Prediction of Flow Pattern in Horizontal Liquid-Liquid Two Phase Flow Using Artificial Neural Network

Flow pattern is one of the main parameters of two-phase liquid-liquid flows. Nevertheless, there is no anaccurate and comprehensive model to predict it. In this paper, artificial neural networks (ANNs) were used for prediction of flow pattern in horizontal liquid-liquid flows. The applied neural networks for this investigation were feed-forward back propagation (FFBP) and probabilistic neural network (PNN).1912 data points from 13 different flow pattern maps reported in literature were collected. Superficial velocity, viscosity ratio and density ratio of oil and water and interfacial tension between them, as well as inner diameter and roughness of pipes were chosen as input variables of both networks and 9 flow patterns were selected as their output variables. The results obtained from optimal structure of networks on their testing data set revealed that the PNN has better performance(withaccuracy of 96.34%) compared to FFBP (with accuracy of73.73%) and can be used as a comprehensive model to predict horizontal liquid-liquid two-phase flow patterns.