Fault detection with little knowledge of system model

Observers-based analytical redundancy is a widely used technique for fault detection. There have been many studies on observer-based fault detection, such as using unknown input observer, disturbance observer and generalized observer. Most of them require a nominal mathematical model of the system. This paper presents a new fault detection technique using improved extended state observer (ESO). Based on the input and output data, the ESO specifically identifies the un-modeled dynamics, a feat no other disturbance observer has ever attempted, and it provides vital information for fault diagnosis with only partial information of the plant, which cannot be easily accomplished with any existing methods. Another advantage of ESO is its simplicity in tuning only a single parameter. This parameter is essentially the observer bandwidth, which is a trade-off between the observer´s tracking speed and the observer´s sensitivity to measurement noise. It is shown in this paper that faults can be detected with little knowledge about the system model. A strongly coupled three-tank nonlinear dynamic system is chosen as a case study. The simulation results indicate that the presented ESO-based techniques effectively detected multiple faults.