FUSE meeting notes: Aug 29 2023

[orso82]

Orso

• Time dependence
  • Brian: What algorithm will we be using
    • staggered time stepping to be 2nd order accurate in time
    • current diffusion in QED is implicit
    • for now we assume transport is adiabatic
• Ovality

Brendan

• no use of @.

David

• We're getting close to optimization
• Errors are in the 1-2 %
• Slowly finds high-Beta solutions at lower cost
• Change the exploration/exploitation parameter?
• We expect cost to go lower at higher beta
  • You can achieve the same amount of power in a lower device
• Most points are not against the constraints --> so there is more to go
• Why optimization is running twice number of cases for each generation?
• Memory issue

Tim+Tom

• FUSE transport (flux matching) on DIII-D data looks AMAZING!
  • Using TGLF-NN (SAT0, SAT1, SAT2)
  • Convergence is a lot better with the NN
  • Evolves everything
  • 6500 cases take 6h on saga cluster (120 cores) 300 iterations each
• Need to spot-check with full TGLF and TGYRO
• Next step is to use the whole steady-state analysis

Tom

• Now training TGLF-NN for DIII-D, ITER

Adriana

• Stuck on Hirshmar-Sigmar
• Joey is helping
• Jeff suggested using only one ion specie for one to one comparison

Giacomo

• Install and first run
• Where is the documentation for the actors