Machine Learning-Assisted Track Morphology Prediction in μ-LPBF with Spray-Based Powder Spreading-data.docx

Micro-scale laser powder bed fusion (μ-LPBF) technology demonstrates significant advantages in manufacturing high-precision components for aerospace and biomedical applications. However, achieving uniform powder spreading remains challenging due to the submicrometer-scale particle size, which critically affects the quality of deposited tracks. This study introduces a spray-based powder delivery method using a spray gun to improve powder spreading in μ-LPBF. To explore the influence of spray and laser parameters on track morphology, a machine learning-assisted framework is developed. A dataset is constructed, and four machine learning algorithms suitable for small datasets are employed to model key morphological features, including single-track height, width, penetration depth, and Ra, with the model demonstrating superior predictive performance (R² ≥ 0.8). SHapley Additive exPlanations (SHAP) are usedto quantify the contribution of each process parameter. A multi-objective optimization framework is then established, targeting surface quality and track uniformity, with the NSGA-II algorithm used to identify Pareto-optimal process parameter sets. Experimental results confirm that the optimized parameters substantially improve melt track consistency while meeting the demanding requirements of micro-scale additive manufacturing. This work demonstrates the potential of integrating machine learning and multi-objective optimization for intelligent process control in μ-LPBF.