Advanced modeling of surface roughness with artificial neural network, Taguchi method and genetic algorithm

Abstract

Modern manufacturing requires reliable and accurate models for the prediction of machining performance. Predicting surface roughness before actual machining plays a very important role in machining practice. This paper presents the modeling methodology for predicting the surface roughness in turning of unreinforced polyamide based on artificial neural networks (ANNs), Taguchi method and genetic algorithm (GA). The machining experiment was conducted based on Taguchi’s experimental design using L27 orthogonal array. Input variables consisted of cutting speed, feed rate, depth of cut and tool nose radius, while surface roughness (Ra) was considered as output variable. To systematically identify optimum settings of ANN design and training parameters, Taguchi method was applied. Furthermore, a simple procedure based on GA for enhancing the ANN model prediction accuracy was applied. Statistically assessed as an accurate model, ANN model equation was graphically presented in the form of contour plots to study the effect of the cutting parameters on the surface roughness.