Using Reduced Support to Enhance Parallel Strong Scalability in 3D Finite-Element Magnetic Vector Potential Formulations with Circuit Equations

被引:0
|
作者
Takala, Eelis [1 ]
Yurtesen, Evren [2 ]
Westerholm, Jan [2 ]
Ruokolainen, Juha [3 ]
Peussa, Tommi [1 ]
机构
[1] Trafotek Oy, Kaarinantie 700, FL-20540 Turku, Finland
[2] Abo Akad Univ, Fac Sci & Engn, Turku, Finland
[3] CSC IT Ctr Sci, Espoo, Finland
关键词
Finite-element method; inductor; magnetic vector potential; reduced support; strong scaling;
D O I
10.1080/02726343.2016.1187107
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Quasi-static electromagnetic problems involving, e.g., inductors, are often solved numerically using the finite-element method with magnetic vector and electric scalar potentials. Coupling the inductors to external circuits may however, lead, to large matrix equations that could become bottlenecks in parallel computation systems, particularly for strong scaling when more processors are introduced to scale down the total computation time. It is argued and shown by numerical simulations that the use of reduced support in the finite-element model improves the strong scalability of multiprocessor simulations due to the reduced communication between the global constraint owner processes and the finite-element equation owner processes.
引用
收藏
页码:400 / 408
页数:9
相关论文
共 50 条