Intelligent Evaluation of Fabrics' Elastic Properties from Simulated

0 downloads 0 Views 2MB Size Report
the mechanical properties of fabric will be identified by ... Identification, Fabric Drape, Finite Element Method. 1. ... Virtual reality in the field of garments animation.
ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

Intelligent Evaluation of Fabrics’ Elastic Properties from Simulated Drape Test

Hassen Hedfi University of Monastir, National Engineering School, Mechanical Engineering Laboratory Ibn Aljazzar Street Monastir 5019, Tunisia [email protected] Adel Ghith University of Monastir, National Engineering School, Thermal and Energetic Systems Studies Laboratory Ibn Aljazzar Street Monastir 5019, Tunisia [email protected] Hédi Bel Hadj Salah University of Monastir, National Engineering School, Mechanical Engineering Laboratory Ibn Aljazzar Street Monastir 5019, Tunisia [email protected]

Abstract

1. Introduction

In this paper we propose an intelligent method for identifying several mechanical parameters of woven fabric. The mechanical fabric characterization requires several expensive experimental tests. An alternative to overcome this problem is to use an inverse identification based on simulation of drape test using the finite element method combined with a learning artificial neural networks. To do this, a database containing a number of fabrics characterized is developed. This database will be used to train the neural network to predict the mechanical properties of textile material from drape’s properties. In this way, the mechanical properties of fabric will be identified by simulating the drape test. By comparing drape properties obtained numerically with those obtained experimentally; the elastic properties will be adjusted in the numerical model. The process is repeated until there is little difference between the simulated and actual drape. This approach is tested on several cases and the results are convincing for a wide range of fabrics. Keywords: Artificial Neural Network, Inverse Identification, Fabric Drape, Finite Element Method

The use of intelligent methods in engineering problems is increasingly important [1]. These methods are known as: Artificial Intelligence (AI) and include: Fuzzy Logic (FL), Artificial Neural Networks (ANN) and Genetic Algorithms (GA). In textile engineering, these methods are increasingly used for many applications [2-5]. Indeed, several researches are invested in this area by implementing, in particular, artificial neural networks and methods based on fuzzy logic. These works can be categorized into:  Decision Support Tools for textile industry,  Prediction and Identification of the textile materials behavior,  Industrial manufacturing processes optimization,  Virtual reality in the field of garments animation and prototyping, and  Fashion trends and mass consumer behavior Hui et al. [6] presented an extended review of application of artificial neural networks in textiles and clothing Industries over last decades. Those applications relate: fiber classification [7-8], yarn manufacture [9-10], yarn-property prediction [11-15] and the relationship between fiber and yarn [16]. For

IJCTA | JULY-AUGUST 2011 Available [email protected]

741

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

applications related to fabric, the authors cited the use of ANN in: fabric manufacture [17-19], fabric-property prediction [20-23], fabric defect inspection [24-26]. Other textile fields, as for examples: pattern fitting prediction [27], clothing sensory comfort [28-29], and textile chemical processing [30-31] benefited also, from the use of ANN. Hui et al., [31] investigated the use of ANN to predict the sewing performance of woven fabrics for efficient planning and control for the sewing operation. Hui and Ng [32] investigated the capacity of back propagation ANN based on an algorithm with weight decay technique and multiple logarithm regression (MLR) methods for modeling seam performance of fifty commercial woven fabrics based on seam puckering, seam flotation and seam efficiency. In this section we focus on some research work employing ANN in fabric mechanical behavior and drape predictions. Chen et al. [33] proposed a neural network computing technique to predict fabric end-use. In this work, instrumental data of the fabric properties and information on fabric end-uses were used as input into neural network software to train a multilayer perceptron model. The prediction error rate from the recognized neural network model was estimated by using a crossvalidation method. Stylios et al. [34-35] modeled the relationship between measured drape attributes and the subjective evaluation of the fabric drape for many enduses on a neural network using back propagation, which can appropriately predict the drape grade of 90 % of the samples. The connections between drape attributes and fabric bending, shear and weight was also modeled using neural networks. The authors found that using the natural logarithm of the material property divided first by the weight of the fabric produced the most predictive model. Lam et al. [37] used neural networks to predict the drape coefficient (DC) and circularity (CIR) of many different kinds of fabrics. The neural network models used were the Multilayer Perceptron using Back propagation (BP) and the Radial Basis Function (RBF) neural network. The authors found that the BP method was more effective than the RBF method but the RBF method was the fastest when it came to training. Comparisons of the two models as well as comparisons of the same models using different parameters are presented. They also found that prediction for CIR was less accurate than for DC for both neural network architectures. Koustoumpardis et al. [38] propose an approach to intelligent evaluation of the tensile test. A robotized system is used that performs the fabrics tensile test and estimates the extensibility of the samples using a feedforward neural network while trying to imitate the human expert estimation.

IJCTA | JULY-AUGUST 2011 Available [email protected]

Hadizadeh et al. [39] introduced an Adaptive NeuroFuzzy Inference System (ANFIS) for predicting initial load-extension behavior of plain-woven fabrics. Input values defined as combination expressions of geometrical parameters of fabric and yarn flexural rigidity. The results show that the neuro-fuzzy system can be used for modeling initial modulus in the warp and weft directions of plain-woven fabrics. Pattanayak et al. [40] proposed a method based on an artificial neural network and multiple regression method for drape profile prediction. In this study, the relationship between the fabric drape parameters (such as drape coefficient, drape distance ratio, fold depth index, amplitude and number of nodes) and low stress mechanical properties was investigated. The authors find that, although both methods are useful for the prediction of the drape parameters, the neural network model proved to be the best between these two. Jedda et al. [41] investigated the relationship between the fabric drape coefficient (DC) and mechanical properties tested on the fabric assurance by simple testing system (FAST). The authors proposed three regression models for each type of pattern and for all fabrics using the multiple linear regressions. A neural model was proposed using the neural networks and it was compared to the regression one. It was be found that ANN prediction is more accuracy then regression models. In this paper, we adopt an inverse approach using simple experimental measurements such as measuring of drape, thickness and mass per unit area of textile fabrics for the prediction of mechanical properties of these fabrics. Indeed, upon learning of neural networks and numerical calculation using finite element method, we identify the mechanical properties of textile fabrics without the need to measure them experimentally. Those mechanical measures are costly in terms of time, mobilization of equipment and technicians.

2. Fabric characterization Characterization of textile fabrics includes the determination of physical, mechanical and aesthetic properties.

2.1. Woven fabric structure A textile fabric is obtained by interlacing two orthogonal sets of yarn called warp and weft. The manner in which the wires are intercrossed is called pattern weave. The two sets of wire define two orthotropic directions in the woven structure: warp and weft directions. The number of warp wire per centimeter of the textile fabric is called the warp count. Similarly, we define the weft count as the number of weft wire per centimeter. The two sets of yarn are also characterized by their linear densities expressed in

742

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

terms of metric Number (Nm). Figure 1 shows a schematic example of plain weave. These design parameters influence fabrics’ mechanical, physical and aesthetic properties.

2.2. Physical properties The physical properties studied in this paper are Fabric Surface Density (FSD) and Fabric Thickness (FT). (Table 1)

2.3. Mechanical properties The mechanical properties studied are Tensile rigidities in warp, weft and bias directions (TR1, TR2, TRB), Poisson’s ratio in warp and weft directions 5PR1, PR2), and Bending rigidities in warp and weft directions (BR1, BR2). (Table 2)

2

Drape Distance Ratio, DDR = (%)

ru 2 − r × 100 ru 2 − rs 2

(2)

rmax − rmin × 100 ru − rs

Fold Depth Index, FDI = (%)

(3)

rmax − rmin 2

Amplitude to Radius, AR =

(4)

Where ru : The radius of undraped fabric, rs the radius of the fabric supporting-disc, rm the average of 16 measured rays between disc centre and projected profile: rm =

1 16 ∑ ri 16 i =1

(5)

= rmax max = ( ri ) , rmin min ( ri )

Table 2: Fabric mechanical parameters Warp Weft Bias tensile tensile tensile Parameters rigidity rigidity rigidity

Figure 1: Fabric plain weave Table 1: Fabric Physical Parameters Fabric Surface Parameters Fabric Thickness Density

Symbol

TR1

TR2

TRB

Unit

MPa

MPa

MPa

Range

0.6599.36

2.5843.25

1.6936.75

Mean

10.60

9.04

17.20

5.25

9.02

Warp Bending rigidity

Weft Bending rigidity

Standard 8.34 deviation (Continued-Table 2) Weft Warp Poisson’s Poisson’s ratio ratio

Symbol

FSD

FT

PR1

PR2

BR1

BR2

Unit

g m-2

mm

Unless

Unless

µNm

µNm

Range

62.61420.34

0.19 1.41

0.090.83

0.13 0.51

1.57160.7

1.989.4

Mean

178.33

0.75

0.29

0.20

22.45

20.3

Standard deviation

65.74

0.35

0.16

0.05

30.54

16.1

3. Fabric drape’s test simulation 3.1. Evaluation of drape attributes The drape attributes are measured using drape-meter (Figure 2-a) according to French standard NF-G-07-109 (Table 3). A schematic representation of the projection of a fabric sample of diameter 250 mm draped over support-disc of diameter 150 mm is given in Figure 2-b. The drape attributes are node number (NN), and: Drape Coefficient, DC = (%)

IJCTA | JULY-AUGUST 2011 Available [email protected]

rm 2 − rs 2 × 100 ru 2 − rs 2

(1)

Table 3: Fabric Drape Attributes Parameter

DC

unit

(%)

Range

23.4390.01

Mean

65.82

NN

210 4

DDR

FDI

ARR

(%)

(%)

mm

1.74106

2.06114.8

0.112.37

16.62

17.49

0.40

743

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

Coefficient of Variation (%)

14.91

1.35

20.02

21.84

0.43

This model is used to simulate the drape-meter test and allows the determination of drape attributes numerically.

3.2. Fabric Model The fabric model used in this paper is based on the general deformable models formulated, resolved using finite element method and validated by Hedfi et al. [42]. In this model, fabric is regarded as continue medium interacting dynamically with environment solid object or fluid flow (wind, water). This dynamic model can be written as follows:    ∂ 2 r  int ∂r ρ 2 +F = ρ⋅g +µ ∂t ∂t

(6)

Where, ρ is the fabric  damping coefficient, r point P, t time ( t ≥ 0 ) ,

surface density (FSD), µ the the instantaneous position of a  g the gravity acceleration, and

 int F : forces due to internal deformation:   2 2  int ∂  ∂r  ∂2  ∂2 r F = −∑  Cαβ ⋅  Sαβ ⋅ + ∑ ∂aα  ∂aα  α , β 1 ∂aα ∂aβ  ∂aα ∂aβ = α , β 1= s Sαβ =wαβ ( gαβt − gαβ0 ) , Cαβ =wαβb ( bαβt − bαβ0 )

  (7) 

(8)

In this model, mechanical properties are introduced via two matrixes:  E1 G   Rf1 b = ( wαβs )1≤α , β ≤ 2 =  , ( wαβ )1≤α , β ≤ 2  0 G E 2  

(g (g (b (b

t

) ) ) )

αβ 1≤α , β ≤ 2 0

αβ 1≤α , β ≤ 2

t

αβ 1≤α , β ≤ 2 0

αβ 1≤α , β ≤ 2

0  Rf 2 

(9)

: The metric tensor at time t : The metric tensor at time t = 0 : The curvature tensor at time t : The metric tensor at time t = 0

Shear rigidity is computed according to following formula: −1

 4 1 − υ2 1 − υ1  (10) − − G =  E1 E2   E45 Where E1 : Tensile rigidity in warp directions (TR1) E2 : Tensile rigidity in weft direction (TR2) E45 : Tensile rigidity in bias direction (TRB)

ν 1 : Poisson’s ratio in warp direction (PR1) ν 2 : Poisson’s ratio in weft direction (PR2)

Figure 2: Measurement of fabric drape attributes: (a) circular fabric sample draped over the drape-meter, and (b) schematic representation of different rays used to calculate the drape attributes

4. ANN-Learning Process 4.1. Data preparations The experimental database used in this study is formed by various textile fabrics. The inputs are: drape coefficient (DC), node number (NN), drape distance ratio (DDR), fabric depth index (FDI), amplitude to radius ratio (ARR), fabric surface density (FSD) and fabric thickness (FT). The outputs are: warp, weft, and bias tensile rigidities (TR1, TR2, and TRB), warp and weft Poisson’s ratios (PR1 and PR2), warp and weft bending rigidities (BR1 and BR2). All these inputs and outputs are normalized before training step using following formula: tin =

ti − ti

σi

(11)

Where, tin the normalized values of parameter i, ti the measured values of parameter i, ti its average, and σ i its standard deviation. After normalization, each parameter has an average of 0 and a standard deviation equal to 1.

G : Shear rigidity (SR) Rf1 : Bending rigidity in warp direction (BR1) Rf 2 : Bending rigidity in weft direction (BR2)

IJCTA | JULY-AUGUST 2011 Available [email protected]

744

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

Figure 3: Neural Network architecture Experimental database is a divided randomly into three subsets: 1. Training subset containing 70% of samples used for gradient computing and for ANN weights and biases updating. 2. Validation subset containing 15% of samples 3.

Test subset containing 15% of samples, used for ANN generalization.

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

4.2 Optimization of ANN Architecture The parameters of the neural network to be optimized are:  the number of neurons Nn  the number of hidden layer HL  the number of iterations N i The neural network architecture is shown in Figure 3. This network is a feed-forward ANN trained with error back-propagation algorithm. The ANN weights and biases updating is carried out using the LevenbergMarquardt optimization algorithm. The ANN optimality criteria are: 1. The correlation coefficient (R) between the predicted and measured values for each output. 2. The mean squares error (mse): = mse

1 N ∑ ( ti − pi ) N i =1

2

(12)

Where: N is the number of samples, ti the target value, pi and the predicted value. The optimization of neurons number is done in an incremental way using following algorithm.

IJCTA | JULY-AUGUST 2011 Available [email protected]

7.

set N i = 50 set HL = 1 set Nn = 1 execute the ANN’s Model compute mse( Nn) repeat a. Nn ← Nn + k b. execute the ANN’s Model c. compute mse( Nn) Until mse( Nn + k ) ≥ mse( Nn) Nopt = Nn − k

The optimal neurons number is the Nopt = Nn − k . In this study, we find Nopt = 35 (Figure 4) The number of hidden layers optimization is also done in an incremental way using following algorithm. 1. set N i = 50 2. set HL = 1 3. set Nn = 35 4. execute the ANN’s Model 5. compute mse( HL) 6. repeat a. HL ← HL + 1 b. execute the ANN’s Model c. compute mse( HL) Until mse( HL + 1) ≥ mse( HL) 7. HLopt = HL − 1

745

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

The optimal hidden layers number is HLopt = HL − 1 . In this study, we find HLopt = 2 (Figure 5) The overfitting problem is treated using early stopping method. In fact, when training process is run, the ANN's performance for validation subset increases until a limit from which these performance are decreasing. The training process should be stopped before the error on the validation set is becoming increasingly important. So that, the optimal number of iterations is found automatically using this method.

Figures 6 and 8 show an example of results obtained

for BR1. 2. A less capacity for the prediction of elastic properties: tensile rigidities in warp and weft directions (TR1 and TR2) and Poisson’s Ratio (PR1 and PR2) Figures 7 and 9 show an example of results obtained for TR1.

Figure 6: Performances obtained for warp bending rigidity (BR1) learning using optimized ANN (Nn=35, HL=2, Ni=12) Figure 4: Optimizing number of neurons

Figure 5: Optimizing number of hidden layers

5. Results & Discussions 5.1. Predictability of mechanical properties By testing the capacity of the optimized neural network for the prediction of mechanical properties from drape attributes we found: 1. A great capacity to predict the bending properties (BR1 and BR2) and bias tensile rigidity (TRB).

IJCTA | JULY-AUGUST 2011 Available [email protected]

Figure 7: Performances obtained for warp tensile rigidity (TR1) learning using optimized ANN (Nn=35, HL=2, Ni=12) We attribute these differences to sensitivity of drape parameters to the flexibility of textile fabrics and their shear behaviour which is strongly correlated with bias tensile rigidity (TRB). To resolve this problem, we decided to increase the size of the training base, but this requires new experimental investigations. The idea is then the use of the finite

746

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

element model to simulate the drape test. In this way, we can expand the learning base by simulated data.

5.2. ANN-FEM Coupling ANN-FEM coupling is performed using the algorithm shown in Figure 10. This algorithm allows the identification of mechanical parameters of fabric from drape attributes and improves the predictability of the parameters related to the stretch behaviour: TR1, TR2, PR1, and PR2. The algorithm is executed until the error between measured drape attributes and those found by the simulation becomes less than a limit value ε = 10−2 .

= Error ( DA )

Predicted DA-Actual DA × 100 Actual DA

(13)

Where, DA=DC, NN, DDR, ARR or FDI

Figure 8: Regressions between warp bending rigidity (BR1) and drape attributes using optimized ANN (Nn=35, HL=2, Ni=12)

Figure 10: Algorithm for ANN-FEM coupling Table 4 shows an example of results obtained on a

Figure 9: Regressions between warp tensile rigidity (TR1) and drape attributes using optimized ANN (Nn=35, HL=2, Ni=12)

IJCTA | JULY-AUGUST 2011 Available [email protected]

plain fabric. The coefficient Error-MP represents the error between the actual values of mechanical properties and those predicted by the neural network without and with ANN-FEM coupling. R denotes the correlation coefficient between the values predicted by the ANN on the test set. This example shows the ability of this approach for the identification of mechanical properties of textile fabrics using the physical parameters and attributes of the drape. Figure 11 shows the results of applying our method to improve the predictability of the parameters TR1, TR2, PR1 and PR2 by coupling ANN-FEM models. The number of iterations varies from one parameter to another. Table 4: Results of ANN-FEM Coupling

747

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

ANN-FEM Coupling Fabric parameters Warp Tensile Rigidity (TR1) Weft Tensile Rigidity (TR2) Bias Tensile Rigidity (TRB) Warp Poisson Ratio (PR1) Weft Poisson Ratio (PR2) Warp Bending Rigidity (BR1) Weft Bending Rigidity (BR2)

Without FEM Simulation

With FEM Simulation

ErrorMP (%)

R

ErrorMP (%)

R

10.26

0.622

1.2

0.887

13.04

0.527

3.6

0.857

0.001

0.998

0.001

0.998

10.04

0.687

1.4

0.874

14.04

0.599

3.67

0.847

-5

0.998

10

-5

0.998

10-5

0.996

10-5

0.996

10

Figure 11: Variation of Error vs. Number of iterations This example shows that the parameters of drape are strongly related to bending and stretch behaviour in bias direction.

6. Conclusion In this paper, an inverse approach is introduced for the identification of mechanical properties of textile fabrics. This investigation combines the prediction by artificial neural networks and simulation by the finite element method. For this purpose, the neural network is used for predicting stretch and bending properties of textile fabrics. It was found that the predictability of the coefficients of bending rigidities and bias tensile rigidity is very important. But, for the coefficients of tensile rigidities and Poisson's ratios in warp and weft directions, predictability is average or even low. Virtual drape meter is then invested and used to improve the predictability of these parameters. Indeed, a coupling between the ANN model and the virtual drape meter was used to predict drape attributes and

IJCTA | JULY-AUGUST 2011 Available [email protected]

enhance the learning database. This strategy was tested and showed a sizable efficiency for predicting the overall mechanical properties of textile fabrics from simple tests. The originality of this work could be displayed clearly in the following fact: several mechanical properties of textile fabrics could be accurately predicted without using mechanical tests usually expensive, tedious and quite complex. This work can be improved by adopting fuzzy rules to automatically determine the relevance of the coefficients predicted by the virtual drape meter and the ANN model and control their inclusion in the learning database.

7. References [1] M., R., G., Meireles, P., E., M., Almeida, M., G., Simoes, “A comprehensive review for industrial applicability of artificial neural networks”, IEEE Transactions on Industrial Electronics, 50 (3), 2003, 585-601 [2] R., Chattopadhyay & A., Guha, “Artificial Neural Networks: Applications to Textiles”, Textile Progress, 35 (1), 2004, 1-46. [3] R., Guruprasad & B., K., Behera, “Soft Computing in Textiles”, Indian Journal of Fibre & Textile Research, 35, 2010, 75-84. [4] Tehran, M-A., & Maleki, M., “Artificial Neural Network Prosperities in Textile Applications”, Chap. 2 in Artificial Neural Networks - Industrial and Control Engineering Applications, edited by K. Suzuki, 35-64. Rijeka, Croatia: InTech, 2011. [5] Vassiliadis, S., Rangoussi, M., Cay, A., & Provatidis, C., “Artificial Neural Networks and Their Applications in the Engineering of Fabrics”, Chap. 6 in Woven Fabric Engineering, edited by P., D. Dubrovski, 112-134. Rijeka, Croatia: Sciyo, 2010. [6] Hui, C. L., Fun, N. S., & Ip, C. (2011, “Review of Application of Artificial Neural Networks in Textiles and Clothing Industries over Last Decades”, Chap. 1 in Artificial Neural Networks - Industrial and Control Engineering Applications, edited by K. Suzuki, 3-34, Rijeka, Croatia: InTech, 2011. [7] T., J., Kang, & S., C., Kim, “Objective Evaluation of the Trash and Color of Raw Cotton by Image Processing and Neural Network”, Textile Research Journal, 72 (9), 776-782, 2002. [8] A., Durand, O., Devos, C., Ruckebusch, &, J., P, Huvenne, “Genetic algorithm optimisation combined with partial least squares regression and mutual information variable selection procedures in near-infrared quantitative analysis of cotton-viscose textiles”, Analytica Chimica Acta, 595 (1-2), 72-79, 2007. [9] R., Beltran, L., Wang, & X., Wang, “Predicting Worsted Spinning Performance with an Artificial Neural Network Model”, Textile Research Journal, 74 (9), 757-763, 2004. [10] A., Farooq, & C., Cherif, “Use of Artificial Neural Networks for Determining the Leveling Action Point at the Auto-leveling Draw Frame”, Textile Research Journal, 78 (6), 502-509, 2008.[11] C., F., Kuo, K., I., Hsiao, & Y., S., Wu, “Using Neural Network Theory to Predict the Properties of

748

ISSN:2229-6093 Hassen Hedfi et al, Int. J. Comp. Tech. Appl., Vol 2 (4), 741-749

Melt Spun Fibers”, Textile Research Journal, 74 (9), 840-843, 2004. [12] Y., C., Zeng, K., F., Wang, & C., W., Yu, “Predicting the Tensile Properties of Air-Jet Spun Yarns”, Textile Research Journal, 74(8), 689-694, 2004. [13] J., J., Lin, “Prediction of Yarn Shrinkage using Neural Nets”, Textile Research Journal, 77 (5), 336-342, 2007. [14] Z., Khan, A., E., Lim, L., Wang, X., Wang, & R., Beltran, “An Artificial Neural Network based Hairiness Prediction Model for Worsted Wool Yarns”, Textile Research Journal, 79 (8), 714-720, 2009. [15] P., G., Ünal, C., Arikan, N., Özdil, & C., Taskin, “The Effect of Fiber Properties on the Characteristics of Spliced Yarns: Part II: Prediction of Retained Spliced Diameter”, Textile Research Journal, 80 (17), 1751-1758, 2010. [16] L., S., Admuthe, & S., Apte, “Adaptive Neuro-fuzzy Inference System with Subtractive Clustering: A Model to Predict Fiber and Yarn Relationship”, Textile Research Journal, 80 (9), 841-846, 2010. [17] G., Yao, J., Guo, and Y., Zhou, “Predicting the Warp Breakage Rate in Weaving by Neural Network Techniques”, Textile Research Journal, 75 (3), 274-278, 2005. [18] B., K., Behera, & B., Karthikeyan, “Artificial Neural Network-embedded Expert System for the Design of Canopy Fabrics”, Journal of Industrial Textiles, 36 (2), 111-123, 2006. [19] B., K., Behera, & Y., Goyal, “Artificial Neural Network System for the Design of Airbag Fabrics”, Journal of Industrial Textiles, 39 (1), 45-55, 2009 [20] S., Ertugrul, & N., Ucar, “Predicting Bursting Strength of Cotton Plain Knitted Fabrics Using Intelligent Techniques”, Textile Research Journal, 70 (10), 845-851, 2000. [21] T., W., Shyr, S., S., Lai, & J., Y., Lin, “New Approaches to Establishing Translation Equations for the Total Hand Value of Fabric”, Textile Research Journal, 74 (6), 528-534, 2004. [22] B., K., Behera, & R., Mishra, “Artificial neural networkbased prediction of aesthetic and functional properties of worsted suiting fabrics”, International Journal of Clothing Science and Technology, 19 (5), 259-276, 2007 [23] C., M., Murrells, X., M., Tao, B., G., Xu, & K., P., Cheng, “An Artificial Neural Network Model for the Prediction of Spirality of Fully Relaxed Single Jersey Fabrics”, Textile Research Journal, 79 (3), 227-234, 2009. [24] R., G., Saeidi, M., Latifi, S., S., Najar, & A., G., Saeidi, “Computer Vision-Aided Fabric Inspection System for OnCircular Knitting Machine”, Textile Research Journal, 75 (6), 492-497, 2005. [25] E., Shady, Y., Gowayed, M., Abouiiana, S., Youssef, & C., Pastore, “Detection and Classification of Defects in Knitted Fabric Structures”, Textile Research Journal 76 (4), 295-300, 2006. [26] C., W., M., Yuen, W., K., Wong, S., Q., Qian, D., D., Fan, L., K., Chan, & E., H., K., Fung, “Fabric Stitching Inspection Using Segmented Window Technique and BP Neural Network”, Textile Research Journal, 79 (1), 24-35, 2009. [27] Z., H., Hu, Y., S., Ding, X., K., Yu, W., B., Zhang, & Q., Yan, “A Hybrid Neural Network and Immune Algorithm

IJCTA | JULY-AUGUST 2011 Available [email protected]

Approach for Fit Garment Design”, Textile Research Journal, 79 (14), 1319-1330, 2009. [28] A., S., W., Wong, Y., Li, P., K., W., Yeung, & P., W., H., Lee, “Neural Network Predictions of Human Psychological Perceptions of Clothing Sensory Comfort”, Textile Research Journal, 73 (1), 31-37, 2003. [29] A., S., W., Wong, Y., Li, & P., K., W., Yeung, “Predicting Clothing Sensory Comfort with Artificial Intelligence Hybrid Models”, Textile Research Journal, 74 (1), 13-19, 2004. [30] C., C., Huang, & I., C., Chen, “Neural-Fuzzy Classification for Fabric Defects”, Textile Research Journal, 71 (3), 220-224, 2001. [31] C.-F., J., Kuo, T.-L., Su, & Y.-J., Huang, “Computerized Color Separation System for Printed Fabrics by Using Backward-Propagation Neural Network”, Fibers and Polymers, 8 (5), 529-536, 2007. [31] C., L., P., Hui, C., C., K., Chan, K., W., Yeung, & S., F., F., Ng, “Application of artificial neural networks to the prediction of sewing performance of fabrics”, International Journal of Clothing Science and Technology, 19 (5), 291-318, 2007. [32] C., L., P., Hui, & S., F., F., Ng, “Predicting Seam Performance of Commercial Woven Fabrics Using Multiple Logarithm Regression and Artificial Neural Networks”, Textile Research Journal, 79 (18), 1649-1657, 2009. [33] Y., Chen, T., Zhao, & B., J., Collier, “Prediction of Fabric End-use Using a Neural Network Technique”, Journal of the Textile Institute, 92 (2), 157-163, 2001. [34] G., K., N., Stylios, J., Powell, & L., Cheng, “An investigation into the engineering of the drapability of fabric”, Transactions of the Institute of Measurement and Control, 24, 33-50, 2002 [34] G., K., N., Stylios, & J., Powell, “Engineering the drapability of textile fabrics”, International Journal of Clothing Science and Technology, 15 (3/4), 211-217, 2003. [37] A., Lam, A., Raheja, & M., Govindaraj, “Neural network models for fabric drape prediction”, 2004 IEEE International Joint Conference on Neural Networks, Australia, IEEE, 25-29 July 2004. 2925-2929. [38] P., N., Koustoumpardis, J., S., Fourkiotis & N., A. Aspragathos, “Intelligent evaluation of fabrics' extensibility from robotized tensile test”, International Journal of Clothing Science and Technology, 19 (2), 80-98, 2007. [39] M., Hadizadeh, M., A., Tehran & A., A., A., Jeddi, “Application of an Adaptive Neurofuzzy System for Prediction of Initial Load/Extension Behavior of Plain-woven Fabrics”, Textile Research Journal, 80 (10), 981-990. 2010. [40] A., K., Pattanayak, A., Luximon, & A. Khandual, “Prediction of drape profile of cotton woven fabrics using artificial neural network and multiple regression method”, Textile Research Journal, 81 (6), 559-566. 2011. [41] H., Jedda, A., Ghith, & F., Sakli, “Prediction of fabric drape using the FAST system” Journal of the Textile Institute 98 (3), 219-225, 2007. [42] H., Hedfi, A., Ghith, & H., BelHadjSalah, “Dynamic fabric modelling and simulation using deformable models”, Journal of the Textile Institute, 102 (8), First published, 647667, 2011.

749