DocumentCode :
2516987
Title :
A Strategy for Parametrization Refinement in the Solution of a Geometric Inverse Problem
Author :
Peters, Franciane Conceição ; da Silva Barra, Luis Paulo
Author_Institution :
Grad. Program on Comput. Modeling, Fed. Univ. of Juiz de Fora, Juiz de Fora, Brazil
fYear :
2009
fDate :
23-25 Nov. 2009
Firstpage :
136
Lastpage :
142
Abstract :
This paper treats the problem of identifying a single inclusion in a conductor domain based on the knowledge of measured electrical potentials on the external boundary of the conductor body due to known injected electrical currents. The inclusion boundary is approximated by an Extended X-Spline, that allows identify inclusions with smooth or sharp boundary. In this work, the forward problem is solved by an implementation of the direct formulation of the Boundary Element Method (BEM) and the inverse one is solved by Levenberg-Marquardt method, which requires the evaluation of the objective function derivatives, here approximated by finite differences. This work presents a methodology to increase the number of optimization variables during the solution of the inverse problem in order to improve the quality of the results.
Keywords :
conductors (electric); electric current measurement; finite difference methods; inclusions; splines (mathematics); voltage measurement; Levenberg-Marquardt method; boundary element method; electrical potential; extended X-spline; finite difference approximation; geometric inverse problem; inclusion boundary; injected electrical current; objective function derivative; parametrization refinement; single inclusion; Current measurement; Electric potential; Electrodes; Geometry; Minimization; Optimization; Shape; Boundary Element Method; Electrical Impedance Tomography; Levenberg-Marquardt Method; inverse problem;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computational Modeling (MCSUL), 2009 Third Southern Conference on
Conference_Location :
Rio Grande
Print_ISBN :
978-1-4244-5980-3
Type :
conf
DOI :
10.1109/MCSUL.2009.26
Filename :
5597974
Link To Document :
بازگشت