Prediction of surface finish model in cutting AISI 4140 steel with different approaches

Abstract

This experimantel study describes the development of surface roughness model with main parameters including tool radius using full-factorial design approach and artificial neural network (ANN). Cutting tests and analysis of variance were used in cutting AISI 4140 steels by coated cutting tools. Factorial design/multi quadratic regression (MQR) were compared to ANN model. The results indicated that surface finish decreased with decreasing feed rate and increasing nose radius. It is showed that both feed rate and tool nose radius were effective while other factors were insignificant effect. For testing stage of both methods, data was selected randomly from the existing experimental runs. Further, both randomly selected ANN and MQR indicated no significant differences for prediction the surface roughness because PE and RMSE were 2.73%, 2.21%, 0.063 and 0.046 for MQR and ANN, respectively. Both approaches can used effectively for prediction of any machinability studies in manufacturing engineering due to high accuracy of results. In the future work, other nonlinear models like support vector machine and principal component analysis would be conducted to improve performance accuracy.