simpleFoam: simple 1-D channel flow, yet very strange convergence behavior
 

Dear FOAMers,

I am new to OpenFOAM - so to start with, tried simpleFoam on one of the simplest of the flow situations - namely laminar flow between two parallel plates. The domain is simple rectangular - 0.5 m long and 0.05 m wide. I started with a 10 x 5 grid (10 cells lengthwise and 5 cells breadthwise) and noticed a very weird convergence behavior. The issue is that the case converges for some inlet velocities, but diverges for some others - even for this coarse grid. More specifically:

Converged perfectly for inlet velocities (V_in) upto 0.05 m/s. Beyond that:

Diverged for V_in upto 0.1 m/s (I tried 0.06, 0.07, 0.08, 0.09)

Diverged for V_in = 0.1 m/s, but CONVERGED for V_in = 0.1085 m/s

Diverged for 0.15 m/s, but converged for 0.2 m/s, again Diverged for 0.25 m/s

Diverged for 0.5 m/s, but Converged for 0.49 m/s

Converged for 1.0 m/s, but Diverged for 0.99 m/s

I see similar pattern for a 30 x 8 grid, only the above numbers are different.

Boundary conditions:

Inlet: fixedValue on U, zeroGradient on p

Outlet: zeroGradient on U, fixedValue (uniform 0) on p

Can someone help me understand what is going on here? Many thanks indeed.

- Kumar

P.S: I've attached my blockMesh dict and fvSchemes/fvSolution.

Sorry: I forgot to mention that the fluid used was water with kinematic viscosity 1.0 x 10^-6

Hello,

You should have provided with the Boundary condition files. Anyways, in your case a proper initialization of the solver is not done.

By initialization I mean to say like:

For velocity = 0.1 m/s,
BC: Inlet: fixedValue on U, zeroGradient on p.
And for initialization: internalField uniform (0.1 0 0);

The problem which you have mentioned is something unrealistic in CFD unless some parameters are missing.

Hi Tushar,

Many thanks for your time and help. Sorry I didn't include the 0/U file - I used U(0 0 0) to initialize; my p and U attached for your reference.

You are absolutely right :)- the case converges without any glitch for the initial U field (0.1 0 0), in a small no. of iterations. Isn't it because the initial guess is so close to the converged solution? Shouldn't the code be able to seek the right solution even for a bad guess, say (0 0 0). Field initialization based on the inlet value may not be ideal always (what if the pipe has a 90 degree bend or worse, if the flow recirculates in the domain?)

I was particularly confused why the instabilities develop so quickly for certain velocities and not for others? for comparison, see the cases of 0.1 (blows up) and 0.1085 m/s (nicely converges)

Thanks and regards
kumar