*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Extreme-scale computing and studies of intermittency, mixing of passive scalars and stratified flows in turbulence
Kiran Ravikumar
Date : 23rd June 2020
Time : 04:15 PM (EST)
Location : https://bluejeans.com/582023325/
Advisor : Prof. P. K. Yeung
Turbulence is well known for the intermittent occurrence of intense strain rates and local rotation acting on individual fluid elements, and for its ability to provide efficient mixing, which is, for instance, essential in aircraft engines. In this work, we focus on understanding the fundamental behavior of such flows using high resolution Direct Numerical Simulations based on a Fourier pseudo-spectral approach. A major motivation for ever-larger simulations is to advance understanding of intermittency and extreme events in turbulence and the mixing of passive scalars, where both high Reynolds number and good small-scale resolution are clearly important. Simulations at world leading problem sizes are enabled using a new batched asynchronous algorithm developed for dense node heterogeneous architecture machines like Summit. With fast computations due to GPUs, optimizing data copies between the CPU and GPU and communication over the network are key to achieve good performance. A batched asynchronous approach helps reaching large problem sizes without being restricted by the limited GPU memory, where data residing on the CPU is processed in small batches on the GPU enabling overlap between computations on the GPU and data copies. Although the new algorithm allows high resolution simulations, they tend to be short due to very high computational resource demands which raises concerns regarding sampling and statistical independence. In this work, “Multiple Independent Simulations” approach is used, where ensemble average over multiple short simulations at high resolution starting from well spaced out snapshots at lower resolution are used to address these limitations. Using this approach, we study the effects of small-scale intermittency through statistics of dissipation rate and enstrophy where high resolution in both space and time helps capture the extreme fluctuations in the velocity gradients accurately. The same approach is employed to study intermittency, through statistics of scalar dissipation rate, in passive scalar turbulent mixing with two passive scalars of moderate molecular diffusivity. However, many atmospheric and oceanic flows are characterized by density variations due to the presence of active scalars that affect the behavior of the flow. Both stable and unstable stratified flows are considered with the most interesting being when the two scalars of different molecular diffusivities have opposing effects with differential diffusion playing a pivotal role. This work involves careful consideration of numerical requirements, especially for both spatial and temporal discretization errors in the case of overall unstable stratification. In order to understand the energy cascade and development of anisotropy the Reynolds-stress budget will be computed.
Committee
