Flow Pattern and Frictional-Pressure-Loss Estimation Using Neural Networks for UBD Operations

Abstract

Abstract Underbalanced drilling is a very commonly used technique in many drilling applications. Usually, gasified fluids are preferred for achieving underbalanced conditions. Proper estimation of hydraulic requirements becomes a more important issue for trajectories with highly-inclined angles and longer sections. Flow of two-phase fluids through circular pipes has, relatively, been well understood. However, flow of two-phase fluids through annular sections is still a challenge. Many attempts have been conducted to model two-phase flow through annular geometries, but still there exists an uncertainty at this area. It has been observed that, for annular geometries, hydraulic diameter concept is not applicable for estimating flow patterns or frictional pressure losses using the models developed for circular pipes. This study approaches to estimate the flow pattern and frictional pressure losses of two-phase fluids flowing through horizontal annular geometries using Artificial Neural Networks (ANN) rather than proposing a mechanistic model. Experimental data collected from experiments conducted at Middle East Technical University, Petroleum & Natural Gas Engineering Multiphase Flow Loop, as well as from literature in order to train the ANN. Flow is characterized using superficial Reynolds numbers for both liquid and gas phase for simplicity. The results showed that ANN could estimate flow patterns with a high accuracy (error is less than ± 5%), and frictional pressure losses with an error less than ± 30%. It is also observed that proper selection of ANN is important for accurate estimations.