Modelling and optimisation of magnetic abrasive finishing process based on a non-orthogonal array with ANN-GA approach

Abstract

ABSTRACT Magnetic abrasive finishing (MAF) is an advanced precise finishing method that achieves micro-level to nano-level surface roughness. In industries, MAF is highly recommended where zero or negligible post-process surface defects are an obligatory requirement. In the same context, process optimisation is essential for making it commercially viable. This study presents an artificial neural network and genetic algorithm (ANN-GA), a robust modelling and optimisation tool (applicable to any sort of data set orthogonal array design or non-orthogonal array design) that is applied to scrutinise and improve the performance of the magnetic abrasive finishing of stainless steel SS302. In addition, the results from ANN-GA modelling and optimisation have been compared with conclusions drawn from conventionally used Taguchi-ANOVA analysis. An L27 non-orthogonal array design has been opted for as per machining set-up restriction. Abrasive size, voltage, machining gap, and rotational speed were the design variables considered in the present research work. It was found that the parametric design used in this study provides a straightforward, methodical, and proficient method of modelling and optimisation of change of surface roughness or finishing behaviour during the MAF process. Modelling and optimisation done with ANN-GA show a maximum value of (ΔR a)max equal to 0.256 µm, which is 7% better than the result obtained from Taguchi-ANOVA analysis.