Keyword: network
Paper Title Other Keywords Page
MOPAB182 Automated Synchrotron Lattice Design and Optimisation Using a Multi-Objective Genetic Algorithm lattice, synchrotron, dipole, superconducting-magnet 616
 
  • X. Zhang, S.L. Sheehy
    The University of Melbourne, Melbourne, Victoria, Australia
  • E. Benedetto
    TERA, Novara, Italy
  • E. Benedetto
    CERN, Meyrin, Switzerland
  
  Funding: This work is partially supported by the Australian Government Research Training Program Scholarship.
As part of the Next Ion Medical Machine Study (NIMMS), we present a new method for designing synchrotron lattices. A step-wise approach was used to generate random lattice structures from a set of feedforward neural networks. These lattice designs are optimised by evolving the networks over many iterations with a multi-objective genetic algorithm (MOGA). The final set of solutions represent the most effi- cient and feasible lattices which satisfy the design constraints. It is up to the lattice designer to choose a design that best suits the intended application. The automated algorithm presented here randomly samples from all possible lattice layouts and reaches the global optimum over many iterations. The requirements of an efficient extraction scheme in hadron therapy synchrotrons impose stringent constraints on the lat- tice optical functions. Using this algorithm allows us to find the global optimum that is tailored to these constraints and to fully utilise the flexibilities provided by new technology. 		 
poster icon Poster MOPAB182 [6.006 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB182  
About • paper received ※ 15 May 2021       paper accepted ※ 23 June 2021       issue date ※ 14 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB281 Research on Resolution Evaluation of Stripline BPM at SXFEL-UF FEL, electron, linac, experiment 892
 
  • B. Gao, J. Chen, Y.B. Leng
    SSRF, Shanghai, People’s Republic of China
  
  48 stripline BPMs are installed in the injection section and linear acceleration section of Shanghai X-ray Free Electron Laser (SXFEL) for electron beam position measurement. These two sections require resolution of 20 µm@100pC. Resolution evaluation is an important step in BPM installation and commissioning. This paper presents BPM resolution evaluation methods based on correlation analysis. Experimental methods, data processing and result analysis will be discussed  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB281  
About • paper received ※ 19 May 2021       paper accepted ※ 27 May 2021       issue date ※ 02 September 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB286 Towards a Data Science Enabled MeV Ultrafast Electron Diffraction System electron, experiment, real-time, laser 906
 
  • M.A. Fazio, S. Biedron, M. Martínez-Ramón, D.J. Monk, S.I. Sosa Guitron
    UNM-ECE, Albuquerque, USA
  • M. Babzien, K.A. Brown, M.G. Fedurin, J.J. Li, M.A. Palmer, J. Tao
    BNL, Upton, New York, USA
  • S. Biedron, T. Talbott
    UNM-ME, Albuquerque, New Mexico, USA
  • J. Chen, A.J. Hurd, N.A. Moody, R. Prasankumar, C. Sweeney
    LANL, Los Alamos, New Mexico, USA
  • D. Martin, M.E. Papka
    ANL, Lemont, Illinois, USA
  
  Funding: US DOE, SC, BES, MSE, award DE-SC0021365 and DOE NNSA award 89233218CNA000001 through DOE’s EPSCoR program in Office of BES with resources of DOE SC User Facilities BNL’s ATF and ALCF.
A MeV ultrafast electron diffraction (MUED) instrument is a unique characterization technique to study ultrafast processes in materials by a pump-probe technique. This relatively young technology can be advanced further into a turn-key instrument by using data science and artificial intelligence (AI) mechanisms in conjunctions with high-performance computing. This can facilitate automated operation, data acquisition and real time or near- real time processing. AI based system controls can provide real time feedback on the electron beam which is currently not possible due to the use of destructive diagnostics. Deep learning can be applied to the MUED diffraction patterns to recover valuable information on subtle lattice variations that can lead to a greater understanding of a wide range of material systems. A data science enabled MUED facility will also facilitate the application of this technique, expand its user base, and provide a fully automated state-of-the-art instrument. We will discuss the progress made on the MUED instrument in the Accelerator Test Facility of Brookhaven National Laboratory. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB286  
About • paper received ※ 20 May 2021       paper accepted ※ 09 June 2021       issue date ※ 25 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB288 Real-Time Edge AI for Distributed Systems (READS): Progress on Beam Loss De-Blending for the Fermilab Main Injector and Recycler real-time, operation, distributed, FPGA 912
 
  • K.J. Hazelwood, M.R. Austin, M.A. Ibrahim, V.P. Nagaslaev, A. Narayanan, D.J. Nicklaus, A.L. Saewert, B.A. Schupbach, K. Seiya, R.M. Thurman-Keup, N.V. Tran
    Fermilab, Batavia, Illinois, USA
  • H. Liu, S. Memik, R. Shi, M. Thieme
    Northwestern University, Evanston, Illinois, USA
  • A. Narayanan
    Northern Illinois University, DeKalb, Illinois, USA
  
  The Fermilab Main Injector enclosure houses two accelerators, the Main Injector and Recycler. During normal operation, high intensity proton beams exist simultaneously in both. The two accelerators share the same beam loss monitors (BLM) and monitoring system. Beam losses in the Main Injector enclosure are monitored for tuning the accelerators and machine protection. Losses are currently attributed to a specific machine based on timing. However, this method alone is insufficient and often inaccurate, resulting in more difficult machine tuning and unnecessary machine downtime. Machine experts can often distinguish the correct source of beam loss. This suggests a machine learning (ML) model may be producible to help de-blend losses between machines. Work is underway as part of the Fermilab Real-time Edge AI for Distributed Systems Project (READS) to develop a ML empowered system that collects streamed BLM data and additional machine readings to infer in real-time, which machine generated beam loss.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB288  
About • paper received ※ 19 May 2021       paper accepted ※ 29 July 2021       issue date ※ 13 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB314 Surrogate Modeling for MUED with Neural Networks electron, experiment, gun, operation 970
 
  • D.J. Monk, S. Biedron, M.A. Fazio, M. Martínez-Ramón, S.I. Sosa Guitron
    UNM-ECE, Albuquerque, USA
  • M. Babzien, K.A. Brown, M.A. Palmer, J. Tao
    BNL, Upton, New York, USA
  • D. Martin, M.E. Papka
    ANL, Lemont, Illinois, USA
  • T. Talbott
    UNM-ME, Albuquerque, New Mexico, USA
  
  Electron diffraction is among the most complex and influential inventions of the last century and contributes to research in many areas of physics and engineering. Not only does it aid in problems like materials and plasma research, electron diffraction systems like the MeV ultra-fast electron diffraction(MUED) instrument at the Brookhaven National Lab(BNL) also present opportunities to explore/implement surrogate modeling methods using artificial intelligence/machine learning/deep learning algorithms. Running the MUED system requires extended periods of uninterrupted runtime, skilled operators, and many varying parameters that depend on the desired output. These problems lend themselves to techniques based on neural networks(NNs), which are suited to modeling, system controls, and analysis of time-varying/multi-parameter systems. NNs can be deployed in model-based control areas and can be used simulate control designs, planned experiments, and to simulate employment of new components. Surrogate models based on NNs provide fast and accurate results, ideal for real-time control systems during continuous operation and may be used to identify irregular beam behavior as they develop.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB314  
About • paper received ※ 20 May 2021       paper accepted ※ 07 June 2021       issue date ※ 15 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB344 Machine Learning Models for Breakdown Prediction in RF Cavities for Accelerators cavity, operation, vacuum, linac 1068
 
  • C. Obermair, A. Apollonio, T. Cartier-Michaud, N. Catalán Lasheras, L. Felsberger, W.L. Millar, W. Wuensch
    CERN, Geneva, Switzerland
  • C. Obermair, F. Pernkopf
    TUG, Graz, Austria
  
  Radio Frequency (RF) breakdowns are one of the most prevalent limits in RF cavities for particle accelerators. During a breakdown, field enhancement associated with small deformations on the cavity surface results in electrical arcs. Such arcs degrade a passing beam and if they occur frequently, they can cause irreparable damage to the RF cavity surface. In this paper, we propose a machine learning approach to predict the occurrence of breakdowns in CERN’s Compact LInear Collider (CLIC) accelerating structures. We discuss state-of-the-art algorithms for data exploration with unsupervised machine learning, breakdown prediction with supervised machine learning, and result validation with Explainable-Artificial Intelligence (Explainable AI). By interpreting the model parameters of various approaches, we go further in addressing opportunities to elucidate the physics of a breakdown and improve accelerator reliability and operation.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB344  
About • paper received ※ 20 May 2021       paper accepted ※ 16 July 2021       issue date ※ 11 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPAB411 Quantifying DNA Damage in Comet Assay Images Using Neural Networks proton, software, experiment, radiation 1233
 
  • S.J.K. Dhinsey, T. Greenshaw, C.P. Welsch
    The University of Liverpool, Liverpool, United Kingdom
  • J.L. Parsons
    Cancer Research Centre, University of Liverpool, Liverpool, United Kingdom
  • C.P. Welsch
    Cockcroft Institute, Warrington, Cheshire, United Kingdom
  
  Funding: This work was supported by the STFC Liverpool Centre for Doctoral Training on Data Intensive Science (LIV. DAT) under grant agreement ST/P006752/1.
Proton therapy for cancer treatment is a rapidly growing field and increasing evidence suggests it induces more complex DNA damage than photon therapy. Accurate comparison between the two treatments requires quantification of the DNA damage the cause, which can be assessed using the Comet Assay. The program outlined here is based on neural network architecture and aims to speed up analysis of Comet Assay images and provide accurate, quantifiable assessment of the DNA damage levels apparent in individual cells. The Comet Assay is an established technique in which DNA fragments are spread out under the influence of an electric field, producing a comet-like object. The elongation and intensity of the comet tail (consisting of DNA fragments) indicate the level of damage incurred. Many methods to measure this damage exist, using a variety of algorithms. However, these can be time consuming, so often only a small fraction of the comets available in an image are analysed. The automatic analysis presented in this contribution aims to improve this. To supplement the training and testing of the network, a Monte Carlo model will also be presented to create simulated comet assay images. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-MOPAB411  
About • paper received ※ 19 May 2021       paper accepted ※ 09 June 2021       issue date ※ 16 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB052 Current Study of Applying Machine Learning to Accelerator Physics at IHEP electron, lattice, target, photon 1477
 
  • J. Wan, Y. Jiao
    IHEP, Beijing, People’s Republic of China
  
  Funding: National Natural Science Foundation of China(No.11922512), Youth Innovation Promotion Association of Chinese Academy of Sciences(No.Y201904) and National Key R&D Program of China(No.2016YFA0401900).
In recent years, machine learning (ML) has attracted increasing interest among the accelerator field. As a complex collection of multiple physical subsystems, the design and operation of an accelerator can be very nonlinear and complicated, while ML is taken as a powerful tool to solve such nonlinear and complicated problems. In this study, we report on several successful applications of ML to accelerator physics at IHEP. The nonlinear dynamics optimization of the High Energy Photon Source (HEPS) that is a 4th-generation light source is a challenging topic. In this optimization, we use a ML surrogate model to fast select the potentially competitive solutions for a multiobjective genetic algorithm that can significantly improve the convergence rate and the diversity among obtained solutions. Besides, we also tried to apply a generative adversarial net to solve one-to-many problems of longitudinal beam current profile shaping. Unlike most supervised machine learning methods than cannot learn one-to-many maps, the generative adversarial net-based method is able to predict multiple solutions instead of one for a 4-dipole chicane to realize several desired custom current profiles. 		 
poster icon Poster TUPAB052 [0.913 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB052  
About • paper received ※ 11 May 2021       paper accepted ※ 21 June 2021       issue date ※ 27 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB060 Machine Learning on Beam Lifetime and Top-Up Efficiency operation, storage-ring, emittance, photon 1499
 
  • Y.P. Sun
    ANL, Lemont, Illinois, USA
  
  Funding: 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.
Both unsupervised and supervised machine learning techniques are employed for automatic clustering, modeling and prediction of Advanced Photon Source (APS) storage ring beam lifetime and top-up efficiency archived in operations. The naive Bayes classifier algorithm is developed and combined with k-means clustering to improve accuracy, where the unsupervised clustering of APS beam lifetime and top-up efficiency is consistent with either true label from data archive or Gaussian kernel density estimation. Artificial neural network algorithms have been developed, and employed for training and modelling the arbitrary relations of beam lifetime and top-up efficiency on many observable parameters. The predictions from artificial neural network reasonably agree with the APS operation data. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB060  
About • paper received ※ 22 May 2021       paper accepted ※ 21 June 2021       issue date ※ 22 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB061 Anomaly Detection by Principal Component Analysis and Autoencoder Approach operation, power-supply, storage-ring, photon 1502
 
  • Y.P. Sun
    ANL, Lemont, Illinois, USA
  
  Funding: 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.
Several different approach are employed to identify the abnormal events in some Advanced Photon Source (APS) operation archived dataset, where dimensionality reduction are performed by either principal component analysis or autoencoder artificial neural network. It is observed that the APS stored beam dump event, which is triggered by magnet power supply fault, may be predicted by analyzing the magnets capacitor temperatures dataset. There is reasonable agreement among two principal component analysis based approaches and the autoencoder artificial neural network approach, on predicting future overall system fault which may result in a stored beam dump in the APS storage ring. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB061  
About • paper received ※ 22 May 2021       paper accepted ※ 18 June 2021       issue date ※ 19 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB198 ESS DTL Tuning Using Machine Learning Methods cavity, DTL, linac, proton 1872
 
  • J.S. Lundquist, N. Milas, E. Nilsson
    ESS, Lund, Sweden
  • S. Werin
    Lund University, Lund, Sweden
  
  The European Spallation Source, currently under construction in Lund, Sweden, will be the world’s most powerful neutron source. It is driven by a proton linac with a current of 62.5 mA, 2.86 ms long pulses at 14 Hz. The final section of its normal-conducting front-end consists of a 39 m long drift tube linac (DTL) divided into five tanks, designed to accelerate the proton beam from 3.6 MeV to 90 MeV. The high beam current and power impose challenges to the design and tuning of the machine and the RF amplitude and phase have to be set within 1% and 1 degree of the design values. The usual method used to define the RF set-point is signature matching, which can be a time consuming and challenging process, and new techniques to meet the growing complexity of accelerator facilities are highly desirable. In this paper we study the usage of Machine Learning to determine the RF optimum amplitude and phase. The data from a simulated phase scan is fed into an artificial neural network in order to identify the needed changes to achieve the best tuning. Our test for the ESS DTL1 shows promising results, and further development of the method will be outlined.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB198  
About • paper received ※ 17 May 2021       paper accepted ※ 21 June 2021       issue date ※ 13 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB215 Novel Non-Linear Particle Tracking Approach Employing Lie Algebraic Theory in the TensorFlow Environment quadrupole, lattice, focusing, operation 1920
 
  • J. Frank, M. Arlandoo, P. Goslawski, J. Li, T. Mertens, M. Ries, L. Vera Ramirez
    HZB, Berlin, Germany
  
  With this paper we present first results for encoding Lie transformations as computational graphs in Tensorflow that are used as layers in a neural network. By implementing a recursive differentiation scheme and employing Lie algebraic arguments we were able to reproduce the diagrams for well known lattice configurations. We track through simple optical lattices that are encountered as the main constituents of accelerators and demonstrate the flexibility and modularity our approach offers. The neural network can represent the optical lattice with predefined coefficients allowing for particle tracking for beam dynamics or can learn from experimental data to fine-tune beam optics.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB215  
About • paper received ※ 12 May 2021       paper accepted ※ 31 August 2021       issue date ※ 21 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB216 Modeling Particle Stability Plots for Accelerator Optimization Using Adaptive Sampling simulation, collider, resonance, dynamic-aperture 1923
 
  • M. Schenk, L. Coyle, T. Pieloni
    EPFL, Lausanne, Switzerland
  • M. Giovannozzi, A. Mereghetti
    CERN, Meyrin, Switzerland
  • E. Krymova, G. Obozinski
    SDSC, Lausanne, Switzerland
  
  Funding: This work is partially funded by the Swiss Data Science Center (SDSC), project C18-07.
One key aspect of accelerator optimization is to maximize the dynamic aperture (DA) of a ring. Given the number of adjustable parameters and the compute-intensity of DA simulations, this task can benefit significantly from efficient search algorithms of the available parameter space. We propose to gradually train and improve a surrogate model of the DA from SixTrack simulations while exploring the parameter space with adaptive sampling methods. Here we report on a first model of the particle stability plots using convolutional generative adversarial networks (GAN) trained on a subset of SixTrack numerical simulations for different ring configurations of the Large Hadron Collider at CERN. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB216  
About • paper received ※ 19 May 2021       paper accepted ※ 17 June 2021       issue date ※ 22 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB285 Broadband Imaging of Coherent Radiation as a Single-Shot Bunch Length Monitor with Femtosecond Resolution radiation, electron, simulation, detector 2147
 
  • J. Wolfenden, R.B. Fiorito, E. Kukstas, C.P. Welsch
    The University of Liverpool, Liverpool, United Kingdom
  • M. Brandin, B.S. Kyle, E. Mansten, S. Thorin
    MAX IV Laboratory, Lund University, Lund, Sweden
  • R.B. Fiorito, C.P. Welsch, J. Wolfenden
    Cockcroft Institute, Warrington, Cheshire, United Kingdom
  • E. Mansten
    Lund University, Division of Atomic Physics, Lund, Sweden
  • T.H. Pacey
    STFC/DL/ASTeC, Daresbury, Warrington, Cheshire, United Kingdom
  
  Funding: This work is supported by the AWAKE-UK project funded by STFC and the STFC Cockcroft core grant No. ST/G008248/1
Bunch length measurements with femtosecond resolution are a key component in the optimisation of beam quality in FELs, storage rings, and plasma-based accelerators. This contribution presents the development of a novel single-shot bunch length monitor with femtosecond resolution, based on broadband imaging of the spatial distribution of emitted coherent radiation. The technique can be applied to many radiation sources; in this study the focus is coherent transition radiation (CTR) at the MAX IV Short Pulse Facility. Bunch lengths of interest at this facility are <100 fs FWHM; therefore the CTR is in the THz to Far-IR range. To this end, a THz imaging system has been developed, utilising high resistivity float zone silicon lenses and a pyroelectric camera; building upon previous results where single-shot compression monitoring was achieved. This contribution presents simulations of this new CTR imaging system to demonstrate the synchrotron radiation mitigation and imaging capability provided, alongside initial measurements and a bunch length fitting algorithm, capable of shot-to-shot operation. A new machine learning analysis method is also discussed. 		 
poster icon Poster TUPAB285 [2.008 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB285  
About • paper received ※ 17 May 2021       paper accepted ※ 24 June 2021       issue date ※ 23 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB297 Data Archive System for Superconducting RIKEN Linear Accelerator at RIBF controls, EPICS, experiment, cyclotron 2178
 
  • A. Uchiyama, N. Fukunishi, M. Kidera, M. Komiyama
    RIKEN Nishina Center, Wako, Japan
  
  At RIKEN Nishina Center, superconducting RIKEN Linear Accelerator (SRILAC) was newly installed at downstream of existing accelerator and upgraded for the search experiments of super-heavy-elements with atomic numbers of 119 and higher. For the data archiving and the data visualization in RI Beam Factory (RIBF) project, we have utilized RIBFCAS (RIBF control archive system) since 2009. For the number of archived data point was expected to increase dramatically for SRILAC, we introduced the Archiver Appliance for improvement of the data archiving performance. On the other hand, to realize a user-friendly system about the data visualization, the data of RIBFCAS and the Archiver Appliance should be visualized on the same system. In this system, by implementing a Web application to convert the RIBFCAS data to JSON format, it became possible to unify the data format with the Archiver Appliance and display the data with the same viewer software. In the SRILAC beam commissioning, it became to useful system for finding anomalies and understanding the behavior of superconducting cavity. In this conference, we report the system implementation, developed tool, and the future plan in detail.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB297  
About • paper received ※ 19 May 2021       paper accepted ※ 10 June 2021       issue date ※ 17 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB309 Alignment Verification and Monitoring Strategies for the Sirius Light Source alignment, monitoring, radiation, survey 2210
 
  • R. Oliveira Neto, R. Junqueira Leão, L.R. Leão
    CNPEM, Campinas, SP, Brazil
  
  The approach for the alignment of Sirius is the use of portable coordinate metrology instruments in a common reference, via a network of stable points previously surveyed. This type of network is composed of a dense distribution of points materialized in the form of embedded target holders on the special slab and radiation shielding. Phenomena such as ground movements, temperature gradients and vibrations could lead to misalignment of the components, possibly causing a degradation in machine performance. Therefore, the relative positions of the accelerator magnets need to be periodically verified along with the structures surrounding it to ensure a good reference to future alignment operations. This paper will present the status of Sirius monitoring systems, including data from the first months of operation of the hydrostatic levelling sensors. Also, possibilities with simplified network measurements for detecting structural deformations and assessing its stability will be presented, along with a proposal of a photogrammetric reconstruction of the alignment profile of the storage ring. Finally, it will be shown a compilation of analysis on the deformation of the Sirius facilities.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB309  
About • paper received ※ 20 May 2021       paper accepted ※ 01 July 2021       issue date ※ 27 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB310 Establishing a Metrological Reference Network for the Alignment of Sirius alignment, survey, controls, laser 2214
 
  • H. Geraissate, G.R. Rovigatti de Oliveira
    LNLS, Campinas, Brazil
  • R. Junqueira Leão
    CNPEM, Campinas, SP, Brazil
  
  Sirius is the Brazilian 4th generation synchrotron light source. It consists of three electron accelerators and it has room for up to 38 beamlines. To make the alignment of Sirius components possible, there is a need for a network of points comprising the installation volume, allowing the location of portable coordinate instruments on a common reference frame. This work describes the development of such networks for the whole Sirius facility. The layout of the networks is presented together with the survey strategies. Details are given on how the calculations combined laser trackers and optical level measurements data and how the Earth curvature compensation was performed. A novel laser tracker orientation technique applied for linking networks on different environments is also presented. Finally, the uncertainty estimation for the resulting network and its deformation history is shown.  
poster icon Poster TUPAB310 [4.084 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB310  
About • paper received ※ 20 May 2021       paper accepted ※ 07 June 2021       issue date ※ 21 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB315 Development of Disaster Prevention System for Accelerator Tunnel radiation, operation, neutron, real-time 2228
 
  • K. Ishii, K. Bessho, M. Yoshioka
    KEK, Ibaraki, Japan
  • Y. Kawabata, H. Matsuda, K. Matsumoto
    Tobishima Corp., Tokyo, Japan
  • S. Tagashira
    Kansai University, Osaka, Japan
  • N. Yamamoto
    J-PARC, KEK & JAEA, Ibaraki-ken, Japan
  
  Funding: This work is supported by Health Labor Sciences Research Grant of Japan
In an enclosed space such as a particle accelerator tunnel, ensuring worker safety during a disaster is an issue of critical importance. It is necessary to have a system in which the manager can know from outside the tunnel whether there is any worker left behind and whether the worker is escaping in the right direction. Because a global positioning system (GPS) is not available in the tunnel, we are developing a disaster prevention system that uses Wi-Fi to transmit the positioning of workers and two-way communication. The Wi-Fi access point (AP) installed in the tunnel should be radiation resistant. Additionally, the equipment carried by the worker is convenient and easy to carry. We tested the radiation hardness of commercial AP devices and developed a smartphone application to perform location information transmission and simultaneous character transmission. In 2019, we installed the system on the J-PARC Main Ring and started its operation. In this paper, the functions of the developed system and its prospects are described. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB315  
About • paper received ※ 19 May 2021       paper accepted ※ 10 June 2021       issue date ※ 25 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB327 Developing Robust Digital Twins and Reinforcement Learning for Accelerator Control Systems at the Fermilab Booster controls, booster, power-supply, FPGA 2268
 
  • D.L. Kafkes
    Fermilab, Batavia, Illinois, USA
  • M. Schram
    JLab, Newport News, Virginia, USA
  
  Funding: This research was sponsored by the Fermilab Laboratory Directed Research and Development Program under Project ID FNAL-LDRD-2019-027: Accelerator Control with Artificial Intelligence.
We describe the offline machine learning (ML) development for an effort to precisely regulate the Gradient Magnet Power Supply (GMPS) at the Fermilab Booster accelerator complex via a Field-Programmable Gate Array (FPGA). As part of this effort, we created a digital twin of the Booster-GMPS control system by training a Long Short-Term Memory (LSTM) to capture its full dynamics. We outline the path we took to carefully validate our digital twin before deploying it as a reinforcement learning (RL) environment. Additionally, we demonstrate the use of a Deep Q-Network (DQN) policy model with the capability to regulate the GMPS against realistic time-varying perturbations. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB327  
About • paper received ※ 18 May 2021       paper accepted ※ 22 June 2021       issue date ※ 20 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPAB413 Rapid Browser-Based Visualization of Large Neutron Scattering Datasets neutron, scattering, experiment, detector 2494
 
  • D.L. Bruhwiler, K. Bruhwiler, P. Moeller, R. Nagler
    RadiaSoft LLC, Boulder, Colorado, USA
  • C.M. Hoffmann, Z.J. Morgan, A.T. Savici, M.G. Tucker
    ORNL, Oak Ridge, Tennessee, USA
  • A. Kuhn, J. Mensmann, P. Messmer, M. Nienhaus, S. Roemer, D. Tatulea
    NVIDIA, Santa Clara, USA
  
  Funding: This work is supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under Award No. DE-SC0021551.
Neutron scattering makes invaluable contributions to the physical, chemical, and nanostructured materials sciences. Single crystal diffraction experiments collect volumetric scattering data sets representing the internal structure relations by combining datasets of many individual settings at different orientations, times and sample environment conditions. In particular, we consider data from the single-crystal diffraction experiments at ORNL.* A new technical approach for rapid, interactive visualization of remote neutron data is being explored. The NVIDIA IndeX 3D volumetric visualization framework** is being used via the HTML5 client viewer from NVIDIA, the ParaView plugin***, and new Jupyter notebooks, which will be released to the community with an open source license.
* L. Coates et al., Rev. Sci. Instrum. 89, 092802 (2018).
** https://developer.nvidia.com/nvidia-index
*** https://blog.kitware.com/nvidia-index-plugin-in-paraview-5-5 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-TUPAB413  
About • paper received ※ 18 May 2021       paper accepted ※ 21 July 2021       issue date ※ 26 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB073 An Overview of the Radio-Frequency System for an Inverse Compton X-Ray Source Based on CLIC Technology klystron, LLRF, controls, laser 2759
 
  • T.G. Lucas, O.J. Luiten, P.H.A. Mutsaers, X.F.D. Stragier, H.A. Van Doorn, F.M. van Setten, H.J.M. van den Heuvel, M.L.M.C. van der Sluis
    TUE, Eindhoven, The Netherlands
  
  Funding: This project is financed by the "Interreg V programme Flanders-Netherlands" with financial support of the European Fund for Regional Development.
Compact inverse Compton scattering X-ray sources are gaining in popularity as the future of lab-based x-ray sources. Smart*Light is one such facility, under commissioning at Eindhoven University of Technology (TU/e), which is based on high gradient X-band technology originally designed for the Compact Linear Collider (CLIC) and its test stands located at CERN. Critical to the beam quality is the RF system which aims to deliver 10-24 MW RF pulses at repetition rates up to 1 kHz with a high amplitude and phase stability of <0.5\% and <0.65~° allowing it to adhere to strict synchronicity conditions at the interaction point. This work overviews the design of the high power and low level RF systems for Smart*Light. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB073  
About • paper received ※ 19 May 2021       paper accepted ※ 23 June 2021       issue date ※ 29 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB184 Optimization of Medical Accelerators proton, medical-accelerators, FEL, detector 3042
 
  • C.P. Welsch
    The University of Liverpool, Liverpool, United Kingdom
  • C.P. Welsch
    Cockcroft Institute, Warrington, Cheshire, United Kingdom
  
  Funding: This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sk’odowska-Curie grant agreement No 675265.
Between 2016 and 2020, 15 Fellows have carried out collaborative research within the 4 MEUR Optimization of Medical Accelerators (OMA) EU-funded innovative training network. Based at universities, research and clinical facilities, as well as industry partners in several European countries, the Fellows have successfully developed a range of beam and patient imaging techniques, improved biological and physical models in Monte Carlo codes, and also help improve the design of existing and future clinical facilities. This paper gives an overview of the research outcomes of this network. It presents results from tracking and LET measurements with the MiniPIX-TimePIX detector for 60 MeV clinical protons, a new treatment planning approach accounting for prompt gamma range verification and interfractional anatomical changes, and summarizes findings from high-gradient testing of an S-band, normal-conducting low phase velocity accelerating structure. Finally, it gives a brief over-view of the scientific and training events organized by the OMA consortium. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB184  
About • paper received ※ 16 May 2021       paper accepted ※ 14 July 2021       issue date ※ 21 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB203 RFQ Beam Dynamics Optimization Using Machine Learning rfq, simulation, focusing, quadrupole 3100
 
  • D. Koser, J.M. Conrad, L.H. Waites, D. Winklehner
    MIT, Cambridge, Massachusetts, USA
  • A. Adelmann, M. Frey, S. Mayani
    PSI, Villigen PSI, Switzerland
  
  To efficiently inject a high-current H2+ beam into the 60 MeV driver cyclotron for the proposed IsoDAR project in neutrino physics, a novel direct-injection scheme is planned to be implemented using a compact radio-frequency quadrupole (RFQ) as a pre-buncher, being partially inserted into the cyclotron yoke. To optimize the RFQ beam dynamics design, machine learning approaches were investigated for creating a surrogate model of the RFQ. The required sample datasets are generated by standard beam dynamics simulation tools like PARMTEQM and RFQGen or more sophisticated PIC simulations. By reducing the computational complexity of multi-objective optimization problems, surrogate models allow to perform sensitivity studies and an optimization of the crucial RFQ beam output parameters like transmission and emittances. The time to solution might be reduced by up to several orders of magnitude. Here we discuss different methods of surrogate model creation (polynomial chaos expansion and neural networks) and identify present limitations of surrogate model accuracy.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB203  
About • paper received ※ 20 May 2021       paper accepted ※ 01 July 2021       issue date ※ 30 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB289 Machine Learning Based Spatial Light Modulator Control for the Photoinjector Laser at FLUTE laser, electron, target, experiment 3332
 
  • C. Xu, E. Bründermann, A.-S. Müller, M.J. Nasse, A. Santamaria Garcia, C. Sax, C. Widmann
    KIT, Karlsruhe, Germany
  • A. Eichler
    DESY, Hamburg, Germany
  
  Funding: C. Xu acknowledges the support by the DFG-funded Doctoral School "Karlsruhe School of Elementary and Astroparticle Physics: Science and Technology".
FLUTE (Ferninfrarot Linac- und Test-Experiment) at KIT is a compact linac-based test facility for novel accelerator technology and a source of intense THz radiation. FLUTE is designed to provide a wide range of electron bunch charges from the pC- to nC-range, high electric fields up to 1.2 GV/m, and ultra-short THz pulses down to the fs-timescale. The electrons are generated at the RF photoinjector, where the electron gun is driven by a commercial titanium sapphire laser. In this kind of setup the electron beam properties are determined by the photoinjector, but more importantly by the characteristics of the laser pulses. Spatial light modulators can be used to transversely and longitudinally shape the laser pulse, offering a flexible way to shape the laser beam and subsequently the electron beam, influencing the produced THz pulses. However, nonlinear effects inherent to the laser manipulation (transportation, compression, third harmonic generation) can distort the original pulse. In this paper we propose to use machine learning methods to manipulate the laser and electron bunch, aiming to generate tailor-made THz pulses. The method is demonstrated experimentally in a test setup. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB289  
About • paper received ※ 19 May 2021       paper accepted ※ 06 July 2021       issue date ※ 26 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB304 Multi-Objective Multi-Generation Gaussian Process Optimizer operation, framework, simulation, storage-ring 3383
 
  • X. Huang, M. Song, Z. Zhang
    SLAC, Menlo Park, California, USA
  
  Funding: DOE, Office of Science, Office of Basic Energy Sciences, DE-AC02-76SF00515 and FWP 2018-SLAC-100469 Computing Science, Office of Advanced Scientific Computing Research, FWP 2018-SLAC-100469ASCR.
We present a multi-objective evolutionary optimization algorithm that uses Gaussian process (GP) regression-based models to select trial solutions in a multi-generation iterative procedure. In each generation, a surrogate model is constructed for each objective function with the sample data. The models are used to evaluate solutions and to select the ones with a high potential before they are evaluated on the actual system. Since the trial solutions selected by the GP models tend to have better performance than other methods that only rely on random operations, the new algorithm has much higher efficiency in exploring the parameter space. Simulations with multiple test cases show that the new algorithm has a substantially higher convergence speed and stability than NSGA-II, MOPSO, and some other recent preselection-assisted algorithms. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB304  
About • paper received ※ 17 May 2021       paper accepted ※ 12 July 2021       issue date ※ 28 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB306 Applying Machine Learning to Optimization of Cooling Rate at Low Energy RHIC Electron Cooler electron, simulation, experiment, emittance 3391
 
  • Y. Gao, K.A. Brown, P.S. Dyer, S. Seletskiy, H. Zhao
    BNL, Upton, New York, USA
  
  Funding: Work supported by Brookhaven Science Associates, LLC under Contract No. DE-SC0012704 with the U.S. Department of Energy.
The Low Energy RHIC electron Cooler (LEReC) is a novel, state-of-the-art, electron accelerator for cooling RHIC ion beams, which was recently built and commissioned. Optimization of cooling with LEReC requires fine-tuning of numerous LEReC parameters. In this work, initial optimization results of using Machine Learning (ML) methods - Bayesian Optimization (BO) and Q-learning are presented. Specially, we focus on exploring the influence of the electron trajectory on the cooling rate. In the first part, simulations are conducted by utilizing a LEReC simulator. The results show that both methods have the capability of deriving electron positions that can optimize the cooling rate. Moreover, BO takes fewer samples to converge than the Q-learning method. In the second part, Bayesian optimization is further trained on the historical cooling data. In the new samples generated by the BO, the percentage of larger cooling rates data is greatly enhanced compared with the original historical data. 		 
poster icon Poster WEPAB306 [1.083 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB306  
About • paper received ※ 12 May 2021       paper accepted ※ 01 July 2021       issue date ※ 24 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB308 Measurement-Based Surrogate Model of the SLAC LCLS-II Injector laser, simulation, controls, cathode 3395
 
  • L. Gupta, Y.K. Kim
    University of Chicago, Chicago, Illinois, USA
  • A.L. Edelen, C.E. Mayes, A.A. Mishra, N.R. Neveu
    SLAC, Menlo Park, California, USA
  
  Funding: This project was funded by the DOE SCGSR Program.
There is significant effort within particle accelerator physics to use machine learning methods to improve modeling of accelerator components. Such models can be made realistic and representative of machine components by training them with measured data. These models could be used as virtual diagnostics or for model-based control when fast feedback is needed for tuning to different user settings. To prototype such a model, we demonstrate how a machine learning based surrogate model of the SLAC LCLS-II photocathode injector was developed. To create machine-based data, laser measurements were taken at the LCLS using the virtual cathode camera. These measurements were used to sample particles, resulting in realistic electron bunches, which were then propagated through the injector via the Astra space charge simulation. By doing this, the model is not only able to predict many bulk electron beam parameters and distributions which are often hard to measure or not usually available to measure, but the predictions are more realistic relative to traditionally simulated training data. The methods for training such models, as well as model capabilities and future work are presented here. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB308  
About • paper received ※ 26 May 2021       paper accepted ※ 27 July 2021       issue date ※ 24 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB309 Study and Design of the Appropriate High-Performance Computing System for Beamline Data Analysis Application at Iranian Light Source Facility (ILSF) software, hardware, experiment, data-analysis 3399
 
  • A. Khaleghi, M. Akbari
    ILSF, Tehran, Iran
  • H. Haedar, K. Mahmoudi, M. Takhttavani
    IKIU, Qazvin, Iran
  • S. Mahmoudi
    Sharif University of Technology (SUT), Tehran, Iran
  
  Data analysis is a very important step in doing experiments at light sources, where multiple application and software packages are used for this purpose. In this paper we have reviewed some software packages that are used for data analysis and design at Iranian Light Source Facility then according to their processing needs, after taking in mind different HPC scenarios a suitable architecture for deployment of the ILSF HPC is presented. The proposed architecture is a cluster of 64 computing nodes connected through Ethernet and InfiniBand network running a Linux operating system with support of MPI parallel environment.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB309  
About • paper received ※ 19 May 2021       paper accepted ※ 23 July 2021       issue date ※ 01 September 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB310 Study and Design of a High-Performance Computing Infrastructure for Iranian Light Source Facility Based on the Accelerator Physicists and Engineers’ Applications Requirements software, hardware, Ethernet, simulation 3402
 
  • K. Mahmoudi, H. Haedar, A. Khaleghi
    IKIU, Qazvin, Iran
  • M. Akbari, A. Khaleghi
    ILSF, Tehran, Iran
  • S. Mahmoudi
    IUST, Narmac, Tehran, Iran
  
  Synchrotron design and operation are one of the complex tasks which requires a lot of precise computation. As an example, we could mention the simulations done for calculating the impedance budget of the machine which requires a notable amount of computational power. In this paper we are going to review different HPC scenarios suitable for this matter then we will present our design of a suitable HPC based on the accelerator physicists and engineers’ needs. Going through different HPC scenarios such as shared memory architectures, distributed memory architectures, cluster, grid and cloud computing we conclude implementation of a dedicated computing cluster can be desired for ILSF. Cluster computing provides the opportunity for easy and saleable scientific computation for ILSF also another advantage is that its resources can be used for running cloud or grid computing platforms as well.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB310  
About • paper received ※ 19 May 2021       paper accepted ※ 19 July 2021       issue date ※ 12 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB318 Prediction and Clustering of Longitudinal Phase Space Images and Machine Parameters Using Neural Networks and K-Means Algorithm FEL, simulation, electron, ECR 3417
 
  • M. Maheshwari
    STFC/DL, Daresbury, Warrington, Cheshire, United Kingdom
  • D.J. Dunning, J.K. Jones, M.P. King, H.R. Kockelbergh, A.E. Pollard
    STFC/DL/ASTeC, Daresbury, Warrington, Cheshire, United Kingdom
  
  Machine learning algorithms were used for image and parameter recognition and generation with the aim to optimise the CLARA facility at Daresbury, using start-to-end simulation data. Convolutional and fully connected neural networks were trained using TensorFlow-Keras for different instances, with examples including predicting Longitudinal Phase Space (LPS) images with machine parameters as input and FEL parameter prediction (e.g. pulse energy) from LPS images. The K-means clustering algorithm was used to cluster the LPS images to highlight patterns within the data. Machine learning techniques can enhance the way large amounts of data are processed and analysed and so have great potential for application in accelerator science R&D.  
poster icon Poster WEPAB318 [1.062 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB318  
About • paper received ※ 17 May 2021       paper accepted ※ 05 July 2021       issue date ※ 21 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB329 LCLS-II Average Current Monitor cavity, vacuum, coupling, simulation 3443
 
  • P. Borchard, J.S. Hoh
    Dymenso LLC, San Francisco, USA
  
  The LCLS-II project at SLAC is a high power upgrade to the existing free-electron laser facility. The LCLS-II Accelerator System will include a new 4 GeV continuous-wave superconducting linear accelerator in the first kilometer of the SLAC linear accelerator tunnel and supplements the existing low power pulsed linac. Average Current Monitors (ACMs) are needed to protect against excessive beam power which might otherwise cause damage to the beam dumps. The ACM cavities are pillbox-shaped stainless steel RF cavity with two radial probe ports with couplers, one radial test port with a coupler, and a mechanism for mechanically fine-tuning the cavity resonant frequency. The ACM RF cavities will be located at points of known or constrained beam energy and will monitor the beam current, a safety system will trip off the beam if the beam power exceeds the allowed value.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB329  
About • paper received ※ 19 May 2021       paper accepted ※ 16 June 2021       issue date ※ 22 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB356 Proposal of an Alignment System for HALF: The Reference Network of Alignment alignment, monitoring, real-time, simulation 3533
 
  • X. Li, J.X. Chen, X.Y. He, W. Wang, Z.Y. Wang
    USTC/NSRL, Hefei, Anhui, People’s Republic of China
  • J.X. Chen, T. Luo
    IHEP CSNS, Guangdong Province, People’s Republic of China
  
  As a fourth-generation light source based on the diffraction-limited storage ring, Hefei Advanced Light Facility (HALF) has higher requirements for magnets alignment in accuracy, efficiency, and reliability. In this paper, the Reference Network of Alignment (RNA) system is proposed to improve the magnetic axis alignment accuracy on the radial direction of the beamline. Herein, we mainly introduce the concept design and the theoretical analysis of the RNA system, which center on the novel fusion method of sensors. A simulation result shows that it is credible to assume the RNA system can achieve an alignment installation accuracy of 20 µm and a displacement monitoring accuracy of 10 µm.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB356  
About • paper received ※ 16 May 2021       paper accepted ※ 21 June 2021       issue date ※ 31 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB373 The Energy Management System in NSRRC operation, controls, MMI, radiation 3585
 
  • C.S. Chen, W.S. Chan, Y.Y. Cheng, Y.F. Chiu, Y.-C. Chung, K.C. Kuo, M.T. Lee, Y.C. Lin, C.Y. Liu, Z.-D. Tsai
    NSRRC, Hsinchu, Taiwan
  
  Taiwan has been suffering from a shortage of natural resources for more than two decades. As stated by the Energy Statistics Handbook 2019 of Taiwan, up to 97.90% of energy supply was imported from abroad. This kind of energy consumption structure is fragile relatively. Not mention to the total domestic energy consumption annual growth rate is 1.97% in twenty years. Either the semiconductor or the integrated circuit-related industry is developed vigorously in Taiwan. All the facts cause us to face the energy problems squarely. Therefore, an energy management system (EnMS) was installed in NSRRC in 2019 to pursue more efficient energy use. With the advantages of the Archive Viewer - a utility supervisory control and data acquisition system in NSRRC, the data of energy use could be traced conveniently and widely. The model of energy use has been built to review periodically, furthermore, it provides us the accordance to replace the degraded equipment and alerts us if the failure occurs.  
poster icon Poster WEPAB373 [0.497 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB373  
About • paper received ※ 21 May 2021       paper accepted ※ 22 July 2021       issue date ※ 11 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB374 The Southern Hemisphere’s First X-Band Radio-Frequency Test Facility at the University of Melbourne electron, klystron, gun, GUI 3588
 
  • M. Volpi, R.P. Rassool, S.L. Sheehy, G. Taylor, S.D. Williams
    The University of Melbourne, Melbourne, Victoria, Australia
  • M.J. Boland
    CLS, Saskatoon, Saskatchewan, Canada
  • M.J. Boland
    University of Saskatchewan, Saskatoon, Canada
  • N. Catalán Lasheras, S. Gonzalez Anton, G. McMonagle, S. Stapnes, W. Wuensch
    CERN, Meyrin, Switzerland
  • R.T. Dowd, K. Zingre
    AS - ANSTO, Clayton, Australia
  
  The first Southern Hemisphere X-band Laboratory for Accelerators and Beams (X-LAB) is under construction at the University of Melbourne, and it will operate CERN X-band test stand containing two 12GHz 6MW klystron amplifiers. By power combination through hybrid couplers and the use of pulse compressors, up to 50 MW of peak power can be sent to any of 2 test slots at pulse repetition rates up to 400 Hz. The test stand is dedicated to RF conditioning and testing CLIC’s high gradient accelerating structures beyond 100 MV/m. It will also form the basis for developing a compact accelerator for medical applications, such as radiotherapy and compact light sources. Australian researchers working as part of a collaboration between the University of Melbourne, international universities, national industries, the Australian Synchrotron -ANSTO, Canadian Light Source and the CERN believe that creating a laboratory for novel accelerator research in Australia could drive technological and medical innovation.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB374  
About • paper received ※ 18 May 2021       paper accepted ※ 06 July 2021       issue date ※ 30 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB398 A C-Band RF Mode Launcher with Quadrupole Field Components Cancellation for High Brightness Applications quadrupole, GUI, brightness, linac 3638
 
  • G. Pedrocchi
    SBAI, Roma, Italy
  • D. Alesini, F. Cardelli, A. Gallo, A. Giribono, B. Spataro
    INFN/LNF, Frascati, Italy
  • G. Castorina
    AVO-ADAM, Meyrin, Switzerland
  • L. Ficcadenti
    INFN-Roma, Roma, Italy
  • M. Migliorati, A. Mostacci, L. Palumbo
    Sapienza University of Rome, Rome, Italy
  
  The R&D of high gradient radiofrequency devices is aimed to develop innovative and compact accelerating stuctures based on new manufactoring techniques and materials in order to produce devices operating with the highest accelerating gradient. Recent studies have shown a large increase in the maximum sustained RF surface electric fields in copper structure operating at cryogenic temperature. These novel approaches allow significant performance improvements of RF photoinjectors. Indeed the operation at high surface fields results in considerable increase of electron brilliance. This requires high field quality in the RF photoinjector and specifically in its poweer coupler. In this work we present a novel power coupler for the RF photoinjector. The coupler is a compact C-band TM01 mode launcher with a fourfold symmetry which minimized both the dipole and the quadrupole RF field components.  
poster icon Poster WEPAB398 [1.799 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB398  
About • paper received ※ 13 May 2021       paper accepted ※ 06 July 2021       issue date ※ 23 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
WEPAB412 Use of a Noise IoT Detection System to Measure the Environmental Noise in Taiwan Light Source monitoring, real-time, site, experiment 3671
 
  • P.J. Wen, S.P. Kao, S.Y. Lin, Y.C. Lin
    NSRRC, Hsinchu, Taiwan
  
  In the past, the method of general noise monitoring altered little; noise was still measured with a human hand-held mobile device, or the measurement at fixed sites was made using traditional analogue data-storage equipment. In recent years, with the rapidly improved network transmission capabilities, the development of a small noise-detection IoT system allows the detection data to be transmitted wirelessly without need for human strength measurements, and records noise information. The statistics of subsequent noise data become a basis for analysis and improvement. Taiwan Light Source (TLS) beamlines have many vacuum pumps, cooling pumps, liquid-nitrogen pressure-relief systems, computer servers etc. that generate much noise. This study is expected to prepare for installation of noise detection. The system uses a noise-detection box to detect, to disclose louder locations, to collect noise data, to determine the source and type of noise source, and to provide information to reduce the noise of the working environment. The TLS noise-detection results find that the inner-ring area has less noise and are more stable than the outer ring area.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-WEPAB412  
About • paper received ※ 14 May 2021       paper accepted ※ 24 June 2021       issue date ※ 27 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THXB06 Results of the First Alignment Run for Sirius alignment, laser, operation, survey 3728
 
  • R. Junqueira Leão, R. Oliveira Neto
    CNPEM, Campinas, SP, Brazil
  • H. Geraissate, F. Rodrigues, G.R. Rovigatti de Oliveira
    LNLS, Campinas, Brazil
  
  It is widely known that the position of particle accelerator components is critical for its performance. For the latest generation light sources, whose magnetic lattice is optimized for achieving very low emittance, the tolerable misalignments are in the order of a few dozen micrometers. Due to the perimeter of these machines, the requirements push the limits of large-volume dimensional metrology and associated instruments and techniques. Recently a fine alignment campaign of the Sirius accelerators was conducted following the pre-alignment performed during the installation phase. To conform with the strict relative positioning demands, measurement good practices were followed, and several 3D metrology procedures were developed. Also, to improve positioning resolution, high rigidity translation devices were produced. Finally, the special target holders designed as removable fiducials for the magnets were revisited to assure maximum reliability. Data processing algorithms were implemented to evaluate the alignment results in a robust and agile manner. This paper will present the final positioning errors for Sirius magnets with an expression of the estimated uncertainty.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THXB06  
About • paper received ※ 20 May 2021       paper accepted ※ 02 July 2021       issue date ※ 01 September 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB068 Denoising of Optics Measurements Using Autoencoder Neural Networks optics, simulation, MMI, controls 3915
 
  • E. Fol, R. Tomás García
    CERN, Meyrin, Switzerland
  
  Noise artefacts can appear in optics measurements data due to instrumentation imperfections or uncertainties in the applied analysis methods. A special type of semi-supervised neural networks, autoencoders, are widely applied to denoising tasks in image and signal processing as well as to generative modeling. Recently, an autoencoder-based approach for denoising and reconstruction of missing data has been developed to improve the quality of phase measurements obtained from harmonic analysis of LHC turn-by-turn data. We present the results achieved on simulations demonstrating the potential of the new method and discuss the effect of the noise in light of optics corrections computed from the cleaned data.  
poster icon Poster THPAB068 [0.881 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB068  
About • paper received ※ 19 May 2021       paper accepted ※ 13 July 2021       issue date ※ 02 September 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB191 Physics-Enhanced Reinforcement Learning for Optimal Control lattice, controls, simulation, alignment 4150
 
  • A.N. Ivanov, I.V. Agapov, A. Eichler, S. Tomin
    DESY, Hamburg, Germany
  
  We propose an approach for incorporating accelerator physics models into reinforcement learning agents. The proposed approach is based on the Taylor mapping technique for simulation of the particle dynamics. The resulting computational graph is represented as a polynomial neural network and embedded into the traditional reinforcement learning agents. The application of the model is demonstrated in a nonlinear simulation model of beam transmission. The comparison of the approach with the traditional numerical optimization as well as neural networks based agents demonstrates better convergence of the proposed technique.  
poster icon Poster THPAB191 [0.846 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB191  
About • paper received ※ 11 May 2021       paper accepted ※ 29 July 2021       issue date ※ 24 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB197 Enhancing Efficiency of Multi-Objective Neural-Network-Assisted Nonlinear Dynamics Lattice Optimization via 1-D Aperture Objectives & Objective Focusing lattice, focusing, simulation, storage-ring 4156
 
  • Y. Hidaka, D.A. Hidas, F. Plassard, T.V. Shaftan, G.M. Wang
    BNL, Upton, New York, USA
  
  Funding: This work is supported by U.S. DOE under Contract No. DE-SC0012704.
Mutli-objective optimizers such as multi-objective genetic algorithm (MOGA) have been quite popular in discovering desirable lattice solutions for accelerators. However, even these successful algorithms can become ineffective as the dimension and range of the search space increase due to exponential growth in the amount of exploration required to find global optima. This difficulty is even more exacerbated by the resource-intensive and time-consuming tendency for the evaluations of nonlinear beam dynamics. Lately the use of surrogate models based on neural network has been drawing attention to alleviate this problem. Following this trend, to further enhance the efficiency of nonlinear lattice optimization for storage rings, we propose to replace typically used objectives with those that are less time-consuming and to focus on a single objective constructed from multiple objectives, which can maximize utilization of the trained models through local optimization and objective gradient extraction. We demonstrate these enhancements using a NSLS-II upgrade lattice candidate as an example. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB197  
About • paper received ※ 20 May 2021       paper accepted ※ 23 June 2021       issue date ※ 10 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB201 A Machine Learning Technique for Dynamic Aperture Computation dynamic-aperture, simulation, hadron, distributed 4172
 
  • B. Dalena, M. Ben Ghali
    CEA-IRFU, Gif-sur-Yvette, France
  
  Currently, dynamic aperture calculations of high-energy hadron colliders are performed through computer simulations, which are both a resource-heavy and time-costly processes. The aim of this study is to use a reservoir computing machine learning model in order to achieve a faster extrapolation of dynamic aperture values. A recurrent echo-state network (ESN) architecture is used as a basis for this work. Recurrent networks are better fitted to extrapolation tasks while the reservoir echo-state structure is computationally effective. Model training and validation is conducted on a set of "seeds" corresponding to the simulation results of different machine configurations. Adjustments in the model architecture, manual metric and data selection, hyper-parameters tuning and the introduction of new parameters enabled the model to reliably achieve good performance on examining testing sets.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB201  
About • paper received ※ 14 May 2021       paper accepted ※ 22 July 2021       issue date ※ 02 September 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB243 Optimizing Mu2e Spill Regulation System Algorithms extraction, controls, simulation, resonance 4281
 
  • A. Narayanan
    Northern Illinois University, DeKalb, Illinois, USA
  • K.J. Hazelwood, M.A. Ibrahim, V.P. Nagaslaev, D.J. Nicklaus, P.S. Prieto, B.A. Schupbach, K. Seiya, R.M. Thurman-Keup, N.V. Tran
    Fermilab, Batavia, Illinois, USA
  • H. Liu, S. Memik, R. Shi, M. Thieme
    Northwestern University, Evanston, Illinois, USA
  
  Funding: The work has been performed at Fermilab. Fermilab is managed by Fermi Research Alliance, LLC (FRA), acting under Contract No. DE-AC02-07CH11359.
A slow extraction system is being developed for the Fermilab’s Delivery Ring to deliver protons to the Mu2e experiment. During the extraction, the beam on target experiences small intensity variations owing to many factors. Various adaptive learning algorithms will be employed for beam regulation to achieve the required spill quality. We discuss here preliminary results of the slow and fast regulation algorithms validation through the computer simulations before their implementation in the FPGA. Particle tracking with sextupole resonance was used to determine the fine shape of the spill profile. Fast semi-analytical simulation schemes and Machine Learning models were used to optimize the fast regulation loop. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB243  
About • paper received ※ 20 May 2021       paper accepted ※ 28 July 2021       issue date ※ 20 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB260 Detection and Classification of Collective Beam Behaviour in the LHC extraction, operation, controls, injection 4318
 
  • L. Coyle, F. Blanc, T. Pieloni, M. Schenk
    EPFL, Lausanne, Switzerland
  • X. Buffat, M. Solfaroli Camillocci, J. Wenninger
    CERN, Meyrin, Switzerland
  • E. Krymova, G. Obozinski
    SDSC, Lausanne, Switzerland
  
  Collective instabilities can lead to a severe deterioration of beam quality, in terms of reduced beam intensity and increased beam emittance, and consequently a reduction of the collider’s luminosity. It is therefore crucial for the operation of the CERN’s Large Hadron Collider to understand the conditions in which they appear in order to find appropriate mitigation measures. Using bunch-by-bunch and turn-by-turn beam amplitude data, courtesy of the transverse damper’s observation box (ObsBox), a novel machine learning based approach is developed to both detect and classify these instabilities. By training an autoencoder neural network on the ObsBox amplitude data and using the model’s reconstruction error, instabilities and other phenomena are separated from nominal beam behaviour. Additionally, the latent space encoding of this autoencoder offers a unique image like representation of the beam amplitude signal. Leveraging this latent space representation allows us to cluster the various types of anomalous signals.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB260  
About • paper received ※ 19 May 2021       paper accepted ※ 19 July 2021       issue date ※ 27 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB268 Hierarchical Intelligent Real-Time Optimal Control for LLRF Using Time Series Machine Learning Methods and Transfer Learning controls, LLRF, cavity, simulation 4329
 
  • R. Pirayesh, S. Biedron
    UNM-ME, Albuquerque, New Mexico, USA
  • S. Biedron, J.A. Diaz Cruz, M. Martínez-Ramón
    UNM-ECE, Albuquerque, USA
  • J.A. Diaz Cruz
    SLAC, Menlo Park, California, USA
  
  Funding: supported by DOE, Office of Science, Office of High Energy Physics, under award number DE-SC0019468, Contract No. DE-AC02-76SF00515, also supported Office of Basic Energy Sciences. ALCF, Element Aero
Machine learning (ML) has recently been applied to Low-level RF (LLRF) control systems to keep the voltage and phase of Superconducting Radiofrequency (SRF) cavities stable within 0.01 degree in phase and 0.01% amplitude as constraints. Model predictive control (MPC) uses an optimization algorithm offline to minimize a cost function with constraints on the states and control input. The surrogate model optimally controls the cavities online. Time series deep ML structures including recurrent neural network (RNN) and long short-term memory (LSTM) can model the control input of MPC and dynamics of LLRF as a surrogate model. When the predicted states diverge from the measured states more than a threshold at each time step, the states’ measurements from the cavity fine-tune the surrogate model with transfer learning. MPC does the optimization offline again with the updated surrogate model, and, next, transfer learning fine-tunes the surrogate model with the new data from the optimal control inputs. The surrogate model provides us with a computationally faster and accurate modeling of MPC and LLRF, which in turn results in a more stable control system.
Machine learning, Surrogate model, control, LLRF, MPC, Transfer learning 		 
poster icon Poster THPAB268 [0.377 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB268  
About • paper received ※ 16 May 2021       paper accepted ※ 13 July 2021       issue date ※ 18 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB273 Spectral Reconstruction for FACET-II Compton Spectrometer photon, electron, site, positron 4346
 
  • Y. Zhuang, B. Naranjo, J.B. Rosenzweig, M. Yadav
    UCLA, Los Angeles, USA
  
  Funding: This work was supported by DOE Contract DE-SC0009914, NSF Grant No. PHY-1549132, and DARPA GRIT Contract 20204571.
The Compton spectrometer under development at UCLA for FACET-II is a versatile tool to analyze gamma-ray spectra in a single shot, in which the energy and angular position of the incoming photons are recorded by observing the momenta and position of Compton scattered electrons. We present methods to reconstruct the primary spectrum from these data via machine learning and the EM Algorithm. A multi-layer fully connected neural network is used to perform the regression task of reconstructing both the double-differential spectrum and the photon energy spectrum incident with zero angular offset. We present the expected performance of these techniques, concentrating on the achievable energy resolution. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB273  
About • paper received ※ 20 May 2021       paper accepted ※ 28 July 2021       issue date ※ 16 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPAB349 Feed-Forward Neural Network Based Modelling of an Ultrafast Laser for Enhanced Control laser, controls, electron, cathode 4478
 
  • A. Aslam, M. Martínez-Ramón, S.D. Scott
    UNM-ECE, Albuquerque, USA
  • S. Biedron
    Argonne National Laboratory, Office of Naval Research Project, Argonne, Illinois, USA
  • S. Biedron
    Element Aero, Chicago, USA
  • S. Biedron
    UNM-ME, Albuquerque, New Mexico, USA
  • M. Burger, J. Murphy
    NERS-UM, Ann Arbor, Michigan, USA
  • K.M. Krushelnick, J. Nees, A.G.R. Thomas
    University of Michigan, Ann Arbor, Michigan, USA
  • Y. Ma
    IHEP, Beijing, People’s Republic of China
  • Y. Ma
    Michigan University, Ann Arbor, Michigan, USA
  
  Funding: Acknowledgements: This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of High Energy Physics, under award number DE-SC0019468.
The applications of machine learning in today’s world encompass all fields of life and physical sciences. In this paper, we implement a machine learning based algorithm in the context of laser physics and particle accelerators. Specifically, a neural network-based optimisation algorithm has been developed that offers enhanced control over an ultrafast femtosecond laser in comparison to the traditional Proportional Integral and derivative (PID) controls. This research opens a new potential of utilising machine learning and even deep learning techniques to improve the performance of several different lasers and accelerators systems. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB349  
About • paper received ※ 20 May 2021       paper accepted ※ 02 July 2021       issue date ※ 17 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
FRXC01 Superconducting Radio-Frequency Cavity Fault Classification Using Machine Learning at Jefferson Laboratory cavity, cryomodule, SRF, radio-frequency 4535
 
  • C. Tennant, A. Carpenter, T. Powers, L.S. Vidyaratne
    JLab, Newport News, Virginia, USA
  • K.M. Iftekharuddin, M. Rahman
    ODU, Norfolk, Virginia, USA
  • A.D. Shabalina
    STFC/DL, Daresbury, Warrington, Cheshire, United Kingdom
  
  Funding: This work supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics under Contract No. DE-AC05-06OR23177.
We report on the development of machine learning models for classifying C100 superconducting radiofrequency (SRF) cavity faults in the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab. Of the 418 SRF cavities in CEBAF, 96 are designed with a digital low-level RF system configured such that a cavity fault triggers recordings of RF signals for each of eight cavities in the cryomodule. Subject matter experts analyze the collected time-series data and identify which of the eight cavities faulted first and classify the type of fault. This information is used to find trends and strategically deploy mitigations to problematic cryomodules. However, manually labeling the data is laborious and time-consuming. By leveraging machine learning, near real-time - rather than postmortem - identification of the offending cavity and classification of the fault type has been implemented. We discuss the performance of the machine learning models during a recent physics run. We also discuss efforts for further insights into fault types through unsupervised learning techniques and present preliminary work on cavity and fault prediction using data collected prior to a failure event. 		 
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-FRXC01  
About • paper received ※ 16 May 2021       paper accepted ※ 01 July 2021       issue date ※ 13 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)