CFD of multiphase pipe flow

Abstract

Pipelines with multiphase flow of gas, oil, and water are commonly used in the oil and gas industry. In the presence of offshore platforms or vessels, the pipeline ends with a vertical riser. A possible new concept is to use a single large diameter pipeline along the sea floor that ends into multiple smaller diameter risers. This concept might be of interest for future Floating Liquefied Natural Gas (FLNG) vessels. This report focusses on detailed numerical simulations for a small scale representation of this industrial flow splitting configuration: a two-phase flow of air and water though a 0.5 [m] long, 0.05 [m] diameter, horizontal pipe with a T-junction to a dual 2.5 [m] high, 0.05 [m] diameter riser system. A comparison has been made between two multiphase flow solvers available in the well-known open-source code OpenFOAM. The interFoam solver utilises a mixture model formulation, while the multiphase- EulerFoam solver uses an Euler-Euler (two continua) formulation for both fluids. Both solvers use Volume of Fluid as interface sharpening method to solve the equations for incompressible flow. Air and water are flowing into the domain with a total volumetric flow rate of 64.2 [m^3/h]. The volumetric flow rate of water in the simulations is taken as 1, 2 and 3 [m^3/h]. Furthermore, a pressure difference between the outlets of the dual risers is applied. This is meant to analyse the difference between solutions of both solvers with respect to maldistribution between the two risers, pressure loss and liquid hold-up. The Pressure Implicit with Splitting of Operator (PISO) algorithm with two corrector loops is used combined with a low Courant-Friedrichs-Lewy condition of 0.25 to obtain sufficient numerical stability. Second order discretisation schemes in space and a first order scheme in time are used. In order to speed up the calculations Geometric Algebraic Multigrid is used to solve the pressure field. Meshing of the domains is done with snappyHexMesh. Three T-junction meshes are generated with increasing fineness of 23042, 47544 and 95292 cells. Two riser meshes are made with 67720 and 276784 cells. The Large Eddy Simulation turbulence model is used with the Smagorinksy Sub-Grid Scale model. The fixed flow rates at the inlet are coupled with a variable pressure. At the outlets the outflow velocity is variable with a fixed pressure. The turbulent viscosity at the wall is governed by wall functions. The calculations are done on the hpc12 cluster at Delft University of Technology. The calculation times are in the order of one to three weeks for a typical simulation. Overall, differences between the production at the outlets of interFoam and multiphaseEulerFoam are small. Solutions from both solvers indicate that the influence of a pressure difference over the outlets has more influence on the non-symmetric production of air than of water. The results from multiphaseEulerFoam look promising and due to its Euler-Euler momentum description it shows realistic flow behaviour in the junction. Improvements in simulating the system can be made by progressing the simulations longer in time, by increasing the entrance length of the horizontal pipe towards the junction and by choosing another Sub-Grid Scale model.