Problem in 3D Navier stokes solver
 

Hi everyone,

I am solving unsteady 3D Navier stokes equations on staggered cartesian grid. The solver is converging but the results are not matching with the Literature. What might be the reason? I have attached the profile obtained for Re Number 1000.


Thanks in advance.

suhas

check if you really reached a steady solution, then check if the divergence-free constraint is satisfied cell-by-cell. The BC is correctly imposed?
Furthermore, you case is a 3D lid-driven, what database are you using to compare your solution?

Hi,


Thank you for the reply.

1. It is reaching steady state.
2. It is divergence-free with residual 10e-10
3. I am doubtful about the boundary condition applied. I am using explicit BC method i.e I am updating the value of ghost cell after the iteration. Is this right?
4. I ran the same solution in OpenFOAM to compare it with mine and referred Feldman paper to match 3D results.


And the more important thing is I am using first order upwind which should be more diffusive. But I am getting a bigger secondary vortex as can be seen in the above attachment. How can this occur?

1. is reaching or reached?
2. cell-by-cell?
3. are you using a first order extrapolation?
4. I don't know that paper, I suggest to check for a more quantitative variable than the snapshot of streamlines. It is also advisable to check if a 2D solver provides the well known solution at Re=1000.

First order upwind can increase the tangential stress, more energy can be transferred to secondary vortices and they can become bigger.

1. It has already reached steady state.
2. I am checking divergence free for every cell and the max value of residual is 10e-10.
3. What exactly do you mean by extrapolation(Is it richardson extrapolation?).
4. I have plotted the centre line value of Ux velocity and tried to match with Ghia's result, but there is small difference due to the presence of that vortex which cuts that centre line.

you can not compare your 3D driven cavity with the Ghia's solution that is 2D... have you tested the 2D OpenFoam solver?

No I ran a 3D OpenFOAM problem. And one more difference is the solution in OpenFOAM was highly unsteady for that Reynolds Number. But for my case it was less fluctuating and reached steady state earlier than OpenFOAM's result.

it has no sense in comparing 3D and 2D solutions due to the strong effect of the lateral walls.... you have to do a 2D run..
A possible chance is to run a lower Re case (Re=400) using the 3D solver but setting periodic condition in the spanwise direction (z).

No I actually ran 3D OpenFOAM code to match it with my solver which is also 3D.

It is probably because of first order upwinding. Try, for example, Quick Scheme and then compare the results.

Yes I am changing that first order upwind now to deferred correction by giving blending between UDS and CDS.

You also can try central differentiating. If you use pressure correction, no wiggles will usually happen, as the pressure correction acts as stabilizer for the momentum. You can try, and see either your model works without wiggles for your applied Re number.