This research is among the core activities of the group since 2002. We use Bayesian inference and machine learning for data integration and sensor fusion, combining heterogeneous data sources to reduce uncertainties on the estimated plasma state. We work on fully probabilistic reconstruction of the magnetic equilibrium and studies of impurity transport in the plasma, based on data from magnetic sensors and spectroscopy. We focus on real-time applicability for plasma control in the next generation of fusion devices.
Tomography of soft X-ray radiation (synthetic data) using Bayesian inference with Gaussian process priors. (H. Wu et al., Plasma Phys. Control. Fusion 67, 085001, 2025)
Current density and plasma position for DEMO obtained with Bayesian techniques, based on external magnetic sensor data only. (J. De Rycke et al., 21st International Congress on Plasma Physics, Ghent, 2024)