基于深度学习的工艺知识图谱构建及其应用  

Construction and application of process knowledge graph based on deep learning

作　　者：王宇东 张琦 马雅丽[1] 王智[1] WANG Yudong;ZHANG Qi;MA Yali;WANG Zhi(School of Mechanical Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]大连理工大学机械工程学院,辽宁大连116024

出　　处：《机电工程》2024年第12期2220-2231,共12页Journal of Mechanical & Electrical Engineering

基　　金：国防科工局车用动力专项(VTPD-3202)。

摘　　要：针对现有的零件工艺知识分散度高、结构性弱、复用性差等问题,提出了一种基于深度学习技术的工艺知识图谱构建和工艺重用方法。首先,分析了工艺知识结构,并建立了工艺知识图谱模式层;其次,搭建了深度学习知识抽取算法,并以工艺知识图谱模式层作为数据模式抽取了工艺知识,建立了工艺知识图谱的数据层;然后,基于图神经网络深度学习算法,搭建了工艺知识推理模型,将其作为工艺推荐基础;最后,搭建了零件工艺知识图谱可视化系统,并以行星架类零件为例,验证了工艺知识的检索和推荐功能。研究结果表明:该方法在工艺知识上的识别准确率达到了80%以上,工艺推荐准确率达到了70%以上,相比以往模型有所提高,证明了该方法在工艺知识图谱自动化构建和工艺重用上的有效性和可行性。In response to the issues of high dispersion,weak structural integrity,and poor reusability of existing part process knowledge,a method was proposed to construct a process knowledge graph based on deep learning technology for process knowledge reuse.Firstly,the structure of process knowledge was analyzed and a model layer of the process knowledge graph was established.Secondly,a deep learning knowledge extraction algorithm was built,the pattern layer was used as the data pattern to guide the process knowledge extraction,and the process knowledge map data layer was established.Then,a process knowledge inference model was constructed based on the graph neural network deep learning algorithm,serving as the foundation for process recommendations.Finally,extensive experiments were conducted to validate the proposed method.The part process knowledge graph visualization system was implemented,providing an intuitive interface for users to retrieve and recommend process knowledge effectively.The planetary gear parts were chosen as a representative example for the validation of the process knowledge retrieval and recommendation functionalities.The research results demonstrate that the process knowledge identification accuracy of the method reaches over 80%,and the process recommendations accuracy exceeds 70%.Comparing with the previous models and traditional methods,the proposed approach shows significant improvements in automating the construction of process knowledge graphs and enhancing process reuse capabilities.This research contributes to the advancement of manufacturing technologies by providing a more efficient and reliable solution for managing and reusing process knowledge.The findings of this study not only validate the effectiveness and feasibility of the proposed method but also pave the way for future research in the field of process knowledge management and utilization.

关 键 词：工艺知识结构 深度学习技术 工艺重用 知识抽取 知识推理模型 图神经网络 模式层和数据层 

分 类 号：TH16[机械工程—机械制造及自动化] TP391[自动化与计算机技术—计算机应用技术]