GPU accelerated lattice Boltzmann method in neutron kinetics problems

摘要

The detailed neutron diffusion simulation is one of the kernels in nuclear reactor engineering, whose application, however, is limited by the high computational cost. In this report, a system of GPU accelerated neutron diffusion lattice Boltzmann method (LBM) is established to improve this condition, including steady-state and space-time kinetics problems. The neutron diffusion equation is solved by using the LBM due to its simplicity and strong parallelism. Both the single-GPU and multi-GPU acceleration techniques are applied to accelerate the neutron diffusion LBM calculations. In addition, to reduce the loop of time evolution, the space-time kinetics problem is solved by using the GPU accelerated neutron diffusion LBM with the predictor-corrector quasi-static method (PCQSM). To further clarify these procedures, the detailed implementations of these methods are outlined. To investigate the accuracy, flexibility, and efficiency of the proposed technique, three benchmark problems are simulated, including the steady-state, eigenvalue and space-time kinetics problems. The results show that the proposed technique can accurately and flexibly simulate the neutron kinetics problems and the GPU acceleration can effectively accelerate the computational speed. In addition, the computational speed can be further accelerated by the multi-GPU technique. This paper may form the basis for a powerful technique for the parallel simulation of large-scale engineering calculation and some alternative perspectives for solving the neutron kinetics problem. (C) 2019 Elsevier Ltd. All rights reserved.