Centrifugal membrane and density separation ( CMDS) is a novel technology pro­posed for treatment of industrial process streams. This cross flow filtration process combines the energy recovery inherent to centrifugal reverse osmosis (CRO) with the potential alleviation of membrane fouling and concentration polarization due to the favourable effects of centrifugal and Coriolis accelerations.

This thesis presents a computational study of CMDS undertaken to understand the basic hydrodynamics of the process and to provide a tool for design optimization. Two distinct models were developed, the porous wall model and the source term model, and incorporated into a commercial CFD code, TASCflow3d which solves the full Navier-Stokes equations coupled to scalar transport equations which account for dissolved species. These models are used to simulate two and three dimensional laminar flows in both static and rotating reverse osmosis membrane cartridges and to predict permeate fluxes.

In the case of static reverse osmosis, the porous wall model is shown to correctly predict the flux decline associated with concentration polarization. Flux predictions are found to be very sensitive to hydraulic permeability, feed concentration and ap­plied pressure. The need for an ad hoc gel layer model proposed by some researchers is not supported by this work.

In the case of centrifugal reverse osmosis, significantly enhanced permeate fluxes over the static cases are predicted. These increases are clue to the mixing associ­ated with secondary motion developed in rotating channel flows clue to Coriolis and centrifugal accelerations. The effectiveness of the secondary motion in improving the reverse osmosis process is shown to depend on the orientation of the membrane cartridge with respect to the axis of rotation.

The current model provides valuable insight into the hydrodynamics of CMDS and concentration polarization limitations . The model is however limited to dissolved second phases, and hence cannot properly model gel layer formation or membrane fouling. It is also limited in the rotation rates that can be simulated clue to problems with the outlet boundary condition, the coupling of the scalar equations and the potential for unsteady flow. Accordingly, it is recommended that a time-dependant model capable of representing discrete phases be developed in future work.