An Experimental and Numerical Investigation of Multiphase Flow Splitting at an Impacting T-Junction Between a Single Flowline And Two Risers

Abstract

In 2011, Royal Dutch Shell plc (Shell) took the final investment decision to develop the Prelude gas field using the world’s first Floating Liquefied Natural Gas (FLNG) facility. FLNG poses some new technical challenges. For example, as the vessel is moored to the seabed, it is subject to movements induced by waves, winds and tides. Therefore, flexible instead of rigid risers are used to bring the gas from the seabed to the surface. However, these risers are limited in diameter. For Prelude each flowline is connected to a single riser. In new future projects it might be beneficial to have a single (larger diameter) subsea flowline at the sea floor which is split into two or more flexible risers. This flow split is complicated as gas wells produce a multiphase mixture of gas, hydrocarbon condensates and/or water. There thus is a need to understand how the different phases will distribute over the multiple risers, and how this will influence the whole operation. In previous lab experiments [1], a horizontal symmetric impacting T-junction was used to divide the multiphase mixture from the horizontal flowline into two vertical risers. It was found that at low gas flow rates, for which the flow regime in the risers is hydrodynamic slug flow, non-symmetric distributions in the phase split are the rule rather than the exception. These maldistributions exhibit transient behaviour, and exhibit hysteresis. However, in gas production from dry fields with low Liquid to Gas Ratios (LGR), the gas flow rates generally are higher and the flow regime in the vertical risers is annular flow. In this flow pattern, liquids flow in the upward direction and no liquid buildup occurs at the riser base. As annular flow is more stable in time, it is expected that for these conditions the phase splits are evenly distributed over the two risers and no transient behaviour is present. The main aim of the present study was to verify the hypothesis that if the gas flow rate is sufficiently high to sustain annular flow in both risers the phases will distribute evenly. Thereto experiments were performed in the Severe Slugging Loop (SSL) at the Shell Technology Centre Amsterdam. The loop consists of a 100 [m] horizontal flowline connected via an impacting T-junction to two 1.25 [in] diameter vertical risers of 16.8 [m] height. The facility is operated with an air-water mixture at atmospheric pressure and ambient temperature. As the SSL is not equipped with flow meters to measure the phase flow in the risers, a new method is developed to identify maldistribution in the phase split. In two phase flow in a vertical riser, the flow rates in the pipe are not uniquely determined by the pressure drop. However, in the current dual riser experiments the total flow rate in the setup is known. It is found that, with the use of single riser benchmark data and visual inspection, one can distinguish between an equal and an unequal phase split. Experiments are performed to test the stability of the phase split, and the effect of hysteresis for various gas and liquid flow rate combinations. It is observed that five different states exist for the flow split in the single flowline dual riser geometry. Out of these, only two are stable, which occur on opposite sides of a certain gas flow rate transition point. This point is identified as the minimum in the pressure gradient curve for a single vertical riser. This minimum can be linked to the transition from churn flow to annular flow. In addition to the experiments, a computational model is proposed to predict the phase split. In this model, the Shell Flow Correlations (SFC) for multiphase pipe flow are used to propagate the boundary conditions of the system from the riser tops to the corresponding outlets of the splitter at the base of the risers. Next, all possible phase splits which satisfy the extended Bernoulli equation and the conservation of momentum in the splitter are identified. The model conserves mass implicitly. The extended Bernoulli equation is implemented as the Advanced Double Stream Model (ADSM) [2], the control volume approach is used for the conservation of momentum. All three sub-models assume 1D steady state fully developed flow in the different parts of the geometry. By imposing different back pressures on the risers, for a constant gas and liquid flow, the model can generate phase split curves. These are compared to experimental data available in the literature and experimental results by Van de Gronden. It is found that the write-up of the derivations in the original ADSM paper [2] contains typos or errors. Therefore we have derived new relations, but they do not exhibit the same behaviour as the modelling results presented in the paper. From the comparison with the experimental results, it is concluded that the proposed computational model, which assumes 1D steady state fully developed flow, predicts a much too symmetric phase split. The modelling effort has not yet resulted in a reliable for the phase split prediction model.