Journal of Xidian University ›› 2021, Vol. 48 ›› Issue (6): 144-150.doi: 10.19665/j.issn1001-2400.2021.06.018

• Computer Science and Technology • Previous Articles     Next Articles

Algorithm for half-space MLFMA domain decomposition utilizing an octree

ZHAI Chang(),LIN Zhongchao(),ZHAO Xunwang(),ZHANG Yu()   

  1. Shaanxi Key Laboratory of Large Scale Electromagnetic Computing,Xidian University,Xi’an 710071,China
  • Received:2020-07-17 Online:2021-12-20 Published:2022-02-24
  • Contact: Xunwang ZHAO E-mail:jione92@163.com;zclin@xidian.edu.cn;xwzhao@mail.xidian.edu.cn;yuzhang@mail.xidian.edu.cn

Abstract:

In order to quickly and accurately analyze the electromagnetic scattering of electrical large objects in half space under the condition of limited resources,a parallel half-space multi level fast multipole algorithm (MLFMA) with the domain decomposition method utilizing an octree is proposed.By using the octree structure formed by the MLFMA,the unknowns are grouped adaptively to realize the domain decomposition,thereby avoiding the creation of artificial interface between domains and reducing the workload of model processing.To ensure the current continuity between domains,the 1/4 impedance on the boundary of the domain is rigorously calculated,with the results more accurate.To deal with the half-space environment,the complex image source is introduced to calculate the near interaction,and the real image source is introduced to calculate the far interaction.A comparison of numerical results from the proposed algorithm and the commercial software FEKO is given,which proves the reliability and accuracy of the algorithm.An out-of-core algorithm is used to store data such as translators in the hard disk,which can significantly reduce the memory consumption.Finally,a numerical example including a 1000-wavelength ship model in half space is given,which demonstrates that the proposed algorithm can simulate electrical large objects under the condition of limited resources.

Key words: domain decomposition method, parallel algorithm, multi level fast multipole algorithm, half-space problem, octree strategy, out-of-core algorithm

CLC Number: 

  • TN820