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.
Exciting different boundaries gives unexplainable several orders of magnitude difference
Posted 12.11.2009, 15:38 GMT-5 2 Replies
Please login with a confirmed email address before reporting spam
Hi
I have two files Comsol_innerboundary_problem.mph and Comsol_outerboundary_problem.mph. They both have identical equations and constants. They differ in which boundary the boundary condition is applied. In both the problems there are two spheres. One little one of R=1 inside of a bigger sphere of R=2.
In the Comsol_innerboundary_problem.mph the inner sphere is excited with a magnetic potential and the resulting velocity profile is in the 'noise' of order 10^-13
Using the same boundary condition but this time applying it to the outer sphere and maintaining a continuity boundary condition on the inner sphere gives a velocity profile of order 10^-3.
I am aware that the radius is larger which lowers the field internal to the sphere, and Im also aware that demagnetizing effects occur. However I dont think that this is enough of a reason to give such a big discrepancy in the order of the flow.
I have narrowed it down to the M_x variable (which drives a FMx force term in the navier stokes module). In the Inner boundary excitation problem the M_x is perfectly constant as a function of space. In the outer boundary excitation problem the M_x variable is almost constant with a variation of 10^-4 which is close to the order of the velocity flow. It doesnt make sense why Comsol is doing this. I have tried increased the mesh and even added more boundary layers but to no avail.
I have attached plots of the profile in the file Comsol_boundary_excitation_problem.docx along with the two Comsol model files.
Any help on this matter will be greatly appreciated.
I have two files Comsol_innerboundary_problem.mph and Comsol_outerboundary_problem.mph. They both have identical equations and constants. They differ in which boundary the boundary condition is applied. In both the problems there are two spheres. One little one of R=1 inside of a bigger sphere of R=2.
In the Comsol_innerboundary_problem.mph the inner sphere is excited with a magnetic potential and the resulting velocity profile is in the 'noise' of order 10^-13
Using the same boundary condition but this time applying it to the outer sphere and maintaining a continuity boundary condition on the inner sphere gives a velocity profile of order 10^-3.
I am aware that the radius is larger which lowers the field internal to the sphere, and Im also aware that demagnetizing effects occur. However I dont think that this is enough of a reason to give such a big discrepancy in the order of the flow.
I have narrowed it down to the M_x variable (which drives a FMx force term in the navier stokes module). In the Inner boundary excitation problem the M_x is perfectly constant as a function of space. In the outer boundary excitation problem the M_x variable is almost constant with a variation of 10^-4 which is close to the order of the velocity flow. It doesnt make sense why Comsol is doing this. I have tried increased the mesh and even added more boundary layers but to no avail.
I have attached plots of the profile in the file Comsol_boundary_excitation_problem.docx along with the two Comsol model files.
Any help on this matter will be greatly appreciated.
Attachments:
2 Replies Last Post 14.11.2009, 16:49 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi
you have a nice, but complex and challenging model, I really like those ;)
Especially complex for us to get into without studying extensively the litterature and your model. With a mix of so many physics you really have to go through it step by step to be sure one understands the results at each level, I'm sure you have, but for us, out here, it's very time consuming !
I would, in youre place, attempt to isolate the boundary discrepancy you show with less physics involved. These are typically the types of issues I submit to COMSOL support, but I believe even for them to follow, either you provide them with much more info, and references, or you manage first to reduce the model complexity drastically, such to get a quick reply.
I agree with you that it looks like a serious gradient step over the boundary that is generated only for one of the two cases. Perhaps, by using week constraints "on" and redefinig some of the physics along the borders it could change, but I reconised I would need a week or two to do that, and really I do not have the free time.
One "point" in the Navier-Stokes section, you define the pressure in one point P9, that I understand, but why "0" for the pressure ? I would use a value >0, but again, that depends on your material parameters, if they do not depend on "p" it's probably OK. In case you choose some of the "Material Library" values, these are often depending on "p" and with "0" pressure you might also have "0" density, "0" conductivity, viscuosity etc
Pls keep us informed and tell us where you expect to publish this, and why not in a future COMSOL Conference procceding too
Good luck
Ivar
you have a nice, but complex and challenging model, I really like those ;)
Especially complex for us to get into without studying extensively the litterature and your model. With a mix of so many physics you really have to go through it step by step to be sure one understands the results at each level, I'm sure you have, but for us, out here, it's very time consuming !
I would, in youre place, attempt to isolate the boundary discrepancy you show with less physics involved. These are typically the types of issues I submit to COMSOL support, but I believe even for them to follow, either you provide them with much more info, and references, or you manage first to reduce the model complexity drastically, such to get a quick reply.
I agree with you that it looks like a serious gradient step over the boundary that is generated only for one of the two cases. Perhaps, by using week constraints "on" and redefinig some of the physics along the borders it could change, but I reconised I would need a week or two to do that, and really I do not have the free time.
One "point" in the Navier-Stokes section, you define the pressure in one point P9, that I understand, but why "0" for the pressure ? I would use a value >0, but again, that depends on your material parameters, if they do not depend on "p" it's probably OK. In case you choose some of the "Material Library" values, these are often depending on "p" and with "0" pressure you might also have "0" density, "0" conductivity, viscuosity etc
Pls keep us informed and tell us where you expect to publish this, and why not in a future COMSOL Conference procceding too
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi Ivar
Yes I completely agree with you that there is a lot going on in my model and it makes it difficult to debug.
I have come up with a very simple model to illustrate the problem. It only uses the magnetostatics module in 3D
I have two files
Comsol_boundarylayer_problem_outer.mph
Comsol_boundarylayer_problem_inner.mph
Both files have the same geometry of an inner sphere of R=1 in a bigger sphere of R=2. In Comsol_boundarylayer_problem_outer.mph I apply a magnetic potential of -x on the outerboundary while on Comsol_boundarylayer_problem_inner.mph, I apply it on the inner boundary.
Now if you plot Vmxx (second derivative of Vm in x) at y=z=0 from x=-1 to 0. You will see that when exciting the inner boundary the order for Vmxx is 10^-7. While exciting the outer boundary gives a jump at the boundary and the order of magnitude is 10^-3. Increasing the mesh doesnt make a difference.
This is a significant problem and I dont know what to do about it. Any help on this is greatly appreciated.
Best
Shahriar
Yes I completely agree with you that there is a lot going on in my model and it makes it difficult to debug.
I have come up with a very simple model to illustrate the problem. It only uses the magnetostatics module in 3D
I have two files
Comsol_boundarylayer_problem_outer.mph
Comsol_boundarylayer_problem_inner.mph
Both files have the same geometry of an inner sphere of R=1 in a bigger sphere of R=2. In Comsol_boundarylayer_problem_outer.mph I apply a magnetic potential of -x on the outerboundary while on Comsol_boundarylayer_problem_inner.mph, I apply it on the inner boundary.
Now if you plot Vmxx (second derivative of Vm in x) at y=z=0 from x=-1 to 0. You will see that when exciting the inner boundary the order for Vmxx is 10^-7. While exciting the outer boundary gives a jump at the boundary and the order of magnitude is 10^-3. Increasing the mesh doesnt make a difference.
This is a significant problem and I dont know what to do about it. Any help on this is greatly appreciated.
Best
Shahriar
Hi
you have a nice, but complex and challenging model, I really like those ;)
Especially complex for us to get into without studying extensively the litterature and your model. With a mix of so many physics you really have to go through it step by step to be sure one understands the results at each level, I'm sure you have, but for us, out here, it's very time consuming !
I would, in youre place, attempt to isolate the boundary discrepancy you show with less physics involved. These are typically the types of issues I submit to COMSOL support, but I believe even for them to follow, either you provide them with much more info, and references, or you manage first to reduce the model complexity drastically, such to get a quick reply.
I agree with you that it looks like a serious gradient step over the boundary that is generated only for one of the two cases. Perhaps, by using week constraints "on" and redefinig some of the physics along the borders it could change, but I reconised I would need a week or two to do that, and really I do not have the free time.
One "point" in the Navier-Stokes section, you define the pressure in one point P9, that I understand, but why "0" for the pressure ? I would use a value >0, but again, that depends on your material parameters, if they do not depend on "p" it's probably OK. In case you choose some of the "Material Library" values, these are often depending on "p" and with "0" pressure you might also have "0" density, "0" conductivity, viscuosity etc
Pls keep us informed and tell us where you expect to publish this, and why not in a future COMSOL Conference procceding too
Good luck
Ivar