Fault Diagnosis And Classifier Design Of Nonlinear Control System

In this thesis, the method of fault diagnosis of nonlinear control system is studied. Todetection the faults of industrial control system many methords had been considered,there are mainly three categories: method based on hardware, method based onmathematical model and method based on signal. This paper presents a fault observerbased on right coprime factorization and design a faults classifier based on support vectormachine.Firstly, the advantages and disadvantages of all kinds of faults diagnosis methord andthe research status of faults diagnosis based on support vector machine are introduced.Then the basic theory of support vector machine and multi-classification method basedon support vector machine are introduced in detailed.Then the nonlinear liquid level contol system has been modeled and control. Themodel of system is established according to the principle of Bernoulli. According to thenonliner characteristic of the system of liquid level control system, operator theory isused to analyze in this paper. The right coprime factorization controller and trackingcontroller are designed according to the robust stability conditions. Then the faultobserver based on right coprime factorization method is designed to detection theactuator fault of the nonlinear liquid level control system. The fault observer is designedaccording to the condition of the controller which designed based on robust right coprimefactorization method can control the system well. Finally the simulation and experimentresults show the validity of the model and the effectiveness of the proposed method.However, when the system has a variety of faults conditions, the fault observerconsidered above can not detection all kinds of faults when faults happened. Consideredthis situation the multi-faults classifier based on support vector machine is designed next.In this paper, the faults of the liquid level control system are analyzed and four kind’s faults are considered. The faults are the fault of pump, the fault of flow meter, the fault ofliquid level sensor and the fault of electromagnetic valve. Then, according to the actualsituation of the system the feature vectors used for support vector machine is extracted.The feature vectors are the output control signal of controller, the value measured byflow meter and the value measured by liquid level sensor. Then four kinds of fault aresimulated on the actual system and the faults data are used to verify the effectiveness ofthe multi-classifier.