Title :
Electromagnetic field inversion using the Quickprop method
Author :
Rekanos, Ioannis T. ; Tsiboukis, Theodoros D.
Author_Institution :
Dept. of Electr. & Comput. Eng., Aristotelian Univ. of Thessaloniki, Greece
fDate :
3/1/1997 12:00:00 AM
Abstract :
The Quickprop method for function minimization that was initially introduced to the training process of multilayer perceptrons is applied to inversion problems in electromagnetism. This method is a variation of the general gradient methods and is combined with the finite elements analysis in order to estimate the inner geometric structure of inaccessible regions via a set of electromagnetic field or potential measurements. An application to the estimation of geometrical parameters of a rectangular cross section conductor imbedded in a conductive half space region is presented. Finally the results are compared to the steepest descent method
Keywords :
backpropagation; electrical engineering; electrical engineering computing; electromagnetic fields; finite element analysis; inverse problems; multilayer perceptrons; parameter estimation; Quickprop method; conductive half space region; electromagnetic field inversion; electromagnetic field measurements; electromagnetism; finite elements analysis; function minimization; general gradient methods; geometrical parameter estimation; inner geometric structure; multilayer perceptrons; potential measurements; rectangular cross section conductor; steepest descent method; training process; Arm; Convergence; Electromagnetic fields; Geometry; Minimization methods; Multi-layer neural network; Multilayer perceptrons; Neural networks; Quadratic programming; Shape;
Journal_Title :
Magnetics, IEEE Transactions on
