[Specialist]

hello guys,

I simulate an air flow through a pipe. Right know I try to find the convergence point of my grid. For this I do following: Step by step I make the grid finer.

The first steps go along with no problems. The solutions convergences. But, at one point, the simulations does not convergence anymore, although I have nothing changed except the grid.

I work with SIMPLE FOAM.

I tried to change the time step, but that does not work

Do you have any idea, how I can stabilize my simulation? Are there any adjustable coefficients?


Regards

Specialist

[RodriguezFatz]

Hey Specialist,
SimpleFoam is a steady-state solver that does not recognize the size of the time step. Except for the names of the directories (such as \0 \100 ...) where the data is stored during calculation. So changing that doesn't work at all for your purpose.

What turbulence model do you use and what boundary conditions do you have?

Generally, the easiest thing is to use under-relaxation.
In your fvSolution file you need to have something like

relaxationFactors
{
    fields
    {
        "(p)"           0.3;
    }
    equations
    {
        "(U)"           0.7;
        "(k)"           0.8;
        "(omega)"       0.8;
    }
}

[Specialist]

so far, I do not use any turbulence model.

Due to the fact that the simulations works with an rough grid, I thought a turbulence model is not important?

I tried:

inletoutlet
zeroGradient

[RodriguezFatz]

What Reynolds number do you have for the pipe flow?

[Googlestuff]

Hi.

I would like to understand under-relaxation. Can you explain more about the relaxationFactors? What do they do exactly and how can i estimate the values for them? Or is it just arbitrary between 0 and 1?

Can i use them for an unsteady simulation as well?

Best regards,
Googlestuff

[RodriguezFatz]

Under-Relaxation is explained in every numerics book, this is a general concept for all iterative solvers. I guess, you will find a much better explanation than I can give you. Anyway:
You want to solve some equation, such as f(x)=0.
Since you don't know the result, you will try some x_0 as intial value.
Now, you put x_0 in your equation and will find out, that it is not satisfied, i.e. f(x_0) = e_0, with e_0 an error.
Hopefully you are already close, so you need to add or substract something from your x_0, such as
x_1 = x_0 + dx_0.
Since you don't know the correct size of dx_0, you will get a new "wrong" value that also does not solve the equation but maybe it is closer, so
f(x_1) = e_1 < e_0.
You keep going until the error e_i is small enough. This is iterative solving.
Under-relaxation means that you don't use the dx_i that are calculated by your solver to correct the x_i, but that you use
x_2 = x_1 + alpha * dx_1, with alpha the under-relaxation factor, if alpha < 1.
It is called over-relaxation factor, if alpha > 1.
Under-relaxation is used, when your new values x_i+1 tend to be even worse than the old values x_i, since your dx_i was too large. You "jumped" over your target value. It stabilizes the simulation but generally makes it slower.

Under-relaxation for SIMPLE algorithm are normally about the values I proposed in my post. Seems to be a good guess for most cases.

[Specialist]

ok guys. you helped me a lot!

now it is time for pizza