摘要
针对零件在线测量过程中多种测量仪器无法协同存储、仪器内数据查询聚合困难等问题,对多工序测量仪器集群的数据分布式存储和共享方法进行了研究。首先,在多个嵌入式测量仪器中,移植非关系型数据库集群对零件数据进行了分布式存储,简化了零件数据存储模型;然后,引入标签感知分片(tag-aware sharding),实现了零件不同工序测量数据的分类存储目的;最后,使用多个搭载嵌入式Linux系统的树莓派3B+开发板作为测量仪器系统平台,搭建了基于MongoDB数据库的在线测量分布式存储集群,通过大量测试数据验证了该集群在多节点存储时数据分布均匀、性能稳定;将存储集群与企业制造执行系统(MES)进行了对接,其可以实时监控分析零件测量数据,快速汇总所有测量节点内的工序数据并生成报表。研究结果表明:当集群中数据量达到7.2×10^(5)条时,单节点查询响应速度稳定在125 ms~208 ms范围内,相比哈希分片,其最高提升了88.15%;多节点协同查询响应速度为1308 ms,相比“升序键+搜索键”方案,其提升了了61.64%;多个测量节点内,1×10^(5)个零件数据聚合统计仅需5 s左右。该存储集群可以监控零件生产情况,在减少零件制造误差、提高制造效率和质量方面具有重要作用。
In order to solve the problem of data association and aggregation query between multiple measuring instruments during the online measurement process for parts,the distributed storage and sharing methods of measuring were studied.Firstly,non-relational database was used in multiple measuring instrument cluster,which realized distributed storage of measuring data and simplified parts data storage model.Then,tag aware sharding model was used to classify and store the measurement data of different processes.Finally,multiple Raspberry Pi 3 Model B+development boards with embedded Linux were used as the system platforms for the measurement instruments,and MongoDB database was used as online measurement distributed data storage to verify the uniformity and stability of distributed data storage on each node using a large amount of test data.The cluster had been integrated with manufacturing execution system(MES),which could monitor and analyze the measuring data of parts in real time,and quickly summarize the working procedure data in all measuring nodes to generate the report forms.The research results show that the single-node query response speed is stable in the range of 125 ms to 208 ms when the data volume reaches 7.2×10^(5) in the cluster,which is 88.15%higher than that of the hashed sharding model,at the same time,the response speed of multi-node cooperative query is 1308 ms,which is 61.64%faster than the scheme of“ascending key+searching key”.Aggregating and summarizing 1×10^(5) parts data within multiple measurement nodes takes approximately 5 s.This storage cluster can monitor the production status of parts and plays an important role in improving manufacturing efficiency and quality.
作者
王佺珅
张爱梅
WANG Quanshen;ZHANG Aimei(School of Mechanical and Power Engineering,Zhengzhou University,Zhengzhou 450001,China)
出处
《机电工程》
北大核心
2024年第1期149-157,共9页
Journal of Mechanical & Electrical Engineering
基金
国家重点研发计划项目(2018YFB0104101-5)。
