Data-Driven Resolvent Analysis

Benjamin Herrmann describes a data-driven algorithm to perform resolvent analysis from fluid mechanics to obtain the leading forcing and response modes, without recourse to the governing equations, but instead based on snapshots of the transient evolution of linearly stable flows. This approach is based on two established facts: 1) dynamic mode decomposition can approximate eigenvalues and eigenvectors of the underlying operator governing the evolution of a system from measurement data, and 2) a projection of the resolvent operator onto an invariant subspace can be built from this learned eigendecomposition.

Benjamin Herrmann, Peter J. Baddoo, Richard Semaan, Steven L. Brunton, and Beverley J. McKeon

This work will be discussed at the APS DFD conference at 8:45 AM, Monday, November 23, 2020
Session K09: Nonlinear Dynamics: Model Reduction (8:45am - 9:30am)

This video was produced at the University of Washington

[kouider]

I am falling in love with this channel, really impressive videos thanks alot

[ethandavis]

Well done, again, Benajmin! Another presentation I was hoping to see. For the future work, differentiating between linear and nonlinear forcing/dynamics, do you plan to use HAVOK (or, maybe, something similar) to extract linear attractors and nonlinear forcing from the data? I've played around with HAVOK a little bit with my turbulence data and find that it is quite difficult to capture a true picture of the dynamics/predict switching between attractors. I also have minimal experience with the nuances of all these methods (for now :) ). I am probably way out in left field, but I would love to hear your thoughts. If you aren't comfortable discussing this or feel it is not the right medium, I understand. Thank you for your time. Cheers.

[quantabot]

Best content for this field. Cant wait for that subscribes to hit 100K

[hamayuncfd]

are code for this analysis available on github ?