-
Validation of a Comprehensive First-Principles-Based Framework for Predicting the Performance of Future Stellarators
Authors:
D. L. C. Agapito Fernando,
A. Bañón Navarro,
D. Carralero,
A. Alonso,
A. Di Siena,
J. L. Velasco,
F. Wilms,
G. Merlo,
F. Jenko,
S. A. Bozhenkov,
E. Pasch,
G. Fuchert,
K. J. Brunner,
J. Knauer,
A. Langenberg,
N. A. Pablant,
T. Gonda,
O. Ford,
L. Vanó,
T. Windisch,
T. Estrada,
E. Maragkoudakis,
the Wendelstein 7-X Team
Abstract:
This paper presents the validation of the $\texttt{GENE-KNOSOS-Tango}$ framework for recovering both the steady-state plasma profiles in the considered radial domain and selected turbulence trends in a stellarator. This framework couples the gyrokinetic turbulence code $\texttt{GENE}$, the neoclassical transport code $\texttt{KNOSOS}$, and the transport solver $\texttt{Tango}$ in a multi-timescale…
▽ More
This paper presents the validation of the $\texttt{GENE-KNOSOS-Tango}$ framework for recovering both the steady-state plasma profiles in the considered radial domain and selected turbulence trends in a stellarator. This framework couples the gyrokinetic turbulence code $\texttt{GENE}$, the neoclassical transport code $\texttt{KNOSOS}$, and the transport solver $\texttt{Tango}$ in a multi-timescale simulation feedback loop. Ion-scale kinetic-electron and electron-scale adiabatic-ion flux-tube simulations were performed to evolve the density and temperature profiles for four OP1.2b W7-X scenarios. The simulated density and temperature profiles showed good agreement with the experimental data using a reasonable set of boundary conditions. Equally important was the reproduction of observed trends for several turbulence properties, such as density fluctuations and turbulent heat diffusivities. Key effects were also touched upon, such as electron-scale turbulence and the neoclassical radial electric field shear. The validation of the $\texttt{GENE-KNOSOS-Tango}$ framework enables credible predictions of physical phenomena in stellarators and reactor performance based on a given set of edge parameters.
△ Less
Submitted 11 March, 2025;
originally announced March 2025.
-
SafeSplit: A Novel Defense Against Client-Side Backdoor Attacks in Split Learning (Full Version)
Authors:
Phillip Rieger,
Alessandro Pegoraro,
Kavita Kumari,
Tigist Abera,
Jonathan Knauer,
Ahmad-Reza Sadeghi
Abstract:
Split Learning (SL) is a distributed deep learning approach enabling multiple clients and a server to collaboratively train and infer on a shared deep neural network (DNN) without requiring clients to share their private local data. The DNN is partitioned in SL, with most layers residing on the server and a few initial layers and inputs on the client side. This configuration allows resource-constr…
▽ More
Split Learning (SL) is a distributed deep learning approach enabling multiple clients and a server to collaboratively train and infer on a shared deep neural network (DNN) without requiring clients to share their private local data. The DNN is partitioned in SL, with most layers residing on the server and a few initial layers and inputs on the client side. This configuration allows resource-constrained clients to participate in training and inference. However, the distributed architecture exposes SL to backdoor attacks, where malicious clients can manipulate local datasets to alter the DNN's behavior. Existing defenses from other distributed frameworks like Federated Learning are not applicable, and there is a lack of effective backdoor defenses specifically designed for SL.
We present SafeSplit, the first defense against client-side backdoor attacks in Split Learning (SL). SafeSplit enables the server to detect and filter out malicious client behavior by employing circular backward analysis after a client's training is completed, iteratively reverting to a trained checkpoint where the model under examination is found to be benign. It uses a two-fold analysis to identify client-induced changes and detect poisoned models. First, a static analysis in the frequency domain measures the differences in the layer's parameters at the server. Second, a dynamic analysis introduces a novel rotational distance metric that assesses the orientation shifts of the server's layer parameters during training. Our comprehensive evaluation across various data distributions, client counts, and attack scenarios demonstrates the high efficacy of this dual analysis in mitigating backdoor attacks while preserving model utility.
△ Less
Submitted 24 February, 2025; v1 submitted 11 January, 2025;
originally announced January 2025.
-
Phantom: Untargeted Poisoning Attacks on Semi-Supervised Learning (Full Version)
Authors:
Jonathan Knauer,
Phillip Rieger,
Hossein Fereidooni,
Ahmad-Reza Sadeghi
Abstract:
Deep Neural Networks (DNNs) can handle increasingly complex tasks, albeit they require rapidly expanding training datasets. Collecting data from platforms with user-generated content, such as social networks, has significantly eased the acquisition of large datasets for training DNNs. Despite these advancements, the manual labeling process remains a substantial challenge in terms of both time and…
▽ More
Deep Neural Networks (DNNs) can handle increasingly complex tasks, albeit they require rapidly expanding training datasets. Collecting data from platforms with user-generated content, such as social networks, has significantly eased the acquisition of large datasets for training DNNs. Despite these advancements, the manual labeling process remains a substantial challenge in terms of both time and cost. In response, Semi-Supervised Learning (SSL) approaches have emerged, where only a small fraction of the dataset needs to be labeled, leaving the majority unlabeled. However, leveraging data from untrusted sources like social networks also creates new security risks, as potential attackers can easily inject manipulated samples. Previous research on the security of SSL primarily focused on injecting backdoors into trained models, while less attention was given to the more challenging untargeted poisoning attacks.
In this paper, we introduce Phantom, the first untargeted poisoning attack in SSL that disrupts the training process by injecting a small number of manipulated images into the unlabeled dataset. Unlike existing attacks, our approach only requires adding few manipulated samples, such as posting images on social networks, without the need to control the victim. Phantom causes SSL algorithms to overlook the actual images' pixels and to rely only on maliciously crafted patterns that \ourname superimposed on the real images. We show Phantom's effectiveness for 6 different datasets and 3 real-world social-media platforms (Facebook, Instagram, Pinterest). Already small fractions of manipulated samples (e.g., 5\%) reduce the accuracy of the resulting model by 10\%, with higher percentages leading to a performance comparable to a naive classifier. Our findings demonstrate the threat of poisoning user-generated content platforms, rendering them unsuitable for SSL in specific tasks.
△ Less
Submitted 2 September, 2024;
originally announced September 2024.
-
The Realization of a Gas Puff Imaging System on the Wendelstein 7-X Stellarator
Authors:
J. L. Terry,
A. von Stechow,
S. G. Baek,
S. B. Ballinger,
O. Grulke,
C. von Sehren,
R. Laube,
C. Killer,
F. Scharmer,
K. J. Brunner,
J. Knauer,
S. Bois,
the W7-X Team
Abstract:
A system for studying the spatio-temporal dynamics of fluctuations in the boundary of the W7-X plasma using the Gas-Puff Imaging (GPI) technique has been designed, constructed, installed, and operated. This GPI system addresses a number of challenges specific to long-pulse superconducting devices like W7-X, including the long distance between the plasma and the vacuum vessel wall, the long distanc…
▽ More
A system for studying the spatio-temporal dynamics of fluctuations in the boundary of the W7-X plasma using the Gas-Puff Imaging (GPI) technique has been designed, constructed, installed, and operated. This GPI system addresses a number of challenges specific to long-pulse superconducting devices like W7-X, including the long distance between the plasma and the vacuum vessel wall, the long distance between the plasma and diagnostic ports, the range of last closed flux surface locations for different magnetic configurations in W7-X, and management of heat loads on the system's plasma-facing components. The system features a pair of "converging-diverging" nozzles for partially collimating the gas puffed locally $\approx$135 mm radially outboard of the plasma boundary, a pop-up turning mirror for viewing the gas puff emission from the side (also acting as a shutter for the re-entrant vacuum window), and a high-throughput optical system that collects visible emission resulting from the interaction between the puffed gas and the plasma and directs it along a water-cooled re-entrant tube directly onto the 8 x 16 pixel detector array of the fast camera. The DEGAS 2 neutrals code was used to simulate the H$_α$ (656 nm) and the HeI (587 nm) line emission expected from well-characterized gas-puffs of H$_2$ and He and excited within typical edge plasma profiles in W7-X, thereby predicting line brightnesses used to reduce the risks associated with system sensitivity and placement of the field of view. Operation of GPI on W7-X shows excellent signal to noise ratios (>100) over the field of view for minimally perturbing gas puffs. The GPI system provides detailed measurements of the 2-dimensional (radial and poloidal) dynamics of plasma fluctuations in the W7-X edge, scrape-off layer, and in and around the magnetic islands that make up the island divertor configuration employed on W7-X.
△ Less
Submitted 15 May, 2024;
originally announced May 2024.
-
Prevention of core particle depletion in stellarators by turbulence
Authors:
H. Thienpondt,
J. M. García-Regaña,
I. Calvo,
J. A. Alonso,
J. L. Velasco,
A. González-Jerez,
M. Barnes,
K. Brunner,
O. Ford,
G. Fuchert,
J. Knauer,
E. Pasch,
L. Vanó,
the Wendelstein 7-X team
Abstract:
In reactor-relevant plasmas, neoclassical transport drives an outward particle flux in the core of large stellarators and predicts strongly hollow density profiles. However, this theoretical prediction is contradicted by experiments. In particular, in Wendelstein 7-X, the first large optimized stellarator, flat or weakly peaked density profiles are generally measured, indicating that neoclassical…
▽ More
In reactor-relevant plasmas, neoclassical transport drives an outward particle flux in the core of large stellarators and predicts strongly hollow density profiles. However, this theoretical prediction is contradicted by experiments. In particular, in Wendelstein 7-X, the first large optimized stellarator, flat or weakly peaked density profiles are generally measured, indicating that neoclassical theory is not sufficient and that an inward contribution to the particle flux is missing in the core. In this Research Letter, it is shown that the turbulent contribution to the particle flux can explain the difference between experimental measurements and neoclassical predictions. The results of this Research Letter also prove that theoretical and numerical tools are approaching the level of maturity needed for the prediction of equilibrium density profiles in stellarator plasmas, which is a fundamental requirement for the design of operation scenarios of present devices and future reactors.
△ Less
Submitted 8 November, 2023; v1 submitted 9 September, 2022;
originally announced September 2022.
-
Design considerations of the European DEMO's IR-interferometer/polarimeter based on TRAVIS simulations
Authors:
K. J. Brunner,
N. Marushchenko,
Y. Turkin,
W. Biel,
J. Knauer,
M. Hirsch,
R. Wolf
Abstract:
Interferometry is the primary density control diagnostic for large-scale fusion devices, including ITER and DEMO. In this paper we present a ray tracing simulation based on TRAVIS accounting for relativistic effects. The study shows that measurements will over-estimate the plasma density by as much as 20 degree. In addition, we present a measurement geometry, which will enable vertical position co…
▽ More
Interferometry is the primary density control diagnostic for large-scale fusion devices, including ITER and DEMO. In this paper we present a ray tracing simulation based on TRAVIS accounting for relativistic effects. The study shows that measurements will over-estimate the plasma density by as much as 20 degree. In addition, we present a measurement geometry, which will enable vertical position control during the plasma's ramp-up phase when gap-reflectometers and neutron cameras are still blind.
△ Less
Submitted 31 January, 2022; v1 submitted 17 December, 2021;
originally announced December 2021.
-
Heat and particle flux detachment with stable plasma conditions in the Wendelstein 7-X stellarator fusion experiment
Authors:
Marcin Jakubowski,
Ralf König,
Oliver Schmitz,
Yuhe Feng,
Maciej Krychowiak,
Matthias Otte,
Felix Reimold,
Andreas Dinklage,
Peter Drewelow,
Florian Effenberg,
Yu Gao,
Holger Niemann,
Georg Schlisio,
Andrea Pavone,
Thomas Sunn Pedersen,
Uwe Wenzel,
Daihong Zhang,
Sebastijan Brezinsek,
Sergey Bozhenkov,
Kai Jakob Brunner,
Daniel Carralero,
Ken Hammond,
Golo Fuchert,
Jens Knauer,
Andreas Langenberg
, et al. (8 additional authors not shown)
Abstract:
Reduction of particle and heat fluxes to plasma facing components is critical to achieve stable conditions for both the plasma and the plasma material interface in magnetic confinement fusion experiments. A stable and reproducible plasma state in which the heat flux is almost completely removed from the material surfaces was discovered recently in the Wendelstein 7-X stellarator experiment. At the…
▽ More
Reduction of particle and heat fluxes to plasma facing components is critical to achieve stable conditions for both the plasma and the plasma material interface in magnetic confinement fusion experiments. A stable and reproducible plasma state in which the heat flux is almost completely removed from the material surfaces was discovered recently in the Wendelstein 7-X stellarator experiment. At the same time also particle fluxes are reduced such that material erosion can be mitigated. Sufficient neutral pressure was reached to maintain stable particle exhaust for density control in this plasma state. This regime could be maintained for up to 28 seconds with a minimum feedback control.
△ Less
Submitted 10 January, 2020; v1 submitted 7 January, 2020;
originally announced January 2020.
-
Compensation of phase drifts caused by ambient humidity, temperature and pressure changes for continuously operating interferometers
Authors:
K. J. Brunner,
J. Knauer,
J. Meineke,
M. Stern,
M. Hirsch,
B. Kursinski,
R. C. Wolf,
the W7-X team
Abstract:
Fusion experiments rely heavily on the measurement of the line-integrated electron density by interferometry for density feed-back control. In recent years the discharge length has increased dramatically and is continuing to rise, resulting in environmentally induced phase drifts to become an increasingly worrisome subject, since they falsify the interferometer's measurement of the density. Especi…
▽ More
Fusion experiments rely heavily on the measurement of the line-integrated electron density by interferometry for density feed-back control. In recent years the discharge length has increased dramatically and is continuing to rise, resulting in environmentally induced phase drifts to become an increasingly worrisome subject, since they falsify the interferometer's measurement of the density. Especially in larger Tokamaks the loss of density control due to uncontrolled changes in the optical path length can have a disastrous outcome. The control of environmental parameters in large diagnostic/experimental halls is costly and sometimes infeasible and in some cases cannot be retro-fitted to an existing machine. In this report we present a very cheap (ca. 100 EUR), easily retro-fitted, real-time capable phase compensation scheme for interferometers measuring dispersive media over long time scales. The method is not limited to fusion, but can be applied to any continuously measuring interferometer measuring a dispersive medium. It has been successfully applied to the Wendelstein 7-X density feed-back interferometer.
△ Less
Submitted 5 November, 2019;
originally announced November 2019.
-
First Results from an Event Synchronized -- High Repetition Thomson Scattering System at Wendelstein 7-X
Authors:
Hannes Damm,
Ekkehard Pasch,
Andreas Dinklage,
Jürgen Baldzuhn,
Sergey Bozhenkov,
Kai Jakob Brunner,
Florian Effenberg,
Golo Fuchert,
Joachim Geiger,
Jeffrey Harris,
Jens Knauer,
Petra Kornejew,
Thierry Kremeyer,
Maciej Krychowiak,
Jonathan Schilling,
Oliver Schmitz,
Evan Scott,
Victoria Winters,
the Wendelstein 7-X Team
Abstract:
The Wendelstein 7-X (W7-X) Thomson scattering (TS) diagnostic was upgraded to transiently achieve kilohertz sampling rates combined with adjustable measuring times. The existing Nd:YAG lasers are employed to repetitively emit "bursts", i.e. multiple laser pulses in a short time interval. Appropriately timing burst in the three available lasers, up to twelve evenly spaced consecutive measurements p…
▽ More
The Wendelstein 7-X (W7-X) Thomson scattering (TS) diagnostic was upgraded to transiently achieve kilohertz sampling rates combined with adjustable measuring times. The existing Nd:YAG lasers are employed to repetitively emit "bursts", i.e. multiple laser pulses in a short time interval. Appropriately timing burst in the three available lasers, up to twelve evenly spaced consecutive measurements per burst are possible. The pulse-to-pulse increment within a burst can be tuned from 2 ms to 33.3 ms (500 kHz - 30 Hz). Additionally, an event trigger system was developed to synchronize the burst Thomson scattering measurements to plasma events. Exemplary, a case of fast electron density and temperature evolution after cryogenic H2 pellet injection is presented in order to demonstrate the capabilities of the method.
△ Less
Submitted 13 August, 2019; v1 submitted 30 June, 2019;
originally announced July 2019.
-
Analysis of trends in experimental observables and reconstruction of the implosion dynamics for direct-drive cryogenic targets on OMEGA
Authors:
A. Bose,
R. Betti,
D. Mangino,
K. M. Woo,
D. Patel,
A. R. Christopherson,
V. Gopalaswamy,
O. M. Mannion,
S. P. Regan,
V. N. Goncharov,
D. H. Edgell,
C. J. Forrest,
J. A. Frenje,
M. Gatu Johnson,
V. Yu Glebov,
I. V. Igumenshchev,
J. P. Knauer,
F. J. Marshall,
P. B. Radha,
R. Shah,
C. Stoeckl,
W. Theobald,
T. C. Sangster,
D. Shvarts,
E. M. Campbell
Abstract:
This paper describes a technique for identifying trends in performance degradation for inertial confinement fusion implosion experiments. It is based on reconstruction of the implosion core with a combination of low- and mid-mode asymmetries. This technique was applied to an ensemble of hydro-equivalent deuterium-tritium implosions on OMEGA that achieved inferred hot-spot pressures ~56+/-7 Gbar [S…
▽ More
This paper describes a technique for identifying trends in performance degradation for inertial confinement fusion implosion experiments. It is based on reconstruction of the implosion core with a combination of low- and mid-mode asymmetries. This technique was applied to an ensemble of hydro-equivalent deuterium-tritium implosions on OMEGA that achieved inferred hot-spot pressures ~56+/-7 Gbar [S. Regan et al., Phys. Rev. Lett. 117, 025001 (2016)]. All the experimental observables pertaining to the core could be reconstructed simultaneously with the same combination of low and mid modes. This suggests that in addition to low modes, that can cause a degradation of the stagnation pressure, mid modes are present that reduce the size of the neuron and x-ray producing volume. The systematic analysis shows that asymmetries can cause an overestimation of the total areal density in these implosions. It is also found that an improvement in implosion symmetry resulting from correction of either the systematic mid or low modes would result in an increase of the hot-spot pressure from 56 Gbar to ~80 Gbar and could produce a burning plasma when the implosion core is extrapolated to an equivalent 1.9 MJ symmetric direct illumination [A. Bose et al., Phys. Rev. E 94, 011201(R) (2016)].
△ Less
Submitted 27 March, 2018;
originally announced March 2018.
-
A Laboratory Investigation of Supersonic Clumpy Flows: Experimental Design and Theoretical Analysis
Authors:
A. Y. Poludnenko,
K. K. Dannenberg,
R. P. Drake,
A. Frank,
J. Knauer,
D. D. Meyerhofer,
M. Furnish,
J. R. Asay
Abstract:
We present a design for high energy density laboratory experiments studying the interaction of hypersonic shocks with a large number of inhomogeneities. These ``clumpy'' flows are relevant to a wide variety of astrophysical environments including the evolution of molecular clouds, outflows from young stars, Planetary Nebulae and Active Galactic Nuclei. The experiment consists of a strong shock (…
▽ More
We present a design for high energy density laboratory experiments studying the interaction of hypersonic shocks with a large number of inhomogeneities. These ``clumpy'' flows are relevant to a wide variety of astrophysical environments including the evolution of molecular clouds, outflows from young stars, Planetary Nebulae and Active Galactic Nuclei. The experiment consists of a strong shock (driven by a pulsed power machine or a high intensity laser) impinging on a region of randomly placed plastic rods. We discuss the goals of the specific design and how they are met by specific choices of target components. An adaptive mesh refinement hydrodynamic code is used to analyze the design and establish a predictive baseline for the experiments. The simulations confirm the effectiveness of the design in terms of articulating the differences between shocks propagating through smooth and clumpy environments. In particular, we find significant differences between the shock propagation speeds in a clumpy medium compared to a smooth one with the same average density. The simulation results are of general interest for foams in both inertial confinement fusion and laboratory astrophysics studies. Our results highlight the danger of using average properties of inhomogeneous astrophysical environments when comparing timescales for critical processes such as shock crossing and gravitational collapse times.
△ Less
Submitted 8 May, 2003;
originally announced May 2003.
