Assessment of tomography signals in view of neural network reconstruction at ISTTOK

Abstract

The detector measurements that serve as a basis for the tomographic reconstruction of plasma radiation profiles can be affected by a number of issues. In this work, we are especially concerned about reflections inside the vessel that appear as direct incidences, when in fact those contributions come from points that are outside the line of sight of a detector. This problem occurs in devices such as ISTTOK, where the surface of the inner vessel acts as an optical system capable of reflecting light. To assess the contribution of those reflections to the detector signals, we used an experimental setup comprising a cylindrical gas-discharge lamp placed at several radial and angular positions inside the vessel. As detectors, we used two photodiode arrays, one placed on the equatorial plane and the other at the top of the vessel. At the same time, we modelled this experimental setup in a ray-tracing simulation software so that, once the experimental data agrees with the simulation data, we can use this simulated environment to study the effect of reflections in realistic plasma profiles. When properly calibrated, such simulated environment can be used to generate large volumes of high-quality data to train a neural network for plasma tomography.