面向高速铁路道岔设备故障处置的知识图谱构建与应用
Construction and Application of Knowledge Graph for Troubleshooting of High-speed Railway Turnout Equipment
林海香 1赵正祥 1卢冉 1白万胜 1胡娜娜 1陆人杰2
作者信息
- 1. 兰州交通大学自动化与电气工程学院,甘肃兰州 730070
- 2. 卡斯柯信号有限公司,上海 200071
- 折叠
摘要
为了解决道岔设备故障处置过程中产生的大量知识难以存储应用,半结构化故障文本数据不能直接检索和推理等问题,提出一种高速铁路道岔设备故障处置知识图谱的构建方法.以少量人工标注故障文本数据为基础,采用基于强化学习的方法搭建BERT-BiLSTM-CRF*深度学习模型抽取实体;根据知识领域特点和数据结构设计关系匹配模板抽取各实体间的关系;针对实体冗余问题,采用文本相似度与结构相似度结合的方法进行故障实体融合;以我国各铁路局近3年的高速铁路道岔设备故障文本为研究对象,累计抽取实体807个且标注准确率在90％以上,构建实体关系2 232条.基于以上知识图谱构建的高速铁路道岔设备故障处置知识图谱直观展示道岔设备故障关系,实现不同实体对象间的数据关联查询,挖掘隐含关系,辅助道岔维护人员进行故障处置决策,实现道岔设备预防性维护,延长设备寿命.
Abstract
In order to solve the problems of difficult storage and application of a large amount of knowledge generated in the process of turnout equipment fault handling as well as the inability to directly retrieve and reason semi-structured fault text data,a method was proposed to construct a knowledge graph for high-speed rail turnout equipment fault han-dling.Firstly,based on a small amount of manually annotated fault text data,a BERT-BiLSTM-CRF*deep learning model based on reinforcement learning was used to extract entities.Then a relationship-matching template was designed to extract the relationship between entities according to the knowledge domain characteristics and data structure.Subse-quently,for the problem of entity redundancy,a combination of text similarity and structural similarity was used to fuse fault entities.Finally,807 entities were extracted with annotation accuracy rate of over 90％,and 2,232 entity relation-ships were constructed based on the fault texts of high-speed railway turnouts equipment in the past three years from each railway bureau in China.The knowledge graph of high-speed railway turnout equipment fault handling constructed based on the above knowledge visually displays the fault relationships of turnout equipment,realizes the data association queries between different entity objects,explores the implied relationships,assists turnout maintenance personnel to make fault handling decisions and realizes preventive maintenance of turnout equipment to extend equipment life.
关键词
高速铁路道岔设备/知识图谱/强化学习/知识抽取/故障处置Key words
high-speed rail turnout equipment/knowledge graph/reinforcement learning/knowledge extraction/fault handling引用本文复制引用
基金项目
甘肃省科技计划(23YFGA0046)
甘肃省高等学校创新基金(2020B-104)
四电BIM铁路行业重点实验室开放基金(BIMKF-2022-02)
出版年
2024
NETL
NSTL
万方数据