lee, D.H. and Carnaby, G.A. (1992), ``Compressional energy of the random fibre assembly, part I: theory'', Textile Research Journal, Vol. 62 No. 4, pp. 185-91. 12. Matsudaira, M. and Qin, H. (1995), ``Features and mechanical parameters of a fabric's compressional property'', Journal of the Textile Institute, Vol. 86 No. 3, pp. 476-86. 13. Pollet, D.M. (1998), ``A study of low force fabric characteristics and vibrational behaviour for automated garment handling'', Doctoral Thesis, University of Hull, Hull. 14. Postle, R. (1971), ``The thickness and bulk density of plain-knitted fabrics'', Journal of the Textile Institute, Vol. 62 No. 4, pp. 219-31. 15. Ukponmwan, J.O. (1994), ``Compressibility analysis of dry abraded woven fabrics'', Textile Research Journal, Vol. 64 No. 12, pp. 756-60. 16. van Wyk, C.M. (1946a), ``A study of the compressibility of wool, with special reference to South African Merino wool'', Onderstepoort Journal of Veterinary Science and Animal Industry, Vol. 21 No. 1, pp. 99-226. 17. van Wyk, C.M. (1946b), ``Note on the compressibility of wool'', Journal of the Textile Institute, Vol. 37, pp. T285-T292.

A preliminary study

25

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,1

26 Received April 1998 Revised July 1999 Accepted July 1999

Development of threedimensional apparel CAD system Part 1: flat garment pattern drafting system Tae J. Kang and Sung Min Kim

Department of Fiber & Polymer Science, Seoul National University, Seoul, South Korea Keywords Apparel, CAD, 3D, Patterns Abstract A comprehensive apparel CAD system was developed to perform automatic garment pattern drafting and the prediction of the final drape shape of designed garment putting on the human body. Three dimensional apparel CAD system starts with a flat garment pattern drafting system. A computerized pattern design script language has been created based on the traditional patterner's principles to develop an automatic draft system of performing basic garment pattern drafting as well as grading rule generation. A pattern modification system was also developed considering functions required in apparel CAD such as auxiliary pattern generation, seam line creation, and dart manipulation to generate engineering patterns which can be used in the three dimensional garment shape prediction system presented later in part II of this paper.

International Journal of Clothing Science and Technology, Vol. 12 No. 1, 2000, pp. 26-38. # MCB University Press, 0955-6222

1. Introduction Automation has been a key factor of success in modern garment manufacturing industries. The application of computer and automated system is widely practiced in every aspect of industrial manufacturing very rapidly. However, for textile and garment industries, the automation seems to be less extent than other industrial segments[1-3]. One of the major reasons for this is despite the vast amount of investment there had been relatively little research on the fundamentals required for the full automation of textile industries[4,8,9]. Unlike other industries, textile and garment manufacturing industries often require more specialized manual intervention, therefore even if an automated system has been introduced, there is still a need for some experts to operate and supervise the system and therefore the goal of the automation ± reduction of labor cost ± cannot be easily achieved. For example, most of the recent studies on the computerized pattern drafting have been focused on the development of the easy pattern modification method for pattern experts to express their skill more efficiently and not on the development of pattern drafting principles[5,6]. In part I of this study, we focus on the development of the automatic basic pattern drafting and grading system by formulating the patterner's drafting principles to make it easy to use even for those with little knowledge of pattern drafting nor grading. Also we develop an automatic engineering pattern generation system to make use of drafted patterns in the garment shape prediction system discussed in part II of this study.

2. Computerized pattern drafting algorithm 2.1. Formulation of the pattern drafting method Apparel CAD system is somewhat different from other CADs used in mechanical or civil engineering. That is, in apparel CAD system, resulting flat garment patterns should follow some explicit rules to form a desired shape of garment while other CADs have less constraints on the objects to be designed[8,9]. Such rules are generally accumulated and monopolized by the skilled patterners and this made it very difficult for an unskilled person to do the job. Many studies on the garment pattern generation so far focused mainly on the development of efficient tools for experts who can draw patterns by themselves and not on the pattern drafting method itself. But it is necessary for the automation of pattern drafting to analyze and formulate those rules. Because such rules are very complex and subjective to the patterner who made them, it hardly defines an optimum pattern for a specific garment. Figure 1 shows an example of drafting rule for bodice patterns. Therefore, it is easier for the production of various sized garment patterns to construct and to utilize a database by collecting and formulating various pattern drafting rules rather than to define optimum rule for each pattern. Recent studies on the computerized pattern drafting mainly used the method which made a program to draft a pattern to apply to the apparel CAD system. However, such a method requires repeated reprogramming each time a new pattern is introduced and has some difficulties in modifying the programs to meet individual specifications of garment manufacturing firms[7-9].

Flat garment pattern drafting system 27

Figure 1. Example of a pattern drafting rule for bodice pattern

IJCST 12,1

28

In this study, we formulated the pattern drafting rules using a script language, a kind of computer programming language which can be written even by simple text editors, and generated patterns by compiling corresponding scripts to make the addition or modification of patterns. To store the pattern geometry defined by the script, a special data structure has been developed. The primary objects defined are the coordinates of the pattern vertices and the combinations of these vertices define the secondary objects, the linear and curvilinear segments of the pattern. A complete pattern includes a series of continuous segments called outline, auxiliary lines for design, and fashion details such as darts and notches. Each element is kept in linked list structure and managed by the object oriented programming (OOP) method which enhances the versatility of the governing CAD system. 2.2. Generation of script language for pattern definition To define the geometrical and topological shape of garment patterns, we developed a script language compiler. This scripts consisted of about 20 commands and variables which are used together with a size specification table to generate garment patterns. Size specification is a series of anthropometric data specific to the individual garment manufacturer which determines the dimension of resulting patterns. The scripts can use either predetermined or customized size specifications and the commands can be grouped to three categories and some examples of the syntax are described as follows. 2.2.1. Constant variable defining group . Length [n,SizeName]: picks the value of specified size from size specification data and assign that to the n-th constant (e.g. Length 1,NeckGirth: neck girth is assigned to the 1st constant denoted by ``C1''). . Constant [n,Expression]: assigns the value of expression to the n-th constant (e.g. Constant 2,C1/3+5: evaluated value of the expression ``C2/ 3+5'' is assigned to the second constant@. . Distance [n,p1,p2]: assigns the distance between points p1 and p2 to the n-th constant. 2.2.2. Geometry defining group . Point [n, x, y]: locate n-th point at specified coordinate. . Move [n, p, dx, dy]: locate n-th point by moving p-th point by specified distance. . Divide [n, p1, p2, a, b]: locate n-th point at the position dividing the distance between two specified points by a/b. . Perpendicular [n, p1, p2, d, l]: locate n-th point perpendicular to the line connecting points p1 and p2 at the distance l from a point on that line away from p1 by distance d.

2.2.3. Topology defining group . Loop [n, p1, p2,... pm]: make the n-th closed loop structure by connecting m specified points. . O_Segment [n, p2, p2, . . ., pm]: make the n-th outline segment in the pattern by connecting m specified points. . A_Segment [n, p1, p2, . . ., pm]: make the n-th auxiliary segment. . Dart [n, s1, s2]: define a dart with two consecutive segments s1 and s2 on a closed loop.

Flat garment pattern drafting system 29

Figure 2 shows the example of a simple script and the resulting pattern. 2.3. Automatic grading rule generation In garment production, especially for the mass production, the grading process is very important. Grading is the adjustment of the standard sized patterns to

Figure 2. Example of a script and the resulting pattern

IJCST 12,1

30

accommodate the various size specifications. The validity of graded patterns is highly dependent on the level of the patterner's skill and even for a skilled craftsman, it may still be very tedious work. The most widely used grading method is the shift amount registration method[6]. The shift amount is defined by the direction and distance assigned to each vertex on the pattern to accommodate the size change. Once the shift amount is determined, various sizes of well fit patterns can easily be generated. However, if performed manually, the calculation is inevitably subjective to the worker and takes a considerable amount of trial and error. Since the newly developed pattern geometry defined by the script is not based on one absolute size but on the selectable anthropometric data, the generation of graded pattern can be relatively easy and accurate. That is, grading rules ± the shift amount of each vertex ± can automatically be determined by calculating the geometrical difference of the drafted patterns based on the series of different sizes. For more compact visualization, graded patterns can be aligned around a fixed point as in Figure 3 showing the examples of the automatic grading. 3. Development of flat pattern modification system The patterns generated automatically by the system mentioned above need some modifications to be used in the real garment manufacture. Such patterns are called the ``basic patterns'' and they just satisfy the minimum requirements needed to

Figure 3. Examples of the automatically graded patterns

form the desired fundamental shape of the garment to be produced[9]. A basic skirt pattern represents only the commonest form and to apply various designs and thus, the pattern needs to be modified. For the industrial apparel CAD system, the modification including the engineering pattern generation should be performed with ease especially for application of the mass production and quick response.

Flat garment pattern drafting system

3.1. Definition of pattern modification It is necessary for the basic patterns to be modified because they have only the basic geometry for a specific type of garment. This can be done with so-called general-purpose CAD systems, but for some garment patterns, a specific modification procedure is required. The modification functions that should be included in an apparel CAD system are as follows.

31

3.1.1. Functions in common with other general purpose CAD systems The examples of such functions are relocation of the vertices and lines, magnification or rotation of the patterns and they can easily be implemented in the system. 3.1.2. Functions for the management of two-dimensional patterns A flat garment pattern generated by script always has a closed outline loop and this topological condition should be maintained during any modifications such as separation, mergence, extension of pattern and especially in the derivation of auxiliary patterns discussed later. Because the pattern topology is stored as a continuous loop of segments, most of those modifications can be achieved by division, insertion, deletion, or substitution of segments followed by the reorganization of the loop, Figure 4 is an example of flat pattern management functions.

Figure 4. Example of a function for the management of flat patterns and pattern separation

IJCST 12,1

32

3.1.3. Apparel CAD specific functions Because there are many special features in a garment pattern to be considered., it is still difficult for a system to be an appropriate apparel design tool even if all such functions described above are implemented. The special features to be considered are darts, notches, grain lines and so on. For example, a dart is used for the transformation of a two dimensional pattern to a three dimensional garment and it must be handled by specially designed functions to be relocated, distributed, or merged. To do this, we examined the relevant methods of a skilled patterner and made them available on the computerized system. These functions are more difficult to make than other functions because there arise not only geometrical but also topological changes in the pattern. For example, to migrate an existing dart over another point as shown in Figure 5, first a new dart of zero width is created on the target point and the segments between two darts are rotated around the pivot point to close and delete the original dart. 3.2. Development of pattern derivation algorithm One of the most frequently used functions in an apparel CAD system is the auxiliary pattern derivation which creates subsidiary patterns by dividing a basic pattern. For a bodice pattern as an example, it is sometimes divided into two parts ± shoulder and back. But, once a pattern is divided, they became two separate entities and thus cannot be easily edited when it is necessary to change a part common to both of them. Thus we have introduced a patternfamily concept. A pattern family has a mother and her daughters and the daughter patterns are associated with their mother by a specially designed data structure.

Figure 5. Example of an apparel CAD specific pattern modification function ± dart migration

By this concept, when a mother pattern needs dividing, it is rather not divided but partially regenerated into subsidiary daughter patterns. Because the daughter patterns inherit the geometrical data of their mother, all of them can exactly keep pace with the changes made on their mother pattern. 3.2.1. The virtual segment concept For the generation of a daughter pattern it is necessary to store and express its pattern geometry efficiently. That is, only the segment numbers rather than the whole geometry of the mother's are stored in a daughter pattern and so the changes made on the mother can affect all her daughters simultaneously. But as shown in Figure 6, sometimes a daughter pattern may partially use the mother's segment. In this case the complete modification cannot be achieved by the mere storage of the mother segment numbers. To solve this problem, we introduced ``the virtual segment concept''. A virtual segment is defined by two points on a mother segment. Their numbers are stored in virtual segment and thus a daughter pattern composed of these virtual segments can maintain the same geometry as their mother's. But there still remains a problem that there are some points that should keep special relations with the mother segment. For example, some points must be at the midpoint of a mother segment, or at the quarter. This can be solved by defining ``the relative point concept''. A relative point has only the data of relationship with their base segment and the exact coordinate is calculated each time on necessity. Figure 7 shows the example of virtual segments using one relative point.

Flat garment pattern drafting system 33

Figure 6. Example of a daughter pattern that uses some part of mother segments

IJCST 12,1

34 Figure 7. Example of virtual segments using relative points

3.2.2. Auxiliary pattern derivation An interactive method has been developed to derive daughter patterns from a mother pattern by selecting the region of interest where at least one closed configuration of virtual segments exists. For a garment pattern to be valid, there should be only one closed outline loop. For this reason, the most suitable closed loop must be selected if there are many available closed loops in the selected region. Figure 8 shows the example of a selected region on a mother pattern with many possible closed loop configurations. As shown above, there are some virtual segment irrelevant to the valid closed loop and thus they should be removed. In this study, a most suitable outline loop is detected by the multiple tree data structure having all the vertex in the selected region as its nodes. The multiply branched tree structure algorithm for the determination of the most valid closed loop can briefly be explained as follows. First, the opened segments in the daughter pattern at least one of its end not attached to others

Figure 8. Example of a region with many possible closed loop configurations

are eliminated. Second, the start and end points of each segment are denoted as ``S'' and ``E'', then a possible selection of virtual segments can be rearranged as shown in Figure 8. To construct a tree structure, the terminal point of the first segment is stored as the first node of the tree. Then, find the segments whose start point is the same as the first node and store them in the second node and fill consequent nodes with the respective terminal points of each segment. Repeating this procedure until the start point of the first segment is reached, a multiply branched tree structure can be obtained as shown in Figure 9. Once the tree structure is obtained, the most valid closed loop can be determined by choosing the longest branch in the tree and a daughter pattern can be defined by this sequence of virtual segments. Because daughter patterns derived by this method are always synchronized with their mother pattern, the modification can be easily applied to both of them as shown in Figure 10.

Flat garment pattern drafting system 35

3.3. Engineering pattern generation The basic and derived patterns must be converted into engineering patterns for the industrial application[8,9]. For example, a symmetrical pattern is usually drawn by half and two patterns of a same shape are drawn once and should be duplicated in desired way. In this study, we have developed some functions to perform such tasks. 3.3.1. Automatic seam line generation One of the most important parts of deriving the engineering pattern generation is the cut line generation. The cut line is the line added along the contour of an

Figure 9. A multiply branched tree structure obtained for a daughter pattern

IJCST 12,1

36

Figure 10. Example of the synchronized modification of mother and daughter patterns

original pattern considering the extra amount needed in the sewing process. This is not a simple procedure because the cutting line is not a mere magnification of the outline of original pattern. It may have different distance from each segment of the pattern and various shapes at the corners. Therefore this process also needs skilled labor and thus it should be supported in an apparel CAD system. In this study, we developed a system that can draw the cut line with different seam amount at each original segment. Once the seam amounts are assigned to each outline segments, the cut lines are generated and the shape of the lines are synchronized with the shape change of original pattern. Figure 11 is the examples of some patterns with cut lines of variable seam amount on each segment. Since the corner shapes of the cut line are also important as the cut line itself in engineering pattern generation, some functions to apply nine different corner shapes to the cut line have been developed. Figure 12 shows the various shape of the corners. Once engineered patterns are generated, they can be used in the automatic cutting process or garment shape prediction system discussed in Part II of this study. 5. Conclusions Some modifications and improvements have been made for the more efficient integration of CAD system into the industrial garment manufacturing process based on the conventional two-dimensional pattern drafting system. The newly developed system has focused on the development of computerized rules for the generation of patterns rather than the generation of individual pattern itself. With the script based pattern generation method developed in this study, even an inexperienced worker can do the job as a skilled patterner. Furthermore, this

Flat garment pattern drafting system 37

Figure 11. Examples of some cutting lines generated by the engineering pattern generator

Figure 12. Examples of cut line corner shapes

method can also generate the complex grading rules automatically and thus can shorten the development period of the garment mass production. The pattern modification system developed in this study has a capability to change the geometry and topology of the patterns and to generate engineering patterns

IJCST 12,1

38

automatically. The resulting patterns will be used for the garment shape prediction system which will be presented in the following report of Part II of this study. Because one of the most serious problems facing in the garment manufacturing industry is the heavy dependency of the highly skilled labor with the current manufacturing system, the development of more efficient CAD/CAM systems can be of great help in increasing productivity as well as lowering the production cost. References 1. Okabe, H., Imaoka, H., Tomiha, T. and Niyawa, H., ``Three dimensional apparel CAD system'', Computer Graphics, Vol. 26 No. 2, 1992, pp. 105-9. 2. Heisey, F., Brown, P. and Johnson, R.F., Three-dimensional pattern drafting: a theoretical framework'', Clothing Textile Res. Journal, Vol. 6 No. 3, 1988, pp. 1-9. 3. Hinds, B.K., McCartney, J., Hadden, C. and Diamond, J., ``3D CAD for garment design'', International Journal of Clothing Science and Technology, Vol. 4 No. 4, 1992, pp. 6-14. 4. Collier, B. and Collier, J., ``CAD/CAM in the textile and apparel industry'', Clothing and Textile Research Journal, Vol. 8 No. 3, 1990, pp. 7-12. 5. Stjepanovic, Z., ``Computer-aided processes in garment production: features of CAD/CAM hardware'', International Journal of Clothing Science and Technology, Vol. 7 No. 2, 1995, pp. 81-8. 6. Liu, Z. and Harlock, S.C., ``A computer-aided grading system for both basic block and adopted clothing patterns'', Textile Res. Journal, Vol. 65 No. 3, 1995, pp. 157-62. 7. Lee, S., Nam, Y. and Kim, J., ``A study of pattern making by computer'', Journal of the Korean Society of Clothing and Textiles, Vol. 9 No. 1, 1985, pp. 37-46. 8. Park, S., ``A study on the possibility of pattern design using CAD system for patternist'', Journal Korean Society of Clothing and Textiles, Vol. 21 No. 4, 1997, pp. 769-81. 9. Shin, S., ``A study on the mass-customization of apparel design by using computer-aided design'', Journal of the Korean Fiber Society, Vol. 33 No. 6, 1996, pp. 544-54.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Development of threedimensional apparel CAD system Part II: prediction of garment drape shape Tae J. Kang and Sung Min Kim

Prediction of garment drape shape 39 Received April 1998 Revised July 1999 Accepted July 1999

Department of Fiber & Polymer Science, Seoul National University, Seoul, South Korea Keywords Apparel, CAD, 3D, Garment Abstract A fast response three dimensional garment drape shape prediction system has been developed. A human body model generator has been established for the garment draping on it. For the mass production of different size of garments for the various sized body models, the cross-sectional value from the anthropometric data was used as the standard for the size accommodation to make a resizable human body model. To construct the cloth drape shape prediction system, the finite element analysis method has been utilized. The designed garment pieces were divided into fine quadrilateral elements using a specially coded mesh generating program, then some appropriate sewing conditions were assigned to transform two dimensional patterns into three dimensional shapes. The final drape shape of the garment was determined from the solutions of the contact condition with human body, deformations, and the weights of the elements constituting the garment pieces, as well as the surface texture of the cloth.

1. Introduction The application of computer and automated system in the manufacturing sectors is widely practiced in every aspect of industry. However for the textile and garment industries, the extent of automation seems to be less than in other industries. Due to the complexity and skilled-labor dependent nature of garment manufacturing processes, it had been regarded as very difficult to automatize the manufacturing process. In this study, we developed a threedimensional apparel CAD system that can assist the unskilled works more efficiently. In Part I we formulated the traditional pattern drafting rules and developed an automatic pattern drafting system based on the pattern defining script language. We also developed a pattern modification system that can adjust the geometry and topology of patterns and draft engineering patterns. In Part II, we develop a garment drape shape prediction system from the engineering patterns obtained from Part I. To make an apparel CAD system more practical, not only the two dimensional but also the three dimensional features should be considered[2,3,7]. Even though the flat patterns can be

International Journal of Clothing Science and Technology, Vol. 12 No. 1, 2000, pp. 39-49. # MCB University Press, 0955-6222

IJCST 12,1

40

effectively generated, the advantage may not be noticeable if the validation of those patterns by putting them on the human body are not realized at reasonable time and cost. For this reason, many recent studies on the apparel CAD system focused on the implementation of the three dimensional design ability and some of them seemed very successful[8,9]. In this study we develop a three dimensional garment drape shape prediction system using the finite element method as well as computer graphics to obtain and demonstrate the 3D drape shapes fast enough to meet one of the general needs of the CAD system as well as the reduction of development time. We also develop a human body model generator that can make various sized body models easily to be used in the garment drape shape prediction as well as in the validation of pattern grading process for the mass-production of industrial application. 2. Human body model generation Because one of the most time consuming and skilled-labor dependent processes in garment design is the formation of a garment preform by sewing each pattern, it may be a serious problem especially when quick response is on demand. We thought that if this process can be done on a CAD system, overall development cycle of garment can be noticeably shortened and the additional cost reduction can also be achieved. Thus we first developed a program that can generate human body models for the garment drape shape prediction system. 2.1 Acquisition of anthropometric data The three-dimensional anthropometric data can be obtained by two major methods, one of which is direct and the other is indirect. The direct method measures body using ruler or gauge while the indirect method is photo measurement or laser-scanner. In this study, for the convenience of the experiment, we adopted the direct measurement method using a sliding gauge as shown in Figure 1. To make a female torso body model, we obtained the serial cross-section images of the body using a sliding gauge and applied image analysis technique to extract the radii of the body sections from them. Figure 2 shows the image

Figure 1. Acquisition of anthropometric data using a sliding gauge

analysis step for the extraction of radial section data. Then a visualized human body model can be reconstructed in the cylindrical coordinate system using a series of radial data. Figure 3 shows an example of reconstructed body model. 2.2 Generation of resizable body model When a different sized body model is needed, the whole data set should be reconstructed from the measurement of each human body and this can be an obstacle in achieving one of our goals ± the quick response. Therefore we developed an algorithm that can generate a different sized body model from a standard one using some statistical data manipulations as described below. First we chose some base sections such as neck, bust, waist, and hip from the standard model to compare with the target sized model. Then we calculated the circumferential ratios between respective selected sections and used them in the interpolation to get those of remaining respective sections. Although this method enables us to obtain acceptable shape of body model, mere proportional calculation cannot make it a realistic model because directional growth of some sections of the body are anisotropic as can be seen in some statistical data. For example, bust and hip sections grow mainly in specific direction as the circumference increases. Therefore it was necessary to assign different growth ratios to the selected base sections considering statistical anthropometric data to obtain a more realistic body model. Figure 4 shows the selection of the base sections and the example of directionally different growth ratios for bust section. Using the rules mentioned, different sized body models can be obtained by changing some parameters and Figure 5 shows some examples of body models.

Prediction of garment drape shape 41

3. Generation of garment preform To predict the garment's draped shape on the human body, flat garment patterns should be transformed into a connected three-dimensional form. Because the patterns drafted by a CAD system have no information about the final draped shape of garment, it is impossible to obtain the answer directly from them. Therefore we developed a method that can make a threedimensional garment preform from flat patterns adopting some finite element

Figure 2. Extraction of radial data from a cross-section image

IJCST 12,1

42

Figure 3. Reconstructed human body model

Figure 4. The selection of base sections and the example of different growth ratio for bust section

methodologies. The flat patterns are divided into fine meshes and assembled together to form a geometrically compatible three-dimensional body. The term ``geometrically compatible'' used here means that there are very small amounts

Prediction of garment drape shape 43 Figure 5. Examples of various sized torso model

of strain between flat patterns and the resulting three-dimensional shape. The total strain on the garment S can be calculated as: 4 X X L ÿ l k l S L k Element k1 Where, l is the length of the side on a deformed element and L is that of an original two dimensional element. 3.1. Automatic quadrilateral mesh generation of flat garment patterns Since the garment patterns have wide variety of irregular shape, it is almost impossible to divide them into fine meshes manually even for experimental purposes. Therefore we developed a specially designed program to generate meshes automatically for garment patterns to make the process more practical. In finite element analysis, the choice of element is very important[1]. Though it is more difficult to generate quadrilateral meshes than triangular ones[4], the quadrilateral mesh elements has been chosen because they more adequately reflect the ``shear'' behavior of the fabric irrespective of their shape as shown in Figure 6.

Figure 6. Different shear behavior between triangular and quadrilateral elements

IJCST 12,1

44

The mesh generation process can be described briefly as follows. First, the boundary of a garment pattern (generated by the CAD system introduced in part I of this paper) is subdivided into an even number of segments ranging from 1 to 2cm in length. Then, the pattern is divided into two parts along one line each of whose ends is on the boundary and has no intersection with the pattern boundary. Finally, whole pattern is divided into quadrilateral elements if this process is repeated on every divided parts by recursive algorithm[5,6,10]. Figure 7 shows the overall procedure of mesh generation. The line along which the pattern is divided is called ``the best splitting line'' in this study. This is the line which has minimum value of an evaluation function F calculated for every line connecting two points on the boundary. 4 X l C2 F C1 ÿ i C3 " where Ci is weight factor 2 DL i1 Since DL is the diagonal length of the smallest rectangle enclosing the pattern and l is the line length, as the line gets shorter, the first term becomes smaller. The second term is the sum of the deviation of terminal angles from right angle and thus becomes zero when the line meets the boundary perpendicularly. The third term is the error caused by the boundary constraints since every divided part should have even number of segments on their boundary for the subdivision into quadrilaterals. Figure 8 is the example of a best splitting line. As there might be some acute and obtuse angles mixed in a garment pattern, the pattern cannot be divided into evenly shaped elements by this method alone. To solve this problem, we generated the offset element which has somewhat regular shapes along the boundary before whole subdivision was made. Figure 9 shows an example of generated elements with and without offset elements. 3.2. Pattern assembling To make a garment from the resulting patterns by the virtual-sewing method, we divided the two facing boundaries to be sewn into the same number of segments and created new points at the middle of each pair of sewing points and reconstructed the boundary elements using those new points. Figure 10 explains this process applied on the bodice patterns.

Figure 7. Schematic diagram of the overall procedure of automatic mesh generation

Prediction of garment drape shape 45

Figure 8. Schematic diagram of the determination of the best splitting line

Figure 9. Example of the mesh generation with and without offset element

But, as shown in the figure, the garment looks very different from the shape which can be assumed from the original patterns because large deformations were introduced by simple sewing. In order to make this garment shape geometrically compatible with original patterns, we developed the strain reduction method that reshapes the elements of resulting garment preform. In this method, the edge lengths and the diagonal distances of each element are adjusted in accordance with the corresponding elements in the original flat pattern until the total geometrical deference between the original flat pattern and the garment preform becomes smaller than a specified tolerable amount (defined as 0.1 percent in this study). Because there are not any complex calculations, the preform can be obtained in a relatively short time. For

IJCST 12,1

46

Figure 10. Schematic diagram of virtual pattern sewing for a bodice

Figure 11. Example of garment preform generation using strain reduction method

example, a bodice form shown in Figure 11 has 300 elements and the process took only 1 second on the Pentium 200 MHz PC system. 4. Final garment drape shape prediction 4.1. Body-garment contact checking algorithm To use the generated body model in the prediction of garment drape shape, it is necessary to develop a quick method to check the contact conditions between the body and the garment. Because the human torso generated in this study could directly be assumed as a cylinder, we defined the model in the cylindrical coordinate system, which made the process easy. First, the coordinates of a point on the garment defined in the Cartesian coordinate system are converted into the cylindrical coordinate system. Then the length of the line connecting the central axis of the body and that position is compared with the corresponding radius of the body section including the line to determine whether that point on the garment model is placed inside or outside the body. Figure 12 describes the contact checking procedure. If a point on the garment turns out to be inside the body, then the point should be expelled and in that case, the expelling direction becomes important. The expelling direction can either be the direction of the normal vector of the

Prediction of garment drape shape 47

Figure 12. Schematic diagram of body-garment contact checking algorithm

body or the garment on the contact point. As we fixed the human body and constantly deformed the garment to keep two models out of touch, the normal vectors of all the points on the garment were also changed constantly. For this reason, the infiltrating points were effectively expelled when we defined the expelling direction as the normal vector on the garment, but the garment model sometimes permanently vibrated when we defined it on the body model. The normal vector on the garment can be defined as the sum of the normal vectors of the neighboring elements of a point on the garment as shown in Figure 13. 4.2. Strain reduction and pseudo-drape algorithm After the body model and the garment preform was prepared, the final shape of garment can be predicted as follows. The two objects are arranged spatially in appropriate positions. At this time, some points on the garment might be inside the human body because there had been no consideration of contact with the body during the garment preform generation mentioned

Figure 13. Definition of the normal vector for a point on the garment

IJCST 12,1

48

above in chapter 3. And expelling those inside points causes new strains on the previously strain-stabilized garment. For this reason, the expelling distance becomes important. In this study, we did not calculate the expelling distance analytically, but we found the best expelled position by moving such points toward the expelling direction until they just cleared the body. Therefore the final shape of garment can be obtained by removing those strains by strain reduction until the total amount of geometrical strain on the garment model gets below a predefined tolerance keeping all the points on the garment outside the body. However there had not been any consideration of some factors such as gravitational effect in the procedure above; the predicted shape may have a different shape from what was initially desired. Therefore we introduced the ``pseudo-drape'' technique which constantly moves the garment model to the direction of gravitation during the strain reduction process to obtain a shape analogous to the one draped by real gravitation. Though there were some restrictions in this method to obtain an exact draped shape of the garment, it was considered acceptable to meet the design-time needs of quick response due its fast processing time. To make the predicted shape more practical for design purposes, we used texture mapping technique to apply fabric texture on the garment. As the calculation was simple, the final shape can also be obtained in a relatively short time; it took less than 30 seconds for an average garment composed of 300 to 500 elements. Figure 14 shows the examples of predicted garment shapes. 5. Conclusions Based on the conventional two-dimensional pattern drafting system, some improvements have been made for more efficient integration of CAD system into the industrial garment manufacturing process. The pattern drafting system developed in part I of this paper had focused on the development of

Figure 14. Examples of the predicted garment shapes

computerized rules for the generation of patterns and tried to make it possible even for an unskilled patterner to work as an experienced skilled worker. In part II, a three-dimensional CAD system has been developed to generate resizable human body models and to visualize the actual drape shape of the garment tried on human body by utilizing the finite element method. In the garment drape shape prediction system we used the strain reduction and pseudo-drape method to meet the need of the goal of an apparel CAD system ± the quick response ± even though the result bears some differences from the exact draped shape. To get more exact draped shape of the garment, it may be necessary to develop a special finite element method. Though it is very important to predict the draped garment shape, for the development of a more complete three-dimensional apparel CAD system, there should be more studies on the direct acquisition of flat patterns from the measured data of the human body. References 1. Cool, R.D., Concept and Applications of Finite Element Analysis, John Wiley & Sons, New York, NY, 1989. 2. Heisey, F., Brown, P. and Johnson, R.F., ``Three-dimensional pattern drafting: a theoretical framework'', Clothing Textile Res. Journal, Vol. 6 No. 3, 1988, pp. 1-9. 3. Hinds, B. K., McCartney, J., Hadden, C., and Diamond, J., ``3D CAD for garment design'', International Journal of Clothing Science and Technology, Vol. 4 No. 4, 1992, pp. 6-14. 4. Kang, T.J., Yu, W.R. and Chung, K.S., ``Drape simulation of woven fabric by using the finite-element method'', Journal of Textile Institute, Vol. 86 No. 4, 1995, pp. 635-48. 5. Lee, C.K. and Lo, S.H., ``A new scheme for the generation of a graded quadrilateral mesh'', Computers & Structures, Vol. 52, 1993, pp. 847-57. 6. Masuda, T. and Imaoka, H., ``3D torso surface curvatures as it relates to clothing design'', Sen'I Gakkaishi, Vol. 15 No. 6, 1998, pp. 299-308. 7. Okabe, H., Imaoka, H. and Tomiha, T., ``Transformation from paper pattern to spatial structure of dress by computer ± simulation of sewing and dressing'', Sen-i Gakkaishi, Vol. 44, 1988, p. 129. 8. Imaoka, H. and Okabe, H., ``Prediction of 3-D shapes of garments from 2-D paper patterns'', Sen'I Gakkaishi, Vol. 45 No. 10, 1989, pp. 420-6. 9. Stylios, G. and Wan, T.R., ``A new collision detection algorithm for garment animation'', International Journal of Clothing Science and Technology, Vol. 10 No. 1, 1998, pp. 38-49. 10. Zhu, J.Z., Zienkiewicz, O.C., Hinton, E. and Wu, J., ``A new approach to the development of automatic quadrilateral mesh generation'', International Journal for Numerical Methods in Engineering, Vol. 32, 1991, pp. 849-66.

Prediction of garment drape shape 49

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,1

50 Received January 1998 Revised September 1999 Accepted September 1999

A study of the roll planning of fabric spreading using genetic algorithms C.L. Hui Patrick and S.F. Ng Frency

Institute of Textiles and Clothing, The Hong Kong Polytechnic University, Hong Kong, ROC, and

C.C. Chan Keith

Department of Computing, The Hong Kong Polytechnic University, Hong Kong, ROC Keywords Fabric, Garments, Manufacturing Abstract In the process of fabric spreading, the variance of fabric yardage between fabric rolls may lead to a difference in fabric loss during spreading. As there are numerous combinations the arrangement of the fabric roll sequences for each cutting lay, it is difficult to construct a roll planning to minimise the fabric wastage during spreading in apparel manufacturing. Recent advances in computing technology, especially in the area of computational intelligence, can be used to handle this problem. Among the different computational intelligence techniques, genetic algorithms (GA) are particularly suitable. GAs are probabilistic search methods that employ a search technique based on ideas from natural genetics and evolutionary principles. This paper presents the details of GA and explains how the problem of roll planning can be formulated for GA to solve. The result of the study shows that an optimal roll planning can be worked out by using GA approach. It is possible to save a considerable amount of fabric when the best roll planning is used for the production.

International Journal of Clothing Science and Technology, Vol. 12 No. 1, 2000, pp. 50-62. # MCB University Press, 0955-6222

Introduction In clothing production, the fabric cost alone is about 35-40 percent of the selling price of a garment, that is the major cost item in clothing product[1]. In recent years, the price of fabric has increased continuously, so a certain percent reduction in fabric cost would affect the total manufacturing cost. The fabric spreading and cutting is the major production process that determines the material utilization as well as the finished quality of the garment. Apart from the fabric loss due to the fabric flaws, there are two causes of fabric loss in the production process: (1) marking loss or marker fallout, which is formed because of the gaps and other non-usable areas that take place between the garment panels of a marker; and (2) spreading loss, which is the fabric loss that exists during the spreading process other than the loss caused by the marker arrangement; these include the end loss, width loss, splicing loss and remnant loss. We are grateful to Mr Lewis Chung for preparing the coding and the results of this experiment in graphic form.

Although marker planning is always the major determinant of material utilisation, it is also important to plan and control the spreading process carefully in order to minimise the spreading loss. In the process of fabric spreading, the yardage quantity of one fabric roll usually is insufficient for the requirement of a cutting lay. Therefore, the yardage quantity required for each cutting lay normally will come from several rolls of fabric. Generally, there is a variance on the fabric yardage on each fabric roll. Garment manufacturers usually accept this pattern of fabric delivery provided that the total amount of fabric delivered is correct. If the yardage of each fabric roll is not the same, the sequence of combining the fabric rolls for each cutting lay may lead to various spreading losses. As there are numerous combinations of the arrangement of the fabric roll sequences, it is very difficult for the operators or supervisors in the cutting room to determine an optimal fabric roll sequence for a cutting lay for minimizing the spreading loss. In practice, the operator or supervisor in the cutting room usually selects the fabric rolls randomly without any roll planning in the spreading process. With reference to the mathematical model[2] that can predict the amount of material wastage of a particular lay during fabric spreading, this paper introduces a new approach to handle the problem in roll planning by using genetic algorithms. This technique could effectively determining the optimal sequence of fabric rolls for each cutting lay in order to minimise the fabric loss during spreading. Genetic algorithms based solution approach The basic principles of genetic algorithms (GA) were first proposed by Holland[3]. The use of GA has been widely adopted for numerous optimisation purposes. Genetic algorithms (GA) are probabilistic search methods that employ a search technique based on ideas from natural genetics and evolutionary principles. GA employs a random directed search for locating the globally optimal solution. They are superior to many ``gradient descent'' techniques as they possess the ability to locate the globally optimal solution for a multimodal objective function. Thus, GA is suited for applications in nonlinear function optimisation and the non-linear programming problem[4,5]. GA works with a population of individuals representing potential solutions to a problem. Each individual is usually represented by a single string of characters. At every iteration of the algorithm, a fitness value, f(i), is calculated for each of the current individuals. Based on this fitness function, a number of individuals are selected as potential parents. Two new individuals can be obtained from two parents by choosing a random point along the string, splitting both strings at that point and then joining the front part of one parent to the back part of the other parent and vice versa. This process is usually called crossover, at an operation rate with a typical value of between 0.6-1.0[5]. Individuals may also change through random mutation when elements within a string are changed directly at a smaller probability with a typical value of less than 0.1[5]. The processes of crossover and mutation are collectively referred to as reproduction. The end result is a new population (or the next

A study of the roll planning

51

IJCST 12,1

52

generation) and the whole process repeats. Over time, this algorithm leads to convergence within a population with fewer and fewer variations between individuals. When a GA works well, the population converges to a good solution of the underlying optimisation problem and the best individual in the population after many generations is likely to be close to the global optimum. The structure of a GA is summarised as below: (1) Create the initial generation. (2) Evaluate the fitness of each individual in the initial generation. (3) Perform the following steps until the termination condition is true: . Create new individuals by mating individuals in the current generation using the genetic operators of select, crossover and mutation. . Evaluate the fitness of each newly created individual. . Create a new generation by inserting new and deleting old individuals in the current generation. (4) Return the best individual(s). We proposed a steady state genetic algorithm to handle the roll planning problem of fabric spreading. A steady state genetic algorithm solution to optimise the sequence of the fabric rolls in the spreading process Chromosome representations Solutions in genetic algorithms are represented by strings of different types (``genomes''). The choice of solution representation (structure) in GA is related to the nature of the problem. It is encoded to form a chromosome. In our case, the fabric roll being laid in the cutting lay is associated with a fixed position on the string and the code is simply the fabric roll number with arbitrary roll length. A chromosome representation for this optimisation problem is shown in Figure 1. In this chromosome representation, we have put the fabric rolls into an order depending on the predetermined fabric yardage of the cutting lay. Initialisation During initialisation of population, random initialisation is adopted. Once duplication is found during initialisation, a new chromosome will be generated to replace the duplicated one. Population size The choice of population size can have an impact on the performance of a GA. If a large population size is chosen, the evolution process may be too slow. But if the size is too small, the population may not sufficient to evolve the best chromosomes. For our GA, we choose a population size of 30.

Figure 1. Example of chromosome representation

The fitness function In a GA, the fitness function provides a way of evaluating the status of each chromosome. It is used to help determine which individual survives into the next generation. Since the objective of this roll planning problem is to minimise the total fabric wastage in the spreading process by optimising the sequences of fabric rolls, the fitness function is defined in terms of the mathematical expression that derived from the study of Ng et al.[2]. The full description of the fitness function is stated as below: Definition of fitness function Prior to the definition of fitness function, the following notation is adopted. In the notation i ranges from 1 to N and j ranges from 1 to N: m Marker length w Marker width n Number of splicing interval Sn Length on nth splicing interval Tj Length of marker from left hand to right hand end of the jth splicing interval Tj' Length of marker from right hand to left hand end of the jth splicing interval Li Length of ith fabric roll W Width of fabric roll h Overlap of mean length during spreading Pi Number of complete plies laid by ith fabric roll Ri Length of remnant laid by ith fabric roll Vi Distant from the end of the marker to the cut edge laid by ith fabric roll xi Cut off length of ith fabric roll g Allowance of mean length made for the fabric in turn between one ply and the next Ni Total no. of turns from 1st fabric roll to ith fabric roll k Proportion of the fabric loss in the kth splicing interval Bi Total length of ``Internal'' loss from 1st fabric roll to ith fabric roll Ai Area of fabric loss in the ith roll Aw Total area of fabric loss over all fabric rolls at Area proportion of fabric loss over all fabric roll The definition of fitness function begins with a theoretical analysis on marker. Suppose a marker is drafted with a length m and width w, and, m is separated into n splice intervals from S1 to Sn by the splice lines as stated in Figure 2. The notation Tj is used to describe the partial sum of the length of the splice intervals S1 to Sj from left to right.

A study of the roll planning

53

IJCST 12,1

54 Figure 2. A theoretical analysis of market (fabric laid from left to right)

On the other hand, Figure 3 used the same marker as shown in Figure 2, but, 0 0 the fabric is laid from right to left. Therefore, m is equal to the sum of S1 to Sn 0 0 0 and Tj is the partial sum of splice interval S1 to Sj from right to left. In this theoretical analysis of marker, fabric loss due to spreading generally derived from two sources: ``internal'' and ``external'' waste. The ``external'' waste is referring to the fabric loss due to laying up fabrics. In Figure 4, an ith fabric roll, of length Li and width W( w) has been laid down. This roll starts at the end of one of the splicing intervals Viÿ1 with an allowance h which is used for fabric alignment since it is difficult to align the roll exactly to the splice line. Being laid the first ply (possibly incomplete), the ith roll can

Figure 3. A theoretical analysis of market (fabric laid from right to left)

Figure 4. A theoretical analysis of ``external'' waste

spread for a certain number of complete plies pi and leaving a remnant length equal to Ri. To continue spreading, Ri will be spliced according to the splice position, which depends on where the end Ri is laying within (j+1)th splicing interval. If the remnant is laid from left to right, it results in T0 j Ri Tj1 . 0 And, if the remnant is laid from right to left, then, Tj Ri Tj1 . The length between the splice line to remnant's end is xi, which is wasted (splice loss). In computing the ``external'' waste of the ith fabric roll, a strip of length Li with width equal to (W-w) will be removed as width loss. Besides, the extra allowance g represents the turns at ends of the cutting lay. The area of each turn is equal to the length of g multiply w, and, totally there are (1+pi) number of them. This is what is named the end loss in spreading. Another waste is the length of allowance h, which is provided in splicing with its area equal to hw. Finally, the area of splice loss of length xi and width w is wasted and termed splice loss. Gathering all these wastages together, the area of ``external'' waste of the ith fabric roll is: LiW ÿ w 1 pi gw hw xi w

1

The ``internal'' waste is pointing to the wastage arising within each splice interval. During the marker making process, it is inevitable for gaps and other fallout areas to exist between garment panels. This is known as marking loss. If a marker is divided into a certain number of splicing intervals, then, ``internal'' waste is referring to the fallout area of each interval. Let k be the proportion of waste fabric in the kth interval. The internal waste in the first ply is equal to that in the intervals J(i-1)+1 to n, i.e. n X k Sk w 2 kJ iÿ11

Similarly, the interval waste in the final ply is equal to that in the first J(i) intervals, or J i X k Sk w 3 k1

Finally, the waste in each of the Pi complete plies is: n X k Sk w

4

k1

Hence the total8area of internal waste is: 9 Ji n n < X = X X w k Sk pi k Sk k Sk wBi ; say :kJiÿ11 ; k1 k1 The final expression for the area of waste in the ith roll is therefore:

5

A study of the roll planning

55

IJCST 12,1

Ai Li W ÿ w 1 pi gw hw xi w Bi w If there are P rolls altogether, the total waste area over all pieces is: Aw

56

6

P X

Ai

i1

W ÿ w

P X i1

P P P X X X Li gw 1 pi Phw w xi w Bi i1

i1

7

i1

By the work of simplification in the paper, the total area of fabric is A=WLt, the final expression for proportion of waste area over all rolls is: ( ) P P X X Aw w 8 1ÿ m 1 pi VP ÿ Bi at A WLt i1 i1 In this study, the equation (8) would become the fitness function to compute the total fabric wastage during spreading. To find the best solution, a GA is set to minimise the total fabric wastage (at) in the spreading process among all solutions. Vice versa, a GA is set to maximise the total fabric wastage (at) for finding the worst solution. Minat ) Best solution of fabric roll plan in a cutting lay Maxat ) Worst solution of fabric roll plan in a cutting lay To make a fast convergence towards the local optimum, the previous research works[5-7] proposed that a linear scaling of the fitness values should be implemented. The scaling is performed in such a way that the average fitness remains constant but the maximum fitness is a multiple (usually 1.5) of this average value[6]. In other words, the actual chromosomes' fitness, f(i), is scaled as a x f(i) + b, where a and b are chosen to enforce the scaled average fitness values[5]. Selection of parents The selection technique used by our GA is the roulette wheel selection technique. The reason for such a name is that it can be viewed as allocating pieshaped slices on a roulette wheel to population members, with each slice proportional to the member's fitness. Selection of a population member to be a parent can be viewed as a spin of the wheel, with the winning population member being the one in whose slice the roulette spinner ends up. Although this selection procedure is random, each parent's chance of being selected is directly proportional to its fitness. Over a number of generations this algorithm will drive out the least fit members and contribute to the spread of the genetic material in the fittest population members. Even though it is possible for the worst population member to be selected by this algorithm each time it is used, the odds of this happening are negligible[7].

Operators: uniform order-based crossover and scramble sublist mutation In our GA, reproduction involves two parents. After two chromosomes are selected from the current population, our GA applied a ``uniform order-based crossover'' operator by recombining the generic materials in the two-parent chromosomes to create two children. A crossover rate that was used in our case is 0.65. The working principles and an illustrative example of the uniform order-based crossover are presented in Figure 5. Working principle: (1) Generate a bit string that is the same length as the parent. (2) Fill in some of positions on child 1 by copying them from parent 1 wherever the bit string of an ordered list contains a 1. (3) Make a list of elements from parent 1 associated with a 0 in the bit string of an ordered list. (4) Permute these elements so that they appear in the same order on parent 2. (5) Fill these permuted elements in the gaps on child 1 in the order generated in statement 4. (6) To make child 2, carry out a similar process.

A study of the roll planning

57

To ensure that a GA is able to escape from a local optima, a scramble sublist mutation operator[6,7] is used. This operator selects a sublist of the items on a parent order-based chromosome and permutes them in the child, leaving the rest of the chromosomes as it was in the parent. An illustrative example of scramble sublist mutation is presented in Figure 6. The rate of mutation that was used in our roll planning problem is 0.008. Delete members of the population to make room for the new chromosome When reproduction occurs, the deletion technique used in our GA is to delete all members of old population except six best members in the population will be remained in the next iteration. All other member of the old generation are replaced by new ones without replication.

Figure 5. Working principle and illustrative example of uniform order-based crossover

Parent = (2 4 j 7 1 4 8 j 3 5 9) Child = (2 4 j 4 8 1 7 j 3 5 9) Remarks: with the beginning and the end of the selected sub-list market by j

Figure 6. Example of scramble sublist mutation

IJCST 12,1

58

Experimental results and discussion Actual production data of 14 cutting lays were collected from a local men's shirt factory for the experimental study (see Appendix (Table AI) for detailed production data of 14 cutting lays). Based on the data collected and using our proposed GA as set out in the previous section, the fabric wastage of each cutting lay under various sequences of fabric rolls was worked out. When a GA is set to minimise the total fabric wastage during spreading, the solution is considered as the best roll planning that can optimise the material utilisation. When a GA is set to maximise the total fabric wastage during spreading, the solution is termed as the worst roll planning. After 300 iterations, the results of the experiment are shown in Table I. The best solution and the worst solution of each of the 14 cutting lays are plotted in a line graph as shown in Figures 7a and 7b. In these graphs, the solid line represents the best/worst solution of each generation. A good estimate of fabric saving would be the average of fabric saving of all the possible sequences of fabric rolls being laid on a cutting lay. In our case, the estimation of possible fabric saving should be half the difference of fabric usage between the best and worst cases. Using this estimation, the amount of fabric that can be saved by optimising the roll planning in a cutting lay could be computed. The results revealed that the amounts of fabric that can be saved for each cutting lay range from 0.13 percent to 1.15 percent. On average, approximately 0.42 percent of fabric can be saved when the fabric roll sequence is arranged by using GA. This is a considerable saving of fabric in production particularly when the volume of cutting order is huge. Conclusions This paper demonstrated the application of GA on the roll planning of fabric spreading in apparel manufacturing. Comparing the use of GA to an Cutting lay no.

1 2 3 4 5 6 7 8 9 10 11 Table I. 12 Experimental results of 13 14 14 cutting lays

Fabric usage under best roll planning (%)

Fabric usage under worst roll planning (%)

Amount of fabric saved (estimation) (%)

78.44 89.55 84.06 87.03 83.32 76.69 78.86 87.65 84.50 86.90 88.63 87.01 88.21 84.48

77.85 89.13 83.63 86.58 82.84 76.05 76.89 87.11 82.20 86.30 88.38 86.58 87.50 82.50

0.30 0.21 0.22 0.23 0.24 0.32 0.99 0.27 1.15 0.30 0.13 0.22 0.36 0.99

A study of the roll planning

59

Figure 7a. Best solution of each cutting lay generated by GA

IJCST 12,1

60

Figure 7b. Worst solution of each cutting lay generated by GA

exhaustive search, GA is much more powerful in searching for a solution. For instance, GA requires 9000 evaluations for 300 generations of 30 samples but exhaustive search requires 2.6 x 1022 evaluations for a 23 rolls' problem. In our experiment, the results showed that the amount of fabric saving could be estimated by about 0.4 percent when the fabric rolls of a cutting lay are planned in an optimal sequence. Reducing 0.4 per cent material has a considerable saving on the cost of apparel production. Previous research[1,9] highlighted that a 2.5 per cent reduction in fabric could save a company 1 per cent in cost, and such amount of reduction in material costs could cause profit to increase by 10 per cent. This implies that a company can increase the profit by 1.25 per cent when they can optimize the roll planning in fabric spreading by using a GA. References 1. Powell, L.Q., ``More efficient marking'', Femme-Lines, November-December, 1977, pp. 13-15. 2. Ng, S.F., Hui C.L. and Leaf G.A.V., ``Fabric loss during spreading: a theoretical analysis and its implications'', Journal of Textile Institute, Vol. 89 No. 4, 1998, pp. 686-95. 3. Holland, J.H., Adaption in Natural and Artificial Systems, MIT Press, Cambridge, MA, 1975. 4. Srinivas, M. and Patnaik, L.M., ``On modelling genetic algorithms for functions of unitation'', IEEE Trans. on Systems, Mans. and Cybernetics, Vol. 26 No. 6, December, 1996, pp. 809-21. 5. Man, K.F., Tang, K.S. and Kwong, S., ``Genetic algorithms: concepts and applications'', IEEE Trans. on Industrial Electronics, Vol. 43 No. 5, October, 1996, pp. 519-33. 6. Zbigniew, M., Genetic Algorithms + Data Structures = Evolution Programs, 3rd ed., Springer-Verlag Berlin Heidelberg, New York, NY, 1996. 7. Goldberg, D.E., Genetic Algorithms in Search, Optimization and Machine Learning, Addison-Wesley, Reading, MA, 1989, pp 122-4. 8. Gustas, L.F., ``Improved profitability from better material utilization'', Femme-Lines, November-December, 1977, p. 12 (Appendix follows over page)

A study of the roll planning

61

Marker length (m) Market width (cm) Marker utilization % Roll width (cm) No. of rolls being laid Sequence of roll Length (m) during Spreading (from bottom to top) 1 2 3 4 5 6 7 8 9 10. 11 12 13 14 15 16 17 18 19 20 21 22 23

20.21 145.50 86.35 147.30 14 126.13 114.25 79.52 114.25 63.07 60.32 71.29 60.32 105.11 112.42 100.54 80.43 96.88 48.44

23.01 112.40 91.52 113.00 16 119.73 115.16 120.65 120.65 120.65 120.65 112.42 108.77 120.65 88.66 113.34 113.34 116.99 112.42 102.37 122.48

4

5

20.98 34.878 23.85 148.0 111.84 111.76 89.92 89.79 86.40 150.00 113.03 113.03 18 17 11 77.69 78.60 110.59 95.51 115.16 119.73 69.92 112.42 101.45 114.25 112.42 114.25 77.23 110.59 110.59 56.00 111.51 110.59 89.50 113.34 111.51 105.20 114.25 112.42 86.00 105.11 115.16 116.00 112.42 112.42 100.60 116.08 110.59 108.30 114.25 109.70 82.26 71.30 113.34 78.20 107.85 66.30 102.37 67.00 109.68 63.00

3 23.19 142.20 86.98 150.00 23 78.60 83.17 26.51 24.68 85.00 82.26 72.21 80.43 84.09 39.30 27.42 31.08 38.39 54.84 84.09 56.67 83.17 84.09 60.32 84.09 80.43 78.60 82.26

6 16.86 142.20 88.54 150.00 23 9.14 14.62 12.80 12.80 83.17 25.59 84.09 71.29 82.26 13.71 83.17 30.16 30.16 37.47 21.02 16.45 29.25 85.00 21.00 84.09 84.09 66.72 83.17

7 23.27 111.84 90.00 113.03 16 79.00 109.68 74.95 108.77 108.77 108.77 108.77 54.84 91.40 109.68 109.68 109.68 109.68 109.68 109.68 109.68

8 25.87 147.30 91.08 150.00 15 81.35 104.65 78.60 76.78 81.35 63.07 76.78 83.17 87.29 75.86 77.69 96.43 97.80 81.35 79.52

9 29.57 111.80 89.50 113.03 17 109.68 109.68 109.68 109.68 112.42 109.68 109.68 109.68 109.28 109.68 109.69 109.68 109.68 109.68 109.68

10 30.25 112.40 91.53 114.30 15 110.59 110.59 110.59 109.68 112.42 109.68 54.84 112.42 110.59 112.42 114.25 108.77 112.42 112.42 109.68

11 34.87 111.84 89.79 113.03 15 126.13 111.51 65.81 111.51 113.34 110.59 110.59 63.98 102.37 73.12 112.42 112.42 111.51 111.51 111.51

12

2

Table AI. Production data of 14 cutting lays collected from a men's shirt factory 1

62

Cutting lay no.

14.47 111.84 90.90 113.03 13 73.12 109.68 70.38 63.98 71.29 54.84 54.84 109.68 109.68 109.68 110.59 109.68 106.94

13

25.87 147.30 91.05 150.00 14 106.94 128.87 92.31 109.68 92.31 100.54 57.58 95.06 94.60 95.97 113.34 113.79 108.77 129.79

14

IJCST 12,1

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Estimation of fused textile systems shrinkage

Fused textile systems shrinkage

M. Gutauskas and V. Masteikaite,

Kaunas University of Technology, Kaunas, Lithuania, and

L. Kolomejec

Burda Moden Showroom, Vilnius, Lithuania Keywords Clothing, Textiles, Twisting

63 Received March 1998 Revised October 1999 Accepted October 1999

Abstract Presents the results of investigation of fabrics' and fused systems shrinkage force and its influence on the shape instability of fused textile systems after various treatments such as wetting and drying. Also the method for determination and prediction of the twisting phenomenon of fused composites is presented. From the analysis of the data the twisting phenomenon was characterised by several parameters. They are system's twisting sign, the shape of twisted system, the twisting coefficient and fused system's twisting direction.

Introduction At present during garments' performance in the sewing industry fusing technology is widely used. The advantages of this technology are good shape, formability and appropriate elasticity of a garment and sufficient stability of fused areas during garment wear. However, the fused system consists of two or more fabrics' layers which are different in structure and composition. So, there exist some problems during garments wear, especially after dry cleaning and washing procedures. For example, the effect in different relaxation shrinkage of interlining and outer fabric cause bubbles, blisters, and puckers in the face fabric. An analysis of surface folding and uneven appearance of fused parts have shown that the main reason of this phenomenon is the sufficient difference between the dimensional instability of face fabric and interlining[1,2,4]. Review of the literature indicates that the unsuitable selection of the face fabric and interlining occurs when the difference in relaxation shrinkage of these layers is higher than 2[3] or 2.5[4] per cent. There is no one single method of testing the dimensional change of fabrics and fused systems. For the determination of the fused textile systems dimensional change the authors often used standard methods such as ISO 5077-1984(E), ASTM D 2724, specified standard procedures[8] or others. For the investigation of the shrinkage phenomenon of the interlinings, outer fabrics and fused textile systems, used often were various sample's treatments: boiling in the water, washing, dry cleaning, steam and hot pressing or fusing[2,3]. To ascertain the behaviour of the interlining in the system during fusing some authors[4] used the pre-shrinking of the face fabric to minimise its residual shrinkage. Thus, any shrinkage during treatment (laundering) would be attributed to the interlining. To determine the individual shrinkage

International Journal of Clothing Science and Technology, Vol. 12 No. 1, 2000, pp. 63-72. # MCB University Press, 0955-6222

IJCST 12,1

64

characteristics during the fusing process other authors offer to place two interlining's layers together with the coated sides in the middle and to fuse them. Usually, in order to receive the shrinkage values after washing or dry cleaning the fused sample must be washed or dry cleaned five times within a period of 24 hours[3,4]. It should be noted that unequal relaxation shrinkage of fused systems layers leads to the other defect which occurs especially for small details. They bend similarly to bimetallic plate after heating. This type of fused system deformation was named as twisting[5,6]. The data in the work[2] have shown that in many cases the shrinkage of the fused fabric is between the composed face fabric and interlining. Review of the literature indicated that there is a lack of published material devoted to the selection rules of fused systems' components taking into account their shrinkage characteristics. Our earlier investigations have shown that sometimes it has been impossible to explain fused textile systems' behaviour after wetting-drying processes with results based on the layers shrinkage values. In some cases a good agreement was not found between the fused system twisting intensity and values of its layers relaxation shrinkage. As a rule many fused systems after their wettingdrying curl up on the side of the component with larger value of its relaxation shrinkage. Contrary to expectations some of them do not behave by this rule. For example, the interlining with high value of relaxation shrinkage had not the strength to curl up the stiffer but less shrinkage face fabric. It was also found that systems with the same interlining and different face fabrics of similar relaxation shrinkage's values were twisted up with various intensity. The possible explanation of these effects may be not only the influence of the relaxation shrinkage of the layers but their shrinkage force, too[5]. So, the shrinkage force characteristic is also necessary for the prediction and determination twisting phenomenon of the fused system. The investigation in threads' and fabrics' shrinkage force during their wetting in the various liquids is presented in work[7]. This study investigated the effect of fabrics' shrinkage force on the shape instability of fused textile system after such treatments, as wetting and drying. Also, in this work we presented the method for determination and prediction the twisting phenomenon of fused composites. Methodology In order to reveal the reason of shape instability of fused systems during their shrinkage, 50 different systems were investigated. For this paper we selected six face fabrics and two interlinings. Table I provides basic information about the tested fabrics used in this work. Before testing, all the samples were conditioned to 65 percent RH at 200C. Fabrics' fusing was performed on a flatbed press machine. The fusing parameters were: temperature T = 1500C, time = 20-40s, pressure p = 0,1-0,02MPa.

Fabric

Structure

Mass g/m2

A-coating

2 twill

356

B-coating

2 twill

438

C-suiting

plain

243

D-suiting

2 twill

263

E-dressing F-dressing G-interlining H-interlining

2 twill plain plain plain

78 143 130 157

Content % 95 wool 5 polyamide 40 wool 29 polyester 21 viscose 55 wool 45 polyester 55 wool 45 polyester 100 acetate 100 cotton 100 viscose 100 cotton

Thickness Relaxation shrinkage % mm Wp Wf 1.30

3.4

0.2

1.35

2.2

2.2

0.59

0.7

0.8

0.48

0.6

1.0

0.13 0.31 0.43 0.48

0.5 1.4 4.5 3.0

0.5 3.1 4.0 0.7

The relaxation shrinkage of the samples was determined by such procedures. The square form sample of the fabrics or fused systems with marks in weft and warp direction (distances between them 20cm) were conditioned to 65 percent RH at 208C. All the samples were then immersed in water (T(1918C) and kept there for 1 hour. After this time was passed the sample was taken out of the water and allowed to dry for 24 hours (in room temperature) in a tensionless state. Calculation of the dimension change for each fabric direction is derived from the measured length differences before and after test as a percentage of the original length. Shrinkage force testing The schematic view of the set-up used for determination of sample's shrinkage force is shown in Figure 1. It was possible to examine sample's force changes continuously using such testing conditions as wetting and drying. The sample sized 2.5 10cm was fixed vertically in the lower immovable clamp and upper clamp that is connected with cantilever. The set-up also contains variable-induction pickup, amplifier, register apparatus, water vessel with vertical movement and heating source, and fixed on the turning round pivot. For more measurement precision initial tension of the sample was applied by attaching the constant weight to the bottom of the sample edge. Result After the sample was immersed into the water (T8200C) the shrinkage force increased gradually. In Figure 2 it can be seen that the violent increase in shrinkage force occurs during the first minute (more than 90 per cent). Similar results were obtained for single fabric's layer as well as for fused systems. According to received results, the time of water soaking in all carried up experiments was one minute. In addition, after soaking, every sample was kept in normal atmosphere conditions for four minutes to

Fused textile systems shrinkage 65

Table I. Characteristics of tested fabrics

IJCST 12,1

4 5

6

7

3 1

66 8

2 9

Figure 1. Principle of fabric's shrinkage force measurement

H2O

permit flow down the water surplus. After this procedure the samples were dried. All these effects on the samples were carried out in the apparatus shown in Figure 1. The influence of the initial tension size on the fabric's shrinkage force is shown in Figure 3a. Results indicate that this influence is insignificant, especially, for the maximum value of shrinkage force during wetting. Therefore, for all our experiments a constant value of initial tension ± 0.1N was chosen. So the weight attached to the sample of the 2.5cm width was 10g. For more heavy fabrics or those inclined to curl after fusing the larger value of initial tension can be used. It is known that owing to higher temperature, the drying process becomes shorter. On the other hand, the drying temperature has an influence on the shrinkage force value (Figure 3b). So, further investigations were carried out at drying temperature 5558C for all fabrics and systems. Therefore at T(558C the system C+G which is presented in Figure 3b reached its maximum P, N/m 50 40 30 20

Figure 2. The shrinkage force changes of fused system C + G during its wetting

10 0 0

1

2

3

4

5 t, min

shrinkage force (P=57,5N/m) and drying process lasted 34min. The sample's warp direction was chosen as the main one because it is more sensitive for shrinkage phenomenon. All the samples of fused systems and separate layers were treated in a shrinkage force tester (Figure 1). The received data reveal that all the curves of various fabrics and fused systems have a similar shape. There are differences only between intervals of and between maximum and minimum values of P (shrinkage force). To analyse the shrinkage force changes during sample wetting-drying procedure we are presented with the results of one of the tested fusing system (B+G) and its layers B and G (Figure 4). It can be seen that after the sample was immersed into the water the shrinkage force increased violently during the first minute and reached its maximum value at point O. The interval 0-1 showed the value of the sample's shrinkage force after it was taken out from the water and kept in the normal atmosphere conditions during 4 min. It can be seen that at the beginning of the drying process (interval 1-2) the shrinkage force decreased slightly for both the face fabric and fused system but more noticeable for the interlining. During more intensive moisture evaporation from the sample (interval 2-3) the shrinkage force increased until its maximum value. The heating source then was removed out and the sample began to cool at room temperature until shrinkage force reached its equilibrium (the horizontal part of curve's beginning at point 4). P, N/m P, N/m

P, N/m 20

τ, min 1

15

1

10

50 2

2

5 0 0,05

0,15

0,25

30 40

0,35 P0, N

55 (b)

(a)

P, N/m

50

80 70 60 50 0 40 0 30 20 10 0 0 0

30 65 T, °C

Fused textile systems shrinkage 67

Figure 3. The influence of: a- the initial fabric's tension P0 (in the weft direction) on the maximum (curve 1) and minimum (curve 2) shrinkage force of interlining G; b-drying temperature T on the shrinkage force P3 (curve 1) and drying duration (curve 2) of system C+G

G 3 3

B+G

1 4

1 2 1 5

2

4 3

B

4

2 15

25

35

45

55 t, min

Figure 4. The changes of samples' shrinkage force during their wetting-drying procedure (fabrics B, G and system B+G)

IJCST 12,1

68

Table II. The shrinkage force (P3) and the twisting coefficient (k) of tested fabrics and fused systems

The results received during wetting-drying experiment have shown that cooling procedure diminishes shrinkage force of samples in 1/2-2/3 of maximum value. It can be seen from Figure 4 that every curve has these important points: (0) ± maximum value of P0 during wetting the sample, (1) ± the value of P1 after keeping the sample in normal atmosphere conditions, (2) ± minimum value of P2 during sample's drying (heating), (3) ± maximum value of P3 during sample's drying (heating), (4) ± minimum value of P4 during keeping the sample in normal atmosphere conditions. As the main characteristic of shrinkage force we have chosen the largest value ± maximum shrinkage force during sample's heating (drying) P3. Experimentally we found that maximum shrinkage force for the investigated face fabrics reached 25.6 N/m. For some type of interlining (for example, G) the parameter P3 increased efficiency. Table II lists the shrinkage force of the tested fabrics and fused systems. It is evident from Table II that the shrinkage force parameter of the fused systems is larger than these values of face fabrics. For some systems the values of P3 exceed the shrinkage force values of interlinings also. It is important to note that for all tested systems the values of relaxation shrinkage are situated between corresponding values of fabric and interlining. It can be seen from Table I that relaxation shrinkage values (in warp direction) for interlinings G and H are both high and differ only by 1.5 per cent. In spite of this, the shrinkage force for these interlinings differs more considerably about nine times (Table II). There are no significant differences between the relaxation shrinkage values of suiting fabrics i.e. C and D (0.1 per cent) but their shrinkage forces differs markedly (about three times). Against all expectations these experiments have shown also that heavy-wool rich samples (A,B) distinguished less shrinkage force than light viscose fabric (G). The possible explanation of this phenomenon could be the different absorption properties of these fabrics. The viscose can take about 75 per cent of its own weight of water by enormous swelling of the fibre, meanwhile the wool absorbs about 40-45 per cent of water. The results of this experiment confirmed that relaxation shrinkage characteristics' values of the fabrics' do not correspond to their shrinkage force,

System

P3 of system layers N/m

C+G D+G E+G F+G A+G B+G A+H B+H

25.6 + 75.2 8.8 + 75.2 6.4 + 75.2 24.8 + 75.2 8.8 + 75.2 10.4 + 75.2 8.8 + 8.4 10.4 + 8.4

P3 of fused system N/m

Relaxation shrinkage of fused systems %

Twisting coefficient k

57.5 77.6 85.4 78.6 53.6 53.4 9.2 19.6

2.43 2.17 2.47 3.02 3.39 3.17 1.63 2.09

0.66 0.71 0.20 0.41 0.98 0.99 0.92 1.00

respectively. In this regard for the prediction of instability phenomenon of fused textile systems during wear, and for selection of the suitable interlining for the face fabric, the characteristic of shrinkage force would be evaluated. For better understanding of the fused systems twisting phenomenon the determination of the shrinkage force of fabrics is very important, too. Presented in this work are methods for determining the shrinkage force of fabrics which may be used not only for estimating the quality of fused systems but also for other types of systems, for example, sewn.

Fused textile systems shrinkage 69

Twisting testing For the analysis of fused systems twisting phenomenon we used circular sample of fused system (diameter 113mm). The drying procedure of the samples was carried out at room conditions in a hanging state. The experiments have shown that after the wetting-drying procedure some samples twisted in various intensities and shape. From the analysis of the obtained data the twisting phenomenon can be characterised by several parameters which are given in Table III.

Table III. The fused system's twisting parameters

IJCST 12,1

70

In sign (1), the fused system can twist on the face fabric side (a) or on the interlining side (b). In our research we assumed that the system has sign (+) when the face fabric is convex (systems D+G, E+G, F+G) and on the contrary it has sign (±) (system A+H). The twisting sign depends on the system's layer with more intensive P 3 or higher relaxation shrinkage value. In shape (2), the shape of the twisted systems can be symmetrical(a) and asymmetrical(b). The results of the test have shown that in case of symmetrical shape of the twisted system (E+G) the stable axis extends to the bias direction of the sample. Perpendicular to this axis, unstable areas of sample curl up in more or less degree. The tubelike shape of fused system is obtained owing to the least sample stiffness in bias direction and insufficient resistance to shrinkage force of one of the layers. On the asymmetrical shape of twisted sample has an influence of several factors. One of them can be non-coincident warp direction of face fabric and interlining[5]. Twisting coefficient (3). The twisting coefficient k is defined as the ratio of twisted sample's projection area (S) and initial sample's area (S0 = 100cm2). The samples projection area was received onto a piece of paper by a parallel light source disposed above the twisted sample. The shadow pattern was traced out and its area (S) was measured in two ways: by planimeter and photoelectronic pattern area measuring setting. Experimentally we have found that for the twisting coefficient determination one wetting-drying cycle is enough. The results indicated that after seven wetting-drying cycles the coefficient k increases only by 2-10 per cent. The investigation reveals that for all tested fused systems the parameter k had value within the range of 0.2 to 1.0 (Table II). According to the results all the fused systems can be divided into the three quality groups: (1) stable, when k > 0.8; (2) medium stable, when 0.5 k 0.8; (3) unstable k 0.5. In case of k > 0.8 the fused system contains successfully chosen layers. The utilisation of systems with k of medium stability has some risk and final decision depends on garment's construction, its details and shape and wear conditions. Undesirable garment quality occurs in case of using for its manufacture the unstable fused systems (k < 5). It was found that the fused systems with lower twisting coefficient showed correct twisting shape in most cases. The characteristic feature for systems of incorrect twisting is their medium value of parameter k. Direction. Angles were used to determine the fused system's twisting direction. In the case of correct sample shape one angle is enough between fabric's weft direction and perpendicular drawn to a twisted sample side (a). The evaluation of twisting direction is more complex when the twisted

sample shape is incorrect. All curl up areas of circular sample must be described. For example, the twisting direction of sample b (Table III) can be presented in such form: 1 = 1708, 2 = 408. It should be noted that the value of the fused systems twisting coefficient depends not only on the fabrics and interlinings' relaxation shrinkage and shrinkage force but also on the other fabric's properties. Especially important are such properties as bending stiffness and thickness of the system's layers (Table I). The larger thickness of the fabric A and B have some influence to the high value of the coefficient k for the corresponding systems. As evident from Table II, the high shrinkage force of the interlining G (P 3 = 75.2 N/m) has not be able to twist systems with thick face fabrics A and B. The thickness' of these fabrics are about three times higher to those of the interlining G (Table I). On the contrary, the fused system E+G distinguishes with interlining of high shrinkage force and thicker face fabric. The thickness of face fabric E is about three times less than such characteristic of interlining G (0.13/ 0.43mm). In this case the shrinkage force of the fused system E+G reaches especially high value of P3 (85.6 N/m) which overcame corresponding values of separate system's layers. Probably it is the main reason for the very low twisting coefficient of this system. These results point out the importance of fused systems layers' thickness especially differences of these values for the prediction of fused systems twisting degree. Conclusions In this work we identified the fabric's shrinkage force importance for the fused systems instability evaluation after wet treatment. This parameter can be used for the better understanding and explanation of the fused system behaviour during different shrinkage degree of its layers. The test method has been developed which can be used to assess fused systems' twisting phenomenon after their shrinkage. In order to characterise the fused systems instability after shrinking we have proposed four parameters: sign, shape, twisting coefficient and direction. The twisting coefficient was chosen as the main parameter of fused system twisting phenomenon. It allows all the fused systems to divide into three quality groups: stable, medium stable and unstable. References

c, A., ``Objective evaluations of stabilised garment parts handle'', 1. Gersak, J. and Sari International Journal of Clothing Science and Technology, Vol. 7 Nos 2/3, 1995, pp. 102-10. 2. Tien-Wei, S. and Shin-Song, L., ``The effect of fusing condition on thermal protective fabric'', The 78th Word Conference of the Textile Institute, Greece, May, 1997, pp. 489-95.

3. Ajemian, R., ``Tailoring the right approach to interlinings'', Bobbin, February, 1994, pp. 89-92. 4. Koenig, S.K. and Kadolph, S.J., ``Comparison of performance characteristics of seven fusible interlinings'', Textile Research Journal, June, 1983, pp. 341-6.

Fused textile systems shrinkage 71

IJCST 12,1

72

5. Kolomejec, L. and Gutauskas, M., ``Skruchivanye dublirovannych materialov I metod evo otsenki, Shveynaya Promyshlennost'' (The twisting of the fused fabrics' and method of their estimation), Clothing Industry, Moscow, No 1, 1981, pp. 29-30. 6. Patent of USSR No 819720, ``Sposob otsenki ustoichivosti k skruchivaniju sloistych materialov'' (The method of the fused fabrics' stability estimation), Int.Cl. G 01 N 33/36, 1980. 7. Bangert, R. and Brederek, K., ``Eine eifache Apparatur zur Messung von SchrumpfkraÈ ften an Garn-und Gurekeproblem'', Melliand Textilberichte, Vol. 54 No. 12, 1973, pp. 1347-50.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

84 Received 20 July 1998 Revised 18 November 1999 Accepted 18 November 1999

Comparison between thermal insulation and thermal properties of classical and microfibres polyester fabrics L. Schacher, D.C. Adolphe and J.-Y. Drean

Ecole Nationale SupeÂrieure des Industries Textiles de Mulhouse, Mulhouse Cedex, France Keywords Thermal analysis, Microfibre, Micro fabrication Abstract A lot of fibre and fabric structures or finishing parameters influence the functional properties of fabrics. In order to assess the thermal properties of conventional polyester and microfibre types of fabrics the plate/fabric/plate method for conductivity or cool/warm feeling was used. Fabrics made of microfibres show lower heat conductance and therefore higher thermal insulation properties. Microfibre fibres exhibit a warmer feeling than conventional polyester fabrics depending on pressure, which may be due to the difference in the fibre and fabric surface in contact with the human skin.

Introduction Even though microfibres are not very recent products, their present popularity comes from two essential qualities: (1) the quality of synthetic fibres in performance; (2) the quality of silk fibres in comfort. The concept of comfort and hand evaluation are subjective and difficult to grade objectively. Nevertheless, the thermal properties of fabrics have to be taken into account, in hand and comfort evaluations. In fact, the feeling of pleasant fabrics is directly induced by the fabric capacity to transfer the heat and humidity of the human body[10]. Two aspects of heat transfer can be taken into consideration: (1) the static aspect, which is the thermal resistance well known by heat specialists; (2) the dynamic one, which is the transitional thermal exchange at the time of contact between the skin and the fabric. It is in relation to the warm/ cool feeling which is very important in textile comfort evaluation.

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 84-95. # MCB University Press, 0955-6222

In this work, the thermal insulation and thermal contact properties of classical and microfibre polyester fabrics will be measured with the thermolab module KES-FB7 of the Kawabata Evaluation System, in order to make an objective comparison of the thermal properties of both microfibres and classical fibres.

Heat transfer through textile materials The heat transfer mechanism through fabrics has been studied for a long time[9,1,12]. These mechanisms are complex and, in the case of dry heat, are composed of the fibre thermal conduction, airtrapped thermal conduction, convection and radiation. Few behaviour models are proposed in the literature. Farnworth[4] has studied low count fabrics and determined the following formula (1): K 1 ÿ f:Ka f:Kf

1

with K

: textile material thermal conductivity

Ka : fibre thermal conductivity Kf

: air thermal conductivity

f

: part of air in the textile material.

This model does not take radiation and convection heat transfer into account and assumes that heat flow is unidirectional and parallel to the fibres orientation. Fibre orientation has been taken into account by Bogaty et al.[2] and two components appear in the K expression, the part of fibre parallel to the heat flow direction is called x and the part of fibre perpendicular to the heat flow direction is called y as shown in formula (2): K x:Vf :Kf Va :Ka

y:Ka :Kf Va :Kf Vf :Ka

2

with Va : volume of air in the textile material; Vf

: volume of fibres in the textile material.

A packing factor has been added by Holcombe[7] in the determination formula of K. This factor is in relation to the structure of the fabric and the density of the material as displayed in formula (3): K 172:packing factor 0:043:fibres conductivity 21:1

3

The aim of garments is to create a micro-climate around the human body. Therefore, the exchanges between the body and the surrounding air follow the classical model given by the equation (4): S M ÿ W Hr Hc Hk ÿ He ÿ HRs with S

: exchange

4

Thermal insulation and properties 85

IJCST 12,2

M

: individual metabolism

W : metabolism part transformed in mechanical energy Hr : radiation exchange

86

Hc : conduction exchange Hk : convection exchange He : evaporation exchange HRs : breathing exchange The thermal balance is obtained when S = 0, with the skin temperature as a variable. This model describes a wet heat transfer through a fabric as Tsushida et al., and Umbach study it[11,13]. The principal mechanisms of wet heat transfer through a fabric are: .

adsorption and migration described as the accumulation and transport of water on the surface of the fibre;

.

capillarity transfer;

.

absorption and desorption of the water by the fibres;

.

radiation and air flow due to the micro-climate between textile material and the human body and generated by the body movements.

These mechanisms are complex and depend on one another. To conclude, the heat transfer mechanisms are complex in the textile material composed of air and fibres. However, some of them are more significant than others. Thus, the thermal conductivity of the air is lower than that of the fibres and this thermal conductivity is considered as the most important transfer parameter. Nevertheless, this heat transfer is very dependent on the fabric structure[2]. The thermal insulation properties of textiles is usually called ``equivalent thermal resistance'' established in accordance with the Fourier law and is defined by the equation (5): Rth

T Q

with T : temperature gradient between the two faces of the textile material Q

: heat flow (W)

5

By analogy with a homogeneous medium, the thermal conductivity can be defined in terms of equation (6): K

h Rth A

6

with h

: thickness of the textile material;

A

: contact area.

87

Measurement devices and methodology There are various devices and methods to measure thermal properties[4-6] and the standard has been defined[15,16]. They are based on the principle of the heat guard plate or heat guard plate with cool plate added (plate/textile/plate method) where the sample is placed on the plate or between the two plates. Thus, the thermal conductivity given by the Thermolab KES-FB7 system, is based on the principle of ``plate/textile/plate'', as defined by equation (7): Q:h K A:T

7

with Q

: heat flow (W)

A

: contact area (standardised at 25cm2)

T : temperature gradient between the two faces h

Thermal insulation and properties

: thickness of the textile material

Thermal insulation is obviously strongly related to fabric thickness. Therefore, the ratio of thermal conductivity to thickness, will give a measure of the intrisic value of the materials studied. It is the heat conductance[8] given by equation (8): K 8 K0 h Its measurement time, using Kes-FB7 is 20 seconds. Few measurement devices, however, are capable of calculating temperature variations with precision. This measurement, which displays the transitional thermal exchange at the time of contact between skin and fabric (dynamic phenomenon), is related to the warm/cool feeling. In his works on fabrics evaluation, Kawabata takes this phenomenon into account and builds a specific device to measure it, the KES-FB7[14]. Based on this approach, Hes and Dolezal[6] developed their own measuring device called Alambeta. The parameters given by the two measuring devices are the following:

IJCST 12,2

88

(1) For the KES-FB7, the maximum value of the heat flow curve after 0.2s of contact between material and plate (Figure 1). The value obtained is called qmax and is expressed in W/m2. (2) The Alambeta first gives the thermal diffusivity, a, defined by equation (9): a

:c

9

with : material density l : thermal conductivity c : specific heat a is expressed in m2/s. Thanks to this ``a'' value, the thermal ``effusivity'', b, specific to the measured material, is defined by equation (10): p b ::c 10 and calculated by means of the instrument computer. Experimental study Table I presents a full description of the fabrics studied. The thicknesses have been measured with the KES-FB4 instrument. They are both plain weave fabrics; one is composed of classical polyester fibres and the other of polyester microfibres.

Figure 1. Qmax curve

Material

Microfibres polyester

Classical polyester

Warp Weft Pattern Warp count Weft count Thickness

0.7dtex (filaments) 1dtex (fibres) Plain 43 yarns/cm 36 yarns/cm 0.62mm (0.5gf/cm2) 0.25mm (50gf/cm2) 61 89.7 (0.5gf/cm2) 74.5 (50gf/cm2) 88g/m2

2.5dtex (filaments) 4dtex (filaments) Plain 29 yarns/cm 23 yarns/cm 0.45mm (0.5gf/cm2) 0.26mm (50gf/cm2) 43 88.4 (0.5gf/cm2) 79.3 (50gf/cm2) 72g/m2

Compressibility (%) Air content (%) Mass per unit area

The air contents of both fabrics are similar under low pressure (0.5gf/cm2), under higher pressure (50gf/cm2) the air content of the microfibres fabric becomes lower than the classical polyester fabric; this is due to their different compressional behaviour which can be explained by the difference of hairiness and yarn compressibility. The thermal properties were measured for each fabric with the KES-FB7 instrument in accordance with the standard procedure (surface: 10cm2, temperature difference between the two sides: 108C), except for the pressure applied to the sample, in order to examine the influence of thickness. A fixed contact temperature of 308C for the hot plate was used to reproduce the temperature of the skin surface. A KES compression tester was used to measure the compressional properties of the samples. Heat conductance analysis The analysis of the experimental results (Figure 2) shows a slight difference in behaviour between the two fabrics: the classical polyester fabric has a higher heat conductance and therefore lower insulation properties. This slight difference can be explained by the minor porosity difference between the two fabrics (cf. Table I), which difference is mainly due to the dense three dimensional lattice of the micofibre fabric which prevents higher convection transfer for fabrics made of conventional fineness polyester yarns. It should be noticed that large holes between fibres or yarns such as those found in classical polyester fabrics can increase heat exchanges by radiation. However, heat transfers by radiation or convection do not contribute a lot to the overall thermal transfer. Analysis of the warm/cool feeling It can be observed (Figure 3) that the cooler sensation is obtained with fabrics made of conventional fineness polyester yarns but that the evolution of the feeling for both fabrics is largely dependent on the pressure applied: when the

Thermal insulation and properties 89

Table I. Material used

IJCST 12,2

90

Figure 2. Heat conductance

Figure 3. Warm/cool feeling

pressure is increasing, then the cold feeling decreases for the classical polyester fabric as it is increasing for the microfibre fabric until a specific pressure is reached. At this point they behave alike. These results show how significant the quantity of the material in contact with the skin is. For classical polyester fabrics, with few fine hairs, the skin is immediately in contact with the fabric. Moreover, an increase of the pressure can partly modify the fibre cross sections. The feeling becomes more dependent on the fibres whose conductivity is, depending on the material, three to ten times more important than the air conductivity. For our fabrics, some tests made on polyester films, show a cool feeling four times that of classical polyester fabric. Comparison between thermal insulation properties We have studied the influence of wind speed on the thermal insulation properties of the two fabrics. We have applied four methods, according to the standard procedure of the instrument: .

dry method with contact;

.

dry space method;

.

wet method with contact;

.

wet space method.

If we leave space between fabrics and heat source, we can simulate a moving body. The main objective of the wet methods is the simulation of a perspiring skin and the analysis of the ability of the fabric to extract humidity from the skin. Comparison between ``dry contact method'' and ``dry space methods'' If we consider Figures 4 and 5, we can observe that heat loss is always more important for fabrics made of classical polyester yarns than those made of polyester microfibres; besides there is a linear relationship with the air flow velocity. Tight fabrics or fabrics with a lot of yarns or small fibres will prevent air passage and then reduce convection heat loss. However, adsorption and migration phenomenon are more important. The difference between the two methods proves the importance of the air film between human skin and textile material. In these methods, fabrics remain in contact with the skin and exchanges mainly take place through conduction.

Thermal insulation and properties 91

IJCST 12,2

92

Figure 4. Dry contact method

Figure 5. Dry space method

Comparison between ``wet contact method'' and ``wet space method'' In Figures 6 and 7, corresponding to the results of the standards tests of the KES apparatus, the values of the amount of water transferred (evaporated water) have been represented. In such a case, exchanges are very complex and moisture transfer has interfered. As stated before, we can observe that there is a linear relationship between loss and air flow. The difference between the ``wet method'' and the ``dry method'' can be explained if we consider that water remaining on fabric confers a better contact effect. If it is difficult to accurately define the surface contact of porous materials in general[3] and textile materials in particular, we know that this parameter is of paramount importance. Moreover, there is an increase of transfer by evaporation with the wet methods.

Thermal insulation and properties 93

Conclusion A lot of fibre and fabric structures or finishing parameters influence the functional properties of fabrics. The measurements are useful for providing objective thermal parameters and therefore wear comfort evaluation. In this study, two plain polyester fabrics were compared. One was composed of microfibres and the other of classical fibres, but both have similar air content under low pressure and similar thickness under higher pressure. Their

Figure 6. Wet contact method

IJCST 12,2

94

Figure 7. Wet space method

difference in compressional behaviour can be easily explained considering their structure. The experiments that were carried out showed the importance of the pressure parameters when measuring the following thermal properties of the fabrics using plate/fabric/plate methods: conductivity or cool/warm feeling. Considering our results, it is possible to establish that fabric made of microfibres show lower heat conductance and therefore, higher thermal insulation properties. It was also found that the microfibre fabric exhibited a warmer feeling than the classical one, depending on the pressure. The reason can be attributable to the difference in the fibre and fabric surface in contact with the human skin (thermal sensor). In order to improve the knowledge of the structure of the tested fabrics, a microscopic analysis will certainly allow a better understanding of heat transfer phenomena. References 1. Barker, R.L. and Anjaria, M.K., ``The thermal insulation properties of low density nonwovens'', International Nonwowen Conference 1988 Proceedings, pp. 367-88. 2. Bogaty, H., Hollies, N. and Harris, M., ``Some thermal properties of fabrics. Part 1: the effect of fiber arrangement'', Textile Research Journal, June 1957, pp. 93-110. 3. Degiovanni, A. and Moyne, C., ``ReÂsistance thermique de contact en reÂgime permanent. Influence de la surface de contact'', Rev. GeÂn. Therm. Fr., No. 334, October 1989.

4. Farnworth, B., ``Mechanisms of heat flow through clothing insulation'', Textile Research Journal, 12 December 1983, pp. 717-25. 5. Farnworth, B. and Dolhan, P., ``Heat and water transport through cotton and polypropylene underwear'', Textile Research Journal, October 1985, pp. 627-30. 6. Hes, L. and Dolezal, I., ``New method and equipment for measuring thermal properties of textiles'', Journal of Textile Machinery Society of Japan, 1989, pp. T124-T128. 7. Holcombe, B. and Hoschke, B., ``Dry heat transfer characteristics of underwear fabrics'', Textile Research Journal, June 1983, pp. 368-74. 8. Kawabata, S., Niwa, S. and Sakaguchi, H., ``Application of the new thermal tester `thermolabo' to the evaluation of the clothing comfort'', Journal of Textile Machinery Society of Japan, 1985, pp. 343-53. 9. Rees, W.H., ``The transition of the heat trough textile fabric'', The Journal of Textile Institute, August 1941, pp. T149-T165. 10. Shishoo, R., ``Technology for comfort'', Textile Asia, June 1988, pp. 93-110. 11. Tsushida, K., Harada, T. and Uchiyama, S., ``Fabrics properties influencing moisture and heat transport through fabrics'', Australian-Japan Symposium on Objective Specification of Fabric Quality, Mechanical Properties and Performances, Kioto, Japan, 1982 Proceedings, pp. 419-26. 12. Umbach, K., ``Messmethoden zur PruÈfung physiologischer Anforderunsprofile an Zivil Arbeits und Schutzbekleidung sowie Uniformen'', Melliand Textilberichte, November 1987, pp. 857-69. 13. Unbach, K., ``Moisture transport and wear comfort in microfibers fabrics'', Melliand Textilberichte, February 1993, pp. 173-8. 14. Yoneda, M. and Kawabata, S., ``A theoretical consideration on the objective measurement of fabrics warm/cool feeling'', Journal of Textile Machinery Society of Japan, 1982, pp. T393-T406. 15. DeÂtermination du pouvoir adiabatique et de l'indice d'isolation thermique NF G07-107. 16. Standard Test Method for Thermal Transmittance of Textile Materials, ASTM 1518.

Thermal insulation and properties 95

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

Convection/diffusion test method for porous textiles Phillip Gibson, Donald Rivin and Cyrus Kendrick

96 Received 2 June 1998 Accepted 2 December 1999

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 96-113. MCB University Press, 0955-6222

US Army Natick Research, Development and Engineering Center, Massachusetts, USA Keywords Convection, Gas, Textiles, Porous materials Abstract Reports on an automated apparatus and test procedure to determine the convective and diffusive gas and vapor transport properties of small pieces of woven and nonwoven fabrics, membranes, and foams. The apparatus allows measurement of these properties in the very small quantities typical of material development programs, where the largest sample available may only be 1-10cm2 in area. The convection/diffusion test method is useful for determining the gas flow resistance property and water vapor diffusion properties from a single experimental run. This eliminates the need for two separate tests, which is the usual procedure. The apparatus may also be used to perform separate tests for the diffusion property or the air permeability property, which may have some advantages when materials exhibit strongly concentration-dependent transport properties. The convection/diffusion test method is well-suited for rapid screening and comparison of the properties of a large number of materials with widely-varying transport properties.

Introduction Vapor transfer through clothing systems may occur due to diffusion (driven by vapor concentration gradients), and convection (driven by air pressure differences). Convective heat and mass transfer in porous media such as textiles is often more important than transport due to diffusion, especially if such materials are used in conditions where a large pressure gradient is present. Laboratory test methods for textiles usually concentrate on one transport mechanism, to the exclusion of the others. Diffusion test methods are particularly easy to perform, and often become the primary ranking and evaluation method for determining the transport properties of textiles. Such test methods can be very misleading for textiles, particularly those which have high air permeability, since a very small pressure gradient can produce large convective flows through the porous structure, far outweighing any diffusive transport which takes place. The usual procedure is to determine the water vapor diffusion properties and the air permeability properties separately. For textiles, water vapor diffusion test methods include the ASTM test method for water vapor transmission of materials (E 96), and the ISO test method for measurement of thermal and water vapour resistance under steady-state conditions (ISO 11092). Air permeability properties may be determined by a textile test method, ASTM D737-75, standard test method for air permeability of textile fabrics. This method, however, is quite limited in that it is more useful for quality-control # US Government.

testing due to its prescription for testing at a single pressure differential (124.5 Pa). A more accurate test method is ASTM F778-88, standard methods for gas flow resistance of filtration media. All the methods mentioned above are time-consuming, require large amounts of material, and are not capable of a very wide range of test conditions. They also require two separate kinds of tests to be run to characterize the potential of a given material to transport water vapor through its structure: a water vapor diffusion test and an air permeability test. It would be very appealing to have a test method available that can determine the diffusion and convection properties from the same test, and to be able to directly compare the results obtained between materials as different as airimpermeable membrane laminates, very air-permeable knitted fabrics, woven fabrics, and complicated nonwoven and polymeric foam structures. It would also be appealing to have this test method able to measure these properties in very small quantities typical of material development programs, where the largest sample available may only be 1-10cm2 in area. The method described here, based on the dynamic moisture permeation cell (DMPC) (Gibson et al., 1995a, 1995b) satisfies the need for a quick, automated method that can test the mass transport properties of very small pieces of woven and nonwoven fabrics, membranes, and foams. The focus of this paper is to describe the use of this test method to obtain the water vapor diffusion property and the air permeability property from a single test. The DMPC may also be used to perform separate tests for the diffusion property or the air permeability property, which may have some advantages when materials exhibit strongly concentration-dependent transport properties. Experimental method A schematic of the DMPC test arrangement is shown in Figure 1. Nitrogen streams consisting of a mixture of dry nitrogen and watersaturated nitrogen are passed over the top and bottom surfaces of the sample. The relative humidity of these streams is varied by controlling the proportion of the saturated and the dry components. By knowing the temperature and water vapor concentration of the entering nitrogen flows, and by measuring the temperature, water vapor concentration, and flow rates of the nitrogen flows leaving the cell, one may measure the fluxes of gas and water vapor transported through the test sample. With no pressure difference across the sample, transport of water vapor proceeds by pure diffusion, driven by vapor concentration differences. If a pressure difference across the sample is present, transport of vapor and gas includes convective transport, where the gas flow through the sample carries water vapor with it, which may add to or subtract from the diffusive flux, depending on the direction of the convective gas flow.

Convection/ diffusion test method 97

IJCST 12,2

98

Figure 1. Schematic of DMPC test arrangement

Review of water vapor diffusion test method The use of the DMPC for determining water vapor diffusion properties will be reviewed first, followed by a description of the diffusion/convection test method. The following equations for calculating water vapor flux apply to either the top or bottom flows in the cell. Strictly speaking, only one measurement on one side of the cell is necessary; the use of two separate humidity transducers for the top and bottom flows allows two measurements of water vapor flux to be made at the same time, using the equations given below for either the top or bottom flow, as appropriate. Further details on this test method are available (Gibson et al., 1995a, 1995b, 1997a). For this type of test, the mass flow rate of water vapor diffusing through the test sample from one side of the cell to the other is given by: _ QC QC2 ÿ C1 m 1 A A A _ m

= mass flux of water vapor across the sample [kg/s]

A

= area of test sample [m2]

Q

= volumetric flow rate through top or bottom portion of the cell [m3/s]

C

= C2 ± C1, water vapor concentration difference between incoming stream (C1) and outgoing stream (C2) in top or bottom portion of the moisture permeation cell [kg/m3]

The incoming water vapor concentration is determined by the ratio of the mass flows of the saturated and the dry nitrogen streams. The mass flow rates are controlled by MKS model 1259C mass flow controllers, with a Model 247C 4Channel Readout (MKS Instruments, Inc.). These mass flow controllers can control mass flow rate at an accuracy of 0.8 per cent of full scale, with a response time of less than two seconds. At constant mass flow, the true volumetric flow rate will vary with temperature; the flow rate set by the MKS controllers is indicated in terms of volumetric flow rates at standard conditions of 08C and atmospheric pressure (1.01325 105Pa) The actual volumetric flow rate at different temperatures may be found from the mass flow rate, the temperature, and the pressure of the actual flow. For water vapor diffusion, the critical measurement is the outgoing flow water vapor concentration C2, which can be measured in a variety of ways. In the work reported here, capacitance-type relative humidity probes (Vaisala HMI 32 or 38) with Type HMP 35 or 37 sensors were used (Vaisala Inc.), which are adequate for materials which have significant vapor flux across them. The advantage of these probes is that they have a relatively fast response time (5-30 seconds: response time slower at higher humidities), which is useful for transient studies. The probes are listed by the manufacturer as having an accuracy of 1 per cent from 0-90 per cent relative humidity, and 2 per cent from 90-100 per cent relative humidity. The measurement accuracy of these probes may be improved to 0.5 per cent by determining a calibration curve in situ. This is done by placing an impermeable aluminum foil sample in the cell and varying the relative humidity of the gas flow in the top and bottom of the cell by means of the flow controllers. The resulting curves (at increments of 10 per cent r.h.) of measured relative humidity versus true relative humidity (set by the flow controllers) are used as calibration factors to correct the measured relative humidity for subsequent tests. Sorption hysteresis of the hygroscopic polymer used in the capacitance probe make any further improvements in probe accuracy difficult. For test materials which have small vapor fluxes, requiring measurements at very low concentrations, an 1100DP Dew Point Hygrometer (General Eastern Instruments, Inc.) may be used. For the highest accuracy, an M200 Gas Chromatograph (MTI Analytical Instruments, Inc.) has also been used as the concentration measurement device, but this is much less convenient in the practical sense of a routine test. To obtain the water vapor concentration in the outgoing air stream, one must be able to convert from the known values of relative humidity and temperature to water vapor concentration. The vapor pressure of saturated water vapor in air is obtained from an empirical formula (or tables) as a function of temperature, and then converted to concentration using the perfect gas law.

Convection/ diffusion test method 99

IJCST 12,2

100

We may express the water vapor transmission rate in terms of the indicated volumetric flow rate at standard conditions, the humidity difference, and the temperature: _ Qs ps Mw m 2 ARTs A Mw = molecular weight of water vapor [18.015kg/kmole] Qs

= volume flow rate at standard conditions of 08C and atmospheric pressure [m3/s]

R

= universal gas constant [8314.5N-m/kg-8K]

Ts

= reference temperature at standard conditions of 08C in degrees K (273.15K)

ps

= saturation vapor pressure of water [Pa]

= 2 ± 1, relative humidity difference between incoming stream (1) and outgoing stream (2) in top or bottom portion of the moisture permeation cell

= pv/ps, relative humidity

pv

= vapor pressure of water [Pa]

For the present test apparatus, various sample holders are available, which have different test sample measurement areas, and which have different downstream locations from the flow inlet. All test results given in this study used a sample measurement area of 1.0 10±3m2, and the sample was located equidistant from the inlet and outlet ports of the cell. The typical volumetric flow rate used was 3.33 10±5m3/s (2,000cm3/min). The dimensions of the DMPC were chosen to assure flow velocities of at least 0.5m/s over the sample to minimize the contribution of boundary air layer resistances to the test measurements. Details of the sample holder are shown in Figure 2. The sample sizes are kept quite small to make it possible to evaluate novel membranes and laminates, which are often produced in quantities too small for testing by some of the standard water vapor diffusion test methods. The small sample area makes it necessary to test at different locations across a typical roll of fabric to adequately characterize a given material. Sample mounting methods vary according to the material being tested. Thin materials, such a laminated materials and woven cloth, were originally tested with rubber sealing gaskets to prevent leakage, but the sealing proved to be unnecessary for most materials; the clamping force provided by the mounting bolts has proven to be sufficient to prevent any leakage. Thick materials which are highly permeable require special sealing methods such as edge sealing by molten wax, or the use of a curable sealant. The testing of thicker materials also requires a larger sample area to minimize factors such as edge effects.

Convection/ diffusion test method 101

Figure 2. Schematic and dimensions of the sample holder for the DMPC

Diffusion test procedure The actual test is conducted under the control of a personal computer (PC) connected to the flow controllers, automated valves, and the various measurement transducers through input and output boards (see Figure 1). Various options exist within the software for operator input setpoint information, or preset files containing the setpoint information. The computer applies the proper setpoint voltage to each controller to produce the desired relative humidity in the upper and lower gas streams entering the DMPC. The A/D board in the PC reads analog voltage outputs of the relative humidity, RTD, thermocouples, differential pressure transducer, mass flow meters, etc., records the data on disk, calculates parameters of interest, and plots results to the PC screen. The software applies operator-determined equilibration criteria to determine when equilibration has been reached for that setpoint. Once equilibration is reached, the results (humidity, calculated flux, etc.) may be output to a printer and to a data file on disk. The computer then proceeds to the next setpoint and repeats the process. The pressure drop across the sample is monitored by means of an MKS Baratron Type 398 differential pressure transducer, with a Type 270B signal conditioner (MKS Instruments, Inc.). For measurement of pure diffusion, especially for materials such as fabrics, which may be quite permeable to

IJCST 12,2

102

convective flows, it is important to make sure that the pressure drop across the sample is zero, so that transport takes place only by pure diffusion. The pressure drop is continuously monitored and displayed, and is controlled by means of two automated valves at the outlets of the cell. For the permeable fabrics, this system also allows one to do testing under controlled conditions of a defined pressure drop across the sample, so that transport takes place by both diffusion and convection (which will be described later). This makes it possible to determine an air permeability value from the apparatus, in addition to the water vapor diffusion properties of the test sample. Materials which have a constant mass transfer coefficient show a linear slope on plots of flux versus concentration difference across the sample. These materials do not change their transport properties as a function of water content or test conditions. For materials which do not have a constant slope, the data points for a test series will not superimpose, but will form a set of curves for each test condition. From an evaluation of the flux versus concentration difference curve at various points we can calculate values for the material diffusion resistance, which will be a function of the concentration of water in the material. We define a total resistance to mass transfer as the simple addition of an intrinsic diffusion resistance due to the sample (Ri ) and the diffusion resistance of the boundary air layers (Rbl): _ C m 3 A Ri Rbl 2

3

7 6 C 7 ÿ Rbl Ri 6 4 m _ 5 A _ m

= mass flux of water vapor across the sample (kg/s)

A

= area of test sample (m2)

4

= log mean concentration difference between top and bottom gas C streams (kg/m3) Ri

= intrinsic diffusion resistance of sample (s/m)

Rbl

= diffusion resistance of boundary air layers (s/m)

The log mean concentration difference across the sample is appropriate since there is a significant change in the concentration of the gas stream both below and above the sample. In addition, the gas streams may not necessarily be in parallel unidirectional (cocurrent) flow, but may be run in counter flow to

maintain a more constant concentration gradient across the sample. The log mean concentration difference (Geankopolis, 1972) is defined as: C

Ca ÿ Cb lnCa =Cb

5

Ca = concentration difference between the two gas streams at one end of the flow cell (kg/m3)

Convection/ diffusion test method 103

Cb = concentration difference between the two gas streams at other end of the flow cell (kg/m3) For parallel cocurrent flow, the concentration differences are between the top and bottom incoming flow at one end of the cell (Ca), and the difference between the top and bottom outgoing flows at the other end of the cell (Cb). For countercurrent flow, the concentration differences are between the incoming and outgoing flows at one end of the cell (Ca), and the incoming and outgoing flows at the other end of the cell (Cb). Use of DMPC for convection/diffusion studies The DMPC may also be run with a specified pressure drop across the sample so that transport takes place by both diffusion (driven by concentration differences) and convection (driven by gas phase pressure differences). The simplest experiment to run is shown in Figure 3. Gas enters the DMPC at a relative humidity of 0.90 (90 percent r.h.) on the top portion of the cell, and 0.0 (0 percent r.h.) on the bottom of the cell. The automated valves are used to restrict the flow on one or the other sides of the cell. This causes the pressure in one side of the cell to be higher than in the other, causing convective flow across the sample, in addition to the diffusion flux taking place due to the concentration gradients. Measurements are taken as a function of pressure drop across the sample, where the convective flow and pressure drop are gradually increased in stepwise increments. In addition to the pressure drop, it is useful to have an actual measurement of gas flow through the sample. An electronic mass flow

Figure 3. Convection/diffusion experiment in the DMPC: example shows bottom outlet flow restricted to force convective flow across sample, which opposes diffusive flux of vapor

IJCST 12,2

104

meter (Model 822 Top-Trak, Sierra Instruments, Inc.), connected to the lower outlet of the cell as shown in Figure 1, is used to record the mass flow rate of gas through the test material. In this paper, the diffusion/convection test method will be demonstrated for three typical materials: (1) A microporous polytetrafluorethylene (PTFE) membrane, with low water vapor diffusion resistance, but a high resistance to convective gas flow (low air permeability). (2) A knit polyester fabric with a slightly greater resistance to water vapor diffusion, but very low air flow resistance (high air permeability). (3) A woven nylon fabric with still higher water vapor diffusion resistance, and an air flow resistance intermediate between the other two materials. Further diffusion/convection test results are available for a wide variety of woven fabrics, nonwoven filter materials, clothing insulation layers, novel electrospun nanofiber membranes, and military chemical protective clothing systems (Gibson et al., 1998). In addition, correlations and comparisons are available for results obtained with the DMPC and standard water vapor diffusion tests (Gibson et al., 1995a, 1995b, 1997a) or standard air permeability tests (Gibson et al., 1998). Typical measurements of water vapor diffusion resistance, and mass flow rate as a function of pressure drop, are shown in Figures 4 and 5 for the three materials. Figure 4, which shows flow rate as a function of pressure drop, is directly related to the air permeability of the material. The greater the slope of the line for a given material, the greater the air permeability. Figure 4 shows that the

Figure 4. Flow rate through fabric as a function of pressure drop

Convection/ diffusion test method 105

Figure 5. Diffusion resistance as a function of pressure drop

microporous PTFE membrane has a very low air permeability (high flow resistance), the knit polyester fabric has a high air permeability, and the woven nylon fabric is intermediate between the other two materials. Calculation of the air flow resistance from the data in Figure 4 will be discussed later. Figure 5 shows the apparent water vapor diffusion resistance as a function of pressure drop. This plot illustrates the interaction of convective and diffusive transport. The intersection of each material's curve with the P = 0 point on Figure 5 defines the true water vapor diffusion value for that material. This is illustrated with an expanded scale in Figure 6. At the condition of 0 pressure drop, the PTFE membrane has the lowest diffusion resistance, followed by the knit polyester, with the nylon fabric having the highest diffusion resistance. This PTFE membrane has previously been shown to have a diffusion resistance of about 6-8 s/m (Gibson et al., 1995a). This means that the boundary layer resistances in this flow cell (defined by flow rates and flow geometry) are approximately 115 s/m. Thus the true diffusion resistance of each material is equal to the difference between its total resistance from Figure 6, and the boundary layer resistance. The resulting intrinsic diffusion resistances are 6 s/m for the PTFE membrane, 96 s/m for the nylon fabric, and 36 s/m for the polyester fabric. These values agree well with those obtained previously for these materials (Gibson, 1996). However, since these materials differ greatly in their air permeability properties, the change in apparent diffusion resistance as the pressure drop increases is quite different for the various materials. The PTFE membrane has a nearly constant diffusion resistance, due to its low air permeability. Because

IJCST 12,2

106

Figure 6. Use of diffusion/ convection data to define true diffusion resistance

of the polyester fabric's high air permeability, even at very low pressure drops (for example at 10 Pa), its apparent diffusion resistance is less than that of the PTFE membrane. The ability to conduct testing over a range of pressure drops increases the accuracy of the water vapor diffusion value. It is clear that slight variations in pressure drop across a sample with high air permeability will greatly influence the measured water vapor diffusion resistance. Diffusion test methods which do not control or monitor the convective flow through the sample are prone to measurement and ranking errors caused by air flow through the sample. The characteristic curve shown in Figure 6, which illustrates the interaction between diffusive and convective transport, is more realistic in terms of the transport processes taking place in clothing systems, where both modes of transport take place at the same time. Water vapor diffusion properties alone, which may rank materials such as the PTFE membrane as superior to more airpermeable materials, can be very misleading when ranking candidate materials for comfortable or breathable clothing systems. Calculating air flow resistance There are many definitions of the permeability or the flow resistance; most often the permeability is given by Darcy's Law (Dullien, 1979) such that: v v

ÿkD p x

= apparent gas flow velocity (m/s)

6

kD

= permeability constant (m2) ±6

= gas viscosity (17.85 10 kg/m-s for N2 @ 208C)

p = pressure drop across sample (N/m2 or Pa) x = thickness (m) For low velocity flows, where the apparent Reynolds number (based on nominal particle diameter or pore sizes) is much less than ten, a plot of pressure drop versus volumetric flow rate or velocity will give a constant value for the permeability constant kD. At higher flow rates, where inertial effects begin to compete with viscous flow effects, pressure drop ± flow rate plots will begin to deviate from linearity, and inertial effects need to be considered. Previous work on air penetration through clothing systems has shown that air pressure differences across textile layers, due to factors such as wind or body movement, are usually less than 100 Pa (Kind et al., 1991; Take-uchi, 1989; Stuart and Denby, 1987; Fedele et al., 1986). For the testing presented in this paper, flow rates and pressure drops are low enough so that inertial effects are not readily apparent in the experimental results. The DMPC operated in the diffusion/convection mode provides plots of pressure drop versus either mass flow rate or volumetric flow rate, seen previously in Figure 4. Volumetric flow rate is the most convenient to use, so the permeability constant may be found from: Q x 7 kD A p

= gas viscosity (17.85 10±6kg/m-s for N2 at 208C)

Q

= total volumetric flow rate (m3/s)

A

= apparent sample flow area (1.0 10±3m2 for DMPC sample holder)

x = thickness (m) p = pressure drop across sample (N/m2 or Pa) For textiles, although thickness measurements seem simple, they are often problematic, and can be a large source of error if they are incorporated into reported measurements of Darcy permeability. It is preferable to present the pressure-drop/flow rate results in terms of an apparent flow resistance defined as: Ap 8 RD total RD

= apparent Darcy flow resistance (m±1)

Convection/ diffusion test method 107

IJCST 12,2

108

The volumetric flow rate shown in Figure 4, measured by the electronic mass flow meter, is the equivalent volumetric flow rate at the reference temperature (T0) of 08C (273.15 K), and reference atmospheric pressure (p0) of 1.01325 105 Pa. The actual volumetric flow rate at a given temperature may be found from the mass flow rate indicated by the electronic mass flow meter (Q0), the ambient temperature (Ta), and the ambient pressure of the actual flow (pa). The pressure correction is negligible (p0/pa 1), so only the temperature correction needs to be made. The correction to obtain the actual volumetric flow rate (Qtotal) from the indicated mass flow rate (Q0) is: Qtotal Q0 Ta =T0 p0 =pa Q0 Ta =T0

9

Note that the quantity (p/Qtotal) is equal to the inverse of the slopes of the curves shown in Figure 4 (after correcting to the proper temperature as shown above). Figure 4 may be replotted as shown below in Figure 7. Thus the air flow resistance can be found from the slope of each curve in Figure 7, using the known flow area and gas viscosity according to equation (8). For the three materials shown in Figure 7, the equivalent air flow resistance is 2.78 107m±1 for the polyester fabric, 1.14 108m±1 for the nylon fabric, and 4.25 109m±1 for the PTFE microporous membrane. These values agree with those obtained previously for these materials by another method (Gibson et al., 1997b).

Figure 7. Pressure drop across sample as function of flow rate

If the material thickness is known, the Darcy permeability (usually reported in units of m2) can be found from the apparent flow resistance as: kD

x RD

10

Complicating factors The convection/diffusion test method is most appropriate for air permeable materials such as porous textiles, membranes, or foams. It can be used for airimpermeable materials, and this is useful to show the relative importance of convective flow versus diffusive flow. The convection/diffusion test method is quite convenient for screening a large number of samples ± particularly developmental materials, since it gives an air permeability and a water vapor diffusion value from a single test. However, for materials such as semipermeable membrane laminates, or porous textiles which have humidity-dependent air permeability, the single air flow resistance and single diffusion resistance number obtained from a convection/diffusion test can be misleading, and it would be preferable to perform a separate diffusion test, and a separate air permeability test to separate out these two effects. Both of these complicating factors are discussed below. Concentration-dependent diffusion in polymer membrane laminates Vapor transport across nonporous hygroscopic polymer membranes and films is often highly dependent on the amount of water present in the polymer. Many commercially available semipermeable membrane laminates such as Gore-Tex and Sympatex exhibit this concentration-dependent behavior to varying degrees. The DMPC, when operated in the pure vapor diffusion mode (no pressure drop across the sample) is capable of showing this concentrationdependent transport behavior (Gibson et al., 1995a, 1995b, 1996, 1997a). For samples which are air-impermeable, convection/diffusion testing does not provide any extra information, and it is better practice to conduct pure diffusion testing which evaluates water vapor transport under a variety of environmental conditions corresponding to different levels of water content in the hygroscopic polymer membrane or membrane laminate. This concentration-dependent behavior is illustrated in Figure 8. Two semipermeable membrane laminates (Gore-Tex and Sympatex) are shown which exhibit concentration-dependent transport behavior. They may be compared to the microporous PTFE membrane, also shown in Figure 8, which does not show the same type of concentration dependence. Further information on concentration-dependent transport is available (Gibson et al., 1995a, 1995b, 1996, 1997a).

Convection/ diffusion test method 109

IJCST 12,2

110

Figure 8. Concentrationdependent water vapor transport behavior of two semipermeable membrane laminates

If one were to blindly apply the convection/diffusion test method to these kinds of materials, one would obtain a diffusion resistance that is only representative of one mean relative humidity, which would not capture the true variability of the transport behavior of these particular materials. Humidity-dependent air permeability The convection/diffusion test method results in a single number for the air flow resistance of a porous material. Porous hygroscopic materials often exhibit humidity-dependent air permeability due to the swelling of the solid matrix as it takes up water vapor from the environment. These effects are most evident in materials such as tightly-woven fabrics, low porosity hygroscopic membranes, and nonwoven fiber mats. Humidity-dependent air permeability is usually evident from the volumetric flow rate versus pressure drop plot. The plot will no longer have a line of constant slope, but will show some curvature according to the relative humidity of the gas flowing through the sample. This is illustrated in Figure 9 for a tightly-woven cotton fabric. For a material such as this, the DMPC can be used to conduct a more traditional air flow resistance test as a function of relative humidity (Gibson et al., 1997b). The corresponding humidity-dependent air flow resistance curve for the tightly woven cotton fabric shown in Figure 9 is given in Figure 10. It is possible to account for the non-linearity of the convection/diffusion test due to this humidity dependence (Gibson et al., 1997c), but it is usually simpler to perform the separate air flow resistance test as a function of relative humidity (Gibson et al., 1997b) to produce plots similar to that shown in Figure 10.

Convection/ diffusion test method 111

Figure 9. Convection/diffusion test exhibiting humiditydependent air permeability

Figure 10. Humidity-dependent air flow resistance of tightly woven cotton fabric

Another minor complicating factor is due to the zero drift of the mass flow sensor. This is usually not a problem since it is the slope of the pressure drop/ flow rate curve which is used to calculate the air flow resistance. A more fundamental factor is the changing composition of the convective gas flow through the sample. The mass flow meter is calibrated to give mass flow rates for pure dry nitrogen. As the nitrogen stream's humidity changes, the indicated

IJCST 12,2

112

mass flow rate is affected by the lower density of the humidified gas stream. In addition, the mass flow meter uses a thermal sensor, which is affected by the changes in heat capacity and thermal conductivity of the humidified gas stream. These effects can be accounted for analytically, since the relative humidity of the gas stream is being measured, but the increased accuracy is usually within the inherent variability of the test anyway, so these minor effects are ignored. Conclusions The convection/diffusion test method outlined in this paper is useful for determining the gas flow resistance property (air permeability) and water vapor diffusion resistance property (water vapor transmission rate), from a single experimental run. This eliminates the need for two separate tests, which is the usual procedure. The convection/diffusion test method is ideal for rapid screening and comparison of the properties of a large number of materials. Because the required test sample size is much less than that required for most other standard types of air permeability and water vapor diffusion test methods, the convection/diffusion test method is particularly well-suited for material development efforts aimed at developing porous woven and nonwoven textiles, coating processes for porous substrates, polymeric foams, and microporous polymer membranes and laminates. References Dullien, F. (1979), Porous Media ± Fluid Transport and Pore Structure, Academic Press, New York, NY, p. 157. Fedele, P., Bergman, W., McCallen, R. and Sutton, S. (1986), ``Hydrodynamically induced aerosol transport through clothing'', Proceedings of the 1986 Army Science Conference, Vol. I, pp. 279-93. Geankopolis, G. (1972), Mass Transport Phenomena, Holt, Rinehart, and Winston, Inc., New York, NY, pp. 277-8. Gibson, P.W. (1996), ``Multiphase heat and mass transfer through hygroscopic porous media with applications to clothing materials'', US Army Natick Research, Development, and Engineering Center Technical Report, Natick/TR-97/005. Gibson, P.W. and Charmchi, M. (1997c), ``Modeling convection/diffusion in porous textiles with inclusion of humidity-dependent air permeability'', International Communications in Heat and Mass Transfer, Vol. 24 No. 5, pp. 709-24. Gibson, P.W., Rivin, D. and Kendrick, C. (1998), ``Convection/diffusion test method for porous materials using the dynamic moisture permeation cell'', US Army Natick Research, Development, and Engineering Center Technical Report, Natick/TR-98/014. Gibson, P.W., Kendrick, C., Rivin, D. and Charmchi, M. (1997a), ``An automated dynamic water vapor permeation test method'', in Stull, J.O. and Schwope, A.D. (Eds), Performance of Protective Clothing, American Society for Testing and Materials (ASTM) Special Technical Publication (STP) 1273, American Society for Testing and Materials, Vol. 6, pp. 93-107. Gibson, P.W., Kendrick, C., Rivin, D., Charmchi, M. and Sicuranza, L. (1995a), ``An automated water vapor diffusion test method for fabrics, laminates, and films'', Journal of Coated Fabrics, Vol. 24, pp. 323-45.

Gibson, P.W., Kendrick, C., Rivin, D., Charmchi, M. and Sicuranza, L. (1995b), ``An automated dynamic water vapor permeation test method'', US Army Natick Research, Development, and Engineering Center Technical Report, Natick/TR-95/032. Gibson, P.W., Elsaiid, A.E., Kendrick, C.E., Rivin, D. and Charmchi, M. (1997b), ``A test method to determine the relative humidity dependence of the air permeability of textile materials,'' Journal of Testing and Evaluation, Vol. 25 No. 4, pp. 416-23. Kind, J., Jenkins, J. and Seddigh, F. (1991), ``Experimental investigation of heat transfer through wind-permeable clothing'', Cold Regions Science and Technology, Vol. 20, pp. 39-49. Stuart, I. and Denby, E. (1987), ``Wind induced transfer of water vapor and heat through clothing'', Textile Research Journal, Vol. 57, pp. 247-56. Take-uchi, M. (1989), ``Analysis of wind effect on the thermal resistance of clothing with the aids of Darcy's law and heat transfer equation'', Sen-i Gakkaishi 39, No. 3, pp. 39-48. Note US Department of Defense Technical Reports are available on order from: National Technical Information Service (NTIS), 5285 Port Royal Road, Springfield, VA 22161-0111; http:// www.dtic.mil/

Convection/ diffusion test method 113

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

114 Received 5 March 1998 Revised 2 October 1999 Accepted 2 October 1999

Grading the crease recovery with twisting of fabric by using image identification technique Xiaohong Wang

China Textile University, ShangHai City, People's Republic of China, and

Mu Yao

Northwest Institute of Textile Science and Technology, Xian City, People's Republic of China Keywords Twisting, Fabric, Image processing Abstract Traditionally, the grade of fabric's crease recovery with twisting is decided by comparing the processed sample fabric with standard sample photograph under some conditions. This method is completely reliant on subjective appraisal, so it is easy to lead to some subjective error and affect the conclusion. In this paper, we use image processing and texture analysis technique to calculate a few of the parameters which describe the fabric's crease recovery properties. At the same time, we use the Fuzzy priority similarity comparison method to assess the fabric's crease recovery properties synthetically.

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 114-123. # MCB University Press, 0955-6222

Introduction Generally speaking, the crease recovery of fabric, sometimes called ``wash and wear'' characteristic refers to the ability of the fabric to return to its original shape because of the act of the fast and slow elasticity after removing the force which causing the crease even without ironing after washing and hydroextracting[1]. Many researchers have studied the testing instrument, testing method and testing parameters of evaluating the fabric's crease recovery at home and abroad. At present, according to the works and the publications, there are three method of assessing the crease recovery: (1) Measuring the fast and slow elastic angle of crease recovery under dry and wet conditions[2,3]. The concrete step is, first, folding the fabric with specified shape and dimension with some instrument (for example, the YG541 fabric crease elasticity tester) and keeping the fabric under the specified loading (10N or 30N) for some time (5min or 30min), then removing the loading, measuring the fast and slow elastic angle of crease recovery after the fabric recovering for some time (15S or 30min); the fast and slow elastic angle of crease recovery are used to express the crease recovery of the fabric directly. The testing method is simple, clear and direct, but there are two shortcomings: first, there is a difference between the practice and the experiment on the loading; second, the angle of the crease recovery is the function of time, especially the fast elastic angle of crease recovery, so any slipping up would affect the correctness of the experiment.

(2) Comparing the surface crease of the fabric with the standard samples or with the photographs of the standard samples[1]. This method is comparing the surface crease of the fabric with the standard samples or with photographs of the standard samples after it is washed and hydroextracted under certain conditions. Many countries have this method as their national standard testing method, for example, China. There are three washing ways: they are wringing, soaking and washing machine. These three washing ways cannot comprehensively and really simulate the practice; at the same time, because the crease recovery of the fabric is evaluated by evaluator, so the evaluating result is thoroughly reliant on the subjective assessment of the evaluator. The evaluator's experience, physiology, psychology and external environment may all affect his drawing a correct conclusion. (3) Using the theory of ``the fuzzy closeness of the best comparison object'' to assess the fabric's crease recovery[4]. This method is the first choice as the best comparison object by many experts, then measuring its slow elastic angle of crease recovery , activity coefficient LP, bending rigidity SB, compressive elasticity RE and stable thickness TS, and use the equation ÿ n X 1 ÿ i0 ÿ i 2 Wi f e i0 i 1g 2 i1 to calculate the closeness. Here Wi is the membership function of each mechanic parameter which affect the fabric's crease recovery, n is the number of the mechanic parameter, i0, i0 are the values of the average and standard error of the ith mechanic parameter of the best comparison object, i, i are the values of the average and standard error of the ith mechanic parameter of the fabric which is to be assessed. From the knowledge of the fuzzy theory, we know that the value of each W is decided by a human being. Because each one has his own view about the importance of the five mechanic parameters, so it is very difficult to decide the exact value of W; furthermore, this method could not assesses the fabric's crease recovery qualitatively. So, in this paper, an objective method is proposed to eliminate the influence of subjective factor. First, calculate a few of parameters which describe the fabric's crease recovery properties by using image processing and texture analysis, then grading the sample's crease recovery by fuzzy priority similarity comparison method. Experimental preparation According to the China Nation Standard testing method, first, cut a piece of fabric into a specified size of sample, immerse it into the solution under the specified temperature, twisting the sample with fixed tension after some time, then put it on a piece of glass. When dry, the sample is photographed under the

Grading the crease recovery

115

IJCST 12,2

116

same conditions as in the standard samples. The photographs were scanned into computer by QuigHua ZiGuang HWF-95FAC4S scanner and the photographs were converted into digital image F(x,y), which can be expressed as matrix F(x,y): 2 3 f 0; 0 f 0; 1 f 0; N ÿ 1 6 f 1; 0 f 1; 1 f 1; N ÿ 1 7 7 1 Fx; y 6 4 5 f M ÿ 1; 0 f M ÿ 1; 1 f M ÿ 1; N ÿ 1 In matrix F, the value of each element is the gray level corresponding to the element of sample fabric. It reflects the unevenness of the corresponding element of sample fabric. The characteristic of some simple image element (pixel) could not be used to describe the fabric's easy care characteristic, texture statistical analysis technique can do this job. Spatial co-occurrence function and texture fineness[8] Because different fabrics have different smoothness or uncreasing characteristics, they may present different uneven strips on the surface of fabric after releasing from twisting. The surface of the fabric which has good crease recovery properties is smooth and has few stripes, the wave crest of the stripes are not high. On the contrary, on the surface of fabric whose crease recovery is bad, there are many stripes and they are chaos, the wave crests are high and uneven. Spatial co-occurrence function and texture fineness can describe the situation of the stripes on the surface of the fabric. (1) Spatial co-occurrence function A iP w

A ij

jw P

m1ÿw njÿw iP w

f m; nf m ÿ ; n ÿ jw P

miÿw njÿw

2 f 2 m; n

where i, j are the center place of the filter window, w is the dimension of the filter window, ; are the moving extent of the filter window along the direction of i and j. (2) Texture fineness ± if ; = 1, 2, T (the largest moving extent of the filter window), the texture fineness can be defined as: t

T X

T X

ÿT ÿT

2 2 A ij

if the texture is more coarse, the value of t is bigger.

3

First-order statistical analysis technique[5,10] Laws had done a great deal work on the first-order statistical analysis technique, especially on the measurement of texture energy[9]. First-order statistical analysis technique is defined as using the distribution of gray level or other properties of the single image element (pixel) and its neighbors to measure the texture. This technique is often used in texture analysis because of its simplicity and correctness. Laws has done many jobs on this area and he also proposed measuring the energy of texture as one of the parameters. There are other first-order parameters which describe the texture, they are: (1) Variance 2 2

Lÿ1 X l ÿ l2 probl

4

l0

(2) Energy E Lÿ1 X

E

probl2

5

l0

(3) Entropy H H ÿ

Lÿ1 X l0

probl log2 probl

6

Lÿ1 X l ÿ l3 probl

7

(4) Skewness S S 1=3

l0

where l is the gray level of each pixel, l is the average gray level of image, prob(l) is the probability of gray level l, and L is the maximum gray level. Second-order statistical analysis techniques[5,10] Generally, second-order statistical analysis technique is based on the estimate of the conditioned probability density function of the second-order of a pair of image element. This method has a long research history, it is one of a group of generally acknowledged important texture analysis methods at present. In the texture image, the statistical regular pattern of a pair of the image elements which have certain distance on some direction could display the texture characteristic of the image concretely. This statistical regular pattern can be described quantitatively by the parameters which are derived from the gray level commensal matrix.

Grading the crease recovery

117

IJCST 12,2

118

The gray level commensal matrix describes the occurrence probability of a pair of image elements at a distance of d that has the gray level i and j on the direction , the element of the matrix can be denoted as p(i, j|d, ); when and d are decided, it can also be denoted as pij. In order to simplify the calculation, we select that is 0 and d is 1. Weszka, Vicker, Unser and other researchers[11] have done a great deal of research work and verified that short distance d could also get the best analysis result. When calculating the parameters of fabric texture, gray level commensal matrix K(L, L) is defined: when a pixel f(i, j)'s gray level is l1(l1 = 0, 1, 2, . . . L ± 1), the probability of f (i + , j + ), whose gray level is l2, can be expressed by gray level commensal matrix. Here l1 and l2 present the row and column of matrix K. Based on gray level commensal matrix, four parameters which are used in texture analysis can be calculated, they are: (1) Texture consistency Q1 Q1

XX l1

(2) Texture contrast Q2 Q2

XX l1 ÿ l2 pl1 ; l2 l1

(3) Texture entropy Q3 Q3

pl1 ; l2

9

l2

XX l1

8

l2

pl1 ; l2 logpl1 ; l2

10

l2

(4) Texture gray occurrence Q4 XX Q4 f l1 ÿ l2 ÿ 1 2 g1 2 l1

11

l2

where p(l1, l2) is normalized gray matrix, each element of p(l1, l2) is acquired by dividing each element of matrix K by the sum of all elements of matrix K, and X X l1 pl1 ; l2 1 l1

2

l2

X X l2 pl1 ; l2 l2

l1

21 22

X

l1 ÿ 1

l1

X l2

X

Grading the crease recovery

pl1 ; l2

l2

l2 ÿ 2

X

pl1 ; l2

l1

119

where l is the gray level of each pixel, and l1, l2 = 0, 1, 2, . . . L ± 1. Fuzzy priority similarity comparison method[7] In practical assessment, the calculating result of parameters are not completely equal to the values of one of the standard grade parameters, they are among these parameters. At the same time, because the process of the assessment relies on the vision and the subjective judgement of human being, there must exist some fuzzy factors during it. In order to simulate the thought process of a human being and eliminate the effect of the subjective factor, there needs to be a further mathematics process if we want to get a correct result. So, in this paper, fuzzy priority similarity comparison method is used to grade the fabric's crease recovery. What is called fuzzy priority comparison method is to select the most similar specified sample from the established set of arguments to the sample. Here, the specified samples are the five grade photographs of the standard fabric which are stated in the national standard of China. The concrete way is: 1. Establish the fuzzy priority comparison matrix First, the five grade fabrics and the selected parameters which are used to analyze the texture form the set U: U fu1 ; u2 ; ; um g

12

where m is the number of specified sample, i is the value of corresponding parameters. Then according to formula (13), to every parameter, comparing the values of each pair of parameters ui, uj of the set U with the values of the sample which is to be assessed uk, the priority similarity comparison rij: rij

jX ÿ Xj j jX ÿ Xi jX ÿ Xj j

13

Here rij indicates the priority comparison of the sample to be assessed uk to the specified samples and uj in other words, that is when comparing the sample to be assessed uk with the samples ui and uj, rij indicates the similarity extent of the sample to be assessed uk to the specified samples ui and uj, rij meets the following conditions (I) rii = 1, rij 2 [0, 1], i, j = 1, 2, . . . 5; (II) rij + rji = 1, i, j = 1, 2, . . . 5, i 6 j.

IJCST 12,2

120

X is the value of corresponding parameter of the sample to be graded, and Xi, Xj are the values of corresponding parameter of the specified sample. So, the priority similarity comparison rij (rij = 1, 2, . . ., 5) form the fuzzy matrix R, it can also be called fuzzy priority similarity comparison matrix. e 2 3 r11 r12 r15 6 r21 r22 r25 7 7 14 R6 4 5 e r51 r52 r55 2. Select the cut set of R and sequence the element of set U Usually, the size of matrix U is very large, in order to judge the similarity between the sample and the specified samples, it can select the cut set R of R e from big to small according to the value of rij for every parameter, R is 2 0 3 0 0 r15 r11 r12 0 0 0 7 6 r21 r22 r25 7 R 6 15 4 5 0 0 0 r51 r52 r55 where rij0 1; when rij0 ; 2 0; 1; 0; when rij0 < ; 2 0; 1 To every parameter, as the value of decreasing, if the values of one row all be 1 appears first, then set the rank of this grade as 1. Do just the same way, other grade's rank are 2, 3, 4 and 5 respectively. After processing all of the parameters, add all of the parameters' ranks of each grade, the specified sample whose sum of rank is the smallest is the most similar sample to the sample we assessed. This method has been successfully used in the assessing the unusualness of chemical prospect. Following are the calculated result of the parameters we selected. (1) The calculated result of texture fineness: .

The texture fineness of standard specified samples t1 906916:44 t4 1278484:13

.

t2 1057433:38 t3 125262:50 t5 1336476:00

The texture fineness of the sample t 622957:00

Grading the U f906916:44 2057433:38 1252616:50 1278484:13 1336476:00g crease recovery According to the calculated result, the set of the specified sample U is

and the fuzzy priority similarity comparison matrix of texture fineness R1 is 2 3 1 0:284 0:275 0:325 0:416 6 0:716 1 0:487 0:547 0:641 7 6 7 6 R1 6 0:725 0:513 1 0:558 0:653 7 7 4 0:675 0:453 0:442 1 0:597 5 0:584 0:359 0:347 0:403 1

121

after calculating the ranks of five grades are S1 5;

S2 2;

S3 1;

S4 3;

S5 4

according to the fuzzy priority similarity comparison method, the grade of the sample's crease recovery is 3. (2) The calculated results of first statistical analysis are shown in Table I. The ranks of each grade are shown in Table II. It is evident, that the grade of this sample's crease recovery is 2. (3) The calculated results of second-order statistical analysis are shown in Table III. The ranks of each grade are shown in Table IV. According to the calculation, the grade of this sample's easy care characteristic is 2.

Grade 1 Variance 2 Energy E Entropy H Skewness S

Variance 2 Energy E Entropy H Skewness S Sum of rank

Grade 2

1,070.3461 450.9931 0.01176 0.01697 6.6289 6.1889 ±1.9369 ±0.1414

Grade 3

Grade 4

Grade 5

Sample

443.9601 0.01716 6.1497 ±0.9237

302.0345 0.01723 6.0524 0.7418

190.1972 0.02179 5.7423 ±0.7146

780.8875 0.01407 6.3532 ±0.1419

Grade 1

Grade 2

Grade 3

Grade 4

Grade 5

1 2 3 3 9

2 1 1 1 5

3 3 2 2 10

4 4 4 4 16

5 5 5 5 20

Table I. The calculating results of first statistical analysis

Table II. The ranks of each grade

IJCST 12,2

122

Result and analysis At the same time, we invited many experts to assess this sample's crease recovery by the traditional method, all of the assessment results are grade 2, this result illustrates that the method we proposed has high consistency with the traditional method. From the above calculation results we could see that the grade of the fabric being assessed has little difference among these three methods. Because the definition of the texture parameter is determined by a human being, its scope of application is not universal, different parameter has different scope of application. So through many times of experiments we drew the conclusion that to grade the fabric's crease recovery only by spatial co-occurrence function and texture fineness is not sufficient. There is inconsistency among the assessing results that were assessed by the first-order statistical analysis techniques parameters, but the assessing results that were assessed by the second-order statistical analysis techniques parameters have good consistency. So from the theory and the experiment results we think that the most commonly used texture analysis technique, the second-order statistical analysis technique, is very suitable for grading the fabric's crease recovery. Conclusion From the experiment we proposed, the low precision and high subjective error of traditional method could be reduced by using image processing and texture analysis technique. By comparison, the second-order statistical analysis technique is more suitable for grading the fabric's crease recovery, and the grade of the fabric's crease recovery can be assessed more objectively and correctly by using the fuzzy priority similarity comparison method. In addition, this method has high consistency with practice. Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Sample

Table III. The results of secondorder statistical analysis

Table IV. The ranks of each grade

Texture Texture Texture Texture

consistency Q1 contrast Q2 entropy Q3 gray level occurrence Q4

Texture consistency Q1 Texture contrast Q2 Texture entropy Q3 Texture gray level occurrence Q4 Sum of ranks

0.02138 17.2938 5.9284 19.8779

0.02013 8.3703 6.9754 54.2934

0.02011 7.3033 7.1997 59.7514

0.02010 6.9726 7.2147 63.9753

0.02009 6.4761 7.2157 64.8692

0.02035 10.1481 7.0857 45.7428

Grade 1

Grade 2

Grade 3

Grade 4

Grade 5

5 5 5 5 20

1 1 1 1 4

2 2 2 2 8

3 3 3 3 12

4 4 4 4 16

References 1. Yoa, M. et al., Textile Mertiaral, 1993, p. 6. 2. GB3819-83, The testing method of the fabric's crease recovery. 3. Li, R. and Song, J., Measuring Techniques for Textile Materials, Vol. 12, 1994, pp. 402-5. 4. Bang, Y., ``Assessing the anti-crease characteristic of ramie by fuzzy assessment'', Textile Standard and Testing, Vol. 2, 1993, pp. 30-4. 5. Wang, R., Image Understanding, 1995, p. 10. 6. Wang, D.C., Language Program of Image Processing, 1993, p. 6. 7. Zhang, Y. et al., Fuzzy Mathematics Methods and Application, 1992, p. 4. 8. Yixiang, F.Z. and Randell, R.B., ``Fabric defect detection and classification using image analysis'', Textile Res. J., Vol. 65 No. 1, 1995, pp. 1-9. 9. Laws, K., Texture image segmentation, USCIPI Report No. 940, University of Southern California, Image Processing Institute, 1980. 10. Haralick, R., ``Statistical and structural approaches to textures'', Proc. IEEE 67, 1979, pp. 786-804. 11. Weszka, J., ``A survey of threshold selection techniques'', CGIP 7, 1978, pp. 259-64.

Grading the crease recovery

123

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

A novel technique to measure stick-slip in fabric P.M. Taylor

124 Received 24 December 1998 Accepted 12 December 1999

Department of Mechanical, Materials and Manufacturing Engineering, The University of Newcastle, Newcastle upon Tyne, UK and

D.M. Pollet

Department of Electronic Engineering, The University of Hull, Hull, UK Keywords Fabric, Friction, Garments Abstract Stick-slip is characterised by an intermittent movement when two materials are sliding across each other at very low speeds. This paper describes a new method of measuring stick-slip on knitted and woven fabrics when sliding over non-fibrous materials. The technique uses emitter-receiver pairs arranged at regular intervals along the full length of the fabric panel, which each visualise the irregular movement of sticking and sliding. It is found that an extensible knitted fabric exhibits either shear or micro-slip during the stick phase, before the actual gross slip movement.

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 124-133. # MCB University Press, 0955-6222

1. Introduction Friction, which is defined as the force resisting motion between two bodies in contact, plays a universal role in the simplest actions of living such as walking and grasping. The classical Amontons' Laws describe the frictional force as being directly proportional to the normal force and the coefficient of friction, , or the proportionality as being independent of the geometric contact area between the two surfaces. In 1785 Coulomb verified the two laws and added the concepts of static and dynamic (or kinetic) friction, which are respectively the force necessary to initiate motion and the force necessary to sustain the movement. In general, the static friction is usually higher than the dynamic friction. Further, Coulomb found what is often regarded as the third friction law, that the dynamic frictional force between dry solids was independent of the velocity. However, the above models only adequately explain the frictional behaviour of materials that deform plastically such as metals and ice. They fail to do so for fibrous materials that deform viscoelastically. The frictional force between fibrous materials or between fibrous-non-fibrous materials is not directly proportional to the normal force but can vary by a factor of between 0.57 and 1.06 and can be area dependent (Wilson, 1963). In addition, friction both in fibrous materials, i.e. fibres and yarns (Kalyanaraman, 1988a, 1988b; RoÈder, 1953), and fabrics (Ajayi, 1992a; Nishimatsu, 1984) and in lubricated solids is, in practice, velocity dependent as indicated in Figure 1. Originally, it was thought that the increase in friction at high speeds was due to frictional heating, affecting the yield-pressure and shear strength of the The authors thank the EPSRC for a studentship award to support this work.

Measure stick-slip in fabric 125 Figure 1. Stribeck curve

fibres (Howell et al., 1959). This heating will have some effect on the friction in fibres and fabrics and will be more pronounced in synthetic fibres, though the main mechanism is now explained by lubrication. The resemblance between the velocity curves for fibrous materials (lubricated or clean yarn) and those for journal bearings is striking (Hansen and Tabor, 1957; Kalyanaraman, 1988a). Friction initially decreases with increasing speed, reaches a minimum, and eventually increases again with increasing speed towards a constant value. This type of curve has been referred to in tribology as the Stribeck curve. Four regimes of lubrication can be distinguished: static friction (velocity independent), boundary lubrication, partial or semi boundary lubrication and hydrodynamic lubrication. Stick-slip occurs at the start of the boundary lubrication area where the sliding speed is very low. From the established shearing theory of Bowden and Tabor (1964), it is clear that when two bodies are sliding against each other, contacts are continuously formed and broken. The frictional force necessary to break this cold adhesion has been successfully modelled for non-fibrous materials by using the intimate contact area, the yield-pressure and the shear strength of the junctions, which are assumed constant. However, the latter is not true. At extremely low speeds of sliding, the resultant strength of the junction is greater than at higher speeds. This means, as often has been found in practice even in fabric-fabric friction, that the static friction is higher than the dynamic friction. If therefore one of the sliding bodies has a certain degree of elasticity, the motion will be intermittent at low speeds and stick-slip appears. Theoretical analysis by Bowden and Leben (1939) shows that, in particular, the difference between the static and dynamic friction causes stick-slip in non-lubricated metallic surfaces. The stick-phase is due to the higher static friction, Fs, whereas the slip-phase is due to the lower kinetic friction, Fd. The larger the difference between the static and dynamic friction, the more regular the stick-slip will be (see Figure 2). This relative difference in static and dynamic friction has been encountered as a good indicator for subjectively ranking the tactile sensation of fibres (RoÈder, 1953) and fabrics (Ajayi, 1992a, Ajayi et al., 1995). Increasing the speed of sliding will abruptly terminate the stick-slip at a certain characteristic velocity (Derjaguin et al., 1957).

IJCST 12,2

126

Stick-slip motion has now been successfully used by several researchers (Ajayi, 1992b; Ajayi et al., 1995; Hosseini Ravandi et al., 1994) in relating the frictional characteristics of fabric to the topographic structure of its surface. Ajayi (1992b) showed that there was a positive relationship between the yarn crown height and the peak to peak values of the slip traces, and that the number of peaks correlated well with the yarn density or number of cords. Hosseini Ravandi et al. (1994), however, explored the periodicity of stick-slip patterns of fabric against a perspex sled by using an autospectral density function. Out of the frequency information, they could easily calculate the fabric density in a particular direction (warp or weft) knowing the velocity of sliding. Furthermore, they noticed that where yarns protrude from the surface of the fabric a repeated stretching and releasing occurred when a sled under normal load was sliding over the fabric. This somewhat lateral shifting produced extra broad peaks in the autospectral density plot. When increasing the normal load on the fabric, the stick-slip amplitude increases (not the friction coefficient) and the lateral shifting moves up in frequency. The few studies on stick-slip in fabrics, mentioned above, all use a tensile tester where the applied force is measured either when a sled is moved over the fabric or when the front edge of the fabric sample is moved. However, this method does not reveal the exact fabric movement over its whole area and considering the limpness of fabric it can be questioned whether a fabric panel moves simultaneously as one piece or whether the fabric moves in a kind of ``caterpillar'' manner. For instance, only shows the behaviour of the front edge of the moving fabric panel though the end of the fabric is moving probably with a time delay compared to the front. Hence, a new measuring technique has been introduced which allows the investigation of the movement of the fabric in different places at the same time. First, the measuring concept is explained then second, experimental results are obtained for a representative knitted and a representative woven material. 2. The concept The testing device is designed to detect the local movement simultaneously at different places of a fabric sample when sliding at very low speed over a surface. In essence, there are two non-contacting approaches to this problem either by using a reflective method (e.g. triangulation laser sensor or cameras)

Figure 2. Regular stick-slip plot from frictional force measurement

or a transmissive method. However, cameras are more costly (for high sampling rates) and would require large magnification, whereas most of the reflective sensors suffer from a low power reflection due to the diffuse optical properties of fabric. On the other hand, acoustical methods (reflective or transmissive) would not give the required accuracy due to the large beam angles of the currently available sensors, thus an optical transmission method seems the best option. The concept of testing is simple, in that the light transmission will alter whenever the underlying fabric moves. So, when the fabric between the sensor is stationary, a constant signal is obtained, but when the fibres or threads move, the signal fluctuates. Thus, the transition from constant signal to fluctuation indicates the transition between the stick to slip phase in between the particular sensor-emitter pairs. Note that the constant signal level is likely to change between each stick phase because fabric is not a perfectly flat continuum. It is further assumed here that the slip movement is large, such that the current stick level is independent of the previous level. If now several sensors are aligned across the fabric as sketched in Figure 3 then one should be able to detect whether the fabric moves as one body or as a kind of ``caterpillar''. For example, when the fabric moves as one body, all sensors, three in this case here, should register the same transitions at the same time as illustrated in Figure 4. Otherwise, if for instance the pulling side of the fabric moves first then a time delay should be visible between the individual sensors as in Figure 5.

Measure stick-slip in fabric 127

Figure 3. Sketch of the diode bench for measuring stick-slip in fabric

IJCST 12,2

128

Figure 4. Theoretical sensor output for a ``rigid'' body under stick-slip movement

Figure 5. Theoretical sensor output for a ``stretchable'' body under stick-slip movement

An offset on the sensor signals (1, 2 and 3) is here just introduced to distinguish better the level. Further, the stick ``levels'' are dependent on the light intensity and therefore vary between the different sensors. 3. Experimental procedure The test rig, pictured in Figure 6, consists of three emitter-receiver units, 150mm apart from each other, embedded in a black perspex ``gantry'' structure through which the fabric sample can be passed. A black structure has been chosen to stop any influence of direct light on the measurements. Each unit comprises a small angle gallium-aluminium-arsenide (GaAlAs OD-880F from Opto Diode Corp) emitter in the top plate and a high-speed silicon photodiode in the bottom plate (AEPX65 from Centronic, Inc) (see Centronic, 1995)) as seen schematically in Figure 7. A 0.8mm diameter pinhole in the table reduces the beam to the photo-sensor ( 0.84mm). A thin sheet of acetate covers the top of the table acting as the sliding surface and avoiding any possible obstructions during sliding due to burrs on the pinholes yet allowing light transmission. The diameter of the receiver and the pinhole is selected in a way not to measure in between the individual gauges of the fabric but to average over a small area. Hence, a change of brightness is measured when the fabric mesh is sliding over the pinholes. The IR emitters (880mm) are supplied from a single DC voltage and tuned to give the same light output. The power of the IR beam after passing through the fabric mesh is still a few milliwatts and is, therefore, only

Measure stick-slip in fabric 129

Figure 6. IR diode bench system to measure stick-slip in fabric

Figure 7. Basic optical arrangement

slightly amplified (factor 30) in a reversed bias mode. Note that the above tuning and amplification levels are not at all critical since only the transitions times between DC and the fluctuating levels are important. The output voltage from the photodiodes is captured with a computercontrolled data acquisition system sampling at 1kHz. Before starting the actual stick-slip tests, the diode bench has been verified electrically for any possible electronic delays by feeding a 500Hz sawtooth signal into the IR emitters and measuring their corresponding output from the photodiodes. At 5kHz sampling, no delay was visible between the three emitter-receiver pairs. Further, the idea of visualising the stick-slip effect by measuring the local fabric deformations has been checked out first with a solid material which,

IJCST 12,2

130

Figure 8. Typical detailed sensor output from photodiodes for perspex sled movement. Perspex sled with gridline pattern (0.5mm wide-0.5mm apart) on top, sensor 1 is closest to the pulling side, sampling frequency: 1,000Hz

Figure 9. Magnified and filtered section from Figure 8. Perspex sled with gridline pattern (0.5mm wide-0.5mm apart) on top, sampling frequency: 1,000Hz, solid line is the result of a fifth order lowpass Butterworth filtering with cut-off frequency at 70Hz

obviously considering the stiffness of the material, should register three identical sensor outputs with no delay. A transparent perspex sled with a printed gridline pattern (0.5mm wide lines 0.5mm apart) on the top has been used as solid material. The perspex allows the IR beam to shine through without deflections and the parallel gridlines imitate a perfect fabric structure. The perspex sled (or fabric strip) is connected with a polyester string to a handoperated linear table, which drags the sled (or fabric strip) very slowly ( 0.1mm/s) through the diode bench. A typical detailed output from the three photodiodes when sliding the perspex sled over the diode table is given in Figure 8 together with a magnification of a specific section in Figure 9. Note the staircase form of the output signals for each sensor, as the transmitted light path becomes gradually more (or less) obscured as the grid lines gradually move across it. When the test is carried out over a longer time, it can be seen that each signal is part of a sinusoidal waveform with a period equal to the velocity of travel divided by the grid pitch. The perspex slips once per second or more frequently indicating a slip distance substantially smaller than the grid pitch. These sinusoidal waveforms are out of phase with each other because the distances between each sensor are incommensurate with the grid pitch.

Figure 9 shows a magnified version of Figure 8 where it can be seen that each transition from stick to slip (or slip to stick) occurs at about the same time instant for all three sensors. This indicates, as expected, that the perspex sheet stick-slips as a solid body. Two traces are shown for each signal; the first is the raw signal, the second, smoother trace, has been filtered using a fifth order Butterworth filter with a cut-off frequency of 70Hz (Little and Shure, 1988). The slip phase occurs over about 10ms.

Measure stick-slip in fabric 131

4. Results The two representative fabrics, the details of which are given in Table I, are chosen deliberately with a rather open structure so that the light beam can easily shine through. When sliding the fabrics across the table, two different outputs are obtained for the woven and knitted material. The signals for the woven fabric in Figure 10 correspond to the previous signal for the perspex with a constant signal during the stick phase and a staircase type waveform showing a small slip distance. The signals for the knitted fabric in Figure 11 follow the overall staircase form but the signal is not constant in the ``stick'' phase. This is particularly apparent in the trace for sensor 1 where it can be seen that the voltage level starts to drop before the actual true slip-phase, characterised by the sudden voltage transition, takes place. Sensor 2 also shows some ``micro'' variation towards the end of the stick phase whereas sensor 3, located at the end of the fabric strip, follows the previous pattern of a constant output in the stick region. Note that the knitted material has a less stiff Fabric code

Fibre content

Fabric structure

KA1 C3

Acrylic Cotton

Rib-stitch Plain weave

Note: P: warp/wale; T: weft/course

Area density (g/m2) 176 153

Sett (thd/cm) P T 9 44

10 30

Longitudinal stifness (N/m) 50 1,193

Table I. Fabric details

Figure 10. Photo-diode outputs for a representative woven material (C3). Sensor 1 is closest to the pulling side, sampling frequency: 1,000Hz, fifth order lowpass Butterworth filtering applied with cut-off frequency at 70Hz

IJCST 12,2

132

construction than the woven material (see Table I) and, therefore, will extend more easily when dragged across a surface. In addition, the knitted structure is likely to exhibit lower shear stiffness across its thickness although no experimental measurements have been made to verify this. Hence, it is hypothesised that the voltage decay is caused by either shear in the lateral dimension or an internal or surface micro movement during the ``stick''. Any delay between the individual slip-phases is less than a few milliseconds (< 3ms). A more precise statement cannot be made, mainly because the signals are smoothed by the filtering, which was applied to remove the noise from the signal. 5. Conclusions A novel technique has been introduced to detect stick-slip in fabric. The method is based on an array of emitter-receiver pairs placed at regular distance intervals across the sample, which register variations in light intensity whenever the fabric strip is moving. The stick-slip phenomenon is clearly observed for the two fabrics, one knitted and one woven, which are used in the test. For the low test-speed used, it is noticed that both fabrics move a small distance during the slip phase, much less than their sett. The slip movement also occurs in approximately 20ms compared with a stick time of about 1-s. The knitted material exhibits a very interesting characteristic in the ``stick'' phase where it is observed that some material movement takes place, which changes the light transmissibility of the fabric. It has been hypothesised that this could be due to lateral shear or micromovement either internal to the fabric structure or at the fabric surface. However, no conclusion could be drawn about whether each fabric sample moves as a rigid body during the slip phase or whether a wave is propagated along the length of the sample. Further research is under way to improve the instrumentation and to investigate the exact method of movements during both the slip and the ``stick'' phases.

Figure 11. Photo-diode outputs for a representative knitted material (KA1)

References Ajayi, J.O. (1992a), ``Fabric smoothness, friction, and handle'', Textile Research Journal, Vol. 62 No. 1, pp. 52-9. Ajayi, J.O. (1992b), ``Effects of fabric structure on frictional properties'', Textile Research Journal, Vol. 62 No. 2, pp. 87-93. Ajayi, J.O., Elder, H.M., Kolawole, E.G., Bello, K.A. and Darma, M.U. (1995), ``Resolution of the stick-slip friction traces of fabrics'', Journal of the Textile Institute, Vol. 86 No. 4, pp. 600-9. Bowden, F.P. and Leben, L. (1939), ``The nature of sliding and the analysis of friction'', Proceedings of the Royal Society, Vol. A169, pp. 371-91. Bowden, F.P. and Tabor, D. (1964), ``The friction and deformation of polymeric materials'', The Friction and Lubrication of Solids, Oxford Clarendon Press, London. Centronic (1995), High Performance Silicon Photo Detectors, European Edition 2, Centronic, Inc., Newbury Park, CA. Derjaguin, B.V., Push, V.E. and Tolstoi, D.M. (1957), ``A theory of stick-slip sliding of solids'', Proceedings of the Conference on Lubrication and Wear, Institution of Mechanical Engineers, London, pp. 257-68. Hansen, W.W. and Tabor, D. (1957), ``Hydrodynamic factors in the friction of fibers and yarns'', Textile Research Journal, Vol. 27, pp. 300-6. Hosseini Ravandi, S.A., Toriumi, K. and Matsumoto, Y. (1994), ``Spectral analysis of the stick-slip motion of dynamic friction in the fabric surface'', Textile Research Journal, Vol. 64 No. 4, pp. 224-9. Howell, H.G., Mieszkis, K.W. and Tabor, D. (1959), Friction in Textiles, Butterworths Scientific Publications, London. Kalyanaraman, A.R. (1988a), ``Coefficient of friction between yarns and contact surfaces'', Indian Journal of Textile Research, Vol. 13, pp. 1-6. Kalyanaraman, A.R. (1988b), ``Yarn-friction studies with the SITRA friction-measuring device'', Journal of the Textile Institute, Vol. 58 No. 1, pp. 147-51. Little, J.N. and Shure, L. (1988), Signal Processing Toolbox for Use with MATLABTM User's Guide, The Mathworks, Inc. Nishimatsu, T. and Sawaki, T. (1984), ``Study on pile fabrics, part IV: investigation of factors affecting frictional properties of pile fabrics'', Journal of the Textile Machinery Society of Japan, Vol. 30 No. 4, pp. 100-6. Opto Diode (1992), Opto-electronics Data Book, Opto Diode Corp., Newbury Park, CA. RoÈder, H.L. (1953), ``Measurements of the influence of finishing agents on the friction of fibres'', Journal of the Textile Institute, Vol. 44 No. 6, pp. T247-T265. Wilson, D. (1963), ``A study of fabric-on-fabric dynamic friction'', Journal of the Textile Institute, Vol. 54 No. 4, pp. T143-T155.

Measure stick-slip in fabric 133

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

134 Received 25 April 1999 Revised 18 November 1999 Accepted 18 November 1999

A garment design system using constrained BeÂzier curves Claudia M. Eckert

Engineering Department, Cambridge University, Cambridge, UK, and

Helmut E. Bez

Department of Computer Science, Loughborough University, Loughborough, UK Keywords CAD, Garment design, Design Abstract A CAD tool for the garment industry is described. The tool generates garment patterns using BeÂzier curves and is currently embedded within an intelligent knitwear design support system. The curves fulfil the design constraints imposed by the domain, are adaptable to individual styles and enable intuitive manipulation by the user. The system described is designed primarily to provide a means of improved communication between designers and technicians but it has the potential to become a key component in a bespoke design system.

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 134-143. # MCB University Press, 0955-6222

Introduction Current CAD systems for knitwear design are highly sophisticated expert systems which support programming by automatically generating knitting machine programs from symbolic representations. They give efficient feedback on the technical realisation of garments through design simulation. The systems however are not intended to be used by designers during the conceptual design stage, as they do not support the creation and evaluation of tentative designs. To receive feedback the users need to commit their ideas and invest time into working designs out in detail. A large empirical study of the knitwear industry (Eckert, 1997) has shown that many design specifications produced by industrial knitwear designers are incomplete, inaccurate and inconsistent, because designers do not have the time or technical knowledge to express their ideas accurately. The technicians who create the physical garments need to interpret these specifications in the light of past designs. As a result many sample garments are not what the designers intended, and a significant bottleneck in the design process is caused. This leads to a high ratio of samples to finished garments and to sub-optimal designs going into production. Designers would benefit from technical feedback during the idea generation process. A CAD system should allow designers to specify tentative designs quickly and receive fast initial feedback, so that they can then explore the design space and hand over a design which is technically plausible (Figure 1). This approach has been introduced in fashion information systems, such as the Gerber system, Gerber Garment Technology (1996), where designers can specify a design through modification of older designs and receive initial

A garment design system

135 Figure 1. Interaction between designers and technicians

costing feedback. A knitted garment however is far more complex then a woven one, because in knitwear the shape and fabric are created at the same time. This communication problem between designers and technicians can partially be overcome by a computer system that can turn a tentative and potentially incompletely specified design into a technically correct version which could be understood by a technician. This paper discusses how a mathematical model of garment shapes can be employed to create technically correct solution suggestions. These suggestions can be evaluated visually and edited by designers while maintaining internal consistency. A similar approach has been applied successfully in design support for architecture by Papamicheal et al. (1996) in the BDA system. Their system uses default values to complete input data, and displays instantly in multiple representations. As in the IDIOM system (Smith et al., 1996) active critiquing and user preferences could be included. The mathematical model should enable: .

the design to start from the designer's customary notations;

.

easy editing of designs and solution suggestion;

.

maintenance of domain constraints;

.

adaptability to individual company styles;

.

easy completion of design parameters;

.

highlighting and modification of inconsistent input measurements.

Designers are used to solution suggestions ± since they are created by technicians as part of the design process. This approach exploits the designers' skills in the perceptual evaluation of designs ± they are able to recognise good, or technically correct, solution suggestions even if they could not specify them. This helps designers to explore design spaces.

IJCST 12,2

136

Figure 2. Body pattern showing neckline curve, set-in sleeve pattern, and sleeve-crown curve detail

Garment pattern shapes The tool described in this paper enables knitwear designers to specify the shape they wish the finished garment to have. This is different to the shape of knitted or cut garment parts ± these allow for the stretch properties of the fabric. The designer needs to communicate the final shape, while the technician needs to create the knitted shape. The two dimensional outline, created from the cutting patterns, is described in (inaccurate) sketches that designers include in their customary design specification. The sketches typically are ignored as shape information, because they are too inaccurate, but are used to express the relative location of pattern elements on the garment shape. An accurate garment outline can fulfil this task better. Tailored garments, as well as most knitted garments, are constructed using cutting patterns; i.e. a set of paper templates of the individual garment pieces that are assembled into garments. The garment pieces can be described by lines and curves. To create the three dimensional garment, curves of different shapes need to be stitched together ± this imposes an equal-length design constraint. The garment curves are subject to a variety of other domain constraints due to tailoring traditions and to fabric properties. This paper concentrates on the most common curves in knitwear design: neck curves, armhole curves and the sleeve-crown curve of a set-in sleeve. Figure 2 shows the cutting pattern from which a knitted garment is constructed. This example is from a standard pattern cutting text book

(Aldrich, 1987), which is widely used across the textile industry. As most garment pieces are symmetrical, only half of it is constructed as a cutting pattern. As the back and front of a knitted garment normally have the same shape, both neck line shapes are drawn on the same cutting pattern (see Figure 2, left illustration). The designers specify the width and depth of the neck line. In knitwear, the armhole is a combination of a straight line and a curve. The construction of the sleeve-crown curves is dependent on the armhole. The most commonly used method in industry to construct sleeve-crown curves is to take the width of the sleeve and the length of the diagonal between the beginning and end of the armhole ± this can be seen as an auxiliary line on the cutting pattern in Figure 2 (right). Experienced pattern cutters draw the sleeve-crown curve freehand; and in most cases get it right at the first attempt with the same length as the armhole curve. The designer defines the armhole depth and width and the sleeve width and length, but does not control the inherent interdependence of these values. Even though the basic form of the set-in sleeve has evolved over centuries, there are subtle differences between all curves. Each company has a specific style and each person who constructs them has a specific ``handwriting''. These individual differences need to be incorporated into the shape of automatically created curves. In this paper we show how straight lines and BeÂzier curves can be used to model garment shapes in a manner that satisfies the domain constraints. Domain constraints Garment curves in tailoring and knitwear have a number of domain constraints derived from the traditional tailoring methods to look right to a practitioner: D1: The curve follows a certain type of shape: it has a specified number (often 0) of points of inflexion ± the relative position of which is known within a certain margin of error. The neckline and armhole curves do not have a point of inflexion and the sleeve-crown curve has one. D2: The curves need to be continuous, smooth and monotonic. D3: The relative lengths of different sections of the curve should be appropriate for the domain ± this aspect is controlled by developing heuristics for parameter values at known interpolation points. Knitted fabric is steamed before cutting to prevent it from unravelling ± however, individual stitches disintegrate when they are cut. Practitioners therefore reduce the length of the cut across a stitch as far as possible: This leads to the following constraints: D4: Instead of one long curve or line, runs along and across stitches are cut, as far as possible, and shorter curves are added. D5: The curves need to have horizontal or vertical end tangent vectors.

A garment design system

137

IJCST 12,2

Constraint D5 is extremely important if a curve is to look right. Finally, it is important that: D6: Curves representing seams that are to be stitched together, e.g. armhole and sleeve-crown curves, are of equal length.

138

Interpolation points for the garment curves The end points of the curves can be directly inferred from measurements by the user. Novices are often trained to construct garment curves using interpolation points, and this is the approach taken here. Figure 3 shows the construction of the interpolation point Pa1 for a neckline curve. The point Pa1 is determined from the user-supplied distance d. The term ``construction triangle'' is used for the right angled triangle having hypotenuse r3 ± r0. The point Pa1 makes it possiblepto incorporate p individual curve characteristics and has coordinates (x0 ± d/ 2, y0 + d/ 2) where r3 = (x3, y3) and r0 = (x0, y0). The construction of the interpolation points for the sleeve curve employs heuristics based on Aldrich (1987). In this case the construction triangle is determined by the sleeve width and the length of the armhole diagonal. Offsets from this line are employed at the specified points c1 and c3 (refer Figure 4).

Figure 3. Construction of the interpolation point Pa1 for armhole curves

Figure 4. Calculation of the interpolation points Pa1, Pa2 and Pa3 for the sleeve-crown curves

These points are respectively at distance 1/6 and 2/3 along the line from r0 to r5; this determines the interpolation points Pa1 and Pa3. The final interpolation point, Pa2 , is also a point of inflexion. Modelling technique The curves need to be modelled to suit the domain constraints D1-D6 and user requirements. Smooth, monotonic curves with horizontal or vertical end tangent vectors are required; also the curves must be easily edited. We show below that BeÂzier curves can be used to fulfil these requirements. In addition a BeÂzier curve is a smooth approximation of its defining polygon ± an important property in this design application. BeÂzier curves are popular in many design applications (Faux and Pratt, 1979) they were chosen here for the following reasons: . the manipulation of BeÂzier curves via control points is intuitive; . BeÂzier curves are mathematically simple; . BeÂzier curves give natural control over the end tangent vectors; and . additional interpolation points can be incorporated ± allowing adaptation to individual characteristics and completion of the design parameters in a natural way. Armhole and neckline curves both have one horizontal and one vertical end tangent vector and have no points of inflexion; they can therefore, be modelled by one cubic BeÂzier curve of the form: rt 1 ÿ t3 r0 31 ÿ t2 tr1 31 ÿ tt2 r2 t3 r3 : 0 t 1: Traditionally, the neckline curve is drawn, monotonically increasing, from left to right; the armhole curve is drawn left to right decreasing. The discussion later will focus on neckline curves. Armhole curves can be treated in the same way. The sleeve-crown curves have two horizontal end tangent vectors and one point of inflexion at the second interpolation point. To allow easy manipulation, the entire curve is modelled by one BeÂzier segment. A cubic curve does not have sufficient degrees of freedom to satisfy these constraints and a quintic BeÂzier segment having the general form rt

5 X

n

Ck 1 ÿ t5ÿk tk rk ; 0 t 1

k0

was chosen. Cubic and quintic BeÂzier curve segments, constrained by appropriate interpolation conditions, proved completely satisfactory for our purposes. In traditional BeÂzier curve applications, the overall curve is created by curve segments that are joined together at known interpolation points. In this application, interpolation points are used to construct the individual BeÂzier segments. The interpolation constraints are of the form:

A garment design system

139

IJCST 12,2

140

rti Pai for some 0 < ti < 1; where Pai is supplied by the user and ti is determined from a geometric heuristic. For armhole and neckline curves cubic curves are appropriate and for these we use one such interpolation point ± shown as Pa1 in Figure 5. The associated parameter t1 influences the geometry of the curve; a large value of t1 gives rise to a greater proportion of the curve length between r0 and Pa1. From the shape constraints of the application domain the following heuristic, that takes both the angle and distance between points into consideration, was developed for the parameter t1 ± see Figure 5. t1

t11 tan 1 tan tan

where t11

d1 : d1 d2

The end points, r0 and r3 of the cubic curves are known and the directions of the end tangents at r0 and r3 are respectively horizontal and vertical. The conditions on the cubic described above determine the remaining control vectors r1 and r2 uniquely and the curve is completely specified. The vectors r1 and r2, shown in Figure 5, can be interactively modified if necessary by the designer ± but only at the expense of losing the interpolation at Pa1. We have seen that the sleeve-curve has three interpolation points: r(ti) = Pai for 1 i 3 and 0 < t1 < t2 < t3 < 1; it cannot therefore be modelled by a single cubic segment. However, a quintic segment can be used. The interpolation

Figure 5. Calculation of the parameter value at the interpolation point Pa1

point Pa2 is also a point of inflexion, i.e. the curvature is known to be zero at Pa2. Heuristic methods, very similar to those for the cubic case, were developed to compute suitable values for the parameters t1, t2 and t3. The end points r0 and r5 ,the interpolation points Pa1, Pa2, Pa3 and the corresponding parameter values t1, t1, t3 are then known. To fully specify the quintic curve under the conditions stated above the following remain to be determined: . the length of the end tangent vectors and ; . the two remaining BeÂzier points r2 and r3.

A garment design system

141

Both end tangent vectors at r0 and r5 are horizontal, i.e. we have: r1 r0 1; 0 and r5 r4 1; 0: Values for the lengths and of the end-tangent vectors can be calculated using domain heuristics; they depend on the ratio of the sleeve width to the sleeve height. These dimensions are the two short sides of the construction triangle defined by r0 and r5 (see Figure 6). The geometry suggests the heuristics: A tan" and B tan"; where A and B are constants, for and . These values of and complete the specification of the curve and give satisfactory results. The parameters A and B can be adjusted to suit individual styles. In this paper we show curves for which A = 1 and B = 2/3. Length constraints Both the cubic and the quintic curves used to model the armhole and sleeve top are completely specified by the applied domain constraints and the applied heuristics. One aspect not yet considered is the equal-length domain constraint D6 ± it applies here since the armhole and sleeve top are to be stitched together. The cubic curve for the armhole defines the style-line of the garment. It is not

Figure 6.

IJCST 12,2

142

appropriate to alter in in any way, but slight adjustments to the endpoint positions of the quintic sleeve top curve are permissible. The following iterative process was satisfactorily applied to satisfy the equal-length constraint D6: determine the length lc of the cubic curve by integration repeat determine the length lq of the quintic curve by integration if lq > lc then reduce the construction triangle for the quintic by moving r0 and r5 along the edges of the triangle and towards the point Pc ± refer Figure 6 if lq < lc then increase the construction triangle for the quintic by moving r0 and r5 along the edges of the triangle and away from the point Pc recompute the quintic curve using the adjusted values of r0 and r5 ± maintaining the domain constraints until lq = lc The prototype system A prototype system (Figure 7) has been developed incorporating the methods described in the paper. The system enables users to select descriptions of body, neckline and sleeve following the domain-pattern of verbal descriptions.

Figure 7. The system prototype

Different domain terms are mapped to fundamental categories of shapes. A measurement selection window for each garment type changes with garment type description to request exactly those measurements required to specify the shape. At present, missing measurements are completed using default values for each garment type. The shape can be displayed as a cutting pattern (top right window) or a two-dimensional outline (bottom right window). A standard extension to the program would enable the user to edit the garment shape, by direct manipulation of control points, while maintaining mathematical consistency. Each garment shape is represented by a complete set of measurements (lower left window), which are translated into a set of vectors that define the garment geometry in terms of straight line segments and BeÂzier curves. Conclusions and further work The paper discusses the geometric domain constraints of the knitwear design process and demonstrates that BeÂzier curves may be used to satisfy them. The application of BeÂzier curves described here is novel in the sense that midsegment interpolatory points and curve length conditions, which arise naturally in the domain, can be tolerated. A further novel feature of the system is its ability to generate automatic solution suggestions from incomplete input ± by intelligent completion of the data. The communication between designers and technicians poses a serious bottleneck in the design process even if designers and technicians work closely together. In recent years the trend in the knitwear industry is to geographically separate production and sampling resources from design centres ± creating a greater communication problem. The tool described in this paper can be Webbased and used between remote sites as required. The authors are currently investigating the use of the tool for the automatic construction of bespoke garment patterns from scanned body measurements or anthropometric data measurements taken from the customer. It is relatively simple to incorporate individual company or designer characteristics via the appropriate positioning of interpolation points in the mathematical models. References Aldrich, W. (1987), Metric Pattern Cutting, 4th ed., Unwin Hyman, London. Eckert, C. (1997), ``Intelligent support for knitwear design'', PhD thesis, Department of Design and Innovation, The Open University, Milton Keynes. Faux, I.D. and Pratt, M.J. (1979), Computational Geometry for Design and Manufacture, Ellis Horwood, Chichester. Gerber Garment Technology (1996), Design and Merchandising, Sales Brochure, Gerber Garment Technology, Inc., Tolland, CT. Papamicheal, K., LaPorta, J., Chauvet, H., Trzcinski, T., Thorpe, J. and Selkowitz, S. (1996), ``The building design advisor'', Proceedings of ACADIA '96. Smith, I., Stalker, R. and Lottaz, C. (1996), ``Creating design objects from cases for interactive spatial composition'', Proceedings of Artificial Intelligence In Design '96, Kluwer Academic Publishers, Dordrecht, Holland.

A garment design system

143

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,2

144 Received 20 May 1998 Revised 4 November 1998 Accepted 4 November 1998

COMMUNICATIONS

Kinanthropometry study of the physique of disciplined personnel Brenda Tsang, C.K. Chan and G. Taylor

Institute of Textiles and Clothing, Hong Kong Polytechnic University, Hunghom, Kowloon, Hong Kong, People's Republic of China Keywords Fire, Physique Abstract This paper applies the study of kinanthropometry to define the physique of 49 Hong Kong-Chinese Fire Services recruits. Kinanthropometry has been widely used in predicting the secular trend in increased body size of and among different ethnic groups world-wide. It has also been applied in the studies of the physiques of Olympic athletes. The criteria developed for this research can be used as standards for physical recruitment in the disciplined forces, as well as for streamlining and improving the basic measurement scale for the manufacture of uniforms. The results showed that the somatotypes of the studied disciplined personnel were distributed in the range of endomorphic mesomorph. Significant correlation was observed between somatotype components and body girths and length measurements.

Introduction Kinanthropometry is defined as ``the study of human size, shape, proportion, composition, maturation, and gross function, in order to understand growth, exercise, performance, and nutrition''[1]. It is similar to the mechanistic approach to human motion, i.e. anthropometry. However, the study of anthropometry is confined to width, length and girth measurements, rather than changes that occur in the human physique as a result of physical training. Somatotyping provides classification of the human body into three basic types: ectomorph, mesomorph and endomorph[2]. Therefore, through the application of kinanthropometry together with somatotyping in the study of the human body, the physique of the disciplined personnel could be identified and categorised as a referred standard for recruitment procedure. The uniform manufacturers could also benefit from consolidated and simplified size charts for different disciplined forces.

International Journal of Clothing Science and Technology, Vol. 12 No. 2, 2000, pp. 144-160. # MCB University Press, 0955-6222

Background Kinanthropometry is a neologism derived from the Greek root words Kinein (to move), Anthropos (man) and Metrein (to measure) [3]. It refers to the dynamic relationship and quantitative interface between human structure and function. In depth, kinanthropometry is a scientific specialisation dealing with the measurement of man from a variety of morphological perspectives. It is applied to movement and those factors which influence movement, including:

components of body build, body measurements, proportions, composition, shape and maturation; motor abilities and cardio-respiratory capacities; physical activity including recreational activity as well as highly specialised sports performance. Somatotyping and anthropometry Hippocrates (460-370BC) was the earliest known person who suggested the classification of human physique as ``short, thick'' and ``long, thin''. He defined the term habitus phthisicus as people susceptible to tuberculosis while habitus apoplecticus were people susceptible to vascular disease. His idea aroused the interest in the development of somatotyping. The early somatotyping involved the study of interrelations of morphology and susceptibility to disease. Since then, many physicians and researchers, such as Aristotle in the fourth century BC, Halle in 1797, Rostan in 1828, and di Giovanni in the late nineteenth century, have endeavoured to further elaborate and develop the idea. In the 1940s, Sheldon[2] examined and photographed thousands of physiques and perfected the method of (photoscopic) somatotyping, which is the description and assessment of the body according to three shapes and composition scales: endomorph (heavy, barrel-shaped fat person), mesomorph (heavy boned, lean, but muscularly massive physique) and ectomorph (very thin person with hardly any depth or width, very thin angular bones with muscle and almost no fat). Three numerals were used to indicate the degree of each component. 1 and 7 represented the two extremes of each component. Sheldon[2] also believed that genes affected human body shape, which was determined right before birth (i.e. genotype). Many researchers, including Tanner and Reily[3], followed Sheldon's[2] method in their somatotyping study. Some of them modified Sheldon's method and developed their own methods, like Parnell[4] in the 1950s. In the 1960s, Carter and Heath[5] adopted Parnell's method and believed that human body shape was affected by the external environment (i.e. phenotype), and was not affected by age and sex differences. Carter and Heath developed their own (anthropometric) somatotyping method [5]. One of the striking things about humans is that they are all basically the same shape. This similarity of shape makes comparative anthropometry much simpler. The development of anthropometry added new dimensions to the study of morphology. Anthropometry was first used in studies of morphology in the seventeenth century when Elsholtz, at the University of Padua, established a method for taking measurements on the body. Somatotyping refers to physique and its appearance, i.e. body shape; anthropometry, on the other hand, defines body size. Anthropometry is the study of the range of human physical dimensions, such as size (e.g. height), breadth (e.g. shoulder width) and distance between anatomical points (e.g. upper arm length). The approach of anthropometry has been applied in the design of work-clothes, apparel sizing system as well as sewing machine workstations in the garment industry. Anthropometry, together with somatotyping were developed since the Middle Ages; however, their joint

Physique of disciplined personnel 145

IJCST 12,2

146

application is very rare in the clothing field. Somatotyping could not be done by anthropometry alone, but the measurements can be useful as supporting evidence for the defined somatotypes. Its integrated study with somatotyping or kinanthropometry can be regarded as a breakthrough in the clothing field. In anthropometric research studies of the US Air Force and Army by Lohman[6] as well as Hertzberg[7], over 70 body girths and lengths were included in the measurement list. However, careful selection and determination of types and numbers of measurements were made according to individual situation especially in this project. Subcutaneous fat layer and body shape The quantity and proportion of the various constituents of the human body are empirically linked to health and quality of life. Knowledge of the interrelationships between constituents within a given level or between levels (i.e. cell, tissue, organ, system and organisation), is essential for a fundamental understanding of body composition, and is useful for indirectly estimating the size of a particular compartment. Environmental influences play an important role in an individual's habitual physical activities, which in turn affect the fat distribution in men. The effect reflects on the fatness and muscularity. These environmental factors formed the ground of Carter-Heath's[5] somatotyping method. On the other hand, fat is deposited and distributed according to a person's sex, and this, together with muscle and bone development, accounts for the difference in body shape between the sexes[8]. The rate of fat deposition differs on different parts of the body so that (in absolute terms) the same individual may exhibit a different fat pattern at a different age[9]. Similarly, older and younger men may differ in the pattern formed by the absolute fat thickness, not because of an age factor, but simply because the older man is fatter[10]. In Tsang's research study, the subjects' ages ranged from 26 to 33; therefore, the difference in absolute fat thickness due to age factor was minimal. Skinfold thickness was measured, as a studied variable to predict the somatotype, the main layer of importance being the subcutaneous layer (also called superficial fascia). This layer anchors the skin to the muscles and the bones. It is composed of adipose tissue (fat cells). Adipose tissue is one of the many different types of connective tissue found in the human body. The adipose tissues in the subcutaneous layer are scattered far apart with many fibres between them. It is a metabolically active tissue that stores fat and releases it in response to a variety of nervous and hormonal stimuli. It also acts as an insulator to help maintain body temperature and acts as a protective padding in certain areas. In obese persons, the fat reserve is substantial ± being less in lean, muscular individuals. Certain parts of the body display more fat than other parts (e.g. the abdomen and buttocks). Such areas are known as fat depots[11]. Actually skinfold thickness measurements (sum of the triceps, subscapular and parathoracic) are used as indices of obesity, like the body mass index[12].

Numerous studies about skinfold thickness in relation to subcutaneous fat of various age groups among different ethnic populations have been carried out in the past 40 years. Skinfold equations for predicting body composition, usually, density, lean body mass, fat body mass, etc., have been derived. Different skinfold sites were chosen, different techniques were applied and different methods were used. In all these studies, subcutaneous skinfold thickness measurements were found to be significantly correlated with the total amount of adipose tissue (internal plus external), indicating a dominance of subcutaneous adiposity (by anatomical approach, skin and adipose tissue are separable, but by chemical approach, they are combined). The mean adipose tissue weight of the males was around 28.1 per cent of body weight[13], and the male skinfolds were slightly more compressible than female skinfolds at five out of the seven sites measured[14]. In practice, the individual subcutaneous adipose tissue patterns vary widely. However, the individual fat patterns hold their shapes even during weight reduction and when the fat thickness changes[9,15]. These results supported Sheldon's single genotype rating theory that the somatotype (body shape) will not change during an individual's lifespan. Variability and validity of the choices of skinfold sites have been researched in several studies[15-16]. Given the easy accessibility of subcutaneous adipose tissue and the fact that it contains a large fraction of the body's total fat content, this application of skinfold caliper appears the most reasonable indirect, noninvasive method. However, what is really being measured is the thickness of a double fold of skin and compressed subcutaneous adipose tissue. Sites where skin thickness is small relative to skinfold might prove better predictors of adiposity. Consequently, the subscapular skinfold should be a poorer predictor than skinfold at arm and leg sites, with the front thigh and medial calf showing the lowest compressibility. The triceps, a highly favoured site for fat prediction, ranked poor[13] since arm fat was found to be more associated with trunk fat. The best predictors found were front thigh, medial calf, rear thigh[15], suggesting that leg fat is important in any index used to describe individual differences in anatomical patterning of subcutaneous fat and supraspinale. The reading should be made as soon as the calipers are applied to the skin, since over 70 per cent of the total compression take place within four seconds[14]. Other factors affecting the variability of the measured skinfold thickness are the different description and definition of the exact skinfold locations, different compressibility (pressure), and different measurers. For skinfolds, the estimate of variation in the relation of the sum of a set of skinfolds to total body fat in a specific population was 3.3 per cent fat which corresponds to 0.0075gm/cc in density units[17]. Triceps and Abdominal are the two best skinfold sites for prediction of body density or measures of fatness, but do not discriminate well among individual differences in anatomical distribution. Since all the gurus used triceps, abdominal (except Carter and Heath[5]) and subscapular as the measured variables in defining subcutaneous adiposity, therefore, medial calf, front thigh, supraspinale were chosen by Tsang as the skinfold sites in deriving endomorphy component[15].

Physique of disciplined personnel 147

IJCST 12,2

148

Muscle and physique Muscularity varies with skeletal sturdiness and reflects habitual physical activity. It is true that exercises can change the body shape since muscles are initially long and lean with very little fat. Then as a person becomes older and more sedentary, intramuscular fat slowly invades the muscle. The shape of the muscle itself becomes short and squat. Exercises can help to burn away the fat and muscle will then go back to its original long and lean shape[18]. Loss of the subcutaneous fat will result in the change of body size, but the change in shape will be minimal.

Application The disciplined forces in Hong Kong can be classified as a homogeneous group in the population. They are always subjectively defined as healthy and muscular people who are actively involved in physical exercises. No objective measurements are ever used to quantify the exact body shape of this group as a whole. In Hong Kong, the disciplined forces consist of a variety of ``streams'' like the police, correctional, fire services, immigration, customs and excise, etc. The recruitment procedures and requirements of these disciplined forces, though not exactly the same, are all similar in nature. The physical requirement is usually in terms of body weight, height and eyesight, which are unclear for classification of the physique of the disciplined personnel in Hong Kong. The lack of an accurate category for defining the required body configuration for each of these disciplined forces, causes difficulties in the development of consolidated size chart. Each disciplined force has its own sizing system. The range of sizes is wide even though the size chart is used only for the same force. There are as many as 37 sizes for the winter blue twill shirt and only nine sizes for summer white twill shirt. From the standpoint of the clothing manufacturers as well as with consideration of the physiology and psychology of these personnel, there is a need to develop alternative criteria with united, objective and measurable standards for the disciplined forces based on kinanthropometry study and somatotyping analysis. Kinanthropometry and somatotyping have never been used in studies involving the Hong Kong-Chinese disciplined personnel, neither have they been employed in the context of clothing manufacture. Differences in size and certain skinfold thickness are likely to occur before and after training. However, the body configuration of the sample group is expected to remain more or less the same. In order for the somatotype to change, the skeleton must change, as well as the shape of the head, the bony structure of the face, the neck, wrists, ankles, calves and forearms, and the relations of stature to measurements made at places where fat does not accumulate. The deposit or removal of fat does not change the somatotype, or any of the measurements except at those fat depots. That is why a reduction in weight will not cause much difference in the

individual fat patterning. The uniform manufacturers can make use of the information of these body configurations to fine-tune their patterns and sizing system. Forty-nine subjects from the Fire Services Department were randomly selected. Their body measurements and photographs were recorded for analysis. It is hypothesised that newly recruited personnel in the disciplined forces may be classified as endomorphic mesomorph or ectomorphic mesomorph. Differences in body size, skinfold and body girth measurements are likely to occur. It is predicted that recruits engaging in the same training activities provided with similar food intake and of the same age and sex should have the same direction of body weight change. However, the body configuration of the studied subjects is expected to remain the same even after prolonged physical training. Three somatotyping methods (Sheldon et al.[28], W.H.'s photographic somatotyping method, CarterHeath's anthropometric somatotyping methods[5] as well as Tsang's modified somatotyping method[19-20]) were adopted in the analytical process.

Physique of disciplined personnel 149

Methodology Apart from the basic metric data (i.e. age, weight and height), six skinfold thickness, two biepicondylar bone breadths and 23 anthropometric parameters were directly measured and recorded. The skinfold thickness and two bone breadths were mainly applied into somatotyping analysis while the 23 anthropometric parameters were included to define certain basic information for completing the size specification of the uniforms. Those dimensions having a more direct and general use for problems associated with clothing are marked by an asterisk. The number of variables for measurements as shown in Table I are to be considered with references to those taken in the anthropometric Metric

Skinfold thickness

Heights and lengths

Widths and circumferences

1. Age 2. Weight 3. Height

1. 2. 3. 4. 5. 6.

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

10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

Triceps Subscapular Supraspinale Medial Front thigh Abdomen

Cervical height Back waist length* Outside leg* Inside leg to knee* Inside leg to ankle* Underarm to elbow* Underarm to wrist* Upper arm length Outer arm length*

Head circumference* Neck circumference* Arm scye Cross back width* Shoulder width* Middle upper arm circumference Wrist circumference* Chest circumference* Waist circumference* Hip circumference* Middle thight Knee circumference Calf circumference Ankle circumference

Note: * Dimensions marked by an asterisk have a more direct and general use for problems associated with clothing

Table I. The 23 basic anthropometric measurements

IJCST 12,2

150

surveys of the military forces, for example, the Australian Army[21-22], Airforce of USA[23-24] and Japan[25]. All measuring techniques are practised extensively before data collection proceeded. Heath-Carter's anthropometric somatotyping method Carter and Heath[5] defined somatotyping based on phenotypes (including anatomical and psychological traits that resulted from both heredity and environment), They modified Parnell's method and used the metric system. Their method was also discussed by other authors[26-27]. The measurements are listed in Table II. Detailed procedures of Carter and Heath's somatotyping methods[2] are included in Appendix 1. Sheldon's photoscopic somatotyping method Sheldon et al.[28] explained somatotyping based on the concept of a genotype with a single rating, irrespective of changes to the physique during the subject's lifespan. Sheldon's somatotyping required standard photographs (with frontal, lateral and dorsal view); 17 photographic body measurements are taken on the photographs except the subjects' weight and height. The 17 photoscopic body measurements in the five categories are defined through the photo images and are used as a calculation base for somatotype body ratio. The locations and descriptions of the 17 diameters according to Sheldon's standard instruction areas are shown in Appendix 2. Detailed procedures of Sheldon's somatotyping methods are included in Appendix 3. Somatotyping method based on Tsang's modifications Modifications are mainly made in the rating of endomorphy and ectomorphy. No modification is made on the rating of mesomorphy. .

Table II. Heath and Carter's ten anthropometric dimensions

Modification of rating endomorphy. Since leg fat is an important indicator of adiposity patterning, therefore, the skinfold site at front

Size

Weight/height

Four skinfold thickness

Triceps skinfold Subscapular skinfold Supraspinale skinfold Medical calf skinfold*

Two limb girths

Upper arm girth, flexed and tensed, right Calf girth, right

Two bone breadths

Biepicondylar breadth of the humerus, right Biepicondylar breadth of the femur, right

Note: * This site is not used in the rating of endomorphy in Carter-Heath's method but is included in Tsang's method

.

thigh is included in the skinfold measuring parts (together with the original sites at triceps, subscapular, supraspinale and medial calf). Modification of rating ectomorphy. By isometric similarity, the arm length (the humerus + radius + hand) should be half of the stature. It is assumed that the humerus is two-fifths of the arm length, and, if the armspread is proportional to the stature, and the cross shoulder is around 1/3 of the armspread, then the humerus should be 2/5 1/3 = 2/15 of the stature. Hence the theoretical humerus length in relation to the stature with the actual humerus measurement and then ectomorph is defined without using the weight measurement.

Equipment for measurements There were two types of equipment being manipulated for conducting data in this research. A detailed description was stated in the following sessions. Somatotyping instruments Five types of equipment were used to obtain the primary data for somatotyping analysis. . Spreading anthropometer (Siber-Hegner GPM, Switzerland) was used for taking measurements of the larger body breadths, depths, and some lengths. . (The property of Institute of Textiles and Clothing (ITC) of Hong Kong Polytechnic University (HKPU). Calibration was done beforehand by the laboratory technician of ITC. . Electrical skinfold caliper (Creative Health Products, 5148 Saddle Ridge Road, Plymouth, MI 48170, England, PAT) was used for taking measurements of various skinfold thickness. . The property of Rehabilitation Science Department (RS) of Hong Kong Polytechnic University (HKPU). Calibration was done beforehand by the laboratory technician of RS. . Harpenden skinfold caliper (H.E. Morse Co., British Indicators, Ltd) was used for taking measurements of various skinfold thickness. . The property of Rehabilitation Science Department (RS) of Hong Kong Polytechnic University (HKPU). Calibration was done beforehand by the laboratory technician of RS. . Various measuring accessories: 150cm tailor measuring ruler-tape, 30cm straight ruler and size No. 10 triangle ruler. . Digital camera modelled ``PhotoPC500'' (Epson, Japan) mounted on a camera stand (to avoid any unnecessary vibration when photographs are taken) was used to take photographs of some of the subjects. The information being extracted was put into photoscopic somatotyping analysis.

Physique of disciplined personnel 151

IJCST 12,2

152

Anthropometric instruments Anthropometric instruments were used during the entire program. Each instrument is graduated in metric system. All measurements were recorded in centimetre (cm) to the nearest 0.5cm (accepting a tolerance of plus or minus 2 millimetre (mm)). . Anthropometer (Siber-Hegner GPM, Switzerland) for taking measurement of the various body heights and lengths. . The property of Institute of Textiles and Clothing (ITC) of Hong Kong Polytechnic University (HKPU). Calibration was done beforehand by the laboratory technician of ITC. . Various measuring accessories: 150cm tailor measuring ruler-tape, 30cm straight ruler and size No. 10 triangle ruler. Methods of equipment handling Care was taken in measurements throughout the survey since the final analysis could not be more accurate than the individual measurements that are used in it. . Length and width measurements: The end of the anthropometer was kept at right angles to the floor. In reading these measurements, care was taken in placing the eye at right angles to the ruler, so that errors of parallax are avoided. . Girth measurements: These measurements were taken using a normal tape measure with adequate pressure or allowance. . Skinfold measurements: The skinfold caliper was placed gently close to touch the contact points of the body and the measurement being read off after two seconds. Preparation for photography for photoscopic somatotyping . Attire: The most appropriate attire for the selected subjects should be a pair of short pants. Subjects refusing to bare their chests were asked to wear a closely fitted T-shirt. The degree of accuracy of the body measurements may be affected; however, it is assumed that the final results would still be meaningful. . Posture: The subject was asked to stand straight with eyes at Frankfort Plane (i.e. looking forward), with feet apart and parallel to shoulders. Results and discussion The mean age, weight and height were 22 years, 66.2kg and 173.8cm respectively. The preliminary mean somatotype ratings ranged from 3.4 to 4.3 in endomorphy, 3.0 to 3.9 in mesomorphy and 2.8 to 3.6 in ectomorphy. The mean somatotype calculated by Sheldon's method and Carter-Heath's method are 3.5-4-3.5 and 4-4-3 respectively. The difference in individual component rating was half only. Whereas, the mean endomorphy and ectomorphy ratings

derived by the Tsang's method were 4 and 3.5 respectively. The mean somatotypes of the 49 subjects derived by the three different somatotyping methods are tabulated in Table III. The observed differences were also within half rating. Hence, the preliminary mean somatotype rating for the 49 subjects was 4-4-3 , which may be classified either having a mesomorph-endomorph (endomorphy and mesomorphy are equal and ectomorphy is smaller) or an average somatotype (no component differs by more than one unit from the other two). The descriptive summary including the means, standard deviations, coefficients of variations, minimum and maximum values as well as the percentile distributions of 5th, 25th, 50th, 75th and 95th percentiles for the three somatotype ratings and anthropometric variables are given in Tables IV and V respectively. The mean somatoplots of the 49 somatotypes derived from Sheldon's, Carter-Heath's[5] and Tsang's[17] methods as well as the overall mean somatoplot are plotted in Figure 1. The somatoplots showed that the general somatotype distribution of the studied disciplined recruits is mainly scattered around the centre, i.e. the average somatotype 4-4-4, and tend towards endomorph-mesomorph. The SDD (Somatotype Dispersion Distance) and SAD (Somatotype Attitudinal Distance) ranged from 1.5 to 1.6 and 3.3 to 3.7 respectively. Close range of distribution of the measured data can be observed in most of the variables except for the height, cervical height, back waist length, head circumference, arm scye and hip circumference. Among these six variables, the measurements of the first four variables will not change in the future since they are confined by the growth of the bones, which have already reached a stable state at the subjects' age. However, the last two variables can be changed by the accumulation of the subcutaneous fat layer or by the change in muscular dimension. Therefore, the range of distribution of the measurements of these two variables may be quite different after the subjects have gone through a longer period of physical training.

Somatotype

Carter-Heath

Somatotype methods Sheldon

Tsang

Endomorphy Mesomorphy Ectomorphy

4 4 3

3.5 4 3.5

4 4 3.5

Measurments

Mean

Endomorphy Mesomorphy Ectomorphy

3.7 3.8 3.8

S.D. C.V.% Min 0.7 0.5 0.6

18.92 13.16 15.79

2.1 2.5 1.9

Physique of disciplined personnel 153

Table III. Mean somatotype of the 49 subjects derived by different somatotyping methods

Max

5%

25%

50%

75%

95%

5.3 4.7 4.5

3.04 2.83 2.35

3.60 3.17 3.00

4.00 3.67 3.38

4.40 4.00 3.63

5.06 4.33 4.38

Table IV. Statistical summary and percentile distribution of somatotype components (n = 49)

IJCST 12,2

154

Table V. Statistical summary and percentile distribution of anthropometry measurement (n = 49)

Measurments

Mean S.D. C.V.% Min

Max

5%

25%

50%

Weight (kg) Height (cm) Cervicale height Back waist Outside leg Inside leg Inside leg to ankle Underam to elbow Underarm to wrist Upperarm Outerarm Cross back Head Neck Arm scye Mid-upperarm Wrist Chest Waist Hip Mid-thigh Knee Mid-calf Ankle

66.2 173.8 143.36 107.74 84.39 39.20 76.33 20.57 44.69 32.70 57.07 38.10 56.84 38.08 45.93 28.36 17.58 89.08 75.24 93.45 46.08 37.51 37.78 25.12

80 183 154 118 90 59 89 28 64 38 77 42 61 40 50 32 20 100 90 100 52 43 43 29

57.7 168.5 138.0 103.0 79.5 34.0 70.0 15.7 39.0 28.8 52.0 36.0 54.0 36.4 42.4 26.0 16.4 83.0 69.4 88.4 41.4 35.0 34.4 19.6

62 170 140 106 82 37 74 19 42 31 54 37 56 38 44 27 17 86 71 91 44 36 36 24

65.0 71.0 77.0 173.5 177.5 181.1 143.0 147.0 148.6 107.5 110.0 112.6 84.0 86.5 89.0 40.0 41.0 44.3 77.0 78.0 82.6 20.5 22.0 24.6 45.0 47.0 49.0 33.0 34.5 36.3 57.0 59.0 61.0 38.0 39.0 41.0 57.0 58.0 59.0 38.0 39.0 40.0 46.0 47.0 49.0 28.0 30.0 31.0 18.0 18.0 19.0 89.0 91.0 95.2 75.0 78.0 83.2 94.0 96.0 98.0 46.0 48.0 50.0 37.0 38.0 41.0 38.0 39.0 41.0 26.0 27.0 27.6

6.1 4.3 3.82 3.16 3.06 4.56 4.34 2.56 4.33 2.48 4.21 1.66 1.57 1.1 2.17 1.62 0.79 3.91 4.64 3.64 2.76 1.95 2 2.66

9.21 2.47 2.66 2.93 3.63 11.63 5.69 12.44 9.69 7.58 7.38 4.36 2.76 2.89 4.72 5.71 4.49 4.39 6.17 3.90 5.99 5.20 5.29 10.59

55 168 137 101 78 27 66 15 37 25 46 33 53 35 42 25 16 81 66 85 40 34 34 15

75%

95%

The differences in somatotype component rating between the different somatotyping methods can be obtained using the conventional somatotyping methods and Tsang's[19] somatotyping method. Furthermore, the paired sample T-test results of individual somatotype component did not show significantly different results between Carter-Heath[5] and Tsang's[19] methods in deriving endomorphy component (t = ±0.628, p = 0.532) as well as ectomorphy component (t = ±1.754, p = 0.083). The mesomorphy component ratings obtained by Carter-Heath and Sheldon's methods were not significantly different (t = ±1.08, p = 0.283). The results showed that individual somatotype components have good correlation with certain body girth and body length measurements. Endomorphy has moderate correlation with chest girth (r = 0.49, p < 0.0005), waist girth (r = 0.59, p < 0.0005) as well as mid-upperarm girth (r = 0.55, p < 0.0005). These girth areas are fat depots where fat is easily accumulated and the range of measurements could vary widely with age. Mesomorphy did not have high correlation with any of the body measurements, while ectomorphy correlated well with underarm to elbow length (r = 0.64, p < 0.0005) as well as with hip girth (r = ±0.52, p < 0.0005) negatively. The correlated girth measurements are the main variables in generating the size chart since their variations occur more frequently than any other body measurements. Alternatively, the underarm to elbow length is a stable body

Physique of disciplined personnel 155

Figure 1. Somatoplot of the distribution of the Fire Service recruits by different somatotyping methods

measurement similar to height since bone growth matures during early adolescence. Hence, emphasis should be put on the body girth variables where sizes are concerned. The somatotype rating of the 49 subjects showed that the newly recruited disciplined personnel are in a range of mesomorph-endomorph body type. Though these 49 subjects might not be a true representation of the whole discipline force, however, they have already passed a series of vetting

IJCST 12,2

156

standards and procedures before being selected as potential personnel. Therefore, their somatotype can be said to be the baseline of the body configuration of regular disciplined personnel. Finally, the sum of the skinfold thickness of the 49 subjects varied widely from 40mm to 114mm. The correlation between the weight and the sum of the skinfold thickness is r = 0.49, which showed medium relationship between the subjects' body mass and their adiposity layer thickness. That is why change in weight will not cause much difference in the individual fat patterning. Conclusion The concept and application of kinanthropometry, somatotyping and anthropometry were introduced in this paper. The inter-correlation in human physiques is revealed by the introduction of a standardised method by which individuals can be somatotyped with precision and reliability. The environmental influences also play an important role in an individual's habitual physical activities. The effects may reflect on the fatness and muscularity. Different body type patterns appear between sports and within sports, according to the specific physical demands of sports. Hence, the application of photoscopic as well as anthropometric somatotyping methods make possible the identification of a certain physique for a given occupation, for example, a career in the disciplined forces. The results showed that the mean somatotype of the studied samples is 4-4-3, which is average in endomorphy, mesomorphy and weaker in ectomorphy. The average somatotype distribution of the studied disciplined recruits are mainly distributed around the centre, i.e. the average somatotype 4-4-4, and tended towards endomorph-mesomorph. There is possibility that the mesomorphy component increases while endomorphy component decreases. The recruits of the disciplined forces are assumed to be of a homogeneous group with a more or less the same somatotype (body type) or within the same physique quartile on the somatochart. The results match this assumption. And even after a prolonged period of strict physical training, it is assumed that their muscle shape may change, their fat thickness may decrease with fat patterning unchanged, but the somatotypes may remain unchanged. The criteria developed for this research can be utilised as referential standards for the physical recruitment in the studied discipline forces. Based on the studied relationship between body shape and body size of the discipline personnel, the clothing industry can also be benefited through streamlining and improving the basic sizing scale for the manufacture of discipline uniforms. References 1. Day, J.A.P., Perspectives in Kinanthropometry, The 1984 Olympic Scientific Congress Proceedings, Vol. 1, Human Kinetics Publishers Inc, IL, 1940. 2. Sheldon, W.H., Stevens, S.S. and Tuck, W.B., The Varieties of Human Physique, Harpers, New York, NY, 1940, p. 5. 3. Reily, T. et al., Kinanthropometry III, E & F.N. Spon, London, 1986.

4. Parnell, R., Behaviour and Physique, Edward Arnold, London, 1958. 5. Carter, J.E.L. and Heath, B.H., Somatotyping ± Development and Applications, Cambridge Studies in Biological Anthropology, Vol. 5, New York, NY, 1990, p. 22. 6. Lohman, T.G. et al., Anthropometric Standardisation Reference Manual, Human Kinetics Books, IL, 1988. 7. Hertzberg, H.T.E. et al., Anthropometry of Flying Personnel ± 1950, Wright - Patterson Air Force Base, OH, 1950. 8. Brooks, S.M., Integrated Basic Science, The C.V. Mosby Company, London, 1979, p. 205. 9. Garn, S.M., ``Relative fat patterning: an individual characteristics'', Human Biology, Vol. 27, 1955, pp. 75-89. 10. Garn, S.M., Sullivan, T.V. and Hawthrone, V.W., ``Differential rates of fat change relative to weight change at different body sites'', International Journal of Obesity, Vol. 115, 1987, pp. 19-25. 11. Brady, R.J., A Programmed Approach to Anatomy and Physiology, 2nd ed., Prentice-Hall, Englewood Cliffs, NJ, 1972, p. 7. 12. Larsson, B. et al., ``Abdominal adipose tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow up of participants in the study of men born in 1913'', British Medical Journal, Vol. 288, 1984, p. 1401. 13. Clarys, J.P., Martin, A.D. and Drinkwater, D.T., ``Gross tissue weights in human body by cadaver dissection'', Human Biology, Vol. 56, 1984, pp. 459-73. 14. Becque, D.M., Katch,V.L. and Moffat, R.J., ``Time course of skin-plus-fat compress in males and females'', Human Biology, Vol. 58, 1986, pp. 33-42. 15. Mueller, W.H. and Stallones, L., ``Anatomical distribution of subcutaneous fat: skinfold site choice and construction of indices'', Human Biology, Vol. 53, 1981, pp. 321-35. 16. Martin, A.D., Ross, W.D., Drinkwater, D.T. and Clarys, J.P., ``Prediction of body fat by skinfold caliper: assumptions and cadaver evidence'', International Journal of Obesity, Vol. 9, Suppl. 1, 1985, pp. 31-9. 17. Lohman, T.G. and Pollock, M.L., ``Which caliper? How much training?'', Journal of Physical Education, Recreation and Dance, Vol. 52, 1981, pp. 27-9. 18. Covert, B., Fit or Fat?, Houghton Mifflin Company, Boston, MA, 1978, p. 48. 19. Tsang, B., Chan, C.K. and Taylor, G., ``Application of kinanthropometry and somatotyping to the study of the physique of Hong Kong-Chinese disciplined personnel [Part 1]'', Federation of Asian Professional in Textiles and Apparel, accepted in 1997. 20. Tsang, B., Chan, C.K. and Taylor, G., ``Application of kinanthropometry to the study of the physique of Hong Kong-Chinese disciplined personnel'', Proceedings of the Joint Conference of China Textiles Engineering Society and Hong Kong Institution of Textiles and Apparel, 1998, pp. 133-8. 21. Carrington, G.M., ``Australian army operational research group memorandum M3'', Anthropometric Survey of Male Members of the Australian Army, Part II Medical Survey, Sponsor DGMS, 1959. 22. Headquarters Army Inspection Service, Australia Army Anthropometric Survey ± Body Dimensions 1970, Design and Development (CLO) HQ AIS Sponsor MGO, 1970. 23. Hertzberg, H.T.E., Daniels, G.S. and Churchill, E., Anthropometry of Flying Personnel ± 1950, Wright-Patterson Air Force Base, OH, 1954. 24. McConville, J.T., College, A. and Alexander, M., Anthropometric Data in 3-dimensional Form: Development and Fabrication of USAF Height-Weight Manikins', Antioch College, Yellow Springs, OH, 1963.

Physique of disciplined personnel 157

IJCST 12,2

158

25. Oshima, M., Fujimoto, T., Oguro, T., Tobimatsu, N., Mori, T., Tanaka, I. and Watanabe T., (Aero-Medical Laboratory, Japanese Air Self-Defence Force), Alexander, M. (Aerospace Medical Research Laboratories), Anthropometry of Japanese Pilot, Behavioral Sciences Laboratory, Aerospace Medical Research Laboratories, Aerospace Medical Division, Air Force Systems Command, Wright-Patterson Air Force Base, OH, 1965. 26. Hebbelinck, M., Duquet, W. and Ross, W.D., ``A practical outline for the Heath-Carter somatotyping method applied to children'', in Bar-Or, O. (Ed.), Pediatric Work Physiology. Proceedings of the 4th International Symposium, Wingate Institute, Israel, 1973, pp. 71-84. 27. Hebbelinck, M. and Ross, W.D., ``Kinanthropometry and biomechanics'', in Nelson, R.C. and Morehouse, C.A. (Eds), Biomechanics IV, International Series on Sport Sciences, University Park Press, Baltimore, MD, Vol. 1, 1974, pp. 537-52. 28. Sheldon, W.H., Dupertuis, C.W. and McDermott, L., Atlas of Men, Harper & Brothers, Publishers, New York, NY. Appendix 1 Carter and Heath's anthropometric somatotyping method[5]. Steps: 1.

Record identification data in the top section of the form.

Start with endomorph rating 2.

Record measurements of the four skinfolds.

3.

Sum up the triceps, subscapular and supraspinale skinfolds.

4.

Circle the closest rating in the row of mid-point.

5.

Circle the endomorph rating directly under the circled rating in step 4.

Mesomorph rating 6.

Record the height, diameters of humerus and femur, girth of biceps and calf.

7.

Circle the nearest height in the height row.

8.

Circle the nearest figures for the measured bone diameter and girth.

9.

Calculate the sum of deviations of the diameters and girths from the height column (right ± positive, left ± negative) Mesomorph = (D/8) + 4

10. Circle the closest value on the mesomorph row. Ectomorph rating 11. Record weight. 12. Obtain height over cube root of weight and fill in the form, 13. Circle the closest value. 14. Circle the ectomorph rating directly below. Anthropometric somatotype 15. Record the circled values for each component in the ``anthropometric somatotype'' box. Appendix 2 Sheldon's[28] somatotyping method begins with photography. The photographic body measurements are taken on the photographs, excluding the subjects' weight and height (Figure A1).

Physique of disciplined personnel 159

Figure A1. Locations of the 17 diameters on the body Region I: Head, face and neck FB1 (facial-breadth-one) photographic diameter taken at the highest level of the junction of the pinna of the ear with the skin line of the head. (Frontal picture). FB2 (facial-breadth-two). Photographic diameter taken at the lowest level of the junction of the lobe of the ear with the skin line. (Frontal picture). Ntap (neck-thickness-anteroposterior). The shortest photographic anteroposterior diameter of the neck. (Lateral picture). Ntt (neck-thickness-transverse). The shortest transverse photographic diameter of the neck. (Frontal picture). Region II: Thoracic region TB1 (trunk-breadth-one). The distance between the uppermost visible points in the photographic lines formed by the posterior auxiliary folds. (Dorsal picture). TT1 (trunk-thickness-one). Horizontal anteroposterior photographic diameter of the trunk taken at a point midway between the level of the centre of the nipple and the most anteriorly projecting point of the sternoclavicular junction. (Lateral picture). TB2 (trunk-breadth-two). Minimum transverse photographic diameter taken at the narrowest level of the waist. (Dorsal picture). Region III: Shoulder, arms and hands ATU (Arm-Thickness-Upper) Anteroposterior photographic arm diameter taken at the level of the mid-point between the photographic centre of the cubital fossa, and a point on the skin overlying the greater tuberosity of the humerus and lying immediately beneath the anterior tip of the acromion process. (Lateral picture) ATL1 (Arm-Thickness-Lower-One) Photographic forearm diameter taken at the level of the greatest thickness below the elbow. (Lateral picture)

IJCST 12,2

ATL2 (Arm-Thickness-Two) Photographic diameter taken in a plane perpendicular to the axis of the forearm, at a level two inches proximal to a point on the skin surface immediately over the most anterior photographic projection of the styloid process of the radius. (Lateral picture)

160

Region IV: Abdomen trunk TT2 (trunk-thickness-two). Minimum horizontal anteroposterior photographic diameter taken at the level of the waistline. (Lateral picture). TTB3 (trunk-breadth-three). Maximum horizontal transverse photographic diameter taken at the waist level of the hips. (Dorsal picture). TT3 (trunk-thickness-three). Horizontal anteroposterior diameter taken at the level of a point on the body surface directly over the symphysis pubis.(Lateral picture). Region V: Legs and feet LTU1 (leg-thickness-upper-one). Horizontal anteroposterior photographic diameter of the leg taken at the level of the centre of the photographic angle formed by the subgluteal fold. (Lateral picture). LTU2 (leg-thickness-upper-two). Horizontal anteroposterior photographic diameter of the leg taken at the level of the photographic centre of the slight fossa or hollow seen immediately above the patella. (Lateral picture). LTL1 (leg-thickness-lower-one). Maximum photographic transverse calf diameter taken at the level of the greatest thickness of the calf, in a plane perpendicular to the axis of the lower leg. (Dorsal picture). LTL2 (leg-thickness-lower-two). Minimum photographic transverse ankle diameter taken at the narrowest point in the ankle. (Dorsal picture). Appendix 3 Sheldon's[28] photoscopic somatotyping method. Steps: 1.

Calculation of height/weight1/3 ratio (HWR).

2.

Calculation of 17 ratios transverse measurements to stature. The transverse measurements were: Four on the head and neck Three on the trunk Three on the arms Three on abdominal trunk Four on the legs

3.

Inspection of the somatotype photograph with reference of the HWR table, and recorded the estimated somatotype.

4.

Comparison of the 17 transverse measurement ratios with a range of scores for each ratio, to give a final score.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Mechanical fabric properties influencing the drape and handle I. Frydrych

Mechanical fabric properties

171

Institute of Textile Architecture, and Technical University of Lodz, Lodz, Poland

G. Dziworska

Institute of Textile Architecture, Lodz, Poland, and

A. CiesÂlinÂska

Technical University of Lodz, Lodz, Poland Keywords Drape, Drape handle, Friction coefficient Abstract The aim of this contribution is to answer the question if and how mechanical fabric parameters being components of a formability coefficient as well as mechanical parameters being components of fabric handle influence the fabric drape. Therefore, all these parameters were measured for ten fabrics and correlated with the drape coefficient.

Introduction Fabric drapeability and drape coefficient describe the way the fabric falls down, takes a special shape on a model or human body under gravitation. Each fabric takes different three-dimensional forms. Chu et al. (1950) and Cusic et al. (1965) contributed towards practical determination of fabric properties by measurement of the three-dimensional form of drape. An existing actual measuring device was created by them. Using it they determined a drape coefficient as a ratio of area of sample projection to the total area of examined sample. The digital value of the drape coefficient enables objective assessment of this fabric property. Drape is related to the rigidity. A low value of drape coefficient is identified with a fabric which deforms easily. Very stiff fabrics have a drape coefficient close to 100 per cent, whereas soft fabrics have one close to 0 per cent. Values of drape coefficients ranged from 30 per cent for fabrics of loose weaves to 90 per cent for rigid fabrics with tight weaves. The values of drape coefficients mentioned above correspond to Cusic's method (1965). According to the Polish Standard PN-73/P-04736 a drape coefficient Ku is also confined in the range 0-100 per cent; but for rigid fabrics its value is close to 0 per cent; whereas for flexible fabrics it is close to 100 per cent. The research on objective measurements of fabric properties was carried out in Japan to predict their behavior in processing and their ability to obtain and retain longterm given shapes as well as the clothes' aesthetic appearance. The KES-F system was invented for this purpose. We in our work used the Instron tester following Pang et al. (1993).

International Journal of Clothing Science and Technology, Vol. 12 No. 3, 2000, pp. 171-183. # MCB University Press, 0955-6222

IJCST 12,3

172

Characterization of tested fabrics The measurements were carried out for ten different woollen or wool-like fabrics designated for coats, costumes, or jackets. Fabric characteristics are shown in Table I. Determination of some mechanical and utility parameters of fabrics In our research the relationship between a drape coefficient and mechanical parameters of fabrics were examined, and we tried to answer the question, if the fabrics which have a good formability are characterized also by a good drape. We also paid attention to such fabric parameters, which are determined during the bending, tensile, compression, shear stresses and friction tests. Three-dimensional drape coefficient Three-dimensional drape was determined according to PN-73/P-04736. A measure of it is the value of drape coefficient Ku. It is expressed by a ratio of difference between sample surface and surface of sample projection and flat sample surface at the beginning of measurement expressed as a percentage. As was mentioned in the introduction, it is confined in the range 0-100 per cent, and its small value characterizes a good drape (just the opposite of Cusic's method (1965)). Three samples of diameter 200mm were cut from each flat fabric without any imperfections. The shape of projection was drawn on a paper and its area Fabric symbol

Table I. Fabric characteristics

Mass (m2)

A

336

B

362

C

258

D

413

E

293

F

401

G

281

H

526

J

311

K

310

Raw material (%) 80 20 50 50 70 30 100

wool elana wool polana wool polana wool

50 20 30 75 25 55 20 25 50 50 80 20

wool flax argona wool polana wool polana anilana wool anilana wool anilana

85 wool 15 polana

Warp linear density (tex)

Weft linear density (tex)

8461 26061 12562

8461 10061 12561

8461

8461

Panama

15061 15062 12561

26062 12561 12561

Plain

15062

15062

Twill

12561

12561

Plain

12563

12563

Plain

26061 12563 9461 10061

16061 12563 9461 10061

Weaves Combined stitches Combined stitches Twill Combined stitches

Twill

was measured by planimeter for each sample. On the basis of three Mechanical measurements the mean area of sample projections (m2) was found and a drape fabric properties coefficient was calculated from the following relationship: KU

r2 r2

s 100% r12

1

173

where: s = area of sample projection (m2), r1 = radius of circle maintaining the sample (r1 = 0.035m), r = sample radius (r = 0.1m). The values of drape coefficient for each of the ten fabrics are set out in Table II. Bending rigidity Bending rigidity was determined by Peirce's method (1930) according to Polish Standard PN-73/P-04631. For this purpose five samples of dimensions 306300mm, successively, in warp and weft directions, were cut from each kind of fabric. For each sample the length of slacking part of sample L (cm) was measured by the constant angle method. In that way 20 results of slacking length in weft and warp direction were obtained. The mean value of L in each (weft and warp) direction was calculated on their basis. Next, the so-called bending length was found from equation (2) for weft and warp direction: c

L cm 2

2

where: Å = mean value of a slacking sample length [cm], L also the unit bending rigidity B [mNmm] from equation (3): B 10 3 Wc3 gmNmm

3

where: W = fabric weight per square meter [g/m2], c = bending length [cm], g = acceleration of gravity [m/s2 ]. General unit bending rigidity was calculated as a geometrical mean value: p BOW BO BW

Ku (%)

A

B

C

D

66.6

76.4

93.3

72.2

Fabric symbol E F 86.7

71.2

4

G

H

J

K

88.9

76.7

86.7

90.9

Table II. The values of drape coefficient

IJCST 12,3

174

where: Bo = unit-bending rigidity in a warp direction [mNmm], Bw = unit-bending rigidity in a weft direction [mNmm]. Determined and calculated values of parameters of bending rigidity, Bo, Bw and Bow respectively, are presented in Table III. Initial tensile modulus The initial tensile modulus was determined based on Polish Standard PN-84/P04669. Initial tensile modulus C [N/mm] is defined as a ratio of force to elongation increment taking place in the initial linear shape of force elongation curve related to the sample width. In order to find the value of initial tensile modulus C, five samples (of the width 3cm), for which bending rigidity was assessed, were taken. Fabric force ± elongation curves were registered on the Instron tester at the following adjustment parameters: .

distance between jaws ± 100 mm,

.

speed of the cross-head ± 100 mm/min,

.

speed of paper ± 200 mm/min.

On the basis of these curves the strain increment, in the initial interval of force elongation curve, and force increment corresponding to this strain were read. Initial tensile modulus was calculated according to equation (5): C

F N=mm "b

5

where: b = sample width [mm] (b = 30mm) F = tensile force [N], e = strain. The values of initial tensile modulus in warp direction Co, weft direction Cw and general Cow are presented in Table IV. General initial modulus was calculated as an arithmetic mean value of Co and Cow.

Table III. The values of bending rigidity

BO [mNmm] BW [mNmm] BOW [mNmm]

A

B

C

D

76 52 63

35 33 34

23 11 16

38 71 52

Fabric symbol E F 21 15 18

48 42 45

G

H

J

K

16 15 15

67 44 54

26 19 22

15 10 12

Fabric formability Mechanical Fabric formability FF is defined as a ratio of bending rigidity B and initial fabric properties tensile modulus: B FF 10 3 mm2 6 C

175

where: B = bending rigidity [mNmm], C = initial tensile modulus [N/mm] On the basis of measurements of bending rigidity and initial tensile modulus fabric formability was calculated as before: . in a warp direction FFO, . in a weft direction FFW , . general FFOW, dividing general unit bending rigidity Bow by the general initial tensile modulus Cow These values are presented in Table V. Discussion of results Drape coefficient and parameters dealt with the fabric formability Examined fabrics differ from each other by fabric weight, weaves, warp and weft linear density. We did not observe a significant influence of raw material for the examined fabrics. Measured fabrics belong to the group of fabrics designated for coats and costumes, so they are characterized by similar raw material components (mainly wool and wool-like fibers). We tried to find the correlation relationships between the value of drape coefficient, and values of the mechanical parameters influencing formability and a formability coefficient (general as well as in warp and weft direction).

CO [N/mm] CW [N/mm] COW [N/mm]

FFo [mm2] FFw [mm2] FFow [mm2]

A

B

C

D

0.696 0.199 0.447

0.335 0.249 0.292

0.208 0.152 0.180

0.471 0.397 0.434

A

B

C

D

0.109 0.261 0.141

0.104 0.133 0.116

0.111 0.072 0.089

0.081 0.179 0.120

Fabric symbol E F 0.525 0.335 0.430

0.615 0.271 0.443

Fabric symbol E F 0.040 0.045 0.042

0.078 0.155 0.102

G

H

J

K

0.229 0.265 0.247

0.950 0.529 0.739

0.538 0.238 0.388

0.195 0.082 0.139

G

H

J

K

0.070 0.057 0.061

0.071 0.083 0.073

0.048 0.080 0.057

0.077 0.122 0.086

Table IV. The values of initial tensile modulus

Table V. The values of fabric formability

IJCST 12,3

176

Figure 1. The relationship between the drape coefficient and fabric weight

On the basis of results contained in Table I, where fabric characteristics are presented, and results contained in Table II, with values of drape coefficient, it was observed that a kind of weave and fabric weight influenced the drape coefficient. It was observed that fabrics of skew weaves and low weight have the high value of drape coefficient; for example, samples C, K, G have values of drape coefficient respectively 93.0 per cent, 90.7 per cent, and 88.9 per cent, i.e. they have much better drape ability than the other samples, all of them have skew weaves and they have much lower weight in relation to the rest of the samples (Figure 1). Fabrics of higher weight generally have worse drape (correlation coefficient r = 0.63). Influence of fabric thickness on the value of drape coefficient (correlation coefficient r = 0.34) was not observed. On the basis of Table III, it was observed that values of bending rigidity in warp direction are higher than the values of bending rigidity in weft direction (Bo > Bw) with only one exception (sample D). The situation is analogous for the values of initial tensile modulus (Table IV). The initial tensile modulus in the warp direction Co is higher than in weft direction Cw (Co > Cw); the exception is sample G. Bending rigidity as well as initial tensile modulus influence the fabric formability. In Table V values of formability are presented, and we can observe that formability in warp direction FFo is smaller than formability in weft direction FFW (with two exceptions for samples C and G, for which the relationship is just the opposite). It was also observed that these samples have high values of drape coefficient, respectively 93.0 per cent and 89.9 per cent. In the case of sample E formability values in weft and warp direction are more or less the same (FFo = 0.040, FFW = 0.045), and the drape coefficient takes also a high value (K= 86.7 per cent). Table VI shows that eight mechanical parameters are well correlated with the drape coefficient. Analysis of correlation coefficients between particular parameters and value of drape coefficient showed the trend rKBow > rKBw > rKBo > rKCo > rKFFow > rKFFw > rKW >. On the basis of the experiment carried out it was found that the highest correlation is between the drape coefficient Ku and general bending rigidity r = 0.9440 (Figure 2).

Correlation coefficient between the drape coefficient and appropriate mechanical parameter

Parameters Initial tensile modulus Bending rigidity Formability Thickness Weight

C0 Cw Cow Bo Bw Bow FFo FFw FFow T W

0.8260 0.4011 0.6019 0.8400 0.8966 0.9440 0.3033 0.8039 0.7177 0.3360 0.6300

Mechanical fabric properties

177 Table VI. Values of correlation coefficient between the drape coefficient and mechanical parameters of fabrics

The analogical relationships for weft and warp directions are also illustrated. It was stated that a better correlation between the drape coefficient and bending rigidity seen is in weft (r = 0.8966) than in warp direction (r = 0.8400) (Figures 3 and 4). An influence of initial tensile modulus in warp direction on the drape coefficient r = 0.8260 was also observed; whereas in weft direction the relationship was not so significant (Figure 5). Simultaneously, there is a relationship between the drape coefficient and general initial tensile modulus (Figure 6), r = 0.6019. The opposite situation is observed if we consider the formability influence on the drape coefficient. For general formability the significant correlation was stated, r = 0.7177 (Figure 7), but a much higher correlation is observed for formability in weft direction (r = 0.8039) (Figure 8). Drape and mechanical fabric parameters obtained from the Instron tester Because of the fact that systems for determination of mechanical fabric parameters under low stresses, which create big potential possibilities for clothes quality control, are unavailable (for example, KES or FAST) in Poland, the measurements of appropriate mechanical parameters were carried out on

Figure 2. The relationship between the drape coefficient and general bending rigidity

IJCST 12,3

178

Instron tester (Pang et al. 1993). A description of eight parameters determined on the basis of Instron measurement is given in Table VII, whereas their appropriate values are given in Table VIII. On the basis of results from Table VIII, it was observed that the fabric drape was influenced by the following parameters: . .

Figure 3. The relationship between the drape coefficient and bending rigidity in weft direction

Figure 4. The relationship between the drape coefficient and bending rigidity in warp direction

Figure 5. The relationship between the drape coefficient and initial tensile modulus in warp direction

RTow = general tensile recovery, rKuRTow = 0.7185 (Figure 9), MIUoo = friction coefficient in the arrangement warp/warp, rKMIUoo = 0.5902 (Figure 10),

.

.

MIUm = mean value of kinetic friction coefficient rKuMIUm = 0.6032 Mechanical (Figure 11), fabric properties 2HBow = general width of bending loop hysteresis, r2HBow = 0.9749 (Figure 12).

In Table IX, values of correlation coefficients between the respective parameters obtained from the Instron tester in all directions (warp, weft, general) are presented. Bold digits denote values of significant correlation.

179

Figure 6. The relationship between the drape coefficient and general initial tensile modulus

Figure 7. The relationship between the drape coefficient and general formability

Figure 8. The relationship between the drape coefficient and formability in weft direction

IJCST 12,3

180

Table VII. Mechanical parameters obtained from Instron tester

Table VIII. The mean values of mechanical fabric parameters obtained from the Instron tester

Conclusions The correlation relationships between the drape coefficient and mechanical parameters were examined for woollen and wool-like fabrics designated for coats, jackets and costumes. On the basis of the experiment carried out it was observed that the influence of applied fabric weaves and their weight W on fabric drape was significant, whereas the influence of fabric thickness T was not observed. The biggest influence on fabric drape was observed in the case of the following mechanical parameters: .

general bending rigidity Bow (especially in the weft directions),

.

general initial tensile modulus Cow (especially in the warp direction Co),

.

general fabric formability (especially in weft direction FFW).

Kind of test

Symbol

Parameters

Units

Tensile

LT WT RT 2HG 2HB RC T MIU

Linearity Tensile energy Tensile recovery Width of shear hysteresis loop Width of bending hysteresis Compression recovery Thickness Surface properties

± [cNmm] [%] [cN] [cN] [%] [mm] ±

Shear Bending Compression Friction coefficient

LTo LTw LTow WTo[cNmm] WTw[cNmm] WTow[cNmm] RTo[%] RTw[%] RTow[%] 2HBo[cN] 2HBw[cN] 2HBow[cN] 2GH [cN] RC [%] T [mm] MIUoo MIUwo MIUww MIUow MIUsÂr

A

B

C

D

0.888 0.856 0.871 32.79 13.98 22.52 50.4 37.1 43.1 14.8 22.4 18.6 1.02 42.6 3.71 1.08 1.20 1.12 1.11 1.12

0.840 0.908 0.874 28.20 9.08 18.64 53.60 41.70 47.60 9.70 13.8 11.8 0.46 37.5 3.86 1.02 1.02 1.04 1.11 1.05

0.893 1.059 0.969 47.63 14.53 32.58 51.4 49.9 50.7 3.6 3.7 3.7 1.74 38.4 2.73 1.30 1.30 1.31 1.33 1.31

0.962 0.881 0.926 25.69 16.12 20.11 42.9 42.3 42.7 11.0 14.1 12.6 0.96 40.6 3.84 1.15 1.19 1.23 1.19 1.19

Symbol E F 0.665 0.915 0.790 98.78 19.64 59.20 65.1 66.2 65.7 6.6 2.3 4.5 0.48 41.0 2.38 1.08 1.08 1.14 1.09 1.10

0.871 0.925 0.898 40.71 21.81 31.25 47.9 39.5 43.6 10.3 15.4 12.9 2.50 44.8 2.30 1.00 1.01 1.02 1.02 1.01

G

H

J

K

0.979 0.879 0.929 16.17 20.65 18.41 53.4 46.1 49.8 4.1 4.4 4.3 1.26 39.7 2.84 1.20 1.22 1.18 1.15 1.19

0.853 1.014 0.934 37.20 12.22 24.71 43.4 48.1 45.8 15.3 7.6 11.5 1.72 38.0 2.28 1.01 1.01 1.05 1.00 1.02

0.788 0.947 0.876 42.04 14.33 28.18 49.2 61.1 55.2 6.0 2.5 4.3 0.63 40.4 2.81 1.08 1.11 1.11 1.12 1.11

0.928 0.970 0.949 22.25 15.71 18.98 59.9 52.9 56.3 2.7 4.4 3.6 1.36 35.9 3.29 1.50 1.56 1.53 1.47 1.52

Mechanical fabric properties

181 Figure 9. The relationship between the drape coefficient and general tensile recovery

Figure 10. The relationship between the drape coefficient and friction coefficient in the following fabric arrangement (warp/ warp)

Figure 11. The relationship between the drape coefficient and mean value of kinetic friction coefficient

Figure 12. The relationship between the drape coefficient and general width of bending loop hysteresis

LTw LTow WTo WTw WTow RTo

1 0.09 0.81 0.88 0.03 0.83 0.47 0.68 0.66 0.04 0.29 0.17 0.40 0.16 0.41 0.42 0.46 0.32 0.38 0.30 0.10

1 0.5 0.15 0.25 0.14 0.19 0.32 0.19 0.27 0.54 0.46 0.41 0.48 0.50 0.31 0.20 0.24 0.32 0.17 0.53

1 0.69 1 0.13 0.25 0.66 0.99 0.47 0.51 0.38 0.67 0.46 0.67 0.19 0.06 0.09 0.33 0.14 0.24 0.58 0.23 0.42 0.25 0.07 0.50 0.54 0.25 0.50 0.31 0.41 0.01 0.51 0.25 0.34 0.01 0.22 0.19

1 0.39 0.23 0.14 0.20 1.00 0.18 0.25 0.39 0.57 0.49 0.26 0.06 0.11 0.07 0.07 0.14

RTw RTow 2HBo 2HBw 2HBow 2GH

1 0.53 1 0.66 0.57 1 0.68 0.84 0.93 1 0.14 0.59 0.53 0.63 1 0.37 0.43 0.84 0.76 0.73 1 0.29 0.53 0.77 0.76 0.90 0.96 1 0.14 0.38 0.40 0.44 0.08 0.13 0.12 0.31 0.18 0.25 0.28 0.38 0.49 0.47 0.57 0.10 0.46 0.35 0.13 0.53 0.40 0.21 0.39 0.20 0.31 0.71 0.41 0.57 0.27 0.35 0.10 0.23 0.60 0.26 0.42 0.05 0.08 0.17 0.15 0.51 0.33 0.43 0.22 0.36 0.16 0.28 0.69 0.34 0.51 0.02 0.04 0.05 0.02 0.42 0.23 0.33 0.23 0.54 0.70 0.72 0.90 0.92 0.97

1 0.26 0.49 0.08 0.05 0.06 0.01 0.08 0.10

RC

T

1 0.21 0.57 0.48 0.23 0.59 0.09 0.56

1 0.17 0.29 0.16 0.32 0.15 0.34

IJCST 12,3

182

Table IX. Correlation coefficients LTo LTw LTow WTo WTw WTow RTo RTw RTow 2HBo 2HBw 2HBow 2GH RC T MIUoo MIUwo MIUww MIUow MIUsÂr Ku

LTo

MIUoo MIUwo MIUww MIUow MIUsÂr Ku

1 0.97 0.62 0.97 0.38 0.65

1 0.58 0.95 0.35 0.51

1 0.63 0.61 0.51

1 0.41 0.60

1 0.41

1

Note: LT = linearity; WT = tensile energy [cNmm]; RT = tensile recovery [%]; 2HG = width of shear hysteresis loop [cN]; 2HB = width of bending hysteresis loop [cN]; MIU friction coefficient; RC = compression recovery [%]; T = thickness [mm]

For these mechanical parameters the values of linear correlation coefficients Mechanical were in the range from 60 per cent to 95 per cent. fabric properties Taking into account parameters obtained on the Instron tester, the fabric drape coefficient is influenced by tensile recovery RT, friction coefficient MIU and width of bending loop hysteresis. Further investigation will continue for fabrics woven from different raw 183 materials. References and further reading Chu, C.C., Cummings, C.L. and Teixeira, N.A. (1950), ``Mechanics of elastic performance of textile materials, Part V: a study of the factory affecting the drape of fabrics ± the development of drape meter'', Textile Res. J, Vol. 20, pp. 539-48. Cusic, G.E. (1965), ``The dependence of fabric drape on bending and shear stiffness'', J. Textile Inst., Vol. 56, pp. T596-T606. De Boos, A.G. and Tester, D.H. (1991), ``The FAST approach to improved fabric performance'', in Stylios, G. (Ed.), Textile Objective Measurement and Automation in Garment Manufacture, CSIRO, Division of Wool Technology, Australia, Ellis Horwood. Hu, J. and Chan Y.F. (1998), ``Effect of fabric mechanical properties on drape'', Textile Res. J., Vol. 68 No. 1, pp. 57-64 . Kawabata, S. (1980), The Standardization and Analysis of Hand Evaluation, The Textile Machinery Society of Japan, Osaka. Lindberg, J. and Co. (1961), ``Fabrics as structural material of clothing'', Textile Research Journal, Vol. 99. Pang, N., Zeronian, S.H. and Ryu, H.S. (1993), ``An alternative approach to objective measurement of fabrics'', Textile Research Journal, Vol. 63 No. 1, pp. 33-4. Peirce, F.T. (1930), ``Clothes handle as a measured value'', J. Textile Inst., Vol. 21, p. T377. Szklarek, R. (1990), ``Possibility of using new methods for assessment of fabrics designated for clothes'', Technik Wll oÂkienniczy, Vol. 2 (in Polish).

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,3

184

Objective hand measurement of materials used for disposable diapers Hiroko Yokura

Faculty of Education, The University of Shiga Prefecture, Otsu, Japan, and

Masako Niwa

Nara Women's University, Nara, Japan Keywords Wearing, Rating evaluation, Fabric, Wettability Abstract The objective hand measurement of the materials used for disposable diapers has been looked into, with consideration given to aspects of both dermatitis and comfort. The subjective hand of the commercially produced diapers both at the leg gatherings and at the center was assessed by mothers and female students. The mechanical and surface parameters of the diapers were measured with the KES-FB system. The equations that connect the material properties to the subjective hand were obtained by using the stepwise block-regression analysis method. It became clear that the hand of diapers could be predicted from the physical parameters, for which the calculated error was within the range of the standard deviation of the subjective hand value of each diaper. In a dry condition, it is expected that the hand of diapers could be satisfactorily evaluated by using only the surface and compression parameters.

Introduction Disposable diapers come in contact with the human skin, and consumers are concerned about whether they cause dermatitis. The water transport properties and handle (softness and smoothness) of diapers are considered to be related to both diaper dermatitis and wearing comfort. Water transport properties of incontinence pads, such as absorption capacity and strike-through times, have been studied in relation to wet comfort (Cottenden, 1988; Cottenden et al., 1998). The relationship between rewet performance of experimental pads and their construction of absorbent cores has been systematically studied (Hodgson, 1991), but little attention has been given to the hand of disposable diapers in terms of their surface and mechanical properties. A method for objective evaluation of the hand of men's suiting has been developed based on precise measurements of certain mechanical and surface properties (Kawabata, 1980). The objective hand evaluation system for suiting was applied to a group of nonwovens used for health and hygiene products (Kawabata et al. 1994). In our previous study, we extended our investigation to the objective hand evaluation of nonwovens used for the center of disposable diapers, where the urine is absorbed in use (Yokura and Niwa, 1996). We examined the measuring International Journal of Clothing Science and Technology, Vol. 12 No. 3, 2000, pp. 184-192. # MCB University Press, 0955-6222

The authors wish to express their thanks to Ms Makiko Seii of Shiga University for her technical assistance. This work was supported in part by a Grant-in-Aid for Scientific Research (project 11680108) from the Ministry of Education, Science and Culture of Japan.

methods for compression and surface properties of disposable diapers that have been applied to the objective evaluation of diaper hand, and confirmed the applicability of the objective hand evaluation system. We also clarified the properties of diapers estimated to have good hand under both dry and wet conditions (Yokura and Niwa, 2000). In this study, we propose the equations that connect the diaper properties to the subjective hand evaluation of diapers both at the center and at the leg gatherings, and discuss the prediction ability of the equations. Experiment The subjective evaluation of diaper hand We collected a total of 68 commercially produced diapers and pads on the following three occasions; sample group I: 1995, sample group II: 1996, sample group III: 1997. A total of 41 baby diapers, 24 adult diapers and three incontinence pads were collected. The subjective hand of the diapers both at the leg gatherings and at the center was assessed by mothers who were using baby diapers, and by female students. They were asked only to judge whether the hand was good or poor, based on the sensations from contact with the materials. They evaluated sample quality in random order using a scale from 1 (poor) to 5 (excellent). We examined the correlation between the individual scores and the mean scores of the subjective hand assessments, excluding evaluations by the few judges whose score had no significant correlation with the mean assessment score at the 0.05 level. Their hand assessments tended to be somewhat different from those of the majority of judges. We used the mean score of the subjective hand assessments as the total hand value (THVsub) of the diaper. The subjective hand of the diaper center was assessed under both dry and wet conditions (Yokura and Niwa, 2000). In the subjective evaluation of dry diapers both at the leg gatherings and at the center, test diapers were set up three-dimensionally, as they would be in actual use. The judges then felt the diapers and evaluated the hand of the materials. The procedure for the sensory test of wet diapers was as follows. A sample was cut from a diaper's center and put into a 7.5610.5cm2 cell, then moistened with 30g of 0.9 per cent NaCl solution (0.4g/cm2) using an injector (constant pressure 16kPa, mean flow rate 3ml/second). The temperature of the solution was 32. In a previous study we determined the volume of the solution for the wetting experiment (Yokura and Niwa, 2000). The wet diaper segments were placed in a conditioning box at 32-33 and 98-100 per cent RH during the sensory test. The judges put their hands into the box to evaluate the diaper's hand. They evaluated sample quality with the same procedure as that of the dry assessment. Physical properties of diapers The mechanical and surface parameters of the diaper were measured with the KES-FB system (Kawabata, 1980). In the case of the nonwovens used in the center of the diapers, we measured the compression, surface, and water

Objective hand measurement

185

IJCST 12,3

186

retention (rewet) properties (Yokura and Niwa, 1996). For the compression and surface measurements, we cut a sample from the center of a diaper and put it into a 7.5610.5cm2 cell. Table I shows the surface and compression parameters and the measurement conditions. Based on results from our previous study (Kawabata et al., 1994; Yokura and Niwa, 1996), we used a single wire sensor for surface friction and a 3.1cm diameter rubber ball indenter for compression. For the wet test, we moistened each diaper segment with 30g of 0.9 per cent NaCl solution (0.4g/cm2) using an injector (constant pressure 16kPa, mean flow rate 3ml/second), and measured the compression and surface properties after immersion in the solution for five minutes. We measured water retention according to the following procedure (Yokura and Niwa, 1996, 2000): a sample was cut from the diaper's center and put into a 7.5610.5cm2 cell, then moistened with 0.4g/cm2 of 0.9 per cent NaCl solution at 32 using the same procedure as described above. After the sample was immersed in the solution for five minutes, a filter paper was placed on the diaper segment, and pressed for 30 seconds under a pressure of 4.5kPa. The weight of the solution absorbed by the filter paper was measured and defined as rewet (Hodgson, 1991; Yokura and Niwa, 1996). In the case of the nonwovens used in the leg gatherings of the diapers, we measured the tensile, compression, and surface properties. A 3cm-wide sample was cut from the outer leg gatherings of the diapers. Table II shows the mechanical parameters and the measurement conditions. For compression and surface measurements, the sample was elongated 40 per cent in length before measurement to simulate wearing conditions. The contact area of the friction sensor was changed from the original 565mm to 10610mm. Parameters

Unit

Description

Measuring conditions

Compression

LC (±)

Linearity of compression/thickness curve

Indenter: 31mm diameter rubber ball

WC (mJ)

Surface

Table I. Measuring conditions for the properties of the materials used in the center of diaper

Thickness and weight

RC (%) MIU (±)

Compression resilience Coefficient of friction

MMD (±) SMD (micron)

Mean deviation of MIU Geometric roughness on a diaper placed parallel to the surface Thickness Mass/unit area

T (mm) W (mg/cm2)

Maximum compression force: 0.2N. Compression energy Rate of compression: 0.2mm/sec A single U-shaped wire 0.5mm in diameter and 5mm in length Measured distance: 20mm; Contact force: 0.05N; Velocity: 1mm/sec. Thickness at a 2mN force

Parameters

Unit

Description

Measuring conditions

Tensile

LT (±)

Maximum load: 1N/cm

Compression

WT (J/m2) RT (%) LC (±)

Surface

WC (J/m2) RC (%) MIU (±)

Linearity of loadextension curve Tensile energy Tensile resilience Linearity of compression/thickness curve Compression energy Compression resilience Coefficient of friction

Thickness and weight

MMD (±) T (mm) W (mg/cm2)

Mean deviation of MIU Thickness Thickness at a 50Pa pressure Mass/unit area

Objective hand measurement

Sample width: 3cm Sample length: 2.5cm Maximum pressure: 5kPa Rate of compression: 0.2mm/sec

Contactor: 20 parallel steel piano wires, 0.5mm in diameter and 10mm long. Contact force: 0.05N

Objective evaluation of diaper hand We used stepwise block-regression analysis (Kawabata, 1980; Draper and Smith, 1966), to derive an equation which connects the mechanical and surface properties directly to the THVsub of diapers, as follows: m X Zi 1 THV C0 il

where Zi represents the contribution of each physical parameter to THV, and can be calculated by equation (2): Zi Ci1 Xi

Mi1 i1 Ci2 Xi2

Mi2 i2

2

where C0, Ci1, Ci2: constant coefficients, Xi: the value of ith parameter, Mi1: population mean value of the ith parameter Xi for diaper materials, si1: population standard deviation of ith parameter Xi, Mi2: population mean value of the square of the ith parameter Xi, si2: population standard deviation of the square of the ith parameter Xi. The stepwise block-regression is performed as follows: variables are grouped into some blocks, each block corresponding to a material property. In the first step, THV is regressed with each of the variable groups separately to choose the block with the highest regression accuracy. The resulting regression error is then regressed with each of the remaining blocks in the same manner as the first step. The first and second regression equations are added to form a new regression equation where the two blocks are regressed. The same procedure is repeated until the last block is completed. The rank of the step also gives us information on the ranking of importance of the blocks into the THV (Kawabata, 1980).

187

Table II. Measuring conditions for the properties of the materials used in the leg gatherings

IJCST 12,3

188

Results Subjective evaluation of diaper hand The results of the subjective hand assessments are summarized in Table III. The mean values of the correlation coefficients R between individuals and the mean scores (THVsub) within each group were 0.83 to 0.54. This Table also shows the mean values of the standard deviation SD on the THVsub of each diaper to clarify the dispersion of evaluation among the judges for the hand assessment. Objective hand evaluation for materials used in the center of diapers The objective evaluation equation for the hand touch feeling of materials used in the center of dry diapers was derived by using the compression and surface properties, thickness, and weight of 46 samples in sample groups I and II. The other 19 samples in sample group III were used to investigate the prediction ability of the equation. The constant parameters of the equation for converting mechanical properties into THV for the materials used in the center of dry diapers are shown in Table IV. The first step was the surface properties and the correlation with THVsub was 0.68. Lower values of logMMD and SMD increase THV. The second step was the compression parameters. Larger values Sample group

Table III. Subjective hand assessments of the materials used in the diapers

Number of samples Number of judges (original) R SD

26

Center (dry) II III 20

19

I 26

Center (wet) II III 18

19

Leg gatherings I II and III 25

30

26 (31) 22 (23) 30 (30) 17 (20) 21 (25) 21 (21) 36 (40) 0.54 0.60 0.70 0.74 0.69 0.83 0.59 1.02 0.85 0.91 0.83 0.85 0.71 0.95

25 (30) 0.74 0.99

Notes: R: the mean value of correlation coefficients between individuals and the THVsub within groups; SD: the mean values of the standard deviation for diaper samples on the THVsub of each diaper

Surface (1st step) Compression (2nd step)

Table IV. Parameters for the THV equation of dry diapers. C0 = 3.0548

I

Construction (3rd step)

i

Xi

1 2 3 4 5 6 7 8

MIU logMMD logSMD LC logWC RC logT logW

C1i

C2i

0.2697 1.1759 ±0.0383 0.0284 ±0.0414 1.2039 ±0.2669 0.8325

±0.3511 1.3438 ±0.2384 ±0.0002 ±0.0651 ±1.0248 0.1967 ±0.7773

M1i

M2i

0.6742 0.5705 ±1.275 1.640 0.2463 0.0693 0.6943 0.4861 0.2403 0.0617 45.52 2,084 0.9631 0.9345 1.931 3.741

s1i 0.3406 0.1192 0.0927 0.0633 0.0629 3.4400 0.0836 0.1093

s2i

r(RMS)a

0.8055 0.677(0.46) 0.3053 0.0420 0.0911 0.768(0.40) 0.0305 319.83 0.1582 0.777(0.40) 0.4130

Notes: ar: correlation coefficient between regressed and experimental values; RMS: root mean square of regression error.

of WC and RC and smaller values of LC give better touch feeling. The third step was the weight and thickness, but this contribution is small. The correlation coefficient between the regressed THV and THVsub was 0.78. The accuracy of the regression increases with an increasing number of blocks, and the regression error RMS decreases to 0.40. The equation for the wet diapers was derived by using the rewet, compression and surface properties, thickness, and weight of 44 samples in sample groups I and II. The constant parameters of the equation for converting mechanical properties into THV for the materials used in the center of wet diapers are shown in Table V. The first step was the rewet properties and the correlation with THVsub was 0.79, showing a high correlation. The wet diapers with good hand showed small rewet values (Yokura and Niwa, 1996, 2000). The second step was the surface parameters. Lower values of logSMD increase THV. The third step was the compression parameters. Larger values of WC and RC and smaller values of LC give better touch feeling. The fourth step was the weight and thickness, but this contribution is small. The correlation coefficient between the THV and THVsub was 0.89, and the regression error RMS decreased to 0.40. The correlation coefficient between the THV and THVsub of the wet diaper showed relatively high value compared with that of the dry diaper. It is clear that the hand of the dry and wet diapers can be predicted from the physical parameters, for which the regressed error is within the range of the standard deviation SD on the THVsub of each diaper listed in Table III.

Objective hand measurement

189

Objective hand evaluation for materials used in the leg gatherings of diapers The objective evaluation equation for the hand touch feeling of materials used in the gatherings of diapers was derived by using the compression, surface and tensile properties and weight of 41 samples in sample groups I and II. The constant parameters of the equation for converting mechanical properties into THV for the materials used in gatherings of diapers are shown in Table VI. The first step was the compression properties and the correlation with THVsub was 0.69. The larger values of WC and RC and the smaller values of LC give better i

Xi

Rewet (1st step) 1 Rewet Surface 2 MIU (2nd step) 3 logMMD 4 logSMD Compression 5 LC (3rd step) 6 logWC 7 RC Construction 8 logT (4th step) 9 logW

C1i ±0.2777 0.4311 2.5587 0.2321 0.9740 0.0131 ±0.4191 0.1204 5.3093

C2i

M1i

M2i

s1i

s2i

±0.4125 0.4384 0.4273 0.4849 0.4758 ±0.3223 1.107 1.383 0.3951 0.9172 2.5355 ±1.165 1.371 0.1205 0.2901 ±0.4797 0.3676 0.141 0.0769 0.0542 ±1.0637 0.8270 0.6907 0.0819 0.1355 0.0614 0.0190 0.0090 0.0930 0.0105 0.5066 36.86 1,394 5.9253 434.46 ±0.1400 1.018 1.055 0.1380 0.2762 ±5.2124 2.688 7.225 0.0182 0.0979

r(RMS)a 0.791(0.52) 0.856(0.44) 0.878(0.41) 0.889(0.40)

Notes: acorrelation coefficient between regressed and experimental values; RMS: root mean square of regression error

Table V. Parameters for the THV equation of wet diapers. C0 = 2.9304

IJCST 12,3

190 Table VI. Parameters for the THV equation of materials used in leg gatherings. C0=2.8955

Compression (1st step) Surface (2nd step) Weight (3rd step) Tensile (4th step)

i

Xi

Ci1

Ci2

Mi1

Mi2

1 2 3 4 5 6 7 8 9

LC 0.8160 ±1.0108 0.5792 0.3400 logWC 0.4992 ±0.2533 0.1955 0.0560 RC 1.1083 ±0.7497 37.62 1,427 MIU 0.6433 ±0.6368 0.2569 0.0681 logMMD ±0.2749 ±0.0867 ±1.698 2.895 logW 0.1100 0.0603 1.008 1.028 LT ±0.1294 0.1314 0.0589 0.0036 logWT 0.6872 ±0.6669 0.2961 0.1042 RT 0.3956 ±0.3247 72.64 5,350

si1

si2

0.0668 0.1332 3.4404 0.0463 0.1073 0.1117 0.0100 0.1285 0.8512

0.0765 0.0535 257.16 0.0259 0.3614 0.2288 0.0012 0.0711 1233.7

r(RMS)a 0.690(0.52) 0.746(0.49) 0.779(0.45) 0.803(0.43)

Notes: a: Correlation coefficient between regressed and experimental values; RMS: root mean square of regression error

touch feeling. The second step was the surface parameters. Lower values of logMMD increase THV. The third step was the weight of materials used in gatherings of diapers. A higher value for weight gives better hand. The fourth step was tensile properties, but this contribution is small. The correlation coefficient between the THV and THVsub of the materials used in the gatherings of diapers was 0.80, showing a high correlation. The accuracy of the regression increases with an increasing number of blocks, and the regression error RMS decreases to 0.43. It is clear that the hand of the materials used in the gatherings of diapers could be predicted from the mechanical parameters, for which the regressed error is within the range of the standard deviation SD on the THVsub of each diaper listed in Table III. Discussion The 19 samples in sample group III were used to investigate the prediction ability of the equations. Table VII shows the mean scores of the subjective hand assessments (THVsub), the calculated hand values (THVcal), and the calculated hand values by using surface and compression parameters (THVcal-SC) as reference for the 19 samples in sample group III. The values of the correlation coefficient between the THVsub and the THVcal and the values of the root mean square of calculated error (RMS') are also shown in Table VII. The correlation coefficient between the THVsub and THVcal of dry diapers was 0.65, and the value of RMS' was 0.64. The correlation coefficient between the THVsub and THVcal of wet diapers was 0.86, and the value of RMS' was 0.55. The correlation coefficient between the THVsub and THVcal of leg gatherings was 0.84, and the value of RMS' was 0.54. It is clear that the hand of the materials used for the disposable diapers can be predicted from the physical parameters, for which the calculated error is within the range of the standard deviation SD on the THVsub of each diaper listed in Table III. In the dry conditions, we calculated THVcal-SC by using the first and second steps (surface and compression parameters) as reference for the 19 samples. The

No. 601 602 603 604 605 606 607 608 609 701 702 703 704 705a 706a 707 708 709 710a R' RMS'

THVsub 2.03 2.77 1.80 2.13 3.07 3.83 3.93 2.67 2.93 2.03 2.97 3.90 4.40 1.63 3.97 4.00 3.27 2.53 2.13

Center (dry) Center (wet) THVcal THVcal-SC THVsub THVcal 3.15 2.85 1.90 3.35 3.53 2.71 3.48 3.29 2.17 1.80 1.98 3.07 3.92 2.06 3.43 4.08 3.38 1.70 2.44 0.65 0.64

3.23 3.05 1.91 3.25 3.57 2.73 3.47 3.26 2.20 1.90 2.07 3.27 4.07 2.27 3.71 4.16 3.50 1.77 2.61 0.67 0.63

3.33 1.05 3.81 3.81 3.24 4.62 3.76 3.81 3.24 1.81 2.38 3.05 2.86 1.00 1.05 4.05 3.52 3.62 1.95

4.29 0.72 4.40 3.73 4.40 4.00 3.76 3.81 3.10 2.93 1.58 1.99 3.61 0.66 1.83 4.23 3.52 3.90 1.84 0.86 0.55

Leg gatherings THVsub THVcal THVcal-SC 3.76 3.12 2.60 2.68 3.00 4.02 1.96 1.72 3.88 3.04 1.28 2.36 4.56

3.87 2.96 2.28 2.94 2.93 3.97 2.88 1.90 3.62 2.07 2.40 2.64 4.10

3.72 2.73 2.15 2.84 2.61 3.96 2.96 1.93 3.44 2.28 2.30 3.33 4.28

3.96 1.84 4.28

3.26 2.07 3.90

3.07 2.83 3.92

0.85 0.51

0.77 0.62

Notes: R': correlation coefficient between the subjective THVsub and the calculated THVcal. RMS': root mean square of calculated error. aThese samples are incontinence pads, so they do not have the outer leg gatherings.

correlation coefficient between THVsub and THVcal-SC of the center of dry diapers was 0.67, and the value of RMS' was 0.63. The correlation coefficient between THVsub and THVcal-SC of the leg gatherings was 0.75, and the value of RMS' was 0.66. We expect that the hand of the diaper materials could be satisfactorily evaluated by using only the surface and compression parameters in a dry condition. Conclusion The equations that connect the material properties of disposable diapers to the subjective hand were obtained by using the stepwise block-regression analysis method. It became clear that the hand of diapers could be predicted from the mechanical and surface parameters, for which the calculated error was within the range of the standard deviation on the subjective hand value of each diaper. In a dry condition, it is expected that the hand of diapers can be satisfactorily evaluated by using only the surface and compression parameters. References Cottenden, A.M. (1988), ``Incontinence pads: clinical performance, design and technical properties'', J. Biomed. Eng., Vol. 10, pp. 506-14.

Objective hand measurement

191

Table VII. The mean scores of the subjective hand assessments (THVsub), the calculated hand values (THVcal), and the calculated hand values by using surface and compression parameters (THVcal-SC) as reference for the 19 samples in sample group III

IJCST 12,3

192

Cottenden, A.M., Thornburn, P.H., Dean, G.E. and Fader, M.J. (1998), ``Wet comfort of small disposable incontinence pads'', Text. Res. J., Vol. 68 No. 7, pp. 479-87. Draper, N.R. and Smith, H. (1966), Applied Regression Analysis, John Wiley and Sons Inc. Hodgson, K.T. (1991), ``Superabsorbent polymers and rewet performance of the internal layers of absorbent disposable products'', Tappi J., pp. 205-12. Kawabata, S. (1980), The Standardization and Analysis of Hand Evaluation, 2nd ed., The Textile Machinery Society of Japan. Kawabata, S., Niwa, M. and Wang, F. (1994), ``Objective hand measurement of nonwoven fabrics Part I: development of equations'', Text. Res. J., Vol. 64 No. 10, pp. 597-610. Yokura, H. and Niwa, M. (1996), ``Objective hand evaluation of non-wovens used for nappies'', International J. of Clothing Science and Technology, Vol. 9 No. 3, pp. 207-13. Yokura, H. and Niwa, M. (2000), ``Changes in disposable diaper properties caused by wetting'', Text. Res. J., Vol. 70 No. 2, pp. 135-42.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Fabric hand property of Polish linen fabrics for ladies' outerwear S. Kawabata

Fabric hand property

193

The University of Shiga Prefecture, Shiga, Japan

Masako Niwa

Nara Women's University, Nara City, Japan

R. Koztowsky and S. Manys

Institute of Natural Fibres, Poland

K. Nakano

The University of Shiga Prefecture, Shiga, Japan and

Takako Inoue

Sugiyama Jogakuen University, Nagoya Keywords Fabrics, Poland, Garments Abstract Recently, the Polish National Fiber Research Laboratory provided linen samples. In addition to these Polish fabric samples, we also collected linen fabric samples which were made in Japan and throughout Europe. We have investigated hand properties of various linen fabrics, and identification of the Polish linen fabrics quality from those of other linen fabrics. The fabric hand of Polish linen fabrics is unique, it possesses a hand just between wool-like and cotton-like fabrics. We have clarified that the Polish linen fabrics are suitable for hari-type and tailored type silhouette designs for women's wear, and the fabrics are well-suited for finishing garment appearance.

Introduction Linen fabrics are one of the most traditional fabrics for garments in Japan, and the most suitable fabric for the tropical climate in the Japanese summer. Only 30 years ago, linen fabrics were still popular for high class suit, underwear and traditional kimono (Japanese style garment) materials for summer use. In recent years, linen garments are very scarce in both men's and women's garment markets in Japan. This may be due to high cost of linen fibers and spread of polyester fabrics. In the markets, the major materials for men's suits and women's outer garments are wool, cotton, silk and polyester fabrics. Niwa et al. have investigated fabric hand character of various fabrics applied to ladies' garments (Koyama et al., 1991), and developed an equation for objective evaluation of fabric quality (Kawabata, 1991) and the equations for discriminating the optimum silhouette design for each fabric (Kawabata and Niwa, 1992; Niwa et al., 1998). Linen fabrics, however, have been almost excluded from the analyses, except for some investigation into the quality of men's linen suitings, which are based on linen and linen blend samples collected from 1985 to 1989 (Liu, 1993) in the Japanese market. The aim of this study was to obtain maximum wearing comfort from materials that have long been used to produce

International Journal of Clothing Science and Technology, Vol. 12 No. 3, 2000, pp. 193-204. # MCB University Press, 0955-6222

IJCST 12,3

194

men's suits for the hot climate of Japan. However, for women's outerwear, it has been difficult to obtain linen fabrics for ladies. Recently, the Polish National Fiber Research Laboratory provided linen samples for further study. In addition to these Polish fabric samples, we also collected new linen fabric samples which were made in Japan and throughout Europe and available on the Japanese market. We have investigated hand properties of various linen fabrics, and identification of the Polish linen fabrics' quality from those of other linen fabrics. Samples The linen fabric samples collected for this study are listed in Table I. All fabrics are for women's outerwear garments and the Polish samples are marked with PL and the other linen samples collected in the Japanese market with JM. These fabrics include top-weight fabrics for jackets, blouses, shirts, women's suits and other similar applications, and bottom-weight fabrics used for skirts, trousers, etc. As a reference, five populations of fabrics are shown in Table II. Fabric analysis Mechanical properties of all linen samples were measured by KES-F AUTO system (Kawabata, 1998; Kawabata et al., 1998) and the following characteristics were derived based on the mechanical properties. Objective evaluation method is shown in Table III. Result and discussion (1) Characteristics of linen fabrics in terms of fabric hand The primary hand values of all samples were derived based on their mechanical parameters by using equation KN202 (Kawabata and Niwa, 1984), then these primary hand values were substituted into the equation (Koyama et al., 1991; Kawabata, 1991) to derive discriminating parameters Z1 and Z2; they distinguish fabric hand by three categories, cotton-like, silk-like and wool-like

Samples Poland linen (PL) Linen fabrics collected

Table I. Linen samples for women's outerwear fabric

Log10T (T: Log10W (W: fabric Number fabric weight, thickness, of mm) mg/cm2) SD Mean SD Fiber samples Mean

Linen 100% Linen 100% Linen In Japanese rich market (JM) blend a

Log10B (B: bending stiffness, gf.cm2/cm) Mean SD

Log10G (G: shear stiffness, gf/ cm.deg) Mean SD

25a

1.251

0.1028 ±0.128 0.1081 ±0.651 0.1672 ±0.518 0.1116

16b

1.194

0.1037 ±0.195 0.0991 ±0.809 0.1854 ±0.497 0.1139

1.189

0.1119 ±0.200 0.1095 ±0.908 0.2484 ±0.461 0.1080

48c b

Notes: Plain weave (24), Twill (1); Plain weave (16); cProduced in Japan (23), Italy (16), Germany (4), Hong Kong (3), Switzerland (1), Austria (1). Linen 100% (16), Linen/Rayon (10), /Cotton (8), /Silk (5), /Others (9). Plain weave (41), Twill (2), Others (5).

Samples (use) Women's suita Women's thin dressb Men's linen suitc Men's summer suitd Men's winter suitd

Log10T (T: Log10W (W: fabric Number fabric weight, thickness, of mm) mg/cm2) samples Mean SD Mean SD 220 120 110 156 214

1.355 0.962 1.245 1.276 1.421

0.1270 0.1768 0.0739 0.0615 0.0591

±0.045 ±0.413 ±0.222 ±0.304 ±0.127

0.1693 0.2058 0.0851 0.0791 0.0797

Log10B (B: bending stiffness, gf.cm2/cm) Mean SD

Log10G (G: shear stiffness, gf/ cm.deg) Mean SD

±0.872 ±1.634 ±0.686 ±0.964 ±0.867

±0.075 ±0.373 ±0.218 ±0.066 ±0.014

0.2565 0.3592 0.2129 0.1081 0.1267

0.2099 0.3044 0.1638 0.1079 0.1287

Fabric hand property

195

Table II. Description of fabric Notes: aWool 100% and blend of wool and synthetic fiber; bCotton, silk, wool, polyester, new group population of the micro-fiber polyester; cLinen 100% and blend of linen and synthetic fiber; dWool 100% and blend properties of some of wool and synthetic fiber different populations Characteristic value

Applied equation

Fabric primary hand values

KN202 (Kawabata and Niwa, 1984) (applied to ladies' thin dress fabrics) Discriminating parameters Z1 and Z2 to Equation developed for discriminating fabric discriminate fabric hand type such as silkhand type for ladies' thin dress fabrics like, wool-like and cotton-like (Koyama et al., 1991; Kawabata, 1991) Discriminating parameters Z1 and Z2 to Equation developed for discriminating the discriminate suitable silhouette of ladies' optimum silhouette design for ladies' garment garments (Niwa et al., 1998) Fabric air permeability AR and heat Air permeability (Kawabata, 1987) and heat conductivity l conductivity (Kawabata, 1984)

Table III. Objective evaluation method

Figure 1. The chart of discriminating the fabric hand of cotton, wool and silk fabrics. The linen group is plotted on this chart. The discrimination parameters Z1 and Z2 are derived from fabric primary hand values

IJCST 12,3

196

hand. In Figure 1, Z1 and Z2 of all linen samples are plotted on the discriminating chart, where the three hand categories are shown by the three zones. It is seen that the Polish linen PL group falls in a zone just between woollike and cotton-like. Among the linen fabrics, the Polish linen PL group fabrics fell on the closer side to wool-like region, while the linen fabrics JM collected from the Japanese market were distributed toward the cotton-like side. (2) Optimum silhouette design for linen fabrics We have developed a discrimination formula (Niwa et al., 1998) that determines the optimum silhouette design for a garment from the basic mechanical properties of the fabric (tensile, bending, shearing properties, and fabric weight). We have also verified that the derived formula has a high hitting ability (Niwa et al., 1998). Figure 2 shows the results of the discrimination of the optimum silhouette design for linen fabrics using this discrimination formula. Nearly all of the linen fabrics were distributed within the hari (anti-drape) type and tailored type zones, indicating that the fabrics are suitable for these types of silhouette designs. Of the fabric samples used in this experiment, a 70 per cent rayon/linen blend fabric was shown to be suitable for drape type silhouettes. (3) Characteristics of hand of Polish linen fabrics Figure 3 shows the mean of the primary hand values of three groups of fabric. These primary hand values were calculated by KN202-LDY (Kawabata and Niwa, 1984) for translating fabric mechanical parameters into hand values applied to women's outerwear thin fabrics. A hand feature of the group of linen PL fabrics is that strong values for koshi and hari, and the higher value of softosa compared with other linen fabric groups, while the value of fukurami is

Figure 2. Fabric discriminating chart for the optimum silhouette design. Linen fabrics are plotted on this chart. The discrimination parameters are derived from fabric mechanical properties

Fabric hand property

197

Figure 3. Primary hand values of linen fabrics are plotted on the chart normalized to women's thin dress fabric population

similar to other linen groups. Figure 4 shows the primary hand value and total hand value THV of three of the linen fabrics, plotted on the hand chart for men's summer suiting population. The shadowed area is the high quality zone for summer suiting. Compared even with groups of men's linen suit fabrics, PL suit fabrics show strong koshi, hari and shari, while fukurami value is near average. The THV is 3.5 or higher, approximately 0.5s higher than average, indicating a high-quality fabric exceptionally well suited for use in men's summer suitings. (4) Mechanical property characteristics Figure 5 shows the average values of the mechanical properties of linen fabrics plotted on a women's suit chart by mean value of each population. A feature of linen fabrics is high B and 2HB, very low G, 2HG and 2HG5, large SMD, and

Figure 4. Hand values and THV of linen fabrics are plotted on the chart normalized to men's summer suiting population

IJCST 12,3

198

Figure 5. The mechanical properties of linen suiting fabrics are plotted on the chart normalized to women's suit population

relatively low MIU. These features appear more strongly in PL fabrics. These properties characterize linen-like property. As seen in Table II, this trend is most noticeable in the PL fabrics. The non-linearity of the PL linen fabrics in the compression force/thickness curve is remarkable (that is, less LC). This property shows that the fabrics give soft touch in the initial compression region. Mechanical properties of high THV (total hand value) and TAV (total appearance value) samples of the PL and JM fabrics evaluated by an expert of textile engineer are shown in Figure 6. The scale of the chart is normalized by the population mean and standard deviation of ladies' outerwear fabrics collected from the Japanese market. The fabrics were

Fabric hand property

199

Figure 6. The mechanical properties of the linen fabrics having both high THV and TAV (*: PL and *: JM). Solid line and broken line show the mean value of all samples of PL and JM respectively

made by various countries in the world. We call this chart ``Global ladies' outerwear fabrics''. In Figure 6, the linen PL suit and JM suit fabrics of relatively higher THV are plotted on this chart by * and * symbols respectively. Solid and broken lines are respectively mean lines of these two fabric groups. The mechanical properties of the higher quality linen fabrics are seen in this Figure. Figures 7 and 8 show a difference in the bending and shear stiffness (B and G) of PL and JM fabrics. For reference, the stiffness parameters of

IJCST 12,3

200

Figure 7. A feature of the mechanical properties of linen fabrics appears on high bending stiffness B and low shear stiffness G

men's suitings and ladies' outerwear fabrics are shown. Figure 8 shows also the difference in the bending hysteresis (2HB) and shear hysteresis (2HG5). A feature of the mechanical property of PL fabrics is high B and very low G, and similarly, hysteresis 2HB is high and 2HG5 is low. The features observed in the mechanical properties of the PL fabrics suggest to us that the fabrics are good quality fabrics. The high B contributes to the beautiful silhouette of garment of hari-type and tailoredtype. The lower shear rigidity G and hysteresis 2HG contribute to comfortable wearing and the forming of smooth and beautiful 3D curve of the garment. The features are the strong point of the PL fabrics; however, the bending-hysteresis is high compared with wool and wool/polyester women's suit fabrics population. This may cause a wrinkling of the garment when the fiber absorbs water in the high humidity condition or sweated condition. The smaller values of G and 2HG are, on the other hand, a minus factor from the bagging problem. This minus factors, however, are also a plus factor to emphasise linen-like feeling. (5) Total appearance value (TAV) of PL fabrics as men's suitings (Kawabata and Niwa, 1989) The PL suit fabrics exhibited the highest values for TAV as men's suitings among all of the comparison sample group populations. Figure 9 shows the average TAV of the PL suiting and the parameters relating to the prediction of men's suitings. As shown in Figure 9, the TAV is over 4.0 and

Fabric hand property

201

Figure 8. A feature of the mechanical properties of linen fabrics appears on high bending hysteresis 2HB and low shear hysteresis 2HG5

this predicts that a high-quality finish suit can be predicted. For the THV used to evaluate quality by hand, as shown in Figure 4, these fabrics showed high values as both hari-type and tailored-type fabrics. Calculating these values using conventional objective evaluation formulas used for men's suiting indicates that these fabrics can be given a high-quality, refined finish (high TAV). These results show suitability of PL to men's summer suitings. (6) Thermal property characteristics Effective thermal conductivity l and air resistance (inverse of air permeability) AR were measured by KES F7 Thermo Labo (Kawabata, 1984) and KES F8 air permeability tester (Kawabata, 1987) to evaluate the thermal properties that affect the coolness and bulk of a linen fabric. Figure 10 shows effective thermal

IJCST 12,3

202 Figure 9. TAV of PL fabrics are related parameters. This chart is normalized by men's suiting population. A snake zone shows the zone in which high quality suitings fall. This plot shows suitability of PL fabrics to men's summer suitings

Figure 10. Effective thermal conductivity l and air resistance AR

conductivity l and air resistance AR of linen fabrics compared with other fabric populations shown in Table II. Linen fabrics possess higher l and lower AR compared with other populations. The PL fabrics, however, strongly exhibited this tendency, clearly indicating that these fabrics are well suited to use in cool clothing for summer wear. This proves that linen is a cool fabric.

Conclusions Characteristics of Polish linen fabrics are as follows: (1) Bending rigidity B and bending hysteresis 2HB are both high; shearing rigidity G and shear hysteresis 2HG/2HG5 are both extremely low. (2) For primary hand values, both koshi and hari are high, and the fabrics also offer high shari and a suitable level of numeri and fukurami. (3) The fabric hand of PL fabrics is unique; it possesses a hand just between wool-like and cotton-like fabrics. (4) We have clarified that the PL fabrics are suitable for hari-type silhouette and tailored type silhouette designs for women's wear. (5) The THV is low for wool-based quality, but high for women's summer suitings. The fabrics show the highest value for the TAV (total appearance value), indicating that the fabrics are well-suited for finishing garment appearance. (6) The effective thermal conductivity is high, while air resistance is low. As such, heat and air pass easily through the fabric, making Polish linen a valuable fabric for use in cool summer clothing. References and further reading Kawabata, S. (1984), ``Development of a device for measuring heat-moisture transfer properties of apparel fabrics'', J. Text. Machinery Soc. Japan, Vol. 37, pp. T130-141. Kawabata, S. (1987), ``Development of an automatic air-permeability tester'', J. Text. Machinery Soc. Japan, Vol. 40, pp. T59-67. Kawabata, S. (1991), ``Shingosen fabric: a new generation of textile and fiber engineering'', Advances in Fabric Technology, Proceedings of the 19th Annual Conference, in conjunction with the HESC Group of the Textile Machinery Soc. Japan, The Textile Institute, New Zealand Section, p. 21-31. Kawabata, S. (1998), ``Development of the design guideline for manufacturing ideal fabrics for apparel fabrics'', Research Project Grant-in Aid for Scientific Research (A), Project Number 08555237, p. 22. Kawabata, S. and Niwa, M. (1984), ``Improvement in the objective evaluation of fabric hand for thin dress fabrics, Part 1: selection of the fabric deformation range in the measurement of mechanical properties'', J. Text. Machinery Soc. Japan, Vol. 37, pp. T113-21. Kawabata, S. and Niwa, M. (1989), ``Fabric performance in clothing and clothing manufacture'', J. Text. Inst., Vol. 80, pp. 19-50. Kawabata, S. and Niwa, M. (1992), ``Objective evaluation of the quality of ladies' garments'', International J. of Clothing Science and Technology, Vol. 4, pp. 34-44. Kawabata, S., Niwa, M. and Yamashita, Y. (1998), ``Recent developments in the evaluation technology of fiber and textiles, towards the engineered design of textile performance'', Proceedings of the 100 years of Modern Fiber Science, The Fiber Society, USA, p. 1-28. Koyama, Y., Niwa, M. and Kawabata, S. (1991), ``An analysis of the hand of SHINGOSEN weave'', The 20th Text. Tech. Symposium, The Text. Machinery Soc., Japan, pp. 136-40.

Fabric hand property

203

IJCST 12,3

204

Liu, C. (1993), ``A study of the objective evaluation of suiting appearance'', Master's thesis at Nara Women's University. Niwa, M., Nakanishi, M., Ayada, M. and Kawabata, S. (1998), ``Optimum silhouette design for ladies' garment based on the mechanical properties of a fabric'', Text. Res. J., Vol. 69, pp. 578-88. Further reading Nakanishi, M. and Niwa, M. (1989), ``Studies on the air permeability of clothing material: Part 1: Air resistance of clothing materials for the different end-uses'', J. Home Economics Japan, Vol. 40, pp. 797-804. Senoo, J., Yoneda, M. and Niwa, M. (1985), ``Study on heat conduction properties of clothing material: Part 1: Measurement of effective thermal conductivity of fabrics'', J. Home Economics Japan, Vol. 36 No. 4, pp. 13-22.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Assessment of the quality of ladies' garment fabrics ± a preliminary report T. Inoue

Assessment of quality

205

Sugiyama Jogakuen University, Nagoya, Japan

M. Niwa

Nara Women's University, Nara City, Japan

Y. Yamashita, Y. Minamide, D. Inoue, A. Ishikawa and S. Kawabata The University of Shiga Prefecture, Shiga, Japan

Keywords Garments, Fabric, Textiles Abstract In order to establish an objective method of evaluating ladies' garment fabrics by connecting the mechanical properties of ladies' garment fabrics to subjective evaluation, subjective assessments were examined by judges who work at textile mills or in the textile trade. We examined a method of deriving objective equations, a total hand value (THV) equation and a total appearance value (TAV) equation. The THV equation was derived directly from the mechanical properties of the fabrics and the TAV equation was derived from three mechanical parameters which are related to the beauty of garment appearance. In the case of the THV objective equation, the accuracy of regression was high within the same groups of judges; however, in the case of the TAV objective equation, accuracy was slightly low. Because there were few subjects and that caused deviation, the accuracy of prediction was slightly low; however, the objective evaluation was adequate.

Introduction The quality evaluation of women's garment fabrics is much more difficult than that of men's suiting fabrics. The fabric design varies greatly, and evaluation is strongly influenced by fashion trends, the preferences of the judges, and different criteria for quality from different points of view. We are confident that real quality factors related to fabric hand must exist which are not influenced by fashion, etc. We believe this because the relationship between garment materials and human feeling does not change with fashion trends, and this is indeed a factor of real quality. During the past five years, we have been conducting surveys to establish an objective evaluation method for ladies' garment fabrics, and this paper reports on the progress made in our recent investigation. Sample collection and subjective evaluation A total of 865 samples of women's dress fabrics were collected from all over the world from 1995 to 1998 (Geick-Mathews, 1998). First, these samples were divided into three groups categorized by suitable garment design according to silhouette (Niwa et al., 1998). These are tailored type, hari type, and drape type

International Journal of Clothing Science and Technology, Vol. 12 No. 3, 2000, pp. 205-213. # MCB University Press, 0955-6222

IJCST 12,3

206

Figure 1. Discriminated map of ladies' garment fabrics

silhouettes. Typical samples of tailored, hari and drape types, 280 samples in total, were sampled and clearly discriminated, as shown in Figure 1. A subjective evaluation was conducted for each group. The judges were asked to evaluate the THV and TAV. THV is the judgement of the fabric handle, the quality which comes from the suitability and comfortable-related properties of a fabric, and is related to fabric touch feeling and the ability to make a beautiful silhouette. TAV is the judgement of the predicted design that emphasizes the beauty of each silhouette type. THV evaluation was standardized following the standardization of the total hand, using ranking numbers 0 (not useful), 1 (poor), 2 (fair), 3 (average), 4 (good), and 5 (excellent) (Kawabata and Niwa, 1989). TAV evaluation was also standardized following the standardization, using ranking numbers 1 (poor), 2 (fair), 3 (average), 4 (good), and 5 (excellent). The first assessment was conducted in 1997 by eight textile experts from textile mills with over 30 years of experience, two garment designers, three people working in cloth trading, and four teachers from a textile-oriented school. There was a total of 17 judges, as shown in Table I. They evaluated the THV only. In the second trial, four experts from textile mills, three of whom also participated in the first trial, conducted evaluations of THV and TAV. In the first trial, a relatively high correlation was observed between the judges, as shown in Table II, but the correlation between the judges of the first and the second trials was relatively low, as shown in Table III. It was found in the second trial that the correlation between THV and TAV was very high, 0.91 for tailored type, 0.70 for hari type, and 0.83 for drape type. This suggests that the experts' evaluation of THV also covers TAV judgement, as shown in Table IV.

Number Evaluation for THV The first assessment 1997.5-10 Textile experts in textile mill Garment designers Cloth trading people Teaching staff in textile oriented school Total Evaluation for THV and TAV The second assessment 1998.12-1999.2 Textile experts in textile mill

Assessment of quality

8 2 3 4 17

207

4

Table I. Subjective judgement

The first assessment Cloth Teaching Experts Designer trade staff Tailored type (N = 83) Textile experts in textile mill (Sn = 8) Garment designer (Sn = 2) Cloth trading people (Sn = 3) Teaching staff in textile oriented school (Sn = 4) Total judges (Sn = 17)

1.000 0.310 0.501 0.526 0.801

1.000 0.847 0.748 0.763

1.000 0.844 0.889

1.000 0.892

Hari type (N = 108) Textile experts in textile mill Garment designer Cloth trading people Teaching staff in textile oriented school Total judges

1.000 0.344 0.478 0.584 0.774

1.000 0.727 0.711 0.771

1.000 0.835 0.878

1.000 0.929

Drape type (N = 89) Textile experts in textile mill Garment designer Cloth trading people Teaching staff in textile oriented school Total judges

1.000 0.475 0.544 0.517 0.825

1.000 0.880 0.783 0.845

1.000 0.847 0.899

1.000 0.867

Table II. Correlation coefficients of the subjective judgement of THV

Table III. Correlation coefficients between the first evaluation and the second evaluation of THV

Notes: N: Number of samples; Sn: Number of subjects

Subjects (textile experts in textile mill)

Tailored (N = 83)

Hari (N = 108)

Drape (N = 89)

S1 S2 S3

± 0.289 0.574 0.828

± 0.536 0.117 0.698

0.112 0.140 0.777

IJCST 12,3

208

Measurement of fabric mechanical properties The mechanical properties of the fabrics were measured with a KESF-B-AUTO (Kawabata, 1996/7; Kawabata et al., 1998). Tensile and compression properties were measured under high-sensitivity measurement conditions (Niwa et al., 1998). Tailored type fabrics were also measured under standard conditions. Next, the measurement results were compared with values for men's suiting fabrics, and the TAV and THV calculated. Results and discussion (1) Characteristics of ladies' garment fabrics The mechanical properties of these fabrics were plotted for each group on a ladies' garment fabric population chart. The scale of the horizontal axis is normalized by the standard deviation of each mechanical property of universal ladies' garment fabrics. The global mean value and standard deviation are shown in Table V. The three groups show different mechanical properties, and m1i and s1i are the population mean value and standard deviation, as shown in Table VI. In the first THV assessment, ten samples obtained a high mean score and ten samples obtained a low mean score for each silhouette type, as plotted on the ladies' global charts shown in Figure 2. The mechanical properties of the good samples and poor samples are clearly separated. In the case of tailored type fabrics, the mechanical properties of high THV are as follows. The LT is low, bending is soft, shearing is stiff, compression is soft, and the surface is smooth. In the case of hari type fabrics, the LT is low, tensile resilience is high, shearing criteria have optimum region, and weight per unit area and thickness are relatively low. For the drape type fabrics, the tensile property is high, bending and shearing is soft, the surface is relatively smooth, and weight per unit area and thickness are low. The tailored type fabrics were measured under standard conditions. The mechanical properties of tailored type ladies' garment fabrics are inside the high quality fabric zone for men's suitings; however, the hand value koshi is relatively low. (2) Derivation of the THV and TAV evaluation equations THV and TAV equations have been derived for each group on the basis of the second trial judgement. The rate of error in prediction is according to the spread of judgement. The men's THV equation translates primary hand values into total hand value; however, for ladies' garment fabrics, previous research suggested that the accuracy of prediction is high for the equation regressed directly with mechanical properties (Niwa et al., 1994; Kawabata and Niwa,

Table IV. Correlation coefficients between experimental THV and TAV

Tailored (N = 144) 0.908

Hari (N = 154)

Drape (N = 182)

0.697

0.831

Global fabrics (N = 280) m s

Mechanical properties Tensilea Bending Shear Compressiona Surface Construction

LT logWT RT logB log2HB logG log2HG log2HG5 LC logWC RC MIU logMMD logSMD logT logW

0.7470 ± 0.3800 65.6136 ±1.2247 ± 1.5157 ± 0.3304 ± 0.2640 0.0243 0.6508 ± 1.3790 49.5757 0.1884 ± 1.7044 0.5993 ± 0.2486 1.1521

0.0932 0.2622 10.3715 0.3700 0.4871 0.2489 0.4811 0.4582 0.0791 0.3512 9.7463 0.0427 0.2019 0.2506 0.2632 0.2242

a

Notes: High sensitivity measurement condition (Niwa et al., 1998)

1992). This assessment obtained the same result in 1998. In the case of THV, the regression was carried out by stepwise regression to derive a regression equation for each silhouette type. This stepwise regression method is a method by which all mechanical properties are grouped (blocked) and normalized by their population means, standard deviations, square means and its standard deviations for each silhouette type . The THV equation, New KN-305 equation, is shown in Table VI. New samples were prepared for the examination of the accuracy of regression equation, with 23 tailored type autumn/winter samples, 46 hari type samples and 93 drape type samples. In Table VII, the accuracy of regression is relatively high; however, the accuracy of prediction is slightly low. This result suggests that the dispersion was the result of using very few subjects, only four. TAV was derived from three mechanical components (Kawabata and Niwa, 1989): formability, elastic potential, and drapability, which are related to garment appearance. The stepwise regression method was also used in the deviation of the TAV, where, logEL is weft directional extensibility, logBS is effective bending stiffness in the weft bending mode, logSS is effective shear stiffness, logBP is bending potential per unit area at K = 2.5cm-1, logSP is shear elastic potential per unit area at F = 5ë, (BS/W)1/3 is stiffness affecting bending length, and (SS/W)1/3 is shear stiffness affecting bending length. These are the same parameters used for men's suiting fabrics. For ladies' garment fabrics, logEP (extension elastic potential) was added. A stepwise block-regression to three mechanical components was applied to the analysis of TAV. The TAV equation is shown in Table VIII. New samples were prepared for the examination of the accuracy of the regression equation, 52 tailored type

Assessment of quality

209

Table V. Global mean value and standard deviation

IJCST 12,3

210

Importance order

Xi

Tailored type C0 = 3.5028 1. Surface MIU logMMD logSMD 2. Shear logG log2HG log2HG5 3. Compression LC logWC RC 4. Bending logB log2HB 5. Construciton logT logW 6. Tensile LT logWT RT Hari type C0 = 3.0296 1. Tensile LT logWT RT 2. Construction logT logW 3. Shear logG log2HG log2HG5 4. Bending logB log2HB 5. Compression LC logWC RC 6. Surface MIU logMMD logSMD

Table VI. THV equation for evaluating total hand value of ladies' garment fabrics (equation new KN-305)

Drape type C0 = 3.4560 1. Bending logB log2HB 2. Shear logG log2HG log2HG5 3. Surface MIU logMMD logSMD 4. Compression LC logWC RC 5. Construction logT logW 6. Tensile LT logWT RT

C1i

0.0502 1.7915 0.2258 0.9112 0.0204 0.8530 0.3399 ± 0.7255 ± 1.6842 0.0535 0.1836 0.0252 2.2994 0.0125 ± 0.0314 0.1001 0.3807 ± 0.0807 0.8190 ± 0.0482 0.5458 0.1919 0.0381 ± 0.0660 ± 0.1114 0.5142 ± 0.2104 0.1254 0.1073 ± 0.0281 0.1951 0.1190 ± 0.9294 0.4346 3.7503 0.1789 ± 0.8023 0.3857 ± 0.0416 0.4006 ± 0.4556 ± 0.5648 ± 0.1490 0.1134 0.6907 0.2095 0.1447 ± 0.2542

C2i

± 0.2454 1.8632 ± 0.6492 0.7324 0.4314 ± 1.0872 ± 0.1836 ± 0.5564 1.7534 0.2081 0.1138 0.0183 ± 2.3269 ± 0.0318 ± 0.0824 ± 0.0437

m1i

The population constants m2i s1i s2i

Tailored type fabricsa (N = 83) 0.1884 0.0366 0.0334 0.0138 ± 1.8051 3.2743 0.1263 0.4449 0.5934 0.3887 0.1910 0.2574 ± 0.2504 0.0941 0.1773 0.0793 ± 0.0788 0.0581 0.2277 0.0765 0.1384 0.0625 0.2081 0.0825 0.6507 0.4259 0.0492 0.0682 ± 1.1019 1.2639 0.2232 0.4581 57.5503 3,343.2507 5.5857 644.5935 ± 0.9241 0.8857 0.1784 0.3041 ± 1.2685 1.6974 0.2970 0.7314 0.0058 0.0305 0.1745 0.0435 1.3714 1.8864 0.0757 0.2103 0.7201 0.5234 0.0703 0.1014 ± 0.2315 0.0802 0.1632 0.0749 68.4863 4,765.0586 8.6397 1,188.2139

0.5057 0.0424 0.4609 0.1268 0.7462 0.0610 0.2351 0.2024 0.0962 0.5164 0.1906 0.0649 0.1082 0.0440 0.2488 0.0999

Hari type fabricsa (N = 108) 0.8147 0.6711 0.0864 0.1377 ± 0.5285 0.3377 0.2416 0.3245 60.7863 3,811.5339 10.7931 1,414.2170 ± 0.2499 0.0943 0.1784 0.1274 1.1398 1.3287 0.1721 0.3587 ± 0.1752 0.0751 0.2108 0.0992 ± 0.0219 0.1993 0.4459 0.3453 0.3308 0.2426 0.3650 0.2709 ± 1.0958 1.2531 0.2288 0.5017 ± 1.2275 1.5891 0.2869 0.7277 0.6154 0.3843 0.0749 0.1104 ± 1.2844 1.6985 0.2210 0.6519 44.0219 2,015.1746 8.7884 918.5056 0.1797 0.0337 0.0370 0.0160 ± 1.6108 2.6443 0.2227 0.7214 0.6862 0.5490 0.2796 0.3691

± 1.2959 0.9984 3.7026 0.3193 ± 0.8872 ± 0.3164 0.0819 ± 0.4472 0.5456 ± 0.5192 0.0628 0.0980 ± 0.6906 ± 0.1935 0.1063 0.2172

Drape type fabricsa (N = 89) ± 1.6615 2.8028 0.2059 0.6779 ± 2.0962 4.4635 0.2637 1.0731 ± 0.5935 0.3595 0.0856 0.0933 ± 0.7305 0.6514 0.3433 0.5980 ± 0.4540 0.3098 0.3221 0.3398 0.1991 0.0425 0.0532 0.0297 ± 1.7242 3.0057 0.1807 0.6128 0.4993 0.2988 0.2224 0.2622 0.6939 0.4887 0.0851 0.1179 ± 1.7523 3.1311 0.2462 0.8210 48.8807 2,468.1306 8.8773 865.0149 ± 0.4841 0.2695 0.1875 0.1594 0.9626 0.9626 0.1896 0.3659 0.6901 0.4804 0.0645 0.0887 ± 0.3381 0.1859 0.2674 0.2194 68.7944 4,814.0781 9.0197 1,209.4424

± ± ± ± ± ± ± ±

± ±

R (RMS)

0.691 (0.589) 0.773 (0.518) 0.822 (0.469) 0.837 (0.457) 0.844 (0.448) 0.850 (0.442)

0.667 (0.477) 0.738 (0.433) 0.763 (0.416) 0.773 (0.410) 0.779 (0.404) 0.780 (0.402) 0.540 (0.534) 0.693 (0.458) 0.766 (0.409) 0.790 (0.392) 0.800 (0.384) 0.805 (0.378)

Notes: aHigh sensitivity coefficient; RMS: root mean square of regression P measurement condition; R: correlation m1i =1i C2i Xi2 m2i =2i where: C1i and C2i are coefficients of error. THV C0 16 i1 C1i Xi

the ith variable terms. Xi is mechanical property of the ith variable term. m1i and s1i are the population mean and standard deviation. m2i and s2i are square mean and standard deviation

Assessment of quality

211

Figure 2. Mechanical properties of ladies' garment fabrics evaluated as good/poor

IJCST 12,3

samples including spring/summer fabrics, 46 hari type samples and 93 drape type samples. The correlation between the subjective and objective evaluations is still lower than that of men's suiting judgements, as shown in Table IX; however, the prediction accuracy is gradually improving. Formability, elastic

212 Table VII. Correlation coefficients between experimental and regressed, predicted THV

Regression

R RMSa

Prediction

R RMSb

Hari

Drape

(N = 83) 0.646 0.571 (N = 52) 0.661 0.639

(N = 108) 0.653 0.624 (N = 46) 0.449 0.891

(N = 89) 0.674 0.577 (N = 93) 0.589 0.773

Notes: R: correlation coefficient; RMSa: root mean square of regression error; RMSb: root mean square of prediction error Importance order

C1i

Pi

Tailored type C0 = 3.4866 1. Formability logEL2 logBS2 logSS 2. Elastic potential logEP logBP logSP 3. Drapeability (BS/W)1/3 (SS/W)1/3 Hari type C0 = 2.8757 1. Elastic potential logEP logBP logSP 2. Formability logEL2 logBS2 logSS 3. Drapeability (BS/W)1/3 (SS/W)1/3

Table VIII. TAV equation for evaluating total appearance value of ladies' garment fabrics. Contribution of mechanical parameters to TAV

Tailored

Drape type C0 = 3.4007 1. Formability logEL2 logBS2 logSS 2. Drapeability (BS/W)1/3 (SS/W)1/3 3. Elastic potential logEP logBP logSP

± 0.2063 0.0652 0.3516 ± 0.0002 ± 0.2737 ± 0.0563 ± 1.3231 ± 0.6238

± ± ± ±

± ± ± ± ± ±

C2i

The population constants m1i m2i s1i s2i

R (RMS)

Tailored type fabrics (N = 83) ± 0.0254 0.6908 0.5260 0.22082 0.31033 0.3826 ± 0.8961 0.8530 0.22372 0.35822 0.604 ± 0.1540 0.0988 0.0459 0.19009 0.06684 (0.596) ± 0.0425 ± 0.3994 0.1870 0.16563 0.11583 ± 0.3474 ± 0.5211 0.3007 0.17056 0.15708 0.622 0.1084 1.0530 1.1390 0.17037 0.38758 (0.586) 1.2523 1.8600 3.5140 0.23365 0.95356 0.646 0.6076 3.8050 14.8000 0.57162 4.73830 (0.571)

Hari type fabrics (N = 108) 0.7685 0.9201 0.2892 0.1850 0.31839 0.20440 1.5275 1.2446 ± 1.0490 1.1980 0.31187 0.67251 0.552 0.4401 0.4607 0.2514 0.1473 0.29005 0.17706 (0.687) 0.0482 ± 0.1778 ± 0.7512 0.6222 0.24055 0.41286 0.3010 ± 0.1883 ± 0.7700 0.6676 0.27316 0.42432 0.640 0.2665 0.3378 0.9245 0.9471 0.30408 0.46496 (0.633) 0.0176 0.0850 2.0430 4.2960 0.35108 1.58590 0.653 0.5003 ± 0.4572 5.2330 29.5800 1.48110 19.4010 (0.624) 0.1474 1.3866 0.0877 1.6913 0.0778 0.0963 0.1298 1.2582

± 0.2312 ± 1.0496 ± 0.1121 ± 1.4913 ± 0.1490 ± 0.0686 ± ± 0.1431 ± ± 1.2663

Drape type fabrics (N = 89) 0.6140 0.4756 0.31388 0.36853 1.6950 2.9360 0.25299 0.83801 0.3493 0.1525 0.17463 0.10825 1.4220 2.0430 0.14314 0.42162 3.6750 13.6500 0.38483 2.79230 0.5045 0.3363 0.28602 0.31696 1.2370 1.5730 0.20507 0.50643 0.7799 0.6098 0.03944 0.06110

0.567 (0.642) 0.669 (0.581) 0.674 (0.577)

Notes: R: correlation coefficient; RMS: root mean square of regression error. P TAV C0 8i1 C1i Pi m1i =1i C2i Pi2 m2i =2i where: C1i and C2i are coefficients of the ith variable terms. Pi is mechanical property of the ith variable term. m1i and s1i are the population

mean and standard deviation. m2i and s2i are square mean and standard deviation

Tailored

Hari

Drape

Regression

R RMSa

(N = 83) 0.850 0.442

(N = 108) 0.780 0.402

(N = 89) 0.805 0.378

Prediction

R RMSb

(N = 52) 0.755 0.743

(N = 46) 0.589 0.564

(N = 93) 0.436 0.846

Notes: R: correlation coefficient; RMSa: root mean square of regression error; RMSb: root mean square of prediction error

potential, and drapeability contribute to TAV; however, the degree of contribution is different with each silhouette type. The formability component contributes greatly to the TAV of tailored type fabrics. Concluding remarks The TAV evaluation overlapped the THV evaluation. The correlation coefficient of TAV and THV is particularly high with tailored type and drape type fabrics. The prediction accuracy of the objective equation depends on the reliability of the subjective judgement. We have to extend the assessment to include garment designers and people working in the clothing trade in the next step of the study. References Geicke-Mathews, E.C. (1998), Elegance Paris, Automne/Hiver '95/96 (1995)-Printemps/EteÂ '98. Kawabata, S. (1996-1997), ``Development of the design guideline for manufacturing ideal fabrics for apparel fabrics'', Research Project Grant-in Aid for Scientific Research (A), Project Number 08555237, pp. 22-4. Kawabata, S. and Niwa, M. (1989), ``Fabric performance in clothing and clothing manufacture'', J. Text. Inst., Vol. 80 No. 1, pp. 37-41. Kawabata, S. and Niwa, M. (1992), ``Objective evaluation of the quality of ladies' garments'', International J. of Clothing Science and Technology, Vol. 4 No. 5, pp. 34-44. Kawabata, S. Niwa, M. and Yamashita, Y. (1998), ``Recent developments in the evaluation technology of fiber and textiles, towards the engineered design of textile performance'', Proc. of the 100 Years of Modern Fiber Science, The Fiber Society, USA, pp. 1-28. Niwa, M., Wang, F. and Kawabata, S. (1994), ``An analysis of the quality assessment of ladies' garment fabrics'', A Collection of Theses for the First CHINA International Wool Textile Conference, pp. 281-90. Niwa, M., Nakanishi, M., Ayada, M. and Kawabata, S. (1998), ``Optimum silhouette design for ladies' garments based on the mechanical properties of a fabric'', Textile Res. J., Vol. 69 No. 8, pp. 578-88.

Assessment of quality

213 Table IX. Correlation coefficients between experimental and regressed, predicted TAV

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,3

214

A structural theory of power-net applied to biaxial extension S. Kawabata, Y. Yamashita and Y. Endo

The University of Shiga Prefecture, Shiga, Japan Keywords Fabrics, Yarns, Stretching Abstract Power-net is an extensible fabric and has a net structure. The biaxial extension property of this fabric is analysed and compared with experimental results obtained by a biaxial tensile tester.

Introduction Power-net is a kind of warp knitted fabric and a distinctive feature of this fabric is its high extensibility in both warp and weft directions. The structure of the power-net is formed from four half-set threaded guide bars, the front two bars producing a net, the remaining two bars laying-in an elastomeric yarn. Although the insertion of elastomeric yarn is only in the warp direction, the extensibility of the power-net is high in biaxial directions. This biaxial extensibility is due to the special net-structure. The main uses of this power-net are supporters, ladies foundation garments etc., but its extended use is expected for the reinforcement of three-dimensionally complex structures of composites due to its high formability property. A phenomenological approach to the biaxial stress/strain relation of power net has been investigated (Kawabata et al., 1998) by applying the linearizing method. This paper presents a structural mechanics approach based on a structure model of the power-net. The structure model The structure is shown in Figure 1 and its simplified model in Figure 2. The net structure of the samples used in the experimental investigation is formed with nylon filaments, and warp insertion yarn is polyurethane filaments. The structure of the power-net can be seen in Plate 1.

International Journal of Clothing Science and Technology, Vol. 12 No. 3, 2000, pp. 214-219. # MCB University Press, 0955-6222

Biaxial extension Figure 3 shows the unit structure deformed by a biaxial extension. Yarn elements in warp direction and weft direction are called warp and weft respectively in this paper. When the fabric is stretched in warp and weft directions, the tensile forces in these directions are derived as follows: . l1, l2: stretch ratio applied to power-net in warp and weft directions respectively. . Ty1, Ty2: Yarn tension of warp and weft respectively . y: Inclination angle of warp after deformation . ly1, ly2: Stretch ratio of warp yarn and weft yarn respectively.

Structural theory

215 Figure 1. Structure of the power-net

Figure 2. Structure model

Plate 1. Structure of the power-net

IJCST 12,3

216

Figure 3. Deformation of unit structure

y1 1 =cos y2 2

L1 =L2 1 tan

1 2

Yarn tension T is derived by the yarn stretch ratio: Ty1 f1 y1 ;

3

Ty2 f2 y2

4

The equilibrium equation of forces is, Ty2 2Ty1 sin

5

By substituting (3) and (4) into (5), the angle y can be obtained by solving (5) for y. Tensile force of the power-net fabric is now obtained by F1 Ty1 cos ; F2 Ty2

6 7

F1 and F2 are the tensile force per warp and weft yarn respectively. General biaxial deformation When shear deformation is overlapped, the forces are derived as shown in Figure 4: ya 1 =cos a ;

Structural theory

217

Figure 4. General biaxial deformation

yb 1 =cos b yc 2

1 L1 =L2 tan a tan b 2

Relation between ya and yb. tan b tan a

2 tan

Tyb fb ya Tyc fc yc Tyb fb ya Tension of element a, b, and c are, equilibrium equation is Tya sin a Tyb sin b Tyc

8

Relation between ya and yb is shown in Figure 5: tan b tan a

2 tan

9

Tension of element a, b, and c are Tyb fb ya

10

Tyb fb ya

11

Tyc fc yc

12

IJCST 12,3

218

By solving (8) and (9) for ya, and yb, la lb and lc, and yarn tensions are derived. Yarn tension is not uniform. It changes from one to another. The average tensile forces are as follows: F1 F1a F1b =2

13

F2 F2a F2b =2

14

F12 F2a

F2b

15

where F12 is shear force per weft yarn. Experimental Angle y and tensile forces are shown in Figures 6 and 7 respectively when an equal biaxial extension (l1 = l2) is applied. Plate 2 shows the deformed net structure with the equal biaxial extension. Yarn property Warp is a composite of nylon and polyurethane filaments. The nylon filament is helical, and weft consists of two nylon filaments and more straightened compared with warp yarn. Extensibility of this yarn structure must be predicted precisely to predict the extensibility of power net structure.

Figure 5. Relation between ya and yb

Figure 6. Angle y under equal biaxial extension

Structural theory

219

Figure 7. Stress/strain relation under equal biaxial extension

Plate 2. Structure of the net by equal biaxial extension

Contribution of elastomeric yarn to tensile force is almost negligible; it contributes to forming helical structure of warp nylon yarn. Weft nylon yarn may be approximated to be rigid yarn in actual cases. Reference Kawabata, S., Niwa, M., Inamura, A., Inoue, M. and Yamashita, Y. (1998), Int. Conf. ``Textiles, engineered for performance'', UMIST, Manchester, April.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,4

Large deflection analysis using an energy method M.T. Grieûer

232 Received June 1997 Revised August 1999 Accepted August 1999

Continental Teves (formerly ITT Automotives), Frankfurt, Germany Keywords Energy methods, Fabric Abstract Alternative differential equations to the heavy elastica equations are given and discussed. These differential equations are used for the solution of large deflection problems of flexible strips such as fabric.

Introduction The use of the heavy elastica differential equation is well accepted and several approximations for special cases have been discussed in the literature (Timoshenko and Gere, 1961). The introduction of two new differential equations allows new ways of approximation to be explored. The analysis is based on the minimum energy principle. Potential and bending energy are given in a general form. The minimum energy solution is then calculated by applying the calculus of variations. The differential equations are calculated in the x/y-space as well as the s/y-space. Definition of variables A piece of strip of length L and width b is clamped as shown in Figure 1. For a strip with constant width b, the width does not play any role in further calculations; thus it is only necessary to consider the x/y-plane. As will be seen later, it is convenient to define a general bending curve according to Figure 1. The curve starts at the point (xo,yo) and ends at (xe,ye). The arc length is calculated from xo, a is the slope angle and g the gravity. At y = 0 the potential energy is defined to be zero. Energy equations For a linear relationship between the curvature k and the torque T, the total energy of potential and bending energy for a strip of unit width is then given by ZL etot 0

International Journal of Clothing Science and Technology, Vol. 12 No. 4, 2000, pp. 232-239. # MCB University Press, 0955-6222

1 2 w ys Bk s ds 2

1

The author would like to acknowledge the support of Mr A. Creed of the Department of Electronic Engineering of the University of Hull and Scott Gobrogge from Continental Teves (formerly ITT Automotives), Frankfurt, Germany.

Large deflection analysis

233 Figure 1. Definition of variables

where w is the weight per unit area and B the bending stiffness which is equivalent to the product of Young's modulus times the second moment of inertia. Going over from the s/y-space to the x/y-space the curvature k(x) is defined by kx

d y00 x ds 1 y02 x3=2

2

p The arc length ds is then replaced by ds 1 y02 x dx; thus the total energy in the interval [xo,xe] can be stated as 8 9 > Zxe > < =q B y00 x2 2 w yx etot 3 3 > 1 y0 x dx > 2 2 : ; 0 x 1 y xo Now the function y(x) has to be found which minimises the total energy etot. In addition to this the condition which defines xe ZL 0

Zxe p ds 1 y02 dx L

4

xo

has to be fulfilled. This is an isoperimetric problem which is solved with the calculus of variations. Solving the energy equations The problem is to find a function y(x) which satisfies the following conditions: the total energy must be minimised with the side condition that the length from the starting point to the end point equals the given length L. The optimisation problem with condition is transferred to a problem without condition. This is done by transforming equation (4) into:

IJCST 12,4

234

1 01 ÿ L

Zxe p 1 y02 dx

5

xo

multiplying with a constant factor m and adding to equation (3): 0 1 ) Zxe ( Zxe p 002 p By 1 1 y02 d x @1 ÿ wy 1 y02 d xA6 F 3 02 L 2 1 y xo

xo

This does not change the total energy, because the added part is zero. Equation (6) is rewritten to: Zxe B y002 p w y ÿ 1 y02 dx 7 F 3 2 1 y02 L xo

The optimisation problem is now in a form without any additional conditions and the function y(x) can be calculated as follows. According to the theory of the calculus of variations it is now assumed that the optimal solution y*(x) is known and a set of functions y(x) is constructed as follows: yx y x " ~yx 8 xo x xe ; j"j < "o

9

yÄ(x) is an arbitrary but exactly defined function. These functions y(x) must still satisfy the general conditions that the term eyÄ(x) vanishes at the endpoints xo and xe for all values of e. Furthermore, the first derivatives of yÄ(x) at xo and xe are required to be zero: ~yxo 0; ~yxe 0 10 ~y0 xo 0; ~y0 xe 0

11

Applying the calculus of variations leads to the implicit differential equation for the bent shape of strips: p y ÿ =w L y02 y00 y02 y ÿ =w L y00 02 ÿ 0w 1y 3=2 1=2 1 y02 1 y02 12 B 2 y0000 5 y003 20 y0 y00 y000 35 y02 y003 ÿ 7=2 9=2 2 1 y02 5=2 1 y02 1 y02 Equation (12) is the final differential equation which describes the bent shape of strips with constant bending stiffness B and weight per unit area w. In a similar

way the calculus of variations can be applied by transforming equation (1) into the s/y-space. The differential equation for the shape of bent strips in the s/yspace is given below: 0

1 y0000 4y0 y00 y000 y003 4y02 y002 B=w 1 ÿ y02 1 ÿ y02 2 1 ÿ y02 3

13

235

To calculate pairs of x/y a Jacobian matrix as given in equation (14) has to be applied: p 02 s cos a siny0 s dx 1 ÿ y ds ds 14 y0 s dy y0 s Both results (equations (12) and (13)) can be used in a similar way. Numerical results The numerical results of a fourth-order Runge-Kutta algorithm for a fixed/free end problem is given. Boundary conditions for standard equation in s/q-space The differential equation for bending with a free and a fixed end is governed by the following differential equation: 3 d2 s cos with W wL and s s ÿ W B ds2 L

15

with w the weight per unit area, B the bending stiffness, and L the total length. Boundary conditions for energy-I-equation in x/y-space The energy-I-model is given by the following differential equation: 0000

y

B ÿ w

ÿ

02 3

1y

ÿ ÿ 2 y y y 1 y02 ÿ y02 y y00 1 y02 02 00

5 y003 20 y0 y00 y000 35 y02 y003 ÿ 2 21 y02 21 y02

Large deflection analysis

16

The initial conditions at the free end for the second and third derivative of y are set to zero. The value y(0) can be chosen arbitrarily, because the bent shape of a heavy elastica is not dependent on the height of the initial starting point y(0). The initial slope y'(0) is varied until the boundary condition at the fixed end is fulfilled.

IJCST 12,4

To allow an equally distributed mesh of nodes the step size of Dx is replaced by the step size Ds along the arc length. The change in x is calculated from the change in s by the following Jacobian: 1 dx p ds 1 y02 x

236

17

Boundary conditions for energy-II-equation in s/y-space The energy-II-model is given by the following differential equation: y0000 ÿ

1 4y0 y00 y000 y003 4y02 y002 ÿ 1 ÿ y 02 ÿ 1 ÿ y 02 B=w 1 ÿ y 02 2

18

The initial conditions at the free end for the second and third derivative of y are set to zero. The value y(0) can be chosen arbitrarily, because the bent shape of a heavy elastica is independent of the height of the initial starting point y(0). Simulation parameter For all approaches the initial slope is varied until the boundary condition at the fixed end is fulfilled. The boundary condition at the fixed end is defined as j y0 L j < "

19

where e is set to 10-6. Figure 2 shows the numerical results of the differential equation for one end fixed horizontally and the other end free to bend under its own weight. The Figure shows an overlay of all three approaches. As expected, all lead to the identical solution. Convergence Figures 3-5 show the convergence of the different approaches for the W = 0.125, 27, and 216 as a function of the step size s. For small values the energy solutions show better numerical convergence than the standard approach. For bigger values the energy-II approach shows the best performance of convergence over the other approaches. Discussion of the differential equations Both differential equations (13) and (14) can be used as alternatives to the standard heavy elastica equation. Problems can occur if the denominator is close to or equals zero. It is known from the standard approach that angles of p/2 can occur for certain shapes such as folds. In the x/y-space an angle of p/2 corresponds to a slope of infinity and thus equation (12) cannot be applied. In

Large deflection analysis

237

Figure 2. Fixed/free end boundary conditions

Figure 3. W = 0.125 covergence

the s/y-space an angle of p/2 corresponds to slope of dy/ds of one. As can be seen from equation (13) for this special case the differential equation tends to infinity as well. One way of solving differential equations is by assuming that certain parts are negligible compared with others. Which approximations can be made are based mainly on intuition. For the standard differential equation for heavy elastica several authors, e.g. Hummel and Morton (1972), Bisshopp (1973), made

IJCST 12,4

238

Figure 4. W = 27 convergence

Figure 5. W = 216 convergence

suggestions as to how approximations for the exact solution can be obtained. The two new forms introduced here will allow a whole new range of approximations to be explored. An approximation to the energy-I differential equation was given for horizontally clamped fabrics by Grieûer and Taylor (1994). The new differential equations allow an entirely new field for the exploration of approximations for the shape of bent flexible strips. Conclusions This paper introduced two new forms of differential equation for the shape of bent flexible strips. Equations for potential energy and bending energy were stated. The shape which is given by the minimum energy solution was found using the calculus of variations. The solution is directly given in the x/y-space compared with the standard model, which is given in the s/q-space and thus

requires a Jacobian matrix to transfer the result into the x/y-space. A different way of representing the differential equation for the solution of the energy equations was derived in the s/y - space. These two new forms of differential equations will allow an entirely new field of approximation to be explored.

Large deflection analysis

References and further reading Bisshopp, K.E. (1973), ``Approximations for large deflections of a cantilever beam'', Quarterly of Applied Mathematics, Vol. 30 No.1, pp. 521-6. Grieûer, M.T. (1994), ``The modelling of fabric behaviour for automated handling'', PhD thesis, Department of Electronic Engineering, The University of Hull. Grieûer, M.T. and Taylor, P.M. (1994), ``The bending behaviour of fabrics: an energy approach'', EURISCON '94, Malaga, Spain. Hummel, F.H. and Morton, W.B. (1927), ``On the large bending of thin flexible stripes and the measurement of their elasticity'', Phil .Mag., Vol. 4, pp. 348-57. Timoshenko, S. and Gere, J.W. (1961), Theory of Elastic Stability, McGraw-Hill, New York, NY.

239

The current issue and full text archive of this journal is available at http://www.emerald-library.com

IJCST 12,4

240 Received July 1999 Revised January 2000 Accepted January 2000

Optimized garment pattern generation based on three-dimensional anthropometric measurement Tae Jin Kang and Sung Min Kim

Department of Fiber & Polymer Science, Seoul National University, Seoul, Korea Keywords CAD, 3D, Apparel Abstract An automatic garment pattern generation system has been developed for the threedimensional apparel CAD system. To substitute the garment fitting process, which requires lots of trial and error in the traditional pattern generation methods, we developed a new direct pattern generation method using body-garment shape matching process. In this method, we first generated a body model using three-dimensionally measured anthropometric data and transformed it to have a convex shape similar to that of a commonly used dummy model in garment design process. Then a typical garment model has been defined by measuring the surface information of a dummy model using stereoscopy and adjusting its shape considering the geometrical constraints of the underlying body model to obtain the optimum fit garment patterns. Finally, we developed a pattern flattening algorithm that flattens the three-dimensionally adjusted garment model into two-dimensional patterns considering the anisotropic properties of the fabric to be used.

International Journal of Clothing Science and Technology, Vol. 12 No. 4, 2000, pp. 240-254. # MCB University Press, 0955-6222

Introduction The apparel industry has long been viewed as the most skilled labor dependent industry and thus the cost saving in production through automation has been necessary for retaining the competitiveness in the market (Stjepanovic, 1995). Although many attempts have been made, some of which are in practical use for this purpose, most of those studies were focused either on the automation of relatively simple inexpensive processes or on the development of complementary tools designed for the already expert patterner, and therefore the benefit of automation is not so clearly visible to date (Collier and Collier, 1990). Our study has been focused on the automation of a basic garment pattern generation process, because we thought that the most important and difficult step in the garment production should be considered first for the automation. The pattern generation is the starting point of garment manufacturing, which is usually performed by highly skilled patterners. Traditionally, to make an optimum pattern, patterners gather anthropometric information of selected people by careful measurement and draft specific flat patterns based on accumulated knowledge and experience. Then they make a test garment and repeatedly adjust it on the body to acquire the desired fitness by trial and error method. Thus, according to the various skill levels and the inevitably subjective fitness evaluation criteria of the patterners, this procedure may take

considerable time to be satisfactory. However, the shortening of the Garment pattern development period has become one of the most important factors in the generation current manufacturing trend of diversified small-lot production or quick response system, and such a time-consuming process may be an obstacle to the further advancement of the garment industry (Ingham, 1983). In this study, we developed a more accurate and faster method of gathering 241 the necessary anthropometric data for the pattern design by utilizing the threedimensionally measured body model in comparison with the traditional gauge measuring or complicated gypsum modeling methods. Then we replicated a typical garment model by applying stereoscopy on a general-purpose dummy model which is usually used in pattern making, and adjusted the shape of the garment model to fit the human body. Finally, we developed a pattern generation system that flattens the adjusted garment model considering the anisotropic properties of the fabric to be used to acquire the optimum-fit patterns. Three-dimensional body measurement Body measurement for pattern generation In the traditional manual pattern generation method, body measurement has been a fundamental but very important step that required considerable time and skill for a patterner to make accurate patterns (Ingham, 1983). In order to make this process more precise and faster, we studied the direct acquisition method of necessary anthropometric data from a three-dimensionally measured body model. Although there have been many studies on the three-dimensional noncontact body measurement, we did not implement any of those up-to-date measuring techniques such as laser profile scanning method (Jones et al., 1989; Cuminato, 1996), since the measurement technology itself was not the main concern of this study but the quantitative human body data were important. Thus we simplified the data acquisition method by dividing the human torso model into 20 virtual cross-sections parallel to the floor from the neck to the thigh and measuring the shape of each section using a sliding gauge. Then we divided each cross-section shape into 60 sectors and obtained the radius by image analysis and reconstructed the body model in the cylindrical coordinate system using those data. Significant point detection on the body model For the body model to be of use in the pattern generation process, some significant points on the body such as neck, shoulder, or bust point can easily be located (Roebuck, 1995). In this study, we used the Fourier series expansion method to find the locations of those points objectively. The Fourier series expansion is often used for the equivalent expression of arbitrary periodic function or sequential data as a series of trigonometric functions as defined in equation (1) (Kreyszig, 1988):

IJCST 12,4

242

f x a0

1 X

an cos nx bn sin nx

1

n1

1 where a0 2

Z

1 f xdx; an ÿ

Z

1 f x cos nxdx; bn ÿ

Z

ÿ

f x sin nxdx

But for the cylindrical characteristics of body data, f(x) was defined rather as a sequence of radius than a continuous function and therefore it was very difficult to perform the integrations in the formula above. For this reason, we used the jump method to calculate the Fourier coefficients more easily, as described in equation (2). " # m m 1 X 1X an ÿ js sin nxs j0 cos nxs n s1 n s1 s " # m m 1 X 1X 0 js cos nxs ÿ j sin nxs bn n s1 n s1 s

2

where js is f xs 0 ÿ f xs ÿ 0 and j0s is f xs 0 ÿ f 0 xs ÿ 0 and where m is the number of radial data along the lateral and the longitudinal body profiles. Although the result of the Fourier series expansion became more and more similar to the original data set as n was increased, we set the value of n around 20 to reduce the small fluctuations along each profile as well as the overall amount of data. The result was fairly satisfactory, as shown in Figure 1. Once all the Fourier coefficients were calculated, cross-section shape of any height or lateral profile of any angle could be obtained using quite a small amount of data. Examining the shapes of transformed functions representing body profiles, we found that most of the significant points on the body were located at the maximum, minimum, or extreme points. For example, the bust section can be

Figure 1. Original and Fourier transformed profile of body model

found at the maximum point of the central lateral profile of the body and the Garment pattern bust point at the maximum point of the cross-sectional profile of the bust generation section. After the determination of significant points, there remained one more process to make use of the measured body model in the pattern generation. Because most garments, except some kinds of bespoke garments, are designed 243 rather to cover the concavity of the body than to closely fit on it, the body model needs to be modified to satisfy this relation (Jaffe and Relis, 1993). This can be done by successive removal of the points that have a ``concave'' relation with neighboring points on the body profile, as shown in Figure 2. Removed points are later rehabilitated on the newly generated convex profile. Figure 3 shows an example of the modified convex body model with 11 significant points determined. Modeling of a garment using stereoscopy Garment modeling for pattern generation Because the body model generated by the three-dimensional measurement is usually defined in the cylindrical coordinate system, direct pattern generation from such a model has been thought to be very difficult so far. There are two major reasons for this. The first is that the garment patterns are defined in the Cartesian coordinate system and therefore it is difficult to project the irregular surface defined in the cylindrical coordinate system into a flat one. The second is that, although it is easy to measure anthropometric data or locate significant points from the body model, it is still very difficult to gather information and map free curves on the body such as neck or arm hole contour (Heisey and Haller, 1988; Heisey et al., 1990a, 1990b). For this reason, we developed an indirect pattern generation method using a deformable garment model. By this method, a typical garment model was generated by stereoscopical reconstruction and subsequently modified to fit the underlying three-dimensionally measured human body. Finally, the adjusted garment model was divided into small panels and projected into the flat patterns. Preparation of typical dummy model Most of the former studies on automatic pattern generation had focused on the direct pattern drafting method for a relatively simple shaped garment such as skirts or sleeves. However, because of the limitation of those methods, rather complicated patterns such as bodices cannot be generated easily (Heisey, 1990). In this study, we selected the women's bodice for the garment model for Figure 2. Reformation of concave profile into a convex one

IJCST 12,4

244 Figure 3. Special point detection after reformation of convex body model

three-dimensional pattern generation because it would be thought more effective for automation if the difficult tasks, not only for computerized method but also for traditional manual method, are considered first. Except some special suits, the shape of a garment rather covers than closely fits the concavity of the underlying body; therefore a convex body model called ``dummy'' is usually used in the custom tailored garment production. For this reason, we also used a standard size dummy model to generate a garment model. Although the garment model can easily be generated using the measuring method introduced in the previous section, such a model might have some problems, because the garment model should differ from the body model in the sense that it must provide information for the subsequent flattening process. However, the model generated by the previously described method is defined in the cylindrical coordinate system and it does not provide any information for later processes. The meshes are generated on the dummy model and scanned by image analysis and spectroscopy to obtain the topological relations as well as approximate spatial coordinates of the intersecting points of the meshes, which can provide information for the flat pattern generation (Gonzalez and Woods, 1993; Russ, 1995). We divided the surface of the dummy model into eight panels and marked 50~60 rectangular meshes on each section with 3~4mm wide black marking tapes as shown in Figure 4. Although some cut lines, called ``dart'' in the garment industry, are necessary for a three-dimensional surface to be projected as flat, too many cut lines may

Figure 4. Schematic diagram of surface segmentation on dummy model

make the subsequent sewing process too difficult or impossible. Therefore we Garment pattern divided the surface using the inter-panel boundaries as darts to reduce the generation number of darts in the pattern flattening procedure (Roebuck, 1995). Reconstruction of the garment model The principles of stereoscopy used in this study can briefly be described as follows. If an object is captured by two CCD (charge coupled device) cameras placed so as to have some parallax angle at the same time, two slightly different images are obtained. Then the distances between the points on the object and the base plane can be calculated by geometrical analysis on the difference of the two images, as shown in Figure 5. Where the distances on the image planes of two cameras differ according to the position of the target points, if the distance between the camera and the center of object D is known, the normal distance between a target point and the base plane can be determined as follows. The coordinates of two cameras can be expressed as (± D tan y, ± D ) and (D tan y, ± D ), while the coordinates of d1 and d2 can be expressed as (d1 cosy, d1 siny) and (d2 cosy, d2 siny). Therefore the distances between a target point and the base plane can be determined using the coordinate of the intersecting point of two segments Ad1 and Bd2 which can easily be obtained by solving a simple Euclidean geometrical equation. In this study, because the acquisition of the overall topological relations and the approximate spatial locations of the mesh points were sufficient, some possible errors caused by the aberration of the cameras, which may lower the accuracy of exact geometry of the measured model, could be neglected. The principle of the stereoscopy itself is rather simple, as explained above, but the most difficult process of the stereoscopy is the extraction and the identification of the same two points on the two different images. We used the mesh vertices as the target points for such identifications. The cameras used in this study had the resolution of 6406480 pixels, and the captured images were converted into 256 gray scales for subsequent image

245

Figure 5. Schematic diagram of stereoscopy

IJCST 12,4

246

processing. But as usual in most processes which convert those images into binary ones, some noises such as shadows or light reflections on the raw images were not easily removed by conventional thresholding technique (treatment that keeps only the points of specific gray value range black and others white). Therefore we developed a special three-step filtering method as follows. (1) Step 1. Selective thresholding technique. Creates binary image by keeping only the pixels with more than four out of eight neighboring pixels of lower gray value than specified as black to detect the relatively dark marking lines. (2) Step 2. Undetected marking line retrieval. Resets the discarded and whitened pixels that have had relatively dark neighboring pixels on the original image into black ones. (3) Step 3. Maintaining the connectivity of marking lines. Connect the locally broken marking lines by series of dilation and erosion processes. The above processes could be summarized as shown in Figure 6. After those processes, we used the image skeletonization technique to attenuate the image into one pixel width lines, preserving the topological relations of the original image to make the determination of the mesh crossing point coordinates possible. After the skeletonization process, the mesh crossing points could be determined by 363 kernel searching method as usual in image processing. Finally, by finding a unique four-step recursive path for each crossing point along the connected edges, a counter-clockwise four-noded element system that was essential for the point identification procedure could be organized, as shown in Figure 7. At last, by specifying one of the corresponding point pairs, the topologically same points on the two different images were successfully identified by neighbor searching algorithm using the information provided by respective element system and we could obtain the absolute geometry of the garment model using stereoscopy. In this study, we developed a general purpose model manipulator to perform all the image filtering processes mentioned above and assemble separately generated multiple panels into one complete garment model in the three-dimensional space.

Figure 6. Schematic diagram of image processes

Garment pattern generation

247 Figure 7. Determination of mesh crossing point coordinate system

To provide information for the subsequent body-garment shape matching process, we determined the positions of some significant points on the garment model corresponding to those on the body model generated in the previous section. In addition, as there were some special boundaries on the garment the shapes of which should be preserved, such as the neck or armhole lines, we found and stored their data by noticing the fact that the points on those boundaries tend to have different relations from neighboring nodes from inner points. Development of the flat pattern generation system Relation between the garment and the body model As mentioned in the above section, most garments, except some special purpose product such as foundation inner wear, are made to convex shapes to cover the concavity of the body rather than to closely fit on it. Therefore we used the convex dummy model in the shape matching between the garment and the body model by modifying the geometry of the garment model close enough to reflect the geometry of the underlying body. For this, we used the information of the significant points obtained by the Fourier series expansion method and the important boundaries defined for the garment model in the previous section. Shape matching between the body and the garment model Figure 8 shows the overall procedure flow of the shape matching process and the detailed explanations of each step are as follows. First, the body and the garment model were arranged properly in the threedimensional space using equation (3): NG NB 1 X 1 X gi0 gi C; where C gi ÿ Bi 3 NG i1 NB i1 in which Bi, Gi, gi respectively denote the position vectors of the significant points on the body, corresponding points on the garment, and every point on

IJCST 12,4

248

Figure 8. Procedure flowchart of the shape matching process

the body, and NB, NG denote the number of the significant points on the body, and on the garment respectively. C is the difference vector between the center positions of garment and body model which can also be regarded as the displacement vector required for the garment model to be properly aligned to the body. However, the resulting arrangement may seem abnormal, as shown in Figure 9(a), because the initial shapes of the two models had been different and therefore some points on the garment were placed inside the body. To maintain a normal arrangement condition, all the points on the garment inside the body should have been expelled out of the body, as shown in Figure 9(b). Among different methods of expelling those points, we selected a method that took the direction of the sum vector of the normal vectors of neighboring elements as the expelling direction. In this method, points inside the body were moved along the expelling direction until all points were placed out of the body.

Figure 9. Alignment and separation of body and garment model

The shape difference between the last modified garment model and the one that Garment pattern had been one step before was calculated using equation (4): generation N P 4 P Lij ÿ L0ij D

i1 j1

N P 4 P i1 j1

100

4

Lij

249

where N = no. of elements, Lij = length of jth side of ith element (0 denotes the current shape). If the difference was larger than a specified tolerance value (set to 1 per cent in this study), the shape modification process was repeated until the difference became smaller than specified; otherwise a final garment shape was determined after adding some ease (about 3~5mm) to be separated from the body. The shape modification process consisted of three steps, as described below. Step 1. Shrinking of the garment. To have an optimum shape, the garment should be in close contact with the underlying convex body model. But in the early stage of the shape matching process, there might be some gaps between the body and the garment due to the different initial geometries. Such gaps can be decreased by repeated garment shrinking followed by the separation process explained in the previous process to make the garment closer to the body surface. The garment shrinking process reduces the overall volume of the garment model by moving all the points on the model toward the volume center by some amount (which is arbitrarily defined as 5 per cent in this study). Step 2. Shape maintenance of the significant boundaries. A special treatment was required for some significant boundaries on the garment such as the neck, arm hole, center, or waist line to maintain their shapes as linear or elliptical, as shown in Figure 10.

Figure 10. Boundary shapes to be maintained during shape matching

IJCST 12,4

250

Since the special boundaries had been defined as the sequence of the nodes having at most three neighboring nodes, the repeated garment shrinking together with the edge smoothing process shown in Figure 11 could maintain the shapes of those boundaries. Step 3. Smoothing of the garment surface. The repeated shrinking and smoothing processes caused some uneven knolls on the garment surface, which were responsible for the abrupt buckling between the elements, and therefore they should be tempered to get a smooth surface. The Laplacian smoothing technique should be used for this purpose. Actually, this is a method that relocates all the nodes in a mesh system one after the other until all the nodes are placed at the center of their neighboring nodes connected by edges. We modified this method to prevent the abrupt change in the interelement angle or buckling, by relocating all the nodes in the mesh to the midpoint of their original position and the central point of all the neighboring connected nodes in the three-dimensional space, as shown in Figure 12. Figure 13 shows the final modified garment shape with some ease added. Projection of the garment model into flat patterns Once a garment model that closely reflects the shape of the underlying body model has been generated, it should be projected into the flat patterns for subsequent production. Although many works had been done on this topic, there were some problems in the application of those flattening methods to the garment pattern generation. One problem was that there had been some desired specifications for the garment pattern generation. For example, the fewer the number of darts for the projection, the easier the subsequent sewing process became. Another was that, because a garment was usually made of fabrics that are anisotropic and less resistant to the shear deformation, flattening methods that had been designed for isotropic continua such as paper or sheet metal are not suitable. To solve those problems, we divided the surface of the dummy model into eight panels (assuming that each panel can be perfectly flattened on a plane) to use the panel boundaries as darts and

Figure 11. Schematic diagram of boundary shaping

Garment pattern generation

251 Figure 12. Schematic diagram of Laplacian smoothing

Figure 13. Final shaped garment with ease

suggested two criteria that can reflect the elastic as well as the shear behavior of various fabrics. The pattern projection method used in this study can briefly be described, as below. When projected flat, the shape of each element on the divided panels would inevitably be more or less distorted. At this, if the lengths of four sides and the diagonal of all the elements in the panel were repeatedly adjusted to the original length as used to be on the three-dimensional garment model, the flattened panels could keep their shapes practically the same as those on the garment. For the conditions which could stop this otherwise permanent adjustment process, we defined two criteria. One was the elastic allowance calculated using the differences in the side length and the other was the shear allowance calculated using the differences in the corner angles of the elements between the original and the projected panels, as equation (5): N X 4 L2D ÿ L3D X ij ij % AModulus 100 L2D ij i1 j1

AShear 100

N X 4 2D ÿ 3D X ij ij i1 j1

2D ij

%

5

IJCST 12,4

252

Figure 14. Clustered front and back patterns of bodice

Figure 15. Example of resulting patterns after rearrangement of dart lines

where N is the number of element, Lij is the length of jth side on ith element, and yij is the size of jth angle on ith element. Therefore, by repeating the flattening process until those two values become smaller than some specified allowance values, different shape of patterns can be obtained for the fabrics of various mechanical properties. In this study, we projected all the panels and grouped them into two patterns, which consisted of four panels sharing one corner node in common, as shown in Figure 14, and used the interpanel boundaries as darts. However, the patterns shown in Figure 14 were still somewhat unsuitable for a typical garment pattern because of too many darts. Therefore the number of darts should be reduced as much as possible, observing the constraint conditions. Through many trials and experiments in this study, we found that the dart between the two panels below the bust line and the dart between the upper and lower two central panels were dispensable, and the remaining two darts toward the neck line and the side line were fairly sufficient for pattern projection, as shown in Figure 15.

Conclusions Garment pattern So far the traditional pattern generation has been a time-consuming and highly generation skilled labor dependent process because the anthropometric measurement by gauge or gypsum method had been very difficult and the subsequent flat pattern generation had depended heavily on the intuition and experience of the patterner. In this study, we developed a method that can obtain various 253 anthropometric data from a three-dimensionally measured body model by the Fourier series expansion method and made a garment model replicator using image analysis and stereoscopy. Then we developed an algorithm that adjusts the shape of the garment model to fit the underlying body model to get an individually optimum fit garment. Finally, we developed a flat pattern generator which projected the garment into flat patterns that can easily be sewn into a garment, considering the properties of the fabric, by suggesting two criteria concerns, the elastic and shear properties, respectively. Considering the fact that the patterns generated by this method look quite different from those drafted by the manual method, we could find that it might have been difficult to express the complex geometry of optimum garment by traditional method only. The automatic three-dimensional pattern generation will be necessary for a garment CAD system to be a more effective tool not only in mass production but also in the custom tailoring production of garments. Further development in the automatic pattern generation can be made by the advancement of the non-contact body measuring system, data accumulation of various garment models, and the vast amount of experiments for the determination of the coefficients used in the pattern flattening process. Moreover, if an objective evaluation method for garment fitness were developed to replace the traditional evaluation method, now depending on subjective human feeling, through continued research, the overall garment development period can dramatically be shortened and subsequently the garment industry would have more competitive power toward the future trends of the industry, the diversified small lot production or quick response system. References Collier, B.J. and Collier, J. (1990), ``CAD/CAM in the textile and apparel industry'', Clothing and Textile Research Journal, Vol. 8 No. 3, pp. 7-12. Cuminato, D.F. (1996), ``Development of an automated body measurement system'', Master's thesis, University of Austin, Texas, USA. Gonzalez, R.C. and Woods, R.E. (1993), Digital Image Processing, Addison-Wesley Publishing Company, Reading. Heisey, F. and Haller, K.D. (1988), ``Fitting woven fabric to surfaces in three dimensions'', J. Textile Inst., No. 2, pp. 250-63. Heisey, F., Brown, P. and Johnson, R.F. (1990a), ``Three-dimensional pattern drafting ± Part I: projection'', Textile Res. J., Vol. 60 No. 11, pp. 690-6. Heisey, F., Brown P. and Johnson, R.F. (1990b), ``Three-dimensional pattern drafting ± Part II: garment modelling'', Textile Res. J., Vol. 60 No. 12, pp. 731-7.

IJCST 12,4

254

Ingham, R. (1983), The Costume Designer's Handbook, Prentice-Hall, New York, NY. Jaffe, H. and Relis, N. (1993), Draping for Fashion Design, Prentice-Hall, New York, NY. Jones, P., West, G., Harris, D. and Read, J. (1989), ``The Loughborough anthropometric shadow scanner (LASS)'', Endeavour, Vol. 13 No. 4, pp. 164-8. Kreyszig, E. (1988), Advanced Engineering Mathematics, 6th ed., John Wiley & Sons, Inc., New York, NY. Roebuck, J.A. (1995), Anthropometric Methods: Designing to Fit the Human Body, Human Factors and Ergonomics Society, Santa Monica, CA. Russ, J.C. (1995), The Image Processing Handbook, 2nd ed., CRC Press, Orlando, FL. Stjepanovic, Z. (1995), ``Computer-aided processes in garment production ± features of CAD/CAM hardware'', International Journal of Clothing Science and Technology, Vol 7. No. 2, pp. 81-8.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Modelling the ``CLOAK'' test: an example of two-dimensional elastica theory

Modelling the ``CLOAK'' test

Department of Textiles, Faculty of Fine Arts, University of Dokuz Eylul, Izmir, Turkey, and

Received February 2000 Revised June 2000 Accepted June 2000

F. Mete

D.W. Lloyd

255

Department of Industrial Technology, University of Bradford, Bradford, UK Keywords Fabric, Fabrication, Testing Abstract Computational elastica theory is used to model a simple test for the bending properties of fabrics. This test, entitled the ``CLOAK'' test, was designed to offer practical experimental advantages over the established cantilever bending, of bending length, test. Computational elastica theory offers a routine method for modelling fabrics in cantilever bending. In this case, the CLOAK test is simulated and shown to be equivalent to both the bending length test and to a related test method proposed in the 1960s.

Introduction The fabric bending length test (BSI, 1949), also known as the fabric cantilever beam test, is the simple, standard test method originally introduced by Peirce (1930) to measure the draping and handling qualities of fabrics. It is used in an updated form in the FAST bending tester (CSIRO, 1989). The test is designed to measure the bending rigidity of a fabric, A, defined by Peirce as the ratio M/C, where M is the bending moment per unit width applied to the fabric and C is the resulting curvature of the cloth. The geometry of the test is shown diagrammatically in Figure 1. Peirce proposed the relationship c3 = a/w, where w is the weight per unit area of the fabric and c is the ``bending length''. This enabled Peirce to calculate an approximate relationship between the overhanging length of the fabric, l, and the angle y: c3 cos 0:5 1 l 8 tan Peirce introduced this approximate (though very accurate) expression, because the non-linear differential equation that he had derived to define the shape of the cantilever proved impossible to solve analytically and because no numerical solution could be obtained. This relationship enables the angle y to be converted into a term called the bending length of the material, defined as The authors gratefully acknowledge the support provided by the Clothworkers' Textile Structures and Mechanics Laboratory in the School of Textile Industries, University of Leeds.

International Journal of Clothing Science and Technology, Vol. 12 No. 4, 2000, pp. 255-264. # MCB University Press, 0955-6222

IJCST 12,4

256

Figure 1. The geometry of the fabric cantilever test

the length of a material that will bend under its own weight to a specific extent. For fabrics, the standard angle of deflection is 41.5ë, at which angle the ratio c/l is equal to 0.5. It is straightforward to confirm these results using computational elastica theory (see Mete, 1996, for example). In practice, the test is conducted by sliding a strip of fabric over an edge and measuring the length of fabric that must be fed into the hanging part to obtain a cantilever whose tip just touches the 41.5ë line. Experimentally, difficulties arise because many fabrics display a tendency to twist, so that the point at which the fabric just hangs down to the 41.5ë line is uncertain. To overcome this difficulty, a simple test was proposed (Stuart and Baird, 1966) in which a strip of fabric was folded back on itself to form a stable, isolated loop. The height of this loop, h, was directly related to bending length, with c/h = 1.10. More recently, the CLOAK system (Cimtex Leicester objective assessment of knitwear) has been proposed (Cassidy et al., 1991, 1993), to permit the bending length of knitted fabrics which are prone to twisting or curling to be measured. The test method is described in more detail below. The purpose of the present paper is to provide a detailed theoretical basis for the CLOAK test and to demonstrate that this test is exactly equivalent to the traditional bending length test and the test proposed by Stuart and Baird. Use of one test method rather than another will then be determined purely by considerations of practical convenience. Detailed discussion of the claimed advantages of the CLOAK test is outside the scope of this paper. Modelling the CLOAK test also provides an interesting example of the application of twodimensional computational elastica theory to a practical situation. The theoretical model presented here has the same basis as previous models of the bending length test, in that it is assumed that the fabric is linearly elastic.

Real fabrics do not possess strictly linear or elastic bending properties; the most important omission here is to neglect the frictional component that leads to bending hysteresis, as this will subtly alter the predicted shape of the bend fabric. This simplification has not been judged to cause significant errors in practice in the case of the traditional bending length test, so it is appropriate to make the same simplification for the CLOAK test model. Computational elastica theory Computational elastica theory has been described in considerable detail elsewhere (see, for example, Konopasek, 1980a, 1980b, 1980c). For the purpose of the present work, the planar elastica was used, as expressed in the bending curve package written by Konopasek. The differential equations for the planar elastica are: 2U 0 Ap ÿ p0 2 2

w0y pwx 6

x 0 wx

3

p0 Fv =A p00 7

y 0 wy

4

Fv0 pFw fy wx 8

w0x ÿpwy

5

Fw0 ÿpFv ÿ fy wy 9

where U A p po x, y, z w x, w y Fv, Fw fy 0

is the strain energy of bending, is the fabric bending rigidity, is the curvature in the deformed state, is the curvature in the undeformed state, are the co-ordinates of a point on the bending curve, are the direction cosines of the unit tangent vector, w, are the v (unit normal) and w components of the internal force, is the distributed force per unit length in the y direction, and indicates differentiation with respect to the arc length, s.

The bending curve package produces both detailed numerical output and graphical output. The package performs the numerical integration required to solve general elastica problems and the iterative approach needed to solve the initial value-boundary value problems that arise in almost all specific cases. Modelling the CLOAK test using bending curve package The CLOAK test was introduced to measure the fabric bending length for knitted fabrics which are prone to twisting or curling. Figure 2 shows the

Modelling the ``CLOAK'' test

257

IJCST 12,4

258 Figure 2. The geometry of the CLOAK bending length test

geometry of the new test apparatus called the ``Bending Box''. The method is based on laying down a strip of fabric in the form of a loop, where the fabric is folded back on to itself. The bending length is calculated from the height of the loop. The fabric is pulled horizontally by a slider that is situated at a fixed height above a horizontal surface. The fabric buckles and is allowed to fold back on to itself, when it forms a stable fold which rests on the horizontal surface. The test can be regarded as a variation of the Stuart and Baird test (1966), in which some elements of operator error are eliminated by the experimental arrangement. In both tests, the mechanics of the completed loop should ensure that the loop settles into the equilibrium shape described below. The formulation and solution of the above cantilever problem using the bending curve package were performed as a series of deformation stages. It was assumed that fabric bends elastically, that the bending moment is linearly related to curvature and that the friction between the horizontal surface and the fabric is negligible. The fabric was modelled as a vertical cross-section normal to the axes of the folds. The distributed weight fy was taken as ± 10 to speed the solution. Dimensionless values were used for all the variables, to facilitate the solution (Lloyd et al., 1978). The deformation leading up to the final geometry can be divided into four deformation stages, each controlled by a different set of boundary conditions. The first stage in the deformation runs from the start of the test to the point at which the fabric starts to form a loop on itself. The second stage runs until the fabric just touches the surface to close the fold. This intermediate ``just touching'' stage is important to evaluate the numerical constants needed in later stages. The third stage involves the shape of the closed fold changing as the curvature at this point decreases to zero. In this stage, the point of contact, where the fabric touches the surface, rolls along the surface, until the fabric loop forms a stable fold. This ``zero curvature'' case at the contact point can be solved as the fourth stage, representing the final stable shape of the closed fold. The whole process was simulated by assigning and reassigning the initial values and boundary conditions for each stage as appropriate. This is done in

an intermediate subroutine and the main program. In each step appropriate numerical constants were calculated for the following steps. The boundary conditions for the first step are as follows: The initial conditions and parameters are: . length l = ? . bending rigidity A = 1. . curvature p(0) = ? . horizontal force Fx(0) = ? . vertical force Fy(0) = ? . distributed force fy = ±10. . direction cosine wx(0) = ±1. . direction cosine sign (wy(0)) = 0. . co-ordinate x(0) = 0.? 3. . co-ordinate y(0) = 1. The boundary conditions are: . curvature p(1) = 0. . direction cosine wy(1) = 0. . co-ordinate y(1) = 0. . co-ordinate x(1) = 0. ? l(1) ± l0 Here, l0 is the initial length of the fabric strip between the slider and the bottom of the box at the start of the movement of the slider. It was calculated as 1.668 for this case and was used to set the x and y co-ordinate values at the beginning of the test. Figure 3 shows the computed profiles of the fabric during stage 1. The boundary conditions for stage 2 are as follows: The initial conditions and parameters are: . length l = ? . bending rigidity A = 1. . curvature p(0) = ? . horizontal force Fx(0) = ? . vertical force Fy(0) = ? . distributed force fy = ± 10. . direction cosine wx(0) = ± 1. . direction cosine sign (wy(0)) = 0. . co-ordinate x(0) = 3.? 3.5. . co-ordinate y(0) = 1.

Modelling the ``CLOAK'' test

259

IJCST 12,4

260

Figure 3. Computed fabric profiles of the CLOAK test ± stage 1

The boundary conditions are: curvature p(1) = 0. direction cosine wy(1) = 0. . co-ordinate y(1) = 0. . co-ordinate x(1) = 0.? l(1) ± l0. The numerical values of the ``just touching'' point are computed at the end of stage 2. This involves a single solution initiated from the nearest solution to the required case. As is clear from the last profile in Figure 4, it is possible to run the solutions past the ``just touching'' point. . .

Figure 4. Computed fabric profiles of the CLOAK test ± Stage 2

After the first unstable loop has formed at the end of stage 2, it is convenient to solve the problem as a compound elastica, in which the fabric is divided into two segments. Stage 3 continues until the curvature where the two segments join decreases to zero (see Figure 5). At that point, the fold becomes stable. The boundary conditions for the two segments in stage three are: Segment 1: The initial conditions and parameters are: . length l(1) = ? . bending rigidity A = 1. . curvature p(0) = ? . horizontal force Fx(0) = ? . vertical force Fy(0) = ? . distributed force fy = ± 10. . direction cosine wx(0) = ± 1. . direction cosine sign(wy(0)) = 0. . arc length S(0) = 0. . co-ordinate x(0) = ? . co-ordinate y(0) = 1.

Modelling the ``CLOAK'' test

261

The boundary conditions are: . direction cosine wy(1) = 0. . co-ordinate y(1) = 0.

Figure 5. Computed fabric profiles of the CLOAK test ± Stage 3

IJCST 12,4

262

Segment 2: The initial conditions and parameters are: . length l2 = ? . bending rigidity A = 1. . curvature p(0) = transferred . . horizontal force Fx(0) = transferred. . vertical force Fy(0) = transferred. . distributed force fy = ± 10. . direction cosine wx(0) = ± 1. . direction cosine sign(wy(0)) = 0. . arc length S(0) = l1 . co-ordinate x(0) = transferred. . co-ordinate y(0) = transferred. The boundary conditions are: . curvature p(1) = 0. . direction cosine wy(1) = 0. . co-ordinate y(1) = 0. . co-ordinate x(1) = 0.? l1 + l2 ± l0. Finally, the boundary conditions for stage 4 are: Segment 1: The initial conditions and parameters are: . length l(1) = ? . bending rigidity A = 1. . curvature p(0) = 0. . horizontal force Fx(0) = 0. . vertical force Fy(0) = ? . distributed force fy = ± 10. . direction cosine wx(0) = ± 1. . direction cosine sign(wy(0)) = 0. . arc length S(0) = 0. . co-ordinate x(0) = 0. . co-ordinate y(0) =0. The boundary condition is: . direction cosine wy(1) = 0.

Segment 2: The initial conditions and parameters are: . length l2 = ? . bending rigidity A = 1. . curvature p(0) = transferred. . horizontal force Fx(0) = transferred. . vertical force Fy(0) = transferred. . distributed force fy = ± 10. . direction cosine wx(0) = transferred. . direction cosine sign(wy(0)) = 0. . arc length S(0) = l1 . co-ordinate x(0) = transferred. . co-ordinate y(0) = transferred.

Modelling the ``CLOAK'' test

263

The boundary conditions are: . direction cosine wy(1) = 0. . co-ordinate y(1) = 0. When the loop reaches the final stable shape in the fourth stage, the tension and curvature are both zero at the ends of the closed loop. The required loop height for the evaluation of the bending length of the fabric is the height of this stable loop top above supporting surface (see Figure 6). It should be noted that the bending length is proportional to the maximum height of the loop, and that this ratio was given by Stuart and Baird (1966) as

Figure 6. Computed fabric profiles of the CLOAK test ± Stage 4

IJCST 12,4

264

bending length/loop height = 1.10. Later, Lloyd et al. (1978) calculated the same constant from computer simulations of a related fabric folding problem. From the numerical results of stage 4, where the shape of the fabric loop is stable, this constant ratio of the bending length to the loop height was calculated as 1.103. The calculation used 40 integration intervals in total. The relationship, c3 = A/ w, was used in the calculation of the bending length of the fabric. Conclusions Computational elastica theory provides a relatively simple and effective method for simulating complex cylindrical bending of fabrics. Applying the method to a bending test suggested for knitted fabrics that are prone to twisting confirms that the method is exactly equivalent to other cantilever-type bending tests. The model gives the same results as simulations of a related, but simpler, bending test proposed in the 1960s. References British Standards Institution (1949), Methods of Test for Textiles, B.S. Handbook, No. 11. Cassidy, T., Cassidy, C., Arkison, M. and Cassie, S. (1991), ``The stiffness of knitted fabrics: a new approach to the measurement of bending ± Part 1: development'', International Journal of Clothing Science and Technology, Vol. 3 No. 5, pp. 14-19. Cassidy, T., Cassidy, C., Arkison, M. and Cassie, S. (1993), ``Objective measurement in the assembly of knitted garments'', in Stylios, G. (Ed.) Objective Measurement Technology in the Textile and Clothing Interface, Proceedings of the 2nd International Clothing Conference, Bradford. CSIRO (1989), FAST Instruction Manual, CSIRO, Ryde, Australia. Konopasek, M. (1980a), ``Classical elastica theory and its generalisations'', in Hearle, J.W.S., Thwaites, J.J. and Amirbayat, J. (Eds), Mechanics of Flexible Fibre Assemblies, Proceedings of the NATO Advanced Study Institute, Kilini, Greece, 1979, NATO ASI Series E: Applied Sciences No. 38, Sijthoff & Noordhoff, The Netherlands, pp. 250-74. Konopasek, M. (1980b), ``Computational aspects of large deflection analysis of slender bodies'', in Hearle, J.W.S., Thwaites, J.J. and Amirbayat, J. (Eds), Mechanics of Flexible Fibre Assemblies, Proceedings of the NATO Advanced Study Institute, Kilini, Greece, 1979, NATO ASI Series E: Applied Sciences No. 38, Sijthoff & Noordhoff, The Netherlands, pp. 275-92. Konopasek, M. (1980c), ``Textile applications of slender body mechanics'', in Hearle, J.W.S. Thwaites, J.J. and Amirbayat, J. (Eds), Mechanics of Flexible Fibre Assemblies, Proceedings of the NATO Advanced Study Institute, Kilini, Greece, 1979, NATO ASI Series E: Applied Sciences No. 38, Sijthoff & Noordhoff, The Netherlands, pp. 293-310. Lloyd, D.W., Shanahan, W.J. and Konopasek, M. (1978), ``The folding of heavy fabric sheets'', International Journal of Mechanical Sciences, Vol. 20, pp. 521-7. Mete, F. (1996), ``The simulation of fabric drape for computer aided design'', PhD thesis, University of Leeds. Peirce, F.T. (1930), ``The handle of cloth as a measurable quantity'', Journal of the Textile Institute, Vol. 21, pp. T377-T416. Stuart, I.M. and Baird, K. (1966), ``A new test for bending length'', Textile Research Journal, Vol. 36, pp. 91-3.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Edge inspection in automatic stitching A.J. Crispin, B. Pokric, M. Rankov, D. Reedman and G.E. Taylor

School of Engineering, Leeds Metropolitan University, Leeds, UK Keywords United Kingdom, Shoes, Research, Machine vision Abstract The paper describes work relating to the laser line triangulation technique which has been used to inspect the edges of overlapping shoe components prior to the sewing operation. The laser line triangulation technique involves projecting a laser line on to a surface which can be viewed using an area camera. A surface height transition (edge) causes a discontinuity in the observed laser line. Different approaches for extracting the edge positions in the image coordinate system have been investigated based on the Hough transform, the spatial histogram, polynomial regression and the discrete first derivative. These edge detection algorithms are compared in terms of speed and precision performance. Three-dimensional scans of typical shoe component parts are presented.

Introduction Microprocessor controlled stitching machines are used for sewing overlapping leather and textile components in the production cycle of apparel. The conventional automatic system comprises an x-y positioning table together with a sewing head. A multi-leafed pallet is used to clamp components prior to the sewing operation to ensure that the positional relationship between the preprogrammed stitch path and component edges exposed by the pallet does not change. Edges and other topological features of overlapping components can be extracted by equipping computerised stitching machines with a laser triangulation based vision system (see Crispin et al., 1997). Such a system is shown in Plate 1. A pre-scan can be performed on a loaded pallet prior to the stitching cycle so that the stitching process will stop if component parts have been badly loaded into a pallet or if component geometry violates a tolerance set by the user. It also allows pre-programmed stitch paths to be adapted according to the edge information obtained (Rankov, 1999). The technique may be applied to the stitching of clothes in garment manufacture. Automatic stitching of materials with high in-plane stiffness requires the point-to-point movement of the component. This is because the component must remain stationary for the duration of needle penetration. The dwell time is the time that the needle spends in the component piece before it is moved to the next position and is directly dependent on the material thickness. The material thickness is usually pre-programmed at the beginning of the stitching cycle and is used to calculate the needle dwell time, which remains fixed even though material thickness can vary from sample to sample. By measuring The authors wish to thank British United Shoe Machinery for their initial funding of this project and also M. Harrison for his help and assistance.

Edge inspection in automatic stitching 265 Received March 2000 Revised June 2000 Accepted June 2000

International Journal of Clothing Science and Technology, Vol. 12 No. 4, 2000, pp. 265-278. # MCB University Press, 0955-6222

IJCST 12,4

266

Plate 1. Computerised stitching machines equipped with a laser triangulation based vision system

material thickness it is possible to adjust the needle dwell time to match the measured material thickness or reject components exceeding an overlap height tolerance. The paper outlines methods for extracting the position and height of surface edges formed by overlapping leather components likely to be of the same colour from laser line images obtained using the triangulation technique. Different approaches for extracting the edge positions in the image co-ordinate system have been investigated based on the Hough transform, the spatial histogram, polynomial regression and discrete first derivative. The challenge of automated pre-inspection of loaded pallets is that of obtaining high spatial resolution combined with high speed. High spatial resolution is required for the inspection of localised defects such as notches or small perforations. High speed is required in order to keep the time overhead to a minimum so as to maximise through-put. Consequently, the edge detection algorithms are compared in terms of speed and precision performance. Edge inspection using laser triangulation The principle of laser line triangulation involves projecting a laser line on to a surface which can be viewed using an area camera. The triangulation technique allows direct measurement of 3-D information and has found applications in robot vision (Domey et al., 1990), surface scanning (Clark et al., 1995) and the Oxford AGV (Reid, 1996). In this application the camera is mounted at an appropriate angle to allow inspection of a desired range of material thickness (see Figure 1). The objective is to determine the surface edge

Edge inspection in automatic stitching 267

Figure 1. Laser line triangulation

position and height from the observed laser image. In Figure 1, the upper-left corner of an image in the camera pixel co-ordinate system is (1,1) which corresponds to the spatial co-ordinate labelled (0,0). The laser system is set up to ensure: (1) the laser line segment projected on to the higher z-plane of the inspected surface is never occluded; and (2) the lower z-plane is constant. The image to world point transformation can be performed using a proportional geometric model by approximating the camera field of view to be a rectangle with constant size. The edge position and material height can be extracted from the laser line image by observing the key distances marked as Px and Py in Figure 1. Taking ratios: Px xe Rx XFOV Py y Ry YFOV and putting y

ze tan

IJCST 12,4

yields equations for the edge world co-ordinate point (xe, ye, ze): xe

268 ze

Px XFOV Rx

1

ye Ya

2

Py YFOV tan Ry

3

Here: Px = pixel position of edge transition point Py = pixel position for calculating edge height XFOV = x-axis field of view YFOV = y-axis field of view a = the camera pitch angle Rx = x-axis range of pixel centres Ry = y-axis range of pixel centres Ya = the vertical position of the laser line relative to the origin. In this work a 1286128 CCD pixel camera with a frame rate of 440fps was used for laser line capture. The field of views is calculated as XFOV = 8.967mm and YFOV = 11.76mm respectively with the camera mounted at a height of 205mm with a = 49.68ë. The average data spatial resolution is dependent on the chosen speed of the x-y table, which can be varied up to a maximum value of 760mm/s using a neural network cross-coupling controller (see Crispin et al., 1999). A typical captured grey-level laser line image is shown in Figure 2. The laser line can be thinned to a single pixel width by tracing the boundary of a binary image and calculating the centre line as shown in Figure 3. The boundary tracing technique is discussed further by Hall (1979) and Sonka et al. (1999). The work here compares techniques for obtaining the co-ordinate point (Px, P1y) and Py in the image given that laser lines can be partially occluded on the lower surface due to the edge orientation with respect to the camera. The worst case of occlusion occurs when the edge presented to the camera has the same orientation as the projected laser line. To overcome this problem a second laser line is projected on to the surface at an angle of 45ë with respect to the first. The two laser lines are switched on and off independently using direction information obtained from the pre-programmed stitch path. Hough transform The Hough transform is a tool for the detection of known shapes in an image. Although it is required to know the analytical equations of the shape to be

Edge inspection in automatic stitching 269

Figure 2. Captured grey-level image with key distances marked

Figure 3. Centre pixel line of laser obtained using binary segmentation and boundary tracing

IJCST 12,4

270

detected, it is not necessary to have prior knowledge of its position within the image. The Hough transform method has the advantage that it is insensitive to imperfect data and noise (Gonzales and Woods, 1992). In this case, the Hough transform is applied to find straight lines of any orientation in a binary image of a thinned laser line. The basic idea of the linear Hough transform is to represent a straight line as a single point using the normal representation given by: x cos y sin and to subdivide the y, r parameter space into accumulator cells (Sonka et al., 1999). For every point in the binary image given by x and y co-ordinates y is incremented in the range 0 to 180ë and values of r are calculated and added to the accumulator. As a result, line detection in the image is transformed to the detection of local maxima in the parameter space, as shown in Figure 4. Positions of local maxima (defined by y and r) in the accumulator represent lines existing in the image. There are two local maxima in Figure 4, representing two straight lines in the binary image. The image space line equations can be found using: y kx c k 90o m c

m sin m

where ym and rm are the positions of local maxima in the parameter space. Figure 5 shows the lines identified by the Hough transform method together with the thinned laser line image. Spatial histogram A spatial histogram can be calculated by dividing the binary image into a set of lines parallel to the laser line such that they cover all pixels in the image. For each line, the number of pixel occurrences on that line is calculated and stored to produce the spatial histogram. Figure 6 shows the spatial histogram in which pixel occurrences are calculated for all horizontal y-axis lines in the image. Two maxima at y1 and y2 can be identified and used to calculate a mid-level line using: y1 y2 ylevel 2 The mid-level line is used to separate the binary thinned laser line image into upper and lower regions. For all pixels in the upper region (above the line) a negative sign is assigned and for all pixels in the lower region a positive sign is assigned. An edge can be detected from the pixel sign change which occurs at the break point. To ensure that the sign change is not due to noise, pixels either side

Edge inspection in automatic stitching 271

Figure 4. Hough transform parameter space

IJCST 12,4

272

Figure 5. Lines recovered by Hough transform method

of a transition point are checked for sign continuity. Edge height (material thickness) can be determined from the distance between the peaks in the spatial histogram. Polynomial regression An edge position in the image can be obtained by fitting second order polynomial segments to a thinned laser line. A polynomial segment is regressed through pixel values which represent the laser line until a calculated error exceeds a pre-specified limit. This point represents a boundary discontinuity, the cause of which can be either an edge or a noise variation. If the regressed segment is very short the discontinuity is considered to be noise and is ignored. The process is repeated until the last point in the image is reached and all boundary positions are found. Polynomial segments representing the laser line are joined at the calculated boundary positions. The first derivative di, at any point i along a binary thinned laser line is calculated using: yi1 ÿ yi di xi1 ÿ xi The x-axis position of the edge is identified by a minimum in the first derivative, as shown in Figure 7. The y position of an edge is calculated from

Edge inspection in automatic stitching 273

Figure 6. Spatial histogram method

IJCST 12,4

274

Figure 7. Polynomial regression method

the next polynomial segment having constant first derivative. The sign of the first derivative can be used to distinguish between either a rising or a falling edge. Discrete first derivative It is possible to calculate the first derivative of a binary thinned laser line without first fitting polynomial segments. However, it has been found that the discrete first derivative di, when used for edge position detection, can give unpredictable results, especially for edges which are not sharp as the laser line data points in the camera image co-ordinate system are prone to noise. To attempt to overcome this problem a modified first derivative can be calculated based on parameters related to edge geometry such as minimum expected width and height. In this work a modified first derivative dmi has been calculated using:

Edge inspection in automatic stitching 275

yid ÿ yi 1 dmi round Q dmin Q where dmin defines the minimum distance between two edges to be detected and Q a quantisation parameter which effectively filters noise. Q is directly dependent on the minimum height of the edge to be detected and noise magnitude and is derived empirically (a typical value is Q = 5). The edge image position is calculated from the first derivative as described in the polynomial regression method. Comparison of edge extraction methods The edge point extraction methods have been applied to 14 different edge profiles having different angle and occlusion characteristics (Pokric, 1997). Two of the profiles were of multiple edges. In each case the actual edge position was measured and recorded. Table I shows the percentage mean of position error obtained from the tests for each method and the standard deviations. The error obtained by calculating the material thickness at an edge was also calculated, with the percentage standard deviations shown in Table I. The algorithm processing times were also recorded for each method in each of the 14 tests and an average processing time calculated. Table I shows the results relative to the spatial histogram method. Mean position (%) x y Hough transform Polynomial method Spatial histogram Discrete first derivative

± ± ± ±

0.4286 1.0000 1.7857 0.7857

± 0.6429 ± 0.6429 2.7857 0.3571

Standard deviation Relative x/y position Thickness processing time 0.6013 1.0160 1.1955 0.8282

1.5012 1.1811 1.1790 1.1790

2.58 3.27 1.00 1.55

Table I. Comparison of edge extraction methods for 14 test samples

IJCST 12,4

276

The Hough transform method proved to be the most reliable method but had one of the highest processing times. The spatial histogram method, although a fast method, can produce significant errors in the edge position when the observed laser line angle changes from the expected value. This is because the spatial histogram method is designed to detect lines of known direction with a small range of tolerance. The polynomial method provided a good approximation to the laser line data and was found to be suitable for more complex edge profiles but has a long processing time. The discrete first derivative method is noise sensitive and produced poor results with the two multiple edge cases, although a relatively fast method. Surface height Surface height ze can be calculated using equation (3) which is based on a constant rectangular field of view approximation. However, the image to world point transformation can be performed by developing camera and laser plane models which take into consideration optical and geometrical properties of the inspection system. Tsai (1986) describes a camera model for the projection of a three-dimensional world point on to a two-dimensional image sensor which takes into consideration the effective focal length, radial lens distortion and image scanning parameters. An image to world point transformation can be performed using the inverse camera model of Tsai (1986) and finding the intersection point with a laser plane model (see Pokric, 1997; Crispin et al., 1997). Calibration can be performed by capturing a set of image points whose world co-ordinates are known. Figure 8 shows the error in obtaining the surface height using the two methods for different camera pitch angles. The error is not a smooth line due to the camera sensor quantisation effect. At small pitch angle both methods produce a small height error. At a larger pitch angle, say 45ë, the surface height error obtained using the constant field of view approximation is similar to that obtained using the camera-laser line intersection method for surface heights less than 5mm. Figures 9 and 10 show a three-dimensional representation of the edge positions and surface height (material thickness) of two typical shoe component parts. The image edge co-ordinates were obtained using the Hough transform method. Surface height has been calculated using the constant field of view approximation outlined in this paper. Conclusions A simple, accurate method based on laser triangulation for obtaining edge position and height of overlapping leather components has been presented. Four methods have been successfully applied to the problem of edge point extraction from binary thinned laser line images. It has been found that the Hough transform provides the most reliable method for edge point extraction, although computationally expensive in terms of processing time. 3-D sampling of some typical leather components has been demonstrated.

Edge inspection in automatic stitching 277

Figure 8. Surface height error for different camera pitch angles

Figure 9. 3D scan of upper shoe component (counter pocket piece)

IJCST 12,4

278

Figure 10. 3D scan of upper shoe component (vamp) References and further reading Chen, C.H. and Kak, A.C. (1987), ``Modelling and calibration of a structured light scanner for 3-D robot vision'', Proc. IEEE Conference Robotics and Automation, pp. 807-15. Clark, J., Zhang, G. and Wallace, A.M. (1995), ``Image acquisition using fixed and variable triangulation'', IEE Conference: Image Processing and its Applications, pp. 539-43. Crispin, A.J., Ibrani, L., Taylor, G.E. and Waterworth, G. (1999), ``Neural network cross-coupling gain controller for bi-axial contouring system'', American Control Conference, Vol. 1, pp. 351-2. Crispin, A.J., Pokric, B., Rankov, M., Reedman, D. and Taylor, G.E. (1997), ``Laser based inspection and path correction system for automatic sewing apparatus'', 3rd International Conference Laser Metrology and Machine Performance, Computational Mechanics Publications, pp. 145-60. Domey, J., Rioux, M. and Blais, F. (1990), ``3-D sensing for robot vision'', in Taylor, P.M. (Ed.), Nato ASI series, F64, Sensory Robotics for the Handling of Limp Materials, SpringerVerlag, Berlin, pp. 159-92. Gonzales, R.C. and Woods, R.E. (1992), Digital Image Processing, Addison-Wesley Publishing Company, New York, NY. Hall, E.L. (1979), Computer Image Processing and Recognition, Academic Press, NewYork, NY. Pokric, B. (1997), ``Laser-based machine vision for three-dimensional surface analysis'', PhD thesis, Leeds Metropolitan University. Rankov, M. (1999), ``Data modelling and analysis for stitch path derivation in automatic stitching'', PhD thesis, Leeds Metropolitan University. Reid, I. (1996), ``Projective calibration of a laser-stripe range finder'', Image and Vision Computing, Vol. 14, pp. 659-66. Sonka, M., Hlavac, H. and Boyle, R. (1999), Image Processing, Analysis and Machine Vision, PWS Publishing, CA. Tsai, R.Y. (1986), ``An efficient and accurate camera calibration technique for 3D machine vision'', Proc. CVPR Computer Vision, pp. 364-74.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

COMMUNICATION

Team-based sewn products manufacturing: a case study Betty G. Dillard

University of Missouri, Columbia, USA Keywords Manufacturing, Management, Sewing

Team-based sewn products manufacturing 279 Received July 1999 Revised May 2000 Accepted May 2000

Abstract Examines a team system using a case study approach in a sewn products plant that transitioned to the team system almost ten years ago. The theoretical framework utilized in the analysis of data was participative management, wherein it has been found that there is a relationship between employee knowledge and performance. Specific themes that arose were successes in the transition to the new team system, including benefits to the plant, and ongoing challenges of the new team system. Data were based on 16 in-depth personal interviews, observations, written documents, and informal conversations with plant employees. The key elements of success in the transition were commitment by upper level managers, education for all employees, and the establishment of open communication among employees and with management. The team system resulted in a number of benefits and challenges to the plant as a whole.

Introduction Upper level executives in apparel and sewn products companies in the USA are facing a number of challenges in order to compete effectively in global markets. Both the retailers and the ultimate consumers are demanding that manufacturers provide goods of high quality, and at the same time offer quick response and competitive prices. Manufacturers often feel intense pressure because of the rising cost of materials and labor, coupled with value-driven consumers who expect little, if any, price increase in products from year to year. Thus, manufacturers must look for new ways of cutting costs to better compete in today's marketplace. One of the cost-cutting strategies employed by a number of companies in the past decade has been to examine closely the activities and processes within their companies, with the goal of reengineering some or all of them. This often means streamlining the process, taking out parts that are not necessary, and even using a different approach to managing these activities, sometimes referred to as ``thinking outside the box''. In large companies, reengineering often takes several years, but the benefit is that managers, as well as employees, in these companies have a better grasp of what it takes to produce and sell a product in a timely manner. Improvements have been made by many companies in a number of different areas including reducing inventory costs, increasing speed of production and delivery, improving product quality, and reducing overhead costs. Managers readily acknowledge that one of their biggest problems in producing sewn products domestically is recruiting, training, and motivating

International Journal of Clothing Science and Technology, Vol. 12 No. 4, 2000, pp. 279-292. # MCB University Press, 0955-6222

IJCST 12,4

280

workers, particularly production workers. This, coupled with the need for a more skilled worker to operate computerized production equipment, often presents a real problem and makes production in other countries even more attractive. Because of the labor shortage and the relatively high wage rate in the USA, many companies are continuing to close factories here and use labor in low-wage countries for assembly. Mexico is particularly attractive to US companies because of the North American Free Trade Agreement (NAFTA). Some production will likely remain in the USA, especially goods produced in small lots for quick delivery ± products that are tailored specifically to customers such as special orders and customized products. A number of companies that have continued to produce domestically have shifted to a team-based environment using a more open, participative management style. In such a system, employees often work together on teams to complete the task whether it is product development, production, or some other activity. They are empowered by managers to make decisions in the areas in which they are knowledgeable and experienced. Using this approach, companies have increased flexibility to meet customers' demands by moving products more quickly through the assembly processes, often changing specific styles throughout the day, and improving the quality of the product. In the past it took many weeks, even months, from the time a garment was cut until it was produced and shipped to retailers. In today's quick response environment, this approach is unacceptable. Today, retailers demand that orders be filled in a matter of days, so they have products in the store at the time the consumer wants to buy them. In addition, retailers often want part of the order delivered at the beginning of the season with smaller shipments arriving throughout the selling season. Using this team approach, a number of companies have implemented flexible manufacturing systems that allow them to effectively compete without taking production offshore. In some cases, an entire factory functions as a team. Employees in such a system understand what is involved in the entire process and how their particular job ties in with the broader issue. Much has been written about the benefits of using team production systems. Fralix (1999) provided a good discussion of the history of team-based systems in sewn products manufacturing in the USA and the overall benefits to companies. However, little information exists about the manner in which companies accomplish the transition to teams, and few have actually approached this issue from both managers' and employees' perspectives. Using a case study approach, the purpose of this research was to examine the effectiveness of the transition from traditional line manufacturing to teambased manufacturing within a sewn products plant. Specific areas of interest in this study were as follows: key elements in the success of the transition from a traditional line system to a team approach, including benefits to the plant as a whole, and ongoing challenges of the new team system.

Review of related research The underlying theory for this new management approach is that of participative management. It was first proposed in the USA as early as the 1920s but received virtually no support or serious attention for two decades (Drucker, 1995). Participative management as an area of organizational research began with the research of Coch and French (1948) in the 1940s. Their research focused on understanding why factory workers in a sewing plant in Virginia resisted change in the production system. They found that those who were involved in group participation in planning changes were positively motivated toward the change while those who were not involved were more resistant. The first comprehensive review of research on participative management was done by Locke and Schweiger (1979), who conducted a meta-analysis of more than 50 studies representing a variety of types of participative management. The theoretical model proposed by Locke and Schweiger (1979) addressed the causal relationships among participation, knowledge, and performance. Research by Miller and Monge (1986) determined that the primary benefit of participation is as a mechanism for improving the knowledge of workers and ultimately their level of performance within the system. Positive outcomes have been reported in a number of studies ranging from manufacturing organizations (Safizadeh, 1991) to service-oriented organizations (Neider, 1980). According to Kantor et al. (1986) the use of employee involvement programs has one common theme: they provide ways for employees to contribute more by allowing greater participation and control over their work. Teams must be able to document their success by reviewing data collected at regular intervals. Participation must be valued and rewarded by upper level managers, and positive results must be recognized and celebrated. Scott and Townsend's (1994) research in team productivity examined teams in a large sewing plant in Virginia. They determined that team performance was related to attractiveness of performance, agreement with team goals, team goal level, willingness to use cross-training, perceived participation, team efficacy, and team commitment. In more recent research, Mills (1995) differentiated between participative management, sometimes known as the Japanese management style, which he described as a traditional directive management system with participative mechanisms included, and true participative management. Mills proposed that true participative management goes beyond the traditional approach because it empowers individuals or teams to accomplish goals in the way that they think is best with the resources that are available. According to Mills, all three of these elements are important to the success of this approach: (1) establishing goals; (2) empowering individuals or teams to take action; and (3) measurement of results. Fralix (1999) reported on the results of a study conducted over the past two years of sewn products plants that had converted from the traditional batch

Team-based sewn products manufacturing 281

IJCST 12,4

282

manufacturing to some form of team-based manufacturing. Results of the study indicate that team-based manufacturing offers consistent work-inprocess improvements and through-put time reductions from weeks to days. Floor space requirements were typically reduced by 25-30 per cent. Operator earnings generally increased while total manufacturing cost was fairly constant. Although quality is often reported to be better, their research showed that some plants experienced improved product quality while others did not. Methods This study was conceived as a qualitative case study, with grounded theory (Strauss and Corbin, 1990) as the orienting approach to the collection of narrative data and analysis. Case studies seek to wring out, from a single unit, all the data that can be had. The unit may be conceived as an individual or as a social unit or social institution. As with other qualitative research approaches, an understanding of the data in their context is critical because, as Yin (1984) explained, ``the boundaries between phenomenon and context are not [always] clearly evident.'' That the researcher seeks to ``examine most or all the potential aspects'' (Crabtree and Miller, 1992) of the case caused Stake (1994) to assert that case studies are inherently triangulated. Grounded theory, initially introduced by Glaser and Strauss (1967), and more recently elaborated by Strauss and Corbin (1990), is an approach to the collection and analysis of narrative data that is inherently a posteriori. That is, the researcher seeks to understand what categories and detail are important to participants, rather than seeks responses to predetermined categories or questions that might limit the density and variation of concepts that can prove valuable to the researcher. Thus, in this study the interviews were conducted as what Patton (1990) called ``informal conversational interviews'' (p. 288). The interviews flowed freely and easily from some form of the request, ``Tell us about working at this plant''. In this way, the resulting interpretation or theory is derived from data that are elicited from participants with minimal structure provided by the researcher. Moreover, data collection, analysis, and interpretation are conducted simultaneously throughout the process. The particular sewn-products plant which was the site of this research was chosen for several reasons. The researchers had knowledge about the change to a team environment in the late 1980s and early 1990s. The plant manager expressed his willingness to discuss openly the shift to a team environment and the implications of using this approach both to the employees and the company. Finally, researchers had the full cooperation from the plant manager to talk with anyone in the plant, review documents of interest and relevance to the study, and spend as much time as needed to fully understand and document the transition to a team-based system. Over the research period the researchers visited the plant seven times and stayed from four to seven hours each day. The data were generated from the following:

(1) in-depth informal conversational interviews with 16 employees of the plant; (2) informal conversations with a number of other employees; (3) observations made at the plant; and (4) written documents provided liberally by the plant management team. The 16 employees interviewed for this study included the plant manager, three supervisors (facilitators), an office manager, two employees from the engineering team, and nine production workers. At the time of the study, age of employees interviewed ranged from mid-20s to mid-70s. Fourteen of the 16 experienced the transition to the team system and the remaining two were hired after the team system was in place. The following paragraph includes a more in-depth description of the employees interviewed; the reader should be aware, however, that the names listed are pseudonyms. John is the plant manager and has worked in this particular plant for seven years, although he has worked for the corporation that owns the plant for 17 years. Beth is the secretary and has worked for this company since the year it was opened in 1960. The three facilitators include the following: (1) Mike, who has been with the company for three-and-a-half years; (2) Amy, an employee of the plant for ten years; and (3) Ken, who has worked at this plant since 1965. Jill, a sample maker on the engineering team, has been with the company for 20 years. The second member of the engineering team, Mary, is a pattern and sample-maker who has worked at the plant for 20 years. Of the nine production workers, five are sewing machine operators, including Donna, Laura, Lucy, Sue and Kate. Listed in chronological order based on length of employment, Donna has been with the company the longest ± 37 years, as opposed to Kate who came after the transition six years ago. Gail, an employee since 1987, is a trainer and floater and Pete works in the garneting area and has been with the company since 1975. Dave, an employee of the company for 25 years, is an accommodator in the warehouse area. Finally, Pam, a member of the quality control team, has been employed by the company for eight years. The interviews normally lasted about an hour, with a range of 35 minutes to one hour and 45 minutes. After initial introductions, participants were assured that the researchers were not working for the plant manager or corporate management, and that their comments would be treated in such a way that the plant manager would not be able to know who had said what. Interestingly, this did not appear to be of much concern to them. One woman responded: There is nothing I have to say that I wouldn't say to [the plant manager's] face.

The tapes were reviewed by the researchers shortly after a day's session and were then transcribed verbatim. The analysis of the data was conducted

Team-based sewn products manufacturing 283

IJCST 12,4

284

following the conceptual guide proposed by Strauss and Corbin (1990), which involves the identification of relevant concepts and their relation to one another through axial, selective, and theoretical coding. Plant history and context This production facility manufactures a recreational sewn product and is part of a division in a large US corporation. It is located in a small town with a population of approximately 1,300 people in a mid-western state. Currently, about 200 people work in the plant, and it has operated continuously for four decades, experiencing growth over the last ten years. The facility had 16 plant managers from the time they opened in 1960 until the current plant manager, who played an integral role in the change-over to a team system, came in 1990. In 1988, this plant began with only one team and eventually restructured the plant of more than 100 employees over a two-year period from 1988-1990. Today, the entire plant functions on a team basis with every one of its 197 employees as a member of some team. Most of the teams comprise five to seven individuals; however, one team has 11 people. Teams include a management team, an engineering team, a raw materials and warehouse team, a quality team, and production teams. Both the upper level executives in the company and the plant manager were committed to this new approach. Prior to the late 1980s this plant followed traditional management practices. Plant managers and production workers had very little contact or interaction; each person had his or her own job and production workers were paid on the piece rate system. Production workers perceived previous management to be totally removed from the workers, and plant managers typically stayed only a few years. Until the change to teams, the plant produced larger lot sizes, 2,000 to 3,000 in a lot or order, compared to small lots today, often fewer than 100 units. The products were less complex in construction compared with those produced today. The work in process was enormous with extremely high stacks of partially sewn product throughout the plant. Employees who went through the transition reported that under the old system it was not a pleasant environment in which to work because they performed the same jobs all day long and seldom talked with anyone else. Results and discussion Transition to the new system Commitment by upper level management. The transition began with a commitment from top company executives in the late 1980s. They read numerous articles in journals about the team system and believed that for this plant to survive in the changing retail environment (demand for high quality, faster delivery, and competitive prices), they needed to change to a team system. The threat of imports from low wage countries such as China was beginning to affect their business. No longer could they afford to have huge piles of inventory on the plant floor. This was described by employees as a

drastic change, and the current plant manager was selected to lead them through the change because of his understanding of the system and what it would take to put it in place. Numerous employees credited the success of the transition to the plant manager, John. Ken described what he felt was a necessary component for the success of such a transition: This type of concept has to start at the very top. The president of the company has to buy into it . . . This change cost a lot of money, so if you are not focused at the top . . . I don't see how it could ever work. John, working for the corporation that owned this plant, was integral in developing that corporate commitment.

Jill also recognized the significant contributions made by John: He always played a big part [in the transition to teams] because he came to help with it before he was plant manager.

Education and training. A number of training sessions, totaling about 18 hours, were held with all employees before the team system began to be implemented. Employees read and discussed books about the team work environment; they also listened to presentations by consultants who were brought in for that purpose. The company invested a great deal of money in training, approximately $2,500 per person, as part of the transition. Amy attributes the success of the transition to training. She claimed: Most people don't want to change because they are afraid . . . Most people had to learn how to change.

Laura explained that the learning took place by: [taking] some classes on teamwork and working together . . . [The training] really made you more aware.

This education and training is one way to involve employees in the transition to a team system, which according to Coch and French (1948) is an effective way to motivate employees toward the change. As part of the training, employees were taught the benefits of teams. Laura explained: [The training] gave you insight as to the advantages of working in a group and how a group could put out more as a whole than [an] individual person could.

Ken offered a caveat to ending the education process at classroom sessions: Even with [the 16-18 hours of classroom training and the plant meetings], I was still apprehensive. I didn't think it would work. It's a learning process as you go; . . . you are not going to buy into it right away.

However, he did admit: You need to start with that, plant meetings and lots of communication [and] classroom work.

Ken's comments clearly indicate the importance of continuing education in the form of communication.

Team-based sewn products manufacturing 285

IJCST 12,4

286

Open communication as continuing education. Most employees feel that this communication did continue past the initial transition. Throughout the transition upper level managers worked to establish better communication with employees than was true before the transition. Amy said: We try to share stuff. [We] give them as much information as we can . . . Sometimes it is scary that [John] can say some of the stuff he says because ten years ago you weren't allowed to say anything.

Ken shares the same opinion concerning John: It used to be that I had to tell my supervisor that I would like to talk to Mr Plant Manager. That's not the way it is now; it's an open door policy. Any of us can talk to John or anybody, anytime.

John said: This is quite a change from the times when the corporate offices mandated the plant manager to spend more time on the floor, and as a response to this [the manager] built an office in the middle of the floor and then sat in that office all day long, which didn't accomplish a thing.

Before the transition, employees were not aware of issues of quality, nor retail customers' expectations, nor did they care. Following the transition, employees were informed about the entire process including how problems in one area affected work in another area; hence, they were motivated to produce first quality garments. For example, Pete explained how communication between teams has increased overall quality: I'll go out there on the floor and communicate with the other groups and if we are [making products] and the [products] aren't [constructed] right, they come back and tell us and we'll try to see what we can do different next time.

The retail customer's expectations about quality and timely delivery are discussed at meetings with team leaders. Laura discussed the change she noticed as a result of these meetings: When I came here you didn't know anything . . . Now we are told everything [about what needs to go out to retailers].

These findings are in congruence with those of Locke and Schweiger (1979) and Miller and Monge (1986); participation, knowledge, and performance are related whereby participation leads to greater knowledge, which in turn leads to better performance. Information about plant efficiency and profit is shared with employees so that they understand better what is involved in shipping a first quality product on time. Bi-weekly team meetings serve as a forum to discuss issues of efficiency: Ken said: At these meetings, one of the things we talked about is improvement ideas. So [employees] are constantly reminded to think about it and try to come up with an easier, better method of doing something.

In addition to discussing issues of efficiency, the company's financial standing is openly shared with employees. Laura claimed that:

every bit of the financial budget is displayed and explained . . . [Employees] are told exactly how much profit or loss [the plant realized].

Benefits of the new system. The plant has seen a number of positive results with the new management approach. In general, it has been positive for both the company and the employees, although the focus of this research is on the benefits to the company. Because there is a learning curve with the team approach, some of the benefits did not occur initially. Employees were crosstrained so that every team member could perform the job of every other team member, thus allowing for the completion of a product even if one team member was absent. Employees who had previously worked at one task all day every day had to learn to work with others to solve problems and to do so in an efficient manner. In some cases this was a real challenge. Benefits to the company have been positive overall; both plant records and discussions with employees show at least six positive results. First is the growth in volume; the number of employees almost doubled from 110 employees before the team system to 197 currently. Second, the turnover rate has decreased from over 50 per cent to approximately 25 per cent. At the time the transition to teams occurred, some employees left because they did not want to work on teams. However, most stayed and, according to John, 90 per cent of the employees bought into and appreciated the added flexibility the new system offers them. Third, efficiency (productivity) of the plant has increased from 85 per cent to over 100 per cent. At the time of the research the efficiency was around 109 per cent, but it fluctuates a great deal due to small lots with numerous style changes and less pay incentive with the hourly wage than the former piece rate system. Each team is expected to maintain 100 per cent productivity, and they are encouraged to set productivity goals. Lucy reported that her team sets a goal of 120 per cent and tries to maintain that. Gail explained one possible reason for the increase: I think efficiency is better [because we change jobs throughout the day] . . . you know, if you get up and do something different it kind of revives you.

With this increased efficiency came a through-put time that has decreased from three to four weeks to one week or less. Ken said: [With the team approach] the turn-around time is faster . . . I mean, you can actually push a work order through [in one afternoon] if you have the materials in stock and it has to go.

This allows the company to capitalize on one of the advantages of domestic production ± quick response. Ken further commented: There used to be mounds of work. We used to have between a week and two weeks' worth of work between operations. Typical work-in-process has gone from three to four weeks to two to four days.

Fifth, not only has efficiency in producing products increased, but the quality of those products has increased as well; Ken said: Quality has greatly improved since the team concept went into effect.

Team-based sewn products manufacturing 287

IJCST 12,4

288

Laura explained that the team approach: encourages better quality [because you] see the product [from] start to finish.

In addition to increased quality, the benefits found in this study are in general agreement with those of Fralix (1999) but also included additional positive results not found in that study, such as the increase in the number of employees and the decrease in employee turnover. A final benefit to the company relates to cost. Although labor costs increased initially due to education and training associated with the transition to the team system, these costs decreased after the new system was fully implemented. Indirect labor has been reduced; for example, they went from five supervisors on the sewing floor to only two with the new system. Labor cost makes up much less of the total cost in their products than the materials cost. John explained that the flexible manufacturing, in addition to implementing just-intime with their materials suppliers, led to much greater cost reduction in materials and inventory. Inventory levels were reduced in all three areas ± raw materials, goods-in-process, and finished goods. Ongoing challenges of the new system The plant under the new team system was not perceived to be perfect by any of the individuals who were interviewed. There have been, and continue to be, many challenges. Three of the workers interviewed had a number of complaints and seemed more negative in general; the others also expressed several concerns. These major concerns that presented challenges for the success of the team system included three specific areas: (1) education and training for new employees; (2) team selection and function; and (3) equitable pay system. The importance of education and training was also discussed as an important element in the successful transition to a team-based system discussed earlier. This implies that once the transition is made, continued attention is needed in order to ensure success. These findings support Fralix's (1999) results that the team-based system is difficult to implement and to maintain, but with continued commitment of management it can be successful. Education and training of new employees. This plant had doubled in size from the time the team system was established to the time of the research. This meant that almost half of the employees had not experienced working in the traditional plant with a piece rate system of pay. It also meant that, although they had training when they started to work for the company, they did not have the in-depth education about teams, how they function, etc. It is important to give new employees a better sense of what the goal is and how they can work together to reach their goals. Employees who went through this transition recognize this need; Gail explained: I think [new employees] need more of the team training.

Some of the employees working there today have never experienced the rigidity and stress of doing piece rate work and they may not fully appreciate how much better the work environment is today than a decade ago. Employees often must be educated about the benefits of the team system to the worker as well as to the company. The flexibility offered under the new system seemed to be taken for granted by Kate, one of the two people interviewed who had not worked there before the team-based system began. The leader of the engineering team, during an informal conversation, discussed the biggest challenge which he believes is getting the employees to change from ``What's in it for me?'' to ``What's in it for the good of the whole company, including myself?'' In order to accomplish this, companies must continue to educate employees about the team-based system, including an understanding of plant productivity, so they will begin to develop a sense of ownership. Team selection and function. When the first team was developed, facilitators assigned workers according to their speed and skill level. Amy explained: We picked out people who at that time were high earners . . . well, that didn't work.

According to Ken: the first team that was put together was hand selected and that bombed . . . [because] I don't think personalities were [considered].

Jill also noted that: ``there were some personality clashes [with that first team]''. All of the facilitators admitted that this was not successful because the faster workers tended to be very competitive and they did not work together well as a team. Managers learned that it was better to configure teams based on both skill and personality fit. ``I'd say personality is just as important as skill'', explained Ken. The facilitators emphasized the importance of the selection process in ensuring the success of a team. Kate, a sewing operator, noticed this importance that facilitators now put on selecting members for a team: When Amy brings in a new person she will try and figure [out] their personality to decide where she is going to put them.

Facilitators learned a valuable lesson from the first team that they formed ± that both skill and personality need to be considered. When a team has problems, team members are expected to try to work them out. One way team members solve these problems is through peer pressure. Ken explained: [If] you have one person that does not do his or her share, it's the team members that somehow will . . . say something about it. Peer pressure is a good thing as far as the team concept is concerned.

If they continue to have problems working together as a team, the facilitator meets with the team members to assist in resolving conflict. If problems persist, the facilitator may move certain individuals from one team to another. Amy explained that:

Team-based sewn products manufacturing 289

IJCST 12,4

290

teamwork is like a marriage. You can be together so long and things are OK, and then something goes wrong and nobody knows what [it is] and then you have to make some changes.

Although the facilitators were trained initially in team supervision and dynamics, they expressed the importance of being flexible and learning from experience. All of the employees interviewed acknowledged that sometimes there was conflict with co-workers. Most of the employees managed it quite well but a few were frustrated about their co-workers. This frustration included taking too many breaks, working too slowly, personality conflicts, and problems arising from outside the workplace. For example, Laura explained: [If] you put five people together in a group, there is going to be one person that's kind of slow or just doesn't want to put out any effort or can't do it. Of course, this kind of ruffles the feathers of the people on the team.

However, even with the team problems, only Gail and Donna were adamant that they would prefer the old system where they were responsible for their own work rather than having to work on a team with other people. The primary reason for this response, however, related to the pay system, perceived as inequitable, and not to problems of team interaction. Equitable pay system. Almost everyone acknowledged that one of the major problems in the current system was the pay structure. During the transition period when the team system was implemented, the piece rate system was no longer appropriate to use. At that time management reviewed each employee's level of skill and ability to perform more than one operation. Two pay levels were established, one at $8.00 an hour and the other at $6.75 an hour. Those who fell into the higher category were workers who were working with specialized equipment and other operations that often require special lifting, turning, or a high skill level. This included a few sewing operators who had high productivity and could operate a number of different machines. In addition to the hourly wage, bonuses were earned based on plant productivity and quality level achieved in the plant. However, the system seemed inequitable to many because team members felt that they were not receiving adequate monetary rewards for their accomplishments. Lucy exclaimed: Whether you are a slow-paced person or a fast-paced person . . . you still get the same thing.

Sue shared the same sentiments as Lucy: Why should I go out there and sew so hard to make such a huge percent and [John] doesn't even acknowledge that [I] do?

According to Kantor et al. (1986) employees must be rewarded for their successes by upper level management. Based on employees' perceptions of the current pay system, this is not happening in the company under study and is an issue that needs to be resolved.

Adding to this problem, employees were confused as to how plant bonuses translated to their individual paychecks. Jill said, speaking for herself and other employees: A lot of people don't have a lot of trust in the pay plan. The feeling [is] that it is a lot of funny numbers.

Gail also spoke of this problem: [Employees] think that they are being cheated . . . They just think [that corporate] is trying to rip them off.

These comments clearly speak to the importance of educating employees on how bonuses are awarded. The plant manager as well as corporate executives acknowledged the problems with the pay plan and were researching other pay systems at the time the research was conducted. Employees were aware of this and one commented: They tell us [that] in January we are going to have a new pay plan, but that doesn't mean it is going to be good. We don't really look forward to it because we think, ``What are they going to do this time?''

Although this plant has experienced many successes with the transition to a team system, employees' perception of inequitable pay is an issue that needs to be addressed before this plant evolves into a truly participative environment. Conclusions Although not perfect, the team approach was perceived by most of the employees interviewed in this case study as being far superior to the traditional bundle system. This new management approach may not work for all companies, but for many it has worked well. Using the case study approach, this research provides insight about key components in the transition to a team system. One of the keys to success in this plant was having upper level managers committed to the team system because it takes a great deal of time and financial resources in the beginning and it does not happen without careful planning. In addition, employees attributed the success of the transition to John, the plant manager. He not only garnered the commitment of those from corporate but those within the plant as well. Second, education and training were provided to all employees before, during, and after the team system was implemented. This needs to continue in the form of open communication and education of new employees. Even though employees were basically more satisfied with the team system than the traditional bundle system, the inequitable pay system was a problem for many. Their earnings represent their livelihood and feeling of self-worth; therefore, companies must work diligently to ensure that employees feel the pay system is an equitable one where they are rewarded for their accomplishments. Findings from this research will lead to a better understanding of the way in which companies can incorporate this type of change in the most positive and productive ways, thereby increasing organizational efficiencies and realizing

Team-based sewn products manufacturing 291

IJCST 12,4

292

optimal cost savings. Although it may be more difficult to implement and maintain than traditional systems, it can offer companies and workers a number of viable benefits. References Coch, L. and French, J.R.P. (1948), ``Overcoming resistance to change'', Human Relations, Vol. 1, pp. 512-32. Crabtree, B.F. and Miller, W.L. (1992), ``Primary care research: a multimethod typology and qualitative road map'', in Crabtree, B.F. and Miller, W.L. (Eds), Doing Qualitative Research, Sage Publications, Newbury Park, CA. Drucker, P.F. (1995), ``Mary Parker Follett: prophet of management'', in Graham, P. (Ed.), Mary Parker Follett ± Prophet of Management, Harvard Business School Press, Boston, MA, pp. 1-9. Fralix, M. (1999), ``Team sewing: the results are in'', Apparel Industry Magazine, February, pp. 73-4. Glaser, B.G. and Strauss, A.L. (1967), The Discovery of Grounded Theory, Strategies for Qualitative Research, Aldine Publishing Co., Chicago, IL. Kantor, R.M., Summers, D.V. and Stein, B.A. (1986), ``The future of workplace alternatives'', Management Review, Vol. 75 No. 7, pp. 30-33. Locke, E.A. and Schweiger, D.M. (1979), ``Participation in decision making: one more look'', Research in Organizational Behavior, Vol. 1, pp. 265-339. Miller, K.I. and Monge, P.R. (1986), ``Participation, satisfaction, and productivity: a meta-analytic review'', Academy of Management Journal, Vol. 29, pp. 727-53. Mills, D.Q. (1995), ``The new management system'', European Management Journal, Vol. 13 No. 3, pp. 251-6. Neider, L.L. (1980), ``An experimental field investigation utilizing an expectancy theory view of participation'', Organizational Behavior and Human Performance, Vol. 26, pp. 425-42. Patton, M.Q. (1990), Qualitative Evaluation and Research Methods, 2nd ed., Sage Publications, Newbury Park, CA. Safizadeh, M.H. (1991), ``The case of workgroups in manufacturing operations'', California Management Review, Vol. 33, pp. 61-82. Scott, K.D. and Townsend, A. (1994), ``Teams: why some succeed and others fail'', HR Magazine, August, pp. 62-7. Stake, R.E. (1994), ``Case studies'', in Denzin, N.K. and Lincoln, Y.S. (Eds), Handbook of Qualitative Research, Sage Publications, Thousand Oaks, CA. Strauss, A. and Corbin, J. (1990), Basics of Qualitative Research, Sage Publications, Thousand Oaks, CA. Yin, R.K. (1984), Case Study Research Design and Methods, Sage Publications, Beverly Hills, CA.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Effect of belt extensibility on variation of the relative position of a needle and a hook in a sewing machine Jerzy Zajaczkowski

Textile Faculty, Lodz Technical University, Lodz, Poland

Effect of belt extensibility

303 Received April 1999 Accepted May 2000

Keywords Sewing machines, Non-linear response, Sewing Abstract Considers the mathematical modelling of sewing machines. The machine with a rotary hook and with a feed mechanism driven by two triangle cams is studied. The behaviour of the system is described by a set of non-linear autonomous ordinary differential equations. It is shown that, due to the timing belt extensibility, the actual relative position of the needle and the hook is different from that expected. This difference is a result of the conflict between the almost constant speed of the hook shaft and the fluctuating speed of the crank-shaft caused by the reciprocating motion of working elements. The vibration of the belt increases with speed. Resonance increase of vibrations takes place when the shaft speed is equal to natural frequency or its fraction 1/n for n = 1, 2, 3, 4, 5.

Introduction The process of sewing is performed by the coordinated action of a needle, a hook, a take up lever and a feed dog. The needle carries the upper thread down putting it through the fabric and then it moves up leaving the thread loop beneath the fabric. The loop is taken by the hook around the bobbin case containing the lower thread. The upper thread is then stretched by the lever of the take up mechanism. At the same time the toothed bar of the feed dog moves forward horizontally pulling the fabric forward. Then it goes down beneath the feed plate and moves to the backward position. The hook can be either oscillatory or rotary. In the machine with the oscillatory hook the slider-crank mechanism driving the needle bar assembly and the crank-rocker-yoke mechanism driving the hook assembly induce different speed fluctuation of driving the cranks. This results in torsional vibration of connecting the shaft (Zajaczkowski, 1996). The main excitation frequency (Zajaczkowski, 1999c) of the crank-shaft is twice its frequency of rotation. The main excitation frequency of the hook shaft is equal to the frequency of rotation of the crankshaft. In a zig-zag sewing machine one cycle of vibrations, resulting from driving belt extensibility, takes two shaft revolutions (Zajaczkowski 1997a). In case of the torsional vibration of the crank-shaft, the difference between the oscillations corresponding to two consecutive rotations is small (Zajaczkowski, 1999b). If the hook is driven by a timing belt then the relative position of the needle and the hook depends upon the belt extensibility (Zajaczkowski, 1997a). The purpose of this paper is to study the dependence of the relative position of the needle and the hook on the speed of the machine taking account of a feed

International Journal of Clothing Science and Technology, Vol. 12 No. 5, 2000, pp. 303-310. # MCB University Press, 0955-6222

IJCST 12,5

304

mechanism. The interaction of a needle and the fabric (Mallet and Ruxu, 1999) is not taken into account. Equations of motion The main mechanisms of a sewing machine are shown in Figures 1 and 2. The crank-shaft () is driven by the motor (m ) with a V-belt. It drives the needle (Y) through the slider-crank mechanism and the hook shaft (ÿ) through the timing belt. The speed of the hook shaft is twice that of the crank-shaft. Two triangle cams (shown in Figure 2) attached to the crank-shaft drive the feed dog, as it is shown symbolically by a broken arrow in Figure 1. The angle 3 can be adjusted to control the stitch length. The sum of virtual works of all forces acting on the crank-shaft, the needle bar and the levers driving the feed dog is equal to zero: 1 MA d FN dY Mv dv Mh dh 0: Here, is the rotation angle of the crank-shaft, Y is the vertical displacement of the needle, v and 1 are the rotation angles of levers driving the feed dog. The forces and the moments resulting from the system inertia and belt extension can be written as: d2 MA ÿA 2 Fb ra ÿ FH RH ; dt dm d rb ÿ Fb Db ra sb m rb ÿ ra ; dt dt d dÿ F H DH RH ÿ rH sH RH ÿ ÿrH ; dt dt FN ÿmN

Figure 1. Mechanisms driving the needle and the hook in a sewing machine

d2 Y d 2 v d 2 1 ; Mv ÿBv 2 ; Mh ÿBh 2 2 dt dt dt

2

Effect of belt extensibility

305

Figure 2. Triangle cam mechanisms driving the feed mechanism ± two relative positions of follower and cam

where A, Bh and Bv are the mass moments of inertia, mN is the mass of the needle bar, Db is the coefficient of viscous damping, sb is the stiffness of the V-belt, ra and rb are the radii of the V-belt pulleys, ÿ is the rotation angle of the hook, DH is the coefficient of viscous damping and sH is the stiffness of the timing belt, RH and rH are the radii of the timing belt pulleys. The friction forces, the thread and fabric reactions are not included here. Instead, the general damping forces are included. Substituting expressions (2) into equation (1) and applying the following transformation: 2 2 d2 d2 d d d 2 3 2 dt dt d dt d2 one obtains the equation governing the motion of the sewing machine:

IJCST 12,5

306

A mN

dY d

!2 Bv

dv d

!2 Bh

d1 d

!2 !

d2 dt 2 !

!2 dY d2 Y dv d 2 v d1 d 2 1 d mN Bv Bh d d2 d d2 d d2 dt ! dm d rb ÿ ÿ Db ra ra ÿ sb m rb ÿ ra ra dt dt ! d dÿ RH ÿ rH RH sH RH ÿ ÿrH RH 0: DH dt dt

4

The sum of the moments of inertia forces and the force resulting from the belt extension is: ! d2 ÿ d dÿ 5 rH rH ÿ sH RH ÿ ÿrH rH 0: RH ÿ BH 2 ÿ DH dt dt dt Here, BH is the mass moment of inertia of the hook shaft. Summation of the moments of forces acting on the motor has the form: ! d 2 m dm d Db rb ÿ ra rb sb m rb ÿ ra rb ÿ Mm 0 Am dt 2 dt dt

6

where m is the rotation angle of the motor, is the rotation angle of the crank-shaft, and Am is the mass moment of inertia. The electromagnetic torque of the motor Mm can be found from the differential equation: dMm dm T Mm Cm m ÿ dt dt

! 7

where T is the motor time constant, Cm is the slope of the motor characteristic and m is the idle run velocity, i.e. the angular velocity for which the motor torque is equal to zero. The motion of the needle mechanism is given by the functions Y = 2 (Lk ± Rk2 sin2 )1/2 ± Rk cos . The triangle cams driving the feed mechanism are similar to the cam driving the zigzag mechanism (Zajaczkowski, 1999b). The distance between the axis of rotation of the cam and its follower can be determined from Figure 2. The function h( ) is symmetric, it starts with value b , then it becomes:

8

Effect of belt extensibility

and finally it equals to R . The rotation angle = () of the lever, v v 1 driving the feed dog in vertical direction, can be found from the following equations:

307

h R2 =a cos ÿ k =2 h a cos ÿ ÿ k =2

k ÿ 2 < < =2; =2 < < ÿ k =2

sin v R1 b =2 ÿ hv =a v =2 v v ; v k q a l2 ÿ d9 2 d5 d10 ÿ l1 2 tan

9

l2 ÿ d9 ; k ÿ d5 ÿ d10 ÿ l1 2

The distance hv = h( v ) is given by expression (8) for = v . In order to find the rotation angle of the lever 1 = 1 () the mechanism, driving the feed dog in horizontal direction, is replaced by two closed vector polygons (Zajaczkowski, 1999a). Equating to zero the components of its resultants, yields: ÿ d1 cos 1 ÿ d2 sin 2 d3 cos 3 d4 0 d1 sin 1 d2 cos 2 ÿ d3 sin 3 ÿ d5 0 d3 cos 3 d4 ÿ d9 d8 sin 2 ÿ d7 cos 2 d6 sin 2 0 ÿ d3 sin 3 d10 ÿ d8 cos 2 ÿ d7 sin 2 ÿ d6 cos 2 0

10

Here, d7 = (R1 + b )/2 ± hh and the distance hh = h( h ) is calculated from (8) for

= h = /2 ± + 2 . Once the angles v and 1 are determined, the angle f , defining the position of the feed dog can be found. Replacing the mechanism by a closed vector polygon and setting its horizontal and vertical components of the resultant to zero gives: lv cos v lf cos f ÿ lh sin h ÿ l2 0 l1 lv sin v ÿ lf sin f ÿ lh cos 1 0

11

The set of equations with the unknowns (v , v ), (1 , 2 , h , d3, d8) and (lf, f ) is solved by using Brent's method (More and Cosnard, 1980). The obtained discrete solutions are then locally replaced by continuous functions using polynomials. Having determined the coefficients in equation (4) the solution of the set of differential equations (4)-(7) may be found numerically.

IJCST 12,5

308

Numerical results Detailed studies were carried out for the mass moments of inertia A = 0.0001kgm2, Am = A/10, BH = A/2, Bv=A/2, Bh = A/2, the mass mN = 0.05g, the stiffness of the timing belt driving the hook sH = 500,000(N/m), the stiffness of the motor V-belt sb = sH/10, the coefficients of viscous damping Db = DH = 0.05(Ns/m) the stiffness of the motor characteristic Cm = 0.06(Nms/rad), the motor time constant T = 1/(10 m ), the radii of the belt pulleys rb = 0.01m, ra = 4rb, rH = 0.015m, RH = 2rH, the dimensions of the slider-crank mechanism Rk = 0.017m, Lk = 0.035m; the dimensions of the cam mechanism k = /2, a = 0.005/(2 ± 2 sin(k /2)), b = 0.014m, the dimensions of the feed mechanism a = 0.195m, d1 = 0.02m, d2 = 0.12m, d4 = d1, d5 = d2, d6 = 0.03m, d9 = 0.02m, d10 = 0.19 ± d5, 3 = /4rad, l1 = 0.01m, l2 = 0.095m, lh = 0.025m, lv = 0.03m. In order to determine approximative values of natural frequencies of vibration, the variable coefficients in differential equations were replaced by their average values. The frequency equation was found to be: 0 2

0 A

0

3

2

31 ÿsb ra rb 0 sb r2b 7 6 7C 0 5 ÿ 4 ÿsb ra rb sb ra2 sH RH2 ÿsH RH rH 5A 0 0 0 BH 0 ÿsH RH rH sH rH2 12 2 2 ! 2 Z2 1 dY dv d1 d mN Bv Bh A A d d 2 d

Am B 26 @! 4 0

0

The calculated natural frequencies were ! = 837 and 2,637rad/s. The integration of differential equations was carried out until the difference between successive periods became negligible. The difference between the actual and expected position of the hook shaft ÿ = ÿ ± RH/rH versus needle position Y for various idle run angular velocity of the crank-shaft

= mrb/ra(rad/s) is shown in Figure 3. Concluding remarks The reciprocating motion of the needle bar and the feed dog cause the fluctuation of the speed of the crank-shaft. The rotary motion of the hook shaft is characterised by a tendency towards rotation with a constant speed. Since both shafts are connected to the belt, the varying strain of the belt occurs. As a result of that strain, the position of the hook in relation to the position of the needle is different from that corresponding to an inextensible belt. That difference increases with the increase of speed. Resonance increase of vibration takes place when the shaft speed is equal to natural frequency or its fraction 1/n for n = 1, 2, 3, 4, 5.

Effect of belt extensibility

309

Figure 3. The difference between the actual and expected position of the hook shaft ÿ = ÿ ± Rh/rH versus needle position Y for various speeds of

IJCST 12,5

310

Figure 3. References Mallet, E. and Ruxu, D. (1999), ``Finite element analysis of sewing process'', International Journal of Clothing Science and Technology, Vol. 11 No. 1, pp. 19-36. More, J.J. and Cosnard, M.Y. (1980), ``BRENTM: a fortran subroutine for the numerical solution of systems of non-linear equations [C5]'', ACM Transactions on Mathematical Software, Vol. 6, pp. 240-51. Zajaczowski, J. (1996), ``Torsional vibrations of shafts in a sewing machine'', Fibres & Textiles in Eastern Europe, Vol. 4 Nos 3/4, pp. 49-50. Zajaczowski, J. (1997a), ``Dynamics of belt drive of a hook in a sewing machine'', Scientific Bulletin of Lodz Technical University ZNPL-Wl, Vol. 778 No. 55, pp. 99-105. Zajaczowski, J. (1997b), ``Dynamics of a zigzzag sewing machine with an oscillatory hook'', Fibres & Textiles in Eastern Europe, Vol. 5 No. 3(18), pp. 60-1. Zajaczowski, J. (1999a), ``Torsional vibration of a hook shaft in a sewing machine'', Fibres & Textiles in Eastern Europe, Vol. 7 No. 1(24), pp. 51-3. Zajaczowski, J. (1999c), ``Vibrations of a crank-shaft in a sewing machine induced by a zigzag mechanism'', International Journal of Clothing Science and Technology, Vol. 11 No. 1, pp. 53-9.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Evaluation of the basic low stress mechanical properties (bending, shearing and tensile) A. Alamdar-Yazdi

Department of Textiles, The University of Yazd, Iran, and

J. Amirbayat

Department of Textiles, UMIST, Manchester, UK

Basic low stress mechanical properties 311 Received March 2000 Revised July 2000 Accepted July 2000

Keywords Mechanical properties, Low stress, Fabric Abstract Objective measurement of fabric mechanical properties has great potential for quality control of clothing materials. However, access to the requisite instruments still remains a problem for many potential users due to their high cost. A new methodology for measuring the basic low stress mechanical properties of woven fabric on a tensile tester is introduced. The results of experimental work on 39 samples are also presented. As a result, new parameters indicating the behavior of woven fabrics are introduced.

Introduction The approaches for evaluation of the basic low stress mechanical properties can be classified into two groups: (1) Those which are not yet practical but have the scientific base and the advantages of simplicity, low cost, and accessibility[1-4]. They are mostly based on extraction of the circular fabric sample through a device such as a nozzle or a ring installed on an Instron tensile tester or other similar universal testing instrument. The guide to evaluating and judging the behavior of the fabric under such complex deformation is the load-displacement curve obtained from the Instron tester. (2) The systems whose approach to measuring the fabric mechanical properties under low load have been acceptable. Among this group, the Kawabata evaluation system is the leading method and widely accepted. This system and its abilities have been discussed and described in detail in many publications[5-12]. However, the above mentioned properties are measured by two testers of Kawabata Instruments which were actually developed by Kato, Tekko Co., Kyoto, Japan in collaboration with Kawabata. KES-FB1 tester measures shearing and tensile properties and the KES-FB2 pure bending tester evaluates bending behavior. The system is designed so that the same piece can be used on both the instruments provided that the tests are run in the appropriate order. Parameters for such evaluation are shown in Table I. Points to consider regarding the KES system are the considerable time required to measure the parameters; the number of the samples for each fabric should be at least three or four. The role of the operator is found to be very

International Journal of Clothing Science and Technology, Vol. 12 No. 5, 2000, pp. 311-330. # MCB University Press, 0955-6222

IJCST 12,5

Instrument

Property

KES-FB1

Tensile

312

KES-FB1

Shear

Table I. KES mechanical parameters

KES-FB2

Bending

Linearity of extension curve Energy in extending fabric to 5N/cm Tensile resilience Fabric extension at 5N/cm. Width Shear rigidity Hysteresis of shear force at 0.58 shear angle Hysteresis of shear force at 58 shear angle Bending rigidity Hysteresis of bending moment

Symbol

Units

LT WT RT EMT G 2HG.5 2HG5 B 2HB

± gf.cm/cm2 percent percent gf./cm deg gf./cm gf./cm gf.cm2/cm gf.cm/cm

important, and the most important point for industry is the cost, which is quite high[13-16]. In this paper, we introduce a method, which is based on concentrating loading as shown in Figure 1, and consider its ability to evaluate the three basic mechanical properties. Woven fabrics, like any other flat sheet can have three basic but independent forms of deformations (tensile (extension or compression), shear, and bending), plus buckling mode, which is the combination of the basic modes of deformation. Figure 2 shows in-plane and out-of-plane deformation of the woven fabric when the load is uniformly applied across the width of the sample. It is noticeable that the intensity of the stress is equal at any point and the direction is the same as the direction of the load applied. When a concentrated load is applied on the boundary of a very large plate it will cause buckling which, in fact, is a combination of bending, shearing and tensile deformations. In fact, a plate buckles when the in-plane load gets so large that the originally flat equilibrium state is no longer stable and the plate deflects into a non-flat configuration. The load at which the departure from the flat equilibrium state takes place is called the critical buckling load[17]. Figure 3 shows three different types of buckling due to the shearing load, the compressive force, and concentrating load.

Figure 1. Concentrated loading on woven fabric

Basic low stress mechanical properties 313

Figure 2. Fabric in-plane and out-of-plane deformations

Buckling of woven fabric due to the compressive loading has been studied by Dahlberg[18], who designed an apparatus to measure plate buckling, and also Grosberg and Swani[19] who analyzed plate buckling of fabrics in terms of their bending behavior. Buckling under combined shear and tensile forces is studied by Ly[20]. Flexural rigidities in warp and weft directions, the ratio of width to the length of the rectangular specimen (aspect ratio), Poisson ratios and the shear modulus are the parameters which would affect the buckling behavior of the fabrics. Buckling due to concentrating load is studied by Amirbayat[17,21]. The ratio of the bending stiffness to the in-plane modulus, Poisson's ratio, and the dimensions of the sheet are important factors in buckling under tension. Amirbayat has discussed and showed that testing three bias fabric samples under concentrated forces (which causes the fabric to buckle) is a method to evaluate the in-plane (tensile and shear) and out-of-plane (bending) deformation of the fabric[22].

IJCST 12,5

314

Figure 3. Buckling of the woven fabrics due to the shearing, compressive, and concentrated loading

Experimental results A total of 39 woven fabrics with the specifications shown in Appendix 1 (Table AI) were randomly chosen for the experimental work. Three rectangular specimens, each 24cm long and 5cm wide were cut from every sample fabric, one at an angle of 22.58 to the warp direction (which is 67.58 to the weft direction), one at 458 to the warp (which has the same angle to the weft direction), and one 67.58 to the warp (which is 22.58 to the weft direction), using a special template. Since bias samples develop shear strain under tensile stress, the strips were folded in half to form a double ply of face-to-face fabrics 12cm long. An eyelet was then punched 1cm from the ply ends opposite the fold. The second eyelet was inserted 10cm from the first one after any possible slack was removed. The samples were then subjected to a single loading-unloading cycle at a rate of 10mm/min with a 200g maximum force using a simple attachment to the jaws of Testometric-micro 350 made in the UK by Shirly developments, with a 10kg load cell. Test results Figure 4(a) shows a typical simple load extension curve of textile fabrics. This simple concave curve is the most general for any testing conditions and for all woven fabrics at low stress conditions. Figure 4(b) shows three types of load extension curves obtained during testing of the samples by the suggested new method. There are two types of difference observed ± between load-extension curves (conventional and new methods) and between the curves obtained by the new method.

Basic low stress mechanical properties 315

Figure 4. (a) A typical simple load extension curve; (b) three types of load extension curves obtained with the new method

Figure 5(a). Features calculated from the curves (slopes and strains)

IJCST 12,5

316

Figure 5(b). Features calculated from the curves (areas)

Symbols

Parameters

Units

AAC ABC AF AGA AGL AH AHF ALF ATH AUD CPD EBC EBCC ECD

Area under the loading curve, above the critical point Area under the loading curve, below the critical point Area under the loading curve, for 50g load Gap area between two curves, above the critical point Gap area between two curves, top 50g load Area under the loading curve, for 100g load Area under the loading curve, for 150g load Area under the loading curve, final 50g load Area under the loading curve, for 200g load Area under the unloading curve Strain at the critical point on the unloading curve Difference in strain between maximum load and end point Strain between two curves at the critical point Difference in strain between maximum load point and the point with the same load as the critical point on the unloading curve Strain at the critical point Strain at end point Strain at 50g load Strain at 100g load Strain at 150g load Strain at the location of MG on the unloading curve Strain at the location of MG on the loading curve Strain at maximum load (200g) Ending slope (final 20g of the unloading curve) Difference in strain between the critical point on unloading curve and end point Area under the unloading curve, above the critical point Area under the unloading curve, below the critical point Load at which maximum gap occurs Maximum distance (strain) between two curves Post buckling slope (20g load after buckling point) Peak slope (first 50g of the unloading curve) Peak slope for final 50g loading Load at the critical point Initial slope (first 10g loading) of the loading curve Slope up to the critical point Gap area between two curves

gf. gf. gf. gf. gf. gf. gf. gf. gf. gf. % % %

ECP EE EF EH EHF EMGD EMGL EML ES ET LAAC LABC LAMG MG PBS PSD PSL SCP SI SUC TGA

Basic low stress mechanical properties 317

% % % % % % % % % gf. % gf. gf. gf. % gf. gf. gf. gf. gf. gf. gf.

Notes: The features are also shown in Figure 5(a) and (b). The curves indicate the behavior of the fabrics under the load. The above extracted features show the differences between the curves and in fact indicate the difference in behavior of the fabrics that are tested under the same conditions

The initial low-stress region of the conventional curve shows the crimp removal and yarn straightening followed by a very rapid increase in the fabric stress. The curve of the new method shows an initial negative curvature, which becomes positive upon further extension, and the end part of the curve (end of loading) becomes almost linear. The differences between the curves are due to the response of the fabric to the method of loading (uniform or applied over a

Table II. Features extracted from the curves

IJCST 12,5

318

small area). In the conventional method we are dealing with a tensile load which is uniformly distributed and the energy used to deform the fabric in its own plane, while in the case of concentrated loading the energy is used to deform both the in-plane and out-of-plane deformation by buckling of the fabric. Unloading curves do not show any significant differences, but due to the method of deformation the return movement could be divided into at least two parts. One is the return movement of the fabric from buckled status to the plate form. The second is the return movement in the plane of the fabric. In the case of conventional method the return movement is in-plane deformation due to the load releasing. Despite the fact that the curves of the new method follow the same trend and look alike, there are some differences between them. These differences are driven from four sources: (1) the buckling point; (2) slopes; (3) strains; and (4) the areas under each part of the curve. Figure 5 shows all these features, thus clearly showing the differences. Features extracted from the tests under concentrated load (load-extension curves) For each curve, features shown in Figure 5(a) and (b) and listed in Table II are measured. Estimation of properties from measured qualities Figures 7-9, 11-14, 16 and 17 give the plots of actual shearing, tensile and bending properties measured by KES against the estimated values from the

Figure 6. The important zones of the curve related to shearing properties

Basic low stress mechanical properties 319

Figure 7. The plot of measured shearing modulus versus the estimated values

Figure 8. The plot of measured shearing hysteresis at 0.58 versus the estimated values

present set of tests. Correlation between the properties and the measured features are also given in Tables AII-AIV in Appendix 2. Charts belonging to all different fabric properties contain 78 points, which is double the actual number of the fabrics tested. This is done by: . considering each fabric in its conventional sense and relating its warp properties to the parameters along 22.58, 458, and 67.58 from the warp direction; and

IJCST 12,5

320

Figure 9. The plot of measured shearing hysteresis at 58 versus the estimated values

Figure 10. The important zones of the curve related to tensile properties .

considering each fabric rotated through 908 and relating its weft properties to the parameters along 67.58, 458, and 22.58 from direction of the warp.

As a result each property along a principal direction is related to the parameters 22.58, 458, and 67.58 from its direction. Note that subscripts 1, 2 and 3 refer to 22.58, 458 and 67.58 from directions along which the features were measured. Discussion and conclusions Comparison between three sets of results shows that estimation of shear properties is more accurate than the other two sets, tensile and bending.

Basic low stress mechanical properties 321

Figure 11. The plot of measured linearity of the tensile curve versus the estimated values

Figure 12. The plot of measured tensile energy versus the estimated values

Shearing modulus has the highest correlation with the post-buckling slopes (PBS1, PBS2 and PBS3). It should be noticed that highest correlation among them belongs to 458 bias samples indicating the same result as previous work done by Amirbayat and Alagha[22]. Comparison of the results also shows that critical load (SCP), post-buckling slope (PBS) and difference in strain between maximum load point and the point with the same load as the critical point on the unloading curve (ECD) have

IJCST 12,5

322

Figure 13. The plot of measured fabric extension versus the estimated values

Figure 14. The plot of measured tensile resilience versus the estimated values

highest correlation among all the other measured features (Figure 6). The estimated equations and the plot of actual values of each shearing parameter versus its estimated values are shown by Figures 7-9. Among the features measured, ABC, SI, PBS and EF (zone 1), EHF, EML and PSL (zone 2), ECD, AUD and LAAC (zone 3), and ES (zone 4) showed highest correlation with tensile properties. This relation highlights the important zones

of the curve related to tensile properties (Figure 10). The estimated equations Basic low stress and the plot of actual values of each parameter versus its estimated values are mechanical also shown by Figures 11-14. properties Even though for bending properties correlations are relatively low, AUD, LAAC, EE, CPD and ECD are the features which gave highest correlation with bending properties indicating the importance of unloading curve (behavior of 323 the fabric during the load release). Figure 15 shows the important zones of the curve related to bending properties. Figures 16 and 17 show the plots of the measured bending parameters versus the estimated values.

Figure 15. The important zones of the curve related to bending properties

Figure 16. The plot of measured bending rigidity versus the estimated values

IJCST 12,5

324

Figure 17. The plot of measured bending hysteresis versus the estimated values References 1. Alley, V.L. Jr and Mc Hatton, A.D., ``A proposed quantitative measure of fabric handle and the relative characterization of some aerospace flexible materials by handle moduli'', AFGL-TR-760306, Special Report No. 200, 1976. 2. Alley, V.L. Jr, ``Revised theory for the quantitative analyses of fabric hand'', Journal of Engineering for Industry, Vol. 102 No. 1, February, 1980, p. 25. 3. Grover, G., Sultan, M.A. and Spivak, S.M., ``Screening technique for fabric handle'', Journal of Textile Institute, Vol. 84 No. 3, 1993, p. 486. 4. Pan, N. and Yen, K.C., ``Physical interpretation of curve obtained through the fabric extraction process for handle measurement'', Textile Research Journal, Vol. 62 No. 5, 1992, p. 279. 5. Harlock, S.C., ``Fabric objective measurement. Part 2: principle of measurement'', Textile Asia, Vol. 20 No. 7, 1989, p. 66. 6. Harlock, S.C. ``Fabric objective measurement. Part 4: production control in apparel manufacture'', Textile Asia, Vol. 20 No. 7, 1989, p. 89. 7. Postle, R. ``Fabric objective measurement. Part 1: historical background and development'', Textile Asia, Vol. 20 No. 7, 1989, p. 64. 8. Postle, R., ``Fabric objective measurement. Part 3: assessment of fabric quality attributes'', Textile Asia, Vol. 20 No. 7, 1989, p. 72. 9. Kawabata, S., ``The development of the objective measurement of fabric handle'', in Kawabata, S., Postle, R. and Niwa, M. (Eds), Proceedings of the 1st Japan-Australia Symposium on Objective Specification of Fabric Quality, Mechanical Properties, and Performance, Kyoto, 1982, Textile Machinery Society of Japan, Osaka, 1982, p. 31. 10. Slater, K., ``Physical testing and quality control'', Textile Progress, Vol. 23 Nos 1-3, 1993. 11. Kawabata, S. and Niwa, M., ``Fabric performance in clothing and clothing manufacture'', Journal of Textile Institute, Vol. 80 No. 1, 1989, p. 19. 12. Kawabata, S. and Niwa, M., ``Objective measurement of fabric mechanical property and quality ± its application to textile and apparel manufactures'', International Clothing Science and Technology, Vol. 3 No. 1, 1991, p. 5.

13. Mahar, T.J., Dhingra, R.C. and Postle, R., ``Measuring and interpreting low stress fabric mechanical and surface properties precision of measurement'', Textile Research Journal, Vol. 57 No. 6, 1987, p. 357. 14. Bishop, D.P., ``Fabric sensory and mechanical properties'', Textile Progress, Vol. 26 No. 3, 1996, p. 31. 15. Ly, N.G. and Denby, E.F., ``CSIRO Inter-laboratory trial of the KES-F for measuring fabric properties'', Journal of Textile Institute, Vol. 79 No. 2, 1988, p. 198. 16. Ly, N.G., ``Error analysis of measurement made with the KES-F system'', Textile Research Journal, Vol. 59 No. 1, 1989, p. 17. 17. Amirbayat, J., ``The buckling of flexible sheets under tension part I'', Journal of Textile Institute, Vol. 82 No. 1, 1991, p. 61. 18. Dahlberg, B., ``Mechanical properties of textile fabrics part II: buckling'', Textile Research Journal, Vol. 31 No. 2, 1961, p. 94. 19. Grosberg, P. and Swani, N.M., ``The mechanical properties of woven fabrics part IV'', Textile Research Journal, Vol. 36 No. 4, 1966, p. 338. 20. Ly, N.G., ``The buckling of fabric under combined shear and tensile loads'', Proceedings of the International Wool Textile Research Conference, Vol. 3, 1985, p. 118. 21. Amirbayat, J. and Bowman, S., ``The buckling of flexible sheets under tension part II. Experimental studies'', Journal of Textile Institute, Vol. 82 No. 1, 1991, p. 71. 22. Amirbayat, J. and Alagha, M.J., ``A new approach to fabric assessment'', International Journal of Clothing Science and Technology, Vol. 7 No. 1, 1995, p. 46.

(Appendices follow over page.)

Basic low stress mechanical properties 325

IJCST 12,5

326

Appendix 1

F. no.

Ends (cm)

Warp count (Nm)

Warp cri. (%)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

40.0 40.0 40.0 46.0 11.5 98.0 18.0 90.0 18.0 100.0 98.0 100.0 18.0 42.0 37.0 100.0 58.9 18.0 50.0 60.0 37.0 55.0 46.0 44.0 100.0 65.0 33.2 40.0 43.2 55.0 56.6 41.2 32.0 53.5 31.0 50.0 78.0 23.0 31.0

180.0 160.0 165.0 130.0 33.0 200.0 34.0 180.0 34.0 180.0 180.0 190.0 34.0 70.0 68.0 165.0 95.0 34.0 90.0 86.0 60.0 80.0 66.0 68.0 150.0 100.0 50.0 65.0 66.0 74.0 81.6 56.0 50.0 68.0 42.0 55.0 88.0 28.0 45.0

3.6 4.0 4.0 10.0 8.0 4.0 6.0 8.0 6.0 6.0 5.0 5.0 6.0 6.0 5.0 5.0 11.0 6.0 4.0 8.0 6.0 8.0 4.0 7.0 5.0 6.0 4.0 4.0 7.0 9.0 11.0 8.0 6.0 9.0 8.0 10.0 12.0 12.0 8.0

Picks/cm

Weft co. (Nm)

Weft cri. (%)

28.0 32.0 30.0 25.0 11.0 38.0 10.0 42.0 12.0 42.0 40.0 42.0 14.0 28.0 28.0 40.0 32.7 16.0 36.0 30.0 30.0 30.0 28.0 28.5 44.0 44.0 23.0 28.0 28.5 30.0 30.6 24.0 26.0 27.5 24.0 27.5 44.0 18.0 26.0

95.0 110.0 100.0 100.0 33.0 120.0 34.0 120.0 34.0 110.0 100.0 102.0 34.0 78.0 64.0 100.0 88.0 34.0 68.0 86.0 68.0 80.0 70.0 68.0 100.0 100.0 50.0 56.0 60.0 84.0 82.0 56.0 48.0 68.0 42.0 50.0 77.0 28.0 26.0

4 6 5 10 12 7 10 8 10 2 8 8 11 8 8 8 6 12 6 10 12 7 6 8 8 10 5 10 10 14 10 10 10 10 8 9 20 15 8

Fab. str.

Fib con

Fab. end use

Fab. wei. (g/m2)

Fab. th. (mm)

Plain Plain Plain Plain Plain Satin Plain Satin Plain Satin Satin Satin Plain Plain Plain Satin Plain Plain Fancy Plain Plain Plain Twill Plain Satin Twill Plain Plain Plain Plain Plain Plain Plain Plain Plain Twill Fancy Fancy Plain

N P P P C P C P C P P P C V P P V C P P C, P V, P C, P C, P P P C, P P P P V C, P C, P C, P C, P C, P P C, P C, P

Shee Shee Shir Shee Shee WW Bond WW Shee WW WW WW Shee Shee Shir WW Shir Shee WW Shir Shir Shir Shir Shir WW WW Shir Shir Shir Shir Shir Shee Shee Shee Shee Shir Shir Shir Shir

54 57 57 67 75 85 89 92 95 98 100 100 102 103 105 106 108 108 114 114 115 115 115 116 118 118 120 120 122 123 125 127 128 130 143 160 167 167 175

0.110 0.100 0.100 0.140 0.380 0.165 0.410 0.190 0.420 0.199 0.190 0.195 0.440 0.198 0.211 0.200 0.200 0.450 0.192 0.200 0.235 0.200 0.285 0.220 0.230 0.195 0.290 0.222 0.202 0.200 0.199 0.260 0.240 0.240 0.250 0.295 0.380 0.335 0.330

Table AI. The specification of the fabrics used in the Notes: Bond = bondage; Shee = sheeting; Shir = shirting; WW = women's wear; experiment C = cotton; P = polyester; V = viscose; N = nylon

Appendix 2

PAR

Shearing modulus G2 G3 G1

AAC ±0.588 ABC 0.593 AF ±0.632 ±0.581 AGA ±0.281 AGL AH ±0.657 ±0.605 AHF ±0.365 ALF ±0.570 ATH ±0.524 AUD ±0.308 CPD EBC ±0.644 ±0.353 EBCC ECD ±0.670 ECP 0.428 EE ±0.258 EF ±0.582 EH ±0.640 EHF ±0.650 EMGD ±0.535 EMGL ±0.391 EML ±0.654 0.787 ES ±0.105 ET LAAC +0.582 LABC 0.515 LAMG 0.736 ±0.448 MG 0.407 PSD PBS 0.916 a 0.539 PSL SCP 0.822 SI 0.446 SUC 0.670 ±0.383 TGA

±0.528 0.668 ±0.725 ±0.515 ±0.063 ±0.689 ±0.596 ±0.164 ±0.507 ±0.498 ±0.270 ±0.692 ±0.328 ±0.754 0.567 ±0.269 ±0.662 ±0.713 ±0.710 ±0.521 ±0.249 ±0.710 0.713 ±0.002 0.559 0.731 0.764 ±0.444 0.353 0.921 a 0.331 0.826 0.459 0.333 ±0.299

±0.588 0.593 ±0.632 ±0.581 ±0.281 ±0.657 ±0.605 ±0.365 0.570 ±0.524 ±0.308 ±0.644 ±0.353 ±0.670 0.428 ±0.258 ±0.582 ±0.640 ±0.650 ±0.535 ±0.391 ±0.654 0.787 ±0.105 0.582 0.515 0.736 ±0.448 0.406 0.916 a 0.549 0.822 0.446 0.683 ±0.387

Shearing hysteresis at 0.58 Shearing hysteresis at 58 2HG.51 2HG.52 2HG.53 2HG51 2HG52 2HG53 ±0.214 0.645 ±0.467 ±0.200 0.040 ±0.341 ±0.239 ±0.022 ±0.191 ±0.631 0.119 ±0.631 0.070 ±0.664 0.531 0.143 ±0.451 ±0.422 ±0.389 ±0.142 ±0.451 ±0.373 ±0.708 ±0.081 0.678 0.463 0.457 ±0.018 0.423 0.543 0.176 0.750 0.256 0.415 0.112

±0.207 0.792 a ±0.620 ±0.214 0.246 ±0.427 ±0.288 0.167 ±0.175 ±0.656 0.127 ±0.719 0.062 ±0.782 a 0.688 0.119 ±0.568 ±0.540 ±0.502 ±0.165 ±0.401 ±0.485 0.676 ±0.008 0.708 0.717 0.506 ±0.041 0.513 0.553 0.010 0.824 a 0.318 0.246 0.159

±0.213 0.646 ±0.471 ±0.201 0.052 ±0.343 ±0.239 ±0.017 ±0.190 ±0.628 0.120 ±0.629 0.070 ±0.666 0.540 0.138 ±0.454 ±0.424 ±0.391 ±0.144 ±0.451 ±0.374 0.701 ±0.067 0.678 0.474 0.463 ±0.020 0.430 0.545 0.186 0.753 0.258 0.508 0.113

±0.376 0.654 ±0.599 ±0.352 ±0.063 ±0.501 ±0.402 ±0.154 ±0.354 ±0.673 ±0.029 ±0.721 ±0.080 ±0.761 0.532 0.003 ±0.576 ±0.568 ±0.546 ±0.302 ±0.471 ±0.535 0.757 ±0.086 0.728 0.515 0.610 ±0.180 0.456 0.706 0.309 0.816 0.347 0.523 ±0.051

±0.367 +0.765 ±0.753 ±0.352 0.151 ±0.590 ±0.453 0.037 ±0.339 0.693 ±0.027 ±0.818 a ±0.091 ±0.885 a +0.661 ±0.025 ±0.698 ±0.694 ±0.664 ±0.327 ±0.380 ±0.654 0.749 ±0.022 0.515 0.753 0.664 ±0.210 0.528 0.733 0.128 +0.875 a 0.416 0.325 ±0.002

±0.376 0.658 ±0.599 ±0.351 ±0.058 ±0.500 ±0.403 ±0.153 ±0.354 0.673 ±0.026 ±0.721 ±0.077 ±0.763 0.538 0.003 ±0.576 ±0.568 ±0.545 ±0.301 ±0.472 ±0.534 0.747 ±0.080 0.753 0.522 0.612 ±0.179 0.456 0.703 0.320 0.819 0.345 0.595 ±0.050

Notes: 78 cases of light-weight fabrics. The italic figures signify the highest correlation coefficient among the three sets of data. a Signifies the three highest correlation values among all the features

Basic low stress mechanical properties 327

Table AII. The correlation between the features and the KES shearing parameters

LT1

±0.123 0.301 ±0.345 ±0.051 0.165 0.255 ±0.166 0.058 ±0.112 ±0.439 0.053 ±0.471 0.025 ±0.494 0.304 0.075 ±0.381 ±0.351 ±0.320

AAC ABC AF AGA AGL AH AHF ALF ATH AUD CPD EBC EBCC ECD ECP EE EF EH EHF

Table AIII. The correlation between the features and the KES tensile properties

PAR

RT1 ±0.068 0.002 ±0.116 ±0.030 ±0.160 0.045 ±0.020 ±0.194 ±0.068 ±0.175 0.180 ±0.080 0.189 ±0.015 ±0.081 0.245 a 0.209 a 0.162 0.130

LT3

±0.222 0.325 ±0.365 ±0.116 ±0.028 ±0.294 ±0.242 ±0.086 ±0.211 ±0.522 0.047 ±0.558 a 0.021 ±0.561 0.287 0.097 ±0.393 ±0.373 ±0.358 ±0.070 0.037 0.017 ±0.100 ±0.139 ±0.013 ±0.045 ±0.124 ±0.070 ±0.092 0.045 ±0.019 0.049 0.015 ±0.039 0.082 0.094 0.070 0.045

RT2

0.140 0.021 0.162 0.099 ±0.001 0.170 0.161 0.060 0.141 0.067 0.143 0.094 0.147 0.139 ±0.031 0.172 0.177 0.187 0.190

RT3 0.130 ±0.111 ±0.004 ±0.050 0.181 0.022 0.073 0.256 0.127 0.575 a ±0.307 0.371 ±0.306 0.346 ±0.041 ±0.354 ±0.026 ±0.009 0.013

WT1 0.051 ±0.190 0.085 ±0.062 ±0.063 0.065 0.054 0.005 0.044 0.432 ±0.226 0.283 ±0.214 0.278 ±0.101 ±0.249 0.054 0.055 0.051

WT2

±0.144 ±0.185 ±0.014 ±0.187 ±0.128 ±0.092 ±0.146 ±0.145 ±0.153 0.201 ±0.235 0.103 ±0.221 0.132 ±0.165 ±0.205 0.026 ±0.021 ±0.051

WT3

0.152 ±0.164 0.081 ±0.036 0.121 0.083 0.111 0.224 0.147 0.627 a ±0.286 0.439 ±0.279 0.431 ±0.106 ±0.325 0.068 0.077 0.090

EMT1

328

±0.185 0.374 ±0.441 ±0.072 0.203 ±0.342 ±0.256 0.101 ±0.169 ±0.512 ±0.023 ±0.547 ±0.049 ±0.579 a 0.303 ±0.003 ±0.482 ±0.465 ±0.446

LT2

±0.082 ±0.243 0.183 ±0.051 ±0.114 0.134 0.103 ±0.025 0.072 0.500 ±0.189 0.376 ±0.173 0.381 ±0.154 ±0.210 0.170 0.164 0.154

EMT2

±0.087 ±0.238 0.083 ±0.148 ±0.116 ±0.013 ±0.080 ±0.127 ±0.097 0.285 ±0.218 0.228 ±0.198 0.253 ±0.222 ±0.204 0.130 0.079 0.046 (continued)

EMT3

IJCST 12,5

±0.110 ±0.312 ±0.300 0.452 ±0.067 ±0.460 0.236 0.316 ±0.025 0.390 0.271 ±0.024 0.415 0.076 0.190 0.103

EMGD EMGL EML ES ET LAAC LABC LAMG MG PBS PSD PSL SCP SI SUC TGA

±0.119 ±0.133 ±0.427 0.474 ±0.052 ±0.522 0.288 0.324 ±0.078 0.403 0.397 ±0.045 0.400 0.157 0.214 ±0.091

LT2

RT1 0.105 ±0.221 a 0.106 0.002 ±0.204 ±0.159 ±0.084 ±0.164 0.175 ±0.088 0.096 0.179 ±0.022 0.054 ±0.017 0.008

LT3

±0.141 ±0.307 ±0.349 0.464 ±0.143 ±0.540 a 0.228 0.321 ±0.059 0.374 0.230 0.128 0.427 0.184 0.249 0.042 0.003 ±0.092 0.036 0.001 ±0.101 ±0.085 ±0.030 0.067 0.031 ±0.077 0.124 0.075 0.038 ±0.047 0.033 ±0.032

RT2

0.182 0.140 0.187 ±0.022 0.100 0.072 ±0.048 ±0.091 0.151 ±0.080 ±0.026 ±0.061 ±0.002 ±0.068 ±0.092 0.128

RT3 ±0.131 0.444 a 0.035 ±0.230 0.171 0.559 a 0.011 0.126 ±0.268 ±0.061 ±0.422 ±0.330 ±0.139 ±0.113 ±0.192 ±0.160

WT1 ±0.099 0.285 0.059 ±0.197 0.089 0.425 ±0.103 0.012 ±0.187 ±0.060 ±0.406 ±0.049 ±0.205 ±0.087 ±0.190 ±0.188

WT2

±0.173 0.084 ±0.063 ±0.102 0.055 0.213 ±0.127 ±0.013 ±0.207 ±0.051 ±0.202 0.093 ±0.196 0.118 ±0.035 ±0.289

WT3 ±0.094 0.470 a 0.107 ±0.300 0.143 0.618 a ±0.049 0.031 ±0.236 ±0.148 ±0.447 ±0.290 ±0.215 ±0.120 ±0.229 ±0.164

EMT1

Signifies the

±0.115 0.143 0.031 ±0.189 ±0.010 0.302 ±0.174 ±0.099 ±0.163 ±0.137 ±0.195 0.074 ±0.273 0.071 ±0.047 ±0.271 ±0.065 0.264 0.156 ±0.274 0.082 0.496 ±0.153 ±0.072 ±0.146 ±0.153 ±0.440 ±0.024 ±0.265 ±0.100 ±0.214 ±0.193 a

EMT3

EMT2

Notes: 78 cases of light-weight fabrics. The italic figures signify the highest correlation coefficient among the three sets of data. three highest correlation values among all the features

LT1

PAR

Basic low stress mechanical properties 329

Table AIII.

IJCST 12,5

330

Table AIV. The correlation between features and the KES bending parameters

PAR AAC ABC AF AGA AGL AH AHF ALF ATH AUD CPD EBC EBCC ECD ECP EE EF EH EHF EMGD EMGL EML ES ET LAAC LABC LAMG MG PBS PSD PSL SCP SI SUC TGA

B1

Bending rigidity B2

B3

2HB1

±0.301 0.347 ±0.321 ±0.228 ±0.145 ±0.310 ±0.297 ±0.220 ±0.290 ±0.569 a 0.022 ±0.525 0.001 ±0.498 0.225 0.099 ±0.291 ±0.310 ±0.318 ±0.344 ±0.161 ±0.323 0.531 ±0.219 ±0.574 a 0.149 0.242 ±0.069 0.403 0.441 0.334 0.428 0.190 0.384 ±0.037

±0.294 0.480 ±0.447 ±0.262 0.017 ±0.390 ±0.338 ±0.052 ±0.276 ±0.515 ±0.009 ±0.545 ±0.048 ±0.550 a 0.337 0.030 ±0.388 ±0.406 ±0.391 ±0.327 ±0.221 ±0.403 0.500 ±0.129 ±0.535 0.373 0.304 ±0.121 0.388 0.428 0.183 0.495 0.162 0.195 ±0.026

±0.109 0.381 ±0.290 ±0.106 0.051 ±0.218 ±0.138 0.042 ±0.095 ±0.353 0.014 ±0.351 ±0.011 ±0.356 0.254 0.042 ±0.293 ±0.275 ±0.249 ±0.129 ±0.080 ±0.233 0.439 ±0.082 ±0.376 0.233 0.338 ±0.047 0.363 0.223 0.053 0.441 0.056 0.317 0.075

0.078 0.068 0.246 0.161 ±0.052 0.164 0.112 ±0.022 0.081 ±0.113 0.308 ±0.061 0.316 0.047 ±0.041 0.413 a 0.304 0.266 0.245 ±0.157 0.210 0.232 0.190 ±0.332 ±0.099 ±0.083 ±0.108 0.266 0.085 0.048 0.071 0.067 ±0.049 0.016 0.151

Bending hysteresis 2HB2 2HB3 0.039 0.163 0.087 0.106 0.014 0.064 0.050 0.034 0.048 ±0.129 0.222 ±0.111 0.210 ±0.044 0.108 0.295 0.120 0.113 0.111 ±0.056 0.164 0.108 0.158 ±0.221 ±0.127 0.040 ±0.025 0.181 0.087 0.044 0.053 0.106 ±0.079 ±0.119 0.124

0.247 0.129 0.301 0.261 0.046 0.283 0.270 0.162 ±0.253 0.039 0.386 a 0.095 0.383 a 0.109 0.091 0.375 0.313 0.319 0.324 0.065 0.346 0.325 0.114 ±0.014 0.028 0.077 ±0.005 0.345 0.067 ±0.081 ±0.066 0.110 ±0.081 ±0.023 0.270

Notes: 78 cases of light-weight fabrics. The italic figures signify the highest correlation coefficient among the three sets of data. a Signifies the three highest correlation values among all the features

The current issue and full text archive of this journal is available at http://www.emerald-library.com

Mechanisms of ultrasonic joining of textile materials

Ultrasonic joining of textile materials

Weihua Shi and Trevor Little

College of Textiles, North Carolina State University, Raleigh, USA

331

Keywords Seams, Ultrasonic, Welding, Bonding, Woven fabrics, Fiber optics

Received April 1999 Abstract Investigates the potential for building smart seams by incorporating optic fibers Accepted February 2000 ultrasonically. The heating and bonding mechanisms of ultrasonic welding process in fabrics were studied. Battle dress uniform (BDU) (50/50 nylon/cotton), 100 percent cotton, 100 percent polyester and Nomex fabrics were used and were bonded ultrasonically with and without polyurethane adhesives. The effects of three important welding parameters, namely weld pressure, weld time and amplitude of vibration, on the joint strength and the temperature profile at the interface were examined. The temperature profiles for different fabrics were measured during ultrasonic welding process. The attenuation degree of signal transition properties of optic fibers incorporated was tested to determine if ultrasonic process provided a possible way of embedding optic fibers into seams and achieving sufficient joint strength while the signal transmission properties of optic fibers incorporated were not changed significantly.

Introduction Apparel is changing in terms of its functions by integrating ``smart'' materials. In addition to current comfort and aesthetic attributes, the garment of the future will include inherent ``circuitry'' to transform the traditional garment into an intelligent and individually protective clothing system. Seams and patches of the garment are latent places to incorporate these functions. This paper looks at the potentials for building ``smart seams'' which incorporate sensors or actuators and data from these paths communicate with a wearable transmitter. Optic fibers and ultrasonic bonded seams offer one way to fabricate intelligent clothing systems. In this paper, the state of the art is reviewed and the research to optimize the fabricating methodology is presented. Joining techniques Fabric, the most common form of textile materials for garments, can change shape by bending, shearing, stretching or contracting. Despite fabric's capability to drape over simple 3D objects, complex 3D shapes of human bodies require that garments must be made by joining panels of various 2D shapes. Many joining methods are available to garment manufacturers, such as sewing, ultrasonic bonding, thermal bonding and laser enhanced bonding (LEB), etc., each with their own advantages and disadvantages. Sewing, which joins individual panels together with another textile element (thread), provides adequate strength, elasticity and aesthetic properties. However, sewing produces discontinuous joints and perforated seams. To This work is sponsored by Army Research Office (Grant No. DAAH04-96-1-0018) and was conducted at the College of Textiles, North Carolina State University.

International Journal of Clothing Science and Technology, Vol. 12 No. 5, 2000, pp. 331-350. # MCB University Press, 0955-6222

IJCST 12,5

332

produce continuous and non-perforated seams, other joining methods are necessary. Thermal bonding (i.e. hot-tool and hot-gas welding), LEB and ultrasonic bonding involve melting and cooling of thermoplastics at the joining interface. These thermoplastic components may be present as hot-melt fibers, powders, films, or the sheath on bicomponent fibers. In the thermal bonding process, surfaces to be joined are melted individually by direct contact with heated elements and then brought together. The interfaces are allowed to cool and solidify under controlled pressure (Stokes, 1989). The main disadvantage of this technique is that excessive heat conducted through fibers can cause fiber degradation. Light Technologies Group and Union Special Corporation (Fraser and Whitwell, 1971) developed LEB as a high-speed, non-contact technique for welding thermoplastics. In this method, a laser (precisely controlled, highly localized infrared CO2 gas) is used to drive polymers into fabrics to create a seam. The laser offers a source of concentrated energy which produces thermal effects without direct physical contact, and thermoplastic materials absorb sufficient energy at this wavelength so that bonding temperatures are readily achieved. Properly controlled LEB produces flexible, impermeable and high strength bonds (Adhesive Age, 1995; Black, 1995), but highly focused and excessively intense laser radiation may degrade some thermoplastics (Potente and Uebbing, 1997). Ultrasonic bonding is an advanced technique for joining synthetic materials and blends to produce continuous and impermeable seams. Fabrics may be 100 percent synthetic (thermoplastic) or blends with up to 40 percent natural fiber content. It is very fast, clean and economic (no thread needed) for apparel fabricators. In ultrasonic bonding process, the material is fed between a vibrating ultrasonic horn and a stationary anvil (or a moving wheel). Highfrequency mechanical vibrations (20-40kHz) are transmitted through thermoplastic parts to generate a frictional heat built up at an interface, and hence to achieve a sufficient temperature to melt and bond the materials. Because the heat necessary for bonding is produced at the interface (not from the outside of materials), ultrasonic welding process starts exactly at the interface (Kuttruff, 1991). For non-synthetic fabrics or blends with more than 40 percent natural fiber content, heat-activated materials are placed between two pieces of fabrics. Ultrasonic vibrations and pressure cause heat-activated materials to melt and penetrate into the inter-fiber spaces of fabrics (Technical Report from Branson Ultrasonic Corp., 1996). Weldability and compatibility of ultrasonic welding of thermoplastics Although thermoplastics can be welded by ultrasound, different thermoplastics have different weldability because of their differences in structure, melt point, modulus and other additives, such as flame retardation and colorants, etc. (Technical Report from Branson Ultrasonic Corp., 1996). Welding dissimilar materials has to meet two requirements (Technical Report from Branson Ultrasonic Corp., 1996):

(1) A similar temperature between materials to be welded, which is a basic Ultrasonic requirement for successful welding of rigid parts (a temperature joining of textile difference of 228C can be sufficient to hinder bonding even for materials materials with like structures). This is because the lower melt temperature material melts and flows first, leveling the micro-irregularities at the interface and thus preventing generation of sufficient heat to melt the 333 higher melt temperature material (Mozgovoi et al., 1968). (2) A like molecular structure (i.e. chemically compatible) with some components of the material, usually a blend. The bonding strength of ultrasonically welded bulky thermoplastics Despite wide-spread applications of ultrasonic bonding techniques in the textile industry, not much research work has been done on the bonding and heating mechanisms of woven fabrics. Most of previous work was focused on ultrasonic welding of bulky polymers (Mozgovoi et al., 1968; Matsyuk and Bogdashevskii, 1960; Mordvinteseva and Druzhinin, 1964; Alosio et al., 1972; Chernyak et al., 1973; Frankel and Wang, 1980; Tolunay et al., 1983; Benatar et al., 1989) and thermal bonding of non-wovens (Chu and Yan, 1994; Dantuluri et al., 1987; Flood, 1992; 1994; Hassan and Rust, 1986; Non-woven Report International, 1982; Britton and Sampson, 1984; Warren and Partenio, 1994; Wei and Vigo, 1985; Lobomir, 1986; Lutzow, 1993). Matsyuk and Bogdashevskii (1960) studied the ultrasonic welding of lap joints between polymer pieces. During a welding cycle, interfacial temperatures were measured for some types of polymers (poly(methyl methacrylate), polyethylene, and poly(vinyl chlorides)). It was found that the highest weld strength was attained at an intermediate value of weld force. Mordvinteseva and Druzhinin (1964) found that ultrasonic bonding strength of polyethylene increased within a range of welding time and welding force at first and then decreased beyond that range. The study on microstructures of ultrasonically welded polyethylene indicated that no new chemical group or bond was formed, and the only noticeable change was a slight increase in molecular weight (Mozgovoi et al., 1968). The joint strength depends on the degree of fusion of the material at the interface. Frankel and Wang (1980) studied the ultrasonic welding of acrylonitrile-butadiene-styrene copolymer (ABS) and observed that, the highest joint strength was obtained in the case of longest weld time and lowest weld force from nine different welding conditions involving three levels of force and time. The rate of temperature rise increased with the increase of amplitude of vibrations. Current research In addition to factors affecting various resins (structure, glass transition temperature, melting point, specific heat capacity, melt index and modulus), ultrasonic welding of textile materials also depends on their thermoplastic content, fabricating types (woven or knit) and desired end results (Technical Report from Branson Ultrasonic Corp., 1996). The main objective of the present

IJCST 12,5

334

study, therefore, was to correlate the resulting bond strength with welding conditions and examine the nature of heat evolution in fabrics during ultrasonic welding. In the experiment, temperature profiles during ultrasonic welding were measured and determined in relation to the properties of the fabrics being welded and welding conditions. The weld strength was tested by fabric grab and tension tests. The microstructure of ultrasonically welded fabrics was examined in terms of chemical and physical characteristics, which gives a direct picture of the joint development in ultrasonic welding of fabrics. Based on these findings, the ultrasonic process for embedding optic fibers into seams to fabricate ``smart structures'' was studied. The morphologies of optic fibers embedded were examined under varying welding conditions and the power loss of signal transition of incorporated optic fibers was tested to determine their attenuation degree of signal transition properties. Experimental techniques Figure 1 shows the experimental configuration for ultrasonic welding and temperature detection, which includes a Branson 901 AES ultrasonic welding machine (actuator, power supply and weld profile controller), SC-2345 signal conditioning box (National Instruments Inc.), data acquisition board and a personal computer. The welding process was set under ``time'' mode. During welding, the microprocessor-controlled welder was used to record weld time, hold time, pressure, power, energy and down speed. The amplitude of vibrations can be adjusted by using different boosters. An anvil (James, 1997) (Figure 2) with a groove in the middle was developed to accommodate the placement of an optic fiber strand between the two fabrics during welding process. In our experiment, 0.008-inch-diameter thermocouples (type T) were placed at the interface of two fabric layers (Figure 3). During welding, the

Figure 1. Experimental configuration for ultrasonic welding and temperature detection

Ultrasonic joining of textile materials 335 Figure 2. Modified anvil for the placement of the optic fiber

Figure 3. Thermocouple placement in temperature profile experiment

temperature change at the interface was detected by the thermocouple and modified by the signal conditioning box, then the data was sent to a data acquisition board and a temperature profile graph was recorded. Ultrasonic welding was performed on five different sets of fabrics. The physical properties of materials used are listed in Table I. For the 50/50 nylon/ cotton blend BDU fabrics, polyurethane adhesive films were used between two fabric pieces. The reason we choose polyurethane is that it has excellent bond strength when ultrasonically bonded (Technical Report from Branson Ultrasonic Corp., 1996) and its soften temperature is close to that of nylon fibers (about 1108C). Factorial experiment design Ultrasonic welding strength is determined by degree of fusion, which is mainly dependent upon the temperature at the interface (Frankel, 1976). Thus, welding strength is correlated with the thermal energy at the interface. An ultrasonic weld is governed by process parameters, illustrated in Figure 4.

IJCST 12,5

Density (g/cm2)

Thicknessa (mm)

0.024 0.021 0.021 0.025 0.015 1.18

0.67 0.43 0.66 0.86 0.50 0.05

Materials

Description

336

100 percent cotton fabrics 100 percent PET fabrics 50/50 nylon/cotton BDU 50/50 nylon/cotton BDU Nomex fabrics Polyurethane filmb Optical fibersc

Table I. Properties of the materials used

Knit Ponte-de-Roma knit Light-weight knit Twill knit Degrading 3108C Soften 115-1358C Core/cladding/jacket diameter: 100/140/250m, SiO2/SiO2/plastic Core/cladding/jacket diameter: 210/230/270m, SiO2/soft plastic/plastic

Notes: a Measured at 0.5g force per cm2; b PUROX form Adhesive Films Inc.; c Fiber Optics and Photonics Manufacturing Centre, Material Engineering Department, Drexel University

Figure 4. Governing parameters for ultrasonic weld

From Figure 4, it can be seen that welding time, amplitude of vibration and welding pressure affect the amount of energy that goes into the weld area. Based on this, a 23 factorial experiment (Figure 5) was carried out to study the effects of weld time, pressure and amplitude of vibration with weld strength. Two levels were chosen for each parameter according to some preliminary experiments. Down speed (2.1 inches/second) and hold time (0.5 seconds) were kept constant. The BDU light-weight fabrics with or without polyurethane films, and BDU twill fabrics with polyurethane films were tested. Individual experiments were also carried out to investigate effects of weld pressure and weld time on the bonding strength. The weld pressures used were 20, 25, 30, 35 and 40psi. The weld times used were 1, 1.5, 2.0, 2.5 and 3.0 seconds. The amplitude (60m), down speed (2.1 inches/second) and hold time (0.5 seconds) were kept constant. The joint strength of various samples was tested in the direction perpendicular to the bond with a Sintech universal testing machine (UTM). The testing speed is 30mm/min. For each set of parameters, 20 welds were

Ultrasonic joining of textile materials 337

Figure 5. Graphical representation of 23 factorial design

performed and the strength was compared with the sewing strength. Strength tests were also carried out on laundered samples to test the durability of ultrasonically welded seams. Study on the microstructure of welded fabrics and optic fibers SEM analysis was carried out to examine the condition and morphology of fabrics and the incorporated optical fibers. Fourier transmission infrared (FTIR) test was performed to study the possible chemical conversions in the weld zone under the influence of ultrasonic vibrations. BDU fabrics and polyurethane films were examined by using attenuated total reflectance (ATR) connector before and after ultrasonic welding and their spectra were compared. The tests were carried out on Nicolet 510P FTIR spectrometer at frequencies from 400-4,000cm±1 with resolution of 32cm±1, scan = 128. Power loss of optic fibers incorporated in fabrics The samples embedded with optic fibers were tested in the Fiber Optics and Photonics Laboratory at Drexel University to see if the enclosed fiber was damaged by checking the continuation of the optic signals propagating within the fiber and the power loss. Results and discussion 23 factorial experiment Preliminary experiments showed that bonding strength increased with weld pressure up to 30psi and then decreased (as shown in Figure 6). It was also observed that a minimum pressure was required to produce satisfactory welds. As mentioned earlier, the purpose of weld pressure is to provide contact between horn and materials to transmit the vibrations and also to bond the heated surfaces. Low pressure provides poor contact and thus poor energy transmission. An increase in pressure will improve the weld strength initially, but higher pressures will orient polymer chains during ultrasonic welding and

IJCST 12,5

338 Figure 6. Effects of weld pressure on bond strength (weld time 2.5s; amplitude 60m)

decrease the weld strength (Mozgovoi et al., 1968; Benatar et al., 1989). Too great a pressure can also damage fibers and reduce the strength of fabrics. A similar trend was observed in the effect of weld time on bond strength (Figure 7). The reason is that energy dissipation increases with the increase of weld time and thus the strength increases until the maximum strength is achieved. If weld time is too long, this may damage cotton fibers (char or burn) and thus decrease the bond strength. The average peak loads for seam strength of BDU light-weight fabrics with polyurethane films are listed in Table II. It can be seen that the seam strength increased up to 80 percent when the amplitude was increased from 42-60m keeping other parameters constant. Compared with the amplitude, the effects of weld pressure and weld time on strength are smaller but important (up to 15 percent). Therefore, amplitude of vibrations seems to be the most important variable affecting weld strength. The maximum weld strength was obtained

Figure 7. Effects of weld time on bond strength (weld pressure 30psi; amplitude 60m)

Exp. no.

Welding conditions Actual setting Coded variables WT Amp X2 X3 (seconds) (m) X1

1 2 3 4 5 6 7 8

± + ± + ± + ± +

± ± + + ± ± + +

± ± ± ± + + + +

2.0 2.5 2.0 2.5 2.0 2.5 2.0 2.5

42 42 60 60 42 42 60 60

Pr (psi) 30 30 30 30 40 40 40 40

Average peak load (psi) 18.0 17.4 24.0 33.2 17.1 20.8 29.4 28.2

Energy (joules)

1.3 2.1 2.8 4.4 2.4 2.5 4.5 3.8

652 800 788 958 827 891 835 1,031

Notes: ±, lower level; +, higher level; WT, weld time; Amp, amplitude of vibrations; Pr, weld pressure

Ultrasonic joining of textile materials 339 Table II. Experimental design and results for BDU light-weight fabrics with polyurethane films

under conditions of the longer weld time (2.5 seconds), the lower weld pressure (30psi) and the higher amplitude of vibration (60m), which had energy input of 957 Joules. The weld strength reached a saturation value at this energy input (957 Joules) and further increasing in energy input inversely affected the strength. This observation is supported by the fact that increasing the energy input from 958 Joules to 1,031 Joules decreased the strength 15 percent. Similar observation has been reported in another paper (Wei and Vigo, 1985). The average peak load for seam strength of BDU light-weight fabrics without additional adhesives and BDU twill fabrics with polyurethane films are listed in Tables III and IV, respectively. Compared with the data in Table II, the average peak loads for seam strengths in Table III are much lower that those in Table II under the same welding conditions. Although the average peak loads in Table IV are higher in most cases than those in Table II, all the bond failure occurred because of fabric

Exp. no.

Welding conditions Actual setting Coded variables WT Amp X2 X3 (seconds) (m) X1

1 2 3 4 5 6 7 8

± + ± + ± + ± +

± ± + + ± ± + +

± ± ± ± + + + +

2.0 2.5 2.0 2.5 2.0 2.5 2.0 2.5

42 42 60 60 42 42 60 60

Pr (psi) 30 30 30 30 40 40 40 40

Average peak load (psi) 9.3 13.3 8.4 10.8 8.5 14.0 9.5 13.3

Notes: ±, lower level; +, higher level; WT, weld time; Amp, amplitude of vibrations; Pr, weld pressure

1.7 1.7 2.0 1.6 1.3 1.7 1.6 2.0

Table III. Experimental design and results for BDU light-weight fabrics without additional adhesives

IJCST 12,5

340 Table IV. Experimental design and results for BDU twill fabrics with polyurethane

Exp. no.

Welding conditions Actual setting Coded variables WT Amp X2 X3 (seconds) (m) X1

1 2 3 4 5 6 7 8

± + ± + ± + ± +

± ± + + ± ± + +

± ± ± ± + + + +

2.0 2.5 2.0 2.5 2.0 2.5 2.0 2.5

42 42 60 60 42 42 60 60

Pr (psi) 30 30 30 30 40 40 40 40

Average peak load (psi) 27.7 30.6 25.4 28.5 26.1 29.5 24.2 34.2

3.0 4.0 2.3 2.2 2.9 3.4 1.9 3.0

Notes: ±, lower level; +, higher level; WT, weld time; Amp, amplitude of vibrations; Pr, weld pressure

failure. Therefore, the peak load obtained here is more indicative of the strength of fabrics. For the results in Table IV, the effect of amplitude of vibrations, weld time and pressure on the strength is not as apparent as observed earlier. Comparison of the strength of ultrasonic weld and sewing seams The optimum ultrasonic bonding strength in our experiment is equivalent to sewing strength (about 30lbf/inch). Strength tests were also carried out on samples after up to five laundry cycles (Figure 8). Machine-washing had no statistically significant effect on bonding strength. Failure modes of the ultrasonic bond Three kinds of failure modes for ultrasonically welded seams were found: (1) failure in bond without damaging fabrics; (2) damage in fabrics; and (3) a combination of (1) and (2).

Figure 8. Effect of washing cycles on ultrasonic bonding strength of BDU light-weight fabrics with polyurethane films (weld time 2.5 seconds; amplitude 60m; pressure 30psi)

The first mode of failure is desirable in that the seam is strong and ductile with Ultrasonic yield points being observed in the stress-strain curves. The correlation between joining of textile the weld parameters, the properties of materials and the failure modes need to materials be investigated further. SEM analysis Samples bonded with adhesive films. Figure 9(a)-(f) show SEM micrographs for the cross section of original fabric and specimens obtained at weld time 0.2, 0.5, 1, 1.5 and 2.5 seconds, with weld pressure 30psi and amplitude of vibrations 60m. Figure 9(a) shows the cross section of BDU fabrics before welding, in which those with larger diameter are cotton fibers and those with smaller diameter are nylon fibers. At weld pressure 30psi and amplitude of vibration 60m, applying the vibration of 0.2 seconds produced the section which is mainly the original structure with the adhesive film softened and deformed (Figure 9(b)). Increasing the weld time to 0.5 seconds and 1 second results in melted or partially melted nylon fibers and polyurethane film (Figure 9(c) and (d)). At weld times of 1.5 up to 2.5 seconds, melting flow is visible and sufficient. Also from Figure 9, the fibers at the interface have undergone more melting than those towards the outsides of the sample (individual fibers are obvious). The ultrasonic vibrations apply rapid impacts and cyclic tangential stress to the sample surfaces against each other. The considerable stress in microroughness peaks brings about their plastic deformation. The material which becomes plastic around micro-roughness peaks has increased sound absorption and continues to be preferaby heated and hence the heating process at these points will be accelerated in an avalanche-like manner (Kuttruff, 1991; Abramov, 1994). At the beginning of a weld cycle, the asperities at the sample interface are pronounced, and they generate heat first, then dissipate heat to regions near the ultrasonic horn and anvil. The thermal energy causes the temperature rise at the interface and hence increases the mobility of polymer chains and promotes intense intermingling of polymer chains at the interface. When the vibration is removed and the interface is cooled down, the intermingling state of polymers is retained and results in a welded seam. Seams incorporating optic fibers. Figure 10(a)-(d) illustrate the seams incorporating optic fibers, which were made at weld times of 0.2 seconds, 0.5 seconds, 1 second and 1.5 seconds, weld pressure 30psi and amplitude of vibration 42m. At a weld time 0.2 seconds, the two fabric layers were separated from each other and no obvious change occurred in the optic fiber (Figure 10(a)). With the increase of weld time, the interface of the two fabric layers began to fuse (Figure 10(b) and (c)). At a weld time 1.5 seconds, although a good weld resulted, the optic fiber was damaged and a crack can be seen clearly along the interface of the core and cladding (Figure 10(d)). Temperature profiles for ultrasonic welding process Figure 11 illustrates the temperature traces for four different fabrics, which were recorded by thermocouples placed at the weld interface. From these

341

IJCST 12,5

342

Figure 9. SEM micrographs of the melting condition at the interface of the BDU fabrics with polyurethane films: (a) before welding; (b)-(f) weld pressure 30psi and amplitude 60m. WT = weld time

Ultrasonic joining of textile materials 343

Figure 10. SEM micrographs for ultrasonic welding seams incorporated with optic fibers (weld pressure 30 psi; amplitude 42m). WT = weld time

Figure 11. Temperature profiles in ultrasonic welding process for four kinds of fabrics (weld time 0.5 seconds; pressure 30psi; amplitude 60m)

IJCST 12,5

344

Table V. Thermal conductances for fabrics used

graphs, some general trends in the thermal profiles recorded are apparent. At the beginning of a weld cycle, interfacial temperatures increased rapidly and reached peak temperatures within about 0.5 seconds, and then temperatures decreased rapidly after the ultrasonic vibration was removed. However, for 100 percent PET fabrics, the temperature rose rapidly at the beginning of a weld cycle and the peak temperature (2258C) was reached in about 0.35 seconds (before end of weld time). The rest time of the weld cycle was characterized by cooling at the interface. Such observation has been reported (Wei and Vigo, 1985). For the other fabrics in Figure 11, the peak temperatures were reached in approximately 0.5 seconds and when ultrasonic vibrations was stopped at 0.5 seconds, the temperatures dropped dramatically. At 0.7 seconds, the interfacial temperature for 100 percent PET was about 608C, and approximately 758C for Nomex, 808C for BDU twill, 1008C for 100 percent cotton. After 0.7 seconds, an unexpected increase in temperature was observed in 100 percent PET fabric. Among the graphs in Figure 11, the rate of temperature change is the highest for 100 percent PET fabric and it is followed by Nomex, 50/50nylon/ cotton BDU and 100 percent cotton fabrics. Thermal conductivity affects the absorption of ultrasonic wave in solids and hence the temperature variation (Bhatia, 1967; Ensminger, 1988). The thermal conductances for four different fabrics used are listed in Table V. It can be seen that the higher the thermal conductance of fabrics, the larger the changing rate of temperature. The difference in measured rates of temperature rise among the thermal graphs of Nomex, BDU twill and 100 percent cotton fabrics was not obvious in the initial period (until 0.2 seconds), which may be caused by the melting condition of weld areas and precision of thermocouples used. Also from Figure 11, different peak temperatures were observed for different fabrics under the same welding condition. Among these fabrics, Nomex achieved the highest peak temperature during welding process and was followed by 100 percent PET and BDU and 100 percent cotton. This may be related with melt temperatures and morphology of fibers under weld and the attenuation of ultrasonic waves, which need to be investigated further. In the experiments, no weld was obtained for 100 percent cotton or Nomex fabrics because these fibers cannot melt and flow although they can be heated by absorbing ultrasonic vibrations. Nomex will degrade above 3718C and cotton will be ``charred'' above 2508C. The function of weld pressure is to bring two surfaces into contact and ensure vibration energy transmitted. The effect of changing weld pressure is Fabrics

PET

Nomex

BDU

Cotton

Thermal conductance (Wm±28C±1)

91.60

80.00

82.00

78.00

Notes: Tested by Thermal and Protective Apparel Lab in North Carolina State University. Thermal conductance = thermal conductivity/thickness of fabrics at 6g force

shown in Figure 12. The peak temperature reached in curve 1 is higher than Ultrasonic that in curve 2. The energy input increases with the increase of weld pressure. joining of textile Increasing the weld pressure from 30psi to 40psi caused the energy input to materials increase from 957 Joules to 1,031 Joules and the peak temperature increased from about 2508C to 2808C. As indicated before, amplitude of vibrations strongly affects the energy 345 dissipation and thus determines the weld strength. The decrease of amplitude from 60m to 42m decreased temperature-rising rates and peak temperatures reached at the interface from 2808C to 1908C (Figures 13 and 14). Thus a decrease in the amplitude of vibrations requires longer weld time to attain the same level of weld strength. During ultrasonic welding, the temperature at the interface of two welded pieces is different in different fabrics and welding parameters used. Thermal conductance is found to be a factor affecting rates of temperature change during welding process. The higher the thermal conductance of fabrics, the larger the rate of temperature change. For the same fabric, the increase in weld pressure or amplitude of vibrations increases the energy dissipated, and thus increases the rate of temperature change and the peak temperature attained. Temperature is critical to a successful joint. If the temperature is too high at the weld area, the materials to be welded will be over-welded or degraded. If the temperature is too low, the materials cannot melt and flow sufficiently and a poor weld strength will result.

Figure 12. The effect of weld pressure on temperature profiles (fabric: BDU light-weight; weld pressure: curve 1, 40 psi; curve 2, 30 psi; weld time 2.5 seconds; amplitude 60m)

IJCST 12,5

346

Figure 13. The effect of amplitude on the welding strength (fabric: BDU twill; weld time 0.5 seconds; pressure 30psi)

Figure 14. The effect of amplitude on the welding strength (fabrics: Nomex; weld time 0.5 seconds; pressure 30psi)

FTIR-ATR spectra Figures 15 and 16 are FTIR-ATR spectra for BDU fabrics and polyurethane films before and after ultrasonic welding. The spectra show that no new chemical groups have been formed under these ultrasonic bonding conditions used. The ultrasonic bonding mechanism is a physical process, rather than a chemical process. It results from molecular diffusion and inter-fiber encapsulation at high temperature and when ultrasonic vibrations are turned off, such state is stiffened and a joint is attained.

Ultrasonic joining of textile materials 347

Figure 15. Comparison of FTIR-ATR spectra for polyurethane films before and after ultrasonic welding

Figure 16. Comparison of FTIR-ATR spectra for BDU fabrics before and after ultrasonic welding

IJCST 12,5

Survival of optic fibers The test on the power loss of optic fibers embedded (El-sherif, 1998) indicated that, the signal transmission properties of the optic fibers were not significantly changed under certain welding condition. Therefore, optic fibers can be integrated into fabrics ultrasonically.

348

Conclusions Ultrasonic welding experiments were performed on three kinds of fabrics as well as BDU fabric incorporating optic fibers as data channels. The effects of changing welding time, welding pressure and amplitude of vibrations were studied in terms of joint strength, thermal profiles at the interface during welding process, and the morphology of fibers at weld areas. Some conclusions can be drawn from the experiments: (1) Increasing the weld pressure (up to 30psi) increases the weld strength initially, but further increases in pressure cause a decrease in the strength. The same effect was observed in changing weld time, the critical value being 2.5 seconds. (2) The amplitude of vibrations strongly affects the bonding strength obtained. Increasing the amplitude of vibrations (from 42 to 60m) enhances the energy dissipated and strength. (3) The strongest welds were obtained in BDU light-weight fabrics with polyurethane films at the condition: weld pressure 30psi; weld time 2.5 seconds; and amplitude of vibrations 60m. (4) Under the same weld condition, the joint strength with polyurethane films was much higher than that without adhesives. No significant decrease in the strength was observed after up to five laundry cycles. (5) The specimens with polyurethane films failed in three modes: . fabric damage; . bond damage; and . a combination of fabric and bond damage. The second mode is desirable because the seam is strong and flexible, and yield point was observed in this mode. (6) SEM analysis indicated that melting of polymeric fibers occurred first at the interface. This is because the asperities at the interface are generally more pronounced and the attenuation of vibration energy there is more severe. (7) The rate of temperature change and the maximum temperature developed in the weld area depend on the thermal properties of fabrics and welding parameters (i.e. pressure and amplitude of vibrations). A reduction in amplitude of vibrations results in a decrease in the temperature rising rate and the maximum temperature achieved at the

interface. An increase in pressure leads to an increase in the maximum Ultrasonic temperature obtained. Under the same weld condition, the peak joining of textile temperatures and temperature rising rates for different kinds of fabrics materials are different because of their different thermal conductances. It was found that higher thermal conductances usually resulted in higher rates of temperature change. 349 (8) FTIR-ATR analysis indicated that, no considerable formation of new chemical groups or bonds was found. (9) A compromise is necessary between the joint strength and the survival of optic fibers, in that too long weld times can melt the cladding of optic fibers and damage their signal transition properties. References Abramov, O.V. (1994), Ultrasound in Liquid and Solid Metals, CRC Press, Boca Raton, FL. Adhesives Age (1995), ``Laser enhanced bonding could replace sewing'', Vol. 38, July, p. 40. Aloisio, C.J., Wahl, D.G. and Whetsel, E.E. (1972), ``A simplified thermoviscoelastic analysis of ultrasonic bonding'', Annual Technical Conference, Society of Plastics Engineers, pp. 445-51. Benatar, A., Eswaran, R.V. and Nayar, S.K. (1989), ``Ultrasonic welding of thermoplastics in the near-field'', Polymer Science and Engineering, Vol. 29 No. 23, pp. 1689-98. Bhatia, A.B. (1967), Ultrasonic Absorption: An Introduction to the Theory of Sound Absorption and Dispersion in Gases, Liquids and Solids, Oxford University Press, New York, NY. Black, S.S. (1995), ``Union special to debut laser-enhanced bonding'', Bobbin, August, pp. 98-104. Britton, P.N. and Sampson, A.J. (1984), ``Computer simulation of the mechanical properties of non-woven fabrics, Part II: bond breaking'', Textile Research Journal, January, pp. 1-5. Chernyak, B.Y. et al. (1973), Welding Production, Vol. 20 No. 87. Chu, C. and Yan, H. (1994), ``A study of high loft nonwoven fabrics using powder bonding technology'', Journal of China Textile University, Vol. 11 No. 1, pp. 1-7. Dantuluri, S.R., Goswami, B.C. and Vigo, T.L. (1987), ``Thermally bonded polyster non-wovens ± effect of fiber morphology'', INDA-TEC., Books of Papers, 1987, pp. 263-79. Ensminger, D. (1988), Ultrasonics, Marcel Dekker, New York, NY. El-sherif, M. (1998), ARO Report, March. Flood, G. (1992), ``Bonding technologies: ultrasonic bonding'', Nonwoven Industry, October, pp. 36-44. Flood, G. (1994), ``Ultrasonic energy: a process for laminating and bonding non-woven web structure'', Journal of Coated Fabrics, October, Vol. 14, pp. 71-81. Frankel, E.J. (1976), ``Ultrasonic welding of thermoplastics'', Project Report, Cornell University. Frankel, E.J. and Wang, K.K. (1980), ``Energy transfer and bond strength in ultrasonic welding of thermoplastics'', Polymer Engineering and Science, Vol. 20 No. 6, pp. 396-401. Fraser, W.A. and Whitwell, J.C. (1971), ``Infrared laser bonding of thermoplastic monofilaments'', Textile Research Journal, December, pp. 1003-5. Hassan, M.B. and Rust, J.P. (1986), ``Effects of production variables on the properties of ultrasonically bonded nonwovens'', INDA-TEC., June 2-5, pp. 120-37. James, A. (1997), ``Joining of textile materials using ultrasonics for fabricating smart material systems'', Masters' thesis, College of Textiles, North Carolina State University, NC. Kuttruff, H. (1991), Ultrasonics Fundamentals and Applications, Elsevier Science, Amsterdam.

IJCST 12,5

350

Lobomir, S. (1986), ``Theory and technology of ultrasonic bonding of non-woven and laminated textiles ± part 2'', Textiles, Vol. 41, pp. 350-4. Lutzow, T. (1993), ``Ultrasonic bonding'', Nonwoven Industry, October, Vol. 24, pp. 48-52. Matsyuk, L.N. and Bogdashevskii, A.V. (1960), Soviet Plastics, Vol. 2 No. 30. Mordvinteseva, A.V. and Druzhinin, N.V. (1964), Soviet Plastics, Vol. 9. Mozgovoi, I.V., Konstantinopol'skaya, M.B., Berestneva, Z.Ya., Kargin, V.A. and Nikolaev, G.A. (1968), ``Mechanism of bonding of plastics by ultrasonic welding'', Soviet Plastics, No. 11, pp. 51-4. Non-woven Report International, Shirley Project on Pinsonic (1982), December, No. 128, pp. 1-2. Potente, H. and Uebbing, M. (1997), ``Friction welding of polyamides'', Polymer Engineering and Science, Vol. 37 No. 4, April, pp. 726-36. Stokes, V.K. (1989), ``Joining methods for plastics and plastic composites: an overview'', ANTEC '89, pp. 442-5. Technical Report from Branson Ultrasonic Corp. (1996). Tolunay, M.N., Dawson, P.R. and Wang, K.K. (1983), ``Heating and bonding mechanisms in ultrasonic welding of thermoplastics'', Polymer Engineering and Science, Vol. 23 No. 13, pp. 726-33. Warren, F.M. and Partenio, A.C. (1994), ``The effects of improved ultrasonic technology on continuous bonding of non-woven fabrics'', Tappi Journal, June, Vol. 77 No. 6, pp. 211-15. Wei, K.Y. and Vigo, T.L. (1985), ``Structure-property relationships of thermally bonded polypropylene non-wovens'', Journal of Applied Polymer Science, Vol. 30, April, pp. 1523-34.

The current issue and full text archive of this journal is available at http://www.emerald-library.com

The quality of fabric knitted from cotton Sirospun1 yarn

Fabric knitted from cotton Sirospun1 yarn

M.N. Sun and K.P.S. Cheng

Institute of Textiles and Clothing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

351

Received November 1998 Revised January 2000 1 Accepted January 2000 Abstract The fine gauge plain weft knitted fabrics knitted from cotton Sirospun yarns are more durable and suitable for summer wear. It was found that they have considerable bursting strength, superior abrasion resistance, superior pilling resistance, greater air resistance, cooler hand-feel and greater thermal conductivity than the fabrics knitted from two-fold yarns. The plain weft knitted fabric composed of coarser Sirospun1 yarn is also better in terms of hand-feel measured by KES instruments. Keywords Cotton yarns, Woven fabrics

Plain weft knitted fabric produced from Sirospun1 yarn for clothing was seldom studied (Helw et al., 1988; Vijayakumar et al., 1991). In the present investigation, plain weft knitted fabrics were knitted from combed cotton Sirospun yarns at three yarn linear densities which are popular in Hong Kong and were compared with those knitted from conventional two-fold yarns. The Sirospun yarns were spun at the optimum condition, with twist multiplier of 31.6tpc tex1/2 and strand spacing of 9mm, which was identified in the previous work (Cheng and Sun, 1998). In addition to testing of the fundamental fabric properties, fabric handle was studied by the KES-F instruments for the compression, surface friction and roughness, bending, shear and tensile properties. Air resistance was measured by the KES-F8 Air Permeability Tester. Warm/cool feeling and thermal conductivity were measured by the KES-F7 Thermo Labo II tester. Materials and method Sirospun yarns of 36.9, 28.1 and 18.5tex were manufactured from 521tex combed roving on Spintester with twist factor of 31.6tpc tex1/2 and strand spacing of 9mm. Two-fold yarns were also spun and doubled to achieve the same linear densities and twist as the Sirospun yarns. They were knitted into plain weft knitted fabric. Sirospun and two-fold yarns of 36.9 and 28.1tex were knitted on the ALBI-Mashinenefabrik Circular Knitting Machine with a diameter of 40.64cm and 804 needles. Positive feed and the same knitting tension were used. Sirospun and two-fold yarns of 18.5tex were knitted on the Precision Fukuhara Circular Knitting Machine with a diameter of 50.8cm and 1,704 needles. The same knitting tension was used. All samples were conditioned and tested under the standard atmospheric condition, that is, temperature at 20 28C and relative humidity at 65 2 per cent. Fabric weight was averaged in grams per square metre by weighing three circular specimens, each of 100cm2, which were cut at random by using the Branca Idealair Cutting Die. Number of wales and courses were counted at

International Journal of Clothing Science and Technology, Vol. 12 No. 5, 2000, pp. 351-359. # MCB University Press, 0955-6222

IJCST 12,5

352

three random places with a pick glass of size 2.54cm 2.54cm. The numbers of wales and courses per centimetre were then averaged. Loop length was obtained by dividing the average course length by 100 loops. Ten consecutive threads consisting of 100 loops were ravelled and were measured on the Shirley Crimp Tester. The tension used was one-fourth of the yarn linear density in tex. Three random places were selected and a total of 30 threads were measured. Shrinkage percentages were averaged in the wale and course direction. Tubular specimens of 65cm long were cut randomly from each sample. Six pairs of 50cm gauge marks along the wales and six pairs of 20cm gauge marks along the courses were marked with indelible ink. Specimens and dummy with 1.8kg load were washed in the Kenmore Automatic Washing Machine at 508C, low water level and permanent press cycle. Then they were dried by the Kenmore Automatic Tumble Dryer at the permanent press cycle. Three washing and drying cycles were completed. In order to find out the fabric spirality, two individual wales were randomly selected and traced on the specimens tested for the shrinkage. Their displacements, if any, from the perpendicular at 50cm length were measured. After three cycles of washing and drying, the displacement was measured again and averaged. Abrasion resistance was carried out on the Martindale Abrasion Tester. Four circular specimens of each sample were cut randomly and were abraded under 9kPa pressure. The number of revolutions was recorded till a single thread was broken. The four readings were then averaged. Pilling resistance was carried out in the ICI Pilling Box. Four specimens, each of 125mm 125mm, were cut randomly from each sample. Two specimens were folded face to face with the wales running in the direction of fold. The other two specimens were folded face to face with the courses running in the direction of fold. They were sewn 12mm from the cut edges to form tubular specimens. After they were turned inside out and wrapped over the polyurethane tubes, they were fixed by adhesive tape. They were pilled at 60rpm for five hours. Then they were rated in the Standard Viewing Cabinet, and the ratings were averaged. Fabric thickness, surface friction and roughness, bending rigidity, shear rigidity and tensile properties were measured by the KES-F Compression, Surface, Bending, Shear and Tensile Testers respectively. Three specimens, each of 20cm 20cm, were measured at knit standard measurement for each tester. The results were then averaged. Air resistance was measured five times for each sample at a piston speed of 20mm/second. Warm/cool feeling and thermal conductivity were measured three times for each sample at an area of 25cm2 and 108C temperature difference between the hot plate and the base plate. Discussion and conclusion Plates 1-6 show the structure of the plain weft knitted fabric knitted from 36.9, 28.1 and 18.5tex two-fold and Sirospun yarns (magnification 20). Table I shows the test results. All plain weft knitted fabrics produced from Sirospun yarns show spirality in the same direction as the yarn twist direction. The finer the yarn, the more is the spirality because of the unbalanced torque of the

Fabric knitted from cotton Sirospun1 yarn 353

Plate 1. Plain weft knitted fabric knitted from 36.9tex two-fold yarn

Plate 2. Plain weft knitted fabric knitted from 36.9tex Sirospun yarn

IJCST 12,5

354

Plate 3. Plain weft knitted fabric knitted from 28.1tex two-fold yarn

Plate 4. Plain weft knitted fabric knitted from 28.1tex Sirospun yarn

Fabric knitted from cotton Sirospun1 yarn 355

Plate 5. Plain weft knitted fabric knitted from 18.5tex two-fold yarn

Plate 6. Plain weft knitted fabric knitted from 18.5tex Sirospun yarn

Notes:

a

kPa revs ± KPa.s/m W/m.K cm3/gf W/cm2

21.7 4.17 3 889.5 32,775 3.30 0.059 0.0464 4.4869 0.072

cm cm % %

0 3.75

g/m2

22.1 4.58 3 900.6 45,000 3.50 0.060 0.0491 4.4490 0.072

4.9 8.9

36.1 11 12 3.81 15.8 171.5

36.9S

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

Loop length Tightness factor Fabric weight Spirality before washing after washing Dimensional stability Walewise shrinkage Coursewise shrinkage Pilling resistance Bursting strength Abrasion resistance Total hand value Air resistance Thermal conductivity Specific volume Warm/cool feeling

36.9F 35.8 11 13 3.76 15.9 167.6

tex wpc cpc mm

Yarn linear density Construction

Table I. Fabric properties Units

14.8 11.3 2-3 725.7 17,400 3.51 0.050 0.0435 4.4588 0.071

3.9 5.7

28.0 12 14.5 3.32 15.9 152.3

28.1F

16.3 8.17 3-4 770 21,200 3.41 0.053 0.0456 4.6489 0.077

6.25 11.2

28.5 11 15 3.40 15.7 160.9

28.1S

12.3 7.08 3 481.1 11,000 3.63 0.024 0.0441 6.0836 0.065

0 5.9

18.9 12 15 3.17 13.7 102.9

18.5F

356

Test items

Yarn typesa

18 8.58 4-5 479 19,350 3.03 0.036 0.0532 6.0565 0.074

19.5 16.9

18.2 14 16 3.16 13.5 124

18.5S

IJCST 12,5

Sirospun yarn. The loop shape is so distorted that the wales are not lying flat and the fabric is slightly thicker. The plain weft knitted fabrics of the two-fold yarns are less spiral. They show balanced loops and flat surface. Referring to Table I, the plain weft knitted fabrics produced from Sirospun yarn shrink more after washing because of relaxation from tension and the increased spirality after washing. The pilling resistance of the plain weft knitted fabrics produced from the Sirospun yarns are good because the Sirospun yarns are less hairy than the two-fold yarns. Plates 2, 4 and 6 reveal that the plain weft knitted fabrics of the Sirospun yarns are less hairy. As a result, less fuzz is produced. The trend in less pilling is obvious as the yarn becomes leaner since the decrease in hairiness in lean yarn is greater in extent. The abrasion resistance of the plain weft knitted fabrics of the Sirospun yarns is superior than that of the plain weft knitted fabrics of the two-fold yarns. This is due to the strand twist and the surface fibre trapping mode in the Sirospun yarn. The surface fibres are trapped into the structure of the yarn more securely so that they are not easily rubbed out. While the fibres of the Sirospun yarn are still twisted in the fabric, the fibres of the two-fold yarn lie parallel to the yarn axis in the fabric and are easier to be rubbed out. All abrasion resistance of the knitted fabrics exceeded the minimum performance requirement of 10,000 rubs for blouse/shirt. Thus the plain weft knitted fabrics produced from the Sirospun yarns spun at the optimum conditions are serviceable. The bursting strength of the plain weft knitted fabrics of the Sirospun yarns is comparable to that of the plain weft knitted fabrics of the two-fold yarns. All bursting strength exceeded the minimum performance requirement of 275kPa. The KES results (see Tables II-VI) show that the plain weft knitted fabric produced from the Sirospun yarn is thicker, softer, of poorer recovery from compression, less smooth, less rough and less stiff. Although it is less easy to be sheared, it is easier to be extended and better in recovery from extension. For finer Sirospun yarn, the shape retention is worse as the hysteresis of bending and shear is greater. The hand-feel of the plain weft knitted fabrics of Sirospun yarn becomes worse while that of the plain weft knitted fabrics of the two-fold yarn improves as the yarn becomes leaner. Only the hand-feel of the fabric knitted from the Sirospun yarn of 36.9tex is better than that of the fabric knitted from the two-fold yarn of 36.9tex. Parameters

Units

36.9F

36.9S

Yarn typesa 28.1F 28.1S

18.5F

a

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

357

18.5S

LC: linearity of compression curve 0.279 0.250 0.274 0.270 0.280 0.302 WC: compression energy g.cm/cm2 0.393 0.374 0.354 0.378 0.358 0.416 RC: compressional resilience % 41.00 34.25 38.57 36.23 40.23 40.98 mm 1.192 1.229 1.123 1.178 1.034 1.178 T0: thickness at 0.5g/cm2 mm 0.628 0.630 0.605 0.618 0.519 0.624 TM: thickness at 50g/cm2 Notes:

Fabric knitted from cotton Sirospun1 yarn

Table II. KES compression results

IJCST 12,5

358 Table III. KES surface friction and roughness results

Parameters

Units

MIU: coefficient of friction

± 0.222 ± 0.286 ± 0.254 ± 0.0107 ± 0.0399 ± 0.0253 micron 2.618 micron 12.192 micron 7.405

Warp Weft Mean MMD: mean deviation Warp of coefficient of Weft friction Mean SMD: mean deviation Warp of surface roughness Weft Mean Notes:

a

B: bending rigidity 2HB: hysteresis of bending Notes:

a

Warp Weft Mean Warp Weft Mean

2HG: hysteresis of shear stiffness at 0.58 2HG5: hysteresis of shear stiffness at 58 a

0.223 0.310 0.267 0.0155 0.0419 0.0287 5.284 13.232 9.258

0.240 0.337 0.289 0.0097 0.0351 0.0224 2.428 15.136 8.782

18.5F

18.5S

0.226 0.300 0.263 0.0111 0.0362 0.0236 4.644 15.693 10.168

0.259 0.352 0.305 0.0087 0.0296 0.0192 2.098 10.699 6.399

Units

36.9F

36.9S

g.cm2/cm g.cm2/cm g.cm2/cm g.cm/cm g.cm/cm g.cm/cm

0.0456 0.0205 0.0331 0.0586 0.0288 0.0437

0.0407 0.0177 0.0292 0.0567 0.0294 0.0431

0.0313 0.0153 0.0233 0.0469 0.0251 0.0360

0.0294 0.0124 0.0209 0.0478 0.0241 0.0359

18.5F

18.5S

0.0221 0.0087 0.0154 0.0265 0.0151 0.0208

0.0234 0.0062 0.0148 0.0412 0.0174 0.0293

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

G: shear stiffness

Notes:

0.219 0.282 0.251 0.0095 0.0350 0.0223 2.491 11.581 7.036

Yarn typesa 28.1F 28.1S

Parameters

Table V. KES shear test results

Yarn typesa 28.1F 28.1S

36.9S

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

Parameters

Table IV. KES bending rigidity results

36.9F

Warp Weft Mean Warp Weft Mean Warp Weft Mean

Units

36.9F

36.9S

Yarn typesa 28.1F 28.1S

18.5F

18.5S

g/cm.deg g/cm.deg g/cm.deg g/cm g/cm g/cm g/cm g/cm g/cm

0.60 0.53 0.56 1.91 1.90 1.91 2.09 2.03 2.06

0.57 0.54 0.56 1.68 1.77 1.72 1.93 1.93 1.93

0.58 0.47 0.52 1.88 1.60 1.74 2.11 1.75 1.93

0.46 0.34 0.40 1.38 1.12 1.25 1.54 1.19 1.37

0.55 0.53 0.54 1.19 1.54 1.37 1.36 1.65 1.51

0.55 0.60 0.58 1.72 1.63 1.68 1.82 1.82 1.82

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

The air resistance of fabrics knitted from the Sirospun yarns of 28.1 and 18.5tex is greater than that knitted from the two-fold yarns of the same tex. The thermal conductivity of the fabrics knitted from the Sirospun yarns is greater than that of the fabrics knitted from the two-fold yarns. Having the fabric thickness measured at 6gf/cm2 (the pressure where the KES thermal

Parameters

Units

36.9F

36.9S

Yarn typesa 28.1F 28.1S

18.5F

18.5S

LT: linearity of tensile Warp curve Weft Mean WT: tensile energy Warp Weft Mean RT: tensile resilience Warp Weft Mean EMT: tensile strain Warp Weft Mean

± ± ± g.cm/cm2 g.cm/cm2 g.cm/cm2 % % % % % %

0.667 0.569 0.618 15.65 29.20 22.42 34.76 32.01 33.38 18.76 41.08 29.92

0.625 0.548 0.586 17.40 29.22 23.31 36.83 37.03 36.93 22.27 42.65 32.46

0.650 0.615 0.575 0.555 0.612 0.585 19.35 21.72 31.80 35.77 25.57 28.74 33.18 35.52 29.96 34.96 31.57 35.24 23.83 28.25 44.27 51.61 34.05 39.93

0.584 0.537 0.561 19.15 34.20 26.67 29.48 25.79 27.63 26.22 50.91 38.57

0.551 0.519 0.535 22.27 46.38 34.55 30.81 25.56 28.18 32.36 72.12 52.24

Notes:

a

Number indicates the nominal yarn in tex; F = two-fold; S = Sirospun

conductivity is measured), the specific volume of the fabrics was calculated. The specific volume of the fabrics knitted from the Sirospun yarns is less than that of the fabrics knitted form the two-fold yarns. As less air is trapped in the fabrics knitted from the Sirospun yarns, the thermal conductivity is better. As the yarns become finer, the hand-feel of the fabrics knitted from the twofold yarns is warmer, and is warmer than that of the fabrics knitted from the Sirospun yarns especially with yarns of 18.5tex. The hand-feel of the fabrics knitted from the Sirospun yarns is cooler because of its greater thermal conductivity. Therefore, in comparing with the fabrics knitted from the two-fold yarns, the fabrics knitted from the Sirospun yarns are more durable and serviceable, with acceptable hand-feel and more suitable for summer wear because of cooler feeling and better thermal conductivity. References Cheng, K.P.S. and Sun, M.N. (1998), ``Effect of strand spacing and twist multiplier on cotton Sirospun yarn'', Textile Research Journal, Vol. 68 No. 7, pp. 520-7. Helw, E.M., Hawary, I. and Okeily, M. (1988), ``Weft knitted fabrics from siro and ring-spun yarns'', Indian Textile Journal, August, pp. 134-42. Vijayakumar, H.L., Muralidhara, J.S., Swamy, J.S.S., Benal, V.S. and Sirnath, K.V. (1991), ``Weft knitted fabrics from siro and ring spun hoisery yarn'', Indian Textile Journal, January, Vol. 101 No. 4, pp. 62-4.

Fabric knitted from cotton Sirospun1 yarn 359

Table VI. KES tensile test results

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

Recommend Documents