Learning Forum

[m.d.lasica]

I'm trying to run WMLES simulation of portion of hydrogen-air combustor (single nozzle pair), where hydrogen is injected perpendicularly into air stream, for TCI I'm using EDC model. At the beginning I would like to add that I'm not the specialist in LES.  First, I've run the RANS cold flow precursor, where the temperature is limited to air and fuel temperatures. Everything converges nicely. I analyzed the mesh using the integral length scale and densified it in regions where the ratio of integral length scale to grid size is less than 10. Y+ is kept below 1, while x+ and z+ are close to 25. Then I switch to LES, set the second order numerics and the second order bounded time discretization.  Due to the high hydrogen velocities (~160 m/s) the timestep has to be very low ~ 1e-8 to keep the Courant number below 1. But for the first few simulations I'm using lower timestep 5e-9. However, when I run simulation, the solution diverges very quickly and triggers floating point error. To prevent from it I reduced to relaxation factors, which helped, but on the other hand solution doesn't converge now.

Does anyone have any idea why it diverges so abruptly or any tips how I could fix it? Is the EDC model suitable for use with LES? So far it is only cold flow and it causes so many problems.

Any help would be greatly appreciated.

[jcooper]

Hi m.d.:

The EDC model seems to cause a lot of convergence problems. It is also tuned for methane combustion in Fluent. 

While TCI seems important in theory, most H2 problems seem to solve better with finite rate chemistry and the thickened flame model.  I would recommend trying that approach with a good hydrogen chemistry mechanism.  I would also look into the LES best practices to ensure that your numerics settings in Fluent are optimized for LES.

https://www.ansys.com/content/dam/product/fluids/cfd/tb-best-practices-scale-resolving-models.pdf