Nonlinear Adaptive Hybrid Compensation for Actuator-Sensor Compound Faults of Non-Gaussian Uncertain Systems

Abstract

This paper presents a hybrid fault-tolerant control (FTC) method to handle actuator-sensor compound faults in non-Gaussian uncertain stochastic systems with approximation errors. Considering that multi-source nonlinearities of a class of stochastic systems cannot be approximated by a unified method, fuzzy logic is used to linearize nonlinear parameters, and the Lipschitz condition helps to proof the stability of FTC systems with nonlinear sensing functions. Moreover, to handle compound faults, an adaptive hybrid fault compensation scheme is devised. When only a sensor fault occurs, the feedback contains the fault information, and a feedback error composite function in the controller direct compensates the fault passively. When multiple actuator faults and a sensor fault occur, an adaptive fusion observer simultaneously implements sensor fault masking and actuator fault estimation, and then an active-passive hybrid FTC algorithm uses a compensation function and fault estimation to perform both passive compensation of the sensor fault and active FTC of actuator faults. Furthermore, an adaptive algorithm that resembles the animal predation behavior makes the controller more sensitive to incipient fault deviations. Lyapunov functions prove the robust stability of the proposed fault tolerant systems with approximation errors, and simulation experiment verifies the performance compared to a state-of-the-art method.