Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data

Pressure swing adsorption (PSA) modeling remains a challenging task since it exhibits strong dynamic and cyclic behavior. This study presents a systematic physics-informed machine learning method that integrates transfer learning and labeled data to construct a spatiotemporal model of the PSA proces...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhiqiang Wu, Yunquan Chen, Bingjian Zhang, Jingzheng Ren, Qinglin Chen, Huan Wang, Chang He
Format: Article
Language:English
Published: KeAi Communications Co. Ltd. 2025-06-01
Series:Green Chemical Engineering
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2666952824000591
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850036396185616384
author Zhiqiang Wu
Yunquan Chen
Bingjian Zhang
Jingzheng Ren
Qinglin Chen
Huan Wang
Chang He
author_facet Zhiqiang Wu
Yunquan Chen
Bingjian Zhang
Jingzheng Ren
Qinglin Chen
Huan Wang
Chang He
author_sort Zhiqiang Wu
collection DOAJ
description Pressure swing adsorption (PSA) modeling remains a challenging task since it exhibits strong dynamic and cyclic behavior. This study presents a systematic physics-informed machine learning method that integrates transfer learning and labeled data to construct a spatiotemporal model of the PSA process. To approximate the latent solutions of partial differential equations (PDEs) in the specific steps of pressurization, adsorption, heavy reflux, counter-current depressurization, and light reflux, the system's network representation is decomposed into five lightweight sub-networks. On this basis, we propose a parameter-based transfer learning (TL) combined with domain decomposition to address the long-term integration of periodic PDEs and expedite the network training process. Moreover, to tackle challenges related to sharp adsorption fronts, our method allows for the inclusion of a specified amount of labeled data at the boundaries and/or within the system in the loss function. The results show that the proposed method closely matches the outcomes achieved through the conventional numerical method, effectively simulating all steps and cyclic behavior within the PSA processes.
format Article
id doaj-art-bc1520cc3209455cb8ecb4914c33aeb7
institution DOAJ
issn 2666-9528
language English
publishDate 2025-06-01
publisher KeAi Communications Co. Ltd.
record_format Article
series Green Chemical Engineering
spelling doaj-art-bc1520cc3209455cb8ecb4914c33aeb72025-08-20T02:57:08ZengKeAi Communications Co. Ltd.Green Chemical Engineering2666-95282025-06-016223324810.1016/j.gce.2024.08.004Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled dataZhiqiang Wu0Yunquan Chen1Bingjian Zhang2Jingzheng Ren3Qinglin Chen4Huan Wang5Chang He6School of Materials Science and Engineering, Sun Yat-Sen University, Guangzhou, 510275, ChinaSchool of Materials Science and Engineering, Sun Yat-Sen University, Guangzhou, 510275, ChinaSchool of Materials Science and Engineering, Sun Yat-Sen University, Guangzhou, 510275, China; Guangdong Engineering Center for Petrochemical Energy Conservation, Guangzhou, 510275, ChinaDepartment of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong Special Administrative Region, ChinaGuangdong Engineering Center for Petrochemical Energy Conservation, Guangzhou, 510275, China; School of Chemical Engineering and Technology, Sun Yat-sen University, Zhuhai, 519082, ChinaState Key Laboratory of Heavy Oil Processing, China University of Petroleum-Beijing, Beijing, 102249, China; CNNC No. 7 Research and Design Institute Co. Ltd., Taiyuan, 030012, China; Corresponding author.Guangdong Engineering Center for Petrochemical Energy Conservation, Guangzhou, 510275, China; School of Chemical Engineering and Technology, Sun Yat-sen University, Zhuhai, 519082, China; Corresponding author.Pressure swing adsorption (PSA) modeling remains a challenging task since it exhibits strong dynamic and cyclic behavior. This study presents a systematic physics-informed machine learning method that integrates transfer learning and labeled data to construct a spatiotemporal model of the PSA process. To approximate the latent solutions of partial differential equations (PDEs) in the specific steps of pressurization, adsorption, heavy reflux, counter-current depressurization, and light reflux, the system's network representation is decomposed into five lightweight sub-networks. On this basis, we propose a parameter-based transfer learning (TL) combined with domain decomposition to address the long-term integration of periodic PDEs and expedite the network training process. Moreover, to tackle challenges related to sharp adsorption fronts, our method allows for the inclusion of a specified amount of labeled data at the boundaries and/or within the system in the loss function. The results show that the proposed method closely matches the outcomes achieved through the conventional numerical method, effectively simulating all steps and cyclic behavior within the PSA processes.http://www.sciencedirect.com/science/article/pii/S2666952824000591Physics-informed machine learningPressure swing adsorptionTransfer learningLabeled dataPartial differential equations
spellingShingle Zhiqiang Wu
Yunquan Chen
Bingjian Zhang
Jingzheng Ren
Qinglin Chen
Huan Wang
Chang He
Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
Green Chemical Engineering
Physics-informed machine learning
Pressure swing adsorption
Transfer learning
Labeled data
Partial differential equations
title Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
title_full Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
title_fullStr Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
title_full_unstemmed Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
title_short Pressure swing adsorption process modeling using physics-informed machine learning with transfer learning and labeled data
title_sort pressure swing adsorption process modeling using physics informed machine learning with transfer learning and labeled data
topic Physics-informed machine learning
Pressure swing adsorption
Transfer learning
Labeled data
Partial differential equations
url http://www.sciencedirect.com/science/article/pii/S2666952824000591
work_keys_str_mv AT zhiqiangwu pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT yunquanchen pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT bingjianzhang pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT jingzhengren pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT qinglinchen pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT huanwang pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata
AT changhe pressureswingadsorptionprocessmodelingusingphysicsinformedmachinelearningwithtransferlearningandlabeleddata