A Class Of Switched Systems Robust Fault Detection

In the last two decades, Fault Detection and Diagnosis have attracted attention of the International automation control field, they exist widely in engineering and social systems. But many actual systems not only include continuous-time systems but also include discrete-time systems, which constitute a hybrid system. Switch system as the most typical of hybrid system, Study fault detection problems of switching system will promote the development of the theory of hybrid systems. This work, which is supported by the National Natural Science Foundation of China, investigates the problem of fault detection for switched systems base on the model method. The main contributions are as follows:Investigating the problem of fault detection for switched systems with continuous-time delay is basing on state estimation method. The fault detection will be put into observer design. By the comparison of the actual value and the estimated value, we will get residual function. By the introduction of a reference model to get the state error of system model, based on multiple Lyapunov functions with no index item and average dwell time approach, the condition that guarantee the state error system to be exponentially stable with a prescribed H∞performance is derived, and the existence conditions of observer are proposed in terms of LMI. Finally, we use simulation examples to prove its feasibility.It is considered that in practical application, there exists nonlinear perturbations, the state of system and its derivatives. When the system cases fault, by the comparison of the actual value and the estimated value, we will get state error of system model, And then through multiple Lyapunov functions with no exponential and average dwell time approach analysis system exponential stability and H∞performance. Under arbitrary switching signals of the average dwell time analysis system exponential stability and H∞performance, and the existence conditions of state error system are proposed in terms of LMI. Finally, through the design of threshold, when the moment of t, the residual evaluation function value is greater than the threshold value, an alarm signal will generate, in the contrary the system will in normal operation.