Proper estimation of surface roughness using hybrid intelligence based on artificial neural network and genetic algorithm

Abstract

The surface roughness is a crucial index that is commonly used in the machining process to evaluate the final product quality. This paper investigates the effect of different machining parameters on the surface roughness of the high-strength carbon fiber composite plate, manufactured by utilizing the vacuum infusion process, under dry end milling conditions. Besides, a hybrid intelligence approach consisting of artificial neural network (ANN) whose parameters are tuned by genetic algorithm (GA) is introduced for accurate estimation of surface roughness. To construct a database for the ANN, the experimental milling tests have been carried out according to the Taguchi optimization method with the design of a mixed orthogonal array L32 (21 × 42). The influence of the machining parameters such as cutting tools, feed rate, and spindle speed on surface roughness have been examined by using analysis of variance (ANOVA). The analyses reveal that the cutting tool and the feed rate are the most effective factors in the surface roughness of the composite material. It is also determined that the experiment with “A1B2C1” combination (TiAlN coated cutting tool, 5000 rpm spindle speed, and 250 mm/rev feed rate) gives the optimal result. The proposed hybrid ANN-GA algorithm provides a good prediction correlation ratio (R = 0.96177) indicating that the estimated and the measured surface roughness values are remarkably close to each other. The mean square error (MSE) specifying the accuracy and adequacy of the network model is obtained as 0.074 during the 33th iteration of the GA.