Operational data for fault prognosis in particle accelerators with machine learning

• DOI: 10.1016/j.dib.2023.109658
• Publication Date: 2023-10-11T22:24:57Z
• Authors: Majdi I. Radaideh; Chris Pappas; Mark Wezensky; Sarah Cousineau
• Keywords: Spallation Neutron Source

This paper presents real operational data collected from the power systems of Spallation Neutron Source facility, which provides most intense neutron beam in world. The authors have used a radio-frequency test facility (RFTF) and simulated system failures lab without causing catastrophic failure. Waveform signals been RFTF normal operation as well during fault induction efforts. dataset significant amount faulty for training statistical or machine learning models. Then, performed 21 experiments, where faults are slowly induced into purpose testing models prognosis to detect prevent impending faults. experiments include interesting combinations magnetic flux compensation start pulse width adjustments, cause gradual deterioration waveforms (e.g., output voltage, current, insulated-gate bipolar transistor currents, fluxes), mimic scenarios. Accordingly, this can be valuable developing predict scenarios general particle accelerators specific. All occurred Oak Ridge National Laboratory Ridge, Tennessee United States July 2022.