Fault isolation and estimation in nonlinear systems: an adaptive filter solution based on multiple model approach

International audience; In this paper, a method based on an adaptive filter is proposed for fault diagnosis in non linear systems. The proposed method makes possible fault detection, isolation and estimation for the whole operating range of nonlinear system. Based on the assumption that the dynamic behaviour of nonlinear system is defined as a multiple model approach, a set of residual vector insensitive to faults is established and thus generating a global model (with probabilistically weighted matrices) exploited in an adaptive filter. The performances of the method are tested in simulation on an academic example