Research On Key Technologies Of Switch Machine Health Management Based On Data Driven

With the rapid development of rail transport,safety problems in the operation of rail transport became more and more prominent,which directly affected the efficiency and operating cost of rail transport.As a key equipment in railway transportation,the safety status of railway turnout switch machine directly affected the normal operation of rail transport.However,there were many problems in the traditional maintenance methods of turnout switch machine,which could not effectively ensure the operation safety of equipment.Looking for effective solutions had become an urgent task.Since the 1960 s,health management system technology had made great progress in many engineering fields.Health management system technology was not only an effective equipment health operation and maintenance technology,but also a feasible way to realize the efficient operation and maintenance of turnout switch machine.This dissertation mainly studied the key technologies such as condition monitoring and fault diagnosis in the health management of turnout switch machine equipment,and developed the health management system software of turnout switch machine,so as to realize the intelligent health management of equipment.In view of the complexity of turnout switch machine design and the randomness of the fault occurrence,the traditional condition monitoring and fault diagnosis methods could not effectively deal with it.Therefore,this paper adopts a data-driven method for the condition monitoring and fault diagnosis methods of the switch machine equipment.Intelligent Algorithm.The condition monitoring algorithm of switch machine based on AOA-XGBoost and fault diagnosis algorithm of switch machine based on ITD-SDP image feature was studied respectively.The main research work of this paper is as follows:（1）Research was carried out on the status monitoring algorithm of switch machine equipment.Aiming at the deficiency of the traditional status monitoring method of switch machine equipment,a status monitoring algorithm of switch machine based on AOA-XGBoost was studied.The algorithm could effectively use the time-domain and frequency-domain features of switch machine switching process oil pressure data to monitor the health state of the equipment and give timely early warning when the equipment was in bad state.The algorithm mainly depended on the XGboost model to judge the equipment state,and the Archimedes optimization algorithm was used to optimize the hyper-parameters of the XGboost model to improve the accuracy of the algorithm model.Simulation results prove that the proposed method has lower false positive rate and false negative rate than other methods.（2）Research was carried out on the turnout switch machine equipment,the equipment fault diagnosis was completed by using ITD-SDP image features and depth separable convolution neural network（DSCNN）model.The simulation results show that the image features can effectively characterize the fault condition of the turnout switch machine,and the average diagnosis accuracy of DSCNN model for the four working states of the turnout switch machine is 98.5%,which verifies the effectiveness and superiority of the method.（3）The health management system software of turnout switch machine was developed.The condition monitoring algorithm of turnout switch machine base on AOA-XGboost and the fault diagnosis algorithm of DSCNN turnout switch machine base on ITD-SDP image features studied in this dissertation are applied to the condition monitoring module and fault diagnosis module in the system respectively.The system has been tested and run to meet the basic needs of the healthy operation and maintenance tasks of the switch machine.