Machine Learning for Linac Improved Performance
Description
In Linacs at FNAL and J-PARC, the current emittance optimization procedure is limited to manual adjustments of a few parameters; using a larger number is not practically feasible for a human operator. Using machine learning (ML) techniques allows li
Detailed example
Outputs are proposed changes to RF system parameters (cavity phase settings and/or field gradients) to counter the effect of daily drift and to stabilize the output energy.
AI / analytics pattern
Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relationships.
Automation level / stage
b) Pilot – The use case has been deployed in a limited test or pilot capacity.
Expected benefit
Daily fluctuations in the Ion Source conditions as well af the effect of environmental changes to RF systems and cavities affect Linac beam. Results include increased beam loss resulting in increased beamline component irradiation, decreased beam intensity to downstream machines affecting Accelerator Complex deliverables, drifts in Linac beam energy directly affecting Booster losses. These drifts are not easily predictable since we do not have environmental control on the RF gallery, not enough instrumentaion in the Ion Source or Linac proper. To counter these effects, we are developing AI-based optimization and modeling, including Bayesian Optimization and surrogate model-based optimization, with the ultimate goal of (near) real-time RF compensation.
Audit / financial statement impact
The use case does not have an effect on civil rights/liberties/privacy, access to education/housing/insurance/credit/employment, access to critical government resources/services, human health/safety, critical infrastructure/public safety
Controls / human review
ATO: No; PIA: Not published
Data needed
Accelerator operations machine data as well as accelerator simulation