Login

WIMAS-2 Weekly Meeting

→ Europe/Berlin
https://kth-se.zoom.us/j/7764082514

https://kth-se.zoom.us/j/7764082514
Thomas Jonsson
Description

WIMAS-2 Weekly Meeting are technical meetings to plan and coordinate the development of the ETS-6 workflow.

 

Remote participation:

https://kth-se.zoom.us/j/7764082514

Thomas
Hide

Minutes from WIMAS-2 Weekly Meeting 29 April 2020

Actions from previous meeting:

There are a few actions not yet completed from 22 April.

There was a request to upload the documents on the JINTRACK-ASTRA benchmark received from Alexei. These are now available at:

Verification_JINTRAC-ASTRA-CORSICA

Active Branches

Branch removeInterpretativeTotalDensity:

  • It was pointed out that it is important to smooth noisy profiles, e.g. experimental profiles or profiles with Monte Carlo noise.

No update on IPramp, nor on revisedImpurity.

How to handle very small or negative densities ot temperatures

Important comments from the discussion:

Dave: We should not extrapolate to remove faulty values. When we have problems at the boundary we need to abort.

Dave, Pär, ... : We should refine time-steps without iterating if the temperature goes negative. 

  • Thomas: Probably a good idea, but this requires bigger changes in the workflow and might be part of a general optimisation of the Convergence Loop.

Jorge: When we get negative values we need to output the virtual source related to the modification of the profile. This can improve the convergence.

  • Answer: To be discussed later on.

Dave: Make sure that the initial (input) profiles satisfies the constraints.

Alexei: Careful with the sign of D/v.

In summary, the proposal was accepted with the exception that we should abort when having unacceptable values at the edge.

Actions:

  1. Thomas: Implement the proposed solution with modifications.
    • Implemented in branch negativeValues. Testing show no error, it causes some convergence problems that we need to understand better.
  2. Thomas: Make sure that the initial profiles satisfies the constraints.
    • Completed. Branch etsInitProfiles has been merged into develop.
  3. Jorge: Formulate a proposal for how to add source terms to compensate for removing negative temperatures, and by this improve the convergence.

rho_tor VS rho_tor_norm

Test that would be useful to add:

  • Compress / Expand (adiabatically) the plasma and check that total energy and number of particles are conserved.
  • Layering: adding/removing layers, a=1m->1.1m and a=1m->0.9m, respectively. Test how the solver maintains conservation laws.

Actions:

  • Jorge: Create test case for testing expansion, compression and layering (both adding and removing layers).

AoB

Impurities:

- Important to deliver something by the end of the year to ensure that the new ETS-6 management gets a working product.

Neutrals:

- Try to get something ready before the end of the year (Task on Irena).

Comments from the Zoom chat

Pär: Typically it is numerical instabilities due to too long time steps and poor control TO some extent it is due to poor edge conditioning (cf the 92436/839 case as shown by Thomas case)

Sun Hee: It can also happen when the radiative loss is high compared with the time-step. In this case, also reducing the time-step helps.

Pär: yes, so this is what I mean with poor Control (upper time step should be limited by physics not conveience)

Jorge: If you need to reduce your dt to 10^-12 then from the point of view of practicality it really doesn’t help

Pär: true but then are you using the correct tool for your physics problem - or your solver is not dealing with the stiff models

Jorge: But you can always try to calculate an approximate solution

Pär: GLF23 is by far the stiffest we have and that can force timesteps to very short but typically not to 10^-6

Pär: Or do the  trick one of out more used code did: switch to BgB when you run into probelms and then switch back when things has stabilised

Jorge: These kind of problems usually occur near to the edge and are usually related with sources/sinks Radiation, equipartition, etc

Pär: right. equilibration terms is not fully implicit in ETS I think

There are minutes attached to this event. Show them.
Powered by Indico v3.3.4