Assessment of observer based fault estimators for TS fuzzy models

Fault detection of nonlinear systems become more feasible when it is conducted over Takagi-Sugeno (TS) approximated fuzzy models. Proportional plus integral observer (PIO) and robust observer (RO) have already been developed for the estimation of the system states and actuator/sensor faults. In this paper, the algorithms are implemented for the detection of valve and level sensor faults of a two-tank system. Our simulation results indicate that both algorithms run well in estimating states and sensor fault, however, there is obvious differences in how they detect actuator fault in the presence of noise. From viewpoint of estimation variance, RO renders cleaner estimate of the fault than PIO, whiles PIO has faster fault tracking speed than RO. According to the achieving result, RO algorithm is recognized to be a more attractive in estimating actuator faults in noisy environments. The results are validated through simulations.