Time taken to converge
 

Does anyone know if there is a way of finding out how long a simulation has taken to converge? I know Fluent gives an indication of how it thinks it should take whilst iterating but is there some way of finding how long it actually took?

Cheers,

Duncan

Re: Time taken to converge
 

hi,

do you really mean the time or the the number of iterations? Iám not very familiar with fluent, but I think it is not possible to say how much iterations you need for convergence. For many cases it is not sure, if it will converge anyway! If you got a simple geometrie and use a small number of equations (e.g. laminar flow), it probably takes less than 100 iterations. If your geometrie/flow pattern is complex and you use more equations (incompressible flow, more accurate turbulence model), it probably takes 5000 iterations or more.

How much time it takes per iteration is mainly depending on the number of grid elements, number of equations, solver (segrated, coupled...), convergence criterion (rms 10^-4, 10^-5,...) and your computer.

greetings

Re: Time taken to converge
 

I have a converged simulation and just wanted to know how long it took to reach this converged state. I would like to do a quantitative analysis of how long different mesh sizes take and wanted to know if there is some way of finding out the duration of simulation (eg hours, mins,seconds).

Many Thanks

D

Re: Time taken to converge
 

ahh, sorry, my mistake. you wrote "has taken". No, sorry, I can´t help you.

Re: Time taken to converge
 

No problem. Thanks for replying

Re: Time taken to converge
 

I've never found a way of doing this one a single processor job, but in parallel, you can go to Parallel->Timer->Usage.

Hope this helps, and good luck, Jason

Re: Time taken to converge
 

No problem. Thanks for replying

Re: Time taken to converge
 

For Linux: Start your Fluent using the /usr/bin/time command (or similar path) - then you get cpu+wallclock time for your process (solving + overhead for loading/saving) when you exit Fluent.

Re: Time taken to converge
 

If you use a journal file to run your jobs, insert a few save points. That way you can compare the time at which each data file was written, which will give you a pretty good idea about the run time.

Re: Time taken to converge
 

You can do it in batch mode You run your cas as a parallel case but for only 1 processor And then you can have your timer usage as jason said I also work on only 1 processor and it does work good luck