AI-powered Geodynamics
Toward the Next-generation of Geodynamic Simulation
Research Overview
Neural operators enable geodynamic models to tackle both forward and inverse mantle convection, greatly accelerating simulations while maintaining accuracy. AI also provides a framework to couple geodynamics with thermodynamics, integrating chemical and physical processes into a unified view of Earth’s interior.
Research Questions
How can neural operators accelerate forward and inverse mantle convection while matching the accuracy of traditional models? How can AI integrate mantle flow with thermodynamic phase changes to build unified models of Earth’s evolution?
Key Findings
We show that the machine learning framework accelerates coupled geodynamic-geochemical modeling by more than 100-fold. The approach paves the way for integrating large-scale 2D and 3D geodynamic and geochemical modeling studies across a broader range of model parameters and at larger spatial and temporal scales than previously possible.
Research Images
Bridging mantle flow and thermodynamics efficiently with Neural Network Acceleration
Publications
Enabling large-scale geodynamic-geochemical modeling via neural network acceleration
Yuan, Q., Asimow, P. D., Gurnis, M., & Antoshechkina, P. M.
AGU (2024) • DOI: https://ui.adsabs.harvard.edu/abs/2024AGUFMV41E.3161Y/abstract
Interested in This Research?
Learn more about our research group, publications, and opportunities to collaborate.