Fault Diagnosis Of Railway Turnout Switch Machine Based On Deep Learning

With the rapid development of China’s railway,the running speed,mileage and density of the train are also increasing.As the key component of railway equipment,the frequency and quantity of turnout are increasing.If there are problems in the process of using the turnout,it may affect the train operation,and even more serious,it may cause personal safety threat.Switch machine is also an important part of railway equipment,which is mainly used to switch switches.The normal and smooth operation of switch machine has an important impact on the safety and efficiency of railway operation.At present,China’s railway operation monitoring system ensures the safe and reliable operation of the switch machine by means of regular maintenance and skylight point maintenance.This traditional way has heavy workload and low work efficiency.Open box operation may bring more security risks due to manual operation errors,so it relies on manual experience to identify the fault of the switch machine,It is easy to miss and misjudge the fault due to experience problems,thus endangering the driving safety.The oil pressure signal of switch machine is an important feature and basis to reflect the normal operation of switch machine.The pressure of the cylinder will vary in different stages of switch machine conversion.If there is a fault,the pressure of the cylinder will be very different from that of normal condition.So,according to the change of the oil pressure of the switch machine and the comparison with the normal situation,it can be judged whether there is any abnormal situation in the process of switch machine conversion.In this paper,switch machine,which is widely used in China’s railway,is taken as the research object.Combined with the advantages of CNN in data local feature extraction and RNN in time series data processing,a fault diagnosis model of switch machine based on cnn-gru is established.The oil pressure data in the process of switch machine conversion is used to train the model,and the influence of different optimizers on the model diagnosis rate and fault recognition rate is explored.This paper uses ROC model evaluation index to evaluate the fault diagnosis model of cnn-gru switch machine.The experimental results show that the CNN Gru fault diagnosis model proposed in this paper has a good diagnosis rate for switch machine,and can identify the fault category accurately.The fault diagnosis method proposed in this paper has reference significance for fault diagnosis of switch machine.Based on the established cnn-gru model,this paper develops a fault diagnosis system of switch machine based on deep learning,which makes the results of fault diagnosis visible.