XGBoost Algorithm for Interpretable AI Prediction of Melt Pool Geometry in IoT-Enabled Additive Manufacturing within Industry 4.0 Utilizing NSGA-II

Additive manufacturing (AM), particularly with Laser Powder Bed Fusion (LPBF), excels in fabricating intricate geometries and custom components through layer-by-layer deposition. However, precise control over melt pool geometry is crucial for ensuring high quality and performance. This study uses Extreme Gradient Boosting (XGBoost) regression techniques to manage melt pool geometry in AM processes. The primary goal is to evaluate these models' effectiveness in predicting and optimizing melt pool depth, width, and length using extensive datasets of process parameters and material properties. The models produced promising results. The best performance was for melt pool length, with an MSE of 0.0036, RMSE of 0.0600, MAE of 0.0300, and an R-squared score of 0.9456. Integrating these models into Multi-Objective Optimization Using Non-Dominated Sorting Genetic Algorithm II (NSGA-II) enhances real-time monitoring, control, and optimization of melt pool geometry during production. This approach leverages IoT-enabled capabilities within Industry 4.0 frameworks, improving decision-making and operational efficiency in AM. By providing interpretable AI predictions, this study aims to enhance the quality, reliability, and efficiency of AM processes, contributing to optimized production outcomes in digital manufacturing environments.