Evaluation of negative ion beam heat flux profile on STRIKE by neural networks

STRIKE (Short Time Retractable Instrumented Kalorimeter Experiment) is an instrumented calorimeter designed to study the SPIDER (Source for Production of Ion of Deuterium Extracted from Radio frequency plasma) negative ion beam. STRIKE is made of 16 one-dimensional Carbon Fiber Composite (CFC) tiles that intercept the whole beam. Since the frontal observation is disturbed, the temperature measures are recorded on the rear side by a set of two infrared cameras. The ultimate goal is the implementation of a method to analyze the thermal images by reconstructing the features of the energy flux associated with the SPIDER beamlets in real-time. The object of this thesis is to propose and compare two different methods based on the use of neural networks in order to evaluate the inverse heat flux. The first method consists of the reconstruction of the whole heat flux image. The second one aims to find only the characteristic parameters, such as the amplitude and the position of the centroid of the peaks and their Half-Width at Half Maximum. While the first method has proven to be expensive in terms of computational power and its results are far from acceptable, the second one presents really promising results while also being more efficient. The comparison between the second method results and the data obtained by a calorimetric analysis of the problem has shown the reliability of this method.