Electromagnetic field inversion based on generalised radial basis functions

The application of the generalised radial basis functions neural networks to the solution of inverse problems in electromagnetism is investigated. A training strategy for the implementation of such networks is presented. The attractive property of these networks to work as universal approximators and the simplicity of their training process is examined in two application problems. In the first problem the evaluation of the conductivity profile of a layered planar structure is performed after the inversion of the impedance of a circular air cored probe coil, of rectangular cross section, which is placed above the considered structure. The second problem involves the introduction of the generalised radial basis functions to the estimation of geometrical parameters of inaccessible structures via a set of field or potential measurements. The results obtained from these applications illustrate the efficiency and accuracy of the estimation and generalisation abilities of the generalised radial basis functions networks.