Process fault prognosis using a fuzzy-adaptive unscented Kalman predictor

摘要

By monitoring the future process status via information prediction, process fault prognosis is able to give an early alarm and therefore prevent faults, when the faults are still in their early stages. A fuzzy-adaptive unscented Kalman filter (FAUKF)-based predictor is proposed to improve the tracking and forecasting capability for process fault prognosis. The predictor combines the strong tracking concept and fuzzy logic idea. Similar to the standard adaptive unscented Kalman filter (AUKF) that employs an adaptive parameter to correct the estimation error covariance, a Takagi-Sugeno fuzzy logic system is designed to provide a better adaptive parameter for smoothing this regulation. Compared with the standard AUKF, the proposed FAUKF has the same strong tracking ability but does not suffer from the drawback of serious tracking fluctuation. Two simulation examples demonstrate the effectiveness of the proposed predictor.