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.
Does the solution depends on the size of the "surroundings"? Weird case!
Posted 12 nov. 2010, 14:09 UTC−5 9 Replies
Please login with a confirmed email address before reporting spam
Hello to all:
In 2D axisymmetry I draw two rectangles and a small circle. A small rectangle representing a magnet, the small circle representing the cross section of a single loop coil and both of these inside a big rectangle representing the "surroundings" for the problem, which in this case is only air.
I needed to do integrations of the field variable along the axis of the coil which is also the axis of the magnet, so before setting the conditions and meshing I also draw several horizontal lines over which I later did the boundary integration (using the compute surface integral option). I also draw a vertical line (parallel to the axis of the coil) to limit the segments of the horizontal lines on which I was going to integrate and every time I finish all the integrations I started a new model, now with a bigger distances from the magnet to the coil. I was moving the circle (the coil ) and this vertical line step by step for increasing distance between the magnet and the coil (the vertical line is tangent to the outer border of the coil). In ALL cases the phisycs settings, the boundary conditions, the subdomain conditions was the same, only the position of the circle and the vetical line were different, but WHEN I GOT TO THE BOUNDARY OF THE SURROUNDINGS, THE VALUES OFTHE INTEGRATION ON ALL THE HORIZONTAL LINES GOT CRAZY. (PLEASE SEE ATACHMENT 5-PLOTS), SO I MADE A BIGGER RECTANGLE FOR THE SURROUNDINGS AND DO THE INTEGRATIONS AGAIN AND THE PLOT WAS NICE, BUT THE VALUES HAD NO-SENSE WITH THE PREVIOUS ONES.
I HAVE DONE THE SAME MODEL WITH DIFFERENT SIZE OF THE "SURROUNDINGS" AND THE ANSWERS ARE DIFFERENT,SO MY QUESTION IS "DOES THE SOLUTIONS DEPEND ON THE SIZE OF THE "SURROUNDINGS" ONES SELECTS?
Thank you for you time and your help, I am attaching pictures of the situation to explain better what I try to say!
In 2D axisymmetry I draw two rectangles and a small circle. A small rectangle representing a magnet, the small circle representing the cross section of a single loop coil and both of these inside a big rectangle representing the "surroundings" for the problem, which in this case is only air.
I needed to do integrations of the field variable along the axis of the coil which is also the axis of the magnet, so before setting the conditions and meshing I also draw several horizontal lines over which I later did the boundary integration (using the compute surface integral option). I also draw a vertical line (parallel to the axis of the coil) to limit the segments of the horizontal lines on which I was going to integrate and every time I finish all the integrations I started a new model, now with a bigger distances from the magnet to the coil. I was moving the circle (the coil ) and this vertical line step by step for increasing distance between the magnet and the coil (the vertical line is tangent to the outer border of the coil). In ALL cases the phisycs settings, the boundary conditions, the subdomain conditions was the same, only the position of the circle and the vetical line were different, but WHEN I GOT TO THE BOUNDARY OF THE SURROUNDINGS, THE VALUES OFTHE INTEGRATION ON ALL THE HORIZONTAL LINES GOT CRAZY. (PLEASE SEE ATACHMENT 5-PLOTS), SO I MADE A BIGGER RECTANGLE FOR THE SURROUNDINGS AND DO THE INTEGRATIONS AGAIN AND THE PLOT WAS NICE, BUT THE VALUES HAD NO-SENSE WITH THE PREVIOUS ONES.
I HAVE DONE THE SAME MODEL WITH DIFFERENT SIZE OF THE "SURROUNDINGS" AND THE ANSWERS ARE DIFFERENT,SO MY QUESTION IS "DOES THE SOLUTIONS DEPEND ON THE SIZE OF THE "SURROUNDINGS" ONES SELECTS?
Thank you for you time and your help, I am attaching pictures of the situation to explain better what I try to say!
Attachments:
- 3. The longest distance from the magnet, the coil is tangent to the problem bounday.jpg
- 4. The bigger surroundings.jpg
- 5. Plots of Magnetic Flux.jpg
- 1. Permanent Magnet in 2D axisymmetry, 1 single looop coil and air surroundings.jpg
- 2. Intgrating over the horizontal lines to get the magnetic flux.jpg
9 Replies Last Post 16 déc. 2010, 04:27 UTC−5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
for me if a solution is depending on the "surrounding" it's because you have an effect due to a gradient that differs depending on the size of the surrounding. This is rather frequent in ACDC as the field lines vary stronlgy if you take into account a little or a large "air/vacuum" area, or mif you also use infinite elements.
Another ieffect is due to gradient concentration typically around sharp edges, this is well known as stress concetration effects in structural (many shafts have broken due to this) but often less known is the field concentration around sharp edges in electromagnetic theory. And a MAxwell tensor theory is based on gradient integration, the correct resolution of these gradients, and the homogenity of the meshing, particulaly around numerical "stress" points can give serious and large errors.
Now I have not had enought time to study your particular case, but I suspect that you are hitting some of these effects. It's worth to test it out as these are the essential skills/knowledge you need to correctly model ACDC effects
furthermore your integration lines are similar to the "projection coupling operator, its wort to take a look at hese too ;)
--
Good luck
Ivar
for me if a solution is depending on the "surrounding" it's because you have an effect due to a gradient that differs depending on the size of the surrounding. This is rather frequent in ACDC as the field lines vary stronlgy if you take into account a little or a large "air/vacuum" area, or mif you also use infinite elements.
Another ieffect is due to gradient concentration typically around sharp edges, this is well known as stress concetration effects in structural (many shafts have broken due to this) but often less known is the field concentration around sharp edges in electromagnetic theory. And a MAxwell tensor theory is based on gradient integration, the correct resolution of these gradients, and the homogenity of the meshing, particulaly around numerical "stress" points can give serious and large errors.
Now I have not had enought time to study your particular case, but I suspect that you are hitting some of these effects. It's worth to test it out as these are the essential skills/knowledge you need to correctly model ACDC effects
furthermore your integration lines are similar to the "projection coupling operator, its wort to take a look at hese too ;)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you Ivar, for your answer.
I did not understand very well when you wrote "It's worth to test it out as these are the essential skills/knowledge you need to correctly model ACDC effects" Can you explain this a little more to me?
And you also wrote "your integration lines are similar to the "projection coupling operator, its wort to take a look at hese too ". I have been looking at the COMSOL 3.5a manual for something related to "projection coupling operator" and I have found nothing with that name. What is a "projection coupling operator"? Can I used it to calculate the integrals I am doing with the horizontal lines?
Thank you,
Igor.
I did not understand very well when you wrote "It's worth to test it out as these are the essential skills/knowledge you need to correctly model ACDC effects" Can you explain this a little more to me?
And you also wrote "your integration lines are similar to the "projection coupling operator, its wort to take a look at hese too ". I have been looking at the COMSOL 3.5a manual for something related to "projection coupling operator" and I have found nothing with that name. What is a "projection coupling operator"? Can I used it to calculate the integrals I am doing with the horizontal lines?
Thank you,
Igor.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
sorry, I'm now using so much v4 so I get confused with the old/new naming convention: operators appear in v4, replacing "coupling variables" of 3.5a, check the projection coupling variable in 3.5a
--
Good luck
Ivar
sorry, I'm now using so much v4 so I get confused with the old/new naming convention: operators appear in v4, replacing "coupling variables" of 3.5a, check the projection coupling variable in 3.5a
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi, I have been reading what you recommended to me about the PROJECTION COUPLING VARIABLES but I still don't understand.
I need to do boundary integration along r (multiplying the integrand by 2*pi*r before integration, because I need magnetic flux, so the integrations are over the area perpendicular to the magnetic flux density lines). I'm working on 2D axisymmetric.
The COMSOL MULTIPHYSIC'S User's Guide (I have 3.5 a) says:
" COMSOL Multiphysics uses a method of the source whereby it first apply a one-to-one transformation to the mesh of the source domain. The last space dimension in the transformed mesh is the one integrated over, so THE LINES USED TO INTEGRATE ARE VERTICAL IN THE TRANSFORMED MESH.
I need the integrations to be done over horizontal lines and need to do a plot of the integrations over for different values of z VS z!
Can you help me with this? It would really help me a lot, because the method I am using requieres a lot of time, since I need to do the integration line by line, and then with the values, make the plot in EXCEL.
Thank you. Thank you very much!
I need to do boundary integration along r (multiplying the integrand by 2*pi*r before integration, because I need magnetic flux, so the integrations are over the area perpendicular to the magnetic flux density lines). I'm working on 2D axisymmetric.
The COMSOL MULTIPHYSIC'S User's Guide (I have 3.5 a) says:
" COMSOL Multiphysics uses a method of the source whereby it first apply a one-to-one transformation to the mesh of the source domain. The last space dimension in the transformed mesh is the one integrated over, so THE LINES USED TO INTEGRATE ARE VERTICAL IN THE TRANSFORMED MESH.
I need the integrations to be done over horizontal lines and need to do a plot of the integrations over for different values of z VS z!
Can you help me with this? It would really help me a lot, because the method I am using requieres a lot of time, since I need to do the integration line by line, and then with the values, make the plot in EXCEL.
Thank you. Thank you very much!
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi, I have been reading about the PROJECTION COUPLING VARIABLES but I still don't understand.
I need to do boundary integration along r (multiplying the integrand, Bz_emqa, by 2*pi*r before integration, because I need magnetic flux, so the integrations are over the area perpendicular to the magnetic flux density lines). I'm working on 2D axisymmetric.
The COMSOL MULTIPHYSIC'S User's Guide (I have 3.5 a) says:
" COMSOL Multiphysics uses a method of the source whereby it first apply a one-to-one transformation to the mesh of the source domain. The last space dimension in the transformed mesh is the one integrated over, so THE LINES USED TO INTEGRATE ARE VERTICAL IN THE TRANSFORMED MESH.
I need the integrations to be done over horizontal lines and need to do a plot of the integrations over for different values of z VS z!
Some help about this would really be good for me a lot, because the method I am using requieres a lot of time, since I need to do the integration line by line, and then with the values, make the plot in EXCEL.
Thank you. Thank you very much!
I need to do boundary integration along r (multiplying the integrand, Bz_emqa, by 2*pi*r before integration, because I need magnetic flux, so the integrations are over the area perpendicular to the magnetic flux density lines). I'm working on 2D axisymmetric.
The COMSOL MULTIPHYSIC'S User's Guide (I have 3.5 a) says:
" COMSOL Multiphysics uses a method of the source whereby it first apply a one-to-one transformation to the mesh of the source domain. The last space dimension in the transformed mesh is the one integrated over, so THE LINES USED TO INTEGRATE ARE VERTICAL IN THE TRANSFORMED MESH.
I need the integrations to be done over horizontal lines and need to do a plot of the integrations over for different values of z VS z!
Some help about this would really be good for me a lot, because the method I am using requieres a lot of time, since I need to do the integration line by line, and then with the values, make the plot in EXCEL.
Thank you. Thank you very much!
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
normally you exchange y for x and you get horizontal, its just it must be done at the right place ;)
havent used 3.5a for some months will try to propose a simple model. But in the mean time, try it out on a rectangular shape with some variable of known gradient. (start simple to debug/understand)
--
Good luck
Ivar
normally you exchange y for x and you get horizontal, its just it must be done at the right place ;)
havent used 3.5a for some months will try to propose a simple model. But in the mean time, try it out on a rectangular shape with some variable of known gradient. (start simple to debug/understand)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
THANK YOU IVAR!!
I DID IT!! I HAVE SAVED A LOT OF TIME DOING IT THE WAY YOU TOLD ME (USING PROJECTION COUPLING VARIABLES) BUT THERE'S A LITTLE DIFFERENCE (MAYBE 6%) IN THE MAXIMUM VALUES OF MY CURVES.
I computed the integration of (Bz_emqa*2*pi*r) over every horizontal line (46 in total) up to the vertical line I also put in the model, at the right, before the right outer bundary, to establish and end for this "boundary integrations)-
then I had the data of MAGNETIC FLUX VS Z FOR -4.5E-3<=Z<=4.5E-3.
I COMPUTED THE INTEGRATION OF (Bz_emqa*2*pi*r)*((r>=0)&&(r<=5E-3)) FOR ALL THE VALUES OF Z IN MY MODEL. THEN I EXPORTED THEM USING THE EXPORT POST PROCESSING DATA OPTION. THE VALUES OF THE INTEGRAL, WHICH I CALLED MFLUX, WERE EXPORTED USING THE SAME COORDINATES OF THE Z VALUE OF THE HORIZONTAL LINES OF THE FIRST MODEL. THEN I PLOT BOTH GRAPHS IN EXCEL.
QUESTIONS:
1. WHY, THERE'S A DIFFERENCE IN THE VALUE OF THE INTEGRAL, AND IT INCREASES AS I GET CLOSER TO THE MAXIMUM VALUE (MIDDLE OF THE MAGNET)
2. IF EVERYTHING IS EXACTLY THE SAME, WHY THE MAXIMUM VALUE OF THE FIELD VARIABLE (NOT OF MY INTEGRAL, BUT THE VALUE OF MAGNETIC FLUX DENSITY) IS DIFFERENT IN BOTH MODELS.
3. IN THE EXAMPLE I SAW IN THE MANUAL FOR PROJECTION COUPLING VARIABLES, THEY MADE THEM AVAILABLE IN THE LEFT BORDER OF THEIR SQUARE. IF I WANT TO MAKE IT AVAILABLE OVER ALL THE DOMAINS ?AS I DID, BECAUSE IF NOT I COULDN'T MAKE A PLOT) IS THERE ANY CONSIDERATION I SHOULD TAKE INTO ACCOUNT? OR JUST CHANGE IN THE DESTINATION TAB THE BOUNDARY OPTION FOR THE SUB DOMAIN OPTION?
I'M ATTACHING PICTURES TO EXPLAIN BETTER THE SITUATION.
THANK YOU, AGAIN, THANK YOU A LOT!!!
I DID IT!! I HAVE SAVED A LOT OF TIME DOING IT THE WAY YOU TOLD ME (USING PROJECTION COUPLING VARIABLES) BUT THERE'S A LITTLE DIFFERENCE (MAYBE 6%) IN THE MAXIMUM VALUES OF MY CURVES.
I computed the integration of (Bz_emqa*2*pi*r) over every horizontal line (46 in total) up to the vertical line I also put in the model, at the right, before the right outer bundary, to establish and end for this "boundary integrations)-
then I had the data of MAGNETIC FLUX VS Z FOR -4.5E-3<=Z<=4.5E-3.
I COMPUTED THE INTEGRATION OF (Bz_emqa*2*pi*r)*((r>=0)&&(r<=5E-3)) FOR ALL THE VALUES OF Z IN MY MODEL. THEN I EXPORTED THEM USING THE EXPORT POST PROCESSING DATA OPTION. THE VALUES OF THE INTEGRAL, WHICH I CALLED MFLUX, WERE EXPORTED USING THE SAME COORDINATES OF THE Z VALUE OF THE HORIZONTAL LINES OF THE FIRST MODEL. THEN I PLOT BOTH GRAPHS IN EXCEL.
QUESTIONS:
1. WHY, THERE'S A DIFFERENCE IN THE VALUE OF THE INTEGRAL, AND IT INCREASES AS I GET CLOSER TO THE MAXIMUM VALUE (MIDDLE OF THE MAGNET)
2. IF EVERYTHING IS EXACTLY THE SAME, WHY THE MAXIMUM VALUE OF THE FIELD VARIABLE (NOT OF MY INTEGRAL, BUT THE VALUE OF MAGNETIC FLUX DENSITY) IS DIFFERENT IN BOTH MODELS.
3. IN THE EXAMPLE I SAW IN THE MANUAL FOR PROJECTION COUPLING VARIABLES, THEY MADE THEM AVAILABLE IN THE LEFT BORDER OF THEIR SQUARE. IF I WANT TO MAKE IT AVAILABLE OVER ALL THE DOMAINS ?AS I DID, BECAUSE IF NOT I COULDN'T MAKE A PLOT) IS THERE ANY CONSIDERATION I SHOULD TAKE INTO ACCOUNT? OR JUST CHANGE IN THE DESTINATION TAB THE BOUNDARY OPTION FOR THE SUB DOMAIN OPTION?
I'M ATTACHING PICTURES TO EXPLAIN BETTER THE SITUATION.
THANK YOU, AGAIN, THANK YOU A LOT!!!
Attachments:
- Integration of (Bz_emqa)(2)(pi)(r) vs z from -4.5 to 4.5 [mm] using horizontal lines for integration AND the same integration using PROJECTION COUPLING VARIABLES.jpg
- Model without the horizontal lines used for integration.jpg
- same model, using horizontal lines for integration, and a vertical line to establish an end for the boundary integration.jpg
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
have a look at the discussion on
www.comsol.eu/community/forums/general/thread/439/#p31885
It might also give you some ideas ;)
--
Good luck
Ivar
have a look at the discussion on
www.comsol.eu/community/forums/general/thread/439/#p31885
It might also give you some ideas ;)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you,
I had made a mistake. The real plot is attached here. (although it oscillates a little near the maximum value)
The name I used for the PROJECTION COUPLING VARIABLE is Mflux. I am trying to export the results of d(Mflux,z) I mean the derivative of Mflux with respect to z, so I can make a plot of d(Mflux,z) vs z but I get 0.00 for all the values. The problem is static, so I don't think I have any problem with time.
I had made a mistake. The real plot is attached here. (although it oscillates a little near the maximum value)
The name I used for the PROJECTION COUPLING VARIABLE is Mflux. I am trying to export the results of d(Mflux,z) I mean the derivative of Mflux with respect to z, so I can make a plot of d(Mflux,z) vs z but I get 0.00 for all the values. The problem is static, so I don't think I have any problem with time.