Scholarlink

<p>This study introduces a data-driven modeling framework to predict the effective interfacial area in rotating packed beds (RPBs), a critical parameter governing mass transfer efficiency in intensified gas–liquid operations. A curated dataset comprising 250 experimental observations from the literature was used to train and evaluate seven machine learning algorithms: Decision Tree, AdaBoost, Random Forest, k-Nearest Neighbors (KNN), Ensemble Learning, Support Vector Regression (SVR), and Multi-layer Perceptron Artificial Neural Network (MLP-ANN). Model performance was assessed using R<sup>2</sup>, mean squared error (MSE), and average absolute relative error (AARE%), with Decision Tree and Ensemble Learning emerging as the most accurate estimators. Crossplots and relative error analyses confirmed the strong agreement between predicted and experimental values. SHAP (SHapley Additive exPlanations) analysis further revealed that rotating speed is the most influential input feature, significantly impacting model output. The proposed ensemble learning framework demonstrates superior predictive accuracy (AARE = 11.7%) and generalization capability (R<sup>2</sup> = 0.81), outperforming traditional empirical correlations and previously reported models. Its enhanced interpretability and robustness further establish it as a reliable and advanced alternative for modeling interfacial area in RPBs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel data driven approach to model effective interfacial area in rotating packed beds

  • 期刊论文

  • 发表日期：

    2026年1月30日

查看全文
  • Ahmad Adel Abu-Shareha,
  • Ibrahim Khersan,
  • Tariq Abdulkader Alrihaim,
  • S Sujai,
  • L Jino,
  • Sikata Samantaray,
  • Ripendeep Singh,
  • Yashwant Singh Bisht,
  • Abhayveer Singh,
  • Merwa Abolhekmat

摘要

This study introduces a data-driven modeling framework to predict the effective interfacial area in rotating packed beds (RPBs), a critical parameter governing mass transfer efficiency in intensified gas–liquid operations. A curated dataset comprising 250 experimental observations from the literature was used to train and evaluate seven machine learning algorithms: Decision Tree, AdaBoost, Random Forest, k-Nearest Neighbors (KNN), Ensemble Learning, Support Vector Regression (SVR), and Multi-layer Perceptron Artificial Neural Network (MLP-ANN). Model performance was assessed using R2, mean squared error (MSE), and average absolute relative error (AARE%), with Decision Tree and Ensemble Learning emerging as the most accurate estimators. Crossplots and relative error analyses confirmed the strong agreement between predicted and experimental values. SHAP (SHapley Additive exPlanations) analysis further revealed that rotating speed is the most influential input feature, significantly impacting model output. The proposed ensemble learning framework demonstrates superior predictive accuracy (AARE = 11.7%) and generalization capability (R2 = 0.81), outperforming traditional empirical correlations and previously reported models. Its enhanced interpretability and robustness further establish it as a reliable and advanced alternative for modeling interfacial area in RPBs.