Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

1D and 2D Model

Please login with a confirmed email address before reporting spam

Hi There,

I have two model one is 1D and the other one is 2D solves same set of equations. I got good results for ID model. For 2D, I suppose to get almost same results. But I am getting results much differnt from ID.

Can any one say what went wrong. Is the problem with BCs oR dimestions or anything elase?

Kind Regards
Ramadas


4 Replies Last Post Sep 1, 2010, 8:42 a.m. EDT
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 1, 2010, 2:14 a.m. EDT
Hi

looks nice, but the equations are somewhat too complex for me to catch immediatly, what you are simulating ?

One trick: turn off the Plot Parameters General tab "Geometry edges" to better visualise your curves, COSMOL has sometimes problems if the value are too different from the vertexes, to plot the curves smoothly.

What slightly astonishes me is the spiky derivatives you have Ygx Ygt ..., this is only a qualitative reply, but I do not like to see oscillations when I expect steps or smooth lines, it remains with a finer mesh.

We had a discussion a few months ago on the forum, where I believe I saw this too, and there there was a missing BC, it's up there somewhere, try a search.

perhaps some more explanations to let us others understand what you are doing here, could help, if not now, publish it on one of the COMSOL conferences, then we other users can learn something too ;)

Have fun Comsoling
Ivar
Hi looks nice, but the equations are somewhat too complex for me to catch immediatly, what you are simulating ? One trick: turn off the Plot Parameters General tab "Geometry edges" to better visualise your curves, COSMOL has sometimes problems if the value are too different from the vertexes, to plot the curves smoothly. What slightly astonishes me is the spiky derivatives you have Ygx Ygt ..., this is only a qualitative reply, but I do not like to see oscillations when I expect steps or smooth lines, it remains with a finer mesh. We had a discussion a few months ago on the forum, where I believe I saw this too, and there there was a missing BC, it's up there somewhere, try a search. perhaps some more explanations to let us others understand what you are doing here, could help, if not now, publish it on one of the COMSOL conferences, then we other users can learn something too ;) Have fun Comsoling Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 1, 2010, 9:00 p.m. EDT
Thanks Ivar , for sharing your thoughts.


I am modelling a hygroscopic wheel channel coated with a desiccant. This is used for dehumidification of air. To dehumidify the air, just pass it through these channels of the wheel. After some time (90 sec), the desiccant in the channel material will be fully loaded with moisture. To remove this moisture and make it ready for next dehumidification process, we pass hot air for some time (90sec). To ensure continuous operation, half of the wheel will be in air which is to be dried and next half will be in the hot air which regenerates and the wheel is rotated between two stream.


I must get reduction in humidity (Yg) for first 90 sec and then increase in humidity for next 90 sec (regeneration). I am getting same trend in the both 1D and 2D models. Temperature profile is also as expected. But ID seems to be more accurate. There must some problem with 2D model.

Here is the 2d model. The model is to meshed with mapped mesh parameters, 'extremely fine mesh’. You can see a global variable AYg and AYg1 which shows the inlet and out let humidity ratio.

What is of primary inertest is the variation of humidity and temperature plots( Yg, Tg) of air.. Yes we had some discussion on step BC and I overcame the problem by using ‘flc2hs’ function. It works very well. Yes, I am interested in publishing it in a COMSOL Conference.


Kind Regards

Your reply will be highly appreciated.
Thanks Ivar , for sharing your thoughts. I am modelling a hygroscopic wheel channel coated with a desiccant. This is used for dehumidification of air. To dehumidify the air, just pass it through these channels of the wheel. After some time (90 sec), the desiccant in the channel material will be fully loaded with moisture. To remove this moisture and make it ready for next dehumidification process, we pass hot air for some time (90sec). To ensure continuous operation, half of the wheel will be in air which is to be dried and next half will be in the hot air which regenerates and the wheel is rotated between two stream. I must get reduction in humidity (Yg) for first 90 sec and then increase in humidity for next 90 sec (regeneration). I am getting same trend in the both 1D and 2D models. Temperature profile is also as expected. But ID seems to be more accurate. There must some problem with 2D model. Here is the 2d model. The model is to meshed with mapped mesh parameters, 'extremely fine mesh’. You can see a global variable AYg and AYg1 which shows the inlet and out let humidity ratio. What is of primary inertest is the variation of humidity and temperature plots( Yg, Tg) of air.. Yes we had some discussion on step BC and I overcame the problem by using ‘flc2hs’ function. It works very well. Yes, I am interested in publishing it in a COMSOL Conference. Kind Regards Your reply will be highly appreciated.


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 2, 2010, 11:14 a.m. EDT
Hi

I cannot say I have undestood the physics yet, but still a couple of comments:

1) your meshing in free mode is thin (also because of your high aspect ratio, I remeshed it with quads, ensuring at least 10 elements in the thickness (per layer) but only 40-80 in the length (in "equal" mode it looks awful but seems to solve OK) Then definitively I get better details, still is it right or not ? not that easy for me to tell

You have some very strong gradients to the right vertical boundary on subdomain 2, perhaps a log or exp progression on your mesh would help for x around 20. to have a denser mesh therearound.

On subdomain 1 you have large gradients along the upper horizontal line (even with some oscilaltions that I wounder if could come perhaps from the internal numerical stabilisation methods, hence are not really "physical")
Here again one should have a non linear meshing to better follow the gradients

To observe the full domain, define an "Expression subdomain expression" Y as Yg respectively Yd and T as Tg Td in the different domains then you can plot everything in one go with Y and T

It would be easier to understsand if you make functions out of your time steps so you could plot them and easier check their shape. If you select a point probe in the middle of SD-1 you can see temperature oscillations between 300 and 335 K (I assume) but your steps are very steep, tricky for the solver, but it pases.

I have tried to control the meshing in this way: first my manually meshing free the edges with the different distributions, then selecting both subdomains and applying square map mesh. It solves in a few minutes with less than a Gb RAM.

But again, I cannot really tell if its phyically correct, check your BC carefully ;) Just ignore the warnings of constraint forces (in my opinion)

I have the impression that the internal boundary coupling on line 4 is not working OK, it does not seem to be continuity, w.r.t the boundary normal

In fact with exponential distributed meshing the gradients become just steeper, and there is less oscilaltions, again for me the oscillations are indicating probably effects from the numerical stabilisation techniques inside COMSOL

Have fun Comsoling
Ivar
Hi I cannot say I have undestood the physics yet, but still a couple of comments: 1) your meshing in free mode is thin (also because of your high aspect ratio, I remeshed it with quads, ensuring at least 10 elements in the thickness (per layer) but only 40-80 in the length (in "equal" mode it looks awful but seems to solve OK) Then definitively I get better details, still is it right or not ? not that easy for me to tell You have some very strong gradients to the right vertical boundary on subdomain 2, perhaps a log or exp progression on your mesh would help for x around 20. to have a denser mesh therearound. On subdomain 1 you have large gradients along the upper horizontal line (even with some oscilaltions that I wounder if could come perhaps from the internal numerical stabilisation methods, hence are not really "physical") Here again one should have a non linear meshing to better follow the gradients To observe the full domain, define an "Expression subdomain expression" Y as Yg respectively Yd and T as Tg Td in the different domains then you can plot everything in one go with Y and T It would be easier to understsand if you make functions out of your time steps so you could plot them and easier check their shape. If you select a point probe in the middle of SD-1 you can see temperature oscillations between 300 and 335 K (I assume) but your steps are very steep, tricky for the solver, but it pases. I have tried to control the meshing in this way: first my manually meshing free the edges with the different distributions, then selecting both subdomains and applying square map mesh. It solves in a few minutes with less than a Gb RAM. But again, I cannot really tell if its phyically correct, check your BC carefully ;) Just ignore the warnings of constraint forces (in my opinion) I have the impression that the internal boundary coupling on line 4 is not working OK, it does not seem to be continuity, w.r.t the boundary normal In fact with exponential distributed meshing the gradients become just steeper, and there is less oscilaltions, again for me the oscillations are indicating probably effects from the numerical stabilisation techniques inside COMSOL Have fun Comsoling Ivar


kyprianos georgkopoulo

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 1, 2010, 8:42 a.m. EDT
Hi there,

I am modeling a sorption wheel in 2d,but i have a problem using the PDE's modes.
I must insert 4 PDE's in comsol via general mode of PDE's(time depending) but i cannot
use the 4 dependent variables that i have in my model as dependent variables in comsol.
In other words the subdomain in PDE's of comsol uses the variable 'u' as the dependent
variable,but in my model this is the indepedent variable.The dependent is in the place of
'?' as shown in the subdomain settings.

Thank you for your time.
Hi there, I am modeling a sorption wheel in 2d,but i have a problem using the PDE's modes. I must insert 4 PDE's in comsol via general mode of PDE's(time depending) but i cannot use the 4 dependent variables that i have in my model as dependent variables in comsol. In other words the subdomain in PDE's of comsol uses the variable 'u' as the dependent variable,but in my model this is the indepedent variable.The dependent is in the place of '?' as shown in the subdomain settings. Thank you for your time.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.