Rapid residual stress simulation and distortion mitigation in laser additive manufacturing through machine learning

Abstract

This study presents a methodology for simulating residual stress and distortion in Laser Powder Bed Fusion (LPBF) additive manufacturing by integrating a simplified finite element analysis (FEA) framework, high-resolution residual stress mapping via the contour method, and machine learning (ML) algorithms to enhance both simulation efficiency and accuracy. Using a two-parameter temperature field, the FEA model reduces computational complexity while maintaining precision. Three ML models, multi-layer perceptron (MLP), gradient boosting (GB) regressor, and random forest (RF) regressor, are trained on FEA simulation data and validated against experimental measurements, showing effective performance with discrepancies ranging from 52 MPa to 84 MPa. The methodology also enables accurate distortion predictions, allowing for a key application on distortion mitigation, where predicted distortions are inversely applied to the CAD model to counteract stress-induced warping. This approach reduces distortion in a bridge sample from 0.94 mm to 0.06 mm - a 94 % improvement. This integrated approach provides a robust tool for predicting residual stresses and mitigating distortions in LPBF, optimizing design and manufacturing practices.