CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Running, Solving & CFD (
-   -   buoyantBoussinesqPimpleFoam gives incorrect results for non-steady laminar flows (

Simba February 28, 2012 14:00

buoyantBoussinesqPimpleFoam gives incorrect results for non-steady laminar flows
Hello everybody,

We consider classical 2D Rayleigh-Benard convection in a domain with aspect ratio sqrt(2)x1 (sqrt(2) is half of the length of the periodicity cell according to linear theory). At the lower and upper boundaries we set slip conditions ("Rayleigh conditions"), at the left and right boundaries there are also slip conditions (so there are no traveling waves and mean horizontal flow).
For the non-dimensional parameters Pr=11.5968, Ra=56931 there must be a periodic solution (we have already obtained it by means of our own pseudospectral method).
The flow modeled with our pseudospectral code (64x64) can be seen here
It can be considered as "exact" solution.

Now we take the hotroom case and transform it to the described problem, setting RASModel to laminar and corresponding dimensional parameters to
h = 0.1 (height)
L = 0.1*sqrt(2) (horizontal length)
blockMeshDict (60 40 1)
nu = 1.567e−6
beta = 3.072e−6
Tb = 277.2 (temperature at the lower boundary)
Tu = 276.8 (at the upper boundary)
Tref = Tb

This gave wrong results in buoyantBoussinesqPimpleFoam, as can be seen here
This regime obviously converges to steady mode.
BuoyantBoussinesqPisoFoam in OF 1.6 gave chaotic motion which does not seem to converge to periodic motion.
Playing with tolerances, nCorrectors, time-integrators, mesh did not give correct results.

Have anybody compared any of these solvers with time-dependent (periodic for instance) laminar flow?
I hope very much there is a solution to this problem...

Thank you!

olivierG February 29, 2012 07:22

This is an interesting test case, so if you can share your final conclusion & tests, this will be very estimate.

In the hotroom case, in fvSchemes, almost all div scheme use Gauss upwind ...
Maybe try with less diffusive 2nd order scheme ?


aaron_lan March 2, 2012 10:06

I also met some problem with buoyantBoussinesqPimpleFoam solver for transient laminar flows.

What I am modelling is the benchmark problem -- flow over a cylinder. I set RASModel to laminar and switch off the turbulent. Make Reynold number to 75 and temperature is set to 289 K in the whole domain. All turbulent parameters in the 0 subdirectory are set to 0, e.g. alphat, epsilon, k, kappat, nut.
Boundaries: left_inlet ( T=289 K, p_rgh buoyantPressure 0); right_outlet ( T zeroGradient, p_rgh buoyantPressure 0), up and down are symmeyPlane.

The weird thing is that the temperature near the backward of the cylinder decreases. And the velocity field also goes wrong.


smajer March 5, 2012 08:52

Which version of the OF do you use? Is it 1.6-ext or 'original' 1.6 supported by OpenCFD?

aaron_lan March 6, 2012 12:17


Originally Posted by smajer (Post 347692)
Which version of the OF do you use? Is it 1.6-ext or 'original' 1.6 supported by OpenCFD?

I am using the latest version 2.1.0. I am still struggling with this mixed convection over cylinder problem. With buoyantBoussinesqPimpleFoam solver, I modeled natural convection in a cavity and natural convection over a cylinder successfully. But when I come to the forced convection over a cylinder, simulations go wrong. The situation for mixed convection (combine natural convection and forced convection over a cylinder) is worse.

All times are GMT -4. The time now is 00:46.