Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : October 25, 2023 - October 27, 2023
A solidification during an additive manufacturing (AM) process is occurred under a highly non-equilibrium condition such as high cooling rate and large thermal gradient. In this study, we use the non-equilibrium multi-phase-field (NEMPF) model coupled with the CALPHAD-based thermodynamic database to simulate the solidification behavior in a SUS316L stainless steel during AM. Although the NEMPF model is capable of describing the microstructure evolution and the partitioning of solute atoms at the solid-liquid interface under the non-equilibrium condition, its computational cost is high. Therefore, we reduce the computational cost of the NEMPF simulation by replacing the CALPHAD-based thermodynamic calculation with the machine learning method where the trained neural network estimates the chemical free energy and chemical potential as a function of temperature and chemical composition at the solid-liquid interface. In this paper, we show the rapid solidification process in SUS316L stainless steel simulated by the NEMPF coupled with the trained neural network. The results show that the dendrites without secondary arms form and the solute atoms segregate at the interfacial region.
