Abstract:

A study to investigate the prediction abilities of BP and RBF neutral networks for worsted yarn performances used the process parameters collected from the fore-spinning and post-spinning as input vectors and yarn unevenness value (CV) and breaking strength (BS) indicating the worsted yarn performances as output vectors. Two software computing tools, i. e., back-propagation (BP) neural network and radial basis function (RBF) neural network, were used to establish the prediction models for the CV and BS of the yarn respectively, and the prediction abilities of the two models were evaluated from the view point of statistics. The results show that the training speed of RBF neural network model is significantly higher than that of the BP neural network model, which are under the conditions of large-scale input samples, high input dimensions and same accuracy, but the forecasting performance of BP neural network model is slightly better than that of RBP neural network model, especially in face of abnormal sample, and BP neural network model shows better fault-tolerant capacity.