Research On Fault Diagnosis Of ZPW - 2000A Tuning Area Based On Wavelet Neural Network

In China, railway construction is being carried out at full speed, and has become the main artery of national economy, as a result, the vigorous development of the railway promotes the healthy development of more industries. Track circuit is one of the necessary equipment to ensure railway system work safely and efficiently, and ZPW-2000 A track circuit has been promoting the use on most of China railways in nowadays. As an important part of ZPW-2000 A track circuit, the working condition of tuning area plays a decisive role in signal transmission quality and electrical insulation between adjacent sections. Once the tuning area fails, not only train control system will be affected, but traffic safety and efficiency of transport are difficult to guarantee. Thus, the fault diagnosis of tuning area is imperative.The thesis aimed at studying ZPW-2000 A track circuit tuning area. Firstly, based on the transmission line theory and the principle of four-terminal network, a model of circuit track circuit has been maken and its track surface voltage has been simulated. Then proposed fault diagnosis method of ZPW-2000 A tuning area based on wavelet neural network(WNN). Combined with the actual test cases in site, the feasibility of the method proposed in this paper is proved.The thesis studies the combination of wavelet transform and neural networks in two different ways. The first type is a auxiliary binding which is called relax-type Wavelet Neural-Networks, that is wavelet transform as a feature extraction method to preprocess the data, then the extracted feature vectors are used as inputs to the neural network for fault diagnosis.The second type is a nested combination which is called close-type Wavelet Neural-Networks. Namely, after preprocessing the data, replace nonlinear activation function of neurons with nonlinear wavelet basis function. This paper choose the Morlet mother wavelet function as the hidden layer activation function for diagnosis. And compare the two kinds of combination way is the effect of diagnosis.Compare track surface voltage of Railway Signal Maintenance Rules and Matlab software simulation results, that shows the correctness of the model. Finally, binding the accident cases in site to validate this diagnostic method, which shows the efficient, accurate, and feasibility the method is.