基于知识图谱的钻井顶部驱动装置故障智能诊断方法  

Intelligent fault diagnosis method of drilling top drive based on knowledge graph

作　　者：陈冬[1,2,3] 肖远山 尹志勇 张彦龙 叶智慧[5] CHEN Dong;XIAO Yuanshan;YIN Zhiyong;ZHANG Yanlong;YE Zhihui(College of Petroleum Engineering,China University of Petroleum-Beijing,Beijing 102249,China;Research Center for Intelligent Drilling&Completion Technology and Equipment//China University of Petroleum-Beijing,Beijing 102249,China;National Key Laboratory of Petroleum Resources and Engineering//China University of Petroleum-Beijing,Beijing 102249,China;CNPC Engineering Technology R&D Company Limited,Beijing 102206,China;College of Safety and Ocean Engineering,China University of Petroleum-Beijing,Beijing 102249,China)

机构地区：[1]中国石油大学(北京)石油工程学院,北京市102249 [2]智能钻完井技术与装备研究中心·中国石油大学(北京) [3]油气资源与工程全国重点实验室·中国石油大学(北京) [4]中国石油集团工程技术研究院有限公司 [5]中国石油大学(北京)安全与海洋工程学院

出　　处：《天然气工业》2025年第2期125-135,共11页Natural Gas Industry

基　　金：中国石油天然气集团有限公司科学研究与技术开发项目“钻井安全控制软件”(编号:2024ZZ46-05);中国石油天然气集团有限公司—中国石油大学(北京)战略合作科技专项项目“物探、测井、钻完井人工智能理论与应用场景关键技术研究”(编号:ZLZX2020-03);中海油科技项目“海上钻井风险智能预警、调控及协同优化技术研究”(编号:CCL2022RCPS2009XNN)。

摘　　要：钻井顶部驱动装置结构复杂、故障类型多样,现有的故障树分析法和专家系统难以有效应对复杂多变的现场情况。为此,利用知识图谱在结构化与非结构化信息融合、故障模式关联分析以及先验知识传递方面的优势,提出了一种基于知识图谱的钻井顶部驱动装置故障诊断方法,利用以Transformer为基础的双向编码器模型(Bidirectional Encoder Representations from Transformers,BERT)构建了混合神经网络模型BERT-BiLSTM-CRF与BERT-BiLSTM-Attention,分别实现了顶驱故障文本数据的命名实体识别和关系抽取,并通过相似度计算,实现了故障知识的有效融合和智能问答,最终构建了顶部驱动装置故障诊断方法。研究结果表明:①在故障实体识别任务上,BERT-BiLSTM-CRF模型的精确度达到95.49%,能够有效识别故障文本中的信息实体;②在故障关系抽取上,BERT-BiLSTM-Attention模型的精确度达到93.61%,实现了知识图谱关系边的正确建立;③开发的问答系统实现了知识图谱的智能应用,其在多个不同类型问题上的回答准确率超过了90%,能够满足现场使用需求。结论认为,基于知识图谱的故障诊断方法能够有效利用顶部驱动装置的先验知识,实现故障的快速定位与智能诊断,具备良好的应用前景。The drilling top drive is structurally complex with multiple fault types,making it difficult for existing fault tree analysis methods and expert systems to effectively address complex and diverse field situations.For this reason,this paper proposes a fault diagnosis method of drilling top drive based on knowledge graph,by making use of the advantages of knowledge graph in the integration of structured and unstructured information,fault mode correlation analysis,and prior knowledge transfer.The hybrid neural network models(BERT-BiLSTM-CRF and BERT-BiLSTM-Attention)are established by using the Bidirectional Encoder Representations from Transformers(BERT),and they achieve named entity recognition and relationship extraction of top drive fault text data.In addition,the effective integration and intelligent question and answering of fault knowledge are realized through similarity calculations.Finally,the fault diagnosis method of drilling top drive is developed.The following results are obtained.First,as for fault entity recognition,the BERT-BiLSTM-CRF model can effectively recognize the information entities in fault texts with an accuracy of 95.49%.Second,as for fault relationship extraction,the BERT-BiLSTM-Attention model accurately establishes the relationship edges of the knowledge graph with an accuracy of 93.61%.Third,the question-answering system developed in this paper realizes the intelligent application of the knowledge graph,and its answer accuracy rate to multiple various questions exceeds 90%,meeting the on-site application needs.In conclusion,the fault diagnosis method based on knowledge graph can effectively utilize prior knowledge of drilling top drive to achieve efficient and accurate fault diagnosis,and it is conducive to improving the efficiency of top drive fault diagnosis,showing promising application prospects.

关 键 词：钻井装备 顶部驱动装置 故障诊断 深度学习 知识图谱 自然语言处理 命名实体识别 智能问答系统 

分 类 号：TE924[石油与天然气工程—石油机械设备]