Diogo R. Ferreira

Diogo R. Ferreira

This document describes an approach used in the Multi-Machine Disruption Prediction Challenge for Fusion Energy by ITU, a data science competition which ran from September to November 2023, on the online platform Zindi. The competition involved data from three fusion devices - C-Mod, HL-2A, and J-TEXT - with most of the training data coming from the last two, and the test data coming from the first one. Each device has multiple diagnostics and signals, and it turns out that a critical issue in this competition was to identify which signals, and especially which features from those signals, were most relevant to achieve accurate predictions. The approach described here is based on extracting features from signals, and then applying logistic regression on top of those features. Each signal is treated as a separate predictor and, in the end, a combination of such predictors achieved the first place on the leaderboard.

Diogo R. Ferreira, Pedro J. Carvalho, Carlo Sozzi et al.

The JET baseline scenario is being developed to achieve high fusion performance and sustained fusion power. However, with higher plasma current and higher input power, an increase in pulse disruptivity is being observed. Although there is a wide range of possible disruption causes, the present disruptions seem to be closely related to radiative phenomena such as impurity accumulation, core radiation, and radiative collapse. In this work, we focus on bolometer tomography to reconstruct the plasma radiation profile and, on top of it, we apply anomaly detection to identify the radiation patterns that precede major disruptions. The approach makes extensive use of machine learning. First, we train a surrogate model for plasma tomography based on matrix multiplication, which provides a fast method to compute the plasma radiation profiles across the full extent of any given pulse. Then, we train a variational autoencoder to reproduce the radiation profiles by encoding them into a latent distribution and subsequently decoding them. As an anomaly detector, the variational autoencoder struggles to reproduce unusual behaviors, which includes not only the actual disruptions but their precursors as well. These precursors are identified based on an analysis of the anomaly score across all baseline pulses in two recent campaigns at JET.

Diogo R. Ferreira, Pedro J. Carvalho, Horácio Fernandes

The use of deep learning is facilitating a wide range of data processing tasks in many areas. The analysis of fusion data is no exception, since there is a need to process large amounts of data collected from the diagnostic systems attached to a fusion device. Fusion data involves images and time series, and are a natural candidate for the use of convolutional and recurrent neural networks. In this work, we describe how CNNs can be used to reconstruct the plasma radiation profile, and we discuss the potential of using RNNs for disruption prediction based on the same input data. Both approaches have been applied at JET using data from a multi-channel diagnostic system. Similar approaches can be applied to other fusion devices and diagnostics.

Diogo R. Ferreira

Nuclear fusion is the process that powers the sun, and it is one of the best hopes to achieve a virtually unlimited energy source for the future of humanity. However, reproducing sustainable nuclear fusion reactions here on Earth is a tremendous scientific and technical challenge. Special devices -- called tokamaks -- have been built around the world, with JET (Joint European Torus, in the UK) being the largest tokamak currently in operation. Such devices confine matter and heat it up to extremely high temperatures, creating a plasma where fusion reactions begin to occur. JET has over one hundred diagnostic systems to monitor what happens inside the plasma, and each 30-second experiment (or pulse) generates about 50 GB of data. In this work, we show how convolutional neural networks (CNNs) can be used to reconstruct the 2D plasma profile inside the device based on data coming from those diagnostics. We also discuss how recurrent neural networks (RNNs) can be used to predict plasma disruptions, which are one of the major problems affecting tokamaks today. Training of such networks is done on NVIDIA GPUs.

Diogo R. Ferreira, Pedro J. Carvalho, Horácio Fernandes

Plasma tomography consists in reconstructing the 2D radiation profile in a poloidal cross-section of a fusion device, based on line-integrated measurements along several lines of sight. The reconstruction process is computationally intensive and, in practice, only a few reconstructions are usually computed per pulse. In this work, we trained a deep neural network based on a large collection of sample tomograms that have been produced at JET over several years. Once trained, the network is able to reproduce those results with high accuracy. More importantly, it can compute all the tomographic reconstructions for a given pulse in just a few seconds. This makes it possible to visualize several phenomena -- such as plasma heating, disruptions and impurity transport -- over the course of a discharge.

Francisco A. Matos, Diogo R. Ferreira, Pedro J. Carvalho et al.

Deep learning is having a profound impact in many fields, especially those that involve some form of image processing. Deep neural networks excel in turning an input image into a set of high-level features. On the other hand, tomography deals with the inverse problem of recreating an image from a number of projections. In plasma diagnostics, tomography aims at reconstructing the cross-section of the plasma from radiation measurements. This reconstruction can be computed with neural networks. However, previous attempts have focused on learning a parametric model of the plasma profile. In this work, we use a deep neural network to produce a full, pixel-by-pixel reconstruction of the plasma profile. For this purpose, we use the overview bolometer system at JET, and we introduce an up-convolutional network that has been trained and tested on a large set of sample tomograms. We show that this network is able to reproduce existing reconstructions with a high level of accuracy, as measured by several metrics.