Extensible Parallel Computing Method with Hundreds of Millions of Freedoms for Large-Scale Engineering Electromagnetic Field
Jin Liang1,2, Li Yuzeng1,2, Yang Qingxin1,2, Zhang Chuang1,2, Yan Shuai3
1. State Key Laboratory of Reliability and Intelligence of Electrical Equipment Hebei University of Technology Tianjin 300130 China; 2. Key Hebei Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability Hebei University of Technology Tianjin 300130 China; 3. Institute of Electrical Engineering Chinese Academy of Sciences Beijing 100081 China
Abstract：Accurate and rapid electromagnetic field calculation is the basis of fine simulation and optimization design of electrical equipment. In this paper, the scalable parallel computing research is carried out on the elastic cluster of high-speed interconnected high-performance cloud platform. OpenMpi is used as the message passing library, and the dual-primal finite element tearing and interconnecting (FETI-DP) method is selected as the domain decomposition algorithm. The parallel computing of eddy current magnetic vector potential A with constant conductivity is achieved by improving the master-slave/peer parallel program framework, which can reduce the complexity of programming and improve the efficiency of parallel computing. The program is written in C language and the scalable parallel computing method is verified by the benchmark problem of international TEAM Problem 7. In this paper, the master-slave/peer-to-peer parallel program framework and the dual-primal finite element tearing and interconnecting FETI-DP method are introduced into the field of electromagnetic computing, which improves the efficiency and scalability of parallel computing, and provides a new practical and theoretical method for large-scale engineering electromagnetic field calculation.
2022, Vol. 37 
