Predictive modelling and optimization of WEDM of nickel aluminium bronze alloy using optimised support vector regression and evolutionary algorithm

The primary objective of this study was to develop an accurate predictive framework and an efficient multi-objective optimisation strategy for wire electric discharge machining (WEDM) of NAB alloy, focusing on Cutting Speed (CS) and Surface Roughness (SR). An optimized Support Vector Regression (OSVR) model was constructed to capture the complex and stochastic input-output relationships inherent to the spark erosion process. The model exhibited excellent predictive accuracy, with MSE = 0.0027 and R² = 0.9970 for CS and MSE = 0.0012 and R² = 0.9924 for SR, validated through scatter and stem plots. To optimise the conflicting objectives of maximising CS and minimising SR, an adaptive offspring generation-driven indicator-based evolutionary algorithm (IBEA-AOG) was applied. The algorithm generated 100 Pareto-optimal solutions and outperformed twelve state-of-the-art algorithms, as confirmed by Friedman mean rank and Nemenyi tests. Spearman correlation analysis was used to assess the influence of process parameters on CS and SR. Surface integrity was further evaluated using field emission scanning electron microscopy (FESEM) and non-contact optical profilometry. High discharge energy settings produced surfaces with globules, large melted deposits, and overlapping craters due to poor dielectric flushing, while low discharge energy led to smoother surfaces with fewer craters and thinner recast layers. Microcracks were observed under all conditions, but were more pronounced at higher energies owing to steeper thermal gradients. Overall, the OSVR-IBEA-AOG hybrid framework proved highly effective for prediction, analysis, and multi-objective optimisation in WEDM.