On Adaptive Fault-tolerant Control Of Output Feedback Systems With Actuator Failures

The improvement of the modern theory of science and technology as well as the level of industrial production prompt control theory and control engineering achieve unprecedented widely used in professional fields such as complex systems modeling, telecommunication technology, military industry, aerospace, agriculture and transportation, economics and statistics. Moreover, it has achieved plenteous research results and remarkable scientific accomplishments in recent years. On the one hand, the application of sophisticated control instruments enables systems to complete increasing complicated control objectives and meet rigorous requirements of various control performances proposed by people. However, on the other hand, it also makes the probability of failures occurrence increasing greatly. Further, major safty accidents and economic loss will be happened if the failures cannot be removed or restrained for some time. Hence, how to strengthen the security and reliability of a dynamical system to make the stability of it can not be influenced by the failures has become an increasingly urgent problem.This paper proposes several fault-tolerant control schemes for some classes of nonlinear systems with actuator failures which are combined by backstepping control, dynamic surface control (DSC), adaptive control, neural network control. It demonstrates the control strategies proposed can ensure tracking errors of all systems are semi-globally uniformly bounded by choosing appropriate Lyapunov functions. The main contributions of this paper are as follows:Firstly, an adaptive output feedback DSC scheme is proposed for a class of multi-input single-output (MISO) nonlinear systems with unmodeled dynamics. In the situations that the external disturbances and system states are unmeasured, by introducing a mersured dynamic signal to overcome the unmodeled dynamics combined with the function separation theorem, using K-filters to estimate unmeasured states at the same time. Moreover, it uses neural networks to approximate the unknown black box continuous function. Compared with the existing research results, removing the premise assumption that the upper and lower bounds of the unknown constant gains should be known, relaxing the assumed condition that the smooth nonlinear functions in state equations are given, using the estimate to the module value of the unknown ideal weight vector for the neural networks instead of estimate it directly. Decreasing the number of the parameters to be estimated and reducing the complexity of the design.Secondly, an adaptive DSC passive fault-tolerant scheme is put forward for a class of output feedback nonlinear systems with unmodeled dynamics and actuator failures. Under the conditions that the fault actuators and time instants are unknown as well as the states are unmeasured, the tracking control problem of such systems have been solved by constructing the virtual control laws and parameter adaptive laws. Moreover, the effectiveness of the proposed control strategy can be demonstrated by choosing appropriate Lyapunov function and the tracking error can converge to a small neighborhood of the origin. Compared with the existing...