Barbara Cannas

Real-time estimation of divertor heat loads is critical for plasma-facing component protection in Wendelstein 7-X (W7-X). The current heat-flux reconstruction tool, THEODOR, is too computationally demanding for real-time use, motivating the development of faster physics-based surrogates. Physics-informed neural networks (PINNs) have recently been shown to model the heat equation and the associated heat-flux partial differential equation, though only for fixed boundary and initial conditions, when the heat potential profile at the top of the tile is represented as a Gaussian function. This choice is motivated by the observation that experimental profiles can be well approximated by a small number of Gaussian peaks in the strike-line region. Within this framework, the present work extends the PINN framework by assessing the sensitivity of the PDE solution to variations in the boundary and initial conditions, using a synthetic dataset with Gaussian boundary-condition profiles. Two approaches are investigated: (i) training multiple PINNs for different initial tile temperatures and Gaussian boundary-condition parameters; (ii) developing a parameterized PINN capable of solving the PDE across a continuous range of conditions. This approach demonstrates the feasibility of PINN-based heat-flux reconstruction with improved flexibility, in view of the application with real-time experimental data at W7-X.

The short time retractable instrumented calorimeter experiment is a critical diagnostic tool for characterizing the source for production of negative ions of deuterium extracted from radio frequency plasma. Due to technical limitations, direct temperature measurement on the calorimeter’s front side, facing the ion source, is infeasible. Instead, reconstructing the front-side heat flux from back-side temperature data formulates an inverse problem. This study introduces a convolutional neural network (CNN) to achieve real-time reconstruction of the heat flux distribution, modeled with Gaussian fitting for each beamlet, directly from infrared (IR) camera images. The CNN’s performance is benchmarked against a multi-layer perceptron (MLP) model, which requires offline parametrization of IR images, limiting real-time applicability. Using experimental IR images and heat flux data generated via iterative finite element method modelling, the CNN demonstrated comparable accuracy to the MLP without the need for feature extraction engineering, offering a robust real-time solution.

Overload detection is crucial in nuclear fusion experiments to prevent damage to plasma-facing components (PFCs) and ensure the safe operation of the reactor. At Wendelstein 7-X (W7-X), real-time monitoring and prediction of thermal events are essential for maintaining the integrity of PFCs. This paper proposes a machine learning approach for developing a real-time overload detector, trained and tested on OP1.2a experimental data. The analysis showed that Self-Organizing Maps (SOMs) are efficient in detecting the overload risk starting from a set of plasma parameters that describe the magnetic configuration, the energy behavior, and the power balance. This study aims to thoroughly evaluate the capabilities of the SOM in recognizing overload risk levels, defined by quantizing the maximum criticality across different IR cameras. The goal is to enable detailed monitoring for overload prevention while maintaining high-performance plasmas and sustaining long pulse operations. The SOM proves to be a highly effective overload risk detector. It correctly identifies the assigned overload risk level in 87.52% of the samples. The most frequent error in the test set, occurring in 10.46% of cases, involves assigning a risk level to each sample adjacent to the target one. The analysis of the results highlights the advantages and drawbacks of criticality discretization and opens new solutions to improve the SOM potential in this field.

This paper introduces a disruption predictor constructed through a fully unsupervised two-dimensional mapping of the high-dimensional JET operational space. The primary strength of this disruption predictor lies in its inherent self-organization capability. Diverging from both supervised disruption predictors and earlier approaches suggested by the same authors, which were based on unsupervised models such as Self-Organizing or Generative Topographic Maps, this predictor eliminates the need for labeling data of disruption terminated pulses during training. In prior methods, labels were indeed required post-mapping to inform the model about the presence or absence of disruption precursors at each time instant during the disrupted discharges. In contrast, our approach in this study involves no labeling of data from disruption-terminated experiments. The Self-Organizing Map, operating without any a priori information, adeptly identifies the regions characterizing the pre-disruptive phase. Moreover, SOM discovers non-trivial relationships and captures the complicated interplay of device diagnostics on the internal plasma states from the experimental data. The provided model is highly interpretable; it allows the visualization of high-dimensional data and facilitates easy interrogation of the model to understand the reasons behind its correlations. Hence, utilizing SOMs across various devices can prove invaluable in extracting rules and identifying common patterns, thereby facilitating extrapolation to ITER of the knowledge acquired from existing tokamaks.

Reliable disruption prediction (DP) and disruption mitigation systems are considered unavoidable during international thermonuclear experimental reactor (ITER) operations and in the view of the next fusion reactors such as the DEMOnstration Power Plant (DEMO) and China Fusion Engineering Test Reactor (CFETR). In the last two decades, a great number of DP systems have been developed using data-driven methods. The performance of the DP models has been improved over the years both for a more appropriate choice of diagnostics and input features and for the availability of increasingly powerful data-driven modelling techniques. However, a direct comparison among the proposals has not yet been conducted. Such a comparison is mandatory, at least for the same device, to learn lessons from all these efforts and finally choose the best set of diagnostic signals and the best modelling approach. A first effort towards this goal is made in this paper, where different DP models will be compared using the same performance indices and the same device. In particular, the performance of a conventional Multilayer Perceptron Neural Network (MLP-NN) model is compared with those of two more sophisticated models, based on Generative Topographic Mapping (GTM) and Convolutional Neural Networks (CNN), on the same real time diagnostic signals from several experiments at the JET tokamak. The most common performance indices have been used to compare the different DP models and the results are deeply discussed. The comparison confirms the soundness of all the investigated machine learning approaches and the chosen diagnostics, enables us to highlight the pros and cons of each model, and helps to consciously choose the approach that best matches with the plasma protection needs.

In view of the future high power nuclear fusion experiments, the early identification of disruptions is a mandatory requirement, and presently the main goal is moving from the disruption mitigation to disruption avoidance and control. In this work, a deep-convolutional neural network (CNN) is proposed to provide early detection of disruptive events at JET. The CNN ability to learn relevant features, avoiding hand-engineered feature extraction, has been exploited to extract the spatiotemporal information from 1D plasma profiles. The model is trained with regularly terminated discharges and automatically selected disruptive phase of disruptions, coming from the recent ITER-like-wall experiments. The prediction performance is evaluated using a set of discharges representative of different operating scenarios, and an in-depth analysis is made to evaluate the performance evolution with respect to the considered experimental conditions. Finally, as real-time triggers and termination schemes are being developed at JET, the proposed model has been tested on a set of recent experiments dedicated to plasma termination for disruption avoidance and mitigation. The CNN model demonstrates very high performance, and the exploitation of 1D plasma profiles as model input allows us to understand the underlying physical phenomena behind the predictor decision.

Wendelstein 7-X (W7-X) is the leading experiment on the path of demonstrating that stellarators are a feasible concept for a future power plant. One of its major goals is to prove quasi-steady-state operation in a reactor-relevant parameter regime. The surveillance and protection of the water-cooled plasma-facing components (PFCs) against overheating is fundamental to guarantee a safe steady-state high-heat-flux operation. The system has to detect thermal events in real-time and timely interrupt operation if it detects a critical event. The fast reaction times required to prevent damage to the device make it imperative to automate fully the image analysis algorithms. During the past operational phases, W7-X was equipped with inertially cooled test divertor units and the system still required manual supervision. With the experience gained, we have designed a new real-time PFC protection system based on image processing techniques. It uses a precise registration of the entire field of view against the CAD model to determine the temperature limits and thermal properties of the different PFCs. Instead of reacting when the temperature limits are breached in certain regions of interest, the system predicts when an overload will occur based on a heat flux estimation, triggering the interlock system in advance to compensate for the system delay. To conclude, we present our research roadmap towards a feedback control system of thermal loads to prevent unnecessary plasma interruptions in long high-performance plasmas.

In tokamak nuclear fusion reactors, one of the main issues is to know the total emission of radiation, which is mandatory to understand the plasma physics and is very useful to monitor and control the plasma evolution. This radiation can be measured by means of a bolometer system that consists in a certain number of elements sensitive to the integral of the radiation along straight lines crossing the plasma. By placing the sensors in such a way to have families of crossing lines, sophisticated tomographic inversion algorithms allow to reconstruct the radiation tomography in the 2D poloidal cross-section of the plasma. In tokamaks, the number of projection cameras is often quite limited resulting in an inversion mathematic problem very ill conditioned so that, usually, it is solved by means of a grid-based, iterative constrained optimization procedure, whose convergence time is not suitable for the real time requirements. In this paper, to illustrate the method, an assumption not valid in general is made on the correlation among the grid elements, based on the statistical distribution of the radiation emissivity over a set of tomographic reconstructions, performed off-line. Then, a regularization procedure is carried out, which merge highly correlated grid elements providing a squared coefficients matrix with an enough low condition number. This matrix, which is inverted offline once for all, can be multiplied by the actual bolometer measures returning the tomographic reconstruction, with calculations suitable for real time application. The proposed algorithm is applied, in this paper, to a synthetic case study.