Application of machine learning techniques in environmentally benign surface grinding of Inconel 625

Abstract

This work explores the grinding performance of Inconel 625 by comparing tangential forces and surface roughness under dry, wet, and minimum quantity lubrication (MQL) environments. Response surface methodology (RSM) and machine learning (ML) techniques establish correlations between inputs and outputs. RSM shows the quadratic regression model effectively captures experimental variability. Four ML models, namely: multilayer perceptron, KNN, and SVM with two kernels, were implemented to predict outcomes, among which KNN appeared as a better-suited model. Lower tangential forces and better ground surface quality support the suitability of MQL grinding compared with dry and wet grinding. Micro-graph investigations of the ground surfaces and chip morphology findings also upkeep the emergence of MQL grinding as a sustainable alternative for Inconel 625 grinding.