Research And Realization Of Fault Diagnosis Of Multi - Component Complex System

With the rapid development of China’s high-speed railway, the high-speed railway has made great contribution to the development of China’s economy. Traction motor as one of the core systems of the Multiple Units, its running state is directly related to the safety of the train. Traction motor itself is complicated in structure, which is composed of a plurality of components. The traditional "Breakdown Maintenance" and "Time Based Maintenance" cannot identify the fault location in the event of a failure. PHM is the abbreviation of Prognostic and Health Management, which can solve the problem timely. The existing Multiple Units traction motor maintenance strategy is still in "Time Based Maintenance", fault handling is not timely. So it is of great significance to carry out fault prediction of traction motor.This paper will take CRH2 as an example to study the failure mechanism of traction motor. The general process and methods of classification and prediction model of PHM are analyzed in this paper. Combined with the fault data of traction motor and integration of PHM and fault prediction of traction motor, find out the relationship between the temperature data and mechanical fault of traction motor during the operation, to establish prediction model.The main research work in the paper included the following aspects:(1) Study on traction motor data acquisition technology, completed the development of Multiple Units data transmission system, including data acquisition module and wireless transmission module. Then we preprocess and mine the data, to find out the correlation within the data and use the data to train the proposed model.(2) The construction of fault prediction model of traction motor is the key point of this paper. The Bayesian Network is proposed, which has powerful learning ability and probabilistic evaluation mechanism to realize the traction motor fault prediction method. Firstly, based on the expert knowledge, the initial Bayesian Network structure is established. In view of the incomplete Bayesian Network structure and contains Latent Variable, an improved Structural Expectation Maximization(SEM) algorithm is proposed. Secondly, the Bayesian Network structure and parameters are studied by SEM algorithm, and a perfect Bayesian Network structure is established. Finally, with the obtained data of the traction motor, especially for temperature data, the fault of motor bearing and stator fault is predicted through the Improved SEM algorithm.Simulation results show that the algorithm proposed in this paper can effectively estimate the relationship between the bearing and the stator fault and the temperature of the traction motor, and can successfully diagnose the fault of the motor through the temperature data.