code_saturne User's Forum

Hello,

I have a problem with my monitoring points. I do a steady calculation and so I have monitoring points in my fluid domain to see if I reach a steady state. The problem is that values of monitoring points oscillate between iteration but I think that I reach a steady state because the average of oscillations is a constant.

Have you already see those oscillations in your monitoring points? Is the steady state really reached?

I join X velocity file of the monitoring points.

Best regards

Romain

Hello,
I did not check your file, but depending on the computational options used (time step, turbulence model, ...) and the mesh, this behavior is quite common and normal (as long as oscillations are not too large).
When postprocessing such a case, using average values ("moments") may be a smarter choice than using instantaneous values, especially if you want to compare results across different versions or computing environments (number of processors, machine, ...), which may modify the rounding errors and hence the exact fluctuations, but should still provide the same averages if there is no error.
Best regards,
  Yvan Fournier

Looking at the file I'd say you haven't achieved a converged solution because the velocity range (by the graph) is less than 5 m/s and some probes fluctuate by over 1m/s. What is the case you're looking at?
James

Hello,
James, I don't know which value of 5 m/s you are talking about. Besides, I know that for one probe, oscillations fluctuate by over 1m/s but I did a more refine mesh and the result for this monitoring point give the same average with much less oscillations.   
Romain

The 5m/s I meant was the overall range given by your plot at a glance - I see the maximum velocity you plot is 2.5 m/s and minimum is approx -1m/s so a range of about 3.5 m/s in fact. I think if you're expecting a fluctuation of around 1m/s in this range then perhaps the test-case is in fact unsteady and you could try that solver.
In any case the idea of steady state is you converge to a solution and so no need to average, if the problem is unsteady you can time-average your results.
Also if you change your results considerably with a refinement you should keep refining until this stops happening - it's probably likely at this point you'll have no more oscillations and you will have achieved steady state.
Cheers,
James