Comprehensive stratified K-fold machine learning for surface roughness prediction with 3D printed electrodes: Analyzing dielectric variability and heat treatment effects

Abstract

Electrical discharge machining (EDM) is popular for its accuracy in cutting difficult materials. Material removal rate (MRR), electrode wear ratio (EWR) and surface roughness (SR) are deemed as the success indicators in EDM. The positive attributes of sustainable manufacturing make it a suitable choice for diverse applications. EDM electrodes used in this work are prepared by 3D printing and further subjected to deep cryogenic treatment (DCT) to investigate the influence of heat treatment on EDM performance. In the second instance, the experiments are conducted with different dielectric fluids viz. conventional dielectric (CD) and lemon peel biodiesel dielectric (LPD) to analyze the impact of dielectric change. When the dielectric is changed to lemon peel biodiesel, the roughness of the electro-discharged machined surface is reduced to a maximum of about 33%. A further improvement of about 11% is possible in the surface finish of machined surfaces when the EDM electrodes are subjected to DCT under an identical operating condition. LPD as dielectric also helps in attaining the primary objective of green manufacturing by reducing specific evaporative emissions. During the EDM process, LPD adoption causes a reduction of 96% in nitrogen oxide (NO) emissions and 86% in hydrocarbon (HC) emissions. A modified adaptive neuro fuzzy inference system (ANFIS) with a stratified k-fold machine learning approach with a prediction accuracy of about 95% is also developed in this work to envisage the surface roughness of electro-discharge machined surfaces.