基于多维级联知识抽取的核电建造经验反馈图谱研究  

Research on Experience Feedback Graph of Nuclear Power Constructions Based on Multi-dimensional Cascading Knowledge Extraction

作　　者：李池璐 熊世权[1] 雷玮剑 刘疏桐 李英敏 易茜[1] 易树平[1] LI Chilu;XIONG Shiquan;LEI Weijian;LIU Shutong;LI Yingmin;YI Qian;YI Shuping(College of Mechanical and Vehicle Engineering,Chongqing University,Chongqing,400044;China Nuclear Power Engineering Co.,Ltd.,Shenzhen,Guangdong,518124)

机构地区：[1]重庆大学机械与运载工程学院,重庆400044 [2]中广核工程有限公司,深圳518124

出　　处：《中国机械工程》2025年第1期177-189,共13页China Mechanical Engineering

基　　金：国家重点研发计划(2020YFB1711700);中央高校基本科研业务费专项资金(2021CDJKYJH022)。

摘　　要：针对核电装备建造过程中经验反馈时效性弱、各参与方对经验知识认知不统一、经验反馈知识复用率低等问题,将经验知识化,构建统一的经验反馈图谱。分析了经验反馈信息的特征,构建了经验反馈关键知识域结构与图谱框架;提出的多维信息感知模型和改进PCNN关系抽取模型在核电企业数据集上取得了良好效果,并基于Neo4j图数据库生成了经验反馈图谱。研究结果表明该方法可完成核电产品的隐性知识深层次检索,辅助核电专业技术与管理人员快速解析、传递经验反馈知识,实现核电建造经验反馈沉淀应用模式从以文档为中心向以知识为中心的转变。Aiming at the problems of weak timeliness of experience feedback in the process of nuclear power equipment constructions,inconsistent perception of experience knowledge by each participant,and low reuse rate of experience feedback knowledge,the experience knowledge was intellectualized,and a unified experience feedback mapping was constructed.The characteristics of experience feedback information were analyzed,and the structure of key knowledge domains and mapping framework of experience feedback were constructed.The proposed multi-dimensional information perception model and improved PCNN relationship extraction model achieved good results on nuclear power enterprise dataset,and the experience feedback mapping was generated based on Neo4j graph database.The results show that the method may complete the deep-level retrieval of tacit knowledge of nuclear power products,assist the nuclear power professional technology and management personnels to quickly parse and transfer the experience feedback knowledge,and realize the transformation of the application mode of nuclear power construction experience feedback precipitation from document-centered to knowledge-centered.

关 键 词：核电装备 经验反馈 知识图谱 自然语言处理 

分 类 号：TL372[核科学技术—核技术及应用]