Deteção online embarcada de sinais raros com baixa relação sinal-ruído em ambientes com alta taxa eventos

Abstract

The ATLAS is one of the main experiments of the powerful particle accelerator ever built, the LHC. Since its conception, this experiment provides great potential for the study of a vast physics program and has already been protagonist of notable discoveries, worthy of a Nobel Prize, such as the observation of the Higgs boson. Currently, the LHC is undergoing an upgrade program that aims at preparing it to operate with a greater amount of energy. This thesis was developed in the ATLAS experiment update program context and presents proposals to improve one of the main challenges found in ATLAS: the detection of signals that, in addition to being corrupted by noise, are immersed in ordinary information. To achieve high eficiency in detecting signals of interest, the experiment employs an online filtering system. Studies hat consider the LHC updates have shown that the available bandwidth, for the filtering system, would be exceeded due to an increase in the background noise. For the purpose of get around this problem, a system based on the information fusion from the ATLAS calorimetry system and the ATLAS muon detector was incorporated in the ATLAS online filtering system, in order to improve the rejection of information that is not of interest to LHC physics. This system uses an algorithm based on a matched filtering approach to detect the signal of interest immersed in noise. Although this algorithm provides good results, studies have shown that it can be improved. Thus, this work aims at studying algorithms to improve the system efficiency. As a result, it was achieved a reduction of 20% in the false alarm rate provided by the algorithm currently implemented in the system, while keeping 98% of the correct classification for the filtering system.