How and which of the under-relaxation factors should I modify to avoid periodicity in residuals and/or forces?

Yours

Miguel

There are two schools of thought on this one, I have heard both from different people.

If on residual is not converging you can lower the U-R factors on 1) the residual that isn't converging or 2) all the other ones. I understand why both work, but have has most sucess with the former.

Thanks Andrew, nevertheless, I am afraid I didn´t get it all of it.

If i am iterating with k-e model and, for example continuity and x-velocity show some periodic movement, which of the following should I modify?

-pressure -density -body forces -momentum -Turbelence kinetic energy -Turbulence dissipation rate -Turbulent viscosity

Miguel

I would try dropping pressure and momentum, and if that dosen't work, try dropping all the other. When you say periodic, if you mean fluctuating, it might not affect your results. If the value of the residual is low enough or the variable you are interested in isn't changing, you can consider a solution "converged"

Well, if you are noticing a periodic fluctuation in your residuals, it could be that your flow is unsteady, and you are trying to use the steady-solver to solve it. If you suspect this is the case, and still want to use the steady solver, go first-order, since second order will not converge very well.

From what I found, if I reduce the momentum under-relaxation factor, the residuals will have a step change downward (about 1/2, but same order of magnitude), and then just level off again.

You were rigth Riaan!!

I turned the under-relaxation factors to the original values and turned the solver options from second order to first order and from second order upwind to first order upwind.

Residuals stayed somehow flutcuating but with a much lower value. But most important, the force values (drag) converged.

thanks!!!

Miguel

You should b careful that the converged force values you claim to have gotten are correct. In most cases I have found first order accuracy to be no good. yo should really strive for a 2nd order accurate solution. If your residuals continue to oscillate you should check your mesh. Plot the residual values for mass imbalance etc to see where the problem is on your grid.

Regards

Shaun.D

[soroush]

my model works well with defualt under relaxation values,but when i increase the pressure urf my droplet becomes distorted and loses it's circular shape!!!
i thought that urfs do not affect final solution and just increase convergence rate!!!
what should i do?