Tuning Area And Compensation Capacitance Fault Diagnosis Of Jointless Track Circuit Based On LVQ Neural Network

With the promotion of Outline on Building a Powerful Transportation Country,high-speed railway has stepped into the era of great development,which puts forward new requirements for railway signaling equipment.Among all,Track circuit is one of basic equipments of railway signal system,which is not only an important part of interlocking,but also the channel of information transmission between train and each station.If track circuit fails,it will directly affect the train control system to judge the condition of the track and transmit the wrong information,thereby endangering the train operation safety and reducing the transportation efficiency.At present,more and more jointless track circuits are put into use.How to detect all kinds of faults of jointless track circuits accurately and in real time is the key research direction in the field of railway signal.The fault of the jointless track circuit mainly includes the compensation capacitance fault of the main track circuit and the equipment fault of tuning area.However,there is still a lack of an efficient and accurate method for the comprehensive diagnosis of the above two faults.Therefore,a study of tuning area and compensation capacitance fault diagnosis of jointless track circuit based on LVQ(learning vector quantification)neural network was proposed in this paper.First,according to the two port network theory,a four port network model of the jointless track circuit was established.The model can accurately express the normal state and fault state(tuning area and compensation capacitance fault)of the jointless track circuit.On the basis of the above model,the parameters of the model were set.Secondly,according to the four port network model of jointless track circuit,the envelope curves of shunt current amplitude in different states(normal state and fault state)of jointless track circuit were simulated and analyzed.Based on that,the corresponding amplitude curve of each state was obtained.Then,EEMD(empirical mode decomposition)and CEEMD(comprehensive ensemble empirical mode decomposition)were respectively used to decompose the shunt current signal into several features and the results from two algorithms were compared.It can be found that CEEMD algorithm can better retain the characteristics of the original signal by changing the white noise introduced by EEMD algorithm into complementary noise,thus improving the operation accuracy.Finally,the LVQ neural network model was introduced into the tuning zone equipment of the jointless track circuit and the fault diagnosis method of compensation capacitance.In the third chapter,the feature vector extracted by ceemd and some field data were used to train LVQ neural network.After the training,the remaining data was used to test LVQ neural network.The test result of LVQ was compared with that of the improved BP neural network method.The results show that the fault diagnosis method based on LVQ neural network can detect the track faults more accurately,which verifies the correctness of the fault diagnosis method based on LVQ neural network.