[Tobi]

Hi all,

is there a thread or paper discussed the SIMPLER algorithm and maybe its implementation into openfoam?

regards
Tobi

[Simurgh]

I know this thread is years old, but I am dealing with a project at hand, and OpenFOAM doesn't seem to have a SIMPLER solver. Have you found anything by any chance?

[Tobi]

Nope. However, FOAM has SIMPLE and SIMPLEC

[Simurgh]

Looking at OpenFOAM SIMPLE implementation (I am looking at openfoam.com, but main openfoam from openfoam.org shouldn't be that different), the issue is that we have 1 relaxation for pressure only:

// Explicitly relax pressure for momentum corrector
p.relax();

Because SIMPLE(C) uses an approximation for correcting velocities, and it has only 1 pressure correction step, and no velocity relaxation (equation relaxation is not the same as field relaxation) then this will cause simpleFoam to diverge quickly for pressure-driven simulations.

To the best of my understanding, the only way to solve pressure-driven simulation is to use pimpleFoam right now. And, pimpleFoam will take much longer to simulate a steady-state than SIMPLER algorithm.

Maybe I should invest some time to implement SIMPLER in OpenFOAM. You are an expert in this area, do you see a use for it?

Thanks

[Tobi]

How much relaxations you would like to have for the pressure field? I am sorry, but I don't get your issue.

[Simurgh]

The issue is not relaxation. I can apply any level of relaxation to p field, and SIMPLEC technically doesn't need relaxation. But, simpleFoam simply diverges for laminar simulation where you have a pressure-driven simulation. SIMPLER should be more stable.

A pressure-driven simulation is when you do not specify velocity at the inlet, and instead you use a stagnation pressure boundary condition (totalPressure for example) at the inlet.

As an example, I have changed boundary condition for Pitz-Daily text book example that comes with OpenFOAM source code. If you run it as is, it will diverge after 70-80 iterations of the simpleFoam solver. If you change the totalPressure for inlet to -1 pa for example, then the solution will converge. Note that pressure relaxation is set to 0.3 here. The solver should not diverge based on small boundary condition changes.

If you use a RAS turbulance model, the simpleFoam will converge. But the advantage of using laminar is that you can see where the flow starts to separate from your surfaces while a turbulance model will hide all these details. Alternative to this (which I have not tried yet) is to use Large Eddy Simulation (LES).

A simple use case is this: imagine you are simulating enterance into a device from a plenum and you want to see where the flow separates and modify the geometry to avoid that.

Regards

[Simurgh]

sorry, the size of the attachment was too big. I just attached directory 0. The rest of the simulation is the same as pitzDaily example except turbulance model is turned off (laminar simulation).

[Tobi]

I know what pressure driven flows are. I did a lot of simulations using pressure boundaries without specifying any velocity. I don't have any issues regarding such problems.



Regarding your problem. I cannot investigate into that as you don't provide any information. The community cannot support you by showing the boundaries. I expect that the problem you have is related to something else. A pressure difference of 10 Pa is nothing and should work out of the box.





Simple example, I am investigating into a cooling system for the Enders 3 v2 and this is only a pressure driven analysis - a bit more complex because it is a fanPressure and a fanPressureJump (as I do have two fans here). simpleFoam works out of the box and is stable. Pressure Differences are around 12 Pa (and is varying). I did not check if laminar works but if you are using a steady-state solver, the separation is not predicted correctly in any case.

[Simurgh]

I guess at this point I will use simpleFoam with a RAS turbulance model as this works well with pressure-driven simulation. However, I still think a SIMPLER solver would be much more stable with laminar model at higher Reynolds.

The boundary condition I attached is for OpenFOAM "tutorial" pitzDaily. It is not my simulation. I just changed the following tutorial case to make it a pressure-driven case:

$FOAM_TUTORIALS/incompressible/simpleFoam/pitzDaily

If you decide to try it, you also need to make it laminar to see the issue. At 10 pa pressure difference simpleFoam will diverge. At 5 pa, it will eventually converge.

Thank you for your help