A generalized coarse mesh finite difference acceleration for 3D method of characteristics in hexagonal fast reactor simulations

Numerical simulation has become essential in nuclear reactor design and safety verification due to the high cost and complexity of physical experiments. The Method of Characteristics (MOC) provides high-fidelity neutron transport solutions with inherent parallelism and geometric flexibility. However, conventional MOC implementations face challenges in memory usage, computational efficiency, and convergence speed, particularly for full-core simulations of fast reactors with complex hexagonal assemblies. To address these challenges, this work extends the self-developed 3D neutron transport solver ANT-MOC by implementing a hexagonally track generation module that reduces track counts and memory demands while improving accuracy. The developed generalized coarse mesh finite difference (GCMFD) method, compared with CMFD, naturally supports unstructured hexagonal/pentagonal meshes and enables efficient mesh indexing, adjacency management, and equivalent width calculations in such geometries. In addition, an energy group coarsening acceleration framework is introduced to alleviate the computational burden caused by the fine energy discretization typical of fast reactors. These features make the method particularly suitable for fast reactor simulations with complex geometries and wide energy spectra. Validation on the China Experimental Fast Reactor (CEFR) full-core benchmark shows the framework achieves an $k_{eff}$ error of 18.4 pcm and fission rate and scalar flux errors below 0.3%. The iteration count decreases from 411 to 144, significantly enhancing convergence efficiency. These results demonstrate a high-precision and efficient neutron transport simulation capability for fast reactor cores, with strong potential for engineering applications.