Analysis of the effect of the ramie fiber properties on the yarn quality by neural network

Abstract

In this paper, the effect of the ramie fiber properties on yarn quality was analyzed. Due to the complex nonlinear relationship between the ramie fiber properties and yarn quality, three methods, the grey analysis combined with BP neural network, principle component analysis combined with BP neural network and pure BP neural network were applied to predict yarn quality on the basis of ramie fiber properties, respectively. The grey analysis and principle component analysis were expected to reduce the input layer node numbers of BP neural network, then the network structure can be simplified, therefore the prediction accuracy and stability can be improved. Compared with that of pure BP neural network, the results got from the other two methods are both better, the mean relative error between the forecast results and measured values of ramie yarn quality, such as the strength, strength irregularity, unevenness and neps, were all reduced greatly.