Chapter 9 - Neural network-based finite-time fault-tolerant attitude control

Abstract

This chapter investigates the finite-time fault-tolerant attitude control for rigid–flexible coupling satellites in large-angle rapid maneuver when some of the actuators are in the fault condition. Based on the first-order approximate method, a more accurate dynamic model for rigid–flexible coupling satellites is set up. Then, Legendre polynomial-based neural network (NN) is introduced to estimate the lump perturbation including the estimation error of the fault control torque, high-order flexible coupling terms, external disturbances, and model uncertainties. With the estimation value as the compensation effort, a finite time fault-tolerant attitude controller is designed based on the nonsingular fast terminal sliding mode. And in order to guarantee that the output of the Legendre polynomial-based NN stays inside the bound of the lump perturbation, a switch mechanism is introduced to generate a switching between the proposed fault-tolerant attitude controller and a robust controller. The proposed fault-tolerant attitude controller is shown to have the finite time stability, with fast convergence rate, high accuracy, disturbance rejection, chattering attenuation, flexible vibration damping through theoretical analysis and simulations, meanwhile the better representation capability of the Legendre polynomial-based NN, whose basic functions are implemented using only the desired attitude, makes the controller design simple and efficient.