[aerost]

Hi,

I am running a transient turbomachinery model with a stator followed by a rotor. I have been slowly decreasing the Timestep and increasing the number of coefficient loops to achieve convergence, but am still not sure that I am getting convergence.

If I switch on "Monitor Coefficient Loop Convergence" and watch a key parameter such as pressure rise over the system, I would expect that an indication that I have adequate timestep size and sufficient coefficient loops would be if at the end of each timestep, the pressure rise is not changing from one coefficient loop to the next.

Am I correct in assuming that the above is an important indicator to watch for? I think I have an adequately small timestep (Max courant # < 10 and TS ~= 1/1600th of a revolution for a 4 bladed device), but yet when I look at the convergence of each timestep individually, the pressure rise across the pump is not steady even after 15+ coefficient loops where the RMS of the residuals has reached ~1e-8.

If anyone knows any sources for guidelines for TRANSIENT turbomachinery convergence, I would appreciate it. It is not covered in either the turbomachinery best practices guide (http://www.cfd-online.com/Wiki/Best_...omachinery_CFD) or the standard convergence criteria documentation.

Thanks in advance,

Saleh

[aerost]

I guess a follow up to that question is, if I am NOT satisfied with convergence, is there a way to know whether the culprit is because I have too large of a timestep OR because I don't have enough coefficient loops...

[ghorrocks]

I think I have an adequately small timestep

This would have to be the most common mistake made by newbies. My timestep looks pretty fine so it must be OK. You have to do a sensitivity check to find out.

The way to check all the parameters you mention (convergence, timestep, coeff loops) is to do a sensitivity check on them. Sweep over a range of values and see if the result changes.

And don't forget convergence and coeff loops are just different ways of expressing the same thing. So they reduce to a single factor as they are coupled. That means a single sensitivity analysis covers both these issues.