Detection, isolation and fault estimation of nonlinear systems using a directional study

In terms of system diagnosis, several studies are generally performed. The diagnosis is composed of three different parts: detecting, isolating and estimating the value of the faults. If many results have been obtained for linear systems with a known model, the situation is quite different in the case of nonlinear systems behavior, especially when the model is not known a priori. This paper proposes to discuss the latter case using a study of the dissimilarities between data. The dissimilarities are evaluated by a nonlinear function of the Euclidean distances. To this end, a radial basis function is used, and a directional study is introduced for fault diagnosis. The relevance of the proposed technique is illustrated on simulated data.