Learning Forum

[vuminhduc280903]

Hello everyone,

I'm simulating combustion in a micro-combustor using EDC model + Species Transport, combined with Realizable k-ε turbulence model + Enhanced Wall Treatment on both poly-hexcore and hexahedral meshes. I encountered significant discrepancies between two approaches:

Approach 1 (FMG → Flashback): FMG initialization (4 levels) → Pressure Standard + Full First-Order scheme → switch to Full Second-Order results in flashback phenomenon with flame root moving upstream ~30% of domain.

Approach 2 (No FMG → Correct): Full Second-Order from start with hybrid initialization gives correct flame position, no flashback.

Could you please help me understand: Does FMG + first-order cause numerical diffusion that smears fuel upstream, creating artificial flashback? Does pseudo-time method only improve convergence, or does it affect steady-state accuracy? (Does CFL need precise tuning?) And what's the optimal strategy for highest accuracy, shortest runtime, and 0% divergence rate for EDC micro-combustor simulations?

Thank you to anyone with EDC/micro-combustor experience for your specific advice!

[abtharpe42]

From what I understand from my own experience doing combustion CFD modeling as a current PhD student, you never want to use the "Standard" option for Pressure, even for a non-reacting flow simulations because it's supposed to be way less accurate. Stick with at least Second-Order or PRESTO! (more computational expensive) for Pressure. I noticed in your post that you changed the discretization for everything when enabling or disabling the FMG initialization. I bet if you keep model settings consistent, you'll get similar simulation results between FMG and hybrid-initialization. 

And honestly, for combustion flow just stick with hybrid or standard initialization as those two easily suffice. I believe FMG has a species limit anyways, which is unfortunate because the EDC model is meant for mechanisms with a lot of species.

For steady-state, the choice for Pressure-Velocity coupling is something that I still struggle with. The Coupled solver would almost always be the preferred option, be it Pseudo-Transient enabled or disabled. The Coupled solver solves the pressure and velocity field together, which is robust for a flow that is very dynamic, like a combusting flow, and is generally much faster to converge than the SIMPLE solver 99% of the time. The problem with the Coupled solver is that it is computationally heavy because of the inner iterations per simulation iteration that it does to resolve the calculations of the flow equations, so it uses a lot of RAM on your computer. For simulations with stiff chemistry combustion (many reactions and species) like what you should be using the EDC model for, which is computationaly intensive on its own, this extra load could really bog down your computer and the simulation could take a while to progress.

The SIMPLE solver would be the alternative to Coupled in this case if your computational resources are low. It solves the pressure and velocity field seperately per iteration, so it uses considerably less RAM. This segregation of the fields makes the simulations less stable and robust, especially with combustion, and the simulations generally take much longer to converge than those using the Coupled solver. But, if you are low on computational resources this option may be the way to go.

If you disable the pseudo-transience of the Coupled solver or enable the Local Time Stepping option for the SIMPLE solver, you have to choose a CFL. I still haven't figured out how to know what optimal values to choose for either solver. For Coupled, I usually reduce the CFL from 200 to 50. For SIMPLE, I usually leave it at the default of 5. I don't ever really mess with the Relaxation Factors, but the defaults may not be optimal for every simulation, especially combustion. Adjust them however you want if you chose.

Finally, could please post pictures of your geometry/mesh, at least a contour or two, and what chemical mechanism you are using? That would really help anybody looking at your post better understand the situation, which would then allow them to better assist you. Hope anything I said above helps you in anyway.