Application of Machine Learning and MOGA for Predictive Modeling and Optimization of Inconel 718 Machining in WEDM

Abstract The present study focuses on the predictive modeling of WEDM responses for Inconel 718 and their simultaneous optimization using a multi-objective genetic algorithm (MOGA). Inconel 718, a nickel-based superalloy known for its exceptional performance at elevated temperatures, was machined using WEDM by varying parameters such as current (I p ), pulse duration (T on ), and pulse interval (T off ). The responses, including machining time (MT), material removal rate (MRR), and surface roughness (SR), were evaluated and modeled using multiple linear regression (MLR) and artificial neural network (ANN). The ANN models provided more accurate predictions, exhibiting lower errors compared to the MLR models, thereby validating the use of ANN for precise prediction. MOGA was applied to the regression equations derived from the ANN models. Multi-criteria decision-making (MCDM) approaches, such as complex proportional assessment and multi-objective optimization on the basis of ratio analysis, were then employed on the Pareto optimal solutions to identify the best settings for achieving lower MT and SR, and higher MRR. Both MCDM methods identified the optimal parameters as I p =2.07 A, T on =47.30 µs, and T off =9.02 µs. The validation experiment at optimum condition yielded error % of 0.22, 3.21, and 0.47, respectively, compared to their theoretical values. The machined surfaces were characterized using scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDS), and X-ray diffraction (XRD). SEM analysis revealed surface defects such as craters, globules, micro-voids, and debris lumps, which became more pronounced at higher discharge energy levels. EDS and XRD confirmed the presence of tool material residues and dielectric decomposition products.