Improved EDM Precision with Advanced Ensemble Machine Learning Prediction for INC-625 Superalloy

Abstract

This research examines the influence of the EDM machining parameters; including servo voltage, current, angle of cut, pulse on and pulse off times, regarding surface roughness (SR) and material removal rate (MRR), on the machining behavior of the INC-625 superalloy. The L42 machining trials were conducted using Response Surface Methodology - Central Composite Design (RSMCCD) to execute the machining trials. According to ANOVA, servo voltage was the most significant factor affecting both, the surface roughness (80.01%) and the material removal rate (77.33%). The SR decreased by 28.13% when the maximum servo voltage (70V) was employed and the MRR increased by 41.36% when the minimum servo voltage (50V) was employed. In addition, Desirability Analysis (DA) have been utilized to construct a response model to estimate the optimal setting values. Additionally, this study introduces a new heterogenous ensemble machine learning model combining Ridge Regression, Random Forest, XG Boost, and Neural Networks to predict EDM performance metrics. The model achieves high accuracy (R 2 =0.936 for surface roughness, R 2 =0.854 for material removal rate) using an optimized weighted average ensemble model. The ensemble method effectively represents the complex relationship between important process parameters and machining outputs. The validated ensemble model enables realtime EDM process optimization and provides actionable parameter recommendations for different manufacturing objectives. For quality-focused machining, a high servo voltage (70 V), low pulseon time (8 μs), and low current (3 A) produced the lowest surface roughness. For productivityfocused machining, higher pulse-on time (16 μs) and current (5 A) maximized material removal rate. A balanced machining strategy was achieved at intermediate parameter settings (PNT = 12 μs, PFT = 12 μs, SRV = 60-65 V, APC = 4 A, AGC = 60°), offering an effective compromise between machining efficiency and surface quality. These recommendations provide practical decision support guidelines for facilitating data-driven decision-making in industrial EDM applications.