Turbulence Modelling of Two Phase Stratified Channel and Pipe Flows

Abstract

Stratified two phase flow is one of the flow regimes that is of importance in multiphase flow transport through pipelines, such as used for example in the oil and gas industry. Its application also extends to chemical production, energy conversion and food processing. The phenomenon of turbulence further complicates the stratified flow behaviour. Having a simulation tool that accurately predicts the pressure gradient and liquid level in a turbulent channel or pipe flow can lead to better designs of multiphase flow systems. The common RANS turbulence models (such as 𝑘−𝜔 and 𝑘−𝜀) artificially produce too much turbulence at the liquid-gas interface. Therefore, these models need to be modified such that turbulent viscosity is sufficiently damped at the interface. To ensure this, in the present study the specific dissipation rate (𝜔) is imposed at the interface. Here 𝜔 is an appropriate function of the surface roughness factor (𝑘𝑠), which represents the effect of interface waves. The present work is a follow up to a previous research project where a MATLAB tool was developed for the prediction of stratified flow in channels. The first and main objective of the present thesis is to find and test a model for 𝑘𝑠 and to apply that to obtain a modified version of the Standard 𝑘−𝜔 (SKW) turbulence model and the Shear Stress Transport (SST) model. MATLAB can be used to find solutions for the channel flow. The simulation results are compared with experimental data. The second objective is to compare the results obtained using the turbulence models in the MATLAB model with predictions using similar models in Fluent. The third objective is to extend this study to a 3D setup of a two phase pipe flow where only the liquid phase is simulated. This is a so-called Segregated Liquid Phase (SLP) simulation. RANS predictions in Fluent are compared with experiments and DNS data. The main conclusions are: . The calculation of 𝑘𝑠 has been automated. . The predictions of the flow rates for the experimental case by Fabre et al. are better than those for the experimental case by Akai et al. . MATLAB predictions of the flow rates for the experimental cases are better than those of Fluent. This is because it is difficult to correctly impose an interface condition in Fluent, whereas this is straightforward in MATLAB.