Optimizing LPBF process parameters and Al2O3 reinforcement in 17-4 PH stainless steel composites using ANN and NSGA-II

This study systematically optimizes Laser Powder Bed Fusion (LPBF) process parameters for the fabrication of 17-4 PH stainless steel composites reinforced with Al2O3 foam, integrating Artificial Neural Network (ANN) modeling and Non-dominated Sorting Genetic Algorithm-II (NSGA-II) optimization. A high-precision fiber laser with an 80 mu m spot size was employed, operating within laser power ranges of 225-300 W, scanning speeds of 800-2000 mm/s, and layer heights of 20-40 mu m. The influence of Al2O3 foam volume fractions (5-15%) and particle sizes (3-10 mu m) on mechanical and tribological performance was evaluated. At the optimized conditions such as laser power of 289.50 W, scanning speed of 1433.33 mm/s, layer height of 30 mu m, Al2O3 foam infill of 13.15%, and Al2O3 particle size of 6.93 mu m, the compressive strength reached 704 MPa, deviating 2.22% from the NSGA-II predicted value of 720 MPa. The wear rate was minimized to 9.3 mg/km, with a 3.13% deviation from the predicted 9.6 mg/km, while the coefficient of friction (COF) was reduced to 0.57, closely matching the predicted 0.55 within 3.51% error. The ANN model exhibited high predictive accuracy, with Mean Absolute Errors (MAE) of 39.72 MPa for compressive strength, 1.82 mg/km for wear rate, and 0.05 for COF, demonstrating its capability in capturing nonlinear dependencies between process parameters and material responses. The multi-objective NSGA-II framework successfully generated Pareto-optimal solutions, with all experimental validation tests exhibiting deviations below 5% from predicted values, underscoring the robustness of the integrated ANN-NSGA-II methodology.