A holistic approach to accelerator reliability modeling

Reliability has been identified as a key factor limiting the development of certain particle accelerator applications, for example Accelerator-Driven Systems (ADS) for energy production and waste-transmutation. Previous studies of particle accelerator reliability have been undertaken using conventional techniques, such as Reliability Block Diagrams (RBD), Fault Tree Analysis (FTA), etc. Although limited data surrounding components and their failure modes limits the applicability of conventional techniques for analysing the reliability of particle accelerators. In addition industrial applications of particle accelerators, i.e. energy production, require a real time response to failure. In this paper we examine a holistic approach to accelerator reliability modelling using Electric Network Frequency (ENF) criterion to look for emergent behaviour of the particle accelerator, from complex datasets, such as beam current/charge, created by the diagnostics systems during the machines operation. To look for predictive characteristics just prior to a machine trip.