CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Why is Fluent unable to deal with my model at the micron scale? (http://www.cfd-online.com/Forums/fluent/95169-why-fluent-unable-deal-my-model-micron-scale.html)

dixylo December 8, 2011 04:47

Why is Fluent unable to deal with my model at the micron scale?
 
My model is a micro turbo rotor spinning in a cylindrical wall. When I scaled the grid by setting the scale factors of xyz to be 0.00001 or lower, the iteration would stop in one step, saying "solution is converged"
At first i thought something was wrong with my grid and settings, so i created a cylindrical pipe model where liquid flows in with a speed magnitude equal to that of the radius through the bottom and out through the top, but this model has the same problem that when the scale factors of xyz are lower than a certain value (0.0009 here), the iteration stops within one or a couple of steps.
however, what has puzzled me is that when the scale factors are greater than a certain value (1e-5 for my rotor model and 0.0009 for the pipe one), the problem doesn't show up and Fluent gives a good solution.
so can Fluent deal with models as small as tens of microns?

Amir December 8, 2011 06:04

Quote:

Originally Posted by dixylo (Post 335094)
My model is a micro turbo rotor spinning in a cylindrical wall. When I scaled the grid by setting the scale factors of xyz to be 0.00001 or lower, the iteration would stop in one step, saying "solution is converged"
At first i thought something was wrong with my grid and settings, so i created a cylindrical pipe model where liquid flows in with a speed magnitude equal to that of the radius through the bottom and out through the top, but this model has the same problem that when the scale factors of xyz are lower than a certain value (0.0009 here), the iteration stops within one or a couple of steps.
however, what has puzzled me is that when the scale factors are greater than a certain value (1e-5 for my rotor model and 0.0009 for the pipe one), the problem doesn't show up and Fluent gives a good solution.
so can Fluent deal with models as small as tens of microns?

Hi,

Did you disable convergence criteria in "solve->monitor->residual" ?
It better to use double precision version for your purpose and also, you can change length dimension to "mm" for example but as you said, this problem may exist even in double precision solvers if the variables would be in the order of round off error which is about 1.0e-16. As there are many simulations with FLUENT for micro channels in literature, I don't think that this can make any issue for your case. But if you couldn't solve this problem, maybe a dimensional analysis can help.

Bests,

dixylo December 8, 2011 14:39

Quote:

Originally Posted by Amir (Post 335106)
Hi,

Did you disable convergence criteria in "solve->monitor->residual" ?
It better to use double precision version for your purpose and also, you can change length dimension to "mm" for example but as you said, this problem may exist even in double precision solvers if the variables would be in the order of round off error which is about 1.0e-16. As there are many simulations with FLUENT for micro channels in literature, I don't think that this can make any issue for your case. But if you couldn't solve this problem, maybe a dimensional analysis can help.

Bests,

I didn't disable convergence criterion (I kept it as "absolute"). Most settings were kept as default. I tried double precision solvers as soon as i met this problem, but they didn't work either. As you said, there are many simulations with FLUENT for micro channels in literature, this is what has puzzled me. If you are interested in this problem and will spend a little time, you may create a similar model, for instance, a cylinder with a height of 5 and a radius of 1, and the bottom as VELOCITY-INLET boundary type, the top as OUTFLOW. Scale it with scale factors 1e-6. Other settings can be done as they should be. And i think you would probably see "solution is converged" in one iteration step. anyway, thank you for your reply and your suggestion.

Amir December 8, 2011 15:31

Quote:

Originally Posted by dixylo (Post 335174)
I didn't disable convergence criterion (I kept it as "absolute"). Most settings were kept as default. I tried double precision solvers as soon as i met this problem, but they didn't work either. As you said, there are many simulations with FLUENT for micro channels in literature, this is what has puzzled me. If you are interested in this problem and will spend a little time, you may create a similar model, for instance, a cylinder with a height of 5 and a radius of 1, and the bottom as VELOCITY-INLET boundary type, the top as OUTFLOW. Scale it with scale factors 1e-6. Other settings can be done as they should be. And i think you would probably see "solution is converged" in one iteration step. anyway, thank you for your reply and your suggestion.

Hi,

I meant toggle off "check convergence" in "solve->monitor->residual"; in this manner you never face "solution is converged"!!! just let it continue to iterate.

Bests,

dixylo December 9, 2011 00:39

Quote:

Originally Posted by Amir (Post 335178)
Hi,

I meant toggle off "check convergence" in "solve->monitor->residual"; in this manner you never face "solution is converged"!!! just let it continue to iterate.

Bests,

Sorry, I'm a newb so i didn't quite understand what you'd meant as well as some of those variables and settings. Your suggestion did work, and i think what i should do now is optimize the grid. Thank you very much for you kind help.:)


All times are GMT -4. The time now is 20:31.