Speaker
Description
This project develops a machine learning–based system to prevent RF cavity trips in the free-electron laser by autonomously controlling the applied RF power. Sudden vacuum and current fluctuations within the cavities can cause reflections that trip the machine, and continuous manual monitoring throughout the conditioning process isn't feasible. To address this and potentially improve the conditioning efficiency, process-variable data was collected and analyzed to identify patterns in cavity behavior across operating power levels. A hybrid model combining clustering methods, linear regression, and a classifiers was designed to categorize current ranges, estimate baseline behavior, and detect anomalies. The resulting control program evaluates the machine state over short intervals, decreases power during unsafe conditions, increases it during prolonged stability, and can automatically reset the RF system after a trip. This approach enables faster and safer conditioning of the RF cavities, reduces operator workload, and provides a pathway toward fully autonomous preventive maintenance.
| In which format do you inted to submit your paper? | LaTeX |
|---|