Non-Equilibrium Phase Field Model Using Thermodynamics Data Estimated by Machine Learning for Additive Manufacturing Solidification

A multi-phase field method using finite interface dissipation model proposed by Steinbach et al. is applied to simulate solidification microstructure evolution of stainless steel composition in the non-equilibrium condition of high cooling rate and temperature gradient of additive manufacturing. The calculation is performed for quinary system in order to simulate solidification of engineering composition. Thermodynamic calculation using CALPHAD database in this multi-phase field method calculation is replaced by machine learning prediction procedure to reduce calculation time. The microstructure evaluated by using machine learning parameter is good agreement with one directly coupled with CALPHAD database. This calculation is approximately five times faster than the direct CALPHAD calculation method. Finally, it is confirmed that this multi-phase field method can be applicable to simulate non-equilibrium phase transformation of additive manufacturing condition with high numerical stabilization.