Prediction of the Surface Tension of Hydrocarbons

Abstract

Prediction of the surface tension of hydrocarbons is required in many chemical engineering calculations. In this work a robust artificial neural network code has been used in MATLAB software (The MathWorks, Natick, MA) to predict the surface tension prediction of 61 hydrocarbons. Experimental data is divided into two parts (70% for training and 30% for testing). Optimal configuration of network is obtained with minimization of prediction error on testing data. The accuracy of our proposed model is compared to four well-known empirical equations. Results showed that artificial neural network was more accurate than these empirical equations. Average relative deviation of our artificial neural network model is 0.93 while average relative deviation of the Brock-Bird, Pitzer, Zuo-Stenby, and Sastri-Rao equations are 6.30, 6.48, 5.73, and 6.33, respectively.

Keywords:

• surface tension
• hydrocarbons
• artificial neural network
• empirical equations
• prediction