19–24 May 2024
Music City Center
US/Central timezone

LSTMs for anomaly detection in the magnet power supply temperatures of APS-U

TUPS51
21 May 2024, 16:00
2h
Blues (MCC Exhibit Hall A)

Blues

MCC Exhibit Hall A

Poster Presentation MC6.D13 Machine Learning Tuesday Poster Session

Speaker

Ihar Lobach (Argonne National Laboratory)

Description

We present an approach for detection of anomalies in the temperatures of magnet power supplies (PSs) in storage rings. We train a Long Short-Term Memory (LSTM) neural network to predict the temperatures of several components of a PS (heatsinks, capacitors, resistors) based on the PS current, PS voltage, and room temperature. An anomaly is detected when the observed PS temperature starts to deviate significantly from the LSTM prediction. A dedicated test stand has been built with a PS and a PS controller of the same kind that will be used in the Advanced Photon Source Upgrade (APS-U). The PS was modified to be able to programmatically create artificial anomalies in the PS temperature, so that the proposed method can be tested. Additionally, we use this test stand to experiment with more advanced PS temperature monitoring techniques employing infrared cameras, which could be used for all APS-U PSs in the future.

Funding Agency

The work is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357.

Region represented North America

Primary author

Ihar Lobach (Argonne National Laboratory)

Co-authors

Jonathan Edelen (RadiaSoft LLC) Kathryn Wolfinger (RadiaSoft LLC) Michael Borland (Argonne National Laboratory)

Presentation materials

There are no materials yet.