The turbo air classifier is one of the most widely used equipment in powder classification. The complex flow behaviour inside it, however, prevents material experiments from providing information about its internal separation mechanisms. A study of the interaction of structural variables is therefore undertaken examining air flow behaviour, specifically the air flow between the blades of the rotor cage. The investigation of these flow field characteristics made use of the computational fluid dynamics (CFD) to simulate the air flow in the classifier. It was found that the inlet velocity of the turbo air classifier and the rotary speed of the rotor cage are two of the dominating, non-structural factors that affect velocity distributions in the region between the rotor cage blades. Once the inlet velocity settles, a critical rotary speed must be present to smoothen the flow field between the blades, resulting in an excellent classification performance. Three-dimensional velocity measurements of the region between the blades by laser Doppler velocimeter (LDV) were performed to test the results of the flow field simulation. This revealed that when inlet velocity is invariable, the velocity distributions in the region between the blades are at its most symmetric with the critical rotary speed of the rotor cage making it more favourable for classification. The velocity measurement results are likewise in good agreement with the results of the flow field simulation. Newly structured rotor cages are also simulated and compared with a conventional turbo air classifier, air flow in the newly structured model is smoother. The distributions of radial and tangential velocities are more symmetric and the trend of the rotating vortex between the blades attenuates, particularly when the rotary speed is high. The newly structured rotor cages can therefore achieve higher classification performances.