Research On Fault Diagnosis Of Main Transformer Of HX Locomotive Based On PHM

As the backbone of China’s railway transportation,locomotives undertake a variety of operations such as passenger transport,freight,shunting and small operation.Since2017,most locomotives have gradually entered advanced repair,the amount of maintenance tasks has gradually increased,and the maintenance operations have faced new challenges.The traditional planned repair method has caused problems such as wasted costs,reduced efficiency and increased potential safety hazards,and it is imperative to transform the locomotive maintenance method and promote the development of intelligent railways.By referring to the research of various types of complex equipment at home and abroad,fault diagnosis and health management（hereinafter referred to as PHM）is taken as the research entry point to promote the transformation of locomotive maintenance from planned repair to state repair.At present,the total number of locomotives allocated to the whole road has reached 20628,of which the number of built locomotives has reached19108,and a closed loop of health management data for the whole life cycle of locomotives has been established,which provides data support for the research of locomotive PHM.First,analyze the characteristics of the data source and study the PHM data governance architecture of the transformer of the harmonious locomotive owner.From the dimensions of data collection,data type,data characteristics and other dimensions,the data source is analyzed,the shortcomings of the existing data governance methods are found,and the key technologies of big data governance are combined,and finally the big data governance architecture of the harmonious locomotive transformer PHM is studied.Second,based on the health management data of the harmonious locomotive owner transformer provided by the data source,combined with the relevant requirements and methods of maintenance,the fault factors that affect the health status of the main transformer are extracted in addition to the maintenance requirements,and the two are combined as the characteristic inputs for establishing the fault diagnosis model,and the data is cleaned to construct the data set available for the fault diagnosis model.Third,using the built data set,establish a fault diagnosis model that is practical for practical applications.Analysis can know that fault diagnosis is a classification problem in machine learning,by comparing the advantages and disadvantages of various classification algorithms,the final choice is to build a harmonious locomotive owner transformer fault diagnosis model based on XGBoost algorithm,and the scientific and feasibility of the model is verified from two angles: machine learning algorithm evaluation and case analysis.The accuracy of the final trained model reached 93.96%,which was significantly improved compared with the accuracy of 81.4% of the current fault diagnosis method of maintenance.Fourth,clarify the locomotive maintenance and management requirements,and design a harmonious transformer fault diagnosis system for locomotive owners based on PHM technology.Fully combined with the main transformer maintenance business process,the overall architecture of the system and the functional architecture of "1+6+N" are designed,and the typical application scenarios in five fault diagnosis: locomotive monitoring,data management,model management,fault information management,and ticketing information management are introduced.In accordance with the standard process of PHM technology,based on the current data conditions and the current maintenance method,this paper systematically studies the transformer fault diagnosis method of harmonious locomotive owners,and designs a system to verify the feasibility of the method proposed in this paper,which has certain reference value for the future comprehensive PHM research of harmonious locomotives.Figure 26,Table 12,Reference 58.