Machine learning approach to CMS RPC HV scan data analysis

Abstract

Resistive Plate Chambers (RPC) are gaseous detectors in the muon system of the Compact Muon Solenoid (CMS) experiment at the European Laboratory for Particle Physics, CERN. The RPC high-voltage scan is a crucial sequence of calibration runs typically conducted at the onset of each data-taking year with the initial collisions of the CERN Large Hadron Collider (LHC) at nominal luminosity in proton–proton collisions 2×1034cm−2s−1, ensuring RPC proper functioning by establishing correct working points. This study applies machine learning algorithms to automate and accelerate previously manual, time-consuming analysis, enhancing efficiency and decision-making. We developed an autoencoder artificial neural network (ANN) in Fourier space (FSAC) to approximate efficiency curves, which are then used to determine working points. This new approach reduces the time for data analysis from over three months to less than a week.