Similar Fault Isolation of Discrete-Time Nonlinear Uncertain Systems Using Smallest Residual Principle
Date of Original Version
This paper investigates the problem of similar fault isolation (sFI) for discrete-time nonlinear uncertain systems. The main challenge lies in that the differences among the so-called "similar" faults could be very small and easily hidden in system uncertainties. To overcome such a challenge, in this paper, the uncertain fault-induced system dynamics is first accurately identified using radial basis function neural network (RBF NNs), where the obtained knowledge can be stored and represented by constant RBF NNs. With the obtained constant networks, a bank of novel fault residual systems are designed by using an absolute measurement of fault dynamics difference, which can effectively measure the match level of the occurred fault from each trained fault. Based on the designed residual systems, real-time fault isolation decision making is achieved according to the smallest residual principle (SRP), i.e., the occurred fault is identified similar to one trained fault when the related residual becomes the smallest one among all the others. Rigorous analysis of the isolatability condition is also given. Extensive simulations have been conducted to demonstrate the effectiveness and advantages of the proposed approach.
Proceedings of the American Control Conference
Zhang, Jingting, Chengzhi Yuan, and Paolo Stegagno. "Similar Fault Isolation of Discrete-Time Nonlinear Uncertain Systems Using Smallest Residual Principle." Proceedings of the American Control Conference 2020-July, (2020): 3176-3181. doi:10.23919/ACC45564.2020.9147429.