Modelization of an injector with machine learning

Angélique Gatera; Mathieu Debongnie; Frédéric Bouly; Maud Baylac; Nicolas Chauvin; Didier Uriot

doi:10.18429/JACoW-IPAC2019-WEPTS006

Modelization of an injector with machine learning

Angélique Gatera, Mathieu Debongnie, Frédéric Bouly, Maud Baylac, Nicolas Chauvin, Didier Uriot

Research output › peer-review

Abstract

Modern particle accelerator projects, such as the accelerator for the Multi-purpose hYbrid Research Reactor for High-tech Application (MYRRHA) project driven by the SCK*CEN in Belgium, have very high stability and/or reliability requirements. This means that new strategies for the control systems have to be developed. For that, having faster beam dynamics simulation could prove to be helpful. In this paper, we report the training of neural networks to model key properties of the beam in the MYRRHA injector as well as in IPHI (“Injecteur de Proton à Haute Intensité”). The trained models are shown to be able to reproduce the general behaviours of the machines while requiring a very low computation time.

Original language	English
Title of host publication	10th International Particle Accelerator Conference
Subtitle of host publication	IPAC2019, Melbourne, Australia
Publisher	JACoW Publishing
Pages	3096-3099
Number of pages	4
ISBN (Electronic)	978-3-95450-208-0
DOIs	https://doi.org/10.18429/JACoW-IPAC2019-WEPTS006
State	Published - 23 May 2019
Event	2019 - IPAC: 10th International Particle Accelerator Conference - Melbourne convention & exhibition centre, Melbourne Duration: 19 May 2019 → 24 May 2019 https://ipac19.org/ https://ipac19.org

Conference

Conference	2019 - IPAC
Abbreviated title	IPAC2019
Country/Territory	Australia
City	Melbourne
Period	2019-05-19 → 2019-05-24
Internet address	https://ipac19.org/ https://ipac19.org

Access to Document

10.18429/JACoW-IPAC2019-WEPTS006License: Unspecified

http://accelconf.web.cern.ch/ipac2019/papers/wepts006.pdfLicense: Unspecified

Cite this

@inproceedings{0ac3cda2c9d446b78a14bd772d83c91e,

title = "Modelization of an injector with machine learning",

abstract = "Modern particle accelerator projects, such as the accelerator for the Multi-purpose hYbrid Research Reactor for High-tech Application (MYRRHA) project driven by the SCK*CEN in Belgium, have very high stability and/or reliability requirements. This means that new strategies for the control systems have to be developed. For that, having faster beam dynamics simulation could prove to be helpful. In this paper, we report the training of neural networks to model key properties of the beam in the MYRRHA injector as well as in IPHI (“Injecteur de Proton {\`a} Haute Intensit{\'e}”). The trained models are shown to be able to reproduce the general behaviours of the machines while requiring a very low computation time.",

keywords = "Machine Learning, neural network, MINERVA, injector transmission",

author = "Ang{\'e}lique Gatera and Mathieu Debongnie and Fr{\'e}d{\'e}ric Bouly and Maud Baylac and Nicolas Chauvin and Didier Uriot",

note = "Score=3; 2019 - IPAC : 10th International Particle Accelerator Conference, IPAC2019 ; Conference date: 19-05-2019 Through 24-05-2019",

year = "2019",

month = may,

day = "23",

doi = "10.18429/JACoW-IPAC2019-WEPTS006",

language = "English",

pages = "3096--3099",

booktitle = "10th International Particle Accelerator Conference",

publisher = "JACoW Publishing",

url = "https://ipac19.org/, https://ipac19.org",

}

TY - GEN

T1 - Modelization of an injector with machine learning

AU - Gatera, Angélique

AU - Debongnie, Mathieu

AU - Bouly, Frédéric

AU - Baylac, Maud

AU - Chauvin, Nicolas

AU - Uriot, Didier

N1 - Score=3

PY - 2019/5/23

Y1 - 2019/5/23

N2 - Modern particle accelerator projects, such as the accelerator for the Multi-purpose hYbrid Research Reactor for High-tech Application (MYRRHA) project driven by the SCK*CEN in Belgium, have very high stability and/or reliability requirements. This means that new strategies for the control systems have to be developed. For that, having faster beam dynamics simulation could prove to be helpful. In this paper, we report the training of neural networks to model key properties of the beam in the MYRRHA injector as well as in IPHI (“Injecteur de Proton à Haute Intensité”). The trained models are shown to be able to reproduce the general behaviours of the machines while requiring a very low computation time.

AB - Modern particle accelerator projects, such as the accelerator for the Multi-purpose hYbrid Research Reactor for High-tech Application (MYRRHA) project driven by the SCK*CEN in Belgium, have very high stability and/or reliability requirements. This means that new strategies for the control systems have to be developed. For that, having faster beam dynamics simulation could prove to be helpful. In this paper, we report the training of neural networks to model key properties of the beam in the MYRRHA injector as well as in IPHI (“Injecteur de Proton à Haute Intensité”). The trained models are shown to be able to reproduce the general behaviours of the machines while requiring a very low computation time.

KW - Machine Learning

KW - neural network

KW - MINERVA

KW - injector transmission

UR - https://ecm.sckcen.be/OTCS/llisapi.dll/overview/39469631

U2 - 10.18429/JACoW-IPAC2019-WEPTS006

DO - 10.18429/JACoW-IPAC2019-WEPTS006

M3 - In-proceedings paper

SP - 3096

EP - 3099

BT - 10th International Particle Accelerator Conference

PB - JACoW Publishing

T2 - 2019 - IPAC

Y2 - 19 May 2019 through 24 May 2019

ER -