Precision simulation and rapid prediction of complex parts formed by laser powder bed fusion based on multi-scale inherent strain method

Laser Powder Bed Fusion (LPBF) achieves layer-by-layer manufacturing of metal parts by melting metal powder using a high-energy laser beam. In response to the low accuracy and efficiency of the traditional inherent strain method in predicting residual stress and structural deformation during LPBF manufacturing, this study establishes a multiscale parameter transfer model for the inherent strain method. At the same time, to improve the efficiency of the inherent strain calculation, a Gaussian process regression algorithm was introduced for fast prediction. Specific methods: This study proposes a dual-domain heterogeneous inherent strain loading method at the microscopic scale for calculating mesoscopic inherent strain, and uses mesoscopic scale unitisation to handle anisotropic microscopic inherent strains. The macroscopic scale performs partition loading of mesoscopic inherent strain values in different scanning feature regions within a single equivalent layer to achieve the final stress and deformation prediction. The comparison among the multi-scale inherent strain method, commercial simulation software, and actual measurement of machined parts shows that the prediction efficiency has been improved by 30 times at the macro scale, the maximum deformation prediction error was reduced to 3.67%, and the accuracy was improved by 60%.