农用电机故障知识图谱的构建与应用  

Construction and application of a knowledge graph for agricultural motor fault diagnosis

作　　者：黄友锐[1,2] 荣雪 徐善永 韩涛[1] 宋奇[1] HUANG Yourui;RONG Xue;XU Shanyong;HAN Tao;SONG Qi(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China;School of Electrical and Optoelectronic Engineering,West Anhui University,Lu’an 237012,China)

机构地区：[1]安徽理工大学电气与信息工程学院,淮南232001 [2]皖西学院电气与光电工程学院,六安237012

出　　处：《农业工程学报》2025年第6期216-226,共11页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金项目(61772033);安徽省高校协同创新项目(GXXT-2023-068)。

摘　　要：随着农业机械化和智能化的发展,农用电机作为主要动力来源,显著提升了农业生产效率,并推动农业向绿色化、智能化和高效化方向发展。为解决农用电机的故障可能导致的作物收获延误、经济损失和安全隐患。尽管传统机器学习和深度学习方法在电机故障诊断中展现出潜力,但其解释性不足和高成本限制了其广泛应用,亟需开发一种能够有效挖掘关键信息以指导故障维修的方法。该研究提出了一种基于多源数据异构融合的农用电机故障诊断知识图谱系统,旨在提升故障诊断效率和降低维修成本。通过实体识别与关系抽取,将非结构化数据转化为结构化数据,使用BERTBiLSTM-CRF模型进行实体识别,模型在实体识别任务中的准确率、召回率、F1值分别达到0.952 3、0.915 7、0.933 6,结合模式匹配与正则表达式进行关系抽取,并嵌入GPT模型构建智能问答系统,采用Neo4j图数据库存储电机故障知识,最终形成包含702个故障实体的图谱。研究表明,农用电机故障诊断知识图谱系统能够提升故障诊断效率,降低维修成本,增强农业生产的智能化水平,为农用电机故障诊断提供了一种高效、智能的解决方案,具有重要的应用前景和研究价值。With the increasing push toward mechanization and digital intelligence in modern agriculture,agricultural motors have become vital components,underpinning productivity by powering various essential equipment.As the sector advances toward greener,smarter,and more efficient operations,the demand for reliable motor performance becomes ever more critical.However,faults in these motors can lead to significant setbacks,disrupting crucial agricultural timelines,inflicting economic losses,and posing potential safety risks.Traditional diagnostic methods,including machine learning and deep learning approaches,though effective,often face challenges in interpretability,scalability,and adaptability to the specific requirements of agricultural systems.These limitations hinder their widespread application in real-world agricultural environments,where conditions can vary greatly,and system transparency is paramount for effective troubleshooting.To address these gaps,this study proposes a novel,knowledge-driven approach by constructing a dedicated knowledge graph tailored for agricultural motor fault diagnosis.The knowledge graph framework integrates information from multiple heterogeneous data sources,such as maintenance logs,technical manuals,and fault records,transforming unstructured textual data into structured,semantically rich,and queryable knowledge.This is achieved through the deployment of advanced natural language processing techniques,specifically leveraging the BERT-BiLSTM-CRF model for entity recognition,which allows the accurate extraction of key faultrelated entities,such as components,fault types,and causes,even from complex and technical descriptions.In addition,relationship extraction between entities is accomplished through a combination of pattern matching and regular expression techniques,capturing the intricate relationships between components and fault causes in a manner that enhances the coherence and utility of the knowledge graph.This structured approach facilitates a more comprehensive understanding of

关 键 词：知识图谱 农用电机 故障诊断 知识抽取 智能问答 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]