FUSE meeting notes: Jul 18 2023

[orso82]

Orso

• Going back in the groove
  • Julia 1.9.2 installed no problem (fast startup ˜15s, highly recommend updating from 1.8.x)
  • Reviewed open FUSE issues and closed some
  • Opened issues related to David's experience (second 1/2 of this meeting dedicated to that)
  • Need to wrap-up two issues I was working before the break
    • When burning plasma, possible multiple core solutions
      • Which one to pick?
      • How to control the burn?
    • Robustness of TEQUILA solver
  • Will be focuing on supporting David's FUSE runs and time-dependence
• We're joining JuliaHub ARPA-e proposal
  • Will discuss with JuliaHub folks at noon today
  • 2024 FUSE IR&D focus will be time dependence and control
  • extra $10k for 2023 to get JH consulting till end of the year
• We're overspending
  • Asked Tyler to pause
  • Kevin to leave GA

Kevin

• Document how to go from the FUSE JSON output to PARAMAK and run the 3D OpenMC simulation
• Wrap up (as much as possible) and document the work you've done on NeutronTransport.jl
• Transfer ownership of NeutronTransport.jl to the ProjectTorreyPines organization (or at least give me access to it so that I can clone it)
• Amber Annecy in diagnostics group but a bit out of her wheelhouse

Leo

• Recently joined (strategic development group)
• Expert in financial modeling (different from costing)
• Mostly EXCEL

Tim

• Generalizing how we couple actors (IMAS has some physical quantities repeated in different places)
  • get_from(dd, :ip, :equilibrium) function abstracts getting these quantities
• Tim went to SOFE and will be at JuliaCon (end of July)

Adriana

• ITER school
• Works for running NEO
  • integrating it with the neoclassical actor
• Working on Hirshmar Sigmar

Amani

• Cyberinfrastructure-Enabled Machine Learning (CIML) training at UCSD
• Discussion with Leo about the cost of materials

Tom

• Retrained TGLF-NN on FPP+DIII-D cases

David

• Instead of using TAUENN, found that is better to set core profiles and then see what confinement looks like, then use global optimizer to converge to specific confinement factor
• Presented lots of progress on the work done towards re-baselining GA FPP design using FUSE
  • Using simplest FUSE models and trying to replicate same assumptions/thoughts that went in the original GASC optimization
• We should improve the multi objective optimization
  • Info from failing runs is not used
  • Results that violate constraints are used