The STCE welcomes Robert Jarolim from the Karl-Franzens-Universitât Graz for a working visit. He will give a seminar on
Title: Physics-informed neural networks for solar magnetic field simulations
Abstract: Physics-informed neural networks (PINNs) provide a novel approach for numerical simulations, tackling challenges of discretization and enabling seamless integration of noisy data and physical models (e.g., partial differential equations). In this presentation, we highlight the new opportunities that are enabled through physics-informed machine learning. We will discuss the results of our recent study where we apply PINNs for coronal magnetic field simulations of solar active regions, which are essential to understand the genesis and initiation of solar eruptions and to predict the occurrence of high-energy events from our Sun.
In order to provide for a non-linear force-free extrapolation of the coronal magnetic field, we optimize our network to match observations of the photospheric magnetic field vector at the bottom-boundary, while simultaneously satisfying the force-free and divergence-free equations in the entire simulation volume. We demonstrate that our method can account for noisy data and deviates from the physical model where the force-free magnetic field assumption cannot be satisfied. We utilize meta-learning concepts to simulate the evolution of the active region (AR) NOAA 11158. Our simulation of 5 continuous days of observations of the AR evolution at full cadence of 12 min of the HMI/SDO observations of the photospheric vector field (601 images), requires about 12 hours of total computation time. The derived evolution of the free magnetic energy and helicity in the active region, demonstrates that our model captures flare signatures, and that the depletion of free magnetic energy spatially aligns with the observed EUV emission from SDO/AIA. With this approach we present the first method that can perform realistic coronal magnetic field extrapolations in quasi real-time, which allows for advanced space weather monitoring.
We further provide an outlook on our ongoing work where we extend this approach to MHD simulations, that can flexibly incorporate additional observational constraints and perform fast computations. We validate our method by using a MURaM simulation as boundary-condition and comparing to the ground-truth reference. We conclude with a data-driven simulation of a solar flare based on SDO/HMI data.
Where: Meridian Room and zoom: https://us06web.zoom.us/j/82017679749?pwd=V05QTitYcEkyTm9FaHFLeFpNZFB6QT09
When: Wednesday June 7, 14:30
Contact: Veronique Delouille
