Physics-embedded graph network for accelerating phase-field simulation of microstructure evolution in additive manufacturing

Abstract The phase-field (PF) method is a physics-based computational approach for simulating interfacial morphology. It has been used to model powder melting, rapid solidification, and grain structure evolution in metal additive manufacturing (AM). However, traditional direct numerical simulation (DNS) of the PF computationally expensive due sufficiently small mesh size. Here, physics-embedded graph network (PEGN) proposed leverage an elegant representation embed classic theory into network. By reformulating problem as unsupervised machine learning task on network, PEGN efficiently solves temperature field, liquid/solid phase fraction, orientation variables minimize loss/energy function. at least 50 times faster than DNS both CPU GPU implementation while still capturing key physical features. Hence, allows simulate large-scale multi-layer multi-track AM build effectively.