一种Dual-LSTM混合模型的产线设备状态预测方法与应用  

Production Line Equipment States Prediction Method and Application Based on Dual-LSTM Hybrid Model

作　　者：马跃 李成蒙[1,2] 尹震宇 李明时 柴安颖[1,2] 赵志浩 MA Yue;LI Cheng-meng;YIN Zhen-yu;LI Ming-shi;CHAI An-ying;ZHAO Zhi-hao(University of Chinese Academy of Sciences,Beijing 100049,China;Shenyang Institute of Computing Technology,Chinese Academy of Sciences,Shenyang 110168,China)

机构地区：[1]中国科学院大学,北京100049 [2]中国科学院沈阳计算技术研究所,沈阳110168

出　　处：《小型微型计算机系统》2020年第12期2470-2474,共5页Journal of Chinese Computer Systems

基　　金：国家重点研发计划项目(2017YFE0125300)资助;辽宁省“兴辽英才计划”项目(XLYC1802056)资助。

摘　　要：智能化生产线设备健康状态、加工过程状态、产品信息等数据具有复杂化、多样化、大容量的特点,传统上对于设备的运行状况主要依靠人工经验来判断,不能及时有效地给出维护意见.针对上述问题,本文提出了一种基于Dual-LSTM(Long Short-Term M emory)混合模型的时序数据预测方法.首先建立LSTM预测模型对设备状态进行初步预测,然后针对多步预测过程中出现的误差"累积"问题,通过预测残差数据建立LSTM辅助模型对初步预测结果进行修正,最后采用循环迭代的方式实现了对数据的多步预测过程.通过与单LSTM模型进行实验对比,该方法在数据的单步预测和多步预测中的表现均优于单模型,验证了所提方法在时序数据预测上的准确性,为分析生产线整体的运行状态趋势提供了有效地判断依据.The data such as health status,processing status,and product information of the equipment in the intelligent production line have the characteristics of complexity,diversification and large-capacity.Traditionally,the operating status of equipment is mainly judged by artificial experience,and maintenance opinions cannot be given timely and effectively.To solve this problem,this paper proposes a time-series data prediction method based on Dual-LSTM(Long Short-Term Memory)hybrid model.Firstly,the LSTM prediction model is established to make a preliminary prediction of the equipment state.Then for the"accumulation"of errors in the multistep prediction process,an LSTM auxiliary model is established based on the prediction residual data to correct the preliminary prediction results.Finally,a multi-step prediction process for the data is implemented in a circular manner.By experimental comparison with single LSTM model,the method performs better than single model in single-step prediction and multi-step prediction of data.And,it verifies the accuracy of the proposed method in the prediction of time series data,and provides an effective judgment criterion for analyzing the running trend of the entire production line.

关 键 词：智能化生产线 Dual-LSTM混合模型 时序数据 多步预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]