Research On Adaptive Fault-Tolerant Control For Nonlinear Systems Based On Neural Network

With the development of science and technology,the scale and complexity of engi-neering control systems are increasing.The occurrence of any kind of failure may lead to a decline in overall system performance and even affect the system stability,which could result in some unpredictable losses.Therefore,it is particularly important to im-prove the safety and reliability of control systems.The emergence and development of the fault-tolerant control(FIC)provide an effective way to solve this problem.Consid-ering the fact that the practical engineering systems are almost nonlinear systems,it is very meaningful to study the fault-tolerant control problem of nonlinear systems.Due to the complexity of the nonlinear system,the control theory for the general nonlinear system is not perfect,and the fault-tolerant control methods for nonlinear systems are not so rich.Moreover,most of the existing results mainly investigate the stability of faulty systems,and on this basis the problem of the system performance optimization is rarely considered.In recent years,the control methods for nonlinear systems based on the neu-ral network approximation have attracted a widespread attention.The introduction of the neural network learning method has greatly promoted the development of the nonlinear system control theory.However,most of the existing control methods for nonlinear sys-tems are applicable to systems with specific structures,and there is almost such a limit for the researches on the fault tolerant control problem for nonlinear systems.On the basis of the previous work,this dissertation is concerned on the fault-tolerant control problem for nonlinear systems using the adaptive dynamic programming method and neural network learning method,where different kinds of faults,performance in-dexes,state constraints and the event-triggered controller design are considered.Based on the Lyapunov stability theory,the fault compensation control algorithms are designed to achieve the adaptive fault-tolerant control for nonlinear systems.The theoretical proofs are given for the main results.Moreover,simulation experiments are carried out on some practical system models,such as an inverted pendulum system,a robot manipulator model and a rocket fairing structural acoustic model.The simulation results demonstrate the ef-fectiveness of the proposed methods.This dissertation is divided into eight chapters.The main contents of each chapter are given as follows:Chapters 1-2 systematically analyze and summarize the background and develop-ment of the fault-tolerant control.Preliminaries and research methods about the consid-ered problem are also given.Chapter 3 investigates the problem of the complementary control design for affine nonlinear control systems with actuator faults.The objective is to design one complemen-tary control policy to improve the performance of a general nonlinear system with faults.Firstly,a novel adaptive scheme is developed to estimate the fault parameters.Then,by using the fault estimations to reconstruct the faulty system dynamics and introducing a cost function as the optimization objective,a nearly optimal complementary control pol-icy can be learnt based on the adaptive dynamic programming method.Unlike most of the existing methods,new adaptive weight update laws are derived to guarantee the con-vergence of neural network weights without the addition of a probing signal.Finally,two simulation examples are given to illustrate the effectiveness of the proposed method.Based on the results in Chapter 3,Chapter 4 investigates the fault-tolerant control problem for affine nonlinear systems with time-varying actuator gain and bias faults.In order to handle the actuator faults and guarantee the approximate optimal performance of the nominal nonlinear dynamics,the adaptive dynamic programming method is employed to design a sliding mode fault-tolerant control policy.Firstly,the actuator faults are es-timated using an adaptive observer technique.Based on the fault estimations,a sliding mode control policy is constructed,which can guarantee the reachability condition and the ideal sliding mode motions.Then,based on the actor-critic control structure,new weight tuning laws are given to learn the nearly optimal control policy for the sliding mode dynamics.Finally,the simulation results are given to verify the effectiveness of the developed method.In Chapter 5,the problem of fault-tolerant control for a class of affine nonlinear sys-tems subject to state constraints is investigated.By modifying an existing transformation technique,the nonlinear optimal control problem with state constraints is converted into an unconstrained one,which facilitates solving the resulted optimal control problem based on the adaptive dynamic programming problem.In order to achieve the fault-tolerant control objective,an NN observer with novel adaptive laws is constructed to identify the actuator faults.An integral sliding mode fault-tolerant control scheme is employed to guarantee the stability of the constrained faulty system.Finally,simulations are presented to illustrate the effectiveness of the proposed method.Chapter 6 investigates the problem of the adaptive fault-tolerant tracking control for a class of nonlinear systems with time-varying tracking error constraints based on the actor-critic control structure.To handle the error constraint problem,the prescribed per-formance method is introduced to transform the error system with constraints to an uncon-strained augmented error system.Then,specific critic functions are designed to supervise the tracking performance and tune the controller adaptively so that the effect of the neu-ral network reconstruction errors,input quantization and actuator faults can be reduced.Finally,the boundedness of the closed-loop signals is proved based on the Lyapunov sta-bility theory.Moreover,the proposed method is applied to an inverted pendulum system,and the effectiveness and superiority are illustrated by comparing the existing methods.Chapter 7 presents a novel event-triggered fault-tolerant control approach for a class of uncertain nonlinear systems.Without using the fault detection and diagnosis apparatus,the purpose of the fault-tolerant control is achieved based on the neural network learning method.By introducing the exponential time-varying gain,one specific event-triggering condition is designed,which is expected to reduce the frequency of the event triggered further on the basis of the existing results.At the same time,one event-based reference dynamics is constructed to improve the neural network learning mechanism in an exist-ing result.Under the premise of ensuring the stability of the system,the new weight tuning law can help to speed up the neural network learning process.In addition,it is proved that there is not the Zeno phenomenon based on the proposed method.Finally,the comparative simulation experiments are carried out on the rocket fairing model and the robot manipulator model,and the simulation results show the effectiveness and better performance of the proposed scheme.Finally,the results of the dissertation are summarized and further research topics are pointed out.