现代防御技术 ›› 2022, Vol. 50 ›› Issue (5): 114-121.DOI: 10.3969/j.issn.1009-086x.2022.05.015

• 综合保障性技术 • 上一篇    下一篇

雷达装备故障原因知识图谱构建研究

谢雨希1, 杨江平1, 孙知建1, 李逸源1, 胡欣1,2   

  1. 1.空军预警学院,湖北 武汉 430014
    2.中国人民解放军93498部队,河北 石家庄,050000
  • 收稿日期:2022-02-20 修回日期:2022-05-30 出版日期:2022-10-28 发布日期:2022-11-03
  • 作者简介:谢雨希(1990-),女,重庆大足人。助理工程师,博士生,研究方向为预警装备管理与保障。

Construction of Radar Fault Cause Knowledge Graph

Yu-xi XIE1, Jiang-ping YANG1, Zhi-jian SUN1, Yi-yuan LI1, Xin HU1,2   

  1. 1.Air Force Early Warning Academy, Hubei Wuhan 430019, China
    2.PLA, No. 93498 Troop, Hebei Shijiazhuang 050000, China
  • Received:2022-02-20 Revised:2022-05-30 Online:2022-10-28 Published:2022-11-03

摘要:

对雷达装备故障原因文本的分析,有助于定位故障部位,便于装备维修,分析装备性能。利用知识图谱技术对雷达故障原因文本进行知识抽取、表示和管理,可有效提升文本的利用效率,快速定位故障部位,及时分析装备性能。在总结了雷达装备故障原因文本特点基础上,运用模式层和数据层相结合的方法构建了雷达装备故障原因知识图谱。自顶向下定义知识图谱的组织架构、概念类型和概念间关系,形成知识图谱的模式层;针对文本特性,自底向上构建知识图谱的数据层:为解决样本量较小和识别由复合词组成的实体的问题,运用Att-ALBERT-BiGRU-CRF模型进行了实体命名识别;运用ALBERT-BiGRU-Att模型进行关系抽取。实验证明了上述抽取方法的有效性,将抽取出的实体与关系构建了雷达装备故障原因领域知识图谱。

关键词: 雷达故障原因文本, 命名实体识别, ALBERT, 关系抽取, 知识图谱

Abstract:

The analysis of radar fault cause text is helpful to locate the fault location, facilitate equipment maintenance and analyze equipment performance. Knowledge extraction, representation and management of radar fault cause text using knowledge graph technology can effectively improve the utilization efficiency of the text, quickly locate the fault location and analyze the equipment performance in time. On the basis of summarizing the characteristics of radar equipment fault cause text, radar equipment fault cause knowledge graph is constructed by a method combines both the scheme layer and the data layer. The scheme layer of the knowledge graph in the top-down style is designed, which defines the knowledge framework, the concept types, and the relationships between the concepts of the knowledge graph. For the text characteristics, the data layer of knowledge map is constructed in the bottom-up style: in order to solve the problems of small sample size and compound words entity recognition, entity naming recognition is carried out by Att-AlBERT-BiGRU-CRF model. The ALBERT-BiGRU-Att model is used for relationship extraction. Experiments show the effectiveness of the above extraction method. Based on the entities and relationship, the knowledge graph of radar fault causes is constructed.

Key words: radar fault cause text, named entity recognition, ALBERT, relationship extraction, knowledge graph

中图分类号: