摘要
针对离散制造车间生产瓶颈预测的实时性差、无法表征车间实际状态的问题,结合定位系统在车间实时监控领域的优势,提出了一种实时定位环境下的生产瓶颈预测方法。首先,围绕实时定位环境下离散制造车间的信息特征,定义并量化表示了车间生产瓶颈;其次,用历史的瓶颈指数序列作为主要输入,在制品、转运车等其他车间生产要素的状态信息作为辅助输入,使用基于长短期记忆(Long Short-Term Memory,LSTM)神经网络预测瓶颈指数;最后,以某航天产品机加车间为例,预测了该车间在给定时延内全部工位的瓶颈指数,以此为依据,瓶颈工位的平均预测准确率为96.73%。实验结果证明了本文方法的有效性。
Considering the poor performance on predicting real-time bottleneck in discrete manufacturing workshops,which is inability to characterize the actual state of workshops,a method for bottleneck prediction in real-time location environment is proposed combined with the advantages of locating system in the workshop real-time monitoring.Firstly,the information characteristics in discrete manufacturing workshops with positioning system are used to defined and quantified the bottleneck.Secondly,using long-short term memory(LSTM)neural network to predict the bottleneck indexes while the historical bottleneck index sequence is used as the main input,and the state information of other factors,such as working-in-processes and transfer vehicles are used as auxiliary.Finally,taking an aerospace machining workshop as an example,the bottleneck indices of all stations in a given period was predicted.Based on this,the average prediction accuracy of the bottleneck stations was 96.73%.The experimental results prove the effectiveness of the method.
作者
杨昊龙
郭宇
方伟光
崔世婷
YANG Hao-long;Guo Yu;FANG Wei-guang;CUI Shi-ting(College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)
出处
《组合机床与自动化加工技术》
北大核心
2020年第10期176-180,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
国防基础科研(JCKY2018605C003、JCKY2017203B071、JCKY2017203C105)。
关键词
实时定位
离散制造车间
生产瓶颈
长短期记忆
real-time location
discrete manufacturing workshops
bottleneck
long-short term memory
