Working meeting – Machine Learning to develop diagnostic and data analysis methods for future thermonuclear reactors

POLONIUM (NAWA) – WORKING MEETING

March 31^st – April 01^st 2026

PAS Scientific Centre in Paris

This project, in collaboration between IFJ PAN and CEA IRFM, aims at using machine learning (ML) methods, in particular evolutionary algorithms (genetic algorithms (GA) or similar), artificial neural networks (NN) and Bayesian networks to support research in the field of plasma physics and thermonuclear fusion.

The research focuses on the following scientific problems:

(1) Design and optimization of plasma diagnostic systems for tokamak devices, e.g. neutron spectrometer or soft X-ray imaging systems;

(2) Development of fast tomographic reconstruction and spectrum deconvolution methods from a limited set of line-integrated measurements;

(3) Development of methods for analyzing data from plasma diagnostics in order to classify and predict plasma events such as impurity accumulation or magnetohydrodynamic instabilities during tokamak plasma discharges.

This dedicated in-person meeting at the PAN Paris stations was a unique opportunity to discuss the recent obtain results and plan the rest of the project activities between French and Polish partners

Institutions involved in the project:

• Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), Krakow, Poland
• Institute for Magnetic Fusion Research (IRFM), French Alternatives Energies and Atomic Energy Commission (CEA), Cadarache, France