Comment by pinkmuffinere
16 days ago
> The motor’s performance on the dyno has exceeded even our most optimistic simulations
Not to take away from the exciting achievement, but I always found comments like this kindof unusual. Really, it exceeded even your _most optimistic_ simulations? If the high end of your simulated performance was below what you actually measured, I am worried that your simulation is seriously neglecting something. I used to work in decent depth with three phase bldc motors, so I feel I can say with some authority that these things _can_ be simulated, and while real world data is hard to exactly predict, getting something outside of the predicted range would generally be interpreted as a sign that your simulation isn't so good. But maybe this is just marketing-speak, and their simulations are actually totally fine.
Corpo speek for the CTO ignored our suggested safety margins and ran it at higher power then we had simulated, but it didn't melt.
+1
I suspect that the marketing guy's eyes glazed over when the engineers tried to explain confidence intervals to him. He demanded a simple figure. They gave him the median projected value to make him go away. The tested value was above median projection and thus you get this wordspew.
> Make him go away
Spoken like a true engineer.
> and while real world data is hard to exactly predict, getting something outside of the predicted range would generally be interpreted as a sign that your simulation isn't so good
Either that, or your measurements are inaccurate.
or, yasa is moving very quickly,and as they are breaking new ground in power density, the numbers are ridonkulous,completly and utterly insane, 30lb's peak over 1000hp,just having the thing not vaporise itself is an achivement, and so there is no actual way to simulate that, outside of rocket motors, which is kind of where they are at. bet the tire guys are shaking there heads
The point is that at this level, you shouldn't just let some engineers go wild on a physical prototype without simulating it first. You sort of have to simulate it before building the real thing or you're likely to waste a ton of time and money.
If they did simulate this design and it exceeded their expectations so much, either something is wrong with the simulation or some engineer worked some voodoo magic into the very late prototype stages.
Extraordinary claims and all that. Of course you can simulate it. If you are this surprised by the outcome of an experimental setup the first thing you should do is distrust the data. Figure out where the discrepancy is. Fix it, re-run the simulation and the experiment until the two are in agreement. Then publish.
The normal scientific reflex is not to hit the marketing guy and say 'we've got a winner, go write it up'.
Depends if you make overly conservative assumptions in your modeling...
If you have simulations with varying levels of optimism, but all of them were too conservative, then you screwed up.
Meh, depends what's the goal. Exceeding predicted performance is not a screw up, it's just providing a minimum guaranteed performance aka playing it safe (under promise, over deliver).
Also, we don't know by how much the most optimistic predicions were exceeded.
Makes for nice marketing ;)
> I used to work in decent depth with three phase bldc motors, so I feel I can say with some authority that these things _can_ be simulated
If I take it literally versus hyperbole and excitement, could there be uncertainty coming from the drive train design choices, system integration details?
Sure, but that sort of uncertainty should only increase the range of values you see out of the simulation. You could say “efficiency may range from 60% to 80%, so the output torque will range from X to Y”. If the real result is bigger than Y, something has gone wrong in your modeling or assumptions
> something has gone wrong in your modeling or assumptions
Maybe they used a "all our previous tests showed 95% of efficiency compared to simulation, so let's multiply all our results, including top efficiency by this". Then their newest motor had 97% efficiency comparing to previous model.
I suspect they correctly simulated everything. Then on the test stand they turned the dial to 11 and saw a number higher than the simulations. Probably not reliable at that power level, but fine for a record announcement.
Is constant power a lot easier to predict than short term peaks?