Speaker
Description
We present a validation study of a Bayesian integrated data analysis (IDA) approach for current tomography, applied to the WEST tokamak using only external magnetic diagnostics. The methodology was originally developed for DEMO but is here adapted to real data, addressing challenges such as noise, currents in structural elements, and iron core effects [1].
A Gaussian prior distribution was constructed for the plasma current density, using the raw second moment from a series of reference equilibria obtained from WEST. The iron core was explicitly modeled using a Gaussian process prior, by segmenting it into six regions, with correlations between segments described by a squared exponential (SE) kernel. No fixed currents were assigned to the iron core; instead, the most probable distribution of core-induced currents was inferred self-consistently from the diagnostic measurements and the prior.
Validation was performed over multiple WEST discharges, focusing on plasma parameters such as the current centroid, total current, and X-point position. As standard WEST reconstructions using the NICE code can suffer from systematic errors [2], the VacTH code was used as an independent reference for the X-point location [3]. Weakly informed priors were used to increase the influence of the diagnostic measurements, identifying systematic discrepancies between our results and those from NICE. A global shift and scaling of the WEST tomograms were then applied to minimize discrepancies for estimates of the current centroid and total current.
The corrected Bayesian reconstructions showed better agreement with VacTH-derived X-point positions compared to the original WEST reconstructions (NICE), confirming the capability of the method to improve plasma boundary determination using external magnetic signals only. This work provides a stepping stone towards integrated plasma reconstruction including additional diagnostics.
J. De Rycke acknowledges the Research Foundation - Flanders (FWO) via PhD grant 1SH6424N
[1] J. De Rycke, et al., Proc. 21st Int. Congr. Plasma Phys. (2024).
[2] T. Radenac, WEST Task Force Mtg. (2022).
[3] P. Moreau, et al., Rev. Sci. Instrum. 89 (2018).
