Accelerating Poisson solvers in front tracking method using parallel direct methods

Parallel direct computation is implemented in multiphase flow simulation of bubbles in a curved duct using a front tracking method to improve the computational cost. To solve the density Poisson equation, three methods including the SOR, FGMRES and PARDISO are tested using two strategies of full domain and masked bubble. The comparison of numerical simulations at the same conditions indicates that the PARDISO scheme under the masked bubble strategy considerably reduces the computational time and RAM usage, which enables simulations on very fine grid resolutions.