Fault detection of nonlinear systems with missing measurements and censored data

In this paper, we present an extended Tobit Kalman filter that deals with fault detection problem in nonlinear systems with missing measurements and censored data. The missing measurements randomly occurring are regulated by individual random variables whose probability distributions are on the interval [0,1]. The censored data are characterized by the Tobit measurement model. The Tobit Kalman filter designed for linear systems with missing measurements is extended to nonlinear systems. The residual signals are generated by the extended Tobit Kalman filter and then evaluated to detect the occurrence of a fault. Finally, the feasibility of the proposed fault detection filter is illustrated by an example of leakage fault of a three-tank system.