Krebs Engineers has investigated the use of computational fluid dynamics (CFD) as a development tool to eliminate the need to fabricate and test each new design concept. This tool would allow the evaluation of an increased scope of hydrocyclone geometries. Krebs Engineers currently follows a methodology that incorporates both CFD analysis as well as traditional fabrication testing methods to evaluate hydrocyclone designs. The model used by Krebs engineers has been validated to the degree that we are confident it can be used to evaluate the expected relative difference in hydrocyclone performance between different hydrocyclone geometries. A higher order Reynolds stress turbulence model provided the best agreement with velocity profiles. The air core has been successfully modeled using this tool and as a result the throttling effect of this on the underflow and overflow orifices has been defined. The flow-splits from an actual hydrocyclone test are very close to predicted flows from the model at two different vortex finder sizes. A CFD tool also provides the ability to track the path of various size particles along the internal hydrocyclone flow field. The particle recovery curve from the CFD model shows good agreement with the actual hydrocyclone performance measured in a controlled test. The CFD model predictions to different vortex finder sizes and the addition of other hydrocyclone geometry changes have also been in the correct direction and magnitude based on validation test results. The CFD model has provided a means to evaluate the effect of design changes on component wear. The tool has the demonstrated ability to keep track of the number, angle and magnitude of particle collisions on the hydrocyclone internal surface. The CFD model does have some notable limitations. The method we have chosen does not consider the effects of the particles on the fluid flow or other particle interaction. Since it is common for the slurry fed to a hydrocyclone to have over 20% solids by volume and a more concentrated 50% solids underflow, this is an important limitation. (C) 2004 Elsevier Ltd. All rights reserved.