文献类型：期刊文献

中文题名：加氢裂化装置工艺操作知识图谱构建与应用

英文题名：Construction and Application of Knowledge Graph for Process Operation of A Hydrocracking Unit

作者：陈思敏[1];汤淳[1];汪张扬[1];李润泽[1];曹跃[1,2]

机构：[1]华东理工大学信息科学与工程学院,上海;[2]华东理工大学能源化工过程智能制造教育部重点实验室,上海

年份：2026

期号：2

起止页码：72

中文期刊名：科学技术创新

外文期刊名：Scientific and Technological Innovation

基金：国家自然科学基金(62203173);华东理工大学大学生创新创业训练计划项目(X202410251199)。

语种：中文

中文关键词：加氢裂化装置;工艺操作;命名实体识别;实体关系抽取;知识图谱构建;知识图谱应用

外文关键词：hydrocracking unit;process operation;named entity recognition;entity relation extraction;knowledge graph construction;knowledge graph application

摘要：加氢裂化装置是中重质油高效转化的核心生产装置,具有生产流程长、多变量耦合、工艺操作规范复杂等特点,操作知识学习成本高,专业性要求强,难以快速普及和有效运用。基于生产报表、工艺卡片、操作技术规程以及HAZOP文档等,提出了一种结合规则与机器学习的混合策略的命名实体识别方法以及利用遍历查找和分类建立的多项实体关系抽取方法,解决了知识文本难以识别实体和抽取关系的问题,构建了加氢裂化装置工艺操作知识图谱。经加氢裂化装置操作实际案例验证,说明了所构建知识图谱解决实际案例的有效性和可行性,为加氢裂化装置工艺操作提供结构化和可视化工具。
Hydrocracking is a core production unit for the efficient conversion of medium and heavy oils.It has the characteristics of long production processes,multivariable coupling,and complex process operation specifications.The cost of learning operational knowledge is high and professional requirements are strong.These factors make it difficult to quickly popularize and effectively apply the knowledge.Based on production reports,process cards,operating technical procedures,and HAZOP documents,this paper proposes a named entity recognition method that combines a hybrid strategy of rules and machine learning,as well as a multiple entity relationship extraction method established by traversal search and classification.It solves the problem that knowledge texts are difficult to identify entities and extract relationships.A knowledge graph is constructed for the process operation of a hydrocracking unit.Verified by actual cases of hydrocracking unit operations,it illustrates the effectiveness and feasibility of the knowledge graph constructed in this paper in solving actual cases,and provides a structured and visual tool for the process operation of the hydrocracking unit.