Please wait a minute...
工程设计学报  2024, Vol. 31 Issue (1): 1-9    DOI: 10.3785/j.issn.1006-754X.2023.03.207
产品创新设计     
融合多主体需求频率特征的复杂产品全生命周期价值链协同设计
何州1,2(),王阳3,蒋翔宇4,洪兆溪4,5(),何利力3,冯毅雄4,6
1.浙江工商大学 萨塞克斯人工智能学院,浙江 杭州 310012
2.杭州州力数据科技有限公司,浙江 杭州 310019
3.浙江理工大学 计算机科学与技术学院,浙江 杭州 310018
4.浙江大学 流体动力基础件与机电系统全国重点实验室,浙江 杭州 310058
5.浙江大学 宁波科创中心,浙江 宁波 315100
6.贵州大学 省部共建公共大数据国家重点实验室,贵州 贵阳 550025
Collaborative design of complex product lifecycle value chain by fusing multi-agent demand frequency characteristics
Zhou HE1,2(),Yang WANG3,Xiangyu JIANG4,Zhaoxi HONG4,5(),Lili HE3,Yixiong FENG4,6
1.Sussex Artificial Intelligence Institute, Zhejiang Gongshang University, Hangzhou 310012, China
2.Hangzhou Zhouli Data Technology Co. , Ltd. , Hangzhou 310019, China
3.School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China
4.State Key Laboratory of Fundamental Components of Fluid Power and Mechatronic systems, Zhejiang University, Hangzhou 310058, China
5.Ningbo Innovation Center, Zhejiang University, Ningbo 315100, China
6.State Key Laboratory of Public Big Data, Guizhou University, Guiyang 550025, China
 全文: PDF(3642 KB)   HTML
摘要:

高周转率和短保质期的复杂产品具有较高订单频率的特征。传统的价值链设计大多基于复杂产品订单数据的时序和销量对销售周期的影响,忽略了其订单频率中蕴含的细节信息,因而难以准确捕捉多主体间快速变化的供需关系。为了解决这一问题,提出了一种融合多主体需求频率特征的复杂产品全生命周期价值链协同设计方法。首先,采用门控卷积的频率序列提取方法识别多主体需求;其次,将基于频率分段的Transformer时序预测模型融合于订单频率信息,根据改进的时序-频率多头自注意力(seq-fre multi-head attention)结构建立全生命周期价值链,不同分段的时序和频率特征对应不同的注意力头,以实现多段时序和频率特征的融合;最后,将新型价值链协同设计方法应用于某复杂产品多主体需求预测问题,进行实验验证。研究表明,所提出的融合需求频率特征的价值链协同设计方法预测准确度较高,具有很好的应用前景。

关键词: 价值链协同设计时序预测Transformer频率特征    
Abstract:

Complex products with high turnover and short shelf life are characterized by a higher order frequency. Traditional value chain design is mostly based on the time series of order data of complex products and the impact of sales volume on sales cycle, while ignoring the detail information contained in the order frequency, which is difficult to accurately capture the rapidly changing supply and demand relationship between multi-agents. In order to solve this problem, a collaborative design method of complex product lifecycle value chain fusing multi-agent demand frequency characteristics was proposed. Firstly, the frequency sequence extraction method of gated convolution was used to identify the multi-agent requirement; Secondly, the Transformer time series prediction model based on frequency segmentation was integrated into the order frequency information, and the lifecycle value chain was built according to the improved time series frequency multi-head self-attention (seq-fre multi-head attention) structure. The time series and frequency characteristics of different segments corresponded to different attention heads to realize the fusion of multi-stage time series and frequency features; Finally, the new value chain collaborative design method was applied to the multi-agent demand prediction problem of a complex product, and experimental verification was carried out. The result showed that the proposed value chain collaborative design method fusing demand frequency features has high prediction accuracy and good application prospects.

Key words: value chain    collaborative design    time series prediction    Transformer    frequency feature
收稿日期: 2023-10-13 出版日期: 2024-03-04
CLC:  TP 24  
基金资助: 浙江省重点研发计划项目(2022C01238)
通讯作者: 洪兆溪     E-mail: zh277@sussex.ac.uk;hzhx@zju.edu.cn
作者简介: 何 州(1995—),男,浙江杭州人,工程师,硕士,从事价值链及人工智能技术研究,E-mail: zh277@sussex.ac.uk
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  
何州
王阳
蒋翔宇
洪兆溪
何利力
冯毅雄

引用本文:

何州,王阳,蒋翔宇,洪兆溪,何利力,冯毅雄. 融合多主体需求频率特征的复杂产品全生命周期价值链协同设计[J]. 工程设计学报, 2024, 31(1): 1-9.

Zhou HE,Yang WANG,Xiangyu JIANG,Zhaoxi HONG,Lili HE,Yixiong FENG. Collaborative design of complex product lifecycle value chain by fusing multi-agent demand frequency characteristics[J]. Chinese Journal of Engineering Design, 2024, 31(1): 1-9.

链接本文:

https://www.zjujournals.com/gcsjxb/CN/10.3785/j.issn.1006-754X.2023.03.207        https://www.zjujournals.com/gcsjxb/CN/Y2024/V31/I1/1

图1  复杂产品制造企业价值链演进升级的基本路径
图2  SFTransformer模型结构
图3  一维门控卷积结构
图4  特征数据序列输入网络
图5  时序-频率多头自注意力模块结构
图6  某复杂产品制造企业销售数据集
参数量值
batchSize32
学习率0.000 5
训练数200轮
多头注意力6个
隐藏数16层
Dropout0.05
优化器Adam
表1  SFTransformer模型参数设置
预测长度评价指标VAR模型LSTM模型Transformer模型SFTransformer模型
4ESM0.1910.1750.0980.084
EMA0.2010.1970.1150.102
ERSM0.4370.4180.3130.290
12ESM0.3970.3560.1980.165
EMA0.4090.3880.2250.201
ERSM0.7300.5960.4450.406
表2  不同模型预测性能的对比
图7  各模型对产品销量的预测结果
1 杨海民,潘志松,白玮.时间序列预测方法综述[J].计算机科学,2019,46(1):21-28. doi:10.11896/j.issn.1002-137X.2019.01.004
YANG H M, PAN Z S, BAI W. Review of time series prediction methods[J]. Computer Science, 2019, 46(1): 21-28.
doi: 10.11896/j.issn.1002-137X.2019.01.004
2 王静,曹春正.基于贝叶斯分层自回归时空模型的北京PM2.5预测[J].南京信息工程大学学报(自然科学版),2023,15(1):34-41.
WANG J, CAO C Z. Prediction of PM2.5 concentration in Beijing based on Bayesian hierarchical autoregressive spatio-temporal model[J]. Journal of Nanjing University of Information Science and Technology (Natural Science Edition), 2023, 15(1): 34-41.
3 陈然.指数平滑与自回归融合预测模型及应用研究[D]. 阜新:辽宁工程技术大学,2021.
CHEN R. Index smoothing and autoregressive fusion forecasting model and its empirical study[D]. Fuxin: Liaoning Technical University, 2021.
4 余廷芳,张浩杰.基于SVM和RBF神经网络的CFB NOx生成预测模型[J].计算机仿真,2020,37(9):209-213. doi:10.3969/j.issn.1006-9348.2020.09.043
YU T F, ZHANG H J. NOx generation prediction model of coal-fired boiler based on SVM and RBF neural network[J]. Computer Simulation, 2020, 37(9): 209-213.
doi: 10.3969/j.issn.1006-9348.2020.09.043
5 高塔.基于改进贝叶斯网络的高维数据本地差分隐私方法的研究[D].北京:北京交通大学,2021.
GAO T. Research on local differential privacy method for high-dimensional data based on improved Bayesian network[D]. Beijing: Beijing Jiaotong University, 2021.
6 MNIH V, HEESS N, GRAVES A, et al. Recurrent models of visual attention[C]// Proceedings of the 27th International Conference on Neural Information Processing Systems. Cambridge. Cambridge, Massachusetts, USA: MIT Press, 2014: 2204-2212.
7 武晋鹏.基于样本精选与BERT模型引导的图像情感分析研究[D].南昌:华东交通大学,2021.
WU J P. Image sentiment analysis based on sample refinement and BERT guided[D]. Nanchang: East China Jiaotong University, 2021.
8 XIANG Y, FOX D. DA-RNN: semantic mapping with data associated recurrent neural networks[C]//13th Conference on Robotics-Science and Systems, Cambridge, Massachusetts, Jul. 12-16, 2017.
9 QIN Y, SONG D, CHEN H, et al. A dual-stage attention-based recurrent neural network for time series prediction[EB/OL]. [2023-10-06]..
10 HOCHREITER S. Long short-term memory[J]. Neural Computation, 1997(9): 1735-1780.
11 SESTI N, GARAU-LUIS J J, CRAWLEY E, et al. Integrating LSTMs and GNNs for COVID-19 forecasting[J]. arXiv preprint arXiv:, 2021.
12 王素.基于深度学习的时间序列预测算法研究与应用[D].成都:电子科技大学,2022.
WANG S. Research and application of time series prediction algorithm based on deep learning[D]. Chengdu: University of Electronic Science and Technology of China, 2022.
13 CHO K, VAN MERRIENBOER B, BAHDANAU D, et al. On the properties of neural machine translation: Encoder-Decoder approaches[EB/OL]. [2023-10-06]. .
14 VASWANI A, SHAZEER N, PARMAR N. Attention is all you need[EB/OL]. [2023-10-06]..
15 LIN L, LUO H, HUANG R, et al. Recurrent models of visual co-attention for person re-identification[J]. IEEE Access, 2019, 7: 8865-8875.
16 陈超.基于神经网络的多变量时间序列预测[D].成都:电子科技大学,2022.
CHEN C. Multivariate time prediction based on neural network[D]. Chengdu: University of Electronic Science and Technology of China, 2022.
17 MIKOLOV T, SUTSKEVER I, KAI C, et al. Distributed representations of words and phrases and their compositionality[C]//NIPS'13: Proceedings of the 26th International Conference on Neural Information Processing Systems: Volume 2. New York: ACM, 2013: 3111-3119.
18 WANGA R, LOUA J, JIANGA Y. Session-based recommendation with time-aware neural attention network[J]. Expert Systems with Applications, 2022, 210: 118395.
19 陈士超,郁滨.面向术语抽取的双阈值互信息过滤方法[J].计算机应用,2011,31(4):1070-1073. doi:10.3724/sp.j.1087.2011.01070
CHEN S C, YU B. Method of mutual information filtration with dual-threshold for term extraction[J]. Journal of Computer Applications, 2011, 31(4): 1070-1073.
doi: 10.3724/sp.j.1087.2011.01070
20 曹昀炀.基于Transformer的选股因子挖掘[D].上海:华东师范大学,2022.
CAO Y Y. Based on Transformer stock factor construction[D]. Shanghai: East China Normal University, 2022.
[1] 罗亚波, 凌鹤. 面向同步协同的协同环境共享策略研究[J]. 工程设计学报, 2006, 13(1): 17-20.
[2] 倪宁, 卢刚, 卜佳俊. 基于Multi-Agent的协同设计中虚拟环境感知研究[J]. 工程设计学报, 2006, 13(1): 21-25.
[3] 陈树勋, 陈伟军. 在互联网上进行结构系统协同优化的信息交流与进程调控[J]. 工程设计学报, 2002, 9(5): 236-240.
[4] 张友良, 汪惠芬. 异地协同设计制造关键技术及系统实现[J]. 工程设计学报, 2002, 9(2): 53-59.
[5] 哈进兵, 张友良. 一种异地协同设计中的安全策略[J]. 工程设计学报, 2001, 8(1): 17-21.
[6] 洪宁, 许端清. 图案协同设计中的数据处理技术[J]. 工程设计学报, 2000, 7(3): 13-17.