潜油电泵故障诊断知识图谱构建与增容研究  

Research on the Construction and Expansion of Knowledge Graph for Fault Diagnosis of Submersible Electric Pumps

作　　者：陆友明 袁向兵 闵哲 孔敏 Lu Youming;Yuan Xiangbing;Min Zhe;Kong Min(Marine Oil Production Plant,Shengli Oilfield Branch,Sinopec Group Company,Dongying,Shandong,China,257237)

机构地区：[1]中石化集团公司胜利油田分公司海洋采油厂,山东东营257237

出　　处：《仪器仪表用户》2024年第7期4-6,共3页Instrumentation

摘　　要：海上油田潜油电泵井具有数量多、产量高、修井成本高。现有的潜油电泵知识图谱故障诊断由于语料数据缺乏,在实际诊断中具有一定的局限性,与国际先进的检泵系统依然存在较大的差距。本文在语料数据增容、图谱本体重构、小样本语料回标等方面,通过对语料数据扩充、图谱结构的改进,实现潜油电泵故障诊断知识图谱的扩充增容。同时,本文从关系抽取、知识融合、知识补全等方面入手,通过文本相似度、皮尔逊相关系数,运用人工筛选、阈值法等多种方法,对图谱中的实体与关系进行融合、消歧、补全,运用专家知识对知识图谱进行进一步审核与纠错,提升图谱质量,增强图谱在实际潜油电泵故障诊断中的应用能力。Marine oilfield submersible pump wells are numerous,productive,and costly to repair.Existing fault diagnosis methods for submersible pumps based on knowledge graphs have certain limitations in practical applications due to the scarcity of corpus data,resulting in a significant gap compared to internationally advanced pump inspection systems.This article focuses on enhancing and expanding the knowledge graph for fault diagnosis in submersible pumps by augmenting corpus data,reconstructing the ontology of the graph,and annotating small-sample corpora.Through the expansion of corpus data and the improvement of the graph structure,the coverage and accuracy of the knowledge graph are significantly improved.Additionally,this paper explores relation extraction,knowledge fusion,and knowledge completion.By utilizing techniques such as text similarity,Pearson correlation coefficient,manual screening,and threshold methods,entities and relationships within the graph are integrated,disambiguated,and completed.Expert knowledge is also employed to further review and correct the knowledge graph,enhancing its quality and enhancing its applicability in practical fault diagnosis for submersible pumps.

关 键 词：知识图谱 潜油电泵 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]