Experimental and computational study of two-phase slug flow

This thesis presents a computational and experimental investigation of horizontal gas-liquid two-phase slug flow. The overall aim of the present project was to carry out analytical and computational studies to model the processes of slug initiation, growth and collapse in horizontal pipes. The literature of two-phase flow, with a focus on slug flow, is reviewed. A “Benchmark exercise” of predicting the horizontal two-phase slug flow on the WASP (Water, Air, Sand and Petroleum) facility was set up for CFD simulations with the aim of investigating the capability of CFD codes in the prediction of slug flow characteristics in large scale pipe (78 mm ID, 37 m long). Six CFD codes (TRIOMPH, LedaFlow, STAR-CD, TransAT, FLUENT and CFX) were participated in this exercise. The complex nature of the slug flow mechanism and the relatively large-scale of the associated modelling domain contribute to the challenges of this CFD exercise; this is particularly challenging for three-dimensional simulations. It has proven difficult to carry out direct comparisons between the performances of the various codes; this is due to the difference in the assumptions and approximations made in each case. Successive slugging was captured by most of the CFD codes, apart from CFX in which the flow was remaining stratified. In terms of flow characteristics, comparison against the measurements was mainly focused on the distribution of slug frequency at various locations. Satisfactory agreement was obtained by the 1D code TRIOMPH, whereas the rest of the codes were failed to reproduce the observed trend of slug frequency distribution. The ability of the TRIOMPH code in the prediction of horizontal slug flow is limited in a restrain region, where the two-fluid equations employed in the code has to be well-posed, therefore a validation case study was examined. For a well-posed system, a unique solution can be obtained, whereas for an ill-posed system, the solution would become mesh dependent. However, the lack of unique solution of the ill-posed case can be altered by applying unsteady inlet condition prescribed in terms of a train of slugs, which can be generated by an alternative code namely slug-tracking code. CFD prediction of horizontal slug flow in a channel with rectangular cross-section (100 mm x 30 mm x 8000 mm) was systematically studied using STAR-CCM+. The experimental data was provided by Forschungszentrum Rossendorf (FZD) from Germany. Mesh sensitivity and parametric study were conducted to determine the settings that optimise accuracy and stability. The polyhedral mesh was found to give a fast and better convergence of the numerical solutions. A sufficient small grid size is important in order to resolve the velocity gradient throughout the two fluids. The onset of interfacial instability, wave growth, and slug generation processes were captured in the simulation. However, the predicted slug initiation site was shifted further downstream than the actual site observed in the experiment, leading to the discrepancy in the prediction of the flow characteristics. Experimental studies on slug initiation and their subsequent development were carried out on the LOWPRSS (Low Pressure) facility at Imperial College London. The motion of the slug is followed using high-speed imaging and conductivity probes for air-oil and air-water systems, respectively. The slug frequency results demonstrate that rapid initiation events occur near the inlet region, and the slug frequency exhibits a maximum at intermediate axial distances before eventually reaching an approximately constant value at sufficiently large distances downstream of the inlet. Particular attention was given to the slug initiation mechanism observed using the LOWPRESS facility. Together with a set of air-water slug flow data obtained on the WASP facility, the influence of inlet geometry was examined. It was shown that the inlet geometry significantly influences slug initiation; however, the effect of inlet geometry on the slug development reduces along the pipe length. Beyond a certain distance, slug frequency approaches a constant value invariant with inlet geometry.Open Acces