AI modeling for high-fidelity heat transfer and thermal distortion forecast in metal additive manufacturing

In this study, a novel AI-based modeling approach is introduced to estimate high-fidelity heat transfer calculations and predict thermal distortion in metal additive manufacturing, specifically for the multi-laser powder bed fusion (ML-PBF) process. The effects of start position and printing orientation on deformation and stress distribution in parts produced using the ML-PBF additive manufacturing process are investigated. A total of 512 simulations are executed, and the maximum and minimum deformation values are recorded and compared. A significant reduction, e.g., 53% in deformation, is observed between the best and worst printing cases. A low-fidelity modeling framework, based on a feedforward neural network, is developed for the rapid prediction of thermal displacement with high accuracy. The model with unknown test cases demonstrates a strong positive correlation (R = 0.88) between the high-fidelity and network-predicted low-fidelity outputs. The simplicity, computational efficiency, and ease of use of the developed model make it a valuable tool for preliminary evaluation and optimization in the early stages of the design process. By adjusting controlling factors and identifying trends in thermal history, the model can be scaled to a high-fidelity model for increased accuracy, significantly reducing development time and cost. The findings of this study provide valuable insights for designers and engineers working in the field of additive manufacturing, offering a better understanding of deformation/thermal displacement control and optimization in the ML-PBF process.