Analysis of effective area and mass transfer in a structure packing column using machine learning and response...

Area of Science:

• Chemical Engineering
• Mass Transfer
• Process Modeling

Background:

• Accurate prediction of mass transfer coefficients is crucial for designing efficient CO2 absorption columns.
• Structured packings significantly influence mass transfer performance.
• Traditional methods like Response Surface Methodology (RSM) have limitations in capturing complex relationships.

Purpose of the Study:

• To develop and compare machine learning (ML) models and RSM for predicting mass transfer coefficients (fractional effective area, gas phase, and liquid phase) in structured CO2 absorption columns.
• To identify the key structural parameters of packing that influence mass transfer.
• To evaluate the predictive superiority of ML models over RSM.

Main Methods:

• Utilized four structured packing characteristics: surface area (ap), corrugation angle (θ), channel base (B), and crimp height (h).
• Developed correlations using Response Surface Methodology (RSM).
• Employed five machine learning models: Random Forest (RF), Radial Basis Function Neural Network (RBF), Multilayer Perceptron (MLP), XGB Regressor, and Extra Trees Regressor (ETR).

Main Results:

• RSM derived correlations with high coefficients of determination (R²): 0.9717 for af, 0.9907 for kG, and 0.9323 for kL.
• ML models showed excellent performance: RBF for af (R²=0.9813) and kG (R²=0.9933), and MLP for kL (R²=0.9871).
• Packing channel base (B) most impacted af and kL, while crimp height (h) most affected kG.

Conclusions:

• Machine learning models, specifically RBF and MLP, offer superior predictive accuracy for mass transfer coefficients compared to RSM.
• The study successfully identified key structural parameters influencing mass transfer in structured packings.
• The developed ML models provide a robust tool for optimizing CO2 absorption column design.