|
The tendency of inertial particles to clusterize is one of the most remarkable properties of turbulence-particles interactions. Possible clustering mechanisms rely on preferential concentration effects, where particles preferentially sample specific structures of the carrier turbulent field (for instance, heavy particles are centrifugated out from turbulent eddies, while light particles move toward the center of the eddies). The main parameter controlling the clustering efficiency is the particle Stokes number St (ratio of particle viscous relaxation time to the turbulent dissipation time). In the present work, we propose an experimental study, where a new approach for the reconstruction of the concentration field of water droplets dispersed in active-grid generated turbulence, allows to investigate clustering properties at scales much larger than previously available data. This method shows that particles not only form clusters, but that clusters themselves tend to assemble in super-clusters. Our experiment runs in a low speed windtunnel where turbulence is generated downstream an active-grid. An array of injectors seeds the flow with small water droplets (50μm in diameter typically). When varying the mean velocity of the wind, we change the Reynolds number of the flow (between 230 and 400, based on Taylor micro-scale) as well as its dissipation time-scale, what results in a variation of particles Stokes number between 2 and 10. Particles are visualized using a high-speed camera in a laser sheet parallel to the mean flow. In a previous work we have shown that a Voronoi tesselation analysis of the recorded images reveals a significant level of droplets clustering. Here, we propose a new analysis of this data, based on a “linear camera” reconstruction combined to a Taylor hypothesis. This allows us to reconstruct the particle concentration field over wide range of scales (from dissipative to metric scales). Using the Voronoi tesselation analysis we first identify clusters of particles. We then iterate the Voronoi analysis to investigate the clustering properties of the center of mass of the identified clusters. Our results show the clear tendency of clusters to form super-clusters with typical dimensions within inertial range scales. Trends of super-clustering with Stokes and Reynolds numbers will be discussed. |