Exploring CO₂ solubility in 1-N-butyl-3-methylimidazolium hexafluorophosphate ionic liquid using neural network...

Area of Science:

• Chemical Engineering
• Computational Chemistry

Background:

• Carbon capture is crucial for mitigating climate change.
• Ionic liquids, such as [Bmim][PF6], show promise for CO2 absorption.
• Experimental CO2 solubility studies in ionic liquids can be challenging.

Purpose of the Study:

• To develop and compare artificial neural network models for predicting CO2 mole fraction in [Bmim][PF6].
• To identify the most accurate model for estimating CO2 solubility under varying conditions.

Main Methods:

• Development of Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) neural network models.
• Training the MLP model using the Levenberg-Marquardt algorithm with specific activation functions.
• Determination of weight matrices, bias vectors, and activation functions for the MLP model.

Main Results:

• The MLP neural network model was identified as the most accurate for estimating CO2 mole fraction.
• The MLP model achieved low relative deviations at lower temperatures (283.15 K).
• The model's predictive accuracy decreased significantly at higher temperatures (298.15 K and 323.15 K).

Conclusions:

• MLP neural networks offer a viable computational approach for modeling CO2 solubility in ionic liquids.
• The accuracy of the developed MLP model is temperature-dependent, showing limitations at higher temperatures.
• Further research may focus on improving model performance across a wider temperature range for CO2 capture applications.