摘要
针对复杂产品装配工艺设计过程中存在的编制效率低、质量波动大、自动化和智能化程度低等问题,提出一种面向语义的知识实例检索和复用的装配工艺快速编制方法。分析复杂产品装配工艺业务流程,建立基于知识检索层和知识实例层的双层装配工艺知识模型,从内容、类型、功能及应用方向对知识进行分类。在此基础上,针对已有的知识实例,提出基于Sentence-BERT句向量模型的工艺文档语义分析方法,结合余弦相似度算法给出了工艺知识的语义检索方法,实现知识在装配工艺设计流程中的快速复用。以某航天产品为例,开发航天产品装配工艺快速设计管理系统,构建装配工艺知识库,并在某制造企业上线运行,验证所提方法的可行性,使工艺编制效率大幅提升。
Aiming at the problems of low preparation efficiency,high quality fluctuation and low automation and intelligence in the process of complex product assembly process design,a fast preparation method for assembly process with semantic-oriented knowledge instance retrieval and reuse is proposed.A two-layer assembly process knowledge model based on the knowledge retrieval layer and the knowledge instance layer is established by analyzing the assembly process of complex product,and the knowledge is classified in terms of content,type,function and application direction.On this basis,a process document semantic analysis method based on the Sentence-BERT sentence embeddings model is proposed for the existing knowledge instances.The semantic retrieval method of process knowledge is given combined with the cosine similarity algorithm,and the rapid reuse of knowledge in the design process of assembly process is realized.Taking aerospace products as an example,a rapid design management system of aerospace product assembly process is developed,and a knowledge base for the assembly process is constructed and put into operation in an aerospace enterprise to verify the feasibility of the proposed method,resulting in a significant increase in process preparation efficiency.
作者
翟思宽
刘检华
庄存波
ZHAI Sikuan;LIU Jianhua;ZHUANG Cunbo(School of Mechanical Engineering,Beijing Institute of Technology,Beijing 100081,China;Yangtze River Delta Research Institute,Beijing Institute of Technology,Jiaxing 314000,Zhejiang,China)
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2024年第4期1332-1343,共12页
Acta Armamentarii
基金
国家自然科学基金项目(52005042)
北京理工大学青年教师学术启动计划项目(2021年)。
关键词
复杂产品
装配
知识模型
句向量
语义检索
知识库
complex product
assembly
knowledge model
sentence embeddings
senmantic retrieval
knowledge base
