助训练策略下的多模型软测量建模

doi:10.16182/j.issn1004731x.joss.22-0976

系统仿真学报 ›› 2024, Vol. 36 ›› Issue (1): 249-259.doi: 10.16182/j.issn1004731x.joss.22-0976

• 论文 • 上一篇

助训练策略下的多模型软测量建模

何罗苏阳¹(), 熊伟丽¹^,²

^1.江南大学物联网工程学院，江苏无锡 214122
^2.江南大学轻工过程先进控制教育部重点实验室，江苏无锡 214122

收稿日期:2022-08-18 修回日期:2022-09-18 出版日期:2024-01-20 发布日期:2024-01-19
第一作者简介:何罗苏阳(1997-)，男，硕士生，研究方向为复杂工业过程建模。E-mail：hlsy1224@foxmail.com
基金资助:
国家自然科学基金(61773182);国家重点研发计划(2018YFC1603705-03)

Multi-model Soft Sensor Modeling under Help-training Strategy

He Luosuyang¹(), Xiong Weili¹^,²

^1.School of Internet of Things Engineering, Jiangnan University, Wuxi 214122, China
^2.Key Laboratory of Advanced Process Control for Light Industry, Ministry of Education, Jiangnan University, Wuxi 214122, China

Received:2022-08-18 Revised:2022-09-18 Online:2024-01-20 Published:2024-01-19

摘要/Abstract

摘要：

由于复杂工业过程中存在强非线性、多阶段耦合以及有标签样本数量偏少的情况，传统的全局软测量模型难以精确描述整个过程。为此，提出一种助训练策略下的多模型软测量建模方法。该方法采用模糊C均值聚类算法挖掘样本集中的相似性样本并建立若干子模型；通过引入助训练策略，形成基于主、辅学习器的协同训练框架，并设计置信度评估机制淘汰误差样本的同时扩充子模型的建模空间；进而将模糊隶属度作为D-S证据理论的概率分配函数计算出子模型权重，对子模型的输出进行融合以得到最终的模型预测结果。通过对脱丁烷塔工业过程的实际数据进行建模仿真，结果表明此模型具有良好的预测性能。

关键词: 软测量建模, 多模型, 助训练, 学习器, 脱丁烷塔

Abstract:

Due to the strong nonlinearity, multi-stage coupling, and the small number of labeled samples in complex industrial processes, it is difficult for traditional global soft sensor models to accurately describe the whole process. Therefore, a multi-model soft sensor modeling method under the help-training strategy is proposed. This method uses a fuzzy C-means (FMC) clustering algorithm to mine similar samples in the sample set and build several sub-models. By introducing the help-training strategy, a collaborative training framework based on main and auxiliary learners is formed, and a confidence evaluation mechanism is designed to eliminate error samples and expand the modeling space of sub-models. Then the fuzzy membership degree is used as the probability distribution function of D-S evidence theory to calculate the weight of the sub-model, and the output of the sub-model is fused to obtain the final model prediction result. Through the modeling and simulation of the actual data of the debutanizer industrial process, the results show that this model has good prediction performance.

Key words: soft sensor modeling, multi-model, help-training, learner, debutanizer

中图分类号:

TP274

何罗苏阳,熊伟丽 . 助训练策略下的多模型软测量建模[J]. 系统仿真学报, 2024, 36(1): 249-259.

He Luosuyang,Xiong Weili . Multi-model Soft Sensor Modeling under Help-training Strategy[J]. Journal of System Simulation, 2024, 36(1): 249-259.

图/表 12

图1

图2

图3

图4

表1

图5

图6

表2

表3

图7

图8

图9

参考文献 20

1	Yuan Xiaofeng, Gu Yongjie, Wang Yalin, et al. A Deep Supervised Learning Framework for Data-driven Soft Sensor Modeling of Industrial Processes[J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 31(11): 4737-4746.
2	Wang Bo, Wang Xingyu, He Mengyi, et al. Study on Multi-model Soft Sensor Modeling Method and Its Model Optimization for the Fermentation Process of Pichia pastoris[J]. Sensors, 2021, 21(22): 7635.
3	Shao Weiming, Ge Zhiqiang, Song Zhihuan, et al. Nonlinear Industrial soft Sensor Development Based on Semi-supervised Probabilistic Mixture of Extreme Learning Machines[J]. Control Engineering Practice, 2019, 91: 104098.
4	Adankon Mathias M, Cheriet Mohamed. Help-training for Semi-supervised Support Vector Machines[J]. Pattern Recognition, 2011, 44(9): 2220-2230.
5	Zhang Lei, Yang Lin, Ma Tianwu, et al. A Self-training Semi-supervised Machine Learning Method for Predictive Mapping of Soil Classes with Limited Sample Data[J]. Geoderma, 2021, 384: 114809.
6	仝小敏, 吉祥. 基于自训练的回归算法[J]. 中国电子科学研究院学报, 2017, 12(5): 498-502.
	Tong Xiaomin, Ji Xiang. Regression Algorithm Based on Self Training[J]. Journal of China Academy of Electronics and Information Technology, 2017, 12(5): 498-502.
7	Adankon Mathias M, Cheriet Mohamed. Help-training for Semi-supervised Discriminative Classifiers. Application to SVM[C]//2008 19th International Conference on Pattern Recognition. Piscataway, NJ, USA: IEEE, 2008: 1-4.
8	Adankon Mathias M, Cheriet Mohamed. Help-training Semi-supervised LS-SVM[C]//2009 International Joint Conference on Neural Networks. Piscataway, NJ, USA: IEEE, 2009: 49-56.
9	Gan Haitao, Sang Nong, Huang Rui, et al. Using Clustering Analysis to Improve Semi-supervised Classification[J]. Neurocomputing, 2013, 101: 290-298.
10	钟伟民, 李杰, 程辉, 等. 基于FCM聚类的气化炉温度多模型软测量建模[J]. 化工学报, 2012, 63(12): 3951-3955.
	Zhong Weimin, Li Jie, Cheng Hui, et al. A Soft Sensor Multi-modeling for Furnace Temperature of Gasifier Based FCM Clustering[J]. CIESC Journal, 2012, 63(12): 3951-3955.
11	嵇小辅, 张翔. 基于FCM与集成高斯过程回归的赖氨酸发酵软测量[J]. 智能系统学报, 2015, 10(1): 156-162.
	Ji Xiaofu, Zhang Xiang. Soft Measurement of Lysine Fermentation Based on FCM and Integrated Gaussian Process Regression[J]. CAAI Transactions on Intelligent Systems, 2015, 10(1): 156-162.
12	李卫, 杨煜普, 王娜. 基于核模糊聚类的多模型LSSVM回归建模[J]. 控制与决策, 2008, 23(5): 560-562, 566.
	Li Wei, Yang Yupu, Wang Na. Multi-model LSSVM Regression Modeling Based on Kernel Fuzzy Clustering[J]. Control and Decision, 2008, 23(5): 560-562, 566.
13	Yuan Xiaofeng, Qi Shuaibin, Wang Yalin, et al. A Dynamic CNN for Nonlinear Dynamic Feature Learning in Soft Sensor Modeling of Industrial Process Data[J]. Control Engineering Practice, 2020, 104: 104614.
14	Wang Hui, Guo Lili, Dou Zheng, et al. A New Method of Cognitive Signal Recognition Based on Hybrid Information Entropy and D-S Evidence Theory[J]. Mobile Networks and Applications, 2018, 23(4): 677-685.
15	刘准钆, 程咏梅, 潘泉, 等. 基于证据距离和矛盾因子的加权证据合成法[J]. 控制理论与应用, 2009, 26(12): 1439-1442.
	Liu Zhunga, Cheng Yongmei, Pan Quan, et al. Combination of Weighted Belief Functions Based on Evidence Distance and Conflicting Belief[J]. Control Theory & Applications, 2009, 26(12): 1439-1442.
16	Xiong Weili, Zhang Wei, Xu Baoguo, et al. JITL Based MWGPR Soft Sensor for Multi-mode Process with Dual-updating Strategy[J]. Computers & Chemical Engineering, 2016, 90: 260-267.
17	Wang Liangjie, Sawada Kazuhide, Moriguchi Shuji. Landslide Susceptibility Analysis with Logistic Regression Model Based on FCM Sampling Strategy[J]. Computers & Geosciences, 2013, 57: 81-92.
18	Meng Yanmei, Lan Qiliang, Qin Johnny, et al. Data-driven Soft Sensor Modeling Based on Twin Support Vector Regression for Cane Sugar Crystallization[J]. Journal of Food Engineering, 2019, 241: 159-165.
19	Niño-Adan Iratxe, Landa-Torres Itziar, Manjarres Diana, et al. Soft-sensor for Class Prediction of the Percentage of Pentanes in Butane at a Debutanizer Column[J]. Sensors, 2021, 21(12): 3991.
20	Kaneko Hiromasa, Funatsu Kimito. Adaptive Soft Sensor Based on Online Support Vector Regression and Bayesian Ensemble Learning for Various States in Chemical Plants[J]. Chemometrics and Intelligent Laboratory Systems, 2014, 137: 57-66

模型	RMSE	COR
TSVR	0.467 1	0.797 4
H-TSVR	0.233 5	0.903 1
HFD-TSVR	0.217 8	0.915 6

主要变量	变量名称
x₁	顶层温度
x₂	顶层压力
x₃	回流流量
x₄	流向下个过程的流量
x₅	第六塔板温度
x₆	塔底温度1
x₇	塔底温度2

模型	RMSE	COR
TSVR	0.141 2	0.510 4
S-TSVR	0.137 4	0.589 7
H-TSVR	0.137 1	0.594 4
HF-TSVR	0.128 9	0.701 2
HFD-TSVR	0.112 8	0.720 5

[1]	代学志, 熊伟丽. 基于概率选择的主动学习智能软测量建模[J]. 系统仿真学报, 2021, 33(6): 1350-1357.
[2]	韩璞, 张婷. 有机朗肯循环余热利用系统的多模型预测控制[J]. 系统仿真学报, 2017, 29(11): 2811-2819.
[3]	周彦, 胡岚, 王冬丽. 基于IMM-UKF-CS的混合环境目标跟踪[J]. 系统仿真学报, 2017, 29(1): 43-48.
[4]	赵小鹏, 章永春. 基于最小熵聚类的多模型在线辨识及仿真[J]. 系统仿真学报, 2016, 28(6): 1306-1312.
[5]	时振伟, 纪志成. 基于多元线性回归模型的永磁同步电机参数辨识的新方法[J]. 系统仿真学报, 2015, 27(8): 1687-1696.
[6]	牛玉广, 李晓明, 王世林, 林忠伟. 分散协调控制在风火混合电力系统中仿真研究[J]. 系统仿真学报, 2015, 27(3): 609-619.

助训练策略下的多模型软测量建模

Multi-model Soft Sensor Modeling under Help-training Strategy

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 20

相关文章 6

编辑推荐

Metrics

本文评价