Prediction of air-jet textured yarn properties using statistical method and neural network

Abstract

Artificial neural network has been used for predicting the air-jet textured yarn properties and the performance of ANN model has been compared with the statistical model based on Box-Behnken response surface design. Leaving apart some stray cases, the artificial neural network is able to predict the properties with reasonably low prediction error. Prediction

ability of the network is better for the instability and physical bulk property as compared to tenacity. For the set of data used for constructing the network, the mean square errors are comparatively higher in the neural network model than the regression model.