View in EDS

Intelligent algorithm-based model for predicting mass transfer performance in CO2 absorption within a rotating packed bed

Saved in:
Bibliographic Details
Title: Intelligent algorithm-based model for predicting mass transfer performance in CO2 absorption within a rotating packed bed
Authors: Wei Zhang, Hao Chen, Ke Huang, Xing Shu, Cheng Fu, Bin Huang
Source: Case Studies in Thermal Engineering, Vol 73, Iss , Pp 106662- (2025)
Publisher Information: Elsevier, 2025.
Publication Year: 2025
Collection: LCC:Engineering (General). Civil engineering (General)
Subject Terms: CO2 absorption, Intelligent algorithms, LSSVM model, Rotating packed bed, Mass transfer, Engineering (General). Civil engineering (General), TA1-2040
Description: The rotating packed bed (RPB) is known for its superior CO2 absorption performance, but accurately predicting the mass transfer coefficient remains challenging due to complex factors like gas-liquid phase properties and operational conditions. Traditional models are limited, and while CFD simulations are precise, they are computationally expensive. To address these challenges, this study introduces three advanced intelligent models aimed at improving the prediction of the mass transfer coefficient. Using dimensional analysis, key factors are transformed into dimensionless numbers, which are then input into models integrating least squares support vector machine (LSSVM) with genetic algorithm (GA), particle swarm optimization (PSO), and a hybrid GA-PSO (HGAPSO). The models demonstrate excellent predictive accuracy, with determination coefficients of 0.9982, 0.9987, and 0.9998 for the training set, and 0.9974, 0.9982, and 0.9983 for the testing set, respectively. Furthermore, the study introduces a Relevancy Factor to quantify the relative significance of each input variable in influencing the output. The Relevancy values for key dimensionless numbers, including ReG, ReL, ScL, GrG, and Mr, are found to be 0.547566, 0.541792, 0.225272, −0.173864, and 0.139201, respectively. Moreover, the occurrence of suspected data points is found to be minimal, with frequencies of 0.64 %, 3.8 %, and 1.28 % for the GA-LSSVM, PSO-LSSVM, and HGAPSO-LSSVM models, respectively. This limited occurrence of suspected data strongly indicates that the dataset employed in this study is robust, ensuring a high degree of reliability in the model predictions. A comparative analysis with three empirical models, a multiple nonlinear regression model and a standard LSSVM model further illustrates the superior predictive accuracy and generalization ability of the proposed intelligent models. The findings of this study suggest that the integration of intelligent optimization algorithms with LSSVM offers a promising alternative to conventional methods for predicting the mass transfer coefficient in RPB systems, thereby enhancing the efficiency and reliability of CO2 absorption processes.
Document Type: article
File Description: electronic resource
Language: English
ISSN: 2214-157X
Relation: http://www.sciencedirect.com/science/article/pii/S2214157X25009220; https://doaj.org/toc/2214-157X
DOI: 10.1016/j.csite.2025.106662
Access URL: https://doaj.org/article/f8d016f72f8d48628f10c93006622afc
Accession Number: edsdoj.f8d016f72f8d48628f10c93006622afc
Database: Directory of Open Access Journals
Description
Abstract:The rotating packed bed (RPB) is known for its superior CO2 absorption performance, but accurately predicting the mass transfer coefficient remains challenging due to complex factors like gas-liquid phase properties and operational conditions. Traditional models are limited, and while CFD simulations are precise, they are computationally expensive. To address these challenges, this study introduces three advanced intelligent models aimed at improving the prediction of the mass transfer coefficient. Using dimensional analysis, key factors are transformed into dimensionless numbers, which are then input into models integrating least squares support vector machine (LSSVM) with genetic algorithm (GA), particle swarm optimization (PSO), and a hybrid GA-PSO (HGAPSO). The models demonstrate excellent predictive accuracy, with determination coefficients of 0.9982, 0.9987, and 0.9998 for the training set, and 0.9974, 0.9982, and 0.9983 for the testing set, respectively. Furthermore, the study introduces a Relevancy Factor to quantify the relative significance of each input variable in influencing the output. The Relevancy values for key dimensionless numbers, including ReG, ReL, ScL, GrG, and Mr, are found to be 0.547566, 0.541792, 0.225272, −0.173864, and 0.139201, respectively. Moreover, the occurrence of suspected data points is found to be minimal, with frequencies of 0.64 %, 3.8 %, and 1.28 % for the GA-LSSVM, PSO-LSSVM, and HGAPSO-LSSVM models, respectively. This limited occurrence of suspected data strongly indicates that the dataset employed in this study is robust, ensuring a high degree of reliability in the model predictions. A comparative analysis with three empirical models, a multiple nonlinear regression model and a standard LSSVM model further illustrates the superior predictive accuracy and generalization ability of the proposed intelligent models. The findings of this study suggest that the integration of intelligent optimization algorithms with LSSVM offers a promising alternative to conventional methods for predicting the mass transfer coefficient in RPB systems, thereby enhancing the efficiency and reliability of CO2 absorption processes.
ISSN:2214157X
DOI:10.1016/j.csite.2025.106662