How to observe transient convergence?
 

Dear All,

I am running a transient simulation and I am not sure how to observe the residual plots. I am assuming that my problem is converging for each iteration loop within each timestep, as I see the residuals drop. What I observe though on the plot is a straight line trend which does not drop (it makes sense because for each timestep the residuals drop to more or less the same level).

However I am in doubt, becuase the normal trend observed for a steady-state plot is one dropping (converging) as the residual error decreases.

My question therefore is how do I observe convergence for a transient simulation. How should the residual plot look like?

N.B. I am using cfx-5

Kind regards, Angelo

Re: How to observe transient convergence?
 

Hi,

You may look at the .out file to see the residuals evolution within each time step.

In CFX-5.7 and 5.7.1, in CFX-Solver, select the Worspace menu, Workspace properties, select the Global Plot Settings tab, and then check the Plot Coefficient Loop Data box to see the evolution of the residuals within each time step.

Regards,

Rui

Re: How to observe transient convergence?
 

Hi Rui,

Thanks for your help.

Angelo.

Re: How to observe transient convergence?
 

Hi Angelo,

The residuals reported during a transient run are different than those reporsted for steady state. In a steady state run, CFX reports the "outer" loop residuals. This is a measure of how well the non-linear equations have been solved. When the solution has converged to a steady state, these residuals are very small.

However, there is no assumption that a steady state will be reached in a transient simulation, so it makes no sense to look at outer loop residuals. Rather, in a transient simulation, CFX reports the "inner" loop residuals, which are a measure of how well the inner, linearized equations have been solved (essentially how well the transient timestep has been solved). These are relevant if you want a time-accurate transient.

So the answer to your question is don't worry about it. A transient is what it is and will not "converge" to a final solution like a steady state simulation will. If you are interested in whether your transient has acheived steady state, the best indicator would be to plot some variables of interest and see if they settle down.

Regards, Robin