Back to Search
Start Over
Massively Parallel Simulation of Large-Scale Electromagnetic Problems Using One High-Performance Computing Scheme and Domain Decomposition Method
- Source :
- IEEE Transactions on Electromagnetic Compatibility. 59:1523-1531
- Publication Year :
- 2017
- Publisher :
- Institute of Electrical and Electronics Engineers (IEEE), 2017.
-
Abstract
- Massively parallel simulation of large-scale electromagnetic problems is performed on a supercomputer using an in-house developed high-performance computing scheme together with a domain decomposition method, where parallel adaptive unstructured mesh generation is at first performed for modeling an arbitrary multiscale structure. The accuracy, speedup, as well as scalability of such parallel scheme are numerically tested and validated during its implementation for simulating electromagnetic environmental effects (E3) on a ship platform using the TianHe-II supercomputer. It is demonstrated that this method can overcome the convergence deficiency of conventional iterative solver. In addition, numerical experiments are also carried out to show its adaptability to the number of subdomains and object electrical size. It is believed that our developed solver can be further used for solving various E3 or electromagnetic compatibility problems for the design of many large-scale platforms.
- Subjects :
- 010302 applied physics
Speedup
Electromagnetics
Computer science
Electromagnetic compatibility
020206 networking & telecommunications
Domain decomposition methods
02 engineering and technology
Parallel computing
Solver
Condensed Matter Physics
Supercomputer
01 natural sciences
Atomic and Molecular Physics, and Optics
Computational science
0103 physical sciences
Scalability
0202 electrical engineering, electronic engineering, information engineering
Electrical and Electronic Engineering
Massively parallel
Subjects
Details
- ISSN :
- 1558187X and 00189375
- Volume :
- 59
- Database :
- OpenAIRE
- Journal :
- IEEE Transactions on Electromagnetic Compatibility
- Accession number :
- edsair.doi...........95e8f2982f37dcea2ca0e081cb996219