Yarn Hairiness Prediction by Generalized Regression Neural Network based on Harris Hawk Optimization

Abstract

Yarn hairiness is an important indicator of yarn quality. It affects not only the quality of yarn but also the woven and knitted performance of yarn and the quality of the fabric produced. The prediction of yarn hairiness index can effectively prevent errors in produced fabrics, but yarn hairiness prediction is a complex nonlinear problem, and the use of a simple prediction model cannot meet the need for yarn hairiness prediction accuracy. Therefore, the main objective of this study is to introduce a new metaheuristic optimization method, namely Harris Hawk Optimization, to improve the accuracy of yarn hairiness prediction by generalized regression neural network. The smoothing factor in the generalized regression neural network is optimally selected by Harris Hawk Optimization, which in turn improves the accuracy of prediction. The experimental results show that the generalized neural network using Harris Hawk Optimization has very high accuracy in predicting yarn hairiness. In this regard, its root mean square error and mean absolute error criteria are 0.05568 and 0.03872, respectively.