Fast and Accurate Solutions of Very Large Electromagnetics Problems with the Parallel Multilevel Fast Multipole Algorithm

Özgür Ergül, Ph.D. Candidate, Bilkent University, Ankara

11 Dec 2008, 17:00; Location: S2|17-103

The multilevel fast multipole algorithm (MLFMA) is a powerful method for the fast and efficient solution of scattering and radiation problems in electromagnetics. Real-life problems formulated with surface integral equations can be solved iteratively, where the required matrix-vector multiplications are performed efficiently by MLFMA. However, many real-life problems require discretizations with tens of millions of unknowns, which may not be handled easily with the sequential implementations of MLFMA running on a single processor. In order to achieve the solution of such large problems, MLFMA is parallelized on relatively inexpensive computing platforms with distributed-memory architectures. Unfortunately, parallelization of this algorithm is not trivial, and it can be difficult to obtain a sufficient parallelization efficiency, especially when the number of processors is large and problems involve complex objects. Recently, we developed a hierarchical partitioning strategy, which provides a higher parallelization efficiency than previous approaches. The resulting parallel implementation has been successfully used to solve very large electromagnetics
problems involving more than 200 million unknowns.

This talk will provide the details of the sophisticated simulation environment, which involves integral-equation formulations, iterative solvers, preconditioners, parallel computing, and a highlyefficient parallelization of MLFMA, developed at Bilkent University Computational Electromagnetics Research Center (BiLCEM).

Category: CE Seminar


Technische Universität Darmstadt

Graduate School CE
Dolivostraße 15
D-64293 Darmstadt

Phone+49 6151/16-24401
Fax -24404

to assistants' office

Open BSc/MSc Theses

Show a list of open BSc/MSc topics at GSC CE.

 Print |  Impressum |  Sitemap |  Search |  Contact |  Privacy Policy
zum Seitenanfangzum Seitenanfang