I do research in the McKeon Research Group
While the group currently operates out of Stanford University, I joined while professor McKeon was at the Graduate Aerospace Laboratories at California Institute of Technology (GALCIT) and continue to do my work there. In the lab, my research has been centered on developing technologies to study complex fluid flows in new ways. Below are some of the specific projects that I have worked on since attending Caltech.
Computer Vision to Identify Flow Features using Natural Tracers
The motion of tracers in a fluid flow indicates the behavior of the flow itself. Often, in experimental fluid mechanics, artificial tracers with favorable properties (small, reflective, neutrally buoyant, and non-inertial) are seeded into the flow and illuminated for the sake of measuring the fluid dynamics. There are, however, many scenarios where it is not possible, practical, or ethical to put artificial tracers into the fluid to study the flow. In many of these cases, natural tracers already exist. For example, think about a river with debris particulate on the surface moving with the current.
In this project we attempt to establish a generalizable framework for using naturally existing particles in a flow to study either the dynamics of the underlying flow or the dynamics of the particles themselves. We accomplish this in a three step process that involves large modern computer vision networks to find debris, Lagrangian particle tracking to stitch trajectories together, and Lagrangian gradient regression to compute flow quantities such as vorticity, LAVD, and FTLE.
Having such a tool allows us to study a broad range of flow phenomenon that is difficult to study by conventional methods. Moreover, because each of the steps in the process are rapidly evolving technologies, there is plenty of room for technical advances in this area. We have presented on this work at the APS DFD conference in 2023, and are in the process of producing publications.
Image: Laboratory debris flow with tracers colored by the dynamic rotation angle (DRA). Brighter colors represent more rotation.
Lagrangian Gradient Regression (LGR)
Lagrangian gradient regression (LGR) offers a derivative and field free alternative to computing velocity gradients and flow map Jacobians for dynamical systems including fluid flows.
While velocity gradients are extremely useful to the analysis of fluid flows (vorticity, for example, requires computation of velocity gradients), they are often expensive to compute and can produce large numerical errors since differentiation schemes typically require that velocities be interpolated to a grid and differentiated in their computation. LGR avoids interpolation and differentiation by fitting the gradient operator through regression of the flow deformation over short time intervals.
While LGR can be used to compute velocity gradients at an instance in time, it can also be applied to the problem of Lagrangian coherent structure (LCS) identification over finite periods of time through the composition of many snapshots. Because deformations over short time intervals are linear even for sparse tracer distributions, quantities like Lagrangian-averaged vorticity deviation (LAVD) and finite-time Lyapunov exponent (FTLE) can be reliably computed from sparse tracers using LGR.
Because of its effectiveness with sparse and noisy data, LGR enables the study of a variety of real-world flows that would typically be difficult to apply traditional tools to. Examples include large-scale Lagrangian particle tracking experiments and geophysical flows characterized by ocean drifter or sea-ice movements.
For more information about LGR, check out our preprint article here: Lagrangian Gradient Regression for the Detection of Coherent Structures from Sparse Trajectory Data (arxiv.org) with associated code: LGR GitHub
I am also slowly compiling an LCS primer for people new to the field, which can be found here: LCS Primer GitHub
Cycle-to-Cycle Variations of Dynamic Stall
Dynamic stall is a complex flow phenomenon that occurs on lifting surfaces (like the wing of an aircraft) that undergo rapid changes in angle of attack. A classic example of this type of flow occurs on rotating blade systems such as helicopters and wind turbines. For example, the blades on a helicopter in forward flight experience a cyclical relative wind velocity (faster incoming speed when the blade is moving in the direction of the vehicle and slower when it is retreating). If the conditions are correct, the flow will separate from the surface of the blade periodically, which constitutes an instance of dynamic stall.
Researchers who study dynamic stall will often consider thousands of pitching cycles on a blade and examine their phase average throughout the cycle. The work that I did during my master’s degree at the University of Wyoming with Dr. Jonathan Naughton explored the accuracy of these phase averages by considering them relative to individual cycles. We found that in many practical cases, the phase average does not reliably represent the true aerodynamic behavior of the blade, and may therefore contribute to errors in numerical models used to simulate the dynamic stall process. We found that, instead of a smooth distribution between attached flow and separation, there is a transitional regime where individual cycles will switch between fully separated and fully attached behavior, even in the same experiment.
Publications from this work include Experimental evaluation of the cycle-to-cycle variation of dynamic stall on the SC1094R8 airfoil and Modal analysis of the cycle-to-cycle variations observed in dynamic stall along with my UW thesis and various co-authored articles.
Selected Publications and Talks
2023
Harms, T. D., Brunton, S. L., & McKeon, B. J. “Lagrangian Gradient Regression for the Detection of Coherent Structures from Sparse Trajectory Data.” arXiv preprint arXiv:2310.10994 (2023)
Harms, T. D., Brunton, S. L., and McKeon B. J. (2023) “Direct Computation of Velocity Gradients from Particle Trajectories.” 15th International Symposium on Particle Image Velocimetry, 1.1 (2023)
APS Annual Meeting of the Division of Fluid Dynamics. Presentation title: “Sparse Detection of Flow Features using Natural Tracers.”
SoCal Fluids Conference. Presentation title: “Identifying Geometric LCS from Sparse Trajectory Data.”
Stanford TFSA Conference. Presentation title: “Efficient Methods for Identifying and Tracking Coherent Structures.”
2022
APS Annual Meeting of the Division of Fluid Dynamics. Presentation title: “Finite-Time Lyapunov Exponent via Locally Linear Regressions.”
2021
Ramasamy, M., Sanayei, A., Wilson, J. S., Martin, P. B., Harms, T. D., Nikoueeyan, P., Naughton, J. “Reducing uncertainty in dynamic stall measurements through data-driven clustering of cycle-to-cycle variations.” Journal of the American Helicopter Society, 66.1 (2021): 1-17
2019
Ramasamy, M., Sanayei, A., Wilson, J. S., Martin, P. B., Harms, T. D., Nikoueeyan, P., Naughton, J. “Data-driven optimal basis clustering to characterize cycle-to-cycle variations in dynamic stall measurements.” Proceedings of the Vertical Flight Society 75th Annual Forum (2019)
Nikoueeyan, P., Harms, T. D., Naughton, J., Ahuja, V. “Experimental assessment of an airfoil optimized to delay the onset of dynamic stall.” Proceedings of the Vertical Flight Society 75th Annual Forum (2019)
Ramasamy, M., Sanayei, A., Wilson, J. S., Martin, P. B., Harms, T. D., Nikoueeyan, P., Naughton, J. “Static and dynamic aerodynamic performance parameters for S814 and S825 airfoils at moderate Reynolds number.” AIAA SciTech Forum (2019): 0802
2018
Harms, T. D., Nikoueeyan, P., Naughton, J. “Modal analysis of the cycle-to-cycle variations observed in dynamic stall” AIAA SciTech Forum (2018): 1267
Harms, T. D., Nikoueeyan, P., Naughton, J. “An experimental evaluation of cycle-to-cycle variations of dynamic stall.” Proceedings of the Vertical Flight Society 74th Annual Forum (2018)
APS Annual Meeting of the Division of Fluid Dynamics. Presentation title: “Characterization and understanding of unsteady airfoil behavior using modal analysis.”
2017
APS Annual Meeting of the Division of Fluid Dynamics. Presentation title: “Analysis of the cycle-to-cycle pressure distribution variations in dynamic stall.”
Want to get in touch?
If you have questions regarding my research, want to collaborate on a project, or have other thoughts to share, send me an email! I will do my best to respond promptly, but if you don’t hear back from me in a few weeks, feel free to resend your message.