Fault Diagnosis For The Discrete Mode Transition Of CTCS-3 Based On Hybrid System Theory

With the expansion of the system size and complexity, single-use of the continuous dynamical system theory or discrete event system theory to describe a system model is insufficient in completely expressing the characteristics of the system.Actually, many systems possess hybrid characteristics, which imply that the dynamic system comprises two types of states:continuous states and discrete states (discrete modes), and the interaction and mutual coupling between them lead to a hybrid characterization of system behavior. To better express the system behavior under the co-driven of time and events, Hybrid System theory is introduced.CTCS-3 is a typical hybrid system. While the train is running, there are continuous states like velocity and displacement, and discrete modes like operating modes, brake level and traction level. Since CTCS-3 is a safety-critical system, its performance can be measured by the important indicators—safety and reliability. Fault diagnosis technology benefits the improvement of system reliability, maintainability and availability. It is significant that the effective work of fault diagnosis is applied in railway transportation to guarantee the efficiency and safety.The main contents of the thesis include:(1) As an important branch of Hybrid System theory, Switched System is used for the modeling of a tank model in the thesis to verify the feasibility of method.Then the theory is applied to a track problem between two trains. The thesis analyses the evolution of the continuous states and discrete states, and the output of discrete states is applied to the fault location of system, for fault is regarded as a special mode in the Hybrid System theory.(2) The thesis describes two kinds of algorithms of particle filter, SIS and SIR. Both of them are used in the state estimation of the tank model and the tracking model. And an improved method of SIR algorithm is proposed. The concept of root mean square error is introduced to value the estimation effect, and the estimation results are used for the fault detection.(3) A model-based approach is applied to the fault diagnosis of CTCS-3.In the thesis, a model adequately considering the hybrid characteristics of CTCS-3 is built firstly, and the work of fault location is done on basis of the output of discrete modes. The state estimation based on the particle filter is used for the fault detection. And a method based on the "Big Value" of weight is proposed to the application of fault detection, and the feasibility of which is verified with the traditional method of residual smoothing. And the multi-model method is introduced to verify the feasibility of fault location method, it is shown that the method is more effective in case of the problem that there is no significant difference of evolution of states under different fault modes, and the discrete modes can be designed to reflect the combination fault of system.The Switched System theory is in application of the modeling of CTCS-3.With the method of dividing subspaces, not only the complexity of the modeling is reduced, but also the accuracy of the model is ensured, making the application of fault diagnosis based on model possible. And this is the first attempt to combine Switched system theory with particle filter, which is applied in the tracking scenario, and the fault diagnosis of the scenario is realized.The thesis proposed a new method of fault detection based on the "Big Value" of weight, and a binary discrete mode classification method is designed to apply to the fault location.