|
Guest >> Sign In |
|
|
DIRECTED
ENERGY
PROFESSIONAL
SOCIETY
Abstract: 25-Symp-002
|
|
|
|
UNCLASSIFIED, PUBLIC RELEASE
Improving Directed Energy Diagnostic Systems with Machine Learning
RadiaSoft has been developing machine learning tools for diagnostics and anomaly detection for compact accelerators in collaboration with multiple national laboratories and industry. Our work includes the adaptation of inverse models for failure point detection as well as implicit detection of hidden variable anomalies using both unsupervised and supervised learning techniques. We also demonstrate the use of machine learning for signal conditioning and noise elimination in RF systems and the associated improvement to control stability using an RF simulator. This talk provides an overview of our techniques and recent successes in their use for laboratory and industrial accelerators with an eye toward generalization to a wide range of directed energy technologies.
UNCLASSIFIED, PUBLIC RELEASE
|