The Joint Accelerator Conferences Website (JACoW) is an international collaboration that publishes the proceedings of accelerator conferences held around the world.
@inproceedings{schirmer:ipac2021-wepab303,
author = {D. Schirmer},
title = {{Machine Learning Applied to Automated Tunes Control at the 1.5 GeV Synchrotron Light Source DELTA}},
booktitle = {Proc. IPAC'21},
pages = {3379--3382},
eid = {WEPAB303},
language = {english},
keywords = {storage-ring, quadrupole, simulation, controls, operation},
venue = {Campinas, SP, Brazil},
series = {International Particle Accelerator Conference},
number = {12},
publisher = {JACoW Publishing, Geneva, Switzerland},
month = {08},
year = {2021},
issn = {2673-5490},
isbn = {978-3-95450-214-1},
doi = {10.18429/JACoW-IPAC2021-WEPAB303},
url = {https://jacow.org/ipac2021/papers/wepab303.pdf},
note = {https://doi.org/10.18429/JACoW-IPAC2021-WEPAB303},
abstract = {{Machine learning (ML) driven algorithms are finding more and more use cases in the domain of accelerator physics. Apart from correlation analysis in large data volumes, low and high level controls, like beam orbit correction, also non-linear feedback systems are possible application fields. This also includes monitoring the storage ring betatron tunes, as an important task for stable machine operation. For this purpose classical, shallow (non-deep), feed-forward neural networks (NNs) were investigated for automated adjusting the storage ring tunes. The NNs were trained with experimental machine data as well as with simulated data based on a lattice model of the DELTA storage ring. With both data sources comparable tune correction accuracies were achieved, both, in real machine operation and for the simulated storage ring model. In contrast to conventional PID methods, the trained NNs were able to approach the desired target tunes in fewer steps. The report summarizes the current status of this machine learning project and points out possible future improvements as well as other possible applications.}},
}