steady state solution
 

Hi,

Does steady state solution depends on your initial guess?

cfxdude

Re: steady state solution
 

A complete converged solution (i.e. constant values in monitoring points, imbalances=0, RMS residuals < 1e-6) should not depend on initial guess.

Gert-Jan

Re: steady state solution
 

Hi,

What you are saying is that If 1e-5 < RMS residuals (pressure and velocities) <1e-4 and 1e-6 < RMS residual (k-e turb) < 1e-3 is not a steady state solution, or the solution depends on the initial guess?

In my case the velocity of the monitored point is fluctuating between 0.25 - 0.35 m/s (11000 iteration), but the volume average velocity is constant. which one is the correct steady state indication.

Thanks.

cfxdude

Re: steady state solution
 

When your monitoring values are fluctuating, you won't have a steady state situation. Looks like your problem is time dependent.

Gert-Jan

Re: steady state solution
 

So constant volume average velocity is not the steady state indication.

Is there any book or paper about these, I am looking for some thing written.

Thanks for your time.

A.A.

Re: steady state solution
 

Hi,

For a small number of situations the steady state solution does depend on the initial guess. I have seen some flows were the Coanda effect caused a jet to stick to one wall or the other depending, then settle down to a steady state solution. Which wall the jet stuck to depended on the initial guess.

In general, however, a fully converged solution is independent of initial guess.

Glenn Horrocks

Re: steady state solution
 

If the velocity of the monitored point is fluctuating between 0.25 - 0.35 m/s after 11000 iteration, try to increase gradually the timescale. In this way you should obtain a non fluctuating velocity value.

Probably now you are reading some fluctuations of your domain; increasing the timescale works as a filter for this numerical fluctuations.

Bye

Alberto

Re: steady state solution
 

Numerical fluctuations? You may damp out physical fluctuations by numerics. If you increase the time step you may stabilize the solution but you will not get a fully converged (grid independant and time step independant) solution.

I agree with Gert-Jan and Glenn