Virtual replication of RomeroUK motor/generator by 3D FEM modeling

[teslaalset]

@ teslaalset

I'm studying your model and the results, as we speak, and I'll comment more intelligently shortly,  but thank you for doing these simulations and models, they look great!.

EM



PS.   Ok, here's some questions teslaalset,   

How do you derive the force that you plot?   Do you first use the calculator inside Maxwell to sum up the surface forces vectoraly, and then take this resultant vector and do a dot product with a unit vector in the direction of the motion?  this is how it should be done from what I remember.

also, it would be nice to produce a smoother force profile, and I would expect it from the software as this is a very simple model.  So I'm wondering how you actualy modeled the coil?  And I mean the coil not the ferrite core?

And most importantly, make sure you use the SAME mesh for all the 3 different scenarios,  and use a highest fidelity mesh generated with the closest gap and the magnet at TDC, and make sure you tell the mesher to use symetry, as the mesh will not always be symetric just because your model is symetric.   Symetry of the mesh, about the TDC line, is very important in this setup as we are trying to look for an inbalance in the forces. 

Anyway, give those comments some thought.    Good work!

@EMdevices,
Good to see some OU veterans having a look at my attempts here.
Forces: I can set up a torque or force measurement while selecting one or more component in the model. In this case I selected the moving magnet for calculating the force on it.
I am not sure how Maxwell processes its internal data for it. It's made available via the GUI of the program as an output parameter (I use Maxwell V14, the latest version of Ansys).

As for defining the coil, this is done via defining the physical area and define a winding + terminal, while also defining single or multi strand. So, even RomeroUKs 7 strand Litze can be used.

I agree that the force output is quite noisy, but like I mentioned, I chose a rather rough mesh to have an acceptable calculation throughput time.
Finer mesh is ending up in a 'night batch', since it will take several hours .
It's not my PC's capabilities, I just assembled a number crusher out of an I7 Intel processor, 16GB of RAM and a SATA 600 SSD.
I used identical meshes for all three force runs.
To allow for fine local meshing I put 'airboxes' around the magnets so the mesh around the magnets have more details locally.
I have to look into symmetrical mesh, that is a good tip and makes sense.
Thanks for helping out here, there are not many FEM specialists around at OU.

What FEM software do you normally use?

[teslaalset]

Some of the things I have on the agenda for simulation is the determination of the delayed BEMF forces.

To explain a bit more, look at the attached graph (just a drawing, no simulation).
Negative vertical axis represent counter force towards the rotational direction, positive vertical values represent pushing (driving) forces in the direction of the movement.

- the blue line represents the force on the magnet without any electrical load to the stator coil while that magnet is passing by a stator coil/core combination. Zero on the timescale is TDC (top dead centre). First the rotor magnet will be attracted to the ferrite core, once it passes the core, it will be attracted in the counter wise direction.
So, attraction in the direction of the motion is has a positive force value, attraction in opposite the direction of motion is negative. Without taking losses into account, the sum (integral) of the forces will be zero in such situation
- The red line represents force caused by non delayed BEMF when a load is connected to the coil. All the time the forces caused by BEMF work as negative force against the direction of motion.
- The grant total of a loaded coil setup is represented by the green line. The integral of this grant total is negative.

So, what I am planning to do is finding out what the delayed BEMF force curve looks like when the bias magnets are in optimum position (sum of all forces are zero over the complete stroke), so we know how the delayed BEMF force curve really looks like.
That will help in understanding the BEMF current in the generator coils, and also the flux behaviour in the core.
This is done via indirect simulation. First I plan to determine the 'blue curve' , so no load.
Then I do the loaded coil simulation again, like the second force graph in my posting #2 of this thread (stator bias magnet position at 0.05). That will give me the sum force graph like the green one in below graph.
Then subtract those two results and the delayed BEMF force curve will be the result (the red graph in below picture).
My expected result will be quite different from the red curve as shown below.

Maybe this is more complicated than that, since the core will also change it's mechanical attraction due to (time and location dependant) saturation. We'll see.....

[teslaalset]

For the guys in the experimenters thread some graphs of the generator coil signals.
First graph indicates the currents induced in all 9 coils time wise
Second graph indicates the flux induced in all 9 coils time wise
Third graph indicates the induced voltages in all 9 coils time wise
Fourth graph has all 3 signal for only coil nr 1 represented

[teslaalset]

Some flux animations within two opposite cores in the complete model.
The AVI videos can be downloaded here:

Vectored version:
http://www.multiupload.com/WEQJNJ7XYM

Cloud version:
http://www.multiupload.com/O72CBIMJN2

The videos are probably best to be viewed in looped mode.

Below a snapshot from vector version video: