Generation of Unusual Plasma Discharge Video by Generative Adversarial Network

Abstract

In nuclear fusion experiments in large helical device (LHD), a lot of videos containing the images of plasma discharge are recorded. An observation of the recorded images of plasma light emission can lead to a new discovery or help to optimize the operational parameters for the experiment. An unusual plasma discharge, which may cause damage to the device, is expected to be foreseen through a prediction method. Due to the shortage of videos having such unusual emissions, the generation of more videos having similar phenomenon is required. However, video generation is a very challenging issue as the videos should have not only similarity in features in the real one but also a plausibility in frame-by-frame transition, especially in the case of plasma discharges. Thus, this paper proposes a method to generate a video containing plasma light emission using generative adversarial network (GAN). It has been confirmed that the proposed generative model can produce a new video having plasma light emission with a very smooth frame transition.