THYGBE —  MC6 Orals   (03-May-18   11:00—12:30)
Chair: L.O. Dallin, CLS, Saskatoon, Saskatchewan, Canada
Paper Title Page
THYGBE1 Applying Artificial Intelligence to Accelerators 2925
 
  • A. Scheinker, R.W. Garnett, D. Rees
    LANL, Los Alamos, New Mexico, USA
  • D.K. Bohler
    SLAC, Menlo Park, California, USA
  • A.L. Edelen, S.V. Milton
    CSU, Fort Collins, Colorado, USA
 
  Particle accelerators are being designed and operated over a wide range of complex beam phase space distributions. For example, the Linac Coherent Light Source (LCLS) upgrade, LCLS-II, is considering complex schemes such as two-color operation [1], while the plasma wake field acceleration facility for advanced accelerator experimental tests (FACET) upgrade, FACET-II, is planning on providing custom tailored current profiles [2]. Because of uncertainty due to limited diagnostics and time varying performance, such as thermal drifts, as well as collective effects and the complex coupling of large numbers of components, it is impossible to use simple look up tables for parameter settings in order to quickly switch between widely varying operating ranges. Several forms of artificial intelligence are currently being investigated in order to enable accelerators to quickly and automatically re-adjust component settings without human intervention. In this work we discuss recent progress in applying neural networks and adaptive feedback algorithms to enable automatic accelerator tuning and optimization.
[1] A. A. Lutman et al., Nat. Photonics 10.11, 745 (2016).
[2] V. Yakimenko et al., IPAC2016, Busan, Korea, 2016.
 
slides icon Slides THYGBE1 [14.261 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2018-THYGBE1  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THYGBE2
Results and Discussion of Recent Applications of Neural Network-Based Approaches to the Modeling and Control of Particle Accelerators  
 
  • A.L. Edelen
    CSU, Fort Collins, Colorado, USA
  • S. Biedron
    University of New Mexico, Albuquerque, USA
  • D.L. Bowring, B.E. Chase, D.R. Edstrom, J. Steimel
    Fermilab, Batavia, Illinois, USA
  • J.P. Edelen
    RadiaSoft LLC, Boulder, Colorado, USA
  • P.J.M. van der Slot
    Mesa+, Enschede, The Netherlands
 
  Here we highlight several examples from our work in applying neural network-based modeling and control techniques to particle accelerator systems, through a combination of simulation and experimental studies. We also discuss where the specific approaches used fit into the state of the art in deep learning for control, including limitations of the present state of the art (for example in efficiently dealing with noisy, time-varying, many-parameter systems, like those found in accelerators). We will also briefly clarify some of the terminology/taxonomy of artificial intelligence, and describe how the neural network approaches used here relate to other classes of algorithms that are familiar to the accelerator community. The particle accelerator applications discussed include resonant frequency control of Fermilab's PIP-II RFQ, fast switching between beam parameters in a compact THz FEL, modeling of the FAST low energy beamline at Fermilab, temperature control for the FAST RF gun, and trajectory control for the Jefferson Laboratory FEL.  
slides icon Slides THYGBE2 [37.663 MB]  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THYGBE3 RF Controls for High-Q Cavities for the LCLS-II 2929
 
  • C. Serrano, K.S. Campbell, L.R. Doolittle, G. Huang, A. Ratti
    LBNL, Berkeley, California, USA
  • R. Bachimanchi, C. Hovater
    JLab, Newport News, Virginia, USA
  • A.L. Benwell, M. Boyes, G.W. Brown, D. Cha, G. Dalit, J.A. Diaz Cruz, J. Jones, R.S. Kelly, A. McCollough
    SLAC, Menlo Park, California, USA
  • B.E. Chase, E. Cullerton, J. Einstein-Curtis, J.P. Holzbauer, D.W. Klepec, Y.M. Pischalnikov, W. Schappert
    Fermilab, Batavia, Illinois, USA
  • L.R. Dalesio, M.A. Davidsaver
    Osprey DCS LLC, Ocean City, USA
 
  Funding: This work was supported by the LCLS-II Project and the U.S. Department of Energy, Contract n. DE-AC02-76SF00515.
The SLAC National Accelerator Laboratory is building LCLS-II, a new 4 GeV CW superconducting (SCRF) Linac as a major upgrade of the existing LCLS. The LCLS-II Low-Level Radio Frequency (LLRF) collaboration is a multi-lab effort within the Department of Energy (DOE) accelerator complex. The necessity of high longitudinal beam stability of LCLS-II imposes tight amplitude and phase stability requirements on the LLRF system (up to 0.01% in amplitude and 0.01° in phase RMS). This is the first time such requirements are expected of superconducting cavities operating in continuous-wave (CW) mode. Initial measurements on the Cryomodule test stands at partner labs have shown that the early production units are able to meet the extrapolated hardware requirements to achieve such levels of performance. A large effort is currently underway for system integration, Experimental Physics and Industrial Control System (EPICS) controls, transfer of knowledge from the partner labs to SLAC and the production and testing of 76 racks of LLRF equipment.
 
slides icon Slides THYGBE3 [14.389 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2018-THYGBE3  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THYGBE4 Early Phase 2 Results of LumiBelle2 for the SuperKEKB Electron Ring 2934
 
  • S. Di Carlo, P. Bambade, D. Jehanno, V. Kubytskyi, C.G. Pang, Y. Peinaud, C. Rimbault
    LAL, Orsay, France
 
  We report on the early SuperKEKB Phase 2 operations of the fast luminosity monitor (LumiBelle2 project). Fast luminosity monitoring is required by the dithering feedback system, which is used to stabilize the beam in the presence of horizontal vibrations. In this report, we focus on the operations related to the electron side of LumiBelle2. Diamond sensors are located 30 meters downstream of the IP, just above, beside, and below the electron beam pipe. During early Phase 2, the sensors are used to measure the background, arising from beam-gas scattering. We present the hardware design, the detection algorithm, and the analysis of the background measurements taken up-to-date. The results are then compared with a detailed simulation of the background, in order to well understand the physical processes involved. The simulation is performed using SAD for generation and tracking purposes, while Geant4 is used to calculate the energy deposition in the diamond sensors.  
slides icon Slides THYGBE4 [3.096 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2018-THYGBE4  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)