Machine learning for early fault detection in accelerator systems

With the development of systems based on a combination of mechanics, electronics and – more and more - software components, increasing system complexity is a de facto trend in the engineering world. Particle accelerators make no exception to this paradigm. The continuous push for higher energies driven by particle physics implies that next generation machines will be at least one order of magnitude larger and more complex than present ones, posing unprecedented challenges in terms of beam performance and availability. The two most promising approaches CERN discusses as next generation projects are the Future Circular Collider (FCC) and the Compact Linear Collider (CLIC), with a size of 100 km and 48 km, respectively (see Fig.1 and Fig. 2)