Finite Element GPU Parallel Computing Method Using Node Preprocessing Matrix

Finite element analysis is one of the important methods for multiphysics calculation of electrical equipment. However, in the face of the increasing complexity of finite element models, the time cost of finite element analysis is getting higher and higher. Using GPU to design parallel structures for finite element programs can significantly improve the speed of finite element calculation. However, due to the characteristics of the finite element data structure, it is easy to encounter the problem of threads competition when designing a parallel structure. This paper solves the problem of threads competition by constructing a node preprocessing matrix to allocate operation areas for threads. The assembly of the overall stiffness matrix and the application of boundary conditions are realized on the GPU. Finally, the linear equation is solved by GPU parallel conjugate gradient method. Through numerical examples, the time required for serial and GPU parallel finite element solutions is compared, and it is verified that the method proposed in this paper can significantly improve the speed of finite element calculations.