|September 3, 2010, 11:04||
Problems with residuals oscillations
Join Date: Aug 2009
Posts: 6Rep Power: 7
i'm using star-cmm+ 5.02.010 and i'm carrying on a simulation (steady model) concerning a cyclone separator, with two phases (air + dust) and segregated model.
boundary conditions: velocity inlet for the inlet tube, pressure outlet for the air outlet tube and wall (escape condition) for the dust outlet tube.
after 2000-3000 iterations ("lagrangian multiphase solver = frozen") the residulas start to oscillate and i cannot solve this problem. i tried to change the following parameters without positive effects:
- "velocity under-relaxation factor" from 0.7 to 0.45
- "pressure under-relaxation factor" from 0.4 to 0.1
- "K-Omega Turbulence under-relaxation factor" from 0.8 to 0.4
- "K-Omega Turbulent viscosity under-relaxation factor" from 1 to 0.5
I left these parameters this way and after this i also changed the following parameter (in the Physics model --> Segregated flow): the convection scheme from 2nd-order to 1-st order.
Then i've done 5000-6000 iterations and the residuals oscillations are very small and the mean values of residuals are much lower then before (i reached the convergence).
But i don't know if this change is correct...
After this convergence do i have to change again the order from 1st to 2nd? In other words, do i have to do 5000-6000 iterations with the 1st order and later other X iterations changing the order to 2nd?
And the parameters i changed (from the list i wrote before) are correct or is better left them as default?
thanks very much
|September 12, 2010, 16:17||
Join Date: Oct 2009
Posts: 637Rep Power: 12
In general, you will get a different solution with 1st order and 2nd order discretisation and the 2nd order is more accurat. So it is recommended to switch back to 2nd order.
The under relaxation factors can stay lower than default values. The solution will be the same, it will only take longer to get to that solution.
Maybe the oscillation occurs because your problem is transient, but you try to get a steady state solution.
|September 13, 2010, 08:57||
Join Date: Mar 2009
Posts: 20Rep Power: 7
i made the experience, that you have to do an unsteady calculation if you want to do a correct cyclone calculation. (You also need a good hex mesh)
So maybe the residuals are oscillating because of an unsteady solution.
|Thread||Thread Starter||Forum||Replies||Last Post|
|Oscillations in Results and Residuals||Freeman||FLUENT||27||December 18, 2010 14:09|
|problems with residuals in fluent.||Tan||FLUENT||3||February 10, 2009 11:05|
|Problems with residuals||Martin||CFX||2||January 2, 2007 23:10|
|Convergence - scaled vs unscaled residuals||HS||FLUENT||1||November 7, 2005 06:45|
|Problems with residuals??||Peter Kolb||FLUENT||0||November 29, 2000 04:48|