topic Re: transient analysis: time step input and output in CFD Forum

transient analysis: time step input and output

AstroJohnPE — Tue, 24 Feb 2015 21:17:22 GMT

Hi,

I am trying to understand the time steps that are output compared to my input for a transient analysis. I am using the results from a steady state analysis as the starting point for my transient analysis. The input for the solver is shown in this image:

These results are available after the analysis completes:

Here are my questions:

Since the "Save Interval" is set to 1 second, where are the results at Time = 1?
I think CFD has sometype of time step control. Is this why the time step size apparently changes? (30 iterations to get to 2 seconds, 30 iterations for the next second, 40 iterations for the next second, 30 iterations for the next 0.6241 second)
Is the time step constant during the "10 Inner Iterations"?
Since I am specifying a "Stop Time" and not a number of "Time Steps to Run", my understanding is that as many time steps as necessary will be performed, regardless of the automatic time step size, in order to reach the stop time. So, what causes the analysis to stop long before it reaches the stop time? This is occurring in another analysis in which I am entering similar time step information. In more than one study, the transient analysis has stopped when the time was less than 0.5 seconds. In one study, there was a message (in the Message Window) about the results have flattened, so I assume it stops rather than outputting the same results for all of the time that I requested. But in two other cases, the analysis stopped with no indication of why. Is there another file and text that I should search for?

Thanks.

Re: transient analysis: time step input and output

AstroJohnPE — Tue, 24 Feb 2015 22:58:05 GMT

Here's some more information related to item 4 in my original post (why does the transient analysis stop before reaching the specified stop time?):

I ran 5 studies of the same model, different boundary condition (flow rate) on one face, and using the steady state results from the appropriate steady state study.
I used the "Solver Manager" to run each one in batch model.
After the new analysis started, I loaded it into Sim CFD and watched each one run.
In 4 of the studies, the "Convergence Plot" extended from 100 iterations (the number from the steady state solution) to 300 iterations. As the number of iterations got to 300, the plot increased by 5 iterations at a time.
In 1 of the studies, the "Convergence Plot" extended from 100 iterations by 5 iterations at a time. It did not jump to 300, then increase by 5!
All 5 studies stopped after reaching iteration 400. The times were between 0.240 and 0.414 seconds. This is well short of the 2 second "Stop Time" that I specified in the solver dialog.

Any clues?

Re: transient analysis: time step input and output

Jon.Wilde — Wed, 25 Feb 2015 11:31:34 GMT

Hi,

Take a look under Solution Controls. Is Intelligent Solution Control on? It would need to be off to keep a continous time step size. You could also go into Advanced and turn off Auto Convergence Assessment.

Also, I would think that in most cases, 10 inner iterations is overkill. I tend to sue 3 and ensure that my timestep is small enough to capture whatever is changing. You might also like to think about utilising the AutomaticInnerIteration flag.

Re: transient analysis: time step input and output

AstroJohnPE — Wed, 25 Feb 2015 14:01:18 GMT

Thanks for the tips Jon,

Yes, the Intelligent Solution Control was on. I'm not concerned about a constant time step size. In fact, I'm glad that the software can decrease (and presumably increase) the time step size based on the convergence. My interest is why the analysis stops before reaching the specified stop time.

It sounds like the Intelligent Solution Control uses the entered "Time Step Size", "Stop Time", and "Inner Iterations" to calculate how many iterations it should take, assuming the time step is not reduced. In my case, (2 sec)/(0.1 sec/step)*(10 iterations/step) = 200 iterations. The solver then runs that many iterations and stops regardless of whether the time has reached the stop time or not. Is this correct? BTW, this behavior is not what the documentation indicates.

This hypothesis would explain why the Convergence Plot started by showing an additional 200 iterations. But it does not explain why the analysis actually performing 300 iterations and then stopped. It did this in 5 different studies, so there must be a setting (either hardcoded or set somewhere in the interface) that controls the maximum number of iterations.

Thanks again.

P.S. I forgot to mention that I am using CFD 2014.

Re: transient analysis: time step input and output

Jon.Wilde — Wed, 25 Feb 2015 14:21:07 GMT

Try turning off the auto Convergence Assessment, is the solver saying it has reached a solution and is stopping?

No, the ISC monitors the convergnce and can tweak timesteps to try as best it can to converge well and avoid any divergence.

If you were running auto forced convection, CFD would run more thermal only iterations after solving for flow. Might that be the issue?