ICALEPCS2019 - Table of Session: WEMPR (Mini-Oral)

Paper	Title	Page
WEMPR001	Data Analysis Infrastructure for Diamond Light Source Macromolecular & Chemical Crystallography and Beyond	1031
WEPHA094	use link to see paper's listing under its alternate paper code
	M. Gerstel, A. Ashton, R.J. Gildea, K. Levik, G. Winter DLS, Oxfordshire, United Kingdom
	The Diamond Light Source data analysis infrastructure, Zocalo, is built on a messaging framework. Analysis tasks are processed by a scalable pool of workers running on cluster nodes. Results can be written to a common file system, sent to another worker for further downstream processing and/or streamed to a LIMS. Zocalo allows increased parallelization of computationally expensive tasks and makes the use of computational resources more efficient. The infrastructure is low-latency, fault-tolerant, and allows for highly dynamic data processing. Moving away from static workflows expressed in shell scripts we can easily re-trigger processing tasks in the event that an issue is found. It allows users to re-run tasks with additional input and ensures that automatically and manually triggered processing results are treated equally. Zocalo was originally conceived to cope with the additional demand on infrastructure by the introduction of Eiger detectors with up to 18 Mpixels and running at up to 560 Hz framerate on single crystal diffraction beamlines. We are now adapting Zocalo to manage processing tasks for ptychography, tomography, cryo-EM, and serial crystallography workloads.
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR001
About •	paper received ※ 30 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR003	Exploring Embedded Systems’ Dedicated Cores for Real-Time Applications	1036
WEPHA107	use link to see paper's listing under its alternate paper code
	P.H. Nallin, J.G.R.S. Franco, R.C. Ito, A.R.D. Rodrigues LNLS, Campinas, Brazil
	Developments and research in high technology leads to powerful and sophisticated machines which are highly important for many scientific fields. Considering real-time applications, however, these systems tend to become non-deterministic and users may find themselves inside a not completely controllable environment. Exploring open-hardware single board computers with a system-on-a-chip which usually runs an operational system on their main processor(s) and also have real-time units is a good alternative. These real-time units are designed as a microcontroller embedded on the chip where a firmware is loaded, runs concomitantly and exchanges data with the main system. As a result, it is possible to achieve performance increase, high temporal resolution and low latency and jitter, features that are widely desired for controls and critical data acquisition systems. This system architecture allows moving real-time data into high level servers, such as Redis (Remote Dictionary Server) and EPICS, easily. This paper introduces and shows uses of Beaglebone Black, an inexpensive single-board computer, its Programmable Real-Time Units (PRUs) and data sharing with Redis data structure.
	Poster WEMPR003 [6.128 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR003
About •	paper received ※ 30 September 2019 paper accepted ※ 18 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR004	Why Should You Invest in Asset Management? A Fire and Gas Use Case	1041
WEPHA109	use link to see paper's listing under its alternate paper code
	H. Nissen, S. Grau CERN, Geneva, Switzerland
	At present, the CERN Fire and Gas detection systems involve about 22500 sensors and their number is increasing rapidly at the same time as the number of equipped installations grows up. These assets cover a wide spectrum of technologies, manufacturers, models, parameters, and ages, reflecting the 60 years of CERN history. The use of strict rules and data structures in the declaration of the assets can make a big impact on the overall system maintainability and therefore on the global reliability of the installation. Organized assets data facilitates the creation of powerful reports that help asset owners and management address material obsolescence and end-of-life concerns with a global perspective Historically preventive maintenance have been used to assure the correct function of the installations. With modern supervision systems, a lot of data is collected and can be used to move from preventive maintenance towards data driven maintenance (predictive). Moreover it optimizes maintenance cost and increase system availability while maintaining reliability. A prerequisite of this move is a coherence on the assets defined in the asset management system and in the supervision system.
	Poster WEMPR004 [0.675 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR004
About •	paper received ※ 27 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR005	The Array Control and Data Acquisition System of the Cherenkov Telescope Array	1046
WEPHA117	use link to see paper's listing under its alternate paper code
	I. Oya, E. Antolini, M. Fuessling CTA, Heidelberg, Germany L. Baroncelli, A. Bulgarelli, V. Conforti, N. Parmiggiani INAF, Bologna, Italy J. Borkowski CAMK, Torun, Poland A. Carosi, J.N. Jacquemier, G. Maurin IN2P3-LAPP, Annecy-le-Vieux, France J. Colome CSIC-IEEC, Bellaterra, Spain C. Hoischen Universität Potsdam, Potsdam-Golm, Germany E. Lyard, R. Walter University of Geneva, Geneva, Switzerland D. Melkumyan, K. Mosshammer, I. Sadeh, T. Schmidt, P.A. Wegner DESY Zeuthen, Zeuthen, Germany U. Schwanke Humboldt University Berlin, Institut für Physik, Berlin, Germany J. Schwarz INAF-Osservatorio Astronomico di Brera, Merate, Italy G. Tosti Università degli di Perugia, Perugia, Italy
	The Cherenkov Telescope Array (CTA) project is the initiative to build the next-generation gamma-ray observatory. With more than 100 telescopes planned to be deployed in two sites, CTA is one of the largest astronomical facilities under construction. The Array Control and Data Acquisition (ACADA) system will be the central element of on-site CTA Observatory operations. The mission of the ACADA system is to manage and optimize the telescope array operations at each of the CTA sites. To that end, ACADA will provide all necessary means for the efficient execution of observations, and for the handling of the several Gb/s generated by each individual CTA telescope. The ACADA system will contain a real-time analysis pipeline, dedicated to the automatic generation of science alert candidates based on the inspection of data being acquired. These science alerts, together with external alerts arriving from other scientific installations, will permit ACADA to modify ongoing observations at sub-minute timescales in order to study high-impact scientific transient phenomena. This contribution describes the challenges, architecture, design principles, and development status of the ACADA system.
	Poster WEMPR005 [3.851 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR005
About •	paper received ※ 30 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR006	Application Development in the Face of Evolving Web Technologies at the National Ignition Facility	1052
WEPHA118	use link to see paper's listing under its alternate paper code
	E.R. Pernice, C.R. Albiston, R.G. Beeler, E.H. Chou, C.D. Fry, M. Shor, J.L. Spears, D.E. Speck, A.A. Thakur, S.L. West LLNL, Livermore, California, USA
	Funding: This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344 The past decade has seen great advances in web technology, making the browser the de-facto platform for many user applications. Advances in JavaScript, and innovations such as TypeScript, have enabled developers to build large scale applications for the web without sacrificing code maintainability. However, this rapid growth has also been accompanied by turbulence. AngularJS arrived and saw widespread adoption only to be supplanted by Angular 2+ a few years later; meanwhile other JavaScript-based languages and developer tools have proliferated. At the National Ignition Facility (NIF), the Shot Setup Tool (SST) is a large web-based tool for configuring experiments on the NIF that is being developed to replace legacy Java Swing application. We will present our experience in building SST during this turbulent time, including how we have leveraged TypeScript to greatly enhance code readability and maintainability in a multi-developer team, and our current effort to incrementally migrate from AngularJS to React.
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR006
About •	paper received ※ 30 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR008	Web Extensible Display Manager 2	1057
WEPHA146	use link to see paper's listing under its alternate paper code
	R.J. Slominski, T.L. Larrieu JLab, Newport News, Virginia, USA
	Funding: Authored by Jefferson Science Associates, LLC under U.S. DOE Contract No. DE-AC05-06OR23177 The Web Extensible Display Manager (WEDM) was first deployed at Jefferson Lab (JLab) in 2016 with the goal of rendering Extensible Display Manager (EDM) control screens on the web for the benefit of accessibility, and with version 2 our aim is to provide a more general purpose display toolkit by freeing ourselves from the constraints of the EDM dependency. Over the last few years WEDM has been extensively used at JLab for 24/7 information kiosks, on-call monitoring, and by remote users and staff. The software has also been deployed to Oak Ridge National Laboratory, and has become more robust as many bug fixes and contributions have been added. However, adoption and utility of the software as a general purpose control system display manager is limited by EDM, which is no longer actively maintained. A new toolkit can be built on modern frameworks, fully embrace web conventions and standards, and support multiple control system data sources. This new version is a result of a technology review and selection, and introduces a web inspired display file format, a web based display builder, new widgets, and a data interface intended to support pluggable data.
	Poster WEMPR008 [1.293 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR008
About •	paper received ※ 24 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR009	Development of Event Receiver on Zynq-7000 Evaluation Board	1063
WEPHA149	use link to see paper's listing under its alternate paper code
	H. Sugimura KEK, Ibaraki, Japan
	The timing system of SuperKEKB accelerator is used Event Timing System developed by Micro Research Finland. In this presentation, we tested the receiver on Zynq7000 evaluation board. The serialized event data are transferred from Event Generator to Event Receiver by using GTX transceiver. So, we selected Zynq7000(7z030) as receiver, because the FPGA has the GTX. And also, Zynq is mounted on arm processor, it is easily able to control received event data stream by using EPICS ICO. Finally we are aiming to combine event system and RF or BPM system in one FPGA board.
	Poster WEMPR009 [0.572 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR009
About •	paper received ※ 17 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR010	Anomaly Detection for CERN Beam Transfer Installations Using Machine Learning	1066
WEPHA155	use link to see paper's listing under its alternate paper code
	T. Dewitte, W. Meert, E. Van Wolputte Katholieke Universiteit Leuven, Leuven, Belgium P. Van Trappenpresenter CERN, Geneva, Switzerland
	Reliability, availability and maintainability determine whether or not a large-scale accelerator system can be operated in a sustainable, cost-effective manner. Beam transfer equipment (e.g. kicker magnets) has potentially significant impact on the global performance of a machine complex. Identifying root causes of malfunctions is currently tedious, and will become infeasible in future systems due to increasing complexity. Machine Learning could automate this process. For this purpose a collaboration between CERN and KU Leuven was established. We present an anomaly detection pipeline which includes preprocessing, detection, postprocessing and evaluation. Merging data of different, asynchronous sources is one of the main challenges. Currently, Gaussian Mixture Models and Isolation Forests are used as unsupervised detectors. To validate, we compare to manual e-logbook entries, which constitute a noisy ground truth. A grid search allows for hyper-parameter optimization across the entire pipeline. Lastly, we incorporate expert knowledge by means of semi-supervised clustering with COBRAS.
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR010
About •	paper received ※ 30 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

Paper

Title

Page

WEMPR001

Data Analysis Infrastructure for Diamond Light Source Macromolecular & Chemical Crystallography and Beyond

1031

WEPHA094

use link to see paper's listing under its alternate paper code

M. Gerstel, A. Ashton, R.J. Gildea, K. Levik, G. Winter
DLS, Oxfordshire, United Kingdom

The Diamond Light Source data analysis infrastructure, Zocalo, is built on a messaging framework. Analysis tasks are processed by a scalable pool of workers running on cluster nodes. Results can be written to a common file system, sent to another worker for further downstream processing and/or streamed to a LIMS. Zocalo allows increased parallelization of computationally expensive tasks and makes the use of computational resources more efficient. The infrastructure is low-latency, fault-tolerant, and allows for highly dynamic data processing. Moving away from static workflows expressed in shell scripts we can easily re-trigger processing tasks in the event that an issue is found. It allows users to re-run tasks with additional input and ensures that automatically and manually triggered processing results are treated equally. Zocalo was originally conceived to cope with the additional demand on infrastructure by the introduction of Eiger detectors with up to 18 Mpixels and running at up to 560 Hz framerate on single crystal diffraction beamlines. We are now adapting Zocalo to manage processing tasks for ptychography, tomography, cryo-EM, and serial crystallography workloads.

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR001

About •

paper received ※ 30 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR003

Exploring Embedded Systems’ Dedicated Cores for Real-Time Applications

1036

WEPHA107

use link to see paper's listing under its alternate paper code

P.H. Nallin, J.G.R.S. Franco, R.C. Ito, A.R.D. Rodrigues
LNLS, Campinas, Brazil

Developments and research in high technology leads to powerful and sophisticated machines which are highly important for many scientific fields. Considering real-time applications, however, these systems tend to become non-deterministic and users may find themselves inside a not completely controllable environment. Exploring open-hardware single board computers with a system-on-a-chip which usually runs an operational system on their main processor(s) and also have real-time units is a good alternative. These real-time units are designed as a microcontroller embedded on the chip where a firmware is loaded, runs concomitantly and exchanges data with the main system. As a result, it is possible to achieve performance increase, high temporal resolution and low latency and jitter, features that are widely desired for controls and critical data acquisition systems. This system architecture allows moving real-time data into high level servers, such as Redis (Remote Dictionary Server) and EPICS, easily. This paper introduces and shows uses of Beaglebone Black, an inexpensive single-board computer, its Programmable Real-Time Units (PRUs) and data sharing with Redis data structure.

Poster WEMPR003 [6.128 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR003

About •

paper received ※ 30 September 2019 paper accepted ※ 18 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR004

Why Should You Invest in Asset Management? A Fire and Gas Use Case

1041

WEPHA109

use link to see paper's listing under its alternate paper code

H. Nissen, S. Grau
CERN, Geneva, Switzerland

At present, the CERN Fire and Gas detection systems involve about 22500 sensors and their number is increasing rapidly at the same time as the number of equipped installations grows up. These assets cover a wide spectrum of technologies, manufacturers, models, parameters, and ages, reflecting the 60 years of CERN history. The use of strict rules and data structures in the declaration of the assets can make a big impact on the overall system maintainability and therefore on the global reliability of the installation. Organized assets data facilitates the creation of powerful reports that help asset owners and management address material obsolescence and end-of-life concerns with a global perspective Historically preventive maintenance have been used to assure the correct function of the installations. With modern supervision systems, a lot of data is collected and can be used to move from preventive maintenance towards data driven maintenance (predictive). Moreover it optimizes maintenance cost and increase system availability while maintaining reliability. A prerequisite of this move is a coherence on the assets defined in the asset management system and in the supervision system.

Poster WEMPR004 [0.675 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR004

About •

paper received ※ 27 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR005

The Array Control and Data Acquisition System of the Cherenkov Telescope Array

1046

WEPHA117

use link to see paper's listing under its alternate paper code

I. Oya, E. Antolini, M. Fuessling
CTA, Heidelberg, Germany
L. Baroncelli, A. Bulgarelli, V. Conforti, N. Parmiggiani
INAF, Bologna, Italy
J. Borkowski
CAMK, Torun, Poland
A. Carosi, J.N. Jacquemier, G. Maurin
IN2P3-LAPP, Annecy-le-Vieux, France
J. Colome
CSIC-IEEC, Bellaterra, Spain
C. Hoischen
Universität Potsdam, Potsdam-Golm, Germany
E. Lyard, R. Walter
University of Geneva, Geneva, Switzerland
D. Melkumyan, K. Mosshammer, I. Sadeh, T. Schmidt, P.A. Wegner
DESY Zeuthen, Zeuthen, Germany
U. Schwanke
Humboldt University Berlin, Institut für Physik, Berlin, Germany
J. Schwarz
INAF-Osservatorio Astronomico di Brera, Merate, Italy
G. Tosti
Università degli di Perugia, Perugia, Italy

The Cherenkov Telescope Array (CTA) project is the initiative to build the next-generation gamma-ray observatory. With more than 100 telescopes planned to be deployed in two sites, CTA is one of the largest astronomical facilities under construction. The Array Control and Data Acquisition (ACADA) system will be the central element of on-site CTA Observatory operations. The mission of the ACADA system is to manage and optimize the telescope array operations at each of the CTA sites. To that end, ACADA will provide all necessary means for the efficient execution of observations, and for the handling of the several Gb/s generated by each individual CTA telescope. The ACADA system will contain a real-time analysis pipeline, dedicated to the automatic generation of science alert candidates based on the inspection of data being acquired. These science alerts, together with external alerts arriving from other scientific installations, will permit ACADA to modify ongoing observations at sub-minute timescales in order to study high-impact scientific transient phenomena. This contribution describes the challenges, architecture, design principles, and development status of the ACADA system.

Poster WEMPR005 [3.851 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR005

About •

paper received ※ 30 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR006

Application Development in the Face of Evolving Web Technologies at the National Ignition Facility

1052

WEPHA118

use link to see paper's listing under its alternate paper code

E.R. Pernice, C.R. Albiston, R.G. Beeler, E.H. Chou, C.D. Fry, M. Shor, J.L. Spears, D.E. Speck, A.A. Thakur, S.L. West
LLNL, Livermore, California, USA

Funding: This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344
The past decade has seen great advances in web technology, making the browser the de-facto platform for many user applications. Advances in JavaScript, and innovations such as TypeScript, have enabled developers to build large scale applications for the web without sacrificing code maintainability. However, this rapid growth has also been accompanied by turbulence. AngularJS arrived and saw widespread adoption only to be supplanted by Angular 2+ a few years later; meanwhile other JavaScript-based languages and developer tools have proliferated. At the National Ignition Facility (NIF), the Shot Setup Tool (SST) is a large web-based tool for configuring experiments on the NIF that is being developed to replace legacy Java Swing application. We will present our experience in building SST during this turbulent time, including how we have leveraged TypeScript to greatly enhance code readability and maintainability in a multi-developer team, and our current effort to incrementally migrate from AngularJS to React.

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR006

About •

paper received ※ 30 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR008

Web Extensible Display Manager 2

1057

WEPHA146

use link to see paper's listing under its alternate paper code

R.J. Slominski, T.L. Larrieu
JLab, Newport News, Virginia, USA

Funding: Authored by Jefferson Science Associates, LLC under U.S. DOE Contract No. DE-AC05-06OR23177
The Web Extensible Display Manager (WEDM) was first deployed at Jefferson Lab (JLab) in 2016 with the goal of rendering Extensible Display Manager (EDM) control screens on the web for the benefit of accessibility, and with version 2 our aim is to provide a more general purpose display toolkit by freeing ourselves from the constraints of the EDM dependency. Over the last few years WEDM has been extensively used at JLab for 24/7 information kiosks, on-call monitoring, and by remote users and staff. The software has also been deployed to Oak Ridge National Laboratory, and has become more robust as many bug fixes and contributions have been added. However, adoption and utility of the software as a general purpose control system display manager is limited by EDM, which is no longer actively maintained. A new toolkit can be built on modern frameworks, fully embrace web conventions and standards, and support multiple control system data sources. This new version is a result of a technology review and selection, and introduces a web inspired display file format, a web based display builder, new widgets, and a data interface intended to support pluggable data.

Poster WEMPR008 [1.293 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR008

About •

paper received ※ 24 September 2019 paper accepted ※ 10 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR009

Development of Event Receiver on Zynq-7000 Evaluation Board

1063

WEPHA149

use link to see paper's listing under its alternate paper code

H. Sugimura
KEK, Ibaraki, Japan

The timing system of SuperKEKB accelerator is used Event Timing System developed by Micro Research Finland. In this presentation, we tested the receiver on Zynq7000 evaluation board. The serialized event data are transferred from Event Generator to Event Receiver by using GTX transceiver. So, we selected Zynq7000(7z030) as receiver, because the FPGA has the GTX. And also, Zynq is mounted on arm processor, it is easily able to control received event data stream by using EPICS ICO. Finally we are aiming to combine event system and RF or BPM system in one FPGA board.

Poster WEMPR009 [0.572 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR009

About •

paper received ※ 17 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEMPR010

Anomaly Detection for CERN Beam Transfer Installations Using Machine Learning

1066

WEPHA155

use link to see paper's listing under its alternate paper code

T. Dewitte, W. Meert, E. Van Wolputte
Katholieke Universiteit Leuven, Leuven, Belgium
P. Van Trappenpresenter
CERN, Geneva, Switzerland

Reliability, availability and maintainability determine whether or not a large-scale accelerator system can be operated in a sustainable, cost-effective manner. Beam transfer equipment (e.g. kicker magnets) has potentially significant impact on the global performance of a machine complex. Identifying root causes of malfunctions is currently tedious, and will become infeasible in future systems due to increasing complexity. Machine Learning could automate this process. For this purpose a collaboration between CERN and KU Leuven was established. We present an anomaly detection pipeline which includes preprocessing, detection, postprocessing and evaluation. Merging data of different, asynchronous sources is one of the main challenges. Currently, Gaussian Mixture Models and Isolation Forests are used as unsupervised detectors. To validate, we compare to manual e-logbook entries, which constitute a noisy ground truth. A grid search allows for hyper-parameter optimization across the entire pipeline. Lastly, we incorporate expert knowledge by means of semi-supervised clustering with COBRAS.

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2019-WEMPR010

About •

paper received ※ 30 September 2019 paper accepted ※ 09 October 2019 issue date ※ 30 August 2020

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)