Physics-Informed Machine Learning Regulated by Finite Element Analysis for Simulation Acceleration of Melt Pool Dynamics in Laser Powder Bed Fusion
Researchers have developed a novel framework called FEA-Regulated Physics-Informed Neural Network (FEA-PINN) to accelerate simulations of melt pool dynamics in Laser Powder Bed Fusion (LPBF). This new approach integrates corrective Finite Element Analysis (FEA) simulations during the inference stage to maintain physical consistency and reduce error drift, particularly in capturing steep gradients. The FEA-PINN framework effectively handles dynamic phase changes, temperature-dependent material properties, and various convection effects, achieving accuracy comparable to traditional FEA methods but with significantly reduced computational costs. AI
IMPACT Accelerates simulation of complex material processes, potentially reducing computational costs for additive manufacturing.