Distentangling turbulent structure with nonlinear dynamics and machine learning
Abstract
The great challenge to understanding and predicting turbulence is its multiscale nature. Thediscovery of exact coherent states (ECS) – nonturbulent solutions to the governing equations offluid dynamics that capture key aspects of turbulent flows – has substantially advanced our understanding of turbulence, especially near solid surfaces where turbulence is most coherent andintense. Meanwhile, advances in machine learning hold promise for yielding insight into the complex patterns of turbulence, but have not been exploited in this context. The proposed work extends the ECS approach to turbulence and integrates it with ideas from machine learning to advance the understanding of turbulent flows.In Aim 1, new multiscale coherent states will be computed and new hypotheses regarding therelationships between coherent states and turbulence will be tested. To do so, advanced computational approaches will be developed and applied, enabling computations in new parameter regimes. These studies will aid us in understanding connections between turbulent flow structures at different scales.In Aim 2, advances in understanding exact coherent states will be complemented by advances inapplying machine learning and pattern analysis to turbulent flow data from direct simulations. Themachine learning paradigm of cluster analysis will generate new data-driven modal decompositionsand dynamic models. Information about exact coherent states will be integrated within the clusteranalysis and dynamic mode decomposition frameworks. The latter framework will be integratedwith dynamic models arising from cluster analysis. These models will be interrogated to yield newinsights into mode and scale interactions in turbulent flows.Extraordinary outcomes from this work include a new generation of results and insights forprediction and understanding of turbulent structure.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 09, 2018
- Source ID
- FA99501810174
Entities
People
- Michael D. Graham
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Wisconsin System