Development of inverse methods for infrared thermography in fusion devices

Abstract

Infrared (IR) thermography system is a key diagnostic in fusion devices to monitor the Plasma Facing Components. Nevertheless, both qualitative and quantitative analysis (i.e. hot spot detection and surface temperature measurement) are challenging due to the presence of disturbance phenomena like variable emissivity and multiple reflections in fully metallic environment. Through the comparison with the experimental IR measurements, simulation is an essential tool for anticipating, quantifying and analysing the effects of the various errors involved in the interpretation of IR images. This paper goes a step further for achieving IR quantitative thermography in developing inverse methods to retrieve the real surface temperature, by taking into account variable emissivity and filtering reflections. Two approaches are studied: (1) using gradients methods through a reduced photonic model (2) using machine learning techniques based on simulated dataset. Applied on WEST-like tokamak numerical prototype, the temperatures are estimated, with these two approaches, with an accuracy better than 6%, which is a clear improvement compared to usual methods (i.e. assuming blackbody object).

Keywords

• Thermography
• Tokamak
• Inverse methods
• Machine learning
• Gradient minimization
• Temperature