Comparison between artificial neural network and response surface methodology in the prediction of the parameters of heat set polypropylene yarns

In the present paper, a response surface model has been introduced to predict the geometrical parameters of heat set polypropylene pile yarns. The input factors of the presented model include yarn twist, initial yarn count, time, and temperature of heat setting and the response factors are yarn count, yarn shrinkage, crimp contraction and packing factor after the heat setting process. To analyse the effect of this process on the yarn parameters, the dry heat setting process has been applied to all samples at different times and temperatures using an oven equipped with air circulation because of better accuracy and control of temperature. The obtained results showed that there is a positive relation between time and temperature and output parameters. Finally, the predicting equations discussions about the optimum points for maximum shrinkage and interactions of parameters have been presented. Hence, due to some disability of the RSM method, an ANN model has been designed to predict the parameters at higher accuracy. The results of the accomplished ANN model represent a higher prediction correlation coefficient compared to RSM.