Author: Ibrahim, M.A.
Paper Title Page
MOPA15 Synchronous High-Frequency Distributed Readout for Edge Processing at the Fermilab Main Injector and Recycler 79
 
  • J.R. Berlioz, J.M.S. Arnold, M.R. Austin, P.M. Hanlet, K.J. Hazelwood, M.A. Ibrahim, A. Narayanan, D.J. Nicklaus, G. Pradhan, A.L. Saewert, B.A. Schupbach, R.M. Thurman-Keup, N.V. Tran
    Fermilab, Batavia, Illinois, USA
  • J. Jiang, H. Liu, S. Memik, R. Shi, M. Thieme, D. Ulusel
    Northwestern University, Evanston, Illinois, USA
  • A. Narayanan
    Northern Illinois University, DeKalb, Illinois, USA
 
  Funding: Operated by Fermi Research Alliance, LLC under Contract No.De-AC02-07CH11359 with the United States Department of Energy. Additional funding provided by Grant Award No. LAB 20-2261
The Main In­jec­tor (MI) was com­mis­sioned using data ac­qui­si­tion sys­tems de­vel­oped for the Fer­mi­lab Main Ring in the 1980s. New VME-based in­stru­men­ta­tion was com­mis­sioned in 2006 for beam loss mon­i­tors (BLM), which pro­vided a more sys­tem­atic study of the ma­chine and im­proved dis­plays of rou­tine op­er­a­tion. How­ever, cur­rent pro­jects are de­mand­ing more data and at a faster rate from this aging hard­ware. One such pro­ject, Real-time Edge AI for Dis­trib­uted Sys­tems (READS), re­quires the high-fre­quency, low-la­tency col­lec­tion of syn­chro­nized BLM read­ings from around the ap­prox­i­mately two-mile ac­cel­er­a­tor com­plex. Sig­nif­i­cant work has been done to de­velop new hard­ware to mon­i­tor the VME back­plane and broad­cast BLM mea­sure­ments over Eth­er­net, while not dis­rupt­ing the ex­ist­ing op­er­a­tions-crit­i­cal func­tions of the BLM sys­tem. This paper will de­tail the de­sign, im­ple­men­ta­tion, and test­ing of this par­al­lel data path­way.
 
poster icon Poster MOPA15 [1.641 MB]  
DOI • reference for this paper ※ doi:10.18429/JACoW-NAPAC2022-MOPA15  
About • Received ※ 03 August 2022 — Revised ※ 04 August 2022 — Accepted ※ 14 August 2022 — Issue date ※ 19 August 2022
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPA28 Semantic Regression for Disentangling Beam Losses in the Fermilab Main Injector and Recycler 112
 
  • M. Thieme, H. Liu, S. Memik, R. Shi
    Northwestern University, Evanston, Illinois, USA
  • J.M.S. Arnold, M.R. Austin, P.M. Hanlet, K.J. Hazelwood, M.A. Ibrahim, V.P. Nagaslaev, A. Narayanan, D.J. Nicklaus, G. Pradhan, A.L. Saewert, B.A. Schupbach, K. Seiya, R.M. Thurman-Keup, N.V. Tran
    Fermilab, Batavia, Illinois, USA
 
  Funding: Operated by Fermi Research Alliance, LLC under Contract No.De-AC02-07CH11359 with the United States Department of Energy. Additional funding provided by Grant Award No. LAB 20-2261, Batavia, IL USA
Fer­mi­lab’s Main In­jec­tor en­clo­sure houses two ac­cel­er­a­tors: the Main In­jec­tor (MI) and the Re­cy­cler (RR). In pe­ri­ods of joint op­er­a­tion, when both ma­chines con­tain high in­ten­sity beam, ra­dia­tive beam losses from MI and RR over­lap on the en­clo­sure’s beam loss mon­i­tor­ing (BLM) sys­tem, mak­ing it dif­fi­cult to at­tribute those losses to a sin­gle ma­chine. In­cor­rect di­ag­noses re­sult in un­nec­es­sary down­time that in­curs both fi­nan­cial and ex­per­i­men­tal cost. In this work, we in­tro­duce a novel neural ap­proach for au­to­mat­i­cally dis­en­tan­gling each ma­chine’s con­tri­bu­tions to those mea­sured losses. Using a con­tin­u­ous adap­ta­tion of the pop­u­lar UNet ar­chi­tec­ture in con­junc­tion with a novel data aug­men­ta­tion scheme, our model ac­cu­rately in­fers the ma­chine of ori­gin on a per-BLM basis in pe­ri­ods of joint and in­de­pen­dent op­er­a­tion. Cru­cially, by ex­tract­ing beam loss in­for­ma­tion at vary­ing re­cep­tive fields, the method is ca­pa­ble of learn­ing both local and global ma­chine sig­na­tures and pro­duc­ing high qual­ity in­fer­ences using only raw BLM loss mea­sure­ments.
 
DOI • reference for this paper ※ doi:10.18429/JACoW-NAPAC2022-MOPA28  
About • Received ※ 02 August 2022 — Revised ※ 05 August 2022 — Accepted ※ 06 August 2022 — Issue date ※ 03 September 2022
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
MOPA75 Machine Learning for Slow Spill Regulation in the Fermilab Delivery Ring for Mu2e 214
 
  • A. Narayanan
    Northern Illinois University, DeKalb, Illinois, USA
  • J.M.S. Arnold, M.R. Austin, J.R. Berlioz, P.M. Hanlet, K.J. Hazelwood, M.A. Ibrahim, V.P. Nagaslaev, D.J. Nicklaus, G. Pradhan, P.S. Prieto, A.L. Saewert, B.A. Schupbach, K. Seiya, R.M. Thurman-Keup, N.V. Tran
    Fermilab, Batavia, Illinois, USA
  • J. Jiang, H. Liu, S. Memik, R. Shi, M. Thieme, D. Ulusel
    Northwestern University, Evanston, Illinois, USA
 
  Funding: Work done partly (READS) collaboration at Fermilab (Grant Award No. LAB 20-2261). Fermilab is managed by Fermi Research Alliance, LLC (FRA), acting under Contract No. DE-AC02-07CH11359.
A third-in­te­ger res­o­nant slow ex­trac­tion sys­tem is being de­vel­oped for the Fer­mi­lab’s De­liv­ery Ring to de­liver pro­tons to the Mu2e ex­per­i­ment. Dur­ing a slow ex­trac­tion process, the beam on tar­get is li­able to ex­pe­ri­ence small in­ten­sity vari­a­tions due to many fac­tors. Owing to the ex­per­i­ment’s strict re­quire­ments in the qual­ity of the spill, a Spill Reg­u­la­tion Sys­tem (SRS) is cur­rently under de­sign. The SRS pri­mar­ily con­sists of three com­po­nents - slow reg­u­la­tion, fast reg­u­la­tion, and har­monic con­tent tracker. In this pre­sen­ta­tion, we shall pre­sent the in­ves­ti­ga­tions of using Ma­chine Learn­ing (ML) in the fast reg­u­la­tion sys­tem, in­clud­ing fur­ther op­ti­miza­tions of PID con­troller gains for the fast reg­u­la­tion, prospects of an ML agent com­pletely re­plac­ing the PID con­troller using su­per­vised learn­ing schemes such as Long Short-Term Mem­ory (LSTM) and Gated Re­cur­rent Unit (GRU) ML mod­els, the sim­u­lated im­pact and lim­i­ta­tion of ma­chine re­sponse char­ac­ter­is­tics on the ef­fec­tive­ness of both PID and ML reg­u­la­tion of the spill. We also pre­sent here nascent re­sults of Re­in­force­ment Learn­ing ef­forts, in­clud­ing con­tin­u­ous-ac­tion soft ac­tor-critic meth­ods, to reg­u­late the spill rate.
 
DOI • reference for this paper ※ doi:10.18429/JACoW-NAPAC2022-MOPA75  
About • Received ※ 03 August 2022 — Revised ※ 08 August 2022 — Accepted ※ 18 September 2022 — Issue date ※ 05 October 2022
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)