Research On Adaptive Fault Diagnosis Method Of A Class Of Nonlinear System

With the development of productivity, the industrial has more and more high safety and reliability requirements for production system. But there are many problems in actual industrial system such as interference, nonlinearity, uncertainty and time-delay etc. So research on fault diagnosis for nonlinear systems is very meaningful. In this paper, the main results are as follows:Firstly, with the help of a high-gain nonlinear observer method, a fault detection scheme is conducted for a kind of multiple input and multiple output nonlinear uncertain systems. When designing the high-gain observer, the value of the gain is presented directly without solving the complex linear matrix inequality (LMI). In this article, the feasibility of this method is proved in detail. Since the size of the gain value is presented directly, the proper gain value is selected through the experimental simulation. If the value of the gain could be choosen appropriately, it can seen that the speed of the convergence is faster, the precision is higher, and it could be ensured that the nonlinear uncertain systems has fixed robustness to outside interference on the premise of guarantee convergence. Finally, the high-gain observer is applied to fault detection appropriately.Secondly, using the square root B-spline model as the output probability density function of the system, a research on fault detection for a class of non-Gaussian time-delay stochastic uncertain systems is made. Since the structure of the system containing unknown items and delay items, a new kind of observer is design based on the thought of high-gain observer and the previous observer design method. Since the study of the non-Gaussian stochastic systems with time delay is just in its infancy, the observer design method given in the paper is relatively novel. In addition the paper not only proved estimation error with exponential convergence stability, but also given out the residual error signal for fault detection. With the 3-D grid curve comparison, the feasibility and effectiveness of this method are illustrated.Thirdly, when the system actuator failure occurs, the faults are estimated through designing adaptive observer for a class of multi-input and multi-output uncertain nonlinear systems. Under the condition of the nonlinear term of system meet the global Lipschitz, the high-gain observer combined with a kind of adaptive observer is used to estimate failure. In this paper, the observer is designed without any linearization or solving the complex dynamic equation. It is also proved that the estimation error of observer can converge to a tiny area.