What is it about?
Understanding the stability of fusion plasmas is essential for developing future fusion energy devices. However, running simulations to determine plasma stability is both computationally demanding and time-consuming. This study describes a proof-of-principle machine learning surrogate model that accelerates the simulation of ideal magnetohydrodynamics (MHD) stability for tokamak plasmas. The ideal MHD stability code used to generate the training data is called MISHKA and the surrogate model is called KARHU. The machine learning model can quickly estimate the ideal MHD stability of a plasma equilibrium. We also show that this model can be integrated into existing workflows (Europed) and demonstrate its use on data from the Joint European Torus (JET), one of the world’s largest experimental fusion devices.
Featured Image
Photo by BoliviaInteligente on Unsplash
Why is it important?
This work describes the development of a surrogate model for ideal MHD stability predictions and demonstrates how a machine learning model can be used to accelerate predictive workflows with reasonable accuracy. A fast predictive model like this enables more efficient research and supports large-scale data analysis.
Perspectives
This work presents the first version of our surrogate model. The model is currently trained on a restricted set of plasmas. Future development will focus on extending its applicability to cover a wider range of plasma conditions, including data from other devices and resistive MHD. Both the model and the training dataset are openly available on GitHub.
Amanda M Bruncrona
VTT Technical Research Centre of Finland
Read the Original
This page is a summary of: Machine learning surrogate model for ideal peeling–ballooning pedestal MHD stability, Physics of Plasmas, September 2025, American Institute of Physics,
DOI: 10.1063/5.0282085.
You can read the full text:
Resources
Contributors
The following have contributed to this page
