LHC beam loss pattern recognition

One of the systems protecting CERN’s Large Hadron Collider (LHC) is the Beam Loss Monitoring system (BLM). More than 3600 monitors are installed around the ring. The beam losses are permanently integrated over 12 different time intervals (from 40 microseconds to 84 seconds). When any loss exceeds the thresholds defined for the integration window, the beam is removed from the machine. Understanding the origin of a beam loss is crucial for machine operation, as it can help to avoid a repetition of the same scenario. The signals read from given monitors can be considered as entries of a vector. This article presents how a loss map of unknown cause can be decomposed using vector based analysis derived from well-known loss scenarios. The algorithms achieving this decomposition are described, as well as the accuracy of the results.