Peter Beyer

Peter Beyer

Auto-generated profile, based on publicly available information

Peter Beyer

Aix-Marseille Université, CNRS, PIIM, UMR 7345 1 , 13397 Marseille

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 September 2025 in Physics of Plasmas
Verified
Journal Article
Open access
DGDavid Garrido González
David Garrido González
NSNathaniel Saura
Nathaniel Saura
PBPeter Beyer
Peter Beyer
WMWaleed MOUHALI
Waleed MOUHALI
SBSadruddin Benkadda
Sadruddin Benkadda
We present an AI-driven reduced order model (ROM) for drift-wave turbulence dynamics based on proper orthogonal decomposition (POD) combined with Galerkin projection. Using the Hasegawa–Wakatani model as a proof of concept, our framework integrates neural ordinary differential equations (NODEs) with POD. In this formulation, a neural network learns the time derivative of the reduced order state, effectively modeling the underlying vector field that governs the system's dynamics. This allows direct access to the phase portrait of the system and provides insights into trends and patterns by capturing its continuous temporal evolution. Results show that the proposed method reconstructs key features of the system and holds promise for significant computational speed-up.

1 December 2024 in Review of Scientific Instruments
Verified
Journal Article
Closed access
NSN. Schoonheere
N. Schoonheere
CRC. Reux
C. Reux
HMH. Meister
H. Meister
PBP. Beyer
P. Beyer
PCP. Carvalho
P. Carvalho
TTthe EUROfusion Tokamak Exploitation Team
the EUROfusion Tokamak Exploitation Team
Massive material injections in the JET tokamak have been observed to substantially affect resistive bolometer measurements, resulting in a spurious radiated power signal proportional to the quantity injected and reaching up to 8 MW. These bolometers are calibrated and designed to operate in near vacuum but certain scenarios requiring large gas injections can push the neutral pressure past nominal values. This study demonstrates that the bolometry measurement can be affected at neutral pressures above 0.1 Pa following injections with standard gas valves, shattered pellet injections, and particularly massive gas injections. The power measurement of resistive bolometers is based on the temperature difference between a measurement sensor exposed to radiation and a shielded reference sensor. We employ a thermal conductivity model to demonstrate that the conduction through the gas and the distinct geometries between the sensors can affect their cooling efficiency. This additional cooling pathway, coupled with the Joule heating from the applied voltage causes the equilibrium temperatures of the sensors to diverge. Being the very basis of the measure, this temperature difference induces a signal that is erroneously interpreted as radiated power. Experiments show large discrepancies in the response to neutral pressure among bolometer channels, attributed to variations in channel physical parameters. Nonetheless, the modeled total radiated power reproduces the experimental measurements within an order of magnitude, affirming the sensitivity of resistive bolometers to neutral pressure and gas species.

1 June 2025 in Physics of Plasmas
Verified
Journal Article
Open access
DGDavid Garrido González
David Garrido González
NSNathaniel Saura
Nathaniel Saura
SBSadruddin Benkadda
Sadruddin Benkadda
PBPeter Beyer
Peter Beyer
Neural ordinary differential equations (NODEs) offer a data-driven approach to modeling chaotic dynamical systems without prior knowledge of governing equations. We show that enhanced phase-space sampling through diverse initial conditions significantly improves NODEs' ability to capture both local and global system behaviors. Using the Lorenz system, we demonstrate that broader phase-space coverage enhances generalization, reduces overfitting, and improves long-term predictive stability. Evaluations based on Lyapunov exponents, the Kaplan–Yorke dimension, return maps, and probability density functions confirm that increased initial condition diversity leads to more accurate reconstructions of chaotic attractors and system dynamics.

1 August 2026 in Plasma Physics and Controlled Fusion
Verified
Journal Article
Closed access
PDF file
DGD Garrido González
D Garrido González
NSN Saura
N Saura
ASA Saleh
A Saleh
WMWaleed MOUHALI
Waleed MOUHALI
PBP Beyer
P Beyer
SBS Benkadda
S Benkadda
Predictive, fast reduced-order models are essential for fusion devices like ITER, where real-time transport forecasting, optimisation, and control cannot depend on costly first-principles turbulence simulations alone. We tackle this problem with the LaQuey Mahajan, Rutherford, and Tang equation, a reduced trapped-ion mode turbulence model equivalent, after nondimensionalisation, to the Kuramoto–Sivashinsky equation. We use a geometry-aware reduced-order framework over a two-parameter space of damping and dissipation. Local proper orthogonal decomposition bases extracted from direct numerical simulation are used to build Galerkin reduced models. Frozen bases remain accurate when turbulent structures vary smoothly, but they deviate near transitional regimes where fluctuation content reorganises. Grassmann-manifold analysis shows that subspace proximity does not always guarantee physical accuracy: transport may be reproduced even when fine-scale structure is not. Adaptive, parameter-dependent bases obtained by manifold interpolation significantly improve robustness and recover both large-scale and fine-scale fidelity. Overall, the results show that transport accuracy, structural fidelity, and basis geometry are complementary validation criteria for plasma surrogates, and they support efficient predictive models for ITER-scale optimisation and control.