Experimental analysis of liquid-liquid flow in a horizontal tube using artificial neural networks

The objective of this work was the application of an artificial neural network in prediction of holdup of two-phase flow (oil-water) in a pipe in horizontal position. To this end, the velocity superficial of water and oil were used as input parameters, meanwhile, the holdups of these two fluids were used as output parameters for the training and testing of the multilayer neural network, the method used was back-propagation. The experimental data (92 data) were taken at LEMI-EESC-USP and were used to develop the artificial neural network model. Finally, it was concluded that the experimental data used in the neural network agreed with the tagsig transfer function with 10 neurons in the hidden layer evaluated from the absolute percentage error of (AAPE= 3,95) and coefficient of determination (R-2= 0,975).