Dynamic mode decomposition coupled with multilevel octree grid algorithm for large-scale discrete ordinates neutron transport calculations

Large-scale discrete ordinates neutron transport calculations are known to be excessively slow due to the large number of grids that need to be solved and the limitations of the source iteration method employed. This paper introduces a novel algorithm called DMD-MLTG as a practical means for solving this problem. The DMD-MLTG algorithm utilizes the dynamic mode decomposition (DMD) method to generate low-order approximation operators from historical iteration calculations, introducing only a minimal amount of additional processing in the multilevel octree grid (MLTG) algorithm. Numerical results from a self-designed one-group problem and crooked pipe problem demonstrate that the DMD-MLTG algorithm shows superior performance and robustness in challenging situations. The HBR-2 benchmark results confirm that the DMD-MLTG algorithm outperforms the MLTG algorithm by reducing computation time by 50%. The DMD-MLTG algorithm offers a promising solution to the issue of slow large-scale discrete ordinates neutron transport calculations, making it a valuable tool for practical shielding problems.