Grid size, convergence issues

franzdrs · June 17, 2009, 06:19

Hi,
I have ran a simulation for a simple geometry, not much more than a parallelepiped with dimensions 3.6x1.2x1.1 m. The air flows in through a channel in one of the 1.2x1.1m faces and leaves through one of the 3.6x1.2m faces. Theres is a porous media between inlet and outlet. My grid has about 120000 cells. First I ran a simulation using the Laminar model and it wouldnt converge at all. Then I changed to the k-epsilon realizable model and it converges in less than 170 iterations. However, from the beginning I got the warning "turbulent viscosity limited to viscosity ratio of 1e+5" in a ever decreasing number of cells (starting at about 2000 cells) until the warning dissapears after about 80 iterations.
So I have some questions about this. Maybe somebody can help me:
1. What does it mean that the simulation doesnt converge with the laminar model and it does with the turbulent, and so fast? Shouldnt be always easier a convergence in Laminar model? Does this has to do with the grid? Is it too coarse?
2. If this warning "turbulent viscosity limited to viscosity ratio of 1e+5" dissapears before convergence, can I still trust my solution or is it flawed already?
3. I expected a different solution. I am interested in the velocity distribution at the outlet. With this simple configuration I expected a very non-uniform distribution, but it turns to be pretty good (and it does not give much chance for improvement

). So, given the above explained issues, can I trust this solution? Should the grid be much finer?
Can anyone help?

Thanks very much,
Franz

vishyaroon · June 17, 2009, 09:17

Hello Franz,

How have you defined the porous media? do you have both the inertial and viscous co-efficients? If your flow is laminar then you'll the pressure drop co-relation is linear and so the inertial co-efficient should disappear. However if you still have defined the inertial co-efficient this might cause convergence problems (based on a couple of my experiences).

The porous media imposes a pressure drop across the region and hence the solution in the porous region is not very sensitive to grid refinement. So refining the grid may not give you a better solution than what you have now.

Again to add to this, you may want to remember that the porous media if dense enough causes enough back-pressure that the flow through it is more uniform than what it will be in the absence of porous media.

franzdrs · June 17, 2009, 13:57

Hi Aroon,
thanks for your answer. Yes, I have defined both coefficients for the porous media. I calculated them from experimental data on pressure drop versus air speed through the medium. The plot points of deltaP vs air speed adjusted much better to a quadratic function than to a linear one, so I judged both coefficients were needed.
I know for sure that the flow is turbulent at the entrance to the system. I have an air speed of 3.6 m/s in the inlet, which is a rectangular aperture with dimensions 0.4x0.3 m. If the flow turns laminar when it flows through the porous media, I really dont know. the superficial velocity through it is about 0.1m/s.
About your last point, yes, the more the flow resistance the porous media imposes, the more uniform the flow through it will be. I thought, though, that the resistance of my porous media was quite low. At 0.1m/s (superficial velocity) the deltaP is about 6 Pa.

What about the warning "turbulent viscosity limited to viscosity ratio of 1e+5"? if it goes away should there be a problem?

Thanks a lot

Franz

vishyaroon · June 18, 2009, 08:57

Hello Franz,

Have you checked the difference in the velocity profiles at the outlet with and without the porous media? Do you see a difference? If you see that the flow is more uniform with the porous media then I'd say that atleast the trend is correct.

For the limiter warning, do you know how many cells are being affected? Can you try plotting the turbulent viscosity and see how large this region is? If this is small then you can trust your results.

June 17, 2009, 06:19	Grid size, convergence issues	#1
franzdrs New Member Franz Roman Join Date: Jun 2009 Posts: 28 Rep Power: 17	Hi, I have ran a simulation for a simple geometry, not much more than a parallelepiped with dimensions 3.6x1.2x1.1 m. The air flows in through a channel in one of the 1.2x1.1m faces and leaves through one of the 3.6x1.2m faces. Theres is a porous media between inlet and outlet. My grid has about 120000 cells. First I ran a simulation using the Laminar model and it wouldnt converge at all. Then I changed to the k-epsilon realizable model and it converges in less than 170 iterations. However, from the beginning I got the warning "turbulent viscosity limited to viscosity ratio of 1e+5" in a ever decreasing number of cells (starting at about 2000 cells) until the warning dissapears after about 80 iterations. So I have some questions about this. Maybe somebody can help me: 1. What does it mean that the simulation doesnt converge with the laminar model and it does with the turbulent, and so fast? Shouldnt be always easier a convergence in Laminar model? Does this has to do with the grid? Is it too coarse? 2. If this warning "turbulent viscosity limited to viscosity ratio of 1e+5" dissapears before convergence, can I still trust my solution or is it flawed already? 3. I expected a different solution. I am interested in the velocity distribution at the outlet. With this simple configuration I expected a very non-uniform distribution, but it turns to be pretty good (and it does not give much chance for improvement ). So, given the above explained issues, can I trust this solution? Should the grid be much finer? Can anyone help? Thanks very much, Franz

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
mesh size and convergence	ertan	Main CFD Forum	2	April 3, 2009 17:15
Phase locked average in run time	panara	OpenFOAM	2	February 20, 2008 15:37
Limitation in grid size	Omkar	Main CFD Forum	6	November 18, 2007 01:16
grid adaptation for better convergence.	co2	FLUENT	9	April 2, 2004 15:08
Grid Independent Solution	Chuck Leakeas	Main CFD Forum	2	May 26, 2000 12:18

June 17, 2009, 09:17		#2
vishyaroon Senior Member Aroon Join Date: Apr 2009 Location: Racine WI Posts: 148 Rep Power: 17	Hello Franz, How have you defined the porous media? do you have both the inertial and viscous co-efficients? If your flow is laminar then you'll the pressure drop co-relation is linear and so the inertial co-efficient should disappear. However if you still have defined the inertial co-efficient this might cause convergence problems (based on a couple of my experiences). The porous media imposes a pressure drop across the region and hence the solution in the porous region is not very sensitive to grid refinement. So refining the grid may not give you a better solution than what you have now. Again to add to this, you may want to remember that the porous media if dense enough causes enough back-pressure that the flow through it is more uniform than what it will be in the absence of porous media.

June 17, 2009, 13:57		#3
franzdrs New Member Franz Roman Join Date: Jun 2009 Posts: 28 Rep Power: 17	Hi Aroon, thanks for your answer. Yes, I have defined both coefficients for the porous media. I calculated them from experimental data on pressure drop versus air speed through the medium. The plot points of deltaP vs air speed adjusted much better to a quadratic function than to a linear one, so I judged both coefficients were needed. I know for sure that the flow is turbulent at the entrance to the system. I have an air speed of 3.6 m/s in the inlet, which is a rectangular aperture with dimensions 0.4x0.3 m. If the flow turns laminar when it flows through the porous media, I really dont know. the superficial velocity through it is about 0.1m/s. About your last point, yes, the more the flow resistance the porous media imposes, the more uniform the flow through it will be. I thought, though, that the resistance of my porous media was quite low. At 0.1m/s (superficial velocity) the deltaP is about 6 Pa. What about the warning "turbulent viscosity limited to viscosity ratio of 1e+5"? if it goes away should there be a problem? Thanks a lot Franz

June 18, 2009, 08:57		#4
vishyaroon Senior Member Aroon Join Date: Apr 2009 Location: Racine WI Posts: 148 Rep Power: 17	Hello Franz, Have you checked the difference in the velocity profiles at the outlet with and without the porous media? Do you see a difference? If you see that the flow is more uniform with the porous media then I'd say that atleast the trend is correct. For the limiter warning, do you know how many cells are being affected? Can you try plotting the turbulent viscosity and see how large this region is? If this is small then you can trust your results.