Deep learning for plasma tomography using the bolometer system at JET

• Published in: Fusion engineering and design 2017-01, Vol.114, p.18-25
• Main Authors: Matos, Francisco A., Ferreira, Diogo R., Carvalho, Pedro J.
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Artificial neural networks; Bolometers; Computed tomography; Deep learning; Image processing; Inverse problems; Levels; Neural networks; Pixels; Plasma diagnostics; Plasma physics; Radiation; Reconstruction; Tomography
• ISSN: 0920-3796; 1873-7196; 1873-7196
• DOI: 10.1016/j.fusengdes.2016.11.006

•Plasma tomography is able to reconstruct the plasma profile from radiation measurements along several lines of sight.•The reconstruction can be performed with neural networks, but previous work focused on learning a parametric model.•Deep learning can be used to reconstruct the full 2D plasma profile with the same resolution as existing tomograms.•We introduce a deep neural network to generate an image from 1D projection data based on a series of up-convolutions.•After training on JET data, the network provides accurate reconstructions with an average pixel error as low as 2%. 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.