Fluids

Fluids

Minimum concentration for reaction calculations – is this variable user-definable?

    • julia.hartig
      Subscriber

      Hey all,

      I've noticed in my reacting porous media simulations, where gas-phase molar concentrations are low and heterogeneous reaction rates are high, that reactions seem to be extremely low or even disabled for precursor_molefraction < 1e-6. I started looking through the ASCII portion of the case file to see if there were any variables that could be setting the cutoff for the species solver and stumbled upon this:

      (species/minimum-concentration 1e-06)

      I have tried repeatedly to find this in the Fluent GUI and cannot. Searches for "limits", "minimum concentration" etc. in the Fluent manuals give me thousands of results that I can't reasonably comb through. Is there a section in the GUI where these kinds of limits are usually set, or at least a portion in the Fluent manuals where these limits are defined?

      Thank you,

      Julia

    • Karthik R
      Administrator
      Hello:
      Unfortunately, there is no easy way of doing this as it cannot be done through TUIs.
      Any reason why you wish to look into this? Perhaps, you can start with a slightly different set of initial conditions to overcome this issue?
      Karthik
    • julia.hartig
      Subscriber
      Karthik That's too bad. The reason I want to adjust this parameter is because our system uses dilute concentrations of precursor gas that react with low surface area substrates (<1 m2/g). I'm concerned that the current cutoff (1e-6) is too high and affecting the species contours, so I'd like to do a parameter sweep and try cutoffs from 1E-9 -> 1E-6 to see how this affects our results. I'm not a huge fan of adjusting the initial conditions because I'm modeling a real system which is fundamentally dilute, so our current initial conditions are representative of the actual commercial process.
      I imagine that directly editing this line of text in the .CAS and .DAT files is not advised? I don't want to break any of the GUI features, but it's important that we are able to do a sensitivity test on this value...
      Julia
    • Rob
      Ansys Employee
      In theory you can alter the values in the case file, we don't recommend it, and if you do change the value it's at your own risk. The reason for the cut off is accuracy. As the volume fractions get small we're into the realms of numerical precision: 1e-6 is also the default to ignore some of the multiphase phase data.
      You won't find most of these values in the GUI as they're not meant to be changed. We do have some commands for this, but given the risks to the case we don't share (and definitely not on here). From a solver perspective, if you copy the case file and then alter the value all that can happen is you break one case/model and get some questionable results.
    • julia.hartig
      Subscriber
      Rob Understandable, I can see the justification in terms of preserving accuracy.
      Do you really feel that 1E-6 is delving into the limits of numerical precision? A 64-bit system should be able to handle floating point values down to 10^-308, correct? I can see the historical argument for a >1E-6 cutoff, but on a modern computer, I would think any number >>1E-12 is sufficiently resolved. However, whether our physical models are accurate enough at such dilute concentrations is a different question that I can certainly see an argument for :)
      The one time I tried modifying the text portion of a case file (i.e. to update the default file path for an animation sequence) the Fluent GUI refused to open and I had to delete the case and start over, so that's why I mentioned the "breaking GUI features" piece.
    • Rob
      Ansys Employee
      If you've got a really good mesh, very good convergence and appreciate the limits of the models then 1e-6 may be a little conservative as a limit. However, we're mostly dealing with less than ideal situations so it's probably sensible. I don't know the theoretical floating point limits, but suspect we don't store data to that precision: the data files would be rather large (I'm English, we tend to be fairly understated).
      Messing with paths in the case file may interfere with other features, and some of those are linked back into sections that aren't human readable. I used to edit mesh files (surface labels) manually so not everything causes failures, hence me suggesting to give it a try on a copied case file.
Viewing 5 reply threads
  • You must be logged in to reply to this topic.