Research On PHM Fault Model And Management Of EMU

With the vigorous development of China's high-speed rail industry,the EMU as the main carrier of railway transportation,the number of on-line along with the growth of demand for transportation capacity has increased year by year.At present,the total number of EMUs on the road has exceeded 3,000,and the huge maintenance task volume and the current plan-based preventive maintenance system have brought huge challenges to the operation and maintenance of EMUs.The whole road urgently needs to realize the reform of the course repair system and deepen the intelligent operation and maintenance of the EMU.EMU prognostics and health management(PHM)is an important means to realize the reform of EMU repair program.Through the use of PHM technology,deepen the application of big data in the professional field,realize the full life cycle data tracking of EMUs and condition-based maintenance based on component status,improve maintenance quality and efficiency,and reduce operation and maintenance costs.As an important component of PHM,the model can directly affect the implementation effect of EMU PHM.Therefore,this article conducts research on the theory and technology involved in the construction and management of the EMU PHM fault model.The main contents include:First,the source data analysis and research of the EMU PHM system.According to the current status of PHM source data of EMUs,the data characteristics are analyzed in terms of data volume,data structure,data source,etc.,and the source data is classified according to different data types to provide a basis for the selection of data processing methods and data processing framework.Second,research on the EMU PHM data processing framework.Aiming at the demand of EMU PHM system business and model for streaming data,combined with related big data technology,a data processing framework based on Spark Streaming and Kafka is proposed to realize the collection,cleaning,conversion processing of EMU PHM source data,support Online processing of streaming data meets the needs of model calculation data.Third,the EMU PHM fault model research.Based on the above research content,the CRH380 BL EMU traction motor is used as the modeling object,and the PSO-BP and random forest algorithms are used to build the traction motor fault prediction model and fault diagnosis model respectively.The feasibility of the model is verified by simulation test.Research provides ideas.Fourth,research on the operation management mechanism of heterogeneous models.Aiming at the characteristics of EMU PHM fault model with wide sources,diverse structures,inconsistent standards and specifications,and different operating environments,the model management method is discussed from the aspects of model access and verification,and a Docker-based model packaging method is proposed to solve the differences Build model cross-platform deployment issues to support the management and application of the EMU PHM fault model.Fifth,based on the research of the heterogeneous model operation management mechanism,the EMU PHM fault model management platform was constructed,and the overall architecture,logical architecture,technical architecture and functional architecture were designed for the platform,and related functions were realized.This paper systematically analyzes and introduces the solutions and related technologies of the problems involved in the EMU PHM fault model and model management research.The feasibility of the method proposed in this paper is verified through the implementation of specific models and systems.In summary,the research in this paper has certain reference value for supporting the construction of Chinese EMU PHM.