Discussion Forum

Hi all,

I am attempting to model an axial flux coreless PMSG in ACDC module of comsol. You can find my model attached.

I continue to have a recurring issue where my model does not respect CoE. Meaning that the calculated output electrical power Pel = VaIa + VbIb + Vc*Ic does not equal the input mechanical power of the generator Pmech = rmm.Taxr wmech - minus the electrical losses in the machine.

I could not upload the model with solutions due to 5MB size restriction, so I try to explain the results via screenshots. You can find a plot of P_el vs P_mech attached. There is a delta around 140W. My computed electrical losses sum up to around 33W. Big difference.

I have been trying to narrow down the issue. I figure that it must be an issue with the force calculation that I am using to compute the rotor torque. The computed results of the rotor torque seem to be pretty unstable. And for the topology of the machine, the large ripple torque that I am seeing doesnt pass a sanity test. I also attached a plot of the rotor torque results.

I have read some blogs/tutorials online and understand the sensitiivity/limitations of the maxwell stress tensor to non-linear reigons, so ive tried to ensure I am computing the torque only on air reigon around the rotor. I can see that I am getting a warning message when computing the torque. So far, I have tried to research this "deprecated behavior" message on the forum, but havent found anything yet. Maybe this is a hint :)

Ive done some kind of mesh refinement study around the domains which I am computing the torque. I refined the mesh in the reigons surrounding the rotor, it doesnt seem to improve the stability of the force calculation. For me this was suprising, I was sure that this would be the issue.

So based on that, I figure that it must be a problem for how I have setup the physics surrounding the force calculation reigon. But I really dont understand it now, and im stuck.

What is also suprising to me is that some of the comsol tutorials for 3D generators exibit the same issues. For example, this tutorial. https://www.comsol.com/model/permanent-magnet-motor-in-3d-47621. This model doesn't seem to repect CoE either.

For me, its important that my model passes the CoE check. It's normally the first thing I am checking for to ensure numerical accuracy/physical reasoning of the model. I have managed to have no issues whatsoever when simulating rotating machines in 2D frame. But so far, every machine ive simulated in 3D frame exhibits this issue.

I am wondering if anyone else has this issue and have some advice for how I can improve the situation? Or maybe you can catch a bigger mistake that I have done in the model which could explain it?

Thanks, BR, Aaron