A unified model-based fault diagnosis scheme for non-linear discrete-time systems with additive and multiplicative faults

Abstract

In this paper, a unified model-based fault diagnosis (MFD) scheme that deals with both multiplicative actuator and additive system faults is designed. For a class of uncertain non-linear discrete-time systems, this MFD scheme is capable of not only detecting both additive and multiplicative actuator faults but also identifying the fault type. Faults are detected by using a novel fault detection observer consisting of two online approximators in discrete-time (OLAD) and a robust adaptive term. Upon detection, a fault diagnosis scheme is introduced to determine the fault type by monitoring the input residual generated via the first OLAD output. Upon performing the diagnosis online, the appropriate OLAD is activated in the observer and the other OLAD is switched off. Thereafter, by using both the parameter update law of the active OLAD and user-selected failure threshold, an online time-to-failure scheme is introduced. In the case of multiplicative faults, boundedness of the detection residual and parameter estimation errors is shown while in the case of additive faults, the asymptotic convergence of the detection residual and parameter estimation errors is guaranteed due to the robust adaptive term. Finally, a simulation example is used to demonstrate the proposed fault diagnosis scheme.