ANN and Mathematical Mass Transfer Modeling of Glycol Amin Liquid Membranes for Separation of Carbon Dioxide from the Air

Aim of this investigation is comparison of mass transfer artificial neural network (ANN) and mathematical model to prediction of carbon dioxide concentration in the exhaust air from the constructed module using glycol-amine liquid membrane. For solving of problem with ANN, command prompt function in Mathlab software was used with following instruction. First, input vector and targets are loaded in Mathlab current directory and a feed-forward tangent sigmoid transfer function in hidden layers and a linear transfer function in output layer was used. Then, the network was trained and Levenberg–Marquardt algorithm was used as training function. For this purpose, 74 input data inclusive input air pressure, input air flow rate, recovery value of CO2 in process was used. The CO2 mole fraction in output air was selected as target. Experimental data was divided in three sectors: 70% for training data, 15% for validation data and 15% for testing network. The optimum number of neurons in the hidden layer network obtained with using try and error method and the best performance of network achieved with four neurons in the hidden layer. Also, the best available models was used to predict mass transfer in liquid membranes, and exponential behavior was seen in modeling of permeating CO2 from membrane. The results of the ANN model were compared with the experimental results and this results were shown that the neural network has great ability to predict values. R-Value for mathematical model obtained 0.9839. Also, R-Value for network training, validation and testing was 0.9899, 0.9910 and 0.9975, respectively. Also, overalls R-Value was 0.9899 that proved a very good validation. Â