Difference between revisions of "GlueX TOF Meeting, March 27, 2019"

From GlueXWiki
Jump to: navigation, search
m
m (TOF ML)
 
(3 intermediate revisions by 2 users not shown)
Line 9: Line 9:
 
# Announcements
 
# Announcements
 
# Review of [[GlueX TOF Meeting, March 13, 2019#Minutes|minutes from the last meeting]]
 
# Review of [[GlueX TOF Meeting, March 13, 2019#Minutes|minutes from the last meeting]]
 +
# [https://halldweb.jlab.org/DocDB/0039/003953/001/TOF_slides.pdf TOF ML (Thomas)]
 
# Upgrade Plan
 
# Upgrade Plan
 
# NIM Paper (Paul)
 
# NIM Paper (Paul)
Line 16: Line 17:
  
 
Talks can be deposited in the directory <code>/group/halld/www/halldweb/html/talks/2019</code> on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2019/ .
 
Talks can be deposited in the directory <code>/group/halld/www/halldweb/html/talks/2019</code> on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2019/ .
 +
 +
== Minutes ==
 +
 +
Present:
 +
 +
* ''' FSU: ''' Paul Eugenio
 +
* ''' JLab: ''' Thomas Britton, Mark Ito (chair), Simon Taylor, Tim Whitlatch, Stephanie Worthington, Beni Zihlmann
 +
 +
There is a [https://bluejeans.com/s/QMKAn/ recording of this meeting] on the BlueJeans site. Use your JLab credentials to access it.
 +
 +
=== Upgrade Plan ===
 +
 +
Stephanie showed some screen shots of the current design, see below. She has been working on modifications to the original STEP files sent from Ian Winger. Known interferences have been addressed.
 +
 +
{|
 +
|-
 +
| [[File:48cm arrangment Full Assembly- DS.png|thumb|Upstream Layer]]
 +
| [[File:48cm arrangment Full Assembly- DS.png|thumb|Downstream Layer]]
 +
|-
 +
|[[File:48cm arrangment Full Assy-close up.png|thumb|Close-up, one view]]
 +
|[[File:48cm arrangment Full Assy-close up Iso.png|thumb|Close-up, another view]]
 +
|}
 +
 +
We discussed recent activities and next steps.
 +
* Holder brackets were adjusted. JLab ordered these pieces for the original TOF.
 +
* Ian is doing a measurement on the mini-TOF to confirm the 2-3/4" OD of the iron shields.
 +
* Paul would like to review the design from first principles with the 2-3/4" dimension in mind (as opposed to 3"). Tim cautioned that there is not much space between the iron shields in the design as it stands.
 +
* FSU will machine the PMT shields and collars as they did for the original TOF.
 +
* Stephanie will send the latest design files to FSU.
 +
* There will be a two-foot gap between bolts used to clamp the support beams together in the center of the array. This has to be done to accommodate the new light guide configuration. This should not cause a problem with support of the counters.
 +
* We should order scintillator and light guide material ASAP. These are on the critical path.
 +
* More tubes will be needed, but not at the beginning of construction.
 +
* The written procedures have been pulled. Mark Stevens is looking at them.
 +
* Stephanie mentioned that an assembly bracket[?] needs redesign.
 +
 +
=== Review of minutes from the last meeting ===
 +
 +
We went over [https://halldweb.jlab.org/wiki/index.php/GlueX_TOF_Meeting,_March_13,_2019#Minutes the minutes].
 +
 +
Beni and Ashley will visit the Hall to confirm that there are sufficient spare channels on the current electronics to accommodate the upgrade.
 +
 +
=== TOF ML ===
 +
 +
Thomas showed
 +
# [https://halldweb.jlab.org/DocDB/0039/003953/001/TOF_slides.pdf slides] describing the Machine Learning system he is using on the monitoring histograms and results from application to the TOF. See his slides for the details. Some highlights:
 +
* The final product will use inputs from histograms for all detector subsystems as well as EPICS variables.
 +
* He is using Keras with a TensorFlow backend.
 +
* He is using the Inception Network, an arrangement of nodes that was developed specifically to do image classification[?].
 +
* The results for the TOF are quite good: "Good" accuracy of 99.4% with a 0.8% rate of false positives and "Bad" accuracy of 90.9% with a 0.8% rate of false positives.
 +
* Future work (from his slide):
 +
** Move to histogram values
 +
** Begin hooking up these nets with the super network (the final product mentioned above)
 +
** Work it into the workflow in counting house

Latest revision as of 09:05, 10 April 2019

GlueX Time-of-Flight Meeting
Wednesday, March 27, 2019
10:00 am
JLab: CEBAF Center, Room F226
BlueJeans: 350 531 998

Agenda

  1. Announcements
  2. Review of minutes from the last meeting
  3. TOF ML (Thomas)
  4. Upgrade Plan
  5. NIM Paper (Paul)
  6. Action Item Recap

Slides

Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2019 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2019/ .

Minutes

Present:

  • FSU: Paul Eugenio
  • JLab: Thomas Britton, Mark Ito (chair), Simon Taylor, Tim Whitlatch, Stephanie Worthington, Beni Zihlmann

There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.

Upgrade Plan

Stephanie showed some screen shots of the current design, see below. She has been working on modifications to the original STEP files sent from Ian Winger. Known interferences have been addressed.

Upstream Layer
Downstream Layer
Close-up, one view
Close-up, another view

We discussed recent activities and next steps.

  • Holder brackets were adjusted. JLab ordered these pieces for the original TOF.
  • Ian is doing a measurement on the mini-TOF to confirm the 2-3/4" OD of the iron shields.
  • Paul would like to review the design from first principles with the 2-3/4" dimension in mind (as opposed to 3"). Tim cautioned that there is not much space between the iron shields in the design as it stands.
  • FSU will machine the PMT shields and collars as they did for the original TOF.
  • Stephanie will send the latest design files to FSU.
  • There will be a two-foot gap between bolts used to clamp the support beams together in the center of the array. This has to be done to accommodate the new light guide configuration. This should not cause a problem with support of the counters.
  • We should order scintillator and light guide material ASAP. These are on the critical path.
  • More tubes will be needed, but not at the beginning of construction.
  • The written procedures have been pulled. Mark Stevens is looking at them.
  • Stephanie mentioned that an assembly bracket[?] needs redesign.

Review of minutes from the last meeting

We went over the minutes.

Beni and Ashley will visit the Hall to confirm that there are sufficient spare channels on the current electronics to accommodate the upgrade.

TOF ML

Thomas showed

  1. slides describing the Machine Learning system he is using on the monitoring histograms and results from application to the TOF. See his slides for the details. Some highlights:
  • The final product will use inputs from histograms for all detector subsystems as well as EPICS variables.
  • He is using Keras with a TensorFlow backend.
  • He is using the Inception Network, an arrangement of nodes that was developed specifically to do image classification[?].
  • The results for the TOF are quite good: "Good" accuracy of 99.4% with a 0.8% rate of false positives and "Bad" accuracy of 90.9% with a 0.8% rate of false positives.
  • Future work (from his slide):
    • Move to histogram values
    • Begin hooking up these nets with the super network (the final product mentioned above)
    • Work it into the workflow in counting house