Application of data-driven system identification algorithms to nonlinear MHD simulations of astrophysical accretion flows

First principles models of plasmas lead to high-dimensional nonlinear systems that require complex MHD or kinetic simulations. Projection-based and data-driven modeling algorithms, such as dynamic mode decomposition (DMD) and sparse identification of nonlinear dynamics (SINDy), provide a potentially powerful approach to build low-dimensional reduced models of plasma systems. Recent advances in the application of these data-driven approaches, including the development of the “trapping SINDy” algorithm [1, 2], open the door to models accurate and small enough to be applied to real-time analysis and control. This poster will present progress on benchmarking these algorithms in magnetized plasmas, with a database of MHD simulations, using the NIMROD [3] code, of astrophysical accretion flows, which exhibit multiscale turbulent dynamics through the onset of the magneto-rotational and magneto-curvature instabilities [4]. System identification algorithms will then be tested on the datasets to determine the quality of the models that can be produced using these techniques. Plans to investigate related tools such as resolvent analysis will also be presented.

[1] - Kaptanoglu, Phys. Rev. E 104, 015206 (2021)

[2] - Kaptanoglu, Phys. Rev. Fluids 6, 094401 (2021)

[3] - A H Glasser et al 1999 Plasma Phys. Control. Fusion 41 A747

[4] - Ebrahimi and Pharr, The Astrophysical Journal, 936:145 (2022)