Enhanced prediction and optimization of WEDM for titanium alloy (grade 5) with hybrid artificial neural network based meta-heuristics

This study presents a novel hybrid artificial neural network (ANN) meta-heuristic model to predict and optimize titanium alloy (Grade 5) in wire electrical discharge machining (WEDM). The model integrates genetic algorithm (GA), particle swarm optimization (PSO), and grey wolf optimization (GWO) separately with ANN, comparing their performance. Additionally, two static tools, a multi-objective optimization method using ratio analysis (MOORA) and a technique for order performance by similarity to the ideal solution (TOPSIS), are employed for comparison. The most effective method is chosen for validation, considering six WEDM process variables and four response variables in the hybrid ANN-based meta-heuristic machine learning model. The MOORA and TOPSIS exhibit maximum response prediction average errors of 37.4% and 7.5%, respectively. In contrast, the ANN-GA demonstrated ~ 2.4% error, whereas ~ 3.5% error is observed by ANN-PSO, and ANN-GWO showed ~ 2.9% error for all response variables. The ANN-GA performed best among the meta-heuristic models with less prediction error and accurate optimum values. The error sensitivity analysis shows that the AO_off and W_F show the highest positive and negative sensitivity on the machining performance. The Pareto front displays that as M_RR increases, K_W simultaneously increases; however, S_R and RLT decrease with K_W increases in these four response variables.