摘要
针对制造业流程工艺质量指标数据波动性强、影响因素繁多,使用传统预测模型难以挖掘其隐含规律实现高精度预测的问题,提出了一种基于VMD-Informer的深度学习质量指标预测模型。首先筛选与质量指标相关的流程工艺参数;接着使用变分模态分解(VMD)将质量指标数据集分解为模态分量与误差项;然后筛选与各分量具有相关性的工艺指标作为输入矩阵;最后使用Informer模型对各分量及误差项分别预测并叠加得到最终预测值。选取国内某制造业企业生产数据,对不同质量指标进行预测,并与LSTM模型和改进前的Informer模型的预测效果进行对比。结果表明:所提的VMD-Informer模型预测误差更小、可决系数较大,预测更为精准,可为制造业企业实现质量预测提供有效方法,并为企业及时调整生产方案提供思路。
To address the problem of high volatility of the process quality indicators data and many complicating influencing factors in the manufacturing industry and the difficulty in mining the hidden laws of the traditional pre‐diction model to achieve high-precision prediction,a prediction model of deep Xi quality index based on VMD-Informer was proposed.Firstly,the process parameters related to the quality indicators were screened.Then,Variational Modal Decomposition(VMD)was used to decompose the quality index dataset into modal compo‐nents and error terms.Thereafter,the process indicators that were related to each component were selected as the input matrix.Finally,the Informer model was used to predict and superimpose the final predicted value of each component and error term.The production data of a domestic manufacturing enterprise were selected to predict different quality indicators,and the prediction effect was compared with that of the LSTM model and the improved Informer model.The results show that the proposed VMD-Informer model has smaller predicted error,larger decision coefficient and more accurate prediction,which can serve as an effective method for manufacturing enterprises to achieve quality prediction and provide ideas as well for enterprises in terms of adjusting their production plans in time.
作者
郑华丽
李志敏
王明君
闫文凯
叶春明
ZHENG Hua-i;LI Zhi-min;WANG Ming-jun;YAN Wen-kai;YE Chun-ming(China Tobacco Zhejiang Industrial Co.,Ltd.,Hangzhou 310008;School of Business,University of Shanghai for Science and Technology,Shanghai 200093)
出处
《制造业自动化》
2025年第5期54-61,共8页
Manufacturing Automation
基金
上海市哲学社会科学一般项目(2022BGL010)。
作者简介
郑华丽(1987-),女,浙江杭州人,工程师,硕士,研究方向为经济管理,质量管理;通讯作者:叶春明(1964-),男,安徽宣城人,教授,博士,研究方向为工业工程、企业战略、企业生产计划与控制、生产调度、企业资源计划(ERP)、供应链管理以及企业信息化。
