Neuro-dynamic programming-based event-triggered fault tolerant control for nonlinear systems with multiple faults

Abstract

Existing neuro-dynamic programming (NDP)-based fault-tolerant control (FTC) methods typically focus exclusively on actuator faults while neglecting sensor faults, and their online implementation is constrained by the strict persistence of excitation (PE) condition and the initial admissible control. This paper presents an online FTC scheme for uncertain nonlinear systems characterized by multiple faults. By integrating two neural networks (NNs) within a neuro-observer, the proposed approach simultaneously reconstructs accurate system states and estimates both actuator and sensor faults. Based on the neuro-observer, a critic NN is built to derive the event-triggered control (ETC) policy indirectly. Then, the NDP-based event-triggered FTC strategy is derived by combining the NDP-based ETC and the actuator fault compensator with significantly reducing computational resource consumption. Meanwhile, an additional stabilizing term and the experience replay technique are introduced to relax the stringent PE and initial control conditions, which enables the online application of our proposed control scheme. The observer errors, fault estimation errors, and the closed-loop system are all shown to be uniformly ultimate boundedness by employing Lyapunov’s direct method. Finally, a simulation example is provided to validate the proposed method.