Separating Transient Statistic Computations
 

In conducting a timestep sensitivity analysis of an unsteady flow, I would like to halve the timestep and continue the simulation.
The simulation has thus far been conducted in two runs. The first run, to equilibrate the domain and provide a valid initial condition. The second run (500 timesteps), wherein the transient statistic, Trnavg (arithmetic average), was enabled for pressure, velocity and wall shear. Both the “Start Iteration List Check Box” and “Start Iteration List Check Box” were not enabled.

I seek to now compute the Trnavg of the next 1000 timesteps without the prior 500 timesteps being included in the arithmetic average. The CFX documentation offers the statement:

“Note: If you want to re-initialize a given statistic (that is, remove the history from the statistic), you must shut down and restart the simulation with a new start (stop) iteration. This step is required to ensure that the new statistic accumulation interval is not included when searching for the earliest and latest start and stop iteration values, respectively.”

I find this vague. Can I achieve my objective using the two transient statistic definitions shown in image A and B?

As an additional question, is there a means to determine the timestep range a transient statistic has been averaged over in CFD Post (or otherwise)? This would go long way in probing the function of the transient statistic feature. Life would be a lot easier if one could specify the range to average over in CFD Post.

If I were to do it, I would use the 2nd setup (B), i.e. explicitly set the Start and End.

There is no mechanism in CFD-Post to determine the range for a given statistics set except by looking at the CCL in the output file for that specific simulation.

On your wish to compute the statistics in CFD-Post over a range of 5000 timesteps, it means the software must store the data for velocity, pressure and any other variable you are interested in 5000 times. Then, you would have been able to compute all kind of averaging ranges within the 5000 limit.

Do you need the statistics for the whole domain? or just a few points in the domain? For the latter, you can monitor the values at those locations only, and compute the statistics in the ANSYS CFX SolverManager, and change the statistics window at will during the run, or after the run has been completed.

Thank you for your reply. I will only use the 2nd transient statistic setup (B) as you recommend.

I had a look at the CFX Command Language (CCL) at the beginning of the output (.out) file

TRANSIENT STATISTICS: Transient Statistics 1
Option = Arithmetic Average
Output Variables List = Pressure,Velocity,Wall Shear
Start Iteration List = 5001
Stop Iteration List = 6000
END

This is useful, however, I’m concerned of the risk of the solver continuing from the prior transient statistics. The documentation confuses me in this regard.

I am storing transient results as well. How may compute averages of a specified range in CFD Post? This ability is valuable as it would allow me to cut off the timestep range to include an exact number of fluctuation cycles. I’ve so far resorted to exporting cgns files and performing such tasks in Paraview but this is time consuming.

I need such statistics over the entire domain. I’d like to be able to analyze any given plane of the flow later in my analysis.

If there is any inconsistency between the setup and the iteration/timestep counters, there should be a warning message in the output file stating the problem.

Unfortunately, there is no mechanism in CFD-Post to compute the statistics over the transient results file. However, a similar question has been posted in this forum a few times, and if you search you may find the approach on how to realize those calculations using a script in CFD-Post.

Understood. Thank you Opaque