Research On Intelligent Fault Diagnosis Of Nonlinear Systems

The increased complexity of modern control system makes the sensors、actuators and system components are inevitable malfunction. The system security and reliability requirements are also increasing, whether it is from the perspective of system security, or from an economic point of view. The fault diagnosis is one of the important techniques to improve the safety、the reliability of dynamic systems. Therefore, further study of fault diagnosis technology, not only has important theoretical significance, but also has great practical value. In addition to modeling errors and external disturbances, the actual system presents nonlinear to some extent, thus making the system fault diagnosis has become more complex and difficult. At present, the fault diagnosis technology of nonlinear systems is one of the hot and difficult contents of the current study. Some problems existing in the nonlinear system fault diagnosis and its development trend are studied with latest research results of relevant theories in this paper.Firstly, the background, research methods and development trend of fault diagnosis are introduced. Our research are focused on the intelligent fault diagnosis of nonlinear systems and its applications, especially at the neural network and support vector machine of intelligent fault diagnosis method.Secondly, the intelligent fault diagnosis based on RBF neural network and support vector machine are researched. First, the RBF neural network and its algorithm are studied. Neural network has the ability to approximate any nonlinear function with arbitrary precision, simulation about a discrete nonlinear system is researched, which indicate this method provides a good fault diagnosis model. Second, the regression algorithm and learning methods about support vector machine are studied, a nonlinear dynamic system is identified using support vector machine, and thus provides a solid theoretical foundation for intelligent fault diagnosis.Again, the robust fault diagnosis is proposed for a class of model uncertainly nonlinear system based on RBF neural network. The system can be measured only input and output, and the input included the modeling errors and other uncertainties. The method to fit the nonlinear system fault characteristics and uncertain part of the system by constructing a neural network online approach, fault diagnosis method applied threshold processing technology to ensure that the algorithm is robust, for a given algorithm, using Lyapunov to prove its stability. Finally, the correctness of this method is verified by simulation experiment.Finally, it is difficult to build the analytical model when the system exist serious nonlinear behavior, so the fault diagnosis method based on the support vector machine is proposed. First of all, for a class of nonlinear systems, the fault diagnosis observer model is established by support vector machine, and use Lyapunov method to analyze the stability of the system. The simulation shows that support machine can accurately detect the fault diagnosis. Then, the research of fault about sensor stuck、constant gain and bias by using support vector machine for a nonlinear system, the simulation indicates the SVR can provide a new method for nonlinear systems.