Prediction of Oxygen Solubility in Organic Solvents Using Artificial Neural Networks

In this paper, solubility of oxygen in organic solvents has been estimated using Artificial Neural Networks (ANN). Solubility data were studied for wide ranges of temperature (298.2-348.29 K) and pressure (0.0535 to 9.2338 MPa). Solvents are included of methanol, n-propanol, octane, toluene, dibutyl ether and 2-methyltetrahydrofuran. Network model consists of four inputs in input layer for acentric factor, molecular weight, TR and PR of the system and one neuron in output layer corresponding to solubility of oxygen. The best structure for feed-forward back propagation neural network is logarithmic sigmoid transfer function for hidden layer, 13 neurons in this layer and linear transfer function for output layer. Results show that optimum neural network architecture is able to predict the solubility of oxygen in organic solvents with an acceptable level of accuracy, R2 of 0.999997, ARD % of 0.8103 and AAD% of 0.0042. Sensitivity analysis shows that TR has the greatest effect on the solubility of oxygen.