Surface Roughness Modelling and Prediction Using Artificial Intelligence Based Models

Abstract

Surface roughness is a significant factor in product quality. Experimental investigation of surface roughness is associated with cumbersomeness, high cost and energy consumption. The results of artificial intelligence (AI) based models namely artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) were presented. Empirical data of hard turning operations containing cutting speed, feed rate and depth of cut and average surface roughness (Ra) were used as inputs and target process parameters respectively to train the AI based models. The performances of the models were evaluated using determination coefficient (DC) and root mean square error (RMSE) criteria. The results of the one-process-parameter influence on Ra donated that both models showed that feed rate was the most influential and dominant process parameter on Ra. However, the sensitivity analysis results indicated that the two different AI based models behaved differently with ANN having cutting speed and depth of cut as the most influential and dominant input-combination process parameters while that of ANFIS showed that feed rate and depth of cut were while describing and predicting the same surface roughness process under the same conditions. The results of the models were in concord with the empirical results of the average surface roughness. Both models predicted Ra but ANN performed better than ANFIS in both scenarios. The models can be used to design for the average surface roughness of hardened steel in hard turning operation thus saving cost, time and energy consumption in the experimental determination of average surface roughness, Ra.