Avis de Soutenance de thèse de Monsieur Enzo Vergnaud

Avis de Soutenance

Monsieur Enzo VERGNAUD

Soutiendra publiquement ses travaux de thèse intitulés :

« Integrated modelling of energy confinement time scaling with

plasma current in tokamak plasmas: from a dedicated study on WEST towards large scale validation »

Soutenance prévue le jeudi 29 octobre 2026 à 14h

Lieu : CEA – IRFM, salle René Gravier, Bâtiment 506

Et par Skype

Composition du jury :

Pascale HENNEQUIN	LPP, CNRS	Rapportrice
Paola MANTICA	ISTP, CNR	Rapportrice
Olivier SAUTER	SPC, EPFL	Examinateur
Eric NARDON	IRFM, CEA	Président du jury
Yann CAMENEN	PIIM, CNRS	Directeur de thèse
Clarisse BOURDELLE	IRFM, CEA	Co-Directrice de thèse

Abstract

Predicting the energy confinement time τ_E is essential for tokamak scenario design and extrapolation towards ITER and future reactors. So far, this relies on scaling laws, statistical regressions with limited extrapolation. Integrated modelling offers an alternative, leveraging scale separation to solve the 1.5D transport equations (current diffusion, heat, particle and impurity transport) coupled to the 2D equilibrium and sources, taking only engineering parameters and separatrix conditions as inputs. This approach is applied to WEST, whose tungsten environment makes impurity transport critical for confinement, with the High Fidelity Pulse Simulator (HFPS) coupled to the turbulent transport model TGLF-SAT2. The framework was first applied to the reference L-mode discharge #57757 from core to separatrix, and is extended here to seven similar discharges to investigate the τ_E dependence on plasma current.

The plasma current I_P is the most influential engineering parameter of these scaling laws. Physically, the plasma current impacts turbulent transport through the safety factor q ∝ 1/ I_P , which determines the winding of the magnetic field around the flux surface and the parallel connection length which plays a key role in micro-instabilities.

The HFPS is applied to 8 WEST L-mode discharges spanning I_P = 0.26−0.5 MA in four type of simulations of increasing fidelity, progressively including current diffusion, heat, particle and impurity transport. Synthetic diagnostics are used to compare the profiles from the predictive simulations with impurities to ECE, interferometry, bolometry and neutron rate measurements, with a relative difference of 9% on T_e, 9% on the line-averaged density and 8% on the total radiated power. The predicted confinement time reproduces the increasing trend with I_P with a relative error of 8%, better than an ITER L96-type scaling law (12%).
For large-scale validation, a Python pipeline automates input preparation, job submission and error handling with corrective relaunch.

As a proof of principle, TGLF-SAT2 is replaced by TGLFNN, a neural network trained via Latin Hypercube sampling in 13 dimensions, speeding up computation by up to two orders of magnitude. Applied to the same 8 discharges, it reproduces the expected trends but shows clear limitations at the edge and on the density profile, due to a training domain of TGLFNN still too narrow for WEST and localised non-physical behaviors in particle transport. Finally the next steps to extend the HFPS to the hundreds of pulses in the WEST database are outlined.

Keywords

Magnetic fusion, tokamak, plasma, turbulent transport, quasi-linear, confinement time prediction, integrated modelling, WEST

………………………………………………………………………………………………………………………..

Participer à une Réunion Skype

Vous n’arrivez pas à rejoindre cette réunion ? Essayer l’App web Skype

Participer par téléphone

+33 1 69 35 55 10,,93392562# (France) Français (France)

Rechercher un numéro local

ID de conférence : 93392562

Avez-vous oublié votre code confidentiel de participation ? |Aide

Pour rejoindre la réunion Skype

Depuis un poste CEA ou non-CEA dans un navigateur Web (Windows, MacOS, Linux, iOS, Androïd) :