Fast and accurate prediction of temperature evolutions in additive manufacturing process using deep learning

Typical computer-based parameter optimization and uncertainty quantification of the additive manufacturing process usually requires significant computational cost for performing high-fidelity heat transfer finite element (FE) models with different settings. This work develops a simple surrogate model using feedforward neural network (FFNN) fast accurate prediction temperature evolutions melting pool sizes in metal bulk sample (3D horizontal layers) manufactured by DED process. Our is trained data obtained from FE model, which was validated experiments. The predicted FFNN exhibit accuracy $$99\%$$ $$98\%$$ , respectively, compared unseen settings studied range. Moreover, to evaluate importance input features explain achieved sensitivity analysis (SA) carried out SHapley Additive exPlanation (SHAP) method. SA shows that most critical enriched impacting predictive capability are vertical distance laser head position material point position.