Diagnostics and Neural Network Control Systems for Linear Accelerators

摘要

Development and validation of robust control systems for linear accelerators is particularly important in the context of free-electron lasers (FELs) due to the high sensitivity of the resultant light quality to perturbations in the electron beam parameters. Recently, there has been growing interest in applying advances in the field of artificial intelligence to the development of a beam-based control system for linear accelerators. The proposed control system would be capable of adapting to and compensating for changes in beam parameters in real time during machine operation, making use of no prior knowledge other than the constraints on the actuators. This work poses a number of interesting challenges and will require efforts in improvement of beam dynamics models, adaptation of artificial intelligence techniques to a wide range of operational regimes, development and implementation of advanced diagnostics, and extensive testing in an operational accelerator environment. Preliminary work toward the development of such a control system will be presented.