Research on Utilizing a Multivariate Feedback Algorithm to Maintain Stable Operation of Variable Energy Electron Accelerators

Measurements performed utilizing particle accelerators rely on the stability and reproducibility of the accelerator operation, most notably for beam flux (beam current) and beam energy. This can be vital when the measurements rely on previous calibrations performed to establish a standard on which to base the end result of the measurement. The authors have designed a monitoring and control virtual instrument (VI) based on National Instruments LabVIEWTM which calls external modeling code that is written in a standard programming language. The LabVIEWTM virtual instrument allows the user to monitor multiple accelerator parameters while permitting user control of the most important four of these parameters which determine machine operation. The external modeling code, the Advanced Data Validation and Verification System (ADVVS), incorporates as a kernel the Universal Process Model (UPM) software from Triant Technologies, Inc. The kernel executes the accelerator modeling function based on previously acquired history data of the relevant parameters for accelerator operation for conditions of interest. This history data is stored as a reference set prior to subsequent operation. During accelerator operation, 11 accelerator parameters are input to ADVVS from the LabVIEWTM virtual instrument. The ADVVS uses the reference files to predict the eleven parameter values that would normally be expected for optimal accelerator operation. The set of 11 model values is the nearest model state. This multivariate modeling approach presented two principal advantages: 1. The ADVVS allowed rapid detection of anomalous accelerator behavior variant from normal accelerator behavior recorded in the reference set of the kernel. 2. The ADVVS presented a methodology to incrementally tune the accelerator back to optimal operation, thus maintaining highly stable and reproducible operation. In the future, this methodology may also be utilized to tune the accelerator initially or move efficiently between separate accelerator tune states. Note that a user-selected setting of the eleven accelerator parameters is termed a tune. (authors)