In this work, a trainable feed-forward back-propagation network was developed by employing the Levenberg-Marquardt training algorithm to predict the surface tension of binary refrigerant mixtures. 1260 experimental data were collected from reliable literature to train and test the network. Temperature, critical pressure, critical temperature, critical volume and acentric factor of the binary mixtures were selected as input variables of the proposed network. The optimum number of hidden layers was determined to be 1, with 19 neurons in the hidden layer. Tan sigmoid and purelin functions was chosen as the transfer functions in the hidden and output layers, respectively. The results revealed that the ANN has the capability to correlate and estimate the surface tension accurately with an overall %AARD and correlation coefficient values of 0.7582 and 0.9997, respectively. In addition, the results were compared to different well-known correlations and models which indicated a better performance of the developed ANN. (C) 2017 Elsevier B.V. All rights reserved.