Predicting the Future of the CMS Detector: Crystal Radiation Damage and Machine Learning at the LHC
This addresses a critical monitoring need for the CMS experiment at CERN, but it is incremental as it applies an existing method (LSTM) to new data.
The paper tackles the problem of predicting optical transparency degradation in CMS detector crystals due to radiation, releasing a public dataset (2016-2018) and demonstrating a solution using an LSTM neural network for short- and long-term predictions.
The 75,848 lead tungstate crystals in CMS experiment at the CERN Large Hadron Collider are used to measure the energy of electrons and photons produced in the proton-proton collisions. The optical transparency of the crystals degrades slowly with radiation dose due to the beam-beam collisions. The transparency of each crystal is monitored with a laser monitoring system that tracks changes in the optical properties of the crystals due to radiation from the collision products. Predicting the optical transparency of the crystals, both in the short-term and in the long-term, is a critical task for the CMS experiment. We describe here the public data release, following FAIR principles, of the crystal monitoring data collected by the CMS Collaboration between 2016 and 2018. Besides describing the dataset and its access, the problems that can be addressed with it are described, as well as an example solution based on a Long Short-Term Memory neural network developed to predict future behavior of the crystals.