[TommyM]

Hi All,
I am studying the flow inside a supersonic nozzle by using rhoPimpleFoam and rhoCentralFoam.
As far as I know, rhoCentralFoam is the correct solver to be used in supersonic flow simulations, but I found out that both solvers give the same solution (N.B. this happens only if there are no shock waves!).
RhoCentralFoam requires an extremely huge computational cost w.r.t. rhoPimpleFoam, so I am wondering if rhoPimpleFoam would be a better choice.
In other words: is rhoCentral the best choice for simulations of supersonic flows or of supersonic flows with strong shocks?
Thanks in advance.


Tommy

[EleGiova]

Hi Tommaso!
I think that as for now you have answered your own question.
But I would like to answer anyway, I am still a student of CFD and I think this could be a nice exercise for me, discuss about CFD with others in order to understand it better.


I think that it depends on what you want to study and with which approximation..
I am now using rhoCentralFoam to run simulations with hypersonic flows and the difference is enormous wrt rhoPimpleFoam.. In this case the standoff distance is completely mispredicted and this leads to a higher Temperature = you get wrong the design of the ablative shield of a re entry vehicle, for example.
Maybe in transonic cases or slightly supersonic this error is not so huge, but if you have a shockwave how can you capture the shock if you don't have a shock capturing method (like KNP you have in rhoCentral)..

[TommyM]

Hi EleGiova,
thanks for your reply.

I do not know your geometry and case but I think that rhoCentralFoam is always able to produce accurate solution for supersonic flows. On the other hand, rhoPimpleFoam is able to reach the same accuracy only in some cases, while it leads to wrong results in other cases.
However, the computational time required by rhoPimpleFoam is extremely smaller than rhoCentralFoam (in my case, 5 hours vs. 5 days), thus when the use of rhoPimpleFoam is allowed and provides good results, I think that it is the best choice.

In my experience, I studied a supersonic nozzle which reaches Ma = 2.5, it is not hypersonic but neither slightly compressible. I found out that rhoPimpleFoam gave me the same accuracy of rhoCentralFoam with a lower cost.
In the case of shock waves, as you suggested, this reasoning fails.

[JBeilke]

You might look at pimpleCentralFoam or rhoPimpleCentralFoam .

https://github.com/unicfdlab/hybridCentralSolvers

[TommyM]

Thanks Joern,
I heard about these hybrid solvers but I did not use them since they were not present in the official release of OpenFOAM. Anyway, I am interested to know more about them since they should be the trade-off between a pressure- and density-based solver.

[dlahaye]

Dear TommyM,

Your observations on the difference between rhoPimpleFoam and rhoCentralFoam are very valuable for me to read!

I am wondering whether you could expand your comments.

1/ Does the difference in computed results between rhoPimpleFoam and rhoCentralFoam increase with Mach number (as I imagine it does)?;

2/ Does the number of iterations of rhoPimpleFoam to reach same convergence levels increase with the Mach number (as I imagine it does)?;

3/ How is the convergence of rhoPimpleFoam affected by the choice of the thermodynamics (rhoThermo vs. psiThermo in constant/thermoPhysicalProperties). I imagine psiThermo to be able to handle larger Mach better (i.e. more similar to rhoCentralFoam).

Thanks for the additional input.

Domenico.

[JBeilke]

Originally Posted by TommyM

Thanks Joern,
I heard about these hybrid solvers but I did not use them since they were not present in the official release of OpenFOAM. Anyway, I am interested to know more about them ...

I used them a lot for transonic flow in airbag-gasgenerators and for the simulation of transonic effects in the cylinder of race car engines. They are very accurate and also very stable, as long as the courant number is around 0.5


For pure subsonic flow the courant number might be larger.

[TommyM]

Hi Domenico,

In my opinion:

1) I never tested hypersonic flows. I tried with Mach from 0.9 to 2.5 and the difference does not increase with the Mach number. I think it regards the flow structure (shock waves, ...) and not simply the Mach, but keep these words with caution because it is only my opinion. It would be interesting to simulate a hypersonic flow without shocks comparing the two solvers.

2) Yes. If the velocity increases you have to decrease the Courant number or, alternatively for rhoPimpleFoam, to increase the corrector loops.

3) I am sorry but I don't know.

I hope these comments are correct and can be useful for you.


Tommy

[febriyan91]

Originally Posted by TommyM

Hi Domenico,

In my opinion:

1) I never tested hypersonic flows. I tried with Mach from 0.9 to 2.5 and the difference does not increase with the Mach number. I think it regards the flow structure (shock waves, ...) and not simply the Mach, but keep these words with caution because it is only my opinion. It would be interesting to simulate a hypersonic flow without shocks comparing the two solvers.

2) Yes. If the velocity increases you have to decrease the Courant number or, alternatively for rhoPimpleFoam, to increase the corrector loops.

3) I am sorry but I don't know.

I hope these comments are correct and can be useful for you.


Tommy

Hi Tommy, Did you use turbulent model or only laminar for both rhoCentralFoam and rhoPimpleFoam ?

[TommyM]

Hi Febriyan,
I used turbulent models (k-omega SST) for both.

[dlahaye]

Dear TommyM,

Thanks again for your valuable input.

I will continue to look into the differences between rhoThermo and psiThermo and share my findings on this forum.

Kind wishes, Domenico.

[shock77]

Hi,


If you want high quality solutions, I would always use a density based solver. For compressible flow, it makes more sense to use a density based solver in general.


Pressure based solvers are guessing and correcting the pressure, instead of calculating it directly. Density based solvers however calculate rho from mass conservation, u from the momentum equation and T from the energy equation. P is then calculated from the ideal gas law, peng robinson or whatever. Since in compressible flow information is also transported by the propagation of waves, it makes more sense to me.




Kind regards,
shock77

[dlahaye]

Yes and no (and more of the latter).

I do agree that in case of unbounded computational resources, density-based is preferred.

In the application that I target however, the Mach number is bounded by 1.5. I am trying to understand the trade-offs in computational efficiency of pressure-based vs. density based. That is, I am trying to understand modeling errors and gains in computational efficiency (in any) in replacing density-based by pressure-based.

D.

[pduytung]

Originally Posted by TommyM

Hi Febriyan,
I used turbulent models (k-omega SST) for both.

Hi TommyM,
When you use turbulent models (k-omega SST) for both, can you generate turbulence with rhoCentralFoam.
In my case, it is same with link below.
rhoCentralFoam cannot generate turbulence while rhoPimpleFoam can do it.
However, the result of rhoCentralFoam is more stable so I would like to simulate turbulence with rhoCentralFoam.
Do you need to modify any particular things?

Best regards,
Pham

Question about rhoCentralFoam and rhoPimpleFoam?

[baratian]

Originally Posted by TommyM

Hi Febriyan,
I used turbulent models (k-omega SST) for both.

hi Tommy
i use sstkomega with rhocentralfoam but turbulent results are very different from the same with fluent!
could you tell me why that is?