Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
1 giu 2010, 05:47 GMT-4
Probably your Pr/Sc number is pretty high and hence very high gradients. Are you using any artificial diffusion?
Probably your Pr/Sc number is pretty high and hence very high gradients. Are you using any artificial diffusion?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
1 giu 2010, 06:25 GMT-4
Thanks for the reply.
I have not changed any setting in the stabilization part of the subdomain settings, so no artificial diffusion. I did try to read something about those from the manuals, but didn't get the idea what to do.'
And yes, there are high concentration gradients in the model so I made very fine mesh in those areas.
Thanks for the reply.
I have not changed any setting in the stabilization part of the subdomain settings, so no artificial diffusion. I did try to read something about those from the manuals, but didn't get the idea what to do.'
And yes, there are high concentration gradients in the model so I made very fine mesh in those areas.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
1 giu 2010, 07:56 GMT-4
Using FEM I can hardly imagine that you will get to a reasonable mesh element size corresponding to Peclet number of 2. Therefore, you will have instabilities in the domain and specially around the boundaries. Hence, you more or less must use an artificial diffusion method, preferably a non-destructive one, like streamline diffusion. The other method is to reduce the time-step to a very small value (typically 1-10e-6).
These are based on my own experience, so it may not be valid completely in your case.
Using FEM I can hardly imagine that you will get to a reasonable mesh element size corresponding to Peclet number of 2. Therefore, you will have instabilities in the domain and specially around the boundaries. Hence, you more or less must use an artificial diffusion method, preferably a non-destructive one, like streamline diffusion. The other method is to reduce the time-step to a very small value (typically 1-10e-6).
These are based on my own experience, so it may not be valid completely in your case.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
2 giu 2010, 05:25 GMT-4
First of all, my model is steady-state, so no time-steps. I have been wondering is there any sense to use the turbulence model in a steady-state problem... but I don't know.
Now I tried few different settings for artificial diffusion, but they seem to make the solution oscillate. I read the part of the stabilization methods in the manual, but I'm still very confused about what settings I really should use. It would help if I found an example about turbulent modelling with molecular diffusion, but I didn't find any.
First of all, my model is steady-state, so no time-steps. I have been wondering is there any sense to use the turbulence model in a steady-state problem... but I don't know.
Now I tried few different settings for artificial diffusion, but they seem to make the solution oscillate. I read the part of the stabilization methods in the manual, but I'm still very confused about what settings I really should use. It would help if I found an example about turbulent modelling with molecular diffusion, but I didn't find any.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
2 giu 2010, 10:36 GMT-4
maybe it has something to do with this:
www.comsol.com/support/knowledgebase/1081/
As long as you do not tell us tech details of your model all we can do is guessing. Above is for Comsol 4.
maybe it has something to do with this: http://www.comsol.com/support/knowledgebase/1081/
As long as you do not tell us tech details of your model all we can do is guessing. Above is for Comsol 4.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
2 giu 2010, 12:19 GMT-4
Now I think I got it :) Streamline diffusion made the reaction rate much more stable. I got a really nice graph out of it now.
I didn't get it at first because I was using logarithmic concentrations instead of real concentrations to avoid negative concentrations. Now with the older model I see that streamline diffusion helped a lot. Maybe I will now just try to ignore the negative (-1e-7) concentrations from the numerical noise or add some more reactants...
Thank you very much for the help.
Now I think I got it :) Streamline diffusion made the reaction rate much more stable. I got a really nice graph out of it now.
I didn't get it at first because I was using logarithmic concentrations instead of real concentrations to avoid negative concentrations. Now with the older model I see that streamline diffusion helped a lot. Maybe I will now just try to ignore the negative (-1e-7) concentrations from the numerical noise or add some more reactants...
Thank you very much for the help.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
2 giu 2010, 16:49 GMT-4
You're welcome.
Also, take care of Artificial diffusion (AD). Try different AD parameter values to see how much it is changing your solution. It can be a real beast sometimes :-)
You're welcome.
Also, take care of Artificial diffusion (AD). Try different AD parameter values to see how much it is changing your solution. It can be a real beast sometimes :-)