Research On Text Big Data Analysis Method Of High-speed Railway Safety

China is the country with the largest construction and operation scale of high-speed railway in the world,as well as the most complicated external environment and operation scenario,in order to ensure the safe operation of high-speed railway,a network transportation organization and safety guarantee technology with Chinese characteristics and super large scale and super passenger and cargo demand has been formed.With the accumulation of the operating mileage of high-speed railway,a large amount of multi-source heterogeneous data is generated and stored for the road network operation and safety,these data carry important safety information about the road network.With the help of big data technology,the value of High-speed railway safety data can be mined to fully understand the safety status and principle of High-speed railway operation,to take scientific and effective measures to prevent and control accidents and their effects is a necessary way to improve the safety level of road network transportation.High-speed railway safety involves various specialties of railway,with many data sources and complex types,among which a large amount of data is stored in unstructured forms such as text,image and audio,High-speed railway safety data stored in text form is the main carrier of unstructured data.Many scholars have studied this kind of text data and put forward many text data analysis methods with reference value,however,there are some problems such as scattered data analysis and poor universality of analysis methods.This paper focuses on High-speed railway safety text data,summarizes the source,characteristics and analysis value of safety text data,and uses the method of knowledge graph construction and application to realize the value mining of safety text data,aiming at the fault data of High-speed railway equipment,this paper proposes a series of text analysis methods to construct the knowledge graph of fault diagnosis and processing,and based on the Railway Data Service Platform,a High-speed railway safety text big data platform is constructed,the main achievements are as follows:(1)Fault classification method for High-speed railway equipment based on deep learning integration.According to the characteristics of high-speed railway equipment fault text data,a combined weighted integration method is proposed to integrate BiGRU and BiLSTM deep learning network,ADASYN adaptive comprehensive oversampling method is used to solve the problem of unbalanced equipment failure data categories,a fault classification model of high-speed railway equipment based on deep learning integration is constructed,automatic fault classification of high-speed railway equipment is realized.Based on the data of equipment failure of high-speed railway from 2008 to 2018,the experiment proves that the equipment failure classification model based on deep learning integration proposed in this paper is a high performance classification model.(2)A method to extract named entity and entity relationship for fault of high-speed railway equipment.A named entity and entity relationship knowledge structure for fault diagnosis and processing is defined according to the device fault text data,the BIOES labeling method was used to annotate the sample data of named entity and entity relationship,the multi-dimensional character feature representation +BiLSTM+CRF is proposed to implement the named entity extraction method,and multi-dimensional word segmentation feature representation +Transformer network to achieve entity relationship extraction method,the automatic extraction of key information from equipment fault text data is realized.The failure data of High-speed railway switch equipment from 2008 to 2018 were used to verify the model,experiments show that the named entity and entity relationship extraction methods presented in this paper have high evaluation indexes.(3)A method of device fault entity alignment based on concept similarity calculation.In order to solve the problem of redundancy of equipment fault named entities extracted by named entity extraction method,an entity alignment method based on similarity calculation of the concept of Hownet is proposed,through the lexical extraction process of equipment fault entity,the calculation process of concept similarity and the entity alignment process,the de-duplication and unification of named entity with redundant equipment fault are realized.The named entity of equipment fault is used to verify the method,the experiment proves that the method of named entity alignment based on concept similarity calculation can effectively solve the problem of named entity redundancy.(4)Design and Implement of High-speed railway safety text big data platform.Propose the general framework of High-speed railway safety text big data platform,as well as the corresponding technical architecture and functional architecture,the business process of safety text data analysis and the dynamic interaction mode of the model are designed,and an intelligent operation platform integrating safety text data from sample annotation,model training and application analysis is realized,the core functions of the platform are demonstrated,and the realization process of the fault text analysis methods on the platform is introduced by taking the knowledge graph of fault diagnosis and treatment of signal equipment as an example.The platform is applied to the subordinate units of the institute of China Academy of Railway Sciences.through the analysis of the safety text data of each High-speed railway in the past ten years by each specialized department,it is proved that the research results of this paper can effectively solve the text data analysis of High-speed railway safety.