Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Rotating magnetic field in eddy current testing

Please login with a confirmed email address before reporting spam

Hi, I'm trying to make this magnetic model : I have a alluminum object with a defect inside. I want to recognize it, applied a rotating magnetic field and plotting the difference between the object with defect and the object without defect.
Unfortunally this plot are much different in ANY point of these, and not only in the position of the defect, such as I want.
Maybe I make a mistake in the phiscal model, I don't know. I'm coming from 3.3, and 4.3 result me confusionary.
Thanks for your support.

EDIT : Sorry I can't upload MPH file (size limit).
This is the setting that I choose :
AC-DC, magnetic filed (MF). I create a square (air material), my object (aluminum) and the defect (air).
I create two magnetic filed:
first magnetic field for inferior and superior border
Hx=cos(omega*t)
Hy=0
Hz=0

second magnetic filed for right and left border :
Hx=0
Hy=sin(omega*t)
Hz=0

This is the only things I do, then I mesh, simulate and plot Bx in function of x and By in function of y (I choose various line x and y, with the option plot 1D line)

6 Replies Last Post 8 oct. 2012, 12:47 UTC−4
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 6 oct. 2012, 16:01 UTC−4
Hi

moving from 3.3 to 4.3 is a step, but I suppose you have already found many interesting improvements ;)

One comment for uploading, as with v4 all steps are stored and a model can bea easily recalculated by any of us, to load your model to the Forum, do first an : Edit Clear All solutions, Clear All Mesh, File reset history, then File save As with a new name and upload this reduced file.

Then for comparing two results, in V4 you have the Data-Set "Join" option to make a difference (and other fucntions) of two models such that you can compare more easily the two

And a last general comment, in ACDC its imporant to avoid "sharp edges" as the derivatives of the fields are not defined at these points (singularties), I alway fillet off all my sharp corners, also of my surrounding air regions, if I simple do not use spherees or circles for the air. When using Inf elements, sometimes its less coputational to use cubic/block shapes for the air

Last point ACDC and MF in time domains is limited, the equations are partly truncaded (from my understanding), the frequency domain analysis seem to be complete (includes 2nd order derivatives of the dependent variable A).
Perhaps your differences is due to transient behaviour and a frequency domain "harmonic" development will work better, use the phasor (complex amplitudes) to offset youre BCs

--
Good luck
Ivar
Hi moving from 3.3 to 4.3 is a step, but I suppose you have already found many interesting improvements ;) One comment for uploading, as with v4 all steps are stored and a model can bea easily recalculated by any of us, to load your model to the Forum, do first an : Edit Clear All solutions, Clear All Mesh, File reset history, then File save As with a new name and upload this reduced file. Then for comparing two results, in V4 you have the Data-Set "Join" option to make a difference (and other fucntions) of two models such that you can compare more easily the two And a last general comment, in ACDC its imporant to avoid "sharp edges" as the derivatives of the fields are not defined at these points (singularties), I alway fillet off all my sharp corners, also of my surrounding air regions, if I simple do not use spherees or circles for the air. When using Inf elements, sometimes its less coputational to use cubic/block shapes for the air Last point ACDC and MF in time domains is limited, the equations are partly truncaded (from my understanding), the frequency domain analysis seem to be complete (includes 2nd order derivatives of the dependent variable A). Perhaps your differences is due to transient behaviour and a frequency domain "harmonic" development will work better, use the phasor (complex amplitudes) to offset youre BCs -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 oct. 2012, 03:54 UTC−4
Thanks Ivar for reply.
First, I upload the file without mesh and solution, and now you can wiev my current work.
Second, I try to make both the strategy that you advice me, but I failed : for Union, I make Union of 2 line but plot show me ever a plot of first solution; then I make union of 2 solution, but not work. Frequency domain don't work, show me an error (of variable t) in the stationary solutor.
Thanks Ivar for reply. First, I upload the file without mesh and solution, and now you can wiev my current work. Second, I try to make both the strategy that you advice me, but I failed : for Union, I make Union of 2 line but plot show me ever a plot of first solution; then I make union of 2 solution, but not work. Frequency domain don't work, show me an error (of variable t) in the stationary solutor.


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 7 oct. 2012, 04:33 UTC−4
Hi

yes the frequency domain does not use "t", not "cos(omega*t)", you define the amplitudes, as H0 and j*H0 for the cosinus respectively sinus components.

Then your results become complex and you need to use the phasor (such as atan2(imag(Bx),real(Bx)) repsectively By to get the relative time =phae angle component within one periode

Then I see that the sharp edges of your model (Al part) are driving the field strength, you should give these a reasonable fillet radius, else I suspect your filed gradients will be too wrong


--
Good luck
Ivar
Hi yes the frequency domain does not use "t", not "cos(omega*t)", you define the amplitudes, as H0 and j*H0 for the cosinus respectively sinus components. Then your results become complex and you need to use the phasor (such as atan2(imag(Bx),real(Bx)) repsectively By to get the relative time =phae angle component within one periode Then I see that the sharp edges of your model (Al part) are driving the field strength, you should give these a reasonable fillet radius, else I suspect your filed gradients will be too wrong -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 oct. 2012, 05:53 UTC−4
Thanks. I change H0cos(omega*t) with H0 and H0*sin(omega*t) with j*H0. Now simulation work, but I don't find the real and imaginary part for plot, and if I plot the magnetic field on plane (Bx and By component) its result such as time=0 in the time domain.

What is the option for make circualar fillet edge? I think a radius of E-4 is enough, consider the dimension of the defect is [mm] (E-3)

This simulation look like simply but make me crazy, I don't know why in the no-defect simulation there are a random variation of magnetic field

EDIT : I do that. I upload an image plotted with matlab, comparison of defect-no defect, for Module(Bx) in function of x.
I don't understand why module increase at 5.5 meter in defect simulation.
Thanks. I change H0cos(omega*t) with H0 and H0*sin(omega*t) with j*H0. Now simulation work, but I don't find the real and imaginary part for plot, and if I plot the magnetic field on plane (Bx and By component) its result such as time=0 in the time domain. What is the option for make circualar fillet edge? I think a radius of E-4 is enough, consider the dimension of the defect is [mm] (E-3) This simulation look like simply but make me crazy, I don't know why in the no-defect simulation there are a random variation of magnetic field EDIT : I do that. I upload an image plotted with matlab, comparison of defect-no defect, for Module(Bx) in function of x. I don't understand why module increase at 5.5 meter in defect simulation.


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 7 oct. 2012, 10:59 UTC−4
Hi

do not underestimate the complexity of even a "simple" simulation, there are many things to think of. If you work in frequency domain you must also define the frequency in the harmonic solver node, and estimate the corresponding skin depth manually as the AC field will not enter into the aluminium material much further than a few "skin depths"

To plot the real and imaginary parts, you need to use the real() and imag() operators in the plot expressions

Fillet is an option of the geometry, you select the points where you want the radius to apply, make them realisitic w.r.t. the meter large model you have, or is there a scaling effect I missed ?

the EM field will concentrate on the edges and the vertex, as you have sharp edges, you get strong fields locally and shielding effects that make the observation of your air slot, less than "evident".

You could test it out on a simple Al (foilleted) rectangle with a round hole in the middle, which is simpler than your L bracket beam profile

You do not need to go to matlab to plot line curves, you can do that either on your edges (boundaries) or on Data Set Cut lines you might define

--
Good luck
Ivar
Hi do not underestimate the complexity of even a "simple" simulation, there are many things to think of. If you work in frequency domain you must also define the frequency in the harmonic solver node, and estimate the corresponding skin depth manually as the AC field will not enter into the aluminium material much further than a few "skin depths" To plot the real and imaginary parts, you need to use the real() and imag() operators in the plot expressions Fillet is an option of the geometry, you select the points where you want the radius to apply, make them realisitic w.r.t. the meter large model you have, or is there a scaling effect I missed ? the EM field will concentrate on the edges and the vertex, as you have sharp edges, you get strong fields locally and shielding effects that make the observation of your air slot, less than "evident". You could test it out on a simple Al (foilleted) rectangle with a round hole in the middle, which is simpler than your L bracket beam profile You do not need to go to matlab to plot line curves, you can do that either on your edges (boundaries) or on Data Set Cut lines you might define -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 8 oct. 2012, 12:47 UTC−4
I have a lot of problem wit this simulation, but there are a primary problem I have. I think this problem is consequence of a mistake in simulation setting, or geometry. I have make 2 MPH, one with defect, one without defect. I have created the second for coping the first and deleted the geometrical element of defect and mesh and simulated again. The problem is : the graph Bx in function of x is IDENTICAL (in the zone of interest). Is it possible? I have done all your hint, filled edge ecc ecc.

I want to try this, but I don't know how : make the SAME IDENTICAL mesh in a defect and no defect simulation. It's possible?

EDIT : I do the test, not work, defect not recognized, I don't know why. It's a simple geometry, wich a 50 hz AC current the filed should recognized the defect.
I have a lot of problem wit this simulation, but there are a primary problem I have. I think this problem is consequence of a mistake in simulation setting, or geometry. I have make 2 MPH, one with defect, one without defect. I have created the second for coping the first and deleted the geometrical element of defect and mesh and simulated again. The problem is : the graph Bx in function of x is IDENTICAL (in the zone of interest). Is it possible? I have done all your hint, filled edge ecc ecc. I want to try this, but I don't know how : make the SAME IDENTICAL mesh in a defect and no defect simulation. It's possible? EDIT : I do the test, not work, defect not recognized, I don't know why. It's a simple geometry, wich a 50 hz AC current the filed should recognized the defect.

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.