Computational study of particle clustering and dispersion in turbulence

This thesis presents a comprehensive investigation into aerosol dispersion and the preferential concentration of particles in turbulent flows, with a focus on computational modelling. It introduces an Eulerian computational method based on the Roe-flux scheme, enhancing the efficiency of particle-laden ow simulations by incorporating second-order velocity moments. This method, compared to traditional Lagrangian particle tracking, is applied to Direct Numerical Simulations (DNS) of both the turbulent ow and particle phase, offering new insights into particle velocity and concentration fluctuations. The research explores the effects of the Stokes (0 < Stk < 10), Reynolds (50 < Re λ < 120), and Froude (0 < F r l < 5) numbers on particle clustering. Particles with Stokes numbers near unity exhibit the strongest clustering, while higher Stokes numbers reveal that increasing Reynolds and Froude numbers influence clustering patterns at different scales. A novel empirical model is developed to predict the spectra of particle concentration fluctuations under varying conditions, providing a practical tool for analysing aerosol behaviour in environments where detailed computations may be infeasible. A significant contribution of this thesis is the investigation into the role of Lagrangian Coherent Structures (LCS) in generating preferential concentration. By using the Lagrangian-averaged vorticity deviation (LAVD), the research shows that rotational coherence in turbulent flows plays a critical role in particle clustering, improving upon previous approaches like the Q criterion. Additionally, it demonstrates that incorporating time delays between fluid structures and particle positions significantly enhances the correlation between particle clusters and turbulent features. These findings contribute to a deeper understanding of particle-turbulence interactions and have important applications in atmospheric aerosol dispersion, pollution control, and industrial processes, offering predictive models that can help manage aerosol behaviour in complex, large-scale environments.