Levenberg-Marquardt, Bayesian-regularization, and scaled conjugate gradient algorithms for predicting surface roughness accuracy on side milling AISI 1045

Abstract

Surface roughness quality is an important requirement for functional machine components such as considerations of wear, lubrication, corrosion, surface fatigue and friction. In machining, this is influenced by machining parameters and it is difficult to develop a theoretical model to describe machining efficiently and completely. In this study, prediction using Artificial Neural Network (ANN) was developed. The Levenberg Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG) algorithms were compared for the AISI 1045 side milling data. Machining parameters consist of cutting speed, feeding rate, radial and axial depth of cut. The network was trained using structures with the number of neurons 1 to 20 in.a hidden layer. It is found that the best network structure for the LM and BR algorithms is 4-10-1 and for the SCG algorithm is 4-9-1. The LM, BR, and SCG algorithms are able to produce predictions that are very close to the experimental results. Based on network performance, the algorithms that produce the best mean square error and coefficient of determination are SCG, LM and BR, respectively.