Research And Implementation Of Knowledge Fusion Entity Alignment Method For Intelligent Maintenance On High Speed Railway

With the increasing number of high-speed railway multiple units and longer operating mileage in China,high-speed railway has become one of the main tools for people to travel.The operation safety of high-speed railway has always been highly valued by relevant departments,which puts forward higher requirements for the operation and maintenance service system to ensure the safe and efficient operation of high-speed railway.Modern information technology supports the transformation of China’s highspeed railway operation and maintenance to digitization and intelligence.In the intelligent construction of high-speed railway operation and maintenance,the application of advanced technologies such as knowledge atlas has begun.Extract knowledge from a large number of operation and maintenance data and form a knowledge map,which is of great significance to improve the intelligent level of operation and maintenance system.However,due to the different construction time and standards of the system related to high-speed railway operation and maintenance,it is inevitable that the expression of the same physical entity is not completely consistent,which leads to the alignment problem of the triplet set extracted from these data from different sources.In view of the above problems,this paper takes the traction motor,the key component of EMU,as the research object,defines the entity class and relationship class in the traction motor operation and maintenance data,constructs the triplet set,and puts forward the embedded entity alignment algorithm suitable for the intelligent operation and maintenance of high-speed railway.The main research work of this paper is as follows:(1)Construct data sets in the field of high-speed railway intelligent operation and maintenance.Firstly,the high-speed railway operation and maintenance data from different sources are collected,and the unstructured text is extracted into triplet form by using joint extraction method.Based on the analysis of high-speed railway operation and maintenance data,seven entity classes and six relational classes are defined in this paper to provide basis for generating triplet set.In order to adapt to the alignment scenario,this paper selects two data sets from different sources to form two triples.On this basis,entity alignment data sets for training and testing are built.(2)The entity alignment model ERGCN is proposed.According to different triple concentrated entities of the same semantic expression inconsistent problem,this paper puts forward the geri weis-corbley based embedded aligned entity model,which not only learn the MRAEA,RREA,RDGCN,HGCN combines relationship information and entity information such as,in turn,can enhance entity represents thoughts,enable entities and relationship between to interact more effectively.In addition,GM-EHD-JEA,CEA and other global matching and alignment methods are combined to obtain the most reasonable matching between entities.It effectively makes up for the problem that most translatation-based methods only train on a single triplet and lack the global view of entities and relations.Most methods based on graph neural network ignore the interaction of relation and entity and only adopt the problem of local alignment.It has a good performance on the high-speed railway intelligent operation and maintenance data set constructed in this paper.(3)Develop knowledge fusion tools.The proposed model is applied to the knowledge fusion tool,and the knowledge fusion tool is integrated into the national key research and development project "Knowledge Graph Building Tool".As an important part of the knowledge graph building tool,knowledge fusion service is provided for downstream tasks.In addition,this paper also realizes the encapsulation of knowledge fusion module,which is convenient for users to call remotely.