Modeling surface tension of pure refrigerants using feed-forward back-propagation neural networks

In this study, a model was proposed to predict the surface tension on the basis of feed-forward back-propagation network by employing different training algorithms including Levenberg-Marquardt, Scaled Conjugate Gradient and Pola-Ribiere Conjugate Gradient. A total of 793 experimental data points from 24 different pure refrigerants were gathered from reliable literature to train, test and validate the proposed network. Temperature, critical pressure, critical temperature, and acentric factor were chosen as input variables of the developed network. The network with 1 hidden layer and 19 neurons with tan-sigmoid and purelin transfer functions in the hidden and output layers was determined to have the optimum performance. The results revealed that the proposed network has the ability to correlate and predict the surface tension accurately with an overall Mean Relative Error (MRE) value of 0.0074 and correlation coefficient (R2) of 0.9996. The obtained results were compared to different well-known correlations in the literature which demonstrated a better performance of the proposed network.