Development of inverse methods for infrared thermography in fusion devices

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