Ph.D. Proposal
Rohini Vaideswaran
(Advisor: Prof. P. K. Yeung)
"Investigations of Lagrangian Intermittency and Inertial Particle Dynamics using
Extreme-scale Computing"
Tuesday, June 3
10:00 a.m.
Montgomery Knight Building 317
Abstract
A Lagrangian view of turbulence naturally lends itself to studies of dispersion and transport of particles by turbulent flows. Particularly interesting are the effects of Lagrangian intermittency, which is known to be stronger than its Eulerian counterpart but is less understood. Investigations in the Lagrangian perspective require tracking individual particle trajectories. This is an experimental and computational challenge that demands high Reynolds numbers, adequate resolution in time and space, and reliable sampling. In this work, we present results from a GPU-accelerated pseudo-spectral code enabled with particle tracking, developed on the world's first Exascale supercomputer, Frontier. The use of a dynamic local particle decomposition and a ghost-layer approach for the spline coefficients leads to scaling independent of the number of particles. Performance results from particle tracking simulations at a resolution up to 327683 and tracking up to 1 billion particles are presented. Using this state-of-the-art simulation data, three main related lines of inquiry are proposed in this thesis.
First, we examine the inertial range Kolmogorov similarity scaling for the Lagrangian velocity increment. This quantity is observed to strongly deviate from theoretical predictions. The proposed work will provide valuable insights into the spatial-temporal dynamics of the Lagrangian velocity increment, and empower a more accurate characterization of its inertial range behavior over a wide range of Reynolds numbers. This study will have direct consequences for Lagrangian stochastic modeling, making it of substantial relevance to several areas of scientific and industrial research.
Secondly, we propose a study of the temporal evolution of the velocity gradient tensor (VGT) following fluid particle trajectories, which is key to determining the local behavior of the small scales in a turbulent velocity field. It is well known that the pressure Hessian plays a primary role in the evolution of the VGT, and is a focus of extensive modeling efforts. The behavior of the VGT is often examined in terms of its second and third order invariants. In this thesis, we will consider conditional statistics based on the particles' initial local condition, and investigate the effects of the pressure Hessian on the temporal evolution of the VGT invariants.
Lastly, the particle tracking effort is extended to analyze the dynamics of inertial particles in turbulent flows, which tend to cluster and preferentially sample certain regions of the flow field, unlike fluid particles. The development of a GPU-enabled algorithm is proposed to compute the Voronoi tessellation of inertial particle data for studying the spatial and temporal dynamics of clustering phenomena from extreme-scale particle datasets. A detailed study of Reynolds number effects on particle clustering will be performed, and the impact of intermittency on the formation and destruction of a particle cluster will be analyzed.
The proposed work will be carried out on leadership-class computational facilities at the U.S. Department of Energy’s Oak Ridge National Laboratory and the National Science Foundation supported Texas Advanced Computing Center.
Committee
- Prof. P. K. Yeung – School of Aerospace Engineering (advisor)
- Prof. Suresh Menon – School of Aerospace Engineering
- Prof. Chris C. K. Lai – School of Civil and Environmental Engineering