Study On Fault Diagnosis Based On Bayesian Network For Air Brake System

Brake equipment is a complex pneumatic control system, there are many uncertainty in appearance of fault, traditional modeling theory and method of fault diagnosis have difficulties in describing uncertainty accurately, which leads precise diagnosis conclusion is obtained difficultly. In order to enhance fault diagnosis efficiency, fault diagnosis models based on Bayesian network is proposed in this paper.The major contribution in this thesis can be separated into three parts:1. The basic theory and methods of Bayesian Networks are introduced and discussed. As a new mechanism for uncertain knowledge representation and manipulation based on probability theory and graph theory the Bayesian Network is a kind of important modeling, inference and learning tool in complex system. it can perfectly quantizing uncertainty generally in complex system and can provide more precise result based on its probability inference, in the same time the system model based on Bayesian Networks has more intelligence.2. To build the 104 air brake system Bayesian network, the nodes of Bayesian network and the states of components of system must be determined firstly. According to reliability of components and field expert experience, prior probabilities are achieved. And then using the classical algorithms based on Bayesian networks infer the fault or work probabilities of components. The fault location could be realized. The model can clearly express the states of system and components and the state probability, and also carry out qualitative analysis and quantitative assessment of air brake system fault cause. The diagnosis results indicate the effectiveness of fault diagnosis of air brake system by using this method.3. Due to the configuration complexity of the diesel locomotive air brake system, it is difficult to realize the fault diagnosis on the brake system with many uncertainty in appearance of fault. In order to enhance fault diagnosis efficiency for diesel locomotive air brake system with uncertain fault, a fault diagnosis model based on Bayesian network is proposed in this paper. According to a priori exact probability or experts estimate that the probability, the classical Expectation-Maximization algorithm calculates the joint fault probability distribution and probability distribution of marginal respectively. Based on joint tree algorithm, Bayesian network is designed to infer the fault probabilities of components. The fault location could be realized. The simulation results indicate that the accurate fault probabilities could be calculated. Therefore, this method is effective for uncertain fault.