A numerical approach for inlet-outlet boundary conditions with least-square moving particle explicit (LSMPE) method on GPU

In severe accidents of nuclear reactors, studying the pressure drop of debris beds is of great significance for delaying the progression of the accident. However, traditional MPS often exhibits significant pressure oscillation issues when dealing with the inlet-outlet boundary. The study introduces a method for managing inlet and outlet boundary conditions using the Least-Square moving particle explicit method (LSMPE). A high-precision least squares discrete method was applied to the moving particle explicit method (MPE) based on predictive-corrected pressure. The code was accelerated using OpenACC for parallelization. Special boundary treatments were applied to minimize errors induced by boundaries, including particle number density correction for inlet particles, addressing slip or nonslip boundaries, and introducing concept particles at the outlet. The stability and accuracy of this method were validated through three cases. Simulation results for Poiseuille flow and circular flow demonstrated the method's good precision, exhibiting stable pressure fields. Results from single-phase flow simulations through a porous medium bed agreed well with predictions from the Ergun equation, highlighting the method's potential for practical engineering applications.