一种顺序式模拟移动床混合建模与优化策略

doi:10.11949/0438-1157.20251079

摘要/Abstract

摘要：

模型是模拟移动床工艺优化的基础。针对机理模型和数据驱动模型各自的不足，提出一种在机理模型中嵌入神经网络的混合建模方法。首先，通过双柱实验采集数据，利用双分支神经网络学习系统状态与吸附速率之间的非线性映射；然后，结合差分进化与网格搜索辨识模型结构与参数，构建顺序式模拟移动床系统；最后，以低聚木糖的纯化过程进行验证。结果表明，该模型能准确预测双柱实验与顺序式模拟移动床的出口浓度，并在工艺优化中获得连续均匀的帕累托前沿，目标产物纯度和收率的预测误差均低于2%。

关键词: 模拟移动床, 色谱, 分离, 混合建模, 工艺优化, 神经网络

Abstract:

The simulated moving bed model serves as the foundation for process optimization. To overcome the respective limitations of mechanistic and data-driven models, a hybrid modeling approach is proposed by embedding neural networks into the mechanistic framework. First, data were collected from twin-column experiments, and a dual-branch neural network was employed to learn the nonlinear mapping between system states and adsorption rates. Then, differential evolution combined with grid search was used to identify the model structure and parameters, upon which a sequential simulated moving bed system was constructed. Finally, the proposed method was validated through the purification of xylo-oligosaccharides. Results demonstrate that the model can accurately predict the outlet concentrations in both twin-column experiments and sequential simulated moving beds. Furthermore, process optimization based on this model yielded a continuous and uniformly distributed Pareto front, with relative prediction errors of product purity and recovery below 2%.

Key words: simulated moving bed, chromatography, separation, hybrid modeling, process optimization, neural networks

中图分类号:

TQ 028.8

刘鹏鹏, 王志国, 栾小丽, 刘飞. 一种顺序式模拟移动床混合建模与优化策略[J]. 化工学报, DOI: 10.11949/0438-1157.20251079.

Pengpeng LIU, Zhiguo WANG, Xiaoli LUAN, Fei LIU. A hybrid modeling and optimization strategy for sequential simulated moving bed[J]. CIESC Journal, DOI: 10.11949/0438-1157.20251079.

图/表 17

参考文献 38

[1]	Broughton D B. Production-scale adsorptive separations of liquid mixtures by simulated moving-bed technology[J]. Separation Science and Technology, 1984, 19(11/12): 723-736.
[2]	Aniceto J P S, Silva C M. Simulated moving bed strategies and designs: from established systems to the latest developments[J]. Separation & Purification Reviews, 2015, 44(1): 41-73.
[3]	Kim K M, Han K W, Kim S I, et al. Simulated moving bed with a product column for improving the separation performance[J]. Journal of Industrial and Engineering Chemistry, 2020, 88: 328-338.
[4]	Barker P E, Abusabah E K E. The separation of synthetic mixtures of glucose and fructose and also inverted sucrose feedstocks using countercurrent chromatographic techniques[J]. Chromatographia, 1985, 20(1): 9-12.
[5]	Nagamatsu S, Murazumi K, Makino S. Chiral separation of a pharmaceutical intermediate by a simulated moving bed process[J]. Journal of Chromatography A, 1999, 832(1/2): 55-65.
[6]	Wu C K, Xu Y Z, Bai Z Y, et al. High-purity tri-alpha-linolenin isolation from silkworm pupae oil using sequential simulated moving bed chromatography[J]. Biomass Conversion and Biorefinery, 2023, 13(18): 16453-16466.
[7]	李艳, 雷美丽, 李鑫钢. 基于分离性能的顺序式模拟移动床结构调控策略[J]. 化工学报, 2025, 76(5): 2219-2229.
	Li Y, Lei M L, Li X G. Regulation strategy of sequential simulated moving bed structure based on separation performance[J]. CIESC Journal, 2025, 76(5): 2219-2229.
[8]	李艳, 凌山, 刘聚明, 等. 顺序式模拟移动床分离低聚木糖多目标优化及其变量调控机制[J]. 现代化工, 2023, 43(1): 240-245.
	Li Y, Ling S, Liu J M, et al. Multi-objective optimization of sequentially-simulated moving bed for separation of xylo-oligosaccharides and its variables regulation mechanism[J]. Modern Chemical Industry, 2023, 43(1): 240-245.
[9]	Zhang X T, Liu J M, Ray A K, et al. Multi-objective optimization of sequential simulated moving bed for multi-component separation based on multi-parameter modeling and its determination[J]. Chemical Engineering Science, 2024, 295: 120172.
[10]	Li Y, Ding Z Y, Wang J, et al. A comparison between simulated moving bed and sequential simulated moving bed system based on multi-objective optimization[J]. Chemical Engineering Science, 2020, 219: 115562.
[11]	Li Y, Xu J, Yu W F, et al. Multi-objective optimization of sequential simulated moving bed for the purification of xylo-oligosaccharides[J]. Chemical Engineering Science, 2020, 211: 115279.
[12]	李洪飞, 孙大庆, 李良玉, 等. 基于顺序式模拟移动床色谱法的两种木糖母液分离工艺比较[J]. 食品与机械, 2019, 35(10): 210-213.
	Li H F, Sun D Q, Li L Y, et al. Comparing of two seperation processes for recovering xylose mother liquor with sequential simulated moving bed technology[J]. Food & Machinery, 2019, 35(10): 210-213.
[13]	曹俊雅, 宋恕哲, 贺鹏, 等. 顺序式模拟移动床分离煤基混合醇的Aspen色谱动态模拟[J]. 化工进展, 2025, 44(3): 1228-1242.
	Cao J Y, Song S Z, He P, et al. Dynamic simulation using Aspen chromatography for sequential simulated moving bed separation of coal-based mixed alcohols[J]. Chemical Industry and Engineering Progress, 2025, 44(3): 1228-1242.
[14]	Bürger R, Mulet P, Rubio L, et al. Linearly implicit-explicit schemes for the equilibrium dispersive model of chromatography[J]. Applied Mathematics and Computation, 2018, 317: 172-186.
[15]	Vajda P, Bocian S, Buszewski B, et al. Examination of the surface heterogeneity of reversed-phase packing materials with solvent adsorption[J]. Journal of Separation Science, 2010, 33(23/24): 3644-3654.
[16]	Shekhawat L K, Rathore A S. An overview of mechanistic modeling of liquid chromatography[J]. Preparative Biochemistry & Biotechnology, 2019, 49(6): 623-638.
[17]	Mao S M, Zhang Y, Rohani S, et al. Chromatographic resolution and isotherm determination of (R, S)-mandelic acid on Chiralcel-OD column[J]. Journal of Separation Science, 2012, 35(17): 2273-2281.
[18]	Xu J, Zhu L, Xu G Q, et al. Determination of competitive adsorption isotherm of enantiomers on preparative chromatographic columns using inverse method[J]. Journal of Chromatography A, 2013, 1273: 49-56.
[19]	Poole C F. Fundamentals of preparative and nonlinear chromatography[J]. Analytica Chimica Acta, 1995, 302(1): 127-128.
[20]	Qamar S, Abbasi J N, Mehwish A, et al. Linear general rate model of chromatography for core–shell particles: Analytical solutions and moment analysis[J]. Chemical Engineering Science, 2015, 137: 352-363.
[21]	Mukhopadhyay S, Tsang Y W, Finsterle S. Parameter estimation from flowing fluid temperature logging data in unsaturated fractured rock using multiphase inverse modeling[J]. Water Resources Research, 2009, 45(4): 2008WR006869.
[22]	Wang X, Liu Y, Ching C B. Kinetic and equilibrium study of enantioseparation of propranolol in preparative scale chromatography[J]. Separation and Purification Technology, 2006, 50(2): 204-211.
[23]	Mueller I, Seidel-Morgenstern A, Hamel C. Simulated-moving-bed technology for purification of the prebiotics galacto-oligosaccharides[J]. Separation and Purification Technology, 2021, 271: 118829.
[24]	Sherstinsky A. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network[J]. Physica D: Nonlinear Phenomena, 2020, 404: 132306.
[25]	Abbasimehr H, Shabani M, Yousefi M. An optimized model using LSTM network for demand forecasting[J]. Computers & Industrial Engineering, 2020, 143: 106435.
[26]	Narayanan H, Luna M, Sokolov M, et al. Hybrid models based on machine learning and an increasing degree of process knowledge: application to capture chromatographic step[J]. Industrial & Engineering Chemistry Research, 2021, 60(29): 10466-10478.
[27]	Ding C, Yang M L, Zhao Y M, et al. Graph convolutional network for axial concentration profiles prediction in simulated moving bed[J]. Chinese Journal of Chemical Engineering, 2024, 73: 270-280.
[28]	Wang Y Q, Wen J W, Wang Z G, et al. Modeling and learning-based predictive control for separation processes of simulated moving beds[J]. Chemical Engineering Research and Design, 2025, 220: 292-299.
[29]	Santana V V, Martins M A F, Loureiro J M, et al. Novel framework for simulated moving bed reactor optimization based on deep neural network models and metaheuristic optimizers: an approach with optimality guarantee[J]. ACS Omega, 2023, 8(7): 6463-6475.
[30]	Raissi M, Perdikaris P, Karniadakis G E. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J]. Journal of Computational Physics, 2019, 378: 686-707.
[31]	Cai S Z, Mao Z P, Wang Z C, et al. Physics-informed neural networks (PINNs) for fluid mechanics: a review[J]. Acta Mechanica Sinica, 2021, 37(12): 1727-1738.
[32]	Narayanan H, Seidler T, Luna M F, et al. Hybrid Models for the simulation and prediction of chromatographic processes for protein capture[J]. Journal of Chromatography A, 2021, 1650: 462248.
[33]	Ding C Y, Gerberich C, Ierapetritou M. Hybrid model development for parameter estimation and process optimization of hydrophobic interaction chromatography[J]. Journal of Chromatography A, 2023, 1703: 464113.
[34]	Deb K, Pratap A, Agarwal S, et al. A fast and elitist multiobjective genetic algorithm: NSGA-II[J]. IEEE Transactions on Evolutionary Computation, 2002, 6(2): 182-197.
[35]	Arora S, Dhaliwal S S, Kukreja V K. Solution of two point boundary value problems using orthogonal collocation on finite elements[J]. Applied Mathematics and Computation, 2005, 171(1): 358-370.
[36]	Arora S, Dhaliwal S S, Kukreja V K. Simulation of washing of packed bed of porous particles by orthogonal collocation on finite elements[J]. Computers & Chemical Engineering, 2006, 30(6/7): 1054-1060.
[37]	魏朋, 陈珺, 王志国, 等. 基于平衡理论的模拟移动床工艺参数鲁棒寻优[J]. 化工学报, 2022, 73(2): 792-800.
	Wei P, Chen J, Wang Z G, et al. Robust optimization of process parameters of simulated moving bed based on equilibrium theory[J]. CIESC Journal, 2022, 73(2): 792-800.
[38]	Piotrowski A P. L-SHADE optimization algorithms with population-wide inertia[J]. Information Sciences, 2018, 468: 117-141.

设备与进料参数		吸附特性参数
柱分布结构	1-1-1-1	理论塔板数	2000
柱直径	2.5 cm	A的传质系数	4.54 min^-1
柱长	100 cm	B的传质系数	0.36 min^-1
空隙率	0.416	A的亨利常数	0.450
最大允许流量	20 ml/min	B的亨利常数	0.167
A的进料浓度	0.09 g/ml	A的吸附常数	1.11
B的进料浓度	0.21 g/ml	B的吸附常数	0.467

设备与进料参数		吸附特性参数
柱分布结构	1-1-1-1	理论塔板数	2000
柱直径	2.5 cm	A的传质系数	4.54 min^-1
柱长	100 cm	B的传质系数	0.36 min^-1
空隙率	0.416	A的亨利常数	0.450
最大允许流量	20 ml/min	B的亨利常数	0.167
A的进料浓度	0.09 g/ml	A的吸附常数	1.11
B的进料浓度	0.21 g/ml	B的吸附常数	0.467

参数	数值
初始种群规模	60
最小种群规模	2
最大迭代次数(G_max)	300
记忆长度	10
p-best比例	0.11
迭代间隔(Δg)	10
相对改善阈值(ε_rel)	2%
重复运行次数	5

参数	数值
初始种群规模	60
最小种群规模	2
最大迭代次数(G_max)	300
记忆长度	10
p-best比例	0.11
迭代间隔(Δg)	10
相对改善阈值(ε_rel)	2%
重复运行次数	5

网络结构	WSSE_total		网络结构	WSSE_total
网络结构	sigmoid	tanh	网络结构	sigmoid	tanh
[1-2]	0.0452	0.0335	[2,2]	0.0417	0.0402
[1,3]	0.0322	0.0286	[2-3]	0.0368	0.0402
[1,4]	0.0414	0.0376	[2,4]	0.0355	0.0342
[1,5]	0.0363	0.0332	[2,5]	0.0364	0.0358