Research On EMU Data Processing And Fault Diagnosis Based On Big Data

The development of our country’s high-speed rail has made remarkable achievements.Behind the high-quality development is the important achievements of the railway’s continuous promotion of scientific and technological innovation,and the overall improvement of the intelligent operation and maintenance capabilities of the high-speed rail.With the substantial increase in the number of high-speed rail EMUs operating online,EMU-related data has become increasingly abundant.These various types of data with different structures truly reflect the operating status and fault conditions of EMUs,and are the only way to carry out intelligent operation and maintenance of EMUs.important foundation.From the perspective of EMU data analysis,processing and application,this paper studies the theories and technologies involved in big data-based EMU data processing and its application in fault diagnosis.The content of the paper includes:（1）Research on EMU data sources.According to the characteristics of the data source of the current EMU business system,a scientific classification method is proposed,and the EMU data of different data types are classified.The characteristics of EMU data are analyzed from the aspects of business scenarios,data sources and data structure,which provide the basis for EMU data processing,analysis and design.（2）EMU data processing and analysis research.The data processing requirements of EMU are analyzed,the existing problems of EMU data are discussed,and the data processing flow is designed.Through the introduction of big data processing methods and related technologies,the EMU data processing architecture is constructed by hierarchical and hierarchical planning,integrating the multi-source and massive business data of EMUs and realizing data collection,cleaning,transformation and storage processing.（3）Research on fault prediction of EMU traction motor.Aiming at the shortcomings of the current traction motor fault diagnosis methods,a traction motor fault prediction model based on long short-term memory neural network（LSTM）is proposed.The temperature data characteristics of the traction motor of the EMU are extracted for prediction,and the real operation data of the traction motor is used to make data.The experimental results show that compared with the traditional network model,the output results of the model designed in this paper are less different from the real value,and are more suitable for the task scenario of temperature prediction of the traction motor of the EMU.（4）Research on image fault diagnosis of EMU skirt plate.At present,the analysis and judgment of the image data of the EMU needs to be done manually.In this paper,a fault diagnosis model of the skirt plate based on the convolutional neural network is proposed for the protective part skirt plate of the running part of the EMU.The target is subjected to cluster analysis,secondly,the YOLO algorithm is selected and the network structure is analyzed,and finally the network is trained using the skirt failure dataset of multiple sets of scale features.The test results show that the model can quickly and accurately realize the fault diagnosis of the skirt plate of the EMU.（5）Based on the research on EMU data processing based on big data and its application in fault diagnosis,this paper clarifies the functional requirements of data processing and fault diagnosis,and proposes functional goals.The test scenarios are designed with functional modules.The EMU data is applied to the system,and the functions of EMU operation monitoring and fault warning are realized.