Machine learning applied at the LHC for beam loss pattern classification

Abstract

Beam losses at the LHC are constantly monitored because they can heavily impact the performance of the machine. One of the highest risks is to quench the LHC superconducting magnets in the presence of losses leading to a long machine downtime to recover cryogenic conditions. Smaller losses are more likely to occur and have an impact on the machine performance, reducing the luminosity production or reducing the lifetime of accelerator systems due to radiation effects, such as magnets. Understanding the characteristics of the beam loss, such as the beam and the plane, is crucial to correct them. Regularly during the year, dedicated loss map measurements are performed to validate the beam halo cleaning of the collimation system. These loss maps have the particular advantage that they are performed in well controlled conditions and can therefore be used by a machine learning algorithm to classify the type of losses during the LHC machine cycle. This study shows the result of the beam loss classification and its retrospective application to beam loss data from the 2017 run.