基于注意力SRU的离散制造过程生产异常分析  

Production Anomaly Analysis of Discrete Manufacturing Process Based on Attention SRU

作　　者：马云骁 郭宇[1] 王胜博 MA Yunxiao;GUO Yu;WANG Shengbo(Collegeof Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《机械制造与自动化》2024年第4期48-53,共6页Machine Building & Automation

基　　金：国防基础科研项目(JCKY2018203A001,JCKY2019204A004)。

摘　　要：为了判断生产异常对生产过程产生何种影响,在对离散制造过程中产生的物联数据进行详细分析的基础上,设计一种基于注意力SRU神经网络的生产任务剩余完工时间预测模型,通过并行化的特征提取过程并结合注意力机制给予不同时刻信息不同权重来对生产任务的延期交付时间进行预测,达到了解量化生产异常对于生产过程影响程度的目的,且有效提高了算法效率和准确率,从而帮助车间工作人员进行后续决策管控工作。In order to judge the impact of production anomalies on production process,a prediction model for the remaining completion time of production tasks based on attention SRU neural network was designed after detailed analysis of the Internet of Things data generated in discrete manufacturing process.Through the parallel feature extraction process and by combination with the attention mechanism,different weights of information at different moments were given to predict the delayed delivery time of production tasks,which achieves the purpose of quantifying the impact of production anomalies on production process,and effectively improves the efficiency and accuracy of the algorithm,so as to help the workshop staff carry out subsequent decision-making control work.

关 键 词：离散制造过程 制造物联数据 生产异常分析 注意力机制 SRU神经网络 

分 类 号：TH164[机械工程—机械制造及自动化]