Abstract

Improved one-dimensional (1D) models—compared to previous work by the authors—are proposed which are able to predict the velocity, length, and position of the liquid column in the rapid emptying and filling of a pipeline. The models include driving pressure and gravity, skin friction and local drag, and holdup at the tail and gas intrusion at the front of the liquid column. Analytical and numerical results are validated against each other, and against experimental data from a large-scale laboratory setup.

Abstract

In this work, the application of an artificial neural network (ANN) is proposed to develop a predicting model for the holdup of a two-phase flow composed of water and mineral oil in a horizontal pipe. For this, the surface velocities of each fluid and the differential pressure in the pipeline are used as input parameters of the multilayer artificial neural network with backpropagation, while the holdup of the fluids is used as the output parameter for the training. A set of 56 experimental data was obtained in the LabPetroCEPETRO-UNICAMP laboratory. The best performing results for the predictive model show a mean absolute error (AAPE) of 3.01% and a coefficient of determination 2 of 0.9964 using 15 neurons in the hidden layer of the network and the TanSig transfer function.