Droplet preferential concentration in homogeneous and isotropic turbulence

Abstract

In particle-laden turbulent flow, it has been found both experimentally and numerically that when the particle response time is similar to a turbulent characteristic timescale, particles tend to preferentially concentrate and form clusters. This phenomenon of non-uniform particle dispersion has been referred as preferential concentration.

The thesis studies experimentally the preferential concentration of poly-dispersed droplets in homogeneous and isotropic turbulence generated in the facility referred to as the `box of turbulence' and includes comparisons with Direct Numerical Simulations (DNS). It discusses the effect of poly-dispersion on droplet preferential concentration, temporal evolution of droplet clustering and the turbulent mechanisms (i.e. topological turbulent flow patterns) that may be responsible for the droplet clustering dynamics.

The thesis is structured into six chapters. Motivations, theoretical background and related literature of this work are discussed in Chapter 1. Chapter 2 describes the experimental setup and the applied laser diagnostic techniques. Chapter 3 focuses on the effect of poly-dispersion on droplet preferential concentration. The techniques used in quantifying the preferential concentration are the Radial Distribution Function (RDF) and Voronoï analysis. An image processing method for locating droplets from droplet Mie-scattering images has been proposed and evaluated. Chapter 4 reports the time-resolved dispersion measurements of poly-dispersed droplets. The fine scale topological turbulent patterns (i.e. zero velocity/acceleration) are extracted from the fluid flow velocity measurements, considering the effect of experimental noise, and are observed to follow a non-uniform spatial distribution and form clusters. The clustering of zero velocity/acceleration points are quantified by RDF and Voronoï analysis and compared with the dispersed droplet clusters. A cluster identification method based on the mean shift pattern space analysis and the Voronoï tessellation has been proposed and applied to all the temporally resolved images to obtain cluster time scale and length scale statistics. Chapter 5 compares the results from experiments and corresponding DNS calculations using the same data processing methods. The clustering of experimentally and numerically acquired zero velocity/acceleration points and dispersed droplets are quantified and compared. Chapter 6 is the conclusion of the thesis and possible directions of future work.