Neural Network-Based Prediction and Sensitivity Analysis of CO₂ Solubility in Ionic Liquids

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Materials Science

Background:

• Ionic liquids (ILs) are promising solvents for gas capture due to their tunable properties.
• Understanding CO2 solubility in ILs is crucial for designing efficient separation processes.
• Artificial Neural Networks (ANNs) offer a data-driven approach to model complex chemical phenomena.

Purpose of the Study:

• To develop and validate Artificial Neural Networks (ANNs) for predicting carbon dioxide (CO2) dissolution in the ionic liquid 1-butyl-3-methylimidazolium tetrafluoroborate ([BMIM]-[BF4]).
• To compare the effectiveness of two distinct ANN training algorithms: Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG).
• To assess the predictive capability and generalization of the trained ANNs across a range of temperatures and pressures.

Main Methods:

• Utilized a feedforward multilayer perceptron design for ANNs, implemented in MATLAB.
• Trained the ANNs using experimental data covering temperatures from 228 to 322 K and pressures up to 90 bar.
• Employed Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) learning algorithms for model training and validation.

Main Results:

• The ANN model trained with the LM algorithm achieved a high regression coefficient (R = 0.9999) and low Mean Squared Error (MSE ≈ 10^-1), significantly outperforming the SCG model.
• The trained ANNs demonstrated accurate predictions on unseen data, indicating successful generalization without overfitting.
• Sensitivity analysis confirmed that temperature and pressure are key factors influencing CO2 solubility, aligning with thermodynamic principles.

Conclusions:

• ANNs, particularly when trained with the LM algorithm, provide a reliable and accurate method for predicting CO2 dissolution in [BMIM]-[BF4].
• The developed models effectively capture the complex, nonlinear relationship between CO2 solubility, temperature, and pressure.
• ANNs are a valuable tool for modeling and optimizing gas-solubility processes in chemical engineering applications.