• DocumentCode
    878902
  • Title

    Interaction matrix localization (IML) permits solution of larger scattering problems

  • Author

    Canning, Francis X.

  • Author_Institution
    Rockwell Sci. Center, Thousand Oaks, CA, USA
  • Volume
    27
  • Issue
    5
  • fYear
    1991
  • fDate
    9/1/1991 12:00:00 AM
  • Firstpage
    4275
  • Lastpage
    4278
  • Abstract
    The interaction matrix localization (IML) replaces the usual N-by-N moment method matrix, with N2 nonzero elements, by a sparse N-by-N matrix with roughly 100 N nonzero elements. It gives a similar reduction in the number of operations required per iteration for iterative solution methods. While these savings are significant, additional reductions are possible by using the structure of the sparse matrix to develop preconditioners which allow a solution in fewer iterative steps. The method considered is preconditioning by incomplete LU (ILU) decomposition. A problem is solved with N=15000 using less memory than that required for N=1500 problems using standard methods. The number of operations necessary to iteratively solve for each incidence angle (i.e., the right hand side of the matrix problem) is approximately equal to the number of operations needed by standard methods to solve for each right hand side, even after the inverse of the moment method matrix has been calculated. Saving orders of magnitude in the storage needed and in the execution time needed extends the range of applicability of the moment method to significantly larger bodies.
  • Keywords
    computational complexity; electromagnetic wave scattering; integral equations; computational complexity; electromagnetic scattering; incidence angle; integral equations; interaction matrix localization; iterative solution methods; sparse matrix; Canning; Fast Fourier transforms; Integral equations; Iterative methods; Matrix decomposition; Message-oriented middleware; Moment methods; Scattering; Sparse matrices; Testing;
  • fLanguage
    English
  • Journal_Title
    Magnetics, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    0018-9464
  • Type

    jour

  • DOI
    10.1109/20.105046
  • Filename
    105046