空军工程大学学报2026，Vol.27Issue(3)：71-82,12.DOI:10.3969/j.issn.2097-1915.2026.03.008
一种基于大模型的装备知识图谱本体自动化构建方法
Research on an Automated Method for Constructing Equipment Knowledge Graph Ontology Based on Large Models
摘要
Abstract
With the continual increase of complexity of equipment maintenance support tasks,construction of high-quality domain ontologies has become a crucial prerequisite between knowledge sharing and intelli-gent reasoning.Aimed at the problems that the traditional ontology construction methods are dependent on manual modeling,are low in efficiency and difficult to meet the needs of large-scale corpora,meanwhile,ex-isting large model-based ontology construction methods are unstable in generation,poor in semantic consis-tency,and lack of domain adaptation,this paper propose an automated equipment knowledge graph ontolo-gy construction method based on large models.This method is met with a challenge in the equipment main-tenance domain,i.e.highly specialized terminology,complex relationship semantics,and sparse cross-docu-ment associations.The paper employs a three-stage process of block parallel extraction(Map)-local merge and redundancy removal(Reduce)-global context refinement(Refine).Through a context-aware relationship completion mechanism and a domain-adaptive configuration module,the automatic transforma-tion is completed from the unstructured text to a highly connected ontology.The experiments show that the OLMM-Refine method can be used to effectively improve the completeness and logical consistency of relationship extraction in ensuring coverage of entities and attributes,reducing the proportion of isolated nodes and forming a knowledge graph with stronger connectivity and more reasonable semantic structure.Compared to the existing large language model-driven ontology construction frameworks,the OLMM-Re-fine is superior at performance in terms of completeness,consistency and accuracy,and the efficiency is within an acceptable range.The results indicate that the OLMM-Refine can take account of the generality and the domain adaptability,providing a new approach and pathway for the construction and application of knowledge graphs in equipment maintenance support.关键词
知识图谱/本体构建/大语言模型/无人机维修/装备保障Key words
knowledge graph/ontology construction/large language model/UVA maintenance/equipment support分类
航空航天引用本文复制引用
李乐源,崔利杰,谢小月,周中良,唐希浪..一种基于大模型的装备知识图谱本体自动化构建方法[J].空军工程大学学报,2026,27(3):71-82,12.基金项目
国家自然科学基金青年基金(NFSC72201276) （NFSC72201276）
中国博士后科学基金(2025M774456) （2025M774456）
