You are on page 1of 10

s92-3,1

Characterization and Control of Fabric Properties in Textile and Apparel
Manufacturing
PIs: P. Banks-Lee, N. C. State; D. R. Buchanan, N. C. State; T. G. Clapp, N. C. State; J.W.
Eischen, N. C. State; T. K. Ghosh, N. C. State; B. S. Gupta, N. C. State; T. J. Little, N. C. State;
A. Seyam, N. C. State.
Post-Doctorate Research Associates: Dr. M. Gunner(EE), Dr. F. Sun

Graduate Students: G. Barrett(EE), S. Deng(ME), Y. G. Kim(ME), D. Kong(TAM), H.
Li(TAM), T. McDevitt(ME), N.Timble(FPS)

Team Leader: T. G. Clapp
RELEVANCE TO NTC GOALS:
The apparel segment of the U.S. textile industrial complex is under tremendous pressure
from foreign competition. The apparel industry is trying to respond to this pressure by changing
its production philosophy to accommodate the consumer demand for ever increasing style
variation and retailer demands for more frequent Just-In-Time (JIT) deliveries of smaller orders.
The apparel industry must be able to respond to these demands to remain competitive in the U.S.
and abroad.
Apparel manufacturing in a Quick Response (QR) environment must have the ability to
quickly optimize its processes when faced with rapidly changing fabrics with variable material
properties. This variability is inherent (statistical) within a given fabric and between different
types of fabric. In order to deal successfully with this challenge, it is necessary to understand
how important material properties permeate through the manufacturing chain. We must develop
technologies to measure these important properties quickly and reliably (not dependent on
traditional laboratory techniques and constraints). It is important to lay the groundwork for
dealing with property variability in the design of apparel fabrics as well as in the manufacturing
and finishing of these products.
An interdisciplinary team of researchers from Textile Engineering (TE), Electrical
Engineering (EE), Mechanical Engineering (ME), and Textile Technology (TT) are conducting
research to collectively address the following research themes: 1) model fabric behavior, 2)
on-line characterization of material properties, and 3) on-line manufacturing control. New
developments in computer modeling and sensor technology are being applied to address these
themes.

GOALS:
LONG TERM: Develop new technologies and methods to characterize fabric properties on-line
(during processing) and control textile and apparel manufacturing processes to optimize quality
and productivity in a Quick Response or Just-in-Time manufacturing environment. Educate
graduate and undergraduate students to help strengthen and lead U.S. fiber, textile, and apparel
companies into the 21st century.
SHORT TERM: Using common fabrics a basis, 1) validate 3-D fabric deformation models, 2)
refine and validate a theoretical model for predicting visco-elastic behavior of fabrics in roll
form, 3) develop and validate a theoretical model for measuring fabric stiffness on-line, 4)
develop sensor technology to characterize fabric on-line during high speed sewing, and 5)
quantify fabric frictional behavior and mechanical properties as a function of orientation.
Involve at least three undergraduates in research activities, and graduate at least two Ph.D. and
two Master’s students.

National Textile Center Annual Report: September, 2994 219
s92-3,2

TECHNICAL QUALITY AND ACCOMPLISHMENTS
An interdisciplinary team of researchers from Textile Engineering, Electrical
Engineering, Mechanical Engineering, and Textile Technology have collectively addressed the
following research themes: 1) model fabric behavior, 2) on-line characterization of material
properties, and 3) manufacturing control. The depth of the research is reflected by theses
submitted. These are listed below. A brief summary of each of the five research tasks is also
provided.

Thesis Completed and in Preparation
1. Kim, Y. G., “Fabric Manipulation Simulation Including Material Non-linearity
and Contact”, PhD Dissertation, Completed June 10, 1993.
2. McDevitt, T., ‘Flexible Fabric Mechanics analysis Using Large Deflection
Beam Theory”, MS Thesis, completed August 1993.
3. Kong, D., “Development of a Control System for Fabric Roll-Making
Operation”, MS Thesis August 1993.
4. Timble, N., ” Structural Factors Affecting Inter-facial Forces Between Fabrics,
PhD Thesis”, submission in Final Form expected by October 25, 1993.
5. Li, H., “Simulated Roll-Like Loading in Laboratory and it’s effects on Fabric
Properties” complete.
6. Deng, S., “Nonlinear Fabric Mechanics Including Material Nonlinearity,
Contact, and Adaptive Global Statics Algorithm,” PhD Dissertation, Final
Defense scheduled August 22, 1994.
7. Barrett, G. R., “On-line Fabric Identification and Adaptive Control of the
Sewing System for Improved Apparel Assembly.”
8. Farrington, C. E., “Deformational Characteristics of Plain Knitted Yarn Loops
Under Shear Loading,” PhD Dissertation.

220 National Textile Center Annual Report: September, 1994
s92-3.3

RESEARCH SUMMARIES:

Deformational Characteristics of Plain Knitted Yarn Loops Under Shear
Loading
Developing an automatic feeding or sewing mechanism requires an understanding of the
behavior of knitted fabrics under stress. A fabric which will be fed by an automatic feeding
mechanism or will be sewn rarely experiences symmetrical loading. To date, however, most
analyses made determine the structural behavior of knitted fabrics under no load or are
symmetrically loaded. Therefore, for practical applications, it is important to examine the
behavior of knitted fabrics under nonsymmetric loading conditions.
S. de Jong and R. Postle, authors of “A General Energy Analysis of Fabric Mechanics
Using Optimal Control Theory”, used an energy analysis based on Optimal Control Theory
(OCT) to determine the key deformational characteristics of symmetrically loaded knitted loops.
This approach assumes that yams in a knitted loop will take a shape such that their internal
energy is minimized while maintaining the force/couple equilibrium. By modifying this model
it is possible to analyze how a knitted fabric deforms under an uniform shearing load.
The first step in this project was to reproduce de Jong’s work. A computer program,
written in C, was developed to determine the deformational behavior of a knitted quarter loop
that is symmetrically loaded. Certain algorithms were developed to reduce run time and increase
solution stability. This computer program will be the basis for any future research involving the
nonsymmetric loading of knitted fabrics.
The second step was to modify the OCT algorithm to analyze the deformational behavior
of knitted loops under nonsymmetrical loading. This involved the expansion of the analysis
from a quarter loop to a half loop. In addition, two subroutines were developed to solve the two
point boundary problem, which arises from the expansion of the analysis to a half loop.
A review of the modified OCT analysis of plain knitted fabrics under a shearing load
confirms the anisotropic behavior of knitted fabrics. A loop subject to counter-clockwise
uniform shear loading will skew in the counter-clockwise direction. To counteract this moment,
the force exchange between interlocking yarns on the left side of the loop becomes higher than
the force exchange between interlocking yarns on the right side of the loop. As the shearing
load increases: 1) the loop becomes more skewed, 2) the tightness of the left side of the loop
increases due to a decrease in yarn curvature; and 3) the tightness in the right side of the loop
decreases as the yarn segment begins to straighten. Sewing or feeding mechanisms add this type
of shearing load to a fabric and therefore require further study.

National Textile Center Annual Report: September, 1994 221
S92-3.4

W

A) Symmetrically Loaded Knitted Loop. B) Nonsymmetrically Loaded Knitted Loop

On-line Fabric Identification and Adaptive Control of the Sewing System
for Improved Apparel Assembly
Analysis of current sewing techniques indicated that higher quality apparel assembly
required research into the fundamental aspects of sewing control. From this analysis, the two
courses of research to be followed were 1) the determination of methods for characterizing the
dynamics of high speed sewing machine and fabric interactions, and 2) the determination of
methods of adaptively controlling the sewing system based on the on-line evaluation of the
fabric/machine interactions.
Needle bar forces occurring during high speed sewing can be used to identify fabrics
on-line. A neural network classifier for on-line identification of fabrics has been created. The
neural network acquires its data from a sewing machine equipped with force transducers. These
transducers measure the forces occurring at the needle during the sewing process. The neural
network is trained to identify fabrics based on their force “fingerprint” (See Figs. 1 and 2). By
implementing the neural network on a microprocessor, a practical way exist to identify not only
the fabric type but also the number of plies sewn. Utilizing this type of classifier, the automated
apparel assembly station can adapt itself to changing sewing conditions and provide quality
checks such as making sure all plies were sewn.
The second course of research pursued was the determination of methods of adaptively
controlling the sewing system based on the on-line evaluation of the fabric/machine interactions.
The fundamental knowledge gained from these investigations has led to a full report on a presser
foot force actuator and patent disclosure to the University. This control method stabilizes the
forces applied to a fabric during feeding. By maintaining an optimai applied force, fabric may ‘be
less likely to pucker during sewing. A prototype to demonstrate the fundamental operation of
this device is currently being designed and built. The digital controller design method uses an
ellipsoid algorithm to optimize controller parameters. Time domain specifications are

222 National Textile Center Annual Report: September, 1994
s92-3,5

implemented as linear inequality constraints within the ellipsoid algorithm (See Figs. 3 and 4).
The ellipsoid algorithm then searches for controller parameters that optimize some (usually
non-linear) objective function subject to the constraints. In this manner, the controlled presser
foot behaves with certain specified time responses.
1.0

0.5

0.0

.

Figure 1. T&ical needle force waveforms for one stitch for five fabrics

Figure 2. Neural network misclassifications of fabric type as a function of training time

National Textile Center Annual Report: September, 1994 223
S92-3,6

Output and t3fott Cmstraints

0 10 20 TimZitep 40 50 60

Figure 3. A unit step respon.se of presser km1 force thal
satisfies the time domain constraints

6-

4-

2-

o-

ci
- 2 -

-4-

-6-

-8’ 1 1 1 I
-6 -4 -2 0 2 4 6 8
Xl
Figure 10. Convergtmcz of a two dimensional parameter *search
to an optimal feasible point using the ellipsoid algorithm

224 National Textile Center Annual Report: September, 1994
s92-3.7

Develop Models for Predicting Fabric Friction and Directional Variation in Fabric
Mechanical and Viscwelastic Properties
Fabric Friction
The friction results obtained on fabrics varying broadly in structure and fiber material,
and reported in previous reports, were rationalized using a modified version of Wilson’s model
which included fabric compressional coefficient, L, estimated experimentally, as a measure of
hardness, and apparent area of contact, Aa, estimated theoretically by Pierce’s model, as a
measure of the number of asperities of contact. Using these modifications, the foal models for
the constants C and n which characterize frictional behavior of fabrics by the equation (P/A) = C
(N/A)” can be given as follows.
C = kSL-7 Aal-Y (1)
n=B(l-y)+ y (2)
In these equations, S is the specific shear strength of the adhesion junctions, yis the factor
related to the shape of the general pressure(P)-area(A) curve given by the relationship P=KAa
where a = p1 - l), and k and B are model constants.
Using these models (Equations 1 and 2) the effects of fabric structure parameters, picks
per inch and yarn twist, on frictional behavior could be rationalized effectively. An example of
the results which give a comparison between the experimentally determined and the theoretically
predicted values of C(Equation 1) is shown in Figure 1. Currently, a paper is being written for
submission to Textile Research Journal for publication on this work.

-0.5 - y = 0.27 + 1.17x
R-square = 0.71, R =0.84
.
-0.6 -

Figure 1. Theoretical values of friction parameter, C , versus experimental values of
friction parameter, C, for all the filament polyester model fabrics. (y = 0.90)
Fabric Mechanical and Viscoelastic Prouerties
The fabrics must endure stresses and strains of variety of levels during conversion into final
products as well as during end use, specially industrial. The sewability, seam quality, style and drape
of garments and performance in which fabrics are subjected to multidirectional forces are
additionally affected by the variation in mechanical and visco-elastic properties of fabrics. The
information available from such a study will allow us to understand and predict (1) the behavior of
a fabric when subjected to pull from different directions, and (2) style and drape, and performance
in manufacturing of fabrics. The focus of activity in this area has been on measuring the variation
in mechanical properties with direction and modelling the behavior using theoretical/empirical
approaches.

National Textile Center Annual Report: September, 1994 225
S92-3.8

The tensile properties of five different fabrics, varying broadly in construction and yam
material (Table l), were measured using ASTM method D5035 (Strip test method). The properties
measured were in seven different directions ( O”, 15’, 30°, 45’, 60”, 75’, 90” >from the warp
direction).
The results obtained clearly show that the properties vary greatly with the direction of testing
and the structure of the fabric. An example of results is given in Figure 1. Currently, the results
obtained are being analyzed and modelled using a finite-deformation theory. Also, the impact the
2.:: variation in mechanical properties has on fabric sewability and seam quality is being
.
Table 1. Fabric Specifications
FABRIC Fabric Counts Yarn Number (Denier) Weave Fiber
EPI PPI Warp filling
A 66 95 150 150 2f2hvill polyester
B 109 60 70 150 plain polyester
C 58 50 200 200 plain nylon
D 212 56 70 150 satin polyester
E 80 69 2/47* 1/34* 2/2 twill wool
* worsted count

70
I

---c- IabncA
- fatmcB
- fatxlCC
- mncD
- labricE

ot0 20
I , j , I
40
, ,, (,.%I

60 60 100
Angle from warp direction (degree)

Figure 1. Stress Variation with Change of Angle

226 National Textile Center Annual Reporf: September, 1994
S92-3,9

Computational/Experimental Determination of Fabric Drape and Motion

T, McDevitt- Flexible Fabric Mechanics Analvsis Using Large Defection Beam
Theorv

The primary objective of this research was to develop a computer method to simulate the
quasi-static motion of fabric parts during certain manufacturing processes in the apparel
industry. Fabric parts are modeled as very flexible elastic beams that can accommodate
stretching and bending in a single plane. Again, the equilibrium equations are solved using the
finite element method. In solving for drape shapes and fabric configurations during
manipulation, numerical problems are encountered when the fabric begins to buckle or wrinkle.
In this project, a technique know as arc-length control has been implemented to treat these
difficulties, and be able to simulate very complex load deflection paths. Attention has also been
focused on generating finite element meshes that adapt to the severity of the deformation.
Generally speaking, a numerical method requires more refinement (nodes) in the areas where the
state variables (displacements) are varying the most rapidly. A method to distribute the finite
element nodes to those regions of the fabric undergoing the most curvature has implemented.

S. Deng- 3D Fabric Drape Mechanics

This project is an ongoing extension of our capability to model three dimensional fabric
drape over complicated surfaces. The use of large displacement shell theory to model the fabrics
behavior has proved to be very successful. A very interesting aspect of this work is the
discovery of multiple numerical solutions for seemingly simple fabric shapes. The solutions of
the highly nonlinear equilibrium equations is non-unique. The challenge is to determine which
of the several solutions is the physical realizable one, or whether multiple physical solutions are
possible. Figure 1 shows the drape of a fabric part over a pair of intersecting cylinders. This
problem demonstrates that the method can treat general part shapes and complex contact
surfaces, We are also able to simulate the drape of fabric parts that show a nonlinear moment
curvature relationship.

(b) Linear Material Solution
(a) Experiment Result

Figure 1. Comparison of Numerical Solutions and Experimental Result for Fabric Draping Over
Intersecting Cylinders

National Textile Center Annual Report: September, 1994 227
s92-3.10

Develop an “Intelligent” Closed Loop System to Control the Roll Making Process
A feed-back control system for fabric roll-making has been developed. The control
system, in principle, would allow production of fabric rolls with optimal levels of internal
stresses. The system is also designed to minimize variation of stresses.
Traditionally, fabric rolls are produced without any particular concern for the winding
tension or roll-size. Their levels are mainly dictated by the unit process or the experiential
knowledge of the operator. Generally, the winding tension remains somewhat constant during
any process. It has been demonstrated in this research that in case of constant tension winding
the fabric inside the roll may actually buckle and may cause permanent deformation of the
fabric. This is particularly undesirable for easily deformable or delicate fabrics such as certain
types of fine woolen or knit fabrics. To demonstrate the levels of stresses developed in fabric
rolls and their variation strain measurement systems for both compressive and tensile strains
have been developed. A miniature compression load cell has been used to, measure the radial
compressive stresses within a roll. The result shows initial rapid increase in compressive stress
near the core followed be a levelling off near the top. This suggests that after a certain number
of layers on the core the hoop stresses are not transmitted to the core from further layers. Maybe
the intermediate layers together act as a non-deformable rigid cylinder. To measure the in-plane
stresses foil-type strain gages have been used. It was necessary to develop special mounting
techniques to obtain any meaningful measurement of strains in fabric layers. The results show
that in some layers the fabric is actually under in-plane compressive stresses, even though, the
web tension in the fabric at the time of winding was positive. These results substantiate earlier
results of theoretical analyses during this project.

1,6: Electric Cylinders 2,4: Guide Cylinders 3: Moving Cylinders
5: Tension Measurement Cylinders

Figure 1. Tension Control System for Roll-making
In the control system developed in this project the winding tension is used as the control
parameter Since our last report the control system has been modified to eliminate noise
vibrations and to improve stability of the control. Figure 1 shows the improved control system.
In this, the vertical position of a control cylinder is precisely controlled to generate various
levels of wrap angle which in turn determines the outgoing tension in the fabric. The measured
tension in the fabric web is used to control the position of the control cylinder. The control
cylinder is driven by two electric cylinders working in unison. A new and improved software
for the control system has been developed by using a graphical programming language
LabView2. Using this program the winding tension can be controlled as function of time.

228 National Textile Center Annual Report: September, 1994