Performance Prediction and Multi-Response Optimization of Si ₃ N ₄ -ZrO ₂ Composites by ECDM Using Hybrid ANN-BWKA and ALMO model

The growing demand for micro-level products in modern industries requires machining processes capable of producing precise micro-features with high strength, hardness, conductivity, and accuracy. This study investigates the machining performance of 5% graphene-reinforced Silicon Nitride-Zirconia (Si 3 N 4 -ZrO 2 -5%Gr) composites using the Electrochemical Discharge Machining (ECDM) process. Moreover, the performance of the Zinc Oxide (ZnO) and Titanium Dioxide (TiO2) nanoparticles, and the Sodium hydroxide (NaOH) and Potassium hydroxide (KOH) electrolytes on the machining of the composite was analyzed. Material Removal Rate (MRR), Tool Wear Rate (TWR), Surface Roughness (SR), and Aspect Ratio (AR) were chosen as the main output responses. Response Surface Methodology (RSM) based on Box–Behnken Design (BBD) was employed to optimize 234 experimental runs and determine the optimal machining parameters. A novel hybrid Artificial Neural Network-Black Winged Kite Algorithm (ANN-BWKA) model was proposed to predict the output response and then compared with XGBoost, RF, SVR, KNN, GPR, and RSM. Furthermore, an Active Learning-based Multi-objective Optimization (ALMO) was used to select the optimal process parameters. The experimental investigation achieved optimal values of MRR (0.921 mm 3 /min), TWR (0.117 mm 3 /min), SR (2.33 μm), and AR (1.461). The RSM-BBD optimization resulted in improved MRR and AR of 0.926 mm 3 /min and 1.453, respectively, and reduced TWR and SR to 0.12 mm 3 /min and 2.378 μm. The proposed ANN-BWKA model performed better than all the baseline models with R 2 values above 0.99 for all the output responses. Moreover, the proposed model achieved a lower MAE, MAPE and RMSE value of less than 0.02 compared to XGBoost, RF, SVR, KNN, GPR and RSM, indicating that the proposed model outperforms in prediction accuracy and robustness. Finally, the ALMO model reveals the optimal parameters with an MRR of 0.93mm 3 /min, AR of 1.455, a TWR of 0.115 mm 3 /min and SR of 2.3 μm.