Prediction of Surface Tension of Pure Hydrocarbons by An Artificial Neural Network System

In this study, a multilayer perceptron neural network with a back-propagation (BP) learning method containing three layers has been developed for prediction of surface tension of pure hydrocarbons. This model was developed using 211 experimental data on surface tension in the literature. Statistical criteria of R2 and root mean square error (RMSE) for this network were 0.9997 and 0.0615, respectively. The accuracy and validity of this model were compared with the experimental data and results obtained from the Pazuki-Nikookar (PN) equation of state (EOS; Pazuki et al., 2007; Nikookar et al., 2008 Nikookar, M., Omidkah, M. R. and Pazuki, G. R. 2008. Prediction of density and solubility parameter of heavy oils and SARA fractions using cubic equations of state. Petrol. Sci. Tech., 26: 1904–1912. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) as well as some semi-experimental correlations such as Pitzer (1995) Pitzer, K. S. 1995. Thermodynamics, , 3rd ed. New York: McGraw-Hill.. [Google Scholar], Sastri and Rao (1999) Sastri, S. R. S. and Rao, K. K. 1999. A new temperature–thermal conductivity relationship for predicting saturated liquid thermal conductivity. Chem. Eng. J., 74: 161–169. [Crossref], [Web of Science ®] , [Google Scholar], and Brock and Bird (1955) Brock, J. R. and Bird, R. B. 1955. Surface tension and the principle of corresponding states. AIChE J., 1: 174–177. [Crossref], [Web of Science ®] , [Google Scholar]. Also, considering the results obtained by the artificial neural network system, a correlation is presented for estimation of the parachor constant and the results are compared with those obtained from equations of state.