Study Of Intelligent Railway Turnout Fault Diagnosis Based On Neural Network

How to get, analyze, process and use the effective feature data of the turnout action circuit and indication circuit and try to classify the reason of turnout fault data is the core study of the railway turnout fault intelligent diagnose. The difficulty is that how to get and analyze the knowledge of turnout typical fault. Through collecting and classifying the actual faults on site, we summarize the typical faults and classification them by using the experimental base at the same time. The problems in the process of obtaining and analyzing the typical fault data of the turnout can be solved effectively. For turnout fault feature information, that includes of different types of turnouts, uncertain failure cause, complex and varied fault curves. We have problems in establishing the reflection relationship between the fault feature vector and the cause of the fault and the model of the turnout fault intelligent diagnosis. Based on this, in order to solve the problem of the turnout intelligent fault diagnosis, we use effective data processing technology and the extraction method of feature vector, study of intelligent diagnosis model parameters and the set of corresponding rules between the fault feature vector and the fault reason.In this paper, based on the existing signal monitors of the interlocking circuit, on condition that the linkage relationship of the existing signal equipment remains unchanged and no bulk of monitors will be added. According to the existing signal monitors of the railway and by using the data provided by the computer system, we have studied the intelligent diagnose of the failure of the turnout. The research main work and results of this paper are as follows:1. The status quo and the existing phenomena of the turnout failure in practice were analyzed, actual demands were proposed and the disadvantages of the frequently-used solutions were concluded. Accordingly, intelligent diagnose of the failure was chosen.2. After carefully analyzing the mechanical and circuitous philosophies of the ZD6 and the speed-up turnouts and combined with the acquisition principle of the computer monitoring system and actual situation on site, analyzed the action current curves concerning the typical failure of both the ZD6 and the speed-up turnouts, laying a good foundation for the intelligent diagnose of the failure of the turnout action circuit.3. To analyze the basic of the action current curves of the ZD6 turnout in fault, three methods to extract vectors of those curves’ characteristics were proposed. Based on that, the intelligent diagnose of the failure was proposed on the of neural network. Finally, intelligent diagnose tests for the living examples of the ZD6 were conducted respectively. offline tests for the actual fault of history. The results of the test is the same as the actual failure.4. Similarly, the action current curves of the speed-up turnout in fault were concluded and two methods to extract vectors of those curves’ characteristics were proposed. And the intelligent diagnose of that failure was accordingly proposed on the basis of neural network.. the voltage curves of the speed-up turnouts in fault were analyzed and concluded, offline tests for the actual fault of history. The results of the test is the same as the actual failure. and then vectors were extracted accordingly before the intelligent diagnose of the failure was finally made on the basis of the neural network.5. The voltage curves of the ZD6 and the speed-up turnouts in fault were analyzed, and the characteristics of those curves were extracted before the intelligent diagnose of the indication circuit was proposed on the basis of neural network. Offline tests for the actual fault of history. The results of the test is the same as the actual failure.The conclusion of this paper not only has theoretical significance and practical application value to the intelligent diagnosis of turnout fault, but also provides a direction for the development of the computer monitoring system in the intelligent diagnosis of turnout fault. Besides, the failure of other signal equipment can benefit from this paper as well.