A Generalized Model for Predicting Yarn Global Quality Index

Abstract

In this work we tried to predict Ring spun yarn quality from fiber properties, yarn count and twist by using a back-propagation network. First, we have proceeded with principal component and path analyses to extract and to visualize the main characteristics of the data set. Then, a model was constructed using neural network to optimize final yarn quality. The study of yarn quality was based on the desirability approach. The present method allowed us to quantify yarn quality by an index belonging to the interval [0, 1] which includes major physical properties of cotton Ring spun yarn. It can be optimized by using an algorithm, which is modified when criteria requirements of yarn quality are changed. All these methods have contributed to establish a convenient model that could predict global yarn quality.