Optimization of laser powder bed fusion process parameters for AlSi10Mg samples towards mechanical strength by using metaheuristic optimization algorithms

Laser Powder Bed Fusion (LPBF) is a highly favoured route for additive manufacturing of AlSi10Mg alloys; however, the maximum mechanical performance of additively manufactured components relies on tight control of a web of interrelated process variables. This research develops a statistical-metaheuristic hybrid model which optimizes the Ultimate Tensile Strength (UTS) of LPBF-fabricated AlSi10Mg within a localized region of the process parameter space. This scientific research was conducted with a limited 3² full-factorial experimental design (n = 9, without replication) that constructed a two-variable quadratic regression model (R² = 0.7475, RMSE = 34.57 MPa) between Laser Power (P) and Exposure Time (E) and tensile response. The Volumetric Energy Density (VED) derived was controlled for out of the model in order to remove the statistical multicollinearity that it brings. Four metaheuristic algorithms (Genetic Algorithm (GA), Simulated Annealing (SA), Teaching-Learning-Based Optimization (TLBO), and JAYA) were then run under a deliberately limited stress test of computational load (N = 12, Iter_max = 100) in MATLAB R2025b. The parameter-dependent solvers, GA and SA, collapsed to an early convergence and all stopped at local optima of 364.30 MPa and 355.59 MPa, respectively. In comparison, JAYA and TLBO, which were parameter-free, found the estimated optimum of 378.74 MPa at P = 350 W and E = 50 µs for the fitted response surface. Two physical experiments that were printed at these settings yielded a real UTS of 380.45 MPa and a prediction error of 0.45%, far less than the 5% engineering threshold. This consensus confirms the regression model and the comparative efficiency of parameter-free algorithms as a valuable proof-of-concept optimization approach within constrained LPBF parameter boundaries.