Textile Research Journal

http://trj.sagepub.com An Artificial Neural Network Model for the Prediction of Spirality of Fully Relaxed Single Jersey Fabrics
Charlotte Marion Murrells, Xiao Ming Tao, Bin Gang Xu and Kwok Po Stephen Cheng Textile Research Journal 2009; 79; 227 DOI: 10.1177/0040517508094091 The online version of this article can be found at: http://trj.sagepub.com/cgi/content/abstract/79/3/227

Published by:
http://www.sagepublications.com

Additional services and information for Textile Research Journal can be found at: Email Alerts: http://trj.sagepub.com/cgi/alerts Subscriptions: http://trj.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.co.uk/journalsPermissions.nav Citations http://trj.sagepub.com/cgi/content/refs/79/3/227

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

Textile Research Journal

Article

An Artificial Neural Network Model for the Prediction of Spirality of Fully Relaxed Single Jersey Fabrics
The present paper proposes an artificial neural network model for the prediction of the degree of spirality of single jersey fabrics made from 100 % cotton conventional and modified ring spun yarns. The factors investigated were the yarn residual torque as the measured twist liveliness, yarn type, yarn linear density, fabric tightness factor, the number of feeders, rotational direction and gauge of the knitting machine and dyeing method. The artificial neural network model was compared with a multiple regression model, demonstrating that the neural network model produced superior results to predict the degree of fabric spirality after three washing and drying cycles. The relative importance of the investigated factors influencing the spirality of the fabric was also investigated.

Abstract

Charlotte Marion Murrells, Xiao Ming Tao1, Bin Gang Xu and Kwok Po Stephen Cheng
Institute of Textiles and Clothing, The Hong Kong Polytechnic University, Hung Hom, Hong Kong

Key words artificial neural networks, fabric spirality, models, multiple regression, prediction, twist liveliness

Fabric spirality affects the a esthetics and quality of knitted fabrics. It is a complex phenomenon arising from many factors and has been studied over the years [1]. However, the accurate modeling or prediction of this phenomenon has not been fully explored because of the complexity of the problem and the large amount of data required to establish quantitative relationships. The few models proposed to predict the degree of spirality have generally been based on the correlation between spirality and yarn twist and various other interacting factors. For instance, Tao et al. [2] determined empirical relationships between the spirality of cotton single jersey fabrics and yarn linear density, twist factor, fabric tightness factor and loop shape factor using statistical techniques. Similarly, Chen et al. [3] used regression techniques to determine empirical relationships between the spirality of plain knitted wool fabrics and parameters of plied yarns and fabrics including the twist factor of plied yarn, loop length and fiber diameter in both dry relaxed and simulated industrial relaxed states. They found that the twist factor of

the singles yarn used to produce the plied yarn had no significant effect. Moreover, they found that the tightness factor did not appear to have a significant correlation with the angle of spirality, which contradicts previous experimental results. The contradiction is expected as the fabrics were not fully relaxed, therefore, the strain stored in the yarns was not fully released. Many intermediate states may occur as the result of temporary setting due to a combination of fiber visco-elasticity and inter-fiber friction.1 However, these models do not directly take into account the predominant factor generally considered to affect the degree of spirality, that is, yarn residual torque often represented by the yarns twist liveliness. The only study to consider this factor is an analysis undertaken by Hepworth [4] who investigated theoretically the mecha-

1 Corresponding author: Institute of Textiles and Clothing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong; email: tctaoxm@inet.polyu.edu.hk

Textile Research Journal Vol 79(3): 227234 DOI: 10.1177/0040517508094091 Figures 25 appear in color online: http://trj.sagepub.com

www.trj.sagepub.com 2009 SAGE Publications Los Angeles, London, New Delhi and Singapore

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

TRJ

228

Textile Research Journal 79(3) investigation, namely yarn type, yarn linear density, tightness factor, the number of feeders on the knitting machine, the machine gauge (needles/inch), the rotational direction of the machine and whether the fabrics had been piece dyed or not. We also examined the relative importance of these factors in contributing to fabric spirality.

nism by which the use of twist lively yarns leads to spirality in a fabric. Through a computer simulation, the shape of a loop in the fabric form could be calculated from consideration of the inter-yarn pressures exerted on a loop by its neighbors. The presence of residual torque in the yarn was simulated by introducing torque acting on the yarn in a loop. Such an analysis yields information about the theoretical interactions between yarn residual torque and fabric spirality; however, practical applications may be limited because of its complexities. It is also based on certain assumptions and the results would be determined by the validity of these assumptions, which do not appear to be fully verified in the study. Therefore, although it is generally agreed that yarn residual torque is the main factor causing spirality, it has not usually been included as a parameter in previous investigations because of the lack of data of accurately measured yarn residual torque and careful selection of the relaxation state of the fabrics. In this work, we employed an accurate method [5, 6] to measure yarn twist liveliness and used the full relaxation state of the fabric as the test condition for fabric spirality. Methodologies suitable for predicting spirality include multiple regression analysis, but artificial neural network (ANN) modeling is increasingly used as a powerful predictive tool. With suitable training data sets, neural networks have been taught to learn well in a wide range of applications, such as in building predictive models in processes where many factors contribute to the eventual outcome, but where there is little knowledge about the exact relationships or interactions between the input and output parameters [7]. The power and advantage of neural networks lie in their ability to represent complex relationships and in their ability to learn these relationships directly from the data being modeled. In the textile field, many applications have been reported. For example, Pynckels et al. [8] used ANN for determining the spinning ability of fibers; Fan and Hunter [9] used ANN for predicting properties of worsted fabrics; Majumdar and Majumdar [10] used ANN models to predict ring yarn elongation from cotton fiber properties and Beltran et al. [11] used ANN compared with multiple regression to model the relationship between a number of fiber, yarn and fabric properties and the pilling tendency of wool knits. In this work, we employed an ANN model and a standard multiple linear regression model to predict the spirality of 100 % cotton single jersey fabrics from a number of factors considered to have the potential to influence fabric spirality. A statistical analysis was undertaken to check the validity of the predictions of spirality and to compare the results obtained from the two types of model. In addition to twist liveliness as a major contributor to fabric spirality, there are other yarn, fabric and knitting machine parameters that may have an influence [2, 3, 12]. Therefore, seven additional factors were included in this

Experimental
Yarn and Fabric Samples
In this work, the data from a total of 66 fabric samples were used. The fabrics were produced from three types of cotton yarn samples including conventional ring yarns, low torque ring yarns and plied yarns. Three yarn linear densities were selected (20, 30 and 37 tex) and the twist multiples of the conventional and low torque yarns were from 1.9 to 3.6. The low torque yarns were produced under the NuTorqueTM singles ring yarn system, which was developed at The Hong Kong Polytechnic University [1317] to directly balance the yarn torque in singles ring spun yarns during the spinning process, and hence reduce the twist liveliness of the yarns. All the fabrics were single jersey fabrics, but of various tightness factors knitted by circular knitting machines rotating both in the clockwise and anti-clockwise directions. The number of feeders included a single feeder, 54 feeders and 90 feeders, and the gauges of the machines were 20, 22 and 24 npi. Some of the fabrics were then piece dyed in the tubular form. The data were randomly divided into 53 and 13 sets of data that were used for training and evaluating the performance of the predictive models, respectively.

Test Methods
A testing methodology and apparatus were developed to evaluate yarn twist liveliness [5, 6] based on the ISO Standard 3343:1984 [18]. Two ends of a known length of yarn were brought together in water under a given tension and the number of snarls in the length of yarn was counted by a yarn twist tester. The propensity of the yarn to snarl indicated the level of yarn residual torque. The yarn samples from cones were measured for twist liveliness and 30 readings were taken for each yarn tested. In the measurements, the twist liveliness was recorded as a positive value if the yarn snarled in the S direction and negative if the yarn snarled in the Z direction. All yarns were conditioned for at least 24 hours prior to testing under standard atmospheric conditions (65 2 % RH, 20 2 C). For the measurement of fabric spirality, the fabric samples were placed on a flat surface for at least 48 hours in

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

An Artificial Neural Network Model for the Prediction C. M. Murrells et al.

229

TRJ

Table 1 Range of variables.

Input parameters Yarn and fabric parameters: Yarn linear density (tex) Yarn type Yarn twist liveliness (turns/25 cm) Tightness factor (tex Piece dyed Knitting machine parameters: Number of feeders Gauge Rotational direction Output parameter Degree of spirality (fully relaxed) 3.76 26.73 1 1 0 3 3 1
1/2

Min

Max

20 1 17 13 0

37 3 61 18 1

cm )

Figure 1 Technique for measuring fabric spirality.

Range of Input Variables for Model Development

standard atmospheric conditions. All fabrics were then subjected to wash and dry relaxation treatment consisting of three 3A cycles of laundering and tumble drying. The washing temperature and drying temperatures were 60 C and 65 C, respectively. After wash and dry relaxation treatment, the dried fabrics were again conditioned at standard atmospheric conditions for 24 hours. The angle of spirality was measured after the washing and drying procedure. A modified IWS test method TM 276 [19] was used to measure the fabric spirality both before and after the relaxation treatment. The angle was measured between the wale line and the line parallel to the machine running direction which in this case, the edge of the circular fabric as shown in Figure 1. The angle of spirality, , was calculated using the following equation: W W1 tan = -----------------L1 Eight yarn, fabric and knitting machine factors were used as input variables to the models. From the measured testing results and relevant parameters, the ranges of the variables used in the models are provided in Table 1. For the discrete parameters, the different levels were defined as follows: Yarn type: 1 = conventional single spun ring yarn, 2 = plied ring yarn, 3 = Nu-Torque singles spun ring yarn Piece-dyed: 0 = no, 1 = yes No. of feeders: 1 = 1 feeder, 2 = 54 feeders, 3 = 90 feeders Gauge: 1 = 20 gauge, 2 = 22 gauge, 3 = 24 gauge Rotational direction: 0 = clockwise, 1 = anti-clockwise

(1)

Data for Testing the Models

After the completion of the model development and training, the selected neural network model and the multiple regression model were evaluated with the testing samples that were separated from the main dataset. The testing data were not used in the model development or training in any way, and hence provided an out-of-sample data set to evaluate the models on. This gives some indication of how well the models will perform when presented with data from the real world. The set of data included 13 fabric samples with parameters shown in Table 2.

As can be seen, four measurements were taken for each fabric sample by substituting W1 with W2, W3 and W4, and L1 with L1 + L2, L1 + L2 + L3 and L1 + L2 + L3 + L4 in the equation. The spirality angle for each sample was taken as the mean of the four measurements. In the measurement, the degree of spirality was recorded as a positive value for the case of Z direction spirality and a negative value for S direction spirality.

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

TRJ

230

Textile Research Journal 79(3)

Table 2 Parameters of the test set.

Sample no. 1 2 3 4 5 6 7 8 9 10 11 12 13 Yarn linear density (tex) 30 30 30 30 30 30 30 30 30 30 30 20 37 Yarn type 1 3 2 2 2 1 2 3 3 3 2 3 1 Twist liveliness Number of (turns/25 cm) feeders 61 21 7 10 17 36 17 22 22 20 0 21 49 2 1 3 3 3 2 3 3 3 3 3 1 1 Gauge 1 2 3 3 3 1 3 3 3 3 3 2 2 Rotational direction 0 1 0 0 1 0 0 0 1 1 1 1 1 Piece dyed 0 0 0 0 0 0 1 1 1 0 0 0 0 Tightness factor 17.03 16.74 16.90 16.83 17.09 16.16 16.96 16.96 17.03 17.03 17.03 16.15 17.75 Measured angle of spirality (degrees) 10.18 2.19 1.54 2.27 2.56 7.30 2.15 2.29 3.17 0.70 1.82 1.48 8.29

Multiple Regression Analysis

Using the specified input and output variables, a standard stepwise multiple regression analysis was performed to predict the angle of spirality for fully relaxed fabrics. The analysis was carried out using Minitab statistical software (Version 13), employing a combination of forward selection and backward elimination. The stepwise regression technique inserts or removes variables from a regression model according to the statistical inclusion criterion (5 % level of significance) until a satisfactory regression equation is reached. The results revealed that twist liveliness, tightness factor and yarn linear density were the important parameters and together accounted for 92 % of the variance in the angle of spirality. The resultant regression equation is given below: Spirality = 56.58 + 0.172T 4.14TF 0.44L (2)

where the units of spirality, twist liveliness (T), tightness factor (TF) and yarn linear density (L) are degrees, turns/ 25 cm, tex1/2 cm1 and tex, respectively.

decisions are required to establish a suitable and stable network. In developing the neural network model, various network structures were tried with one hidden layer. It has been shown that multilayer feedforward networks with one hidden layer are universal approximators, able to model any complex linear function provided there is a sufficient number of hidden neurons available [20]. Therefore, one of the primary aspects of the neural network training process is the selection of the optimal number of hidden neurons. A network with too few hidden neurons would be incapable of differentiating between complex patterns, leading to only a linear estimate of the actual trend. Whereas, if a network has too many hidden neurons, it will follow the noise in the data, leading to poor generalization for untrained data. To establish the optimal number of neurons required in the hidden layer, an experimental process was conducted by systematically varying the number of neurons. The Levenberg-Marquardt algorithm was chosen as the basic learning procedure for the network. The hyperbolic tangent sigmoid transfer function was assigned as the activation function in the hidden layer and the linear function was used in the output layer.

Artificial Neural Network Model

ANN design
Neural network design is a complex process, as there are numerous possible architectures available and important

Inputs
There were eight inputs (Table 1) and a single output the degree of fabric spirality after wash and dry relaxation. However, from the multiple regression analysis it was found that only twist liveliness, tightness factor and yarn linear density were the important parameters. Therefore,

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

An Artificial Neural Network Model for the Prediction C. M. Murrells et al. in order to ensure that this difference in number of factors would not affect a valid comparison of predictive results from the different models, we trained the neural network using an additional set comprised of only three inputs, namely twist liveliness, tightness factor and yarn linear density. In both cases, the data sets were normalized between limits of 1 and +1, with the average value set to zero. The normalized variable xi, norm is represented by: x i x i, avg x i, norm = -------------------R i, max and R i, max = Maximum ( x i, max x i, avg ), ( x i, avg x i, max ) (4) (3)

231

TRJ

where xi is an input or output variable, xi, avg is the average value of the variable over the data set, xi, min is the minimum value of the variable, xi, max is the maximum value of the variable and Ri, max is the maximum range between the average value and either the minimum or the maximum value.

Figure 2 Relationship between the actual and predicted values of spirality for the training set of data (neural network model) three factors.

ANN training
During the process, the data were divided into training, validation and test sets. The training set consisted of 60 % of the original data, 20 % of the data was set aside for crossvalidation set and the remaining 20 % of the data was used as the test set to evaluate the performance of the network. The mean square error (MSE) the average squared error between the network outputs and the target outputs was used to measure the performance of the model.

Results and Discussion

For the ANN models, the effect of the number of hidden neurons from two to 20 was studied for the networks trained with three and eight inputs. We found that with three inputs, the network with 14 hidden neurons gave the best prediction results with a MSE of 0.00461, whereas with eight inputs, the network with six hidden neurons gave the best prediction results with a MSE of 0.00327. The performance of a trained network can be measured to some extent by the errors on the training, validation and test sets, but it is often useful to investigate the network response in more detail. One option is to perform a regression analysis between the network response and the corresponding targets. Therefore, the accuracy of the neural network models for three and eight inputs was judged by the regression coefficient of the line that relates the fabric spirality predicted by the model and the actual tested val-

Figure 3 Relationship between the actual and predicted values of spirality for the training set of data (neural network model) eight factors.

ues of the spirality of the fabric used for training. This is shown in Figures 2 and 3, respectively. The prediction was

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

TRJ

232

Textile Research Journal 79(3) considered good if all the data points were aligned on the line of perfect fit, whereas the amount of scattering around the line gave an indication of the quality of the modeling. The results showed that, for both the three input and the eight input models, the predicted values, A, tracked the measured or actual values, T, very well, as the correlation coefficients were high (for three inputs R = 0.963 and eight inputs R = 0.967). Similar to the results of the neural network models in Figures 2 and 3, the prediction accuracy of the multiple regression model was judged by the regression coefficient of the line that relates the fabric spirality predicted by the model and the actual tested values of the spirality of the fabric used for training, as shown in Figure 4. The results showed that the predicted values, A, tracked the measured or actual values, T, very well, as the correlation coefficient was sufficiently high (R = 0.959).

Prediction Performance
Figure 4 Relationship between the actual and predicted values of spirality for the training set of data (multiple regression model). The testing data were used as an out-of-sample set in order to evaluate the predictive performance of the finalized models. A simple comparison of the results, as in Figure 5, showed the spirality values predicted by the multiple regression and neural network models versus the actual target measured

Figure 5 Predicted and measured degree of spirality for the test fabrics.

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

An Artificial Neural Network Model for the Prediction C. M. Murrells et al.

233

TRJ

Table 3 Comparison of prediction performance of the three models.

Statistical parameter Correlation coefficient, R Mean square error Mean absolute error Regression 0.970 1.201 0.941 ANN: three inputs 0.915 2.875 1.028 ANN: eight inputs 0.976 0.906 0.622

Table 4 Relative importance of the input variables.

Excluded input variable Yarn linear density Yarn type Yarn twist liveliness Piece dyed Tightness factor Number of feeders Gauge Rotational direction Mean square error 0.446251 0.003535 0.616312 0.004661 0.306652 0.168523 0.165278 0.063617 % Increase in mean square error 13545.11 8.09 18745.10 42.52 9276.56 5052.96 4953.73 1845.23 Rank 2 8 1 7 3 4 5 6

values for each of the samples in the test data set. It indicated that the eight input neural network model produced the best predictive performance, whereas there was a wider discrepancy between predicted and actual values by both the three input neural network and the multiple regression models. A comparison of the results from the neural network models showed that the predicted values from the model with three inputs were not as good as the model with eight inputs. The results were also subjected to further analysis where the predictive accuracy was judged by statistical parameters, such as the correlation coefficient between the actual and predicted degree of spirality (R), mean square error and the mean absolute error. The results are shown in Table 3. The correlation coefficients (R) between the actual and predicted degree of spirality were very high for all models, but the neural network model with eight inputs had a slightly higher value of 0.976 compared with 0.970 for the regression model and 0.915 for the neural network with three inputs. This order of predictive ability was also confirmed by the mean square and mean absolute errors, as both the resulting errors were lower for the neural network with eight inputs.

input saliency test by eliminating only one designated input from the optimized neural network model at a time. The increase in the mean square error value compared to the optimized ANN model was considered as the indicator of the importance of the excluded input. The results are shown in Table 4. It can be seen that yarn twist liveliness was the dominant variable, as expected. Yarn linear density and tightness factor were the next two variables in the hierarchy of contributors to the angle of spirality, which concurs with the results of the multiple regression analysis. They were followed by number of feeders, gauge, rotational direction, piece dyed and yarn type were indicated to have relatively minor influence on the angle of spirality for single jersey knitted fabrics made from pure cotton ring spun yarns.

Conclusions
Using an artificial neural network model, we predicted the degree of spirality of fully relaxed single jersey fabrics made from 100 % cotton ring spun yarns from eight yarn, fabric and knitting machine parameters, i.e., the measured twist liveliness, yarn linear density, yarn type, fabric tightness factor, piece dyeing and the number of feeders, rotational direction and gauge of the knitting machine. We achieved a relatively good agreement between our predictions and actual measured values of fabric spirality with a correlation coefficient, R, of 0.976 in out-of-sample testing. To assess the validity of this approach, the models results were compared to those produced by use of a multi-

The Relative Importance of the Input Variables

To provide a general indication of the relative importance of the various input variables using ANN, we conducted an

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009

TRJ

234

Textile Research Journal 79(3)

ceedings of the Textile Institute 83rd World Conference, Shanghai, China, pp. 401404 (2004). Xu, B. G., Murrells, C. M., and Tao, X. M., Automatic Measurement and Recognition of Yarn Snarls by Digital Image and Signal Processing Methods, Textile Res. J. 78(5), 439456 (2008). Beltran, R., Wang, L., and Wang, X., Predicting Spinning Performance with an Artificial Neural Network Model, Textile Res. J. 74, 757763 (2004). Pynckels, F., Kiekens, P., Sette, S., Van-Langenhove, L., and Impe, K., Use of Neural Nets for Determining the Spinnability of Fibres, J. Textile Inst. 86(3), 425437 (1995). Fan, J., and Hunter, L., A Worsted Fabric Expert System Part II: an Artificial Neural Network Model for Predicting the Properties of Worsted Fabrics, Textile Res. J. 68(10), 763771 (1998). Majumdar, P. K., and Majumdar, A., Predicting the Breaking Elongation of Ring Spun Cotton Yarns Using Mathematical, Statistical and Artificial Neural Network Models, Textile Res. J. 74, 652655 (2004). Beltran, R., Wang, L., and Wang, X., Measuring the Influence of Fibre-to-fibre Properties on the Pilling of Wool Fabrics, J. Textile Inst. 97, 197204 (2006). Primentas, A., Mechanical Methods for the Reduction of Spirality in Weft Knitted Fabrics, PhD Thesis, University of Leeds, Leeds, England (1995). Tao, X. M., and Xu, B. G., Manufacturing Method and Apparatus for Torque-free Singles Ring Yarns, U.S. Patent 6,860,095 (2005). Tao, X. M., Xu, B. G., and Wong, S. K., Method and Apparatus for Manufacturing a Singles Ring Yarn, U.S. Patent 7,096,655 (2006). Yang, K., Tao, X. M., Xu, B. G., and Lam, J., Structure and Properties of Low Twist Short-staple Singles Ring Spun Yarns, Textile Res. J. 77(9), 675685 (2007). Hua, T., Tao, X. M., Cheng, K. P. S., and Xu, B. G., Effects of Geometry of Ring Spinning Triangle on Yarn Torque: Part I: Analysis of Fiber Tension Distribution, Textile Res. J. 77(11), 853863 (2007). Xu, B. G., and Tao, X. M., Techniques for Torque Modification of Singles Ring Spun Yarns, Textile Res. J. (in Press). ISO Standard 3343:1984, Textile Glass-yarns-determination of Twist Balance Index. IWS Standard TM 276, Test Method for the Angle of Spirality (in Plain Knitted Garments). Babay, A., Cheikhrouhou, M., Vermeulen, B., Rabensolo, B., and Castelain, J. M., Selecting the Optimal Neural Network Architecture for Predicting Cotton Yarn Hairiness, J. Textile Inst. 96, 185192 (2004).

ple regression technique. The neural network model outperformed the multiple regression model in predicting the angle of spirality using data that were not used to train the network. This indicates that it is worthwhile using the more complex ANN technique if a large amount of different types of data are available. Both the ANN and the regression approach showed that twist liveliness, tightness factor and yarn linear density were the most important factors in predicting fabric spirality, with the other factors having a minor influence. In particular, twist liveliness was the most important parameter. This confirms the generally held view that twist liveliness is the major contributor to spirality.

6.

7.

8.

9.

10.

Acknowledgements
We would like to acknowledge the funding support from the Hong Kong Research Grants Council and from the Commissioner of Innovation and Technology, the Government of the Hong Kong SARs Innovation and Technology Fund and sponsorships from Central Textiles (H.K.) Limited, Chip Tak Weaving Factory Limited, Fountain Set (Holdings) Limited, Perfecta Dyeing Printing & Weaving Works Limited. C. M. Murrells would also like to thank The Hong Kong Polytechnic University for providing her with a postgraduate scholarship.
11.

12.

13.

14.

Literature Cited
1. 2. Lau, Y. M., Tao, X. M., and Dhingra, R. C., Spirality in Singlejersey Fabrics, Textile Asia 8, 95102 (1995). Tao, J., Dhingra, R. C., Chan, C. K., and Abbas, M. S., Effects of Yarn and Fabric Construction on Spirality of Cotton Single Jersey Fabrics, Textile Res. J. 67, 5768 (1997). Chen, Q. H., Au, K. F., Yuen, C. W. M., and Yeung, K. W., Effects of Yarn and Knitting Parameters on the Spirality of Plain Knitted Wool Fabrics, Textile Res. J. 73, 421426 (2003). Hepworth, B., Spirality in Knitted Fabrics Caused by Twistlikely Yarns: a Theoretical Investigation, Melliand Textilber. 74, 515520 E212E213 (1993). Murrells, C. M., Wong, K. K., Hua, T., Leung, C. L., Xu, B. G., Yang, K., Wong, S. K., Cheng, K. P. S., and Tao, X. M., Study of Yarn Snarling in Nu-Torque Singles Ring Yarns, in Pro-

15.

16.

17. 18. 19. 20.

3.

4.

5.

Downloaded from http://trj.sagepub.com by Subrata Saha on April 29, 2009