[sah2105]

I am trying to solve steady state incompressible 3d lid driven cavity using SIMPLE algorithm. I have implemented CDS, FOU, SOU and QUICK schemes. I keep getting divergence for Re=400. The code converges for Re=100. Can someone tell me where might be the problem?

[FMDenaro]

Originally Posted by sah2105

I am trying to solve steady state incompressible 3d lid driven cavity using SIMPLE algorithm. I have implemented CDS, FOU, SOU and QUICK schemes. I keep getting divergence for Re=400. The code converges for Re=100. Can someone tell me where might be the problem?

how about your grid resolution?

[sah2105]

Hi FMDenaro,

Thanks for your reply.

I tried refining the grid and ran simulations till 120x120x120 with no convergence. I came across some literature where the author uses third order Upwinding and has obtained results for 81x81x81. The only difference being in the formulation of equations; he has used Poisson's equation whereas I have used SIMPLE algo. Does this difference sound reasonable or is there something wrong with my code?

[arjun]

Originally Posted by sah2105

Hi FMDenaro,

Thanks for your reply.

I tried refining the grid and ran simulations till 120x120x120 with no convergence. I came across some literature where the author uses third order Upwinding and has obtained results for 81x81x81. The only difference being in the formulation of equations; he has used Poisson's equation whereas I have used SIMPLE algo. Does this difference sound reasonable or is there something wrong with my code?

If your flow is incompressible then you are also solving Poisson equation in SIMPLE.
The main thing is you might be using pressure correction method while he may be using projection method.

[sah2105]

Yeah, you are right. So, is this discrepancy based on grid resolution possible?

[FMDenaro]

Consider the fact that while the 2D case at Re=400 is steady, the 3D can not and upwind can simply add numerical viscosity and makes unphysically steady the solution... Have you tried to use periodic spanwise condition to check the 2D solutions up to Re=1000?

[sah2105]

Hi FMDenaro

Sorry for this late reply, I was busy with my coursework.

I tried out your suggestion and it worked out. The QUICK solver for 3d lid driven with periodic spanwise condition is working well for Re=400 (50x50x50 grid) and Re=1000 (100x100x100).

Should I try changing my discretization scheme or add some term to decrease artificial dissipation for the 3D case?

[FMDenaro]

You can try the QUICKEST and refine the grid in the confined 3D case