Hydrocylone for oil-water separations with high oil content: Comparison between CFD simulations and experimental data

Abstract

Production separators for gas-oil-water separation are huge, require long residence times, and present several internals to improve oil-water separation. As an alternative to production separators, specially for subsea facilities, this work presents a hydrocyclone which separates a 40% oil-in-water mixture with 93% ± 2% total efficiency and flow ratio of 7%. Such high oil concentration is typical of mature wells and hydrocyclones have not been designed for such purpose before. In order to design this hydrocyclone, fractional factorial techniques have been used to consider the influence of seven geometrical variables. CFD has been used to assess the performance of each hydrocyclone of the factorial design. The selected hydrocyclone has been tested experimentally in order to validate its performance. PIV measurements have also been carried out in order to assess tangential and axial mean velocity profiles for only water flowing in the hydrocyclone. FBRM has been used to measure droplet size distribution at inlet, overflow and underflow streams. As results, computational reduced total efficiency has been experimentally confirmed. Numerical simulations captured tangential velocity near the centerline of the hydrocyclone, but failed in predicting axial velocity profiles and tangential velocity peaks and near wall regions. It has been discussed though that these are probably not responsible for oil droplets separation in the hydrocyclone when water is the continuous phase. A methodology to analyze the effect of droplet breakup and coalescence, using measured inlet, overflow and underflow cumulative size distribution, has been presented. It has been shown that for the inlet cumulative undersize distribution used, breakup was negligible and coalescence occurred only for droplet diameters higher than those separated with 100% grade efficiency. In this case, no breakup and coalescence models are thus needed in the mathematical model and this explains why reduced total efficiency has matched in both numerical simulations and experiments.