A multilayer neural-network-based fault estimation and fault tolerant control scheme for uncertain system

Abstract

This work introduces a new actuator fault estimation approach coupled with a fault-tolerant control (FTC) strategy for uncertain systems in an output feedback framework. The proposed method involves constructing a Multi-Layer Neural Network (MLNN) observer-based fault estimation unit to accurately predict system states and potential faults in the actuator channel. An online control allocation (CA) scheme is then developed, utilizing the derived estimates to actively reconfigure the virtual control signals among the healthy redundant actuators in the event of actuator malfunction. Furthermore, an adaptive neural network-based output integral sliding mode control scheme is designed based on the virtual control. This integration enhances the overall system's robustness and significantly reduces the chattering effect. The stability analysis of the proposed fault estimations scheme is initially performed using MLNN structure, followed by a comprehensive closed-loop stability analysis to establish the stability of the entire system. Finally, the effectiveness of the proposed method is validated on a nonlinear six-degree-of-freedom model of multirotor unmanned aerial vehicle aircraft. Numerical simulations under different fault and failure scenarios validate the efficacy of the proposed method. The comparative analysis of the proposed scheme is conducted with the static output feedback control allocations and adaptive allocation strategy. This analysis focuses on evaluating performance using metrics such as root mean square error and mean square deviation, particularly in the presence of faults and failures. The results demonstrate the superior performance of the proposed scheme in fault/failure conditions.