本文针对烟草业务系统日常运维中,对生产异常,特别是物料损耗异常发现难、追溯排查难的问题,设计并实现了一种基于双向长短期记忆模型(bi-directional long short-term memory,Bi-LSTM)和自注意力机制的损耗异常分析模型.以烟丝损耗异常检查为例介绍该模型,以卷包系统的时序剔除数据为输入,判断原材料损耗是否存在异常.该分析模型可用于烟草业务系统日常监控运维,自动识别各生产阶段物料损耗异常,并通过注意力权重从空间和时间维度解释分析结果,为人工排查提供先验,辅助生产管理,提升运维系统的智能化.
Application of Self Attention Model in Operation and Maintenance of Tobacco Business System
In the daily operation and maintenance of the tobacco business system,it is difficult to find and trace production abnormalities,especially material loss abnormalities.A loss anomaly analysis model based on bi-directional long short-term memory(Bi-LSTM)and self attention mechanism is designed and implemented.This paper introduces the model by taking tobacco loss anomaly inspection as an example,Using the time series elimination data of the packaging system as input,determine whether there is any abnormality in raw material loss.The analysis model can be used for the daily monitoring,operation and maintenance of the tobacco business system,automatically identify the abnormal material loss in each production stage,and interpret the analysis results from the space and time dimensions through the attention weight,providing a priori for manual troubleshooting,assisting production management,and improving the intelligence of the operation and maintenance system.
tobaccobusiness systemintelligent operations and maintenanceself attention
NETL
NSTL
万方数据
