FEA question

[opster]

Hello,

I'm designing a magnetic device very similar to a speaker circuit (i.e. a bottom and top plate, a central pole, and a ring magnet). The only exception is that I'm adding an adjustable leakage path on the outside diameter, and inserting a coil in the empty space between the ring magnet and central pole. Also instead of an air gap where the voice coil usually is there will be a somewhat magnetic material.

OK with that said I did a magnetic circuit analysis and then made a spreadsheet so that I could quickly look at the effect of changing variables. After that I used Ansoft's "Maxwell" software to run a 2D (radial) model. The FEA results and my spreadsheet were in very good agreement. Am I right to expect that the 2D model in this case is better than a 3D model?

So my other question is how accurate should I expect the model to predict the physical device? Assuming all the B-H curves I used in the analysis accurately reflect the construction materials, and I do indeed construct the device as I modeled it. I've heard that flux densities usually end up being half of what the model predicts.

 

[MagMike]

Since your model is rotationally symmetric, the 2D analysis is just as valid (not better) as a 3D analysis would be.

Regarding your question of accuracy: I've found my rotationally symmetric 2D models tend to overestimate forces by 20% to 30%. I attribute most of that difference to small misalignments in the construction. It's impossible to keep things perfectly concentric and small shifts can lead to large changes in performance.

As you mentioned, it's also very important to make sure the B-H curves are accurate too.

 

[MJR2]

My analysis is generally as close as the accuracy of data can be taken. However that is with circuits with which I have familiarity. 2D is very acceptable if the circuit is as MagMike says rotationally symmetric. Accurate B-H cuves are important. We have developed our own and that improves accuracy. Continue to refine and run your model to improve accuracy. A higher order model and much greater mesh desity is needed when looking at force. The neat part of this is 2D can be run so much faster than 3D. My results are using Infolytica Magnet.


Mike