【作者】 杨春晖；

【导师】 吴国伟；

【作者基本信息】 大连理工大学 ， 软件工程（专业学位）， 2021， 硕士

【摘要】 为保障石油天然气运输安全,我国展开了针对油气管道视频、图像、雷达等各类数据的分析研究工作。然而现阶段油气管道数据的管理组织方式刻板繁琐,存在数据管理低效,再利用困难,数据关系分析、历史追溯操作繁琐等问题。应中石化大连研究院需求,本文旨在开发一个满足自动化数据注入、数据清洗、数据质量分析需求,并提供风险数据分析和定制化检索服务的石油天然气管道数据管理分析系统。本文的重点在于将知识图谱概念与油气管道数据相结合,对数据进行内容划分与知识抽取,分别构建了以数据实体为中心的数据知识图谱和巡检事件为中心的风险事件知识图谱。利用图结构特性,实现数据关联搜索、数据质量分析、数据一致性检验等功能。结合改进的En-TransE知识表示学习模型,实现对管道风险事件的推理和预测,探究了知识图谱学习在油气管道安全领域应用的可行性。同时本文设计了一种基于混合指纹的相似度算法实现定制化的图像检索功能,提升系统检索的准确度。系统整体采用B/S的架构模式与前后端分离的开发策略。依据需求分析,将系统划分为数据统计分析模块、管道数据注入模块、数据关系图谱模块、数据检索模块、系统用户管理模块、系统监管模块六个部分。此外系统提供Nginx+redis访问拦截以及Jwt权限认证等安全保障措施。系统前端采用Vue框架配合element ui模块化开发,并结合Echarts实现各类数据的图形化展示。后端业务服务主要采用SpringBoot,结合MVC三层结构实现服务端开发。数据库以Neo4j图数据库为主体辅以Mysql关系型数据库,通过路径存储的方式与FastDFS分布式文件系统关联实现大量非结构化数据的存储。经过部署与测试验收证明,该系统满足实际场景下的业务需求,所提供的数据自动注入、数据清洗、图谱分析、图像检索等功能具有实用价值,为油气管道数据的组织管理、数据分析等科研工作提供了巨大帮助。更多还原

【Abstract】 In order to ensure the safety of oil and gas transportation,China has carried out the analysis and research work on oil and gas pipeline video,image,radar and other data.However,at this stage,the management of oil and gas pipeline data organization is cumbersome.There are problems such as inefficient data management,reuse difficulties,data relationship analysis,and cumbersome historical retrospective operation.Therefore,in response to the needs of Sinopec Dalian Research Institute,this paper aims to develop an management and analysis system of oil and gas pipeline data that meets the needs of automated data injection,data cleaning,data quality analysis,and provides risk data analysis and customized retrieval services.The focus of this paper is to combine the concept of knowledge graph with oil and gas pipeline data,divide the data content and extract knowledge,and construct the knowledge graph of risk events centered on the data entity and the inspection event.Using the characteristics of graph structure,the functions of data association search,data quality analysis and data consistency test are realized.Combined with the improved En-TransE knowledge representation learning model,the reasoning and prediction of pipeline risk events are realized,and the feasibility of knowledge graph learning in the field of oil and gas pipeline safety is explored.In addition,a similarity algorithm based on mixed fingerprint is designed to realize the custom image retrieval function and improve the accuracy of system retrieval.The system as a whole adopts a B/S architectural model with a development strategy that separates the front and back ends.According to the requirements analysis,the system is divided into six parts: data statistical analysis module,pipeline data injection module,data graph module,data retrieval module,system user management module and system supervision module.In addition,the system provides Nginx + Redis access interception and Jwt permission authentication and other security measures.The front end of the system is modularly developed with a Vue framework in conjunction with the elementUI,and the graphical presentation of various types of data is achieved in conjunction with Echarts.Back-end business services mainly use SpringBoot,combined with MVC three-tier structure to achieve service-side development.The database is supplemented by the Neo4j graph database with mysql relational database,which is associated with fastDFS distributed file system by way of path storage to realize the storage of a large amount of unstructured data.After deployment and test acceptance,the system meets the business needs of the actual situation,provides the functions of automatic data injection,data cleaning,graph analysis,image retrieval and so on,and provides great convenience for the organization and management of oil and gas pipeline data and data analysis.更多还原