[Mat_fr]

Dear Foamers,

I am investigating the adiabatic compression of air (assumed to be a perfect gas).
During this process, the pressure must follow the law PV^Gama = constant.

To simulate this problem, I am squeezing my computational domain (DyM).
I use zero pressure gradient everywhere and zero temperature gradient also (adiabatic).

When using rhoCentralDyMFoam, the pressure satisfies approx. the PV^Gama = constant law.
As it is a density based solver, the drawback of this approach is the appearance of pressure wave and the need of small time step.

When using compressibleInterFoam however, there is a large underestimation of the pressure.

For rhoCentralDyMFoam, the thermophysical properties are:

thermoType
{
type hePsiThermo;
mixture pureMixture;
transport const;
thermo hConst;
equationOfState perfectGas;
specie specie;
energy sensibleInternalEnergy;
}
with standard properties (Cp, Pr, M) for air.

For compressibleInterFoam, the thermophysical properties are similar except for the type which is "heRhoThermo" in this case.
Would the type explain the difference in pressure between both solvers for the same simulation case ?
hePsiThermo is not available with compressibleInterFoam though.

I will gladly welcome any feedback on this issue.

Kind Regards,
Mat

[olivierG]

Hello,
You almost find your answer: compressibleInterFoam is a pressure based solver (heRhoThermo type), rhoCentral is a density based solver (hePsiThermo type). You can't swap a pressure based solver with a density based thermo type.

regards,
olivier

[Mat_fr]

Hi Oliver,

Thank you for your answer.
Would you know why the air is not following a perfect gas behavior with the pressure based solver then ? (although I specified "equationOfState perfectGas;")
Regards,
Mat

[olivierG]

hello,
Why do you use compressibleInterFoam, and not rhoPimpleFoam ?
because I guess you may have artificial second phase creation, which can explain a wrong PV^g conservation.
But the first candidate is that with pressure based solver, you should have a good convergence on pressure to get a good perfect gaz behaviour.
regards,
olivier

[Mat_fr]

Hi Olivier,
Thanks for your time!


I am using compressibleInterFoam, because I want to run two-phase simulation afterwards.
For the moment, it is single-phase only indeed, and I'm sure there is no second phase creation.


However, looking at the residuals, you are right, I have an unsatisfactory convergence behavior for pressure.

I've tried playing with the time step and the number of corrections steps, but without success at the moment.


Regards,
Mat

[dygu]

Hi Mat,

I am working on a similar problem where I am using compressibleInterDyMFoam to squeeze a box of air using a moving wall approach and a sinusoidal pointDisplacement. In the images below, a pressure probe in the middle of the domain is compared to:
p = p0*(V0/V)^(1.4) where V=V0-amp*sin(2*pi*f*t) is the instantaneous volume based on a sinusoidal motion of one wall of the box.

I have found that I can get the anticipated PV^gamma behavior but only with very small time step size using the Euler dt scheme or by compiling the solver with the backward dt scheme option and running with a more reasonable dt at least 4x larger than with Euler. The images below compare the Euler and backward schemes using the same dt=0.01s.




I understand that the backward scheme can be unbounded and this is likely why it is not an option for the multiphase solvers. In a case of a cube of air with perfect hex cells, this scheme works very well. On more realistic industries grids with non-orthogonality around 65 deg and non-zero skew, the pressure / T solve blows up quite quickly with negative temperature when using the backward scheme.

I wonder if anyone has any hits as to how to stabilize the compressible multiphase solver with a 2nd-order time discretization scheme? I have tried the CrankNicolson scheme but it is very unstable unless the Euler-biasing is quite high and thus the benefit of the 2nd-order scheme is mostly lost.