基于特征选择和SSA-LSTM的车间订单剩余完工时间预测  

Workshop Order Remaining Completion Time Prediction Based on the Feature Selection and SSA-LSTM

作　　者：朱海华[1] 王健杰 李霏 刘长春 蔡祺祥[1] ZHU Haihua;WANG Jianjie;LI Fei;LIU Changchun;CAI Qixiang(College of Mechanical&Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016;Beijing Institute of Electronic System Engineering,Beijing 100854)

机构地区：[1]南京航空航天大学机电学院,南京210016 [2]北京电子工程总体研究所,北京100854

出　　处：《机械工程学报》2024年第20期388-402,共15页Journal of Mechanical Engineering

基　　金：国家重点研发计划(2021YFB1716304);中央高校基本科研基金(NT2021021);成飞-南航“智汇蓝天”工程实践计划六期(CF202206)资助项目。

摘　　要：客户多变的需求和激烈的市场竞争使得车间产品生产种类不断增加、批量不断减小、交货期不断缩短,导致业务流程和制造系统环境愈发复杂与不稳定,对车间生产管控能力提出了更高的要求。订单剩余完工时间(Order remaining completion time,ORCT)精准预测可以及时发现生产计划交期波动,为后续优化车间生产调度计划、改进工艺流程提供判断依据。针对传统数据分析方法在复杂制造环境下大规模制造数据利用分析能力较弱的问题,提出一种大数据驱动的车间ORCT预测方法。基于皮尔森相关系数和正则化的特征选择算法从海量的制造数据中挑选出关键数据;设计一种基于改进麻雀搜索算法-长短期记忆神经网络模型和注意力机制的预测方法实现车间ORCT的快速、精准预测,其中的预测模型以长短期记忆神经网络为结构基础,并使用改进麻雀搜索算法对模型的超参数进行优化;最后,以典型机加车间为对象进行实例验证,通过与三种常用预测方法进行对比,试验结果表明所提方法在准确性和效率上均有明显优势。The ever-changing demands of customers and fierce market competition have led to an increase in the number of products produced in the workshop,a decrease in batch sizes and a shortening of delivery times,which has led to an increasingly complex and unstable business process and manufacturing system environment,placing greater demands on the workshop's production control capabilities.Accurate forecasting of order remaining completion time(ORCT)allows for the timely detection of production schedule fluctuations and provides a basis for subsequent optimisation of the workshop production scheduling plan and process improvements.A big data-driven workshop ORCT prediction method is proposed to address the weak ability of traditional data analysis methods to utilise large-scale,long-time series manufacturing data in the complex manufacturing environment.A feature selection algorithm based on Pearson's correlation coefficient and regularisation is used to select key data from a huge amount of manufacturing data;a prediction method based on the improved sparrow search algorithm-long and short-term memory neural network model and the attention mechanism is designed to achieve fast and accurate prediction of ORCT in the workshop,where the prediction model is supported by the long and short-term memory neural network structure,which is optimised by the improved sparrow search algorithm;finally,a typical machine workshop is used as an example for verification,by comparing three common prediction methods,the experimental results show that the proposed method has obvious advantages in terms of accuracy and efficiency.

关 键 词：订单剩余完工时间 数据驱动 特征选择 预测模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]