【作者】 郭静宇； 王雅琳； 李淑贤；

【Author】 Guo Jingyu;Wang Yalin;Li Shuxian;Central South University, School of Automation;

【机构】 中南大学自动化学院；

【摘要】 产品质量分析是制造业生产的重要环节,包括关键工艺参数状态分析、产品质量缺陷溯因分析及关键工艺参数优化调整等环节。知识图谱为制造业产品质量分析的知识自动化提供了理论及技术支持。但传统知识图谱侧重于定性知识表示,忽略了生产过程关键参数参考值、实时值以及知识可信度等定量知识,且难以描述多种原因共同作用导致某一缺陷或针对某一缺陷需同时进行多种操作的复杂关系,由此本文提出了基于虚拟节点与定量-定性知识融合的产品质量分析知识图谱构建方法,首先,基于产品质量分析相关知识的特点及定量、定性知识融合表达需求,构建了产品质量分析本体元模型,并在此指导下完成知识图谱三元组类别初步设定;其次,提出了基于关系属性的知识可信度表达方式,引入了虚拟触发节点以表示多原因同时作用或同时执行多种操作的复杂关系;最后,提出了本体引导下考虑定量知识与虚拟节点的知识推理机制,并以注塑产品缺陷溯因及工艺调整为例进行应用验证。更多还原

【Abstract】 Product quality analysis is an important link in manufacturing production, including key process parameter status analysis, product quality defect traceability analysis, and key process parameter optimization and adjustment. The knowledge graph provides theoretical and technical support for the knowledge automation of manufacturing product quality analysis. However, traditional knowledge graphs focus on qualitative knowledge representation, neglecting quantitative knowledge such as reference values, real-time values, and knowledge credibility of key parameters in the production process. Moreover, it is difficult to describe the complex relationship between multiple reasons that cause a certain defect or require multiple operations to be carried out simultaneously for a certain defect. Therefore, this article proposes a method for constructing product quality analysis knowledge graphs based on the fusion of virtual nodes and quantitative qualitative knowledge. Firstly, Based on the characteristics of product quality analysis related knowledge and the fusion expression requirements of quantitative and qualitative knowledge, a product quality analysis ontology metamodel was constructed, and under this guidance, the preliminary setting of knowledge graph triplet categories was completed; Secondly, a knowledge credibility expression method based on relationship attributes was proposed, introducing virtual trigger nodes to represent complex relationships where multiple reasons act or perform multiple operations simultaneously; Finally, a knowledge inference mechanism that considers quantitative knowledge and virtual nodes under the guidance of ontology was proposed, and application validation was conducted using injection molding product defect tracing and process adjustment as examples.更多还原