Saxe-Coburg Publications
Computational Technology Publications
|
|
COMPUTATIONAL METHODS FOR ACOUSTICS PROBLEMS
Edited by: F. Magoulès
Chapter 10
Inverse Acoustic Problems R. Djellouli
Department of Mathematics, California State University, Northbridge CA, United States of America Keywords: acoustic scattering, inverse obstacle problem, ill-posed problem, Newton
method, Tikhonov regularization, Fréchet derivative, sensitivity analysis, finite element
method, domain decomposition method.
We report on the performance of a regularized Newton solution methodology for retrieving the shape of an impenetrable three-dimensional obstacle from the intensity measurements of its corresponding acoustic far-field pattern. The main features of this optimization procedure are: (a) a sensitivity-based and frequency-aware multi-stage solution strategy, (b) a computationally efficient usage of the exact sensitivities of the far-field pattern to the specified shape parameters, and (c) a numerically scalable domain decomposition method for the fast solution in a frequency band of three-dimensional direct acoustic scattering problems. Numerical results obtained in the case of three-dimensional inverse mockup submarine problems are presented to illustrate the salient features of this computational methodology and highlight its performance characteristics. |
|