Guaranteed method for the estimation of dielectric relaxation model parameters

In this paper, a guaranteed method for the estimation of dielectric relaxation model parameters is presented. The only assumption used is that data uncertainty is bounded. The main difference with classical methods based on least squares minimisation is that the solution is not a point but leads to a set. When least squares method are used, the model structure should be known. In some cases, this information is not available (for instance because there is not enough data); in this case the identification process can lead to unacceptable results. In this paper, a new technique, based on interval analysis, allowing rejection of models that are not consistent with data is presented.