Prediction and optimization of yarn properties using genetic algorithm/artificial neural network

Abstract

Relative performance of the back propagation neural network (BPN) algorithm combined with genetic algorithm (GA) approach for the prediction/optimization of the properties of yarn produced on jet ring spinning system has been studied. Yarn samples of various linear densities have been produced on ring spinning machines using air-jet nozzles as retrofit by varying the nozzle parameter and the yarn properties studied. The hybrid application is used to predict selected yarn properties based on the effect of certain nozzle parameters. The network trained for a set of training vectors is found to predict the yarn properties for a compacting method with minimum error percentage. The proposed GA/BPN model could be extended to suggest a suitable compacting method for the desired yarn properties.