Memory efficient indexing algorithm for physical properties in OpenMC

[摘要] OpenMC is an open source Monte Carlo code designed at MIT with a focus on parallel scalability for large nuclear reactor simulations. The target problem for OpenMC is a full core high-fidelity multi-physics coupled simulation. This encompasses not only nuclear physics, but also material science and thermohydraulics. One of the challenges associated with this problem is efficient data management, as the memory required for tallies alone can easily enter the Terabyte range. This thesis presents an efficient system for data storage which allows for physical properties of materials to be indexed without any constraints on the geometry. To demonstrate its functionality, a sample depletion calculation with 4 isotopes is completed on the BEAVRS benchmark geometry. Additionally, a temperature distribution assembly layout is presented.

[发布机构] Massachusetts Institute of Technology