Part II. Predicting the Pilling Tendency of the Cotton Interlock Knitted Fabrics by Artificial Neural Network

Abstract

Artificial neural network (ANN) is a mathematical model inspired by biological neural networks and it processes information using a connectionist approach to computation. The aim of the second part of the study is to determine models for estimating the pilling propensity of the interlock knitted fabrics produced from yarns of different yarn counts (Ne 20, Ne 30, Ne 40) and yarn twist coefficients (αe=3.2, αe=3.6, αe=4.0) spun by using seven different cotton types harvested from different regions. The fabrics were manufactured in three different tightness factors, including dense, medium, and loose, by changing the yarn length utilized in each course of the fabrics. The models for pilling degree, total pill number, total weighted pill number, average pill area, and average pill height of the fabrics evaluated by PillGrade Objective Pilling Grading System, were derived by using a neural network method. In order to define the effective properties on pilling formation, sensitivity analysis was carried out. All models indicated relatively good estimation power. Fabric cover factor and short fiber content were found as the most significant parameters influencing the pilling propensity feature of the interlock knitted fabrics.