The principal objectives of the work described in this thesis were: 1. To establish the effect of angle of inclination on hydrocyclone performance. 2. To develop an improved predictive model of hydrocyclones, incorporating angle of inclination and slurry viscosity. 80 cyclone tests were conducted in pilot plants both at the JKMRC (5. 1 and 10.2 cm cyclones) and in the Mount Isa Mines Copper Concentrator (50.8 cm cyclones) to study the effect of different cyclone inclined positions on hydrocyclone separation performance at different feed solids concentrations. The experimental results indicated that cyclone inclination significantly affects the classification performance of hydrocyclones, the bigger diameter cyclones being more affected than the smaller ones. Inclined cyclones can generally be operated with dilute feed solids concentrations to achieve the same cut-point as a vertically operated cyclone, but with an improved performance. For the 10.2 cm cyclone, a cyclone inclination greater than 60°, and with feed solids concentrations higher than 30 wt.%, had a significant effect on the cyclone performance. For the 50.8 cm cyclone, an inclination greater than 45° had a significant effect on its performance. To be able to achieve the maximum effect on cyclone performance in an inclined position the feed solids concentrations should be less than about 60 wt.%. Generally, a coarser cut-point, a reduced water recovery (fines recovery) to the underflow and a denser underflow solids concentration can be achieved with inclined cyclones. Some limited tests with different minerals confirmed that denser minerals cut finer than lighter ones. Using the experimental data and additional data from the JKMRC hydrocyclone database (117 data sets in all), a new empirical cyclone model has been developed to predict the performance characteristics of hydrocyclones. The new model incorporates the effects on cyclone performance of cyclone inclination, slurry viscosity and feed size (P40). These variables are absent in the existing published predictive cyclone models. The new model includes an equation for the sharpness of separation. The model comprises equations to predict the following cyclone performance characteristics: • the classification cut-point, d50c, • the water recovery to underflow, Rf (1 - C), • the sharpness of separation, a. and • the pressure, P, or throughput, Q. With the new model, it is possible to predict the relative changes in these performance characteristics caused by changes in the following variables: a) feed throughput, b) spigot diameter, c) vortex finder diameter, d) feed solids concentration e) feed pulp density, f) slurry viscosity, g) cyclone inclination and h) 40% passing feed size (P40). The influence of cyclone diameter, inlet diameter, cone angle and cyclinder length of a cyclone on the cyclone performance have also been quantified in the equations. Application-dependent system constants have been introduced in the equations to account for other variables which were either not represented or whose effects could not be fully qualified and quantified in the equations, in particular the feed characteristics. The predictive capability of the model was assessed by comparing its predictions with results from five independent datasets, and with the predictions of the models of Plitt and Nageswararao.