Design And Development Of Fault Diagnosis Support System Towards Massive Bearing Temperature Data Of High-Speed Trains

With the continuous development of high-speed trains,the safety of high-speed trains has also received more and more attention.As one of the key components of high-speed trains,the status of bearing is directly related to the safety of the entire train.It is of great significance to diagnose faulty bearings in time and take measures to ensure the safe operation of the train.The existing research on bearing fault diagnosis of high-speed trains is mostly through the modeling and analysis of bearing temperature data to achieve the purpose of fault monitoring.Traditional research methods are mainly based on small sample data,In recent years,some scholars have found that the massive operational data of high-speed trains has great value for fault diagnosis of high-speed trains.By mining and modeling the massive data of high-speed trains,it can provide more accurate and rich information for the monitoring and diagnosis of bearing faults.Because the traditional fault diagnosis system and method can’t effectively model and analyze the massive bearing temperature data of high-speed trains,how to effectively model and diagnose the massive bearing temperature data of high-speed trains is a problem that needs to be solved in current research.In recent years,the development of big data processing technology has provided a new way to solve the problem of modeling the massive bearing temperature data of high-speed trains.However,fault diagnosis researchers still face great technical challenges in diagnosing the massive bearing temperature data at present.It is embodied in the following aspects:1)the big data processing framework itself has complex distributed features and is very difficult for researchers to use;2)existing big data processing frameworks do not have a distributed approach to time series pattern retrieval;3)existing big data processing frameworks do not have basic distributed algorithms that support massive bearing temperature data for fault modeling,making it difficult to develop distributed fault diagnosis algorithm.Therefore,in order to assist and support fault diagnosis researchers to carry out fault diagnosis related research on massive bearing temperature data of high-speed trains and to reduce the technical difficulty of fault diagnosis researchers using big data processing technology to study the related research of bearing fault diagnosis,It is urgent to study the fault diagnosis support system towards massive bearing temperature data for high-speed trains.Based on the research of "Fault Modeling theory and method of high speed train information control system based on big data and knowledge",this paper puts forward a fault diagnosis support system for massive bearing temperature data of high-speed train.The main research work of this paper is as follows:(1)The existing fault diagnosis systems and methods at home and abroad and the development status of big data processing technology are studied.Because the bearing temperature data of high-speed trains has large data volume,high dimension and fast update speed,it has obvious characteristics of big data.However,there is currently no system that can effectively support the fault analysis modeling and diagnosis of massive bearing temperature data of high-speed trains.Therefore,this paper proposes a fault diagnosis support system for massive bearing temperature data of high-speed trains,and carries out detailed requirements analysis on system functions and distributed algorithms combined with the characteristics of bearing temperature data of high-speed trains.(2)Design the system according to demand analysis.According to the modular design idea,the overall function structure and system architecture of the system are designed.The system is mainly composed of large data base support layer,fault modeling support layer and fault diagnosis support layer.The big data foundation support layer provides basic management functions of bearing temperature data,including big data storage,big data query and visualization functions;The fault modeling support layer provides the function of feature extraction of the bearing fault,and supports the distributed bearing temperature time series retrieval method to retrieve the fault characteristics of the massive bearing temperature time series and provide data support for subsequent fault modeling;Fault diagnosis support layer provides distributed fault diagnosis algorithm library,and supports fault diagnosis personnel to research and develop distributed fault diagnosis algorithm.(3)According to the design scheme,based on the industrial cloud platform,the development of the system is completed by using Java development language,Scala development language,big data processing framework such as Spark,Hadoop,etc.Among them,HDFS is used as the underlying storage technology of the supporting platform to solve the high-reliability storage of massive data generated during the operation of high-speed trains.Spark is used as the computing engine of the supporting platform,which can effectively calculate the massive data because of the in-memory computing technology.Impala is used to design big data query modules to solve the query requirements of massive high-speed train data.Zeppelin is used to design the system’s visual functional module to solve the problem of data visualization for high-speed trains.Based on the Spark platform,a distributed time series retrieval method is designed,for the data association problem between different computing nodes,a distributed time series data set is constructed by decoupling data redundancy.And based on the optimal calculation performance,the optimal length of the elements of the distributed time series data set is solved by minimizing the time function of the calculation process.Aiming at the problem that there is no distributed algorithm library for fault diagnosis at present,the distributed fault diagnosis algorithm library is designed and implemented based on the reusable and scalable ideas by abstracting the existing data-driven fault diagnosis algorithm,which can help researchers to carry out fault diagnosis modeling research on massive bearing temperature data of high-speed trains on the basic support platform.(4)The system is deployed in the Industrial Cloud Computing Center of the State Key Laboratory of Integrated Automation for Process Industry.The distributed system consists of 10 nodes on the cloud platform.Firstly,the basic functions such as data storage,data query and visualization of the support platform of the system are verified.Then the consistency and the parallel efficiency of the distributed time series retrieval algorithm were tested using the bearing temperature data of the high speed trains,and the parallel acceleration effect of distributed PCA fault modeling algorithm and distributed DPCA fault modeling algorithm was tested.The experimental results show that the functions of the basic support platform proposed in this paper meets the design requirements.And distributed time series similarity retrieval algorithm and distributed fault modeling algorithm have good parallel acceleration performance,so the design purpose of the system is achieved.