Modeling of carbon dioxide absorption into aqueous alkanolamines using machine learning and response surface...

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Process Modeling

Background:

• Alkanolamine solvents are crucial for capturing carbon dioxide (CO2) from industrial emissions.
• Accurate modeling of CO2 absorption is essential for optimizing capture processes.
• Various computational techniques offer potential for developing predictive models.

Purpose of the Study:

• To develop and compare models for predicting CO2 absorption into alkanolamine solvents.
• To evaluate the performance of Multilayer Perceptron (MLP), Radial Basis Function (RBF), Support Vector Machine (SVM), and Response Surface Methodology (RSM).
• To identify optimal network architectures and training algorithms for accurate CO2 mass flux prediction.

Main Methods:

• Utilized MLP, RBF, SVM, and RSM to model CO2 absorption.
• Input parameters included solvent density, mass fraction, temperature, equilibrium constant, CO2 loading, and partial pressure.
• Trained neural networks using trainlm, trainbr, and trainscg algorithms.

Main Results:

• Optimized MLP network structures were determined for one, two, and three layers.
• The best spread for RBF was found to be 2.202.
• Coefficients of determination (R²) reached 0.9996 for MLP and 0.9940 for RBF, indicating high accuracy.
• The trainlm function demonstrated superior performance.

Conclusions:

• MLP and RBF networks, trained with specific algorithms, effectively model CO2 absorption.
• These computational models provide accurate predictions of CO2 mass flux.
• The findings support the use of advanced machine learning techniques in chemical process design and optimization.