Particle colliders, such as the Large Hadron Collider (LHC) at the European Organization for Nuclear Research (CERN), have been the main drivers of discoveries in particle physics for several decades. They are highly complex dynamical systems with an immense number of control parameters and interdependent subsystems that need to be carefully adjusted to maximize the overall collider performance. This study employs Big Data and Machine Learning techniques to create a surrogate model of the particle beam physics of the LHC using the vast amounts of data acquired during machine operation, and explores whether and how the model can be employed to infer the optimal parameter settings to minimize the beam losses. The first results look promising which indicates that Machine Learning techniques are indeed valuable tools also for the particle accelerator physics field.

Download the slides for this talk.Download ( PDF, 3676.11 MB)