Multiple machine learning models for prediction of CO2 solubility in potassium and sodium based amino acid salt solutions

Reduced Order Machine Learning Models for Accurate Prediction of CO2 Capture in Physical Solvents

“…A drastic drop in the RMSE from 0.05 to 0.003 (train data) and from 0.07 to 0.006 (test data) was observed for r-KRR with 4 KDs as compared to that of the KRR with 7 KDs, which indicates that the additional descriptors, including boiling point of the solvent, temperature, and the partial pressure of CO 2 , significantly improved the prediction accuracy of the KRR model. These results are in good agreement with those reported in the literature, in which partial pressure of CO 2 and temperature were identified as the important descriptors for the prediction of CO 2 solubility in amine solutions and potassium- and sodium-based amino acid solutions . Further, from the zoomed part of Figure b, an inflection point in the RMSE can be observed for the r-KRR model with 9 KDs for test data, beyond which it remained almost stable.…”

“…These results are in good agreement with those reported in the literature, in which partial pressure of CO 2 and temperature were identified as the important descriptors for the prediction of CO 2 solubility in amine solutions 33 and potassium-and sodium-based amino acid solutions. 32 Further, from the zoomed part of Figure 2b, an inflection point in the RMSE can be observed for the r-KRR model with 9 KDs for test data, beyond which it remained almost stable. Therefore, the r-KRR model with 9 KDs can be chosen to be the optimum model for predicting the CO 2 solubility in physical solvents.…”

“…52−54 Physical absorption deals with the mass transfer between the gas and liquid phases, and it is driven majorly by the gas−liquid interactions. 32 And the descriptors such as No. of C atoms, molecular weight, No.…”

“…Further, the validity and relevance of the KDs resulting from this study were compared with the findings of different studies reported in the literature. All these KDs significantly influence the gas–liquid surface interactions and absorption rates of solvent at the molecular level that in turn affect the CO 2 solubility. − Physical absorption deals with the mass transfer between the gas and liquid phases, and it is driven majorly by the gas–liquid interactions . And the descriptors such as No.…”

“…Such reduced-order models are very useful, especially for advanced real-time process optimization and inferential control applications. 31 Considering the progress reported on ML models for the prediction of CO 2 solubility, very few studies include the KD ranking for the cases of salt, 32 aqueous amine solutions, 33 and ionic solvents, 34 and most importantly, none of them report the reduced order models.…”