A Parameterized Data-Driven Neural Network with Physics-Based Regularization for Thermal Field Prediction in Metal Additive Manufacturing

<title>Abstract</title> <p>Laser Powder Bed Fusion (L-PBF) is a metal additive manufacturing process characterized by highly transient and localized thermal behavior that strongly influences melt pool evolution and the resulting material properties. Accurate prediction of the spatiotemporal temperature field is therefore essential for process optimization, but conventional numerical simulations remain computationally expensive for large parametric studies. In this work, a parameterized data-driven neural network with physics-based regularization is developed for predicting transient thermal fields in SS316L across varying laser power and scan speed conditions. The framework incorporates process parameters directly into the input space together with spatiotemporal coordinates, enabling continuous thermal prediction within the bounded process domain. High-fidelity finite element method (FEM) thermal simulations provide the primary supervision, while the transient heat conduction equation is incorporated as a physics-based regular- ization term during training. The proposed framework achieves an RMSE of 9.01 K while accurately predicting thermal fields and melt pool geometry.</p>