不完备数据下的聚酯熔体特性黏度预测方法

doi:10.16183/j.cnki.jsjtu.2023.011

摘要/Abstract

摘要：

特性黏度是衡量聚酯熔体质量的关键指标,对其进行精准预测有利于提前发现聚酯熔体潜在的质量问题,及时调整工艺参数,减少企业损失.考虑到聚酯熔体生产过程的数据不完备性、数据时序性以及高维冗余性,提出了不完备数据下聚酯熔体的特性黏度预测方法.针对聚酯熔体极端生产环境造成的数据不完备问题,设计了以卷积神经网络判别器和注意力长短期记忆神经网络生成器为架构的缺失数据生成对抗网络(MDGAN),通过对抗生成机制实现了缺失数据的填充.针对聚酯熔体生产过程中高维冗余和时序双向因果特性,设计了基于极端梯度提升双向门控循环单元(XGBoost-BiGRU)的特性黏度预测模型,通过极端梯度提升算法进行特征筛选,获取预测模型输入变量,再利用双向门控循环单元捕捉数据的时序双向因果关系,实现特性黏度的精准预测.浙江某聚酯纤维生产企业的实际数据测试结果表明,MDGAN算法在不同缺失率数据集下的填充精度均优于KNN、RF、MICE、GAIN数据填充算法,XGBoost-BiGRU特性黏度预测方法较STL-GPR、CAGRU、BiGRU算法优势显著,结合MDGAN的特性黏度预测方法能有效解决数据不完备下的聚酯熔体特性黏度预测问题.

关键词: 特性黏度预测, 不完备数据, 生成对抗网络, 循环神经网络

Abstract:

Characteristic viscosity is a key indicator of the quality of polyester melts, whose accurate prediction can help to identify potential quality problems of polyester melts in advance, adjust the process parameters in time and reduce enterprise losses. Considering the data incompleteness, data time series and high dimensional redundancy of the polyester melt production process, a method is proposed to predict the characteristic viscosity of polyester melt under incomplete data. A missing data generative adversarial nets (MDGAN) with a convolutional neural network discriminator and an attention long short-term memory neural network generator is designed to address the data incompleteness problem caused by the extreme production environment of polyester melts, and the missing data is filled by the adversarial generation mechanism. The extreme gradient boosting-bidirectional gated recurrent unit (XGBoost-BiGRU) is designed to predict the viscosity of polyester melts based on high dimensional redundancy and temporal characteristics prediction. The actual data test results of a polyester fiber manufacturer in Zhejiang show that the filling accuracy of the MDGAN algorithm at different missing rate data sets is better than that of data filling algorithms such as KNN,RF,MICE,and GAIN. The XGBoost-BiGRU characteristic viscosity prediction method has significant advantages over STL-GPR, CAGRU, BiGRU. In combination of MDGAN characteristic viscosity prediction, the method proposed can effectively solve the problem of predicting the characteristic viscosity of polyester melts under incomplete data.

Key words: intrinsic viscosity prediction, incomplete data, generative adversarial networks (GAN), recurrent neural networks (RNN)

中图分类号:

TP183

毕金茂, 张朋, 张洁, 赵春财, 崔利. 不完备数据下的聚酯熔体特性黏度预测方法[J]. 上海交通大学学报, 2024, 58(4): 534-544.

BI Jinmao, ZHANG Peng, ZHANG Jie, ZHAO Chuncai, CUI Li. Polyester Melt Characteristic Viscosity Prediction Method Under Incomplete Data[J]. Journal of Shanghai Jiao Tong University, 2024, 58(4): 534-544.

图/表 14

图1

图2

图3

图4

图5

图6

表1

图7

表2

图8

图9

图10

图11

图12

参考文献 26

[1]	WU C, REN L, HAO K. Modeling of aggregation process based on feature selection extreme learning machine of atomic Search algorithm[C]// Data Driven Control and Learning Systems Conference. Suzhou, China: IEEE, 2021: 1453-1458.
[2]	GAO J, CHEN L, HAO K, et al. Copula-based JITL for polyester fiber polymerization process modeling[C]// Chinese Automation Congress. Shanghai, China: IEEE, 2020: 2628-2633.
[3]	PENG H, CHEN L, HAO K. Deep transfer model with source domain segmentation for polyester esterification processes[C]// Chinese Control and Decision Conference. Hefei, China: IEEE, 2022: 293-298.
[4]	RAVINDRANATH K, MASHELKAR R A. Modeling of poly (ethylene terephthalate) reactors: 4. A continuous esterification process[J]. Polymer Engineering and Science, 1982, 22(10):628-636. doi: 10.1002/pen.v22:10 URL
[5]	ZHU X, DAMARLA S K, HAO K, et al. Parallel interaction spatiotemporal constrained variational autoencoder for soft sensor modeling[J]. IEEE Transactions on Industrial Informatics, 2021, 18(8): 5190-5198. doi: 10.1109/TII.2021.3110197 URL
[6]	毕佳俊. 数据驱动的聚酯纤维聚合过程特性黏度预测方法[D]. 上海: 东华大学, 2022.
	BI Jiajun. Data-driven intrinsic viscosity prediction method for polyester fiber polymerization process[D]. Shanghai: Donghua University, 2022.
[7]	GENG J, CHEN L, HAO K, et al. Fractal-based combined kernel function model for the polyester polymerization process[C]// Chinese Control and Decision Conference. Kunming, China: IEEE, 2021: 656-661.
[8]	XIE R, HAO K, HUANG B, et al. Data-driven modeling based on two-stream λ gated recurrent unit network with soft sensor application[J]. IEEE Transactions on Industrial Electronics, 2019, 67(8): 7034-7043. doi: 10.1109/TIE.41 URL
[9]	PAN J, LI C B, TANG Y, et al. Energy consumption prediction of a CNC machining process with incomplete data[J]. IEEE/ CAA Journal of Automatica Sinica, 2021, 8(5): 987-1000.
[10]	LIU W, REN C, XU Y. PV generation forecasting with missing input data: A super-resolution perception approach[J]. IEEE Transactions on Sustainable Energy, 2021, 12(2): 1493-1496. doi: 10.1109/TSTE.2020.3029731 URL
[11]	JIANG C, ZHONG W, LI Z, et al. Real-time semi-supervised predictive modeling strategy for industrial continuous catalytic reforming process with incomplete data using slow feature analysis[J]. Industrial & Engineering Chemistry Research, 2019, 58(37): 17406-17423. doi: 10.1021/acs.iecr.9b03119 URL
[12]	ZHU G, ZHOU K, LU L, et al. Partial discharge data augmentation based on improved Wasserstein generative adversarial network with gradient penalty[J]. IEEE Transactions on Industrial Informatics, 2022, 19(5): 6565-6575. doi: 10.1109/TII.2022.3197839 URL
[13]	肖竹, 钱鑫, 蒋洪波, 等. 基于双向RNN的私家车轨迹重构算法[J]. 通信学报, 2020, 41(12): 171-181. doi: 10.11959/j.issn.1000-436X.2020227
	XIAO Zhu, QIAN Xin, JIANG Hongbo, et al. Bidirectional RNN-based private car trajectory reconstruction algorithm[J]. Journal on Communications. 2020, 41(12): 171-181. doi: 10.11959/j.issn.1000-436X.2020227
[14]	WANG M, ZHOU T, WANG H, et al. Chinese power dispatching text entity recognition based on a double-layer BiLSTM and multi-feature fusion[J]. Energy Reports, 2022, 8: 980-987.
[15]	朱秀丽. 聚酯纤维聚合过程的智能建模与多目标优化[D]. 上海: 东华大学, 2022.
	ZHU Xiuli. Intelligent modeling and multi-objective optimization of polyester fiber polymerization process[D]. Shanghai: Donghua University, 2022.
[16]	ZHOU F, ZHAO L, ZHU J, et al. Research on short-term wind power forecasting method based on incomplete data[J]. Journal of Renewable and Sustainable Energy, 2022, 14(3): 036103.
[17]	COLLOBERT R, WESTON J, BOTTOU L, et al. Natural language processing (almost) from scratch[J]. Journal of Machine Learning Research, 2011, 12: 2493-2537.
[18]	倪扬帆, 杨媛媛, 谢哲, 等. 基于LSTM与注意力结构的肺结节多特征抽取方法[J]. 上海交通大学学报, 2022, 56(8): 1078-1088. doi: 10.16183/j.cnki.jsjtu.2021.113
	NI Yangfan, YANG Yuanyuan, XIE Zhe, et al. Multi feature extraction method for pulmonary nodules based on LSTM and attention structure[J]. Journal of Shanghai Jiao Tong University, 2022, 56(8): 1078-1088.
[19]	CHEN Y. Convolutional neural network for sentence classification[D]. Waterloo, Canada: University of Waterloo, 2015.
[20]	KARAN B. Speech-based Parkinson’s disease prediction using XGBoost-based features selection and the stacked ensemble of classifiers[J]. Journal of The Institution of Engineers (India): Series B, 2023, 104(2): 475-483. doi: 10.1007/s40031-022-00851-2
[21]	康守强, 周月, 王玉静, 等. 基于改进SAE和双向LSTM的滚动轴承RUL预测方法[J]. 自动化学报, 2022, 48(9): 2327-2336.
	KANG Shouqiang, ZHOU Yue, WANG Yujing, et al. RUL prediction method of a rolling bearing based on improved SAE and Bi-LSTM[J]. Acta Automatica Sinica, 2022, 48(9): 2327-2336.
[22]	杨海柱, 江昭阳, 李梦龙, 等. 基于CS-GRU模型的短期负荷预测方法研究[J]. 传感器与微系统, 2022, 41(9): 54-57.
	YANG Haizhu, JIANG Zhaoyang, LI Menglong, et al. Study on short-term load forecasting method based on CS-GRU model[J]. Transducer and Microsystem Technologies. 2022, 41(9): 54-57.
[23]	BATISTA G E, MONARD M C. An analysis of four missing data treatment methods for supervised learning[J]. Applied Artificial Intelligence, 2003, 17(5/6): 519-533. doi: 10.1080/713827181 URL
[24]	TANG F, ISHWARAN H. Random forest missing data algorithms[J]. Statistical Analysis and Data Mining: The ASA Data Science Journal, 2017, 10(6): 363-377. doi: 10.1002/sam.2017.10.issue-6 URL
[25]	RATOLOJANAHARY R, NGOUNA R H, MEDJAHER K, et al. Model selection to improve multiple imputation for handling high rate missingness in a water quality dataset[J]. Expert Systems with Applications, 2019, 131: 299-307 doi: 10.1016/j.eswa.2019.04.049 URL
[26]	YOON J, JORDON J, SCHAAT M. Gain: Missing data imputation using generative adversarial nets[C]// International Conference on Machine Learning. Stockholm, Sweden: IEEE, 2018: 5689-5698.

缺失率/%	KNN	RF	MICE	GAIN	MDGAN
5	0.179	0.182	0.175	0.136	0.082
10	0.236	0.221	0.223	0.169	0.113
15	0.315	0.307	0.279	0.204	0.168
20	0.390	0.368	0.328	0.263	0.201
25	0.488	0.447	0.364	0.313	0.257
30	0.561	0.522	0.425	0.380	0.316

编号	误差
编号	η/ (dL·g^-1)	STL-GPR	BiGRU	CAGRU	XGBoost- BiGRU
1	50.44	0.108	0.077	0.087	0.047
2	50.40	0.088	0.087	0.078	0.048
3	50.31	0.096	0.083	0.081	0.042
4	50.21	0.097	0.078	0.061	0.055
5	50.10	0.121	0.088	0.075	0.036
6	50.13	0.110	0.076	0.083	0.017
7	50.16	0.103	0.071	0.070	0.040
8	50.34	0.139	0.077	0.079	0.024
9	49.98	0.139	0.085	0.086	0.011
︙	︙	︙	︙	︙	︙
91	49.74	0.149	0.099	0.072	0.014
92	49.70	0.102	0.128	0.071	0.038
93	50.00	0.140	0.106	0.080	0.025
94	50.13	0.146	0.123	0.076	0.039
95	49.74	0.091	0.119	0.084	0.057
96	49.77	0.124	0.092	0.050	0.047
97	49.99	0.116	0.121	0.056	0.042
98	49.97	0.101	0.104	0.057	0.051
99	50.00	0.127	0.125	0.067	0.057
100	49.83	0.103	0.102	0.042	0.055