基于迁移学习的微反应器大气液比Aspen模型构建

doi:10.16085/j.issn.1000-6613.2024-1212

摘要/Abstract

摘要：

微反应器作为过程强化的关键设备，已在多个领域应用。然而，多数流程模拟软件缺乏微反应器模块，在一定程度上阻碍了其在工业中的应用。本文提出了一种基于迁移学习构建的适用于大气液比的Aspen微反应器的产物收率预测模型，以微反应器中气液磺化合成十二烷基苯磺酸过程为例，验证了该模型的准确性。首先，通过大气液比［（2000∶1）~（3000∶1）］磺化实验，在T形微反应器中收集38组十二烷基苯磺酸的收率数据，作为迁移学习的目标域。基于微通道环状流特征采用Aspen Plus的平推流反应器（PFR）模块初步建立该过程的产物收率预测模型，并生成29700组源域数据。考虑流体动力学和微通道结构等特征，本文采用迁移学习中的条件对抗域适应网络（CADAN）并对其进行调整，包括采用深度ReLU网络架构及优化对抗损失函数。随后利用模拟数据训练特征提取器，并利用实验数据进行条件对抗域适应训练。最终建立的模型拟合系数（R²）可达0.9346，相较人工神经网络提升了14.6%，较PFR模型提升了98.18%。同时，该模型在20%的噪声水平下仍保持R²大于0.78，均方根误差低至5.07，优于同等条件下的人工神经网络，显示出较高的预测精度和强鲁棒性。

关键词: 微反应器, 过程建模, Aspen Plus, 迁移学习, 算法

Abstract:

Microreactors, as key devices for process intensification, have been applied in several fields. However, the lack of a microreactor module in most process simulation software has hindered its application in industry to some extent. In this paper, based on transfer learning, a product yield prediction model for Aspen microreactors applicable to large gas-liquid ratio is proposed, and the accuracy of the model is verified by taking the process of synthesizing dodecylbenzene sulfonic acid by large gas-liquid sulfonation in a microreactor as an example. First, 38 sets of yield data of dodecylbenzene sulfonic acid are collected in a T-type microreactor through sulfonation experiments with gas-liquid ratios ［(2000∶1)—(3000∶1)］, which serve as the target domain for transfer learning. A preliminary product yield prediction model for the process is developed based on the microchannel annular flow characteristics using the flat push-flow reactor (PFR) module of Aspen Plus, and 29700 sets of source domain data are generated. Considering the features such as fluid dynamics and microchannel structure, this study adopts and adapts the conditional adversarial domain adaptation network (CADAN) in transfer learning, including the adoption of a deep ReLU network architecture and optimization of the adversarial loss function. Subsequently, the feature extractor is trained using simulated data and the conditional adversarial domain adaptation is trained using experimental data. The final model fit coefficient (R²) can reach 0.9346, which is improved by 14.6% compared to the artificial neural network and 98.18% compared to the PFR model. Meanwhile, the model maintains an R² greater than 0.78 at 20% noise level, and the root mean square error is as low as 5.07, which is better than that of the artificial neural network under the same conditions, showing high prediction accuracy and strong robustness.

Key words: microreactor, process modeling, Aspen Plus, transfer learning, algorithm

中图分类号:

TP183

秦睦轩, 张玮, 王盈锦, 李梓良. 基于迁移学习的微反应器大气液比Aspen模型构建[J]. 化工进展, 2025, 44(9): 4908-4916.

QIN Muxuan, ZHANG Wei, WANG Yingjin, LI Ziliang. Construction of Aspen model for large gas-liquid ratio in microreactors based on transfer learning[J]. Chemical Industry and Engineering Progress, 2025, 44(9): 4908-4916.

图/表 8

参考文献 29

[1]	PENG Zhengbiao, WANG Guichao, MOGHTADERI Behdad, et al. A review of microreactors based on slurry Taylor (segmented) flow[J]. Chemical Engineering Science, 2022, 247: 117040.
[2]	OUYANG Yi, HEYNDERICKX Geraldine J, VAN GEEM Kevin M. Development of intensified reactors: A process intensification methodology perspective[J]. Chemical Engineering and Processing: Process Intensification, 2022, 181: 109164.
[3]	HAFEEZ Sanaa, MAHMOOD Sabbir, ARISTODEMOU Elsa, et al. Process simulation modelling of the catalytic hydrodeoxygenation of 4-propylguaiacol in microreactors[J]. Fuels, 2021, 2(3): 272-285.
[4]	WANG Yancheng, WU Qiong, MEI Deqing, et al. Development of highly efficient methanol steam reforming system for hydrogen production and supply for a low temperature proton exchange membrane fuel cell[J]. International Journal of Hydrogen Energy, 2020, 45(46): 25317-25327.
[5]	SHEN Mengqi, BENDEL Christoph, VIBBERT Hunter B, et al. Maximizing hydrogen utilization efficiency in tandem hydrogenation of nitroarenes with ammonia borane[J]. Green Chemistry, 2023, 25(18): 7183-7188.
[6]	CÓRDOBA Rodríguez Maria, PORTILLO Esmeralda, NAVARRETE Benito. Simulation and economic analysis of hydrogen production from dry reforming of methane[J]. Available at SSRN 4708003.
[7]	R V Tona VÁSQUEZ, JIMÉNEZ ESTELLER L, BOJARSKI A D. Multiscale modeling approach for production of perfume microcapsules[J]. Chemical Engineering & Technology, 2008, 31(8): 1216-1222.
[8]	LIU Shuaifeng, ZHANG Xiaojuan, ASSELIN Edouard, et al. A new process for peracetic acid production from acetic acid and hydrogen peroxide based on kinetic modeling and distillation simulation[J]. Industrial & Engineering Chemistry Research, 2022, 61(1): 339-348.
[9]	DING Zhiming, ZHANG Yaheng, RUAN Jian, et al. Development of an electrophotochemical flow microreactor for efficient electrophotocatalytic C-H hydroxylation of benzene to phenol[J]. Chemical Engineering Science, 2024: 119900.
[10]	SATO Eisuke, FUJII Mayu, TANAKA Hiroki, et al. Application of an electrochemical microflow reactor for cyanosilylation: Machine learning-assisted exploration of suitable reaction conditions for semi-large-scale synthesis[J]. The Journal of Organic Chemistry, 2021, 86(22): 16035-16044.
[11]	KARAN Dogancan, CHEN Guoying, JOSE Nicholas, et al. A machine learning-enabled process optimization of ultra-fast flow chemistry with multiple reaction metrics[J]. Reaction Chemistry & Engineering, 2024, 9(3): 619-629.
[12]	RACKAUCKAS Christopher, MA Yingbo, MARTENSEN Julius, et al. Universal differential equations for scientific machine learning[EB/OL]. 2020: 2001.04385..
[13]	WU Hao, ZHAO Jinsong. Fault detection and diagnosis based on transfer learning for multimode chemical processes[J]. Computers & Chemical Engineering, 2020, 135: 106731.
[14]	ZHUANG Fuzhen, QI Zhiyuan, DUAN Keyu, et al. A comprehensive survey on transfer learning[J]. Proceedings of the IEEE, 2021, 109(1): 43-76.
[15]	GUO Jingjing, DU Wenli, NASCU Ioana. Adaptive modeling of fixed-bed reactors with multicycle and multimode characteristics based on transfer learning and just-in-time learning[J]. Industrial & Engineering Chemistry Research, 2020, 59(14): 6629-6637.
[16]	XIAO Ming, VELLAYAPPAN Keerthana, PRAVIN P S, et al. Optimization-based multi-source transfer learning for modeling of nonlinear processes[J]. Chemical Engineering Science, 2024, 295: 120117.
[17]	BARDEENIZ Santi, PANJAPORNPON Chanin, FONGSAMUT Chalermpan, et al. Digital twin-aided transfer learning for energy efficiency optimization of thermal spray dryers: Leveraging shared drying characteristics across chemicals with limited data[J]. Applied Thermal Engineering, 2024, 242: 122431.
[18]	BI Kexin, BEYKAL Burcu, AVRAAMIDOU Styliani, et al. Integrated modeling of transfer learning and intelligent heuristic optimization for a steam cracking process[J]. Industrial & Engineering Chemistry Research, 2020, 59(37): 16357-16367.
[19]	LONG Mingsheng, CAO Zhangjie, WANG Jianmin, et al. Conditional adversarial domain adaptation[J]. Advances in Neural Information Processing Systems, 2018, 31.
[20]	CAMERON Ian T, GANI Rafiqul. Product and process modelling: A case study approach[M]. Oxford: Elsevier, 2011.
[21]	王晋东, 陈益强. 迁移学习导论[M]. 北京：电子工业出版社， 2021.
	WANG Jindong， CHEN Yiqiang. Introduction to transfer learning [M]. Beijing: Publishing House of Electronics Industry, 2021.
[22]	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification[C]//Proceedings of the IEEE International Conference on Computer Vision. 2015: 1026-1034.
[23]	Zimeng LYU, ELSAID AdbEIRahman, KARNS Joshua, et al. An experimental study of weight initialization and lamarckian inheritance on neuroevolution[C]//International Conference on the Applications of Evolutionary Computation (Part of EvoStar). Cham: Springer International Publishing, 2021: 584-600.
[24]	赵德普, 张吉顺, 罗传义. 用三氧化硫直接磺化DDB的宏观动力学研究[J]. 化学反应工程与工艺, 1991, 7(2): 191-197.
	ZHAO Depu, ZHANG Jishun, LUO Chuanyi. Macroscopic Kinetic study of direct sulfonation of DDB with sulfur trioxide[J]. Chemical Reaction Engineering and Technology, 1991, 7(2): 191-197.
[25]	GRETTON Arthur, BORGWARDT Karsten M, RASCH Malte J, et al. A kernel two-sample test[J]. Journal of Machine Learning Research, 2012, 13: 723-773.
[26]	Stanislaw WĘGLARCZYK. Kernel density estimation and its application[C]//ITM Web of Conferences. EDP Sciences, 2018, 23: 00037.
[27]	CAI Tony T, MA Rong. Theoretical foundations of t-SNE for visualizing high-dimensional clustered data[J]. Journal of Machine Learning Research, 2022, 23(301): 1-54.
[28]	HUANG Lei, QIN Jie, ZHOU Yi, et al. Normalization techniques in training DNNS: Methodology, analysis and application[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(8): 10173-10196.
[29]	NAKAGAWA Shinichi, JOHNSON Paul C D, SCHIELZETH Holger. The coefficient of determination R ² and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded[J]. Journal of the Royal Society Interface, 2017, 14(134): 20170213.

编号	反应器温度/℃	SO₃-DDB摩尔比	SO₃体积分数/%	气体体积流量/mL·min^-1	DBSA收率/%
1	30	1.2	10	400	73.11
2	40	1	8	400	77.75
3	40	1.2	6	400	88.13
4	40	1.2	8	200	92.36
5	40	1.2	8	600	73.19
…
34	60	1.2	8	400	90.8
35	60	1.2	8	600	79.91
36	60	1.2	10	400	86.85
37	60	1.4	8	400	97.8
38	70	1.2	10	400	85.37

编号	反应器温度/℃	SO₃-DDB摩尔比	SO₃体积分数/%	气体体积流量/mL·min^-1	DBSA收率/%
1	30	1.2	10	400	73.11
2	40	1	8	400	77.75
3	40	1.2	6	400	88.13
4	40	1.2	8	200	92.36
5	40	1.2	8	600	73.19
…
34	60	1.2	8	400	90.8
35	60	1.2	8	600	79.91
36	60	1.2	10	400	86.85
37	60	1.4	8	400	97.8
38	70	1.2	10	400	85.37

编号	反应器温度/℃	SO₃-DDB摩尔比	SO₃体积分数/%	气体体积流量/mL·min^-1	DBSA收率/%
1	20	1	2	100	53.31
2	20	1.1	2	100	55.09
3	20	1.2	2	100	56.71
4	20	1.3	2	100	58.11
5	20	1	2	100	53.31
…
15150	55	1.2	16	650	82.43
15151	55	1.3	16	650	84.47
15152	55	1.4	16	650	86.2
15153	55	1.5	16	650	87.67
15154	55	1.6	16	650	88.91
…
29696	90	1.6	20	950	88.96
29697	90	1.7	20	950	89.87
29698	90	1.8	20	950	90.64
29699	90	1.9	20	950	91.3
29700	90	2	20	950	91.88

编号	反应器温度/℃	SO₃-DDB摩尔比	SO₃体积分数/%	气体体积流量/mL·min^-1	DBSA收率/%
1	20	1	2	100	53.31
2	20	1.1	2	100	55.09
3	20	1.2	2	100	56.71
4	20	1.3	2	100	58.11
5	20	1	2	100	53.31
…
15150	55	1.2	16	650	82.43
15151	55	1.3	16	650	84.47
15152	55	1.4	16	650	86.2
15153	55	1.5	16	650	87.67
15154	55	1.6	16	650	88.91
…
29696	90	1.6	20	950	88.96
29697	90	1.7	20	950	89.87
29698	90	1.8	20	950	90.64
29699	90	1.9	20	950	91.3
29700	90	2	20	950	91.88

评价指标	Aspen Plus中PFR模型	基于模拟数据的ANN模型	基于实验数据的ANN模型	CADAN模型
R²	-2.5915	-2.1430	0.8153	0.9346
MSE	432.7011	378.6666	22.2497	7.8798
RMSE	20.8015	19.4594	4.7170	2.8071