Classification of LHC beam loss spikes using support vector machines

Abstract

The CERN Large Hadron Collider's (LHC) collimation system is the most complex beam cleaning system ever designed. It requires frequent setups to determine the beam centres and beam sizes at the 86 collimator positions. A collimator jaw is aligned to the beam halo when a clear beam loss spike is detected on a Beam Loss Monitor (BLM) downstream of the collimator. This paper presents a technique for identifying such clear loss spikes with the aid of Support Vector Machines. The training data was gathered from setups held during the first three months of the 2011 LHC run, and the model was tested with data from a machine development period.