Research On Train Operation Plan For Network Operation Of Urban Rail Transit Based On Interoperability

In recent years, due to the accelerated process of urbanization in China, the subway serving as the infrastructure of urban public transport has been constantly expanded. Automatic train protection system is the indispensable safety equipment which could ensure traffic safety under the subway trains’running status of high density and small interval, and has been widely used in the subways and high-speed railways globally. The train delays and even major accidents caused by the breakdown of ATP on-board device occurred in the process of running are related to the travel of millions of passengers. At present, the maintenance system such as the periodic maintenance and posterior maintenance adopted by subway corporation is ineffective and unable to maintain, repair and troubleshoot those key equipment or device prone to failure in advance. So realizing the intelligent fault diagnosis and prediction of on-board device and ensuring the safer and more stable working of on-board device is significant for improving operational efficiency of subways.This thesis takes ATP on-board device of TBS-100 model on No.13 metro line as the studying object, makes full use of failure logging provided by vehicle maintenance department of No.13 metro line, combines with the service manual offered by manufacturers and rich work experience of on-site maintenance personnel, a new method for fault diagnosis and prediction of the vehicle equipment is proposed, which combines the data statistics with the BP neural network, the method can not only solve the problem that based on reliability prediction method can not describe the specific circumstances of the fault and the accuracy is not high,but also overcome the difficulty of modeling based on model prediction method. A method of combining wavelet transform and BP neural network is proposed to solve the problem that BP neural network training is easy to fall into local minimum point, and the results are compared by an example.Independently designs a fault diagnosis and prediction system of on-board device which could be applied in No.13 metro line. Subway technical personnel to give a high degree of evaluation.The specific work done in this thesis is as follows:Firstly, it studies the structure and function of on-board device of TBS-100 model, analyzes the relationship between its working principle and each function module, sums up the common fault types and fault phenomena and analyzes the causes of the failures, conducts the data mining of above failure loggings and treats it as the data sample of fault prediction, and makes a fault prediction by analyzing the existing fault prediction methods and choosing artificial neural networks method combined with the fault characteristics and data feature of on-board device.Secondly, it studies the theoretical knowledge of artificial neural network and confirms its feasibility of its application to the fault prediction of on-board device. Through theoretical analysis and processing data sample, it establishes the failure prediction model based on BP neural network by taking the R110 module in on-board device as an example; choosing ten fault codes, service time and maintenance frequency as input and equipment failure rate and replacement rate as the output. Also, it learns about the data sample through MATLAB, analyzes the factors that affect the BP network learning rate and the prediction accuracy, optimizes the failure prediction model of on-board device from aspects of number of layers, number of hidden layer neurons, learning speed, transfer function and training function. The validity and accuracy of the model are verified by comparing the predicted results with the actual results of test sample. The wavelet theory and BP neural network are combined to improve the model. The validity and accuracy of the model are verified by an example, and the results of BP network and wavelet network are compared and the results show that the performance of the model is better.In order to realize the diagnostic function of the system, this thesis proposes a fault diagnosis method based on BP neural network, and the accuracy of the generation criterion is verified by simulation. Lastly, we designed and developed a fault diagnosis and prediction system equipped with a friendly human-computer interaction interface by using MATLAB, MySQL database and Visual Studio, for use by the maintenance department. And the requirements for the whole system, structures, development environment and functions of each module are described in detail.