Yarn Quality Indexing Using a Mechanical Stylus

Abstract

Fabric hand, an important characteristic to the textile industry, is influenced by such fiber parameters as flexural rigidity and friction and by yarn parameters such as count, twist, CV%, hairiness, stiffness, and softness. This study deals primarily with predicting the softness of knitted T-shirts from yarn quality parameters. The work consists of a short literature review on the existing yam parameters as well as fabric hand evaluation and prediction techniques. The latest developments in measuring the roughness of textile material surfaces are, also covered. Using the surface profile as tested by the mechanical stylus surface analyzer (MSSA), developed at North Carolina State Uni versity, a novel yarn surface analysis parameter called the "surface response average" (SRA) is developed, along with a model for the fiber/stylus tip interaction. Ten T-shirts produced from ten different yarn samples are ranked based on their softness by a panel of judges. The yams used to make these T-shirts are tested by Uster III and MSSA, and standard roughness parameters are calculated. The results show no significant corre lation between standard roughness parameters and fabric softness. The correlation be tween hand and SRA is about -0.6, which suggests that a higher SRA corresponds to a softer fabric.

Data mining and machine learning in textile industry

“…Another type of problem for which the NN technique was used in the textile industry, was to predict the utility properties of textile materials, such as the moisture and heat transfer rate in fabrics 14 and air permeability of fabrics. 15 The use of neural networks for color science 16 and textile printing 17 was also investigated. For example, Golob et al 17 proposed to determine the correct pigment combinations of dyes for textile printing using the ANN technique.…”

“…Figure 3 presents three different textile application examples of NN. While the first example 5 focuses on predicting the thread consumption of jean trousers, the others 15,18 focus on determining the product parameters (the air permeability of woven fabrics) and process parameters (fabric coating process).…”

“…Their study clearly showed that MLP models predicted more successfully when compared with RNNs. Another textile study 15 that used MLP was performed to predict the air permeability of woven fabrics. A radial basis function (RBF) network is another type of FFNN that uses RBFs as its activation functions.…”

“…While some textile studies used a single activation function for all NN layers (e.g., sigmoid, 25 linear 19 ), most of the studies 3,5,7,15,20,21,23,24 proposed the usage of different activation functions for different layers (i.e., sigmoid and linear activation functions in the hidden and output layers, respectively). Several textile studies 26,27 aimed to perform the analysis of different activation functions to figure out the optimal function for a problem.…”

“…Determining the value of the parameters (i.e., momentum, learning rate, bias value) is equally important for accurate work with ANNs to obtain reliable results. Some textile studies 6,15,[20][21][22][23]27 focused on the design issues in the building of an ANN classifier. For example, Bahadir et al 27 tested the NN efficiency by changing the number of neurons for the problem of predicting the drape behavior of woolen fabrics treated with different finishing processes.…”