Implementing New Chemical Mechanism

[Me-Goosa]

I'm trying to simulate DDT combustion with ethylene and oxygen. I initially was using grimech30 as the mechanism however grimech is optimized for methane and specifies that it shouldn't be used for ethylene. Bit of googling and I found aramco1.3 and aramco2.0.

I'm trying to implement aramco1.3 and so I downloaded the files and producted a .yaml file and then used ceptr to make it usable by PeleC.

My trouble is that I'm getting convergence errors and I'm not sure what to do about it. Note that I wasn't getting this error with grimech30. I see that there are ethylene mechanisms available out of the box with PeleC, but they are skeletal and I am interested in the flame itself as it accelerates.
[ERROR][rank 0][/home/akresnya/PeleC/Submodules/PelePhysics/Submodules/sundials/src/cvode/cvode.c:3682][cvHandleFailure] At t = 0 and h = 4.98615e-18, the corrector convergence test failed repeatedly or with |h| = hmin.

One of the things I noted is that the the aramco13 mechanism has 'C3H51-2,3OOH' and 'C3H52-1,3OOH' in them which I got warnings for because of the , was making an issue with the defines. So I went back in and changed them to C3H5_123_OOH and C3H5_213_OOH and all of the references. This built with no warnings. Do you think this could be creating a species that it is hanging on and no converging or does the name not matter as long as it is consistent? (I did replace all on them).

I've attached a zip of my files. Have you implemented a new mechanism before?

PoreTube_Outlet.zip

[baperry2]

If you are refering to defines for the species IDs in mechanism.H, e.g. like

#define H2_ID 0
#define H_ID 1
#define O_ID 2
...

These are mostly just defined as a convenience and I don't think should impact anything within the reaction solve. You can verify that things are set up properly by looking at the species and reaction source term output in your plot files.

In general, when moving to a new mechanism, particularly a larger/more detailed mechanism, you may encounter new/different numerical stiffness and may need to select different reaction solver parameters. If using Cvode/GMRES, you can try tightening the tolerances or moving to a direct solve, which may be slower but more robust.

[Me-Goosa]

Thank you for your advice Bruce!

I have 2 follow up questions.
CEPTR responds by saying that there are PLOG reactions (pressure-dependent logarithm) in the mechanism and it asks for a reference pressure in Pa. When it produces the mechanism for PeleC it presents the units of CGS.

Is there a potential for a unit discrepancy, or are chemkin and cantera converted files using ck2yaml standardized unit wise?

Are the PLOG reactions compatible with a compressible flow solver?

I'm suspecting that the issue might have more to do with the mechanism because my inputs and tolerances all worked using grimech3.0 and it is returning this error when trying to solve for the very first time step.

[baperry2]

The way PLOG reactions are handled in Pele is to evaluate coefficients for a specific pressure, so the mechanisms converted for use in Pele are valid only at the provided reference pressure. This does present potential issues for using these mechanisms with PeleC rather than PeleLMeX. In practice, if your simulation is subsonic and has relatively small pressure variation, the error incurred by this approach in PeleC is likely to be manageable. For DDT, once detonations form you may have significant pressure variation that would make this approximation less appropriate. In that case, I'd recommend using a mechanism without PLOGs or running some tests for simple flamespeed and IDT cases across different pressures to quantify the error incurred at your particular conditions (similar to here: AMReX-Combustion/PelePhysics#549 (comment)).

If you use standard automated tools to go from chemkin to cantera to ceptr/Pele, any needed unit conversions should be handled automatically and you shouldn't have to worry about that.