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.
convergence
Posted 20 gen 2011, 04:42 GMT-5 5 Replies
Please login with a confirmed email address before reporting spam
hi,
Its Amol Avhad. My geometry consists of 2 rectangular subdomains, one inside the other. The dimensions are
3*3*3 m^3 and 0.60*0.10*5e-5 m^3. From inner domain, diffusion of species in outer domain is taking place.
In outer domain velocity source is there. So i want to solve diffusion equation in inner domain and diffusion+convection
and momentum balance equation in outer domain.
Initially it was very difficult to even form the mesh inside the inner domain, but now i could form the mesh. But i am only able to solve diffusion equations in both the domains. As soon as i insert the convection term in outer domain, the solution doesn't converge. The same thing happens when i add momentum balance equation also. I am solving this in 2D.
I suspect this might be because of the large dimension difference between 2 domains. So i reduced the outer domain to 1*1*1 m^3, and even further divided it into 2 domains, still its not converging.
I don't know how to proceed this. Can anybody suggest on this.
Any help would be greatly appreciated!!!
Thank You.
Its Amol Avhad. My geometry consists of 2 rectangular subdomains, one inside the other. The dimensions are
3*3*3 m^3 and 0.60*0.10*5e-5 m^3. From inner domain, diffusion of species in outer domain is taking place.
In outer domain velocity source is there. So i want to solve diffusion equation in inner domain and diffusion+convection
and momentum balance equation in outer domain.
Initially it was very difficult to even form the mesh inside the inner domain, but now i could form the mesh. But i am only able to solve diffusion equations in both the domains. As soon as i insert the convection term in outer domain, the solution doesn't converge. The same thing happens when i add momentum balance equation also. I am solving this in 2D.
I suspect this might be because of the large dimension difference between 2 domains. So i reduced the outer domain to 1*1*1 m^3, and even further divided it into 2 domains, still its not converging.
I don't know how to proceed this. Can anybody suggest on this.
Any help would be greatly appreciated!!!
Thank You.
Attachments:
5 Replies Last Post 25 gen 2011, 01:50 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello,
Can anyone please help me out?????
Can anyone please help me out?????
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I had a quick look, your model is not evidnt, first w.r.t. the mesh ratio:
do you really need to study what happens in the thikcness of your thin volume ?
1) if yo you need at least 5-7 mesh elements in the thickness, hence elements of the size of 7-10 um, that means very many elements in the total volume, even if you allow the mesh elements of external volumes to grow
2) if you do not need the internl details, it's probably better to simulate the domain as a "thin volume = boundary with specific properties as a defined height, but then you need to adapt the physics behind.
Next I would start with a domain of 2 rectangles of about equal size ad then set up a simialr model to get the model running and to undersand the interconnection needed, then I would turn to your model, essentially because of the meshing issue, in your case you end up with many elements so the solving is slow and debugging the same way
--
Good luck
Ivar
I had a quick look, your model is not evidnt, first w.r.t. the mesh ratio:
do you really need to study what happens in the thikcness of your thin volume ?
1) if yo you need at least 5-7 mesh elements in the thickness, hence elements of the size of 7-10 um, that means very many elements in the total volume, even if you allow the mesh elements of external volumes to grow
2) if you do not need the internl details, it's probably better to simulate the domain as a "thin volume = boundary with specific properties as a defined height, but then you need to adapt the physics behind.
Next I would start with a domain of 2 rectangles of about equal size ad then set up a simialr model to get the model running and to undersand the interconnection needed, then I would turn to your model, essentially because of the meshing issue, in your case you end up with many elements so the solving is slow and debugging the same way
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you for your reply.
But i think that was the best possible way i could generate mesh inside the inner domain. Do you think it is only bcoz of large no of elements that the solution is not converging? Is it possible to further coarsen the mesh inside the inner domain to reduce no of elements? What should be the minimum element quality and element area ratio?
I am much more interested in my inner domain than the outer one. I want a concentration profile inside the inner domain.
Thank you.
But i think that was the best possible way i could generate mesh inside the inner domain. Do you think it is only bcoz of large no of elements that the solution is not converging? Is it possible to further coarsen the mesh inside the inner domain to reduce no of elements? What should be the minimum element quality and element area ratio?
I am much more interested in my inner domain than the outer one. I want a concentration profile inside the inner domain.
Thank you.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
is "inner" is the very thin domain ? if so you you should have >= 5-10 elements across the thickness hence a few microns sized elements, hence very many elements for a several square meter area. Still possible on a powerful WS as one have today.
if the inner domain is the medium rectangle, then probably you could replace the thin domain by a boundary with physical properties, including thickness. But this means equation writing, and getting them right
I agree the mesh is probably not the culprit, but makes life a bit tougher to identify really what is wrong. I did not see any evident error at my first glance, but small things are quickly missed as physics mixing are complex affairs, and your domain is by far not my specialty.
When I'm stuck like that, I start to solve a very simple problem, of the type 2 domains and check that I fully understand the physics coupling, then I apply that to a more complex (even if yours seems simple) case
--
Good luck
Ivar
is "inner" is the very thin domain ? if so you you should have >= 5-10 elements across the thickness hence a few microns sized elements, hence very many elements for a several square meter area. Still possible on a powerful WS as one have today.
if the inner domain is the medium rectangle, then probably you could replace the thin domain by a boundary with physical properties, including thickness. But this means equation writing, and getting them right
I agree the mesh is probably not the culprit, but makes life a bit tougher to identify really what is wrong. I did not see any evident error at my first glance, but small things are quickly missed as physics mixing are complex affairs, and your domain is by far not my specialty.
When I'm stuck like that, I start to solve a very simple problem, of the type 2 domains and check that I fully understand the physics coupling, then I apply that to a more complex (even if yours seems simple) case
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
hi,
Thank you for your reply!!!
But I am not able to form 5-10 elements across the thickness of inner domain. I will have to follow some different approach.
Anyway thanks for your help...
Thank you for your reply!!!
But I am not able to form 5-10 elements across the thickness of inner domain. I will have to follow some different approach.
Anyway thanks for your help...