Physics-Informed Graph Neural Operator for Melt-Pool Thermal Field Prediction and NSGA-II-Based Optimization in SS316L Directed Energy Deposition

This dataset supports a physics-informed framework combining a Graph Neural Operator (GNO) with NSGA-II for rapid prediction and optimization of melt-pool behavior in SS316L Directed Energy Deposition. High-fidelity thermal fields are generated using a nonlinear heat-transfer PDE and used to train a surrogate model that maps process parameters (laser power, scanning speed, thermal conductivity) to full temperature fields. The trained GNO enables fast and accurate estimation of melt-pool geometry and cooling rates, while integration with NSGA-II allows efficient identification of Pareto-optimal trade-offs between thermal and geometrical objectives. The proposed work provides a computationally efficient alternative to repeated simulations and supports process optimization and microstructure-aware design in additive manufacturing. This work was supported by the Czech Science Foundation (GAČR), project No. 24-11505S, and by the EU & Ministry of Education, Youth and Sports (MEYS), Czech Republic, under project CZ.02.01.01/00/22_011/0008569 "Czech Technical University- International postdoctoral programme".