Study of a Method for Muon Detection in the ATLAS Experiment Based on Convolutional Neural Network

Abstract

The development of new digital signal processing techniques is crucial in high-energy physics experiments. In this context, the ATLAS experiment upgrade for Phase-II of the high-luminosity LHC, particle detection system of the biggest hadrons collider of the world, will allow experiments with higher event rates, leading to the signal pile-up effect. The ATLAS hadronic calorimeter (TileCal) can be used to help the detection of muon particles resulted from proton-proton collisions. Classical techniques for signal reconstruction in this new phase suffer from decreased performance, requiring new strategies that has better results in this new environment, and the least square(LS) method has already been explored as an alternative. In this paper, a Convolutional Neural Network (CN) is proposed for simulated muon detection system using selected cells from TileCal, and its performance is compared with detection using LS. The CN was able to reduce the false alarm rate with 98% of detection probability at the two cells studied compared to LS, reaching reductions up to 84% in a specific module.