RadiaSoft has been collaborating with multiple facilities on the development of new machine learning techniques for the improvement of accelerator operations. These efforts include improved signal processing for RF systems, automation of alignment for neutron beamlines, and the development of smart alarm systems for linear accelerators. This talk will provide an overview of RadiaSoft and the wide range of machine learning projects under our purview. We will then detail our efforts to apply ML for noise reduction in RF waveform data, the application of convolutional neural networks for sample identification in neutron beamlines, and our results in developing inverse models for anomaly detection in electron LINACs at two different facilities. For each topic we will provide high level context followed by an overview of our ML methods and our results.
The meeting was also streamed via Zoom