船舶装备故障知识图谱构建及智能诊断方法

Construction of a Ship Equipment Fault Knowledge Graph and Intelligent Diagnosis Method

摘要

针对船舶装备故障诊断中可解释性弱、过度依赖专家经验、历史数据利用率低等问题,提出一种融合知识图谱与深度学习的船舶装备故障智能诊断方法.首先,以船舶维修手册、历史故障报告为数据源,采用自顶向下策略构建故障知识图谱模式层.其次,引入BERT-BiLSTM-CRF模型实现故障现象、原因、部位及维修方法的实体精准抽取,结合Jaccard相似度算法完成实体消歧与知识融合,基于Neo4j图数据库存储结构化故障知识.最后,集成朴素贝叶斯分类器与马尔可夫链归因算法,开发故障智能检索与根因追溯系统.实验结果表明,所提模型实体识别F₁值达83.2%,可高效挖掘船舶故障关联关系.案例验证显示,该方法能显著提升故障诊断效率与可解释性,降低对专家经验的依赖,为船舶装备智能运维提供技术支撑.

Abstract

To overcome the limitations of poor interpretability, strong dependence on expert experience, and insufficient utilization of historical data in ship equipment fault diagnosis, an intelligent diagnosis method integrating knowledge graphs and deep learning is proposed. Ship maintenance manuals and historical fault reports are first used as data sources to construct the schema layer of a fault knowledge graph using a top-down approach. A BERT-BiLSTM-CRF model is then employed to extract entities related to fault phenomena, causes, locations, and maintenance methods. To improve data consistency and quality, entity disambiguation and knowledge fusion are performed based on the Jaccard similarity algorithm, and the resulting structured fault knowledge is stored in the Neo4j graph database. Furthermore, an intelligent fault retrieval and root cause analysis system is developed by integrating a Naive Bayes classifier with a Markov chain attribution algorithm. Experimental results demonstrate that the proposed model achieves an F₁ score of 83.2% for entity recognition and effectively captures the associations among ship faults. Case studies further verify that the proposed method enhances both the efficiency and interpretability of fault diagnosis while reducing reliance on expert experience, providing effective support for the intelligent operation and maintenance of ship equipment.

关键词

船舶装备 / 故障诊断 / 知识图谱 / BERT-BiLSTM-CRF / 知识融合 / 归因分析

Key words

ship equipment / fault diagnosis / knowledge graph / BERT-BiLSTM-CRF / knowledge fusion / root-cause analysis

参考文献

[1] Hocek H, Yazır D, Aygün C, et al. The effect of failure on energy efficiency in maritime vessels autopilot systems[J]. Ocean and Coastal Management, 2024, 259: 107451.

[2] 文怡. 基于知识图谱与大模型的船舶问答系统研究与应用[D]. 成都: 电子科技大学, 2025: 56.

[3] 聂同攀, 曾继炎, 程玉杰, 等. . 面向飞机电源系统故障诊断的知识图谱构建技术及应用[J]. 航空学报, 2022, 43(08): 46-62.

[4] 吴闯, 张亮, 唐希浪, 等. . 航空发动机润滑系统故障知识图谱构建及应用[J]. 北京航空航天大学学报, 2024, 50(04): 1336-1346.

[5] Zhang Z J, Sun Y F, Zhang L S, et al. Enabling online search and fault inference for batteries based on knowledge graph[J]. BATTERIES-BASEL, 2023, 9(02): 124.

[6] Zhou B X, Gao D Q, Yan L C, et al. Research on key technologies for fault knowledge acquisition of power communication equipment[J]. Procedia Computer Science, 2021, 183: 479-485.

[7] 许驹雄, 李敏波, 刘孟珂, 等. . 发动机故障领域知识图谱构建与应用[J]. 计算机系统应用, 2022, 31(07): 66-76.

[8] 刘建湘, 陈晓慧, 刘海砚, 等. . 船舶活动知识图谱可视分析方法[J]. 计算机辅助设计与图形学学报, 2024, 36(08): 1194-1204.

[9] 司聿宣. 基于电子知识库的船舶故障诊断技术及应用[J]. 舰船科学技术, 2024, 46(16): 153-157.

[10] 邱凌, 张安思, 张羽, 等. . 面向无人机故障诊断的知识图谱构建应用方法[J]. 计算机工程与应用, 2023, 59(09): 280-288.

[11] 张栋梁, 周伟, 马刚, 等. . 面向水利防汛抢险的知识图谱构建与应用[J]. 水利学报, 2025, 56(03): 341-353.

[12] Chen J, Zhuang C, Shi J, et al. Risk factors extraction and analysis of Chinese ship collision accidents based on knowledge graph[J]. Ocean Engineering, 2025, 322: 120536.

[13] 涂芳, 周华涛. 基于人工智能的船舶故障检测结果智能推荐系统[J]. 舰船科学技术, 2024, 46(11): 173-176.