A robust predictive tool for estimating CO2 solubility in potassium based amino acid salt solutions

The acid gas absorption in four potassium based amino acid salt solutions was predicted using artificial neural network (ANN). Two hundred fifty-five experimental data points for CO2 absorption in the four potassium based amino acid salt solutions containing potassium lysinate, potassium prolinate, potassium glycinate, and potassium taurate were used in this modeling. Amine salt solution's type, temperature, equilibrium partial pressure of acid gas, the molar concentration of the solution, molecular weight, and the boiling point were considered as inputs to ANN to prognosticate the capacity of amino acid salt solution to absorb acid gas. Regression analysis was employed to assess the performance of the network. Levenberg-Marquardt back-propagation algorithm was used to train the optimal ANN with 5:12:1 architecture. The model findings indicated that the proposed ANN has the capability to predict precisely the absorption of acid gases in various amino acid salt solutions with Mean Square Error (MSE) value of 0.0011, the Average Absolute Relative Deviation (AARD) percent of 5.54%, and the correlation coefficient (R2) of 0.9828.