axial flux FEMM simulation

[dinges]

After a discussion earlier this week on IRC with Oztules we ran a simulation to determine how flux density changes as a function of airgap in the common axial flux alternator. It gave some surprizing insights (at least, they were surprizing to me). So much so that I decided to run a more complete simulation, the results of which are presented here.

I've simulated flux density for the common axial flux generator, over a range of 5 mm to 25 mm airgap. The magnets are grade N40, size 2"x1"x.5". The spacing between the magnets is one magnet width, 1". The rotor plates consist of 10 mm thick AISI-1010 steel.

Below is a screenshot of one particular simulation, in this case for an airgap of 13 mm (about one magnet thickness). There are 11 of such screenshots in total which I haven't bothered to resize and upload to fieldlines. I figure most are just interested in the results anyway. For those that are interested in the individual screenshots, they are all uploaded to the IRC gallery:

http://www.anotherpower.com/gallery/album89

An example image:

The results of the simulation can be seen plotted in the graph below. On the vertical axis is flux density (Tesla), on the horizontal axis airgap (mm).

A high-resolution image of the graph can be found here:

http://www.anotherpower.com/gallery/album89/axial_flux_airgap_sim_graph

I've uploaded the CAD drawing (.dxf) and the FEMM .fem file to my FL files, if anyone cares to play with it some more:

http://www.otherpower.com/images/scimages/3538/axial_flux_gap_simulations_10mm_steel_N40.dxf

http://www.otherpower.com/images/scimages/3538/axial_flux_gap_simulations_10mm_steel_N40.FEM

I present this 'as is' to the community. Perhaps it's of use to anyone else too.

Regards,

Peter.

[Flux]

That makes sense. From zero gap to about 1 magnet thickness it is virtually linear. beyond that the linearity falls, making prediction more difficult.

I am not sure what flux it predicts, with smaller gaps the flux is nearly constant across the gap but with wide gaps the flux density near the magnets is far higher than half way between them. One effect of this is that much of the useful flux is wasted in the mechanical gap that can't be avoided. It tends towards two single rotors on either side of a coil if you make the stator too thick.

It would be interesting if you could do a simulation on a single rotor then others would see how bad it is especially with stators of significant thickness.

Perhaps you could include the single rotor and a single rotor with return flux disc, that would be really enlightening.

Flux

[dinges]

Good ideas.

I'll give it (single rotor and single rotor with return disc) a go when I've got some more time. Making the above simulation took up the best part of a morning.

Peter.

[oztules]

Good stuff Peter

Thanks for spending your time on this material

It appears that the leakage is significant with this setup rather than the simulations for the round magnets

http://www.anotherpower.com/gallery/dinges?page=21

.........oztules

[dinges]

The simulation for a single rotor axial flux was quick and easy; see picture below for the result.

It becomes immediately obvious how much lower the flux densities are and how much flux gets shorted.

Full resolution picture can be found here:

http://www.anotherpower.com/gallery/album89/axial_flux_single_rotor

The simulation with dual rotor (but magnets on one rotor only) will take some more time to do.

Peter.

[dinges]

Hello Oztules,

Not sure I understand this remark:

"It appears that the leakage is significant with this setup rather than the simulations for the round magnets"

Notice that the two simulation runs (the one I did today and the one we did earlier this week) are essentially the same, since FEMM is a 2D package. All sims were done using 2"x1"x.5" magnets, viewed 'head on', i.e. looking at the 1"x.5" side.

The link you gave is to simulation of earlier this week; today I've modeled for more magnets (thus reducing the end-effects of the 2D model) and a larger airgap range.

Again, please notice that FEMM is a 2D package and that what we simulate is essentially a cross-section of the rotor. So round or square magnets is not an option in these simulations.

Peter.

[dinges]

Perhaps something else is confusing you.

These simulations were run at much higher resolution than earlier this week (it took a lot longer to compute too) and I've plotted much more fieldlines; the earlier plots used just 19 lines, the new ones 59 or even more (can't remember). This means that we get a much better idea of how the fieldlines flow. Maybe this is throwing you off w.r.t. leakage flux.

Peter.

[tecker]

Might be interesting to see your test setup if you are of a mind .

[oztules]

Ah ha

Yes that would explain the extra leakage lines in the later simulations. Thanks for clearing that up. There was such a difference, it was the only possible explanation (aside from spacing)...... except for resolution, didn't see that one coming.

Thanks Peter

.......oztules

[dinges]

I've just ran the simulations for the dual rotor axial flux with one bare rotor (no magnets on the 2nd rotor). The resulting individual images can be found here:

http://www.anotherpower.com/gallery/album89?page=2

Below is one sample image, showing the situation for an airgap of 13 mm (about one magnet thickness):

The graph below shows the results of both simulations. The lower line represents the latest simulation of the dual rotor with magnets on only one rotor. As can be seen, not  only is flux density a lot smaller for a given airgap, it also decreases more rapidly as airgap increases. Keep in mind that the dual rotor with magnets on one rotor has only half the magnets of a normal dual rotor axial flux genny though.

The full resolution graph can be found here:

http://www.anotherpower.com/gallery/album89/graph_simulations?full=1

Peter.

[Flux]

Thanks Peter.

In fact the single rotor with return path does well. It will need to work with half the gap to be a reasonable comparison. Its biggest drawback is that the thing still needs two completely useless mechanical gaps and that wastes more winding space with a narrow total gap.

For a given number of magnets it is more effective than using a dual rotor with half the number of poles.

I appreciate your effort and am reluctant to ask for anything else but there is one other thing that interests me. What would the effect be of a single rotor with return disc if the return disc has mild steel polar projections instead of a smooth surface.

I don't think the projections need to be large. Perhaps the equivalent of blank magnets 2 x 1 x 1/4",  instinct makes me think the leakage will be less but the gain may not be worth the effort.

Flux

[vawtman]

Thanks for doing these tests Peter :>)

[dinges]

...Your wish is my command...  

I intuitively doubt it will be an improvement over the plain rotor, but we'll soon know.

Yes, I noticed the same: for a given amount of magnets, the dual rotor (with bare 2nd disc) is most effective. (ignoring everything else).

Peter.

[dinges]

Flux,

I've run the simulations using your suggestion: a 2nd rotor disc with protrusions.

The protrusions have a height of 6mm (~1/4") and the same width as the magnets.

Below is an example image, for an airgap of 13 mm (approximately one magnet high).

I didn't bother to plot the flux densities in the graph, as the curve is exactly the same as for the plain (bare) return disc. The explanation for that is the flux densities I've plotted are at the middle of the protrusions (near the steel disc (Y-position just below the steel disc/protrusion), at the X-position of the middle of the magnet). The effect of the protrusions, however, is most markedly present at the edges of the magnets, not the middle.

Still, one can see an improvement using protrusions on the return disc. It definitely helps to focus the flux lines a bit more and prevents them from wavering out. It even makes sense now, in hindsight Whether it's worth the extra expense and effort of the machining may be another matter though.

The full resolution images can be found here:

http://www.anotherpower.com/gallery/album89?page=3

Hope this was of help,

Peter.

[Flux]

Thanks Peter that is great.

It does confirm my suspicions. I don't have any 2 x 1  x 1/2" magnets to do a direct comparison with your plots, but from my tests with a fluxmeter and 46 x 30 x 10 mm magnets it all seems to tie up well enough.

I am not quite sure what the effects will be with the variable gap between magnets caused by the spacing on a disc but in general the results come close to my predictions and your graphs of flux density against air gap should be a good starting point for those who want to play with different air gaps.

I usually base things on 600mT in a gap 1.4 times magnet length for a dual rotor. That lets you use a stator about magnet thickness and have safe mechanical gaps.

On the single rotor with return path I didn't include the polar projections but it looks as though it would be worthwhile. I always pot the magnets to prevent corrosion in our wet climate but the fan effect of the projections would help cooling and increase the flux linkage.

Thanks again for your efforts.

Flux

[SparWeb]

Peter,

I just caught up with your thread.  FEMM teaches a lot with its graphical presentation.

But, you still haven't put it to its full use, yet.  Try integrating the Sum of Flux across each pole.  Try to make the integration path the same size as a coil, to be realistic.

Have you compared your results with the results I posted last year?  They seem to be about the same.  If you're intested, I can dig up the old links...

[SparWeb]

Sorry, Peter, in my last comment I was a bit rushed.

Last year I was trying to correlate various voltage results, from spinning the

rotors over one coil, against the predictions from FEMM.  The chart above was

made from test results from my 16-pole rotor.

I also used a feature of FEMM that plots the flux across a line on the screen

- see below:

In my first models I only measured from magnet to magnet, then across two magnets.

Later I realized that what I really wanted was a line of integration equal to the

diameter of the coil, so if the average coil diameter is 2" in diameter, then draw

a line centered on a magnet pole 2" long.

The diagram below shows the integral across 2 poles, and it reveals a troublesome

reversal of our fortunes:

So it's not obvious when you look at the field lines in the diagram above, but

it's there, and you can find it if you play with the resolution of the FEMM Solution.  

If you look at the scale legend, in the upper right of the colour solution,

I was tweaking the scale solely to make it fit whole numbers.  When

you change the parameters a little you can find the field lines that don't get

to the other rotor.

Tonight, I've stacked the various windows together as best I can into the same

screen-shot.  I find that there are still a few lines short-circuited between the

magnet and the rotor it's on, and the chart shows extra peaks, though it looks empty

of field lines.

When compared to the theory, I kept reading much higher open-circuit voltages

than the model predicted.  I suppose I shouldn't complain, usually theory over-

estimates performance, and reality is a let-down - not the other way around!  

I never figured out if it was a problem with the model, with my analysis, or

the voltage readings.  If you have more success, please share it because I,

for one, will be very happy to see it.

Thanks