Within the research project, multi-scaled approaches on the microscopic (particle movement) and macroscopic scale (separation process) are necessary. To achieve the challenging numerical simulations, solutions of both parts have to be combined.

On the macroscopic scale, comprehensive separation experiments will be drawn into focus. In addition to classifying (separation by particle size) of sand-gravel mixtures, sorting (separation by solid density) of soft-air-particles was performed. Finally, separation experiments with more complex mixtures (e.g. specifically formed plastic granules) will be executed with particle size, density and shape as separation attributes. Using different mass fluxes of the feed at fixed air velocities, the effect of mass loading of the process air on the efficiency of the separation operation will be investigated. In this case, the determination of the separation attribute and as well as several energetic values (such as pressure drop, energy consumption, etc.) is of utmost importance. Furthermore, the variation of the channel geometry, the number of stages and the determination of the feed stage will help to find an optimal design.

On the microscopic scale, the velocity field and the particle movement inside the turbulent flow will be investigated. In addition to measurements like the performed Laser-Doppler-Velocimetry (LDV) and Particle Image Velocimetry (PIV) to visualize the flow field, particle trajectories will be captured by Particle Tracking Velocimetry (PTV). For simultaneous acquisition of different particle collectives color coding will be applied.

The preconditions for the measurements were met by extensive alterations of the test rig. A new fan was installed for the variation of the volumetric flow rate of air. For a continuous controllable solid mass flow the screw conveyer was replaced by a vibrating conveyer. To enable optical access over the complete channel height from three sides a new modular zigzag channel was designed and built (see Figure 2). Through the modular design the number of stages can be varied with little effort.

Using the determined fluid properties single-phase CFD simulations where executed and validated using the flow field measurement data to find an optimal turbulence model. In this case, three dimensional models using Reynolds-averaged Navier-Stokes-equations (RANS) and unsteady Reynolds-averaged Navier-Stokes-equations (URANS) where used. On this basis, multi-phase flows will be numerically simulated with respect to the occurring mechanisms (e.g. collisions, interactions); an increased precision will be achieved by using one-, two- and four-way coupling (Figure 3).