Towards a New Understanding of the Cyclone Separator

"The cyclone would have a claim to be the commonest, least understood and most abused piece of mineral processing equipment available for study by researchers. It has certainly generated its fair share of publications – 1,000 in the last 60 years would be a fair estimate. However this remarkable degree of interest and scale of investigation is not reflected in our understanding of how they work; cyclones are still installed inappropriately and run poorly, which suggests that there are still things to learn about this deceptively simple process.In mineral processing there are two broad classes of cyclone: the hydrocyclone, used in classification, thickening and dewatering, and the dense medium cyclone, used for concentration by density. Although rarely studied together, the laws of physics presumably apply equally to both and so it seems that there would be interesting things to learn by considering them as a single class of device. The good news is that in recent years there have been spectacular advances in the computational and measurement tools available to study these devices, including computational fluid dynamics (CFD), tomography, and laser doppler anemometry (LDA), and some good progress has been made in an understanding of the fundamental principles involved.The JKMRC has made many experimental studies of hydrocyclones and DM cyclones, leading to useful empirical models. However some recent CFD and tomographic work, in association with work of other institutions, has thrown important light on the workings of these processes. Progress has also been made in novel designs of cyclone to achieve improved separating efficiency, based on an attack on the subprocesses believed to contribute to inefficient operation. This paper describes the work to date.Modelling of the dense medium and classifying cyclone was conducted using the Fluent computational fluid dynamics code. Three phase simulations of the dense medium application (air/water/medium) showed appropriate medium segregation effects but over-predicted the level of segregation compared to segregation measured by gamma-ray tomography. This is believed to be because the basic Algebraic Slip mixture model in Fluent does not model any back-mixing of the dispersed phase due to turbulence."