A scalable and efficient model to predict fabric width of single jersey finished cotton knitted fabrics using statistical and computational techniques

Abstract

Knitted fabrics score over woven ones in terms of many advantages. They are easier to handle, better in comfort and possess more crease recovery property. They are also easily washable and more preferred as outerwear. The dimensional stability of knitted fabric depends on fabric width, thickness, and other geometrical properties. The quality and performance variation of fabrics is attributed to fabric width which is chiefly governed by the diameter of the knitting machine. The most important problem in the knitting industry is attaining optimum fabric width during the garment manufacturing process. The textile industry needs to find and fix a flawless model to predict the best possible fabric width accurately to overcome the unwanted textile wastage and provide the most advantageous technology in knitted fabric manufacturing. With the multiplicity of variables, new types of machines, yarn counts and relaxation processes, the number of parameters to predict fabric properties has shown a phenomenal increase. Statistical and computational models such as neural network, data mining, fuzzy logic and genetic algorithm support in improving the accuracy of prediction of fabric properties. This research work is principally concerned with the prediction of the most ideal fabric width using statistical and computational techniques known as data mining and rough set theory. Although some work has been done earlier on the prediction of fabric width using the data mining technique, there were some shortcomings in the work because the number of samples considered was small and not adequate to validate the technique. Therefore, to overcome these shortcomings extensive data were collected from the knitting industries in and around Tirupur for two years. The input attributes were selected as per the knowledge gained from the resources of previous studies, a newline