Prediction and optimization of hardness in AlSi10Mg alloy produced by laser powder bed fusion using statistical and machine learning approaches.

The primary objective of this study is to evaluate the influence of critical process parameters on the hardness of AlSi10Mg alloy fabricated via the Laser Powder Bed Fusion (LPBF) technique. To optimize these parameters, a Taguchi-based signal-to-noise (S/N) ratio analysis was employed. Furthermore, Analysis of Variance (ANOVA) was conducted to quantify the statistical significance and contribution ratio of each parameter to the observed variation in hardness. In addition, Machine Learning algorithms were applied to predict hardness values. A multidisciplinary approach was adopted to optimize production processes, reduce manufacturing costs, and shorten processing times. Taguchi and ANOVA analyses were performed using Minitab 21, while Machine Learning implementations were carried out in R statistical software. The applied Machine Learning techniques include Artificial Neural Networks (ANN), Random Forest (RF), Support Vector Machines (SVM), and Multiple Linear Regression (MLR). The results indicate that scan speed exerts the most significant influence on hardness, followed by laser power. Notably, the highest hardness was achieved using a laser power of 200 W, a hatch distance of 0.20 mm, and a scan speed of 2000 mm/s. Among the models, SVM achieved the highest coefficient of determination (R = 0.73). This study highlights the importance of integrating Machine Learning and statistical analysis methods for the effective modeling and optimization of LPBF processes. The findings contribute significantly to the literature and serve as a valuable reference for future research aimed at improving LPBF process efficiency and performance.