The present article describes a numerical study of multiphase, strongly swirling How in a cylindrical separator with double vortex generators to predict the separation efficiency of mixture consisting of oil as the primary phase and sand with several sizes as the secondary phases. Solids are classified by the separation process and the feed is split into coarse and fine fractions (with the fine fraction usually remaining in suspension). The mixture-granular multiphase and renormalization group (RNG) of k - epsilon turbulence models are implemented in this study. The predictions are compared against the experimental data of mean tangential velocity and mean radial pressure profiles. The overall agreement between the measurements and the predictions obtained with RNG k - epsilon model is reasonably good. The present models based on computational fluid dynamics (CFD) techniques have ability to capture a narrow localized residence zone for solid particles at location near the mid-cylinder where the two vortices merge and the suspension process can take place. Furthermore, they have proven to be useful in predicting the internal flow structure of the continuous phase thus the separation of the particulate phases. Most of the larger particles to be separated are forced to remain near the periphery of the separator and concentrated at the mid-separator as a result of strong centrifugal force. On the other hand, the finer particles with less inertia are not affected too much and are subsequently dispersed over the separator. The separation process does not end by the arrival of the particles onto a collecting zone. If the process is to be continuous, the collected particles have to be transported and discharged from the separator. (C) 2009 Elsevier Ltd. All rights reserved.