An inverse geometric problem: Position and shape identification of inclusions in a conductor domain

This work presents a methodology for identifying inclusions in a conductor domain. The methodology is based on electrical potential measurements on the external boundary of a conductor body subjected to a prescribed set of electrical current injections. The boundary of each inclusion is approximated by a special kind of spline, whose control points have an extra parameter related to the distance between the control point and the curve. Such special feature allows identification of smooth or sharp inclusions with the same initial guess. The identification is an inverse problem that, in this work, is solved by the Levenberg–Marquardt Method. This iterative method tries to locate the minimum of an objective function, the square of the norm of a residual vector function, given by the differences between electrical potential measurements and the computed ones. The computation of the electrical potential is called forward problem and it is solved by an implementation of the direct formulation of the Boundary Element Method (BEM). The present paper addresses two approaches for computing the derivatives of the residual function with respect to the minimization parameters, required by the Levenberg–Marquardt Method. The first one is based on finite differences and the second one is based on the direct differentiation of the integral equation of the BEM for potential problems. Performance comparisons of these two approaches are presented, based on numerical experiments of identification of inclusions with noisy datasets computationally obtained.