neural network modeling for development of high-pressure measurement of carbon dioxide solubility in the aqueous aeea+sulfolane

Due to increasing concerns about global warming regarding co2 release to the atmosphere, various methods are used to capture co2, among which chemical absorption via amine mixture solutions is very well developed. a set of 179 data related to co2 absorption in a mixture, including a physical absorbent (sulfolane) and a chemical absorption (aeea) in a wide range of temperature, pressure and solvent concentration is used to develop two artificial neural networks (ann). in multi-layer perceptron (mlp), the levenberg-marquardt method is used to train the network. most important factors such as regression analysis value (r2) of 0.99963, mean squared error (mse) value of 1.22e-05 and average absolute relative deviation value (%aard) of 0.2671 factors reveal that the mlp network has a high capability to predict co2 loading (αco2). also, a radial basis function (rbf) network was developed. rbf network with a spread value of 2.2 and 138 neurons had an outstanding performance and achieved an mse value of 2.53e-05 along with an r2 value of 0.99993, 11 seconds, and a %aard value of 0.1460. according to experimental and predicted data, the neural networks are well trained and are able to predict co2 loading precisely in an economic and optimized way.