High-Viscosity Liquid/Gas Flow Pattern Transitions in Upward Vertical Pipe Flow

Abstract

Summary High-viscosity liquid two-phase upward vertical flow in wells and risers presents a new challenge for predicting pressure gradient and liquid holdup due to the poor understanding and prediction of flow pattern. The objective of this study is to investigate the effect of liquid viscosity on two-phase flow pattern in vertical pipe flow. Further objective is to develop new/improve existing mechanistic flow-pattern transition models for high-viscosity liquid two-phase-flow vertical pipes. High-viscosity liquid flow pattern two-phase flow data were collected from open literature, against which existing flow-pattern transition models were evaluated to identify discrepancies and potential improvements. The evaluation revealed that existing flow transition models do not capture the effect of liquid viscosity, resulting in poor prediction. Therefore, two bubble flow (BL)/dispersed bubble flow (DB) pattern transitions are proposed in this study for two different ranges of liquid viscosity. The first proposed transition model modifies Brodkey's critical bubble diameter (Brodkey 1967) by including liquid viscosity, which is applicable for liquid viscosity up to 100 mPa·s. The second model, which is applicable for liquid viscosities above 100 mPa·s, proposes a new critical bubble diameter on the basis of Galileo's dimensionless number. Furthermore, the existing bubbly/intermittent flow (INT) transition model on the basis of a critical gas void fraction of 0.25 (Taitel et al. 1980) is modified to account for liquid viscosity. For the INT/annular flow (AN) transition, the Wallis transition model (Wallis 1969) was evaluated and found to be able to predict the high-viscosity liquid flow pattern data more accurately than the existing models. A validation study of the proposed transition models against the entire high-viscosity liquid experimental data set revealed a significant improvement with an average error of 22.6%. Specifically, the model over-performed existing models in BL/INT and INT/AN pattern transitions.