Error when running multiple temperatures for adiabatic flame simulation

[ajuluc]

Error when running multiple temperatures for adiabatic flame simulation

[ischoegl]

Thanks for some external testing of this code! (it runs fine on my machine, but it hasn't been used much at this point yet). Hopefully it'll get you the results that you need.

The issue is that you had your YAML input set up for a variation of a large number of equivalence ratios, where some parameters may have poor convergence.

The default field specifies scalar inputs, whereas variation selects the entry that is varied; in your case:

# YAML file example for the `adiabatic_flame` module
# file specifes a parameter variation of temperature
variation:
  entry: upstream.T 
  values: [300.0, 360.0, 400.0, 403.0, 470.0, 500.0]
output:
  format: h5
defaults:
  upstream:
    T: [300., kelvin, 'temperature']
    P: [1., atmosphere, 'pressure']
    phi: [1.0, dimensionless, 'equivalence ratio']
    fuel: 'nC10H22(1)'
    oxidizer: 'O2(2):0.21,N2:0.79'
  chemistry:
    mechanism: chem_annotated.cti
  domain:
    width: [30, millimeter, 'domain width']
ctwrap: 0.1.0

I'm not familiar with SLURM, but I'd suggest to run a couple of tests on a standard Linux system (or a virtual machine/container, which are easy to set up), so you can troubleshoot prior to releasing jobs that are more difficult to debug on a cluster.

[ajuluc]

Hi @ischoegl thanks for the help with this and debugging the yaml file. I didn't realise the variation and defaults field until now. Thanks for pointing that out to me. What I was really trying to get was a code that had a fixed temperature and varied equivalence ratios (phis) but I guess I could do that by adding more lines where default is fixed then another section of the file where default is incremented to a different value with that fixed and varying phis.

So SLURM is really the Simple Linux Utility Resource Management for high performance computation. So it's a bunch of supercomputers that we access to do heavy computational calculations that otherwise cannot be performed on a simple local system like a Mac or Windows.I'll try it out again on my local then send it back to SLURM.

[ischoegl]

@ajuluc ... ok, got it. At the moment, I have only implemented a single variation, so the approach would be to have one YAML file per temperature.

[ischoegl]

PS: my own experience with marginal cases (phi close to flammability limit) is that they oftentimes do not converge easily, which can cause extremely long runs (without ever converging). So some workers may finish, while others may time out.

[ajuluc]

Ah thanks for letting me know about the flammability limit convergence issues. I was planning on starting off at 0.75 and stopping around 1.8 maybe that might help

[ischoegl]

Not sure whether the issue persists - closing.

[ischoegl]

Opened #25 to address variation of multiple parameters.