Global warming has become a very popular subject of our time, developing a great need to capture carbon from our atmosphere. One method is through Direct-Air Capture (DAC), where a chemical absorbent is used to selectively absorb carbon dioxide (CO2) from air or flue gas. However, developing solvents for carbon capture is a challenging task due to complex molecular structures and costly experimentation. This study investigates the use of an optimized Artificial Neural Network (ANN) model combined the Group Contribution Method (GCM) to predict the best ionic liquid (IL) structures, with capabilities of achieving high CO2 solubility. Increasing their effectiveness in carbon capture applications. A database has been previously compiled containing 10,310 data measurements of CO2 solubilities in various ILs. These solubility values have been experimentally tested at different temperatures and pressures. Based on this database, the relationship between IL structure, CO2 solubility, temperature, and pressure are correlated using GCM. By utilizing GCM, the IL molecules are separated into smaller functional groups, incorporated into an ANN model, permitting an enhanced prediction process. The ANN model utilizes a tree-based algorithm and demonstrates robust test accuracy in predicting CO2 solubility. The performance of the model was analyzed using statistical metrics such as the coefficient of determination (R2) and mean absolute error (MAE), resulting in values of 0.986 and 0.0155, respectively. It is believed that the computational cost of this model is significantly lower than the essential computational time of other models. Using this newly developed model, the ultimate goal would be to propose a novel IL structure. One that can maximize CO2 solubility, while maintaining cost-effectiveness and ease of production. Future research will be needed to find ways to synthesize and test this proposed molecule.
