Simulation of 3-D Mold-Filling and Solidification Processes on Distributed Memory Parallel Architectures

Abstract This work presents industrial mold-filling applications of a three-dimensional stabilized finite element solver on distributed memory parallel architectures. The paper focuses on the solution algorithm and parallel implementation for complex multiphasics problems involving high Reynolds number flows with free surfaces, turbulence modeling and heat transfer. Standard domain decomposition methods (Chaco, Metis) are applied to the graph of nodes obtained from the finite element mesh, and a distributed-memory MPI programming model is used. An implicit time integration scheme and a segregated iterative algorithm are used to solve the momentum, energy, turbulence variables and front tracking equations. The equations are dis-cretized using a stabilized SUPG finite element method on linear elements. The resulting sparse system of linear equations is solved using parallel preconditioned iterative methods.