CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Cyclone Simulation Convergence Problem (http://www.cfd-online.com/Forums/fluent/45520-cyclone-simulation-convergence-problem.html)

Sal July 26, 2007 07:31

Cyclone Simulation Convergence Problem
 
Hello every one.

I am simulating a cyclone seperator and I have easily got convergence using K-E, K-E RNG both first and second order discritization. Using the RSM model i can only get convergence for first order but cant seem to get convergence for second order. I am using 517000 tetrahedral cells in the mesh. Can some one help me on this please Thank you in Advance.

Sal

Joe July 26, 2007 08:07

Re: Cyclone Simulation Convergence Problem
 
Restart the RSM solution from a converged two equation solution. Adjust your under relaxation factors.

Sal July 26, 2007 08:15

Re: Cyclone Simulation Convergence Problem
 
Should i start it from the first order discretization. Steady, unsteady, boundary conditions bottom outlet as wall or pressure outlet and if unsteady what would u recomend the time step to be and should I use transient controls and which one? I appreciate this.

Sal

Bubble July 26, 2007 12:00

Re: Cyclone Simulation Convergence Problem
 
I would recommend that u start using steady state first to build up pressure to create a swirling flow. Once the residual gets stable change to unsteady state because the vortex will move in quasi periodic movement. You can adjust the time step to get better convergence or u can adjust your relaxation factor.


Sal July 26, 2007 18:18

Re: Cyclone Simulation Convergence Problem
 
For the bottom wall what do you recomend me to use as a boundary condition. At the moment it is defined as a wall. Do i need to change? if yes what to and would I have to enter any other parameters. Thank you for the help in advance.

Sal

Sal July 27, 2007 08:16

Re: Cyclone Simulation Convergence Problem
 
While running the boundary conditions i get no reverse flow. Does any one know how to correct this. Thank You. Sal

JSM July 28, 2007 01:53

Re: Cyclone Simulation Convergence Problem
 
Hi,

For cyclone, there are some basic steps in CFD.

1. For meshing - Always use structured mesh (hexahedral elements) with aspect ratio within the range 5 to 10. 2. solver setup - RSM turbulence model is recommended. First solve with default setup with RSM. Once converged, then go to higer order discretization with increased converge monitor 1e-6. 3. Choose time setup size as analysis should converge in every time step.

Note: first order converged solution always gives some appropriate values. dont worry about. Cyclone phenomena is purely unsteady. So unsteady with higher order discretization, you can more accurate solution.

fatih-ITS December 17, 2014 04:32

Another viscous model besides rsm
 
Iam still running Cyclone with double inlet with section angle 30 degree, 2 phase coal and air
Iam going to asking too, can I use another viscous model like k-epsilon realizable to get correct separation efficiency and pressure drop ??
Can you give me advice setting for double inlet cyclone to me ?
because Iam very depressed with my final assignment,:o:o
Thank you for your attention :)




Quote:

Originally Posted by JSM
;143944
Hi,

For cyclone, there are some basic steps in CFD.

1. For meshing - Always use structured mesh (hexahedral elements) with aspect ratio within the range 5 to 10. 2. solver setup - RSM turbulence model is recommended. First solve with default setup with RSM. Once converged, then go to higer order discretization with increased converge monitor 1e-6. 3. Choose time setup size as analysis should converge in every time step.

Note: first order converged solution always gives some appropriate values. dont worry about. Cyclone phenomena is purely unsteady. So unsteady with higher order discretization, you can more accurate solution.


ghost82 December 17, 2014 08:46

As suggested above rsm turbulence model is the most appropriate.
If you want to switch to k-epsilon I suggest the rng model with the swirl option switched on.

PS: don't be depressed :D


All times are GMT -4. The time now is 19:08.