面向大规模并行的全隐式托卡马克MHD数值模拟

Large-scale parallel simulation of MHD in Tokamak based on a fully implicit method

提出了一种基于环形托卡马克模型的MHD (magneto hydrodynamic)方程并行求解算法,开展了等离子体的非理想MHD不稳定性及其演化过程的数值模拟,采用了全隐式的离散方法,并运用Newton-Krylov方法求解非线性系统。相对传统算法,文章提出的基于全隐格式的求解算法具有时间步长限制小、并行可扩展性高的优势,对于大规模的并行计算具有良好的适应性。根据在超级计算平台上的并行测试结果,文章开发的求解器在大规模并行中具有良好的并行效率与并行可扩展性,与传统求解算法结果具有良好的数值一致性,可适用于大规模的并行磁流体仿真计算。

In this paper,a parallel solving algorithm for the MHD equations based on the toroidal tokamak model is proposed,and the algorithm focuses on the numerical simulation of the nonideal MHD instability and its development process.In the proposed algorithm,we selected a fully-implicit scheme in the discretization,which has less limited time steps than the classic methods,and Newton-Krylov method for solving the nonlinear systems.Furthermore,according to the scalability test on the HPC platform,the solver based on the proposed algorithm has high parallel efficiency in the large-scale parallel computation,and the computational result has a good numerical consistency with the classic solvers.Therefore,the proposed algorithm has excellent adaptability for large-scale numerical simulation of MHD in the Tokamak plasma.