Robust Fault Diagnosis And Fault-Tolerant Control For Several Nonlinear Systems

The demands for plant reliability, safety and efficiency are higher since the faults onsensors, actuators and components occur inevitably with the running of complexitycontrol systems. Unfortunately, these demands are satisfied difficultly because of theexistence of nonlinearity in the practical plants and uncertainties of plant models.Therefore, the fault detection and diagnosis (FDD) and fault-tolerant control (FTC) foruncertain nonlinear systems has been led to greater attention and became one of theadvanced objectives in control field, and also is one of the objectives focused byinternational scientific association. In the thesis, the problems of robust FDD and FTCfor several types of uncertain nonlinear systems are studied by using observer approach,adaptive and neural network technique. The main research contents are given as follows:Based on sliding-mode/adaptive observers and neural network, the problems of robustFDD for two kinds of uncertain nonlinear systems are discussed, respectively. Under thecondition that the model of the plant faults are known, A global diffeo-morphismtechnique is used to make the original nonlinear system be transformed into two differentsubsystems with uncertainty. Due to the transformation, it can be shown that the proposedscheme includes two kinds of observers as sliding-mode/adaptive observers. Under thetwo observers, a decision logic algorithm for fault diagnosis corresponding to theconsidered fault cases, is proposed. The proposed algorithm avoids the requirements ofderivation to output and full rank to fault parameter matrixes. A simulation is given toillustrate the design procedures and the efficiency of the proposed algorithm. Further, inthe cases of complete and part measurability of states, a novel robust FDD approach thatis suited for abrupt and incipient faults, is presented by using adaptive neural networkobserver, respectively. When the states can be observed completely, a generalizeddynamic recurrent neural network (GDRNN) whose inputs is the system states is onlyutilized to approximate the fault part of the nonlinear systems, while the GDRNN whoseinputs is the estimation state of the observer is only utilized to approximate the fault partof the nonlinear systems when the states can't be observed completely. The robustness offault detection and stability of diagnosis systems are analyzed and proved, respectively.The efficiency of the proposed algorithms is illustrated by simulation results.We proposed an active robust FTC based on the adaptive estimation of the upperbounded of uncertainties for a class of uncertain nonlinear systems with structured andparameter dismatched uncertainties. A robust adaptive controller is designed firstly tocompensate for the effective of uncertainties by estimating the unknown parameters ofparameterizable part and the upper bound of unparameterizable part of uncertainties. Onthe basis of estimation information, a novel FDD scheme that has better robustness thanthe previous researches, is presented by using a fault detection estimator in which anonline approximator in the form of neural network is used. After a fault is detected, afault-tolerant controller is reconfigured and the stability of the closed-loop system isrigorously investigated. It is shown that the system signals remain uniformly ultimatelybounded under the proposed fault-tolerant controller. A simulation example was given toshow the effectiveness of the proposed scheme and the superiority compared with theprevious work.An approach to robust fault-tolerant control for two classes of uncertain nonlinearsystems, is proposed based on a novel parameter projection technique which handles theunknown uncertain parameters with a priori available upper and lower bounds ofpotential faults on actuators/system components such that the adaptive speed of theunknown fault parameters is rapid but not to go beyond the scale of uncertain bounds. Onthe basis of the fault information obtained by fault identification procedure, an onlineadaptive fault-tolerant control component is designed to enhance the trackingperformance. A simulation example associated to Van der Pol oscillator was given toshow the effectiveness of the proposed fault-tolerant tracking control methodology. Anovel integration algorithm of robust FDI and FTC in which includes two controllers thatare reconfigured by using the information of fault diagnosis and fault isolation,respectively, for a class of uncertain nonlinear systems whose states are measuredcompletely, is proposed. In the algorithm, the plant is supervised at first under the abovementioned robust adaptive controller;when a fault is detected but not isolated, areconfigured adaptive FTC, which is designed by using the fault information that isobtained by using neural network, is used to compensate the influence to the plantstability made by fault;once the fault is isolated by using multiple models fault isolationfilters in which a method of adaptive estimation based on the above parameter projectiontechnique is used, a reconfigured adaptive FTC based on the isolated fault information isdesigned again to enhance the control performance of the fault systems. To evaluate theperformance of the fault detection and isolation and fault-tolerant control scheme, aparticular example, the magnetic levitation system that is modeled by a simple nonlinearsystem, was chosen to illustrate the effectiveness.A novel robust FTC for a class of uncertain systems with nonlinear faults, which isbased on time-delay-optimized reconstruction model of the fault system, is presented.After a fault is detected, a time-delay-optimized reconstruction model is made, by usingB-splines function neural network and scroll-optimized delay method. At the same time,an adaptive FTC is designed aiming at the incipient faults. As the approximation error isstabilized in a given bounded, a LMI-based robust FTC controller is designed tocompensate the fault aiming at the incipient faults. Both the cases of the linearity anddiscrete are considered and the robust FTC algorithms were given, respectively. Todemonstrate the proposed fault detection and fault-tolerant control scheme, an F-16helicopter case study was considered. Simulation results illustrated the applicability andeffectiveness.