HYBRID MACHINE LEARNING AND GREY WOLF OPTIMIZER FRAMEWORK FOR MULTI-RESPONSE PARAMETRIC OPTIMIZATION IN MICRO-EDM DRILLING OF NIMONIC C263 SUPERALLOY USING SILVER-COATED BRASS TUBULAR ELECTRODES

Abstract

This study presents a machine learning (ML)-assisted prediction and Grey Wolf Optimizer (GWO)-based multi-objective optimization framework for Micro-Electrical Discharge Machining Drilling (Micro-EDMD) of Nimonic C263, a nickel-cobalt superalloy widely employed in aerospace and high-temperature structural components. Experiments were conducted following a Taguchi L9 orthogonal array with three process parameters — Pulse-ON time (Ton), Pulse-OFF time (Toff), and Discharge Current (DC) — at three levels each, comparing uncoated Brass Tubular Electrodes (BTE) against Silver-Coated Brass Tubular Electrodes (SCBTE). Three critical performance responses were evaluated: Material Removal Rate (MRR), Tubular Electrode Wear Rate (TEWR), and Hole Overcut (OC). ANOVA-based Signal-to-Noise (S/N) analysis confirmed that Discharge Current is the dominant parameter for MRR (PCR = 68.4%) and TEWR (PCR = 71.2%), while Pulse-ON time governs hole overcut behaviour (PCR = 52.7%). Linear Regression (LR) models augmented with leave-one-out cross-validation (LOOCV) achieved R² values of 0.91 (MRR-BTE), 0.93 (MRR-SCBTE), 0.87–0.89 (TEWR), and 0.83–0.85 (OC), demonstrating reliable predictive accuracy for Taguchi-scale datasets. The hybrid ML-GWO optimization framework identified continuous-domain optimal settings (Ton = 76 μs, Toff = 12 μs, DC = 5.6 A for SCBTE) with prediction errors below 3.5% relative to experimental observations. SCBTE achieved an average MRR improvement of 12.6%, TEWR reduction of 12.3%, and competitive overcut performance versus BTE. This work demonstrates the first application of a hybrid ML-GWO framework for parametric optimization in Micro-EDMD of Nimonic C263 using silver-coated tubular electrodes