Abstract Details


Name: Prateek Mayank
Affiliation: University of Colorado, Boulder
Conference ID: TVS202510141
Title: Next-Generation MHD Modeling Of Solar Wind Using Neural Operators
Authors and Co-Authors: Enrico Camporeale, Zhenguang Huang, Gabor Toth
Abstract Type: Contributory Presentation
Abstract: Traditional magnetohydrodynamic (MHD) solvers remain indispensable for modeling heliospheric plasma dynamics, yet their high computational cost and limited scalability hinder ensemble simulations and real-time forecasting. In this study, we propose a new framework employing neural operators to efficiently emulate 3D solar wind conditions learned directly from MHD simulation data. Our approach integrates observationally-derived multi-channel inputs and utilizes a hybrid training scheme, combining data-driven supervised learning with physics-informed constraints through embedded conservation laws. The neural operator demonstrates robust generalization capabilities, capturing complex heliospheric structures such as stream interaction regions and heliospheric current sheets. By ensuring bot