Fault prediction of the nonlinear systems with uncertainty

Fault prediction which can forecast the fault in advance to avoid large calamity has attracted more and more attention. However, the current filter based fault prediction methods for the nonlinear systems are all based on the framework of the probability theory, and cannot realize fault prediction of the nonlinear systems with fuzzy uncertainty. Based on the extended fuzzy Kalman filter (EFKF) and the extended orthogonality principle, an improved fuzzy Kalman filter (IFKF) is firstly proposed to estimate the system states or the parameters in this paper. Then, according to the IFKF, a multi-step improved fuzzy Kalman predictor (MIFKP), which can be considered as an adaptive predictor, is obtained. Once the characteristic parameter is chosen, the MIFKP can be used to implement the multi-step fault prediction. Simulation results demonstrate that the proposed approach has the better prediction ability and stronger robustness than the traditional multi-step extended fuzzy Kalman predictor (MEFKP).