So here we are again, running cogging torque simulations.
This time I automated the thing based on the percentage of stator circumference occupied by the stator coil cores. I gave the cores a “hammerhead” shape to make the design more realistic, and then specified an occupation from 50% to 90%.
Here are 2 pictures, showing the difference between a 50% occ. stator and a 90% occ. stator, respectively:
On the upper image, you can see the hammerheads are narrower than the stator core; on the lower image, you can see the hammerheads are wider than the stator core.
I ran the batch for a good day or so, and accumulated the data. The script made the motor rotate 90 degrees in 1 degree steps, calculating the torque at each step. The output data is very noisy (the thing is yucky, click to see in full size): raw_cogging_torque_vs_occupation.png .
So I made use of a little statistical power to get some insight. First I turned all values into absolute values (removed the minus signs), and then applied the Max(), Median(), and Average() functions. I’m no maths or stats wizard, so I just threw everything at it and drew a graph.
Now that’s more like it! 🙂
Once again we can see that the worst case maximum cogging torque will always be very small (3 Nm!), but I can’t be sure of the remaining interpretation of the data. There seems to be a “nice spot” at 70% occupation, but I’m not so sure… the data noise could come from the precision of calculation (FEMM mesh size vs. rotation step)… maybe if I pump up the volume and specify a super-fine mesh I get a close to zero torque?…
Anyway, I’m satisfied for the time being, regarding the cogging torque of this model. I think it is good enough to be tested again only in the prototype phase. Of course, I could run more cogging simulations ad nauseum, but I’ve got bigger fish to fry right now.