A neural network-based actuator gain Fault Detection and Isolation strategy for nonlinear systems

Abstract

The problem of robust fault detection and isolation (FDI) for nonlinear systems is investigated in this paper. The proposed FDI scheme employs a neural network-based observer for detecting and identifying the severity of actuator gain faults in the presence of disturbances and uncertainties in the model and sensory measurements. The neural network weights are updated based on a modified dynamic backpropagation scheme. The proposed FDI scheme does not rely on the availability of full state measurements. In most work in the literature, the fault function acts as an additive (bias) term, whereas in our scheme the fault function acts as a multiplicative (gain) term. This representation makes the stability analysis of the overall FDI scheme rather challenging. Stability properties of the proposed fault detection scheme in the presence of unknown actuator gain faults as well as plant and sensor uncertainties are demonstrated by using Lyapunov's direct method with no restrictive assumptions on the system and/or the FDI algorithm. The performance of our proposed FDI approach is evaluated through simulations performed on a reaction wheel type actuator that is commonly employed in the attitude control subsystem (ACS) of a satellite.