Skip to main content
Photon Sciences

Applying Machine Learning to the Advanced Photon Source Accelerator Complex

Argonne researchers are applying artificial neural networks to improve accelerator performance for the APS Upgrade.

Argonne’s Advanced Photon Source is undergoing an $815 million upgrade project. As a part of this upgrade, the Advanced Photon Source accelerator complex will feature a new multi-bend achromat lattice design, with many more bending magnets and magnet-focusing cells than the present machine, resulting in brighter x-ray production. Stable operation of the facility is based on tens of thousands of settings and measurements taken at intervals from seconds to minutes. In the APS Upgrade era this will increase by multiple orders of magnitude. Argonne researchers are working on applying artificial neural networks trained with simulated data and over twenty years of historical data to predict anomalies and failure modes, and to maintain proper orbit motion and extend beam lifetime.