Estimating Flow Patterns and Frictional Pressure Losses of Two-Phase Fluids in Horizontal Wellbores Using Artificial Neural Networks

Abstract

Underbalanced drilling achieved by gasified fluids is a very commonly used technique in many petroleum-engineering applications. This study estimates the flow patterns and frictional pressure losses of two-phase fluids flowing through horizontal annular geometries using artificial neural networks rather than using conventional mechanistic models. Experimental data is collected from experiments conducted at METU-PETE Flow Loop as well as data from literature in order to train the artificial neural networks. Flow is characterized using superficial Reynolds numbers for both liquid and gas phase for simplicity. The results showed that artificial neural networks could estimate flow patterns with an accuracy of 5%, and frictional pressure losses with an error less than 30%. It is also observed that proper selection of artificial neural networks is important for accurate estimations.