CFD modeling of hydrocyclones: Prediction of particle size segregation
Graphical abstract
.
Highlights
► We model hydrocyclones using multiphase CFD to simulate the particle size segregation. ► The CFD data compared with sampling probe data and found in good agreement near wall and cone section. ► The predictions are deviating with the probe data in the unstable forced vortex. ► The particle size segregation phenomena due to centrifugal and turbulent dispersive forces are explained.
Introduction
Hydrocyclones are widely used in the mining and chemical industries for the separation of solids or droplets based on their size and density. A typical hydrocyclone consists of a cylindrical section with a central upward flow discharge tube connected to a conical section with a downward flow discharge tube. An inlet conduit is attached tangentially to the top section of the cylinder. The fluid being injected tangentially into hydrocyclone causes swirling and thus generates centrifugal force within the device. This centrifugal force field brings about a rapid classification by size of the particulates suspended within the fluid.
The flow in a hydroyclone is a multiphase structure which consists of solid particles which are dispersed throughout the fluid, generally water. In addition, an air core is present. Such multiphase flows can be studied using a number of Computational Fluid Dynamics (CFD) techniques. These include the full Eulerian Multiphase approach, simplified Eulerian approaches such as the Mixture (Manninen et al., 1996) and Volume of Fluid (VOF) models (Hirt and Nichols, 1981) and the Lagrangian approach (Crowe et al., 1998).
Most of the previous numerical studies which have adopted the Lagrangian frame were not comprehensive. They only include the drag and centrifugal forces in the calculation of the particle trajectory, with or without particle dispersion effects (Hsieh, 1988, Hsieh and Rajamani, 1991, He et al., 1999, Rajamani and Millin, 1992, Boysan et al., 1982; Griffiths and Boysan, 1996). Also, these studies are limited to very dilute particle concentrations in cyclones. The Lagrangian approach has been extended to modeling cyclones at large particle concentrations by Rajamani and Millin (1992) and Devulapalli (1996). They couple the effect of solid concentration with fluid viscosity but were limited to Prandtl-mixing turbulence models. A similar model (Rajamani and Millin, 1992) has been re-investigated using LES turbulence model by Delgadillo and Rajamani (2005). They found that the prediction of particle classification follows closely the experimental values at low feed solids, whereas the predictions for the high feed solid concentrations overestimate the mean coarse size particle classification when compared to the experimental classification data in a larger cyclone.
There have been few studies where various cyclones have been modelled using a full Eulerian approach in conjunction to RANS turbulence models (Suasnabar, 2000, Nowakowski et al., 2004, Nowakowski and Dyakowski, 2003; Cokljat and Slack, 2003, Huang, 2005). All these studies have used the Fluent based full Eulerian models while simulating dilute solids flow in cyclones. The disadvantage of the full Eulerian multiphase modeling has been its high computational cost. Further implementations in commercial CFD codes have until recently been limited to using the k-epsilon/RSM models for turbulence.
Initially Pericleous and Rhodes (1986) and Davidson (1994) coupled the particle and fluid equations by modifying the mixture density and effective viscosity using an algebraic slip mixture model for 2D hydrocyclone simulations. Suasnabar (2000) and Brennan (2003) studies have adopted the mixture model (Manninen et al. (1996)) for dense medium cyclone simulations in which the mono average size of dispersive phase was considered as particulate phases and the particle classification mechanism was explained qualitatively.
There have been numerous studies where hydrocyclones have been modelled using modified algebraic slip mixture (ASM) model by the author Narasimha (2010). This CFD work has been validated by either Laser Doppler Anemometry (LDA), conducted on a water flows in clear Perspex models (Brennan, 2006; Delgadillo and Rajamani, 2005; Narasimha et al., 2006), or Gamma Ray Tomography (GRT) measurements of density profiles in a plastic cyclone (Narasimha et al., 2007). Whilst both LDA and GRT have generated useful data for validation, they are laboratory techniques which investigate the internal flow structure and have been primarily used on small cyclones. LDA and GRT do not provide any information on how the particles are distributed inside the cyclone by size. Recent numerical studies on various cyclone geometries also successfully used multi-fluid models (Rajamani et al., 2010, Hsu and Wu, 2010, Wang and Yu, 2010, Brennan et al., 2009, Davailles et al., 2012). ASM model was successfully used to model the multiphase flows in dense medium cyclones (Wang et al., 2009, Wang et al., 2011, Chu et al., 2009a, Chu et al., 2009b, Chu et al., 2012), in particularly simulating the magnetite segregation levels inside the dense medium cyclone.
In this paper the multiphase CFD model validation is attempted against the Renner’s (1976) data on particle size segregation inside a hydrocyclone. Renner’s data was measured using a high-speed sampling probe at different precisely controlled positions (Renner, 1976, Renner and Cohen, 1978). The work compares the particle size distributions predicted by CFD model to Renner’s data. This work also looks at the reliability of both the sampling probe and the multiphase CFD model for hydrocyclones.
Section snippets
Cyclone geometries and grid generation
The simulations use Fluent with 3-D body fitted grids which were generated in Gambit (a pre-processor for FLUENT). The grids typically have the size between 120,000 and 200,000 nodes. The feed port is a velocity inlet boundary condition and the overflow and underflow are pressure outlet boundary conditions. All other boundary conditions are no slip at the wall. Two geometries are selected for the CFD study, which were a Hsieh75 mm hydrocyclone, a Renner (1976) 150 mm cyclone. The key dimensions
CFD modeling
The momentum equations are discretized using the bounded central differencing method. PRESTO is used for pressure and QUICK is used for the dispersed phase transport equations. The SIMPLE method is used for the pressure–velocity coupling. The simulations are run with a fixed time step of 5.0 × 10−4 s. A velocity boundary condition is used on the feed and pressure outlet boundary conditions, with an air back flow volume fraction of 1.0, for both the overflow and underflow. Setting the air backflow
Two phase water/air velocity simulations
Case studies for the 75 mm cyclone with only water and air, were run with a water feed rate of 1.116 kg s−1, which is equal to the feed flow rate used in Hsieh (1988) series I LDV measurements. A comparison between the mean velocities predicted from the CFD and Hsieh’s measurements was reported by authors elsewhere (Brennan et al., 2007, Narasimha et al., 2007, Brennan, 2006). The mean tangential and axial velocity components predicted by the LES model agree with the measured velocities quite
Conclusions
The CFD modeling of hydrocyclones attempts to understand the particle size segregation. In two-phase simulations the mean and turbulent stresses predicted by the LES are in good agreement with the data measured by the LDA, whilst the DRSM model over-predicts the stresses and is consistent with the DRSM model under-predicting the tangential velocities. Simulated multiphase data analysis in a 150 mm cyclone design (Renner, 1976) indicates that the overall classification curve follows the
Acknowledgments
The authors would like to express their sincere thanks to Prof. Emmy Manlapig, Manager, AMIRA P9N, JKMRC, University of Queensland, Australia, and AMIRA-P9N research sponsors and management, for their keen interest, encouragement and funding for undertaking these studies. Specially acknowledged Dr. Aubrey Mainza, CMR group, University of Cape Town, for his constant encouragement and support in pursuing this piece of work.
References (42)
CFD simulations of hydrocyclones with an air core: comparison between large eddy simulations and a second moment closure
Chem. Eng. Res. Des.
(2006)- et al.
Multiphase modelling of hydrocyclones – prediction of cut-size
Miner. Eng.
(2007) - et al.
CFD–DEM modeling of multiphase flow in dense medium cyclones
Powder Technol.
(2009) - et al.
CFD–DEM study of the effect of particle density distribution on the multiphase flow and performance of dense medium cyclone
Miner. Eng.
(2009) - et al.
Computational study of multiphase flow in dense medium cyclones: effect of particle density
Chem. Eng. Sci.
(2012) Analysis of swirling flow in hydrocyclones operating under dense regime
Miner. Eng.
(2012)- et al.
A comparative study of three turbulence-closure models for the hydrocyclone problem
Int. J. Miner. Process.
(2005) - et al.
Computational fluid dynamics and empirical modelling of a number of cyclone samplers
J Aerosol Sci.
(1996) - et al.
A numerical simulation of hydrocyclones
Chem. Eng. Res. Des.
(1999) - et al.
Volume of fluid (VOF) method for the dynamics of free boundaries
J. Comput. Phys.
(1981)