Definitive Tests on the Heating and Cooling of Hub Motors

[justin_le]

Hey everyone and just getting back from a fairly long absence on this thread but am very encouraged by the ongoing discussions, innovations, and experiments that continue to be posted here.

For those who don't know we had to move our shop to a new location last year, and we've put a lot of effort into setting up the wind tunnel into a permanent fixture in it's own R&D room and writing all kinds of automation tools both for running experiments and doing the analysis of motor thermal performance. And we're hopefully about to see all of that work pay off.


In the previous wind tunnel tests from 2015, what I'd do is run the motor at with a constant field weakening current, rpm, and wind speed and wait like 1-2 hours for the motor core temperature to exponentially approach steady state. Then we'd record the temperatures motor core and shell and power in the hub, change the rpm and wind speed, wait another 1-2 hours for temperatures to become steady state again and repeat. It would take an entire day to get just a half dozen data points like this.
https://endless-sphere.com/forums/viewtopic.php?p=1082371#p1082371

With this new setup, instead of running more or less constant watts into the motor and waiting for temperatures to stabilize into a steady state value, what we're doing is setting the target temperature for the motor core and then letting a PID loop adjust the field weakening currents to achieve that temperature. The motor core will quickly get to temperature, then the power scales back to whatever value is needed to sustain that temp. We then set the conditions for the motor side cover temp to not change more than X degrees after Y minutes as a steady state condition, and once this is achieved the key data (motor power and temperatures) is saved and the wind tunnel automatically updates the motor RPM and wind speed to the next line in the list.



The result is that an entire experiment can take place in just a fraction of the time that used to be needed. In the above test arrangement on a Crysatlyte H3540 hub motor we had seven steady state datapoints to achieve spanning from 70 to 500 rpm and 7 to 50 kph wind speeds. The entire experiment completed on its own in just a tad over 2 hours:



Now not only can we characterize motors for their thermal model much faster, we can also do tests with many more data points over the RPM and wind speed range to really understand and model the nuances of heat dissipation.

 

[justin_le]

Now not only can we characterize motors for their thermal model much faster, we can also do tests with many more data points over the RPM and wind speed range to really understand and model the nuances of heat dissipation.

On that front, one of the things we noticed in the above test on the H3540 motor with 8mL of statorade is that the core to shell conductivity started off pretty high (above 8 watts/degree), increased a little with RPM at first, but then decreased to a below 6 watt/degree at 400 rpm before increasing again at the 500 rpm point.


That seemed peculiar, so we decided to repeat the test with a lot more data points for a finer resolution. Going from 100 to 566 rpm in 14 steps required just a little over 3 hours for the experiment to complete and the results are fascinating. You can see the graph below, blue data is with the 8mL of Statorade.


What we see is that up to about 200 rpm the Statorade had a huge effect on the core to shell conductivity, increasing it from ~3.5 to over 8 watts/degree. However, above 200 rpm the conductivity definitely starts to drop steadily with increasing RPM. At 360 RPM it looks like it hits a minimum value and then most abruptly starts to rise again. To me this data looked like a very clear sign that the above 200 rpm the centripetal forces start to exceed the magnetic forces on the ferrofluid which gets increasingly flung out of the air gap or squished flat so that it no longer bridges to the stator core as often for wiping the heat off. We'd hypothesized about that happening in this thread before but never seen it so clearly on display or quantified the rpm's at which that starts to take place.

Going from 200 to 360 rpm the Statorade is having less and less of an effect, but meanwhile the normal air turbulence inside the motor shell is playing a larger roll in the heat transfer at these speeds. Once we get above 360 rpm, then the air convection transfer dominates and takes over, with the natural and expected increase in conductivity with RPM from that point onward.

This theory seemed to make a lot of sense, and so I repeated the exact same test on the motor with the Statorade removed to see if the plots above 360 rpm would start to line up to the motor with no fill at all. You can see from the red data points that they don't quite fit this way; the increase in conductivity with RPM above 360 watts is still a fair bit higher in the motor with 8mL of Statorade than the one without, and some more digging will be needed to understand just why this is. But it is close in principle.

 

[macribs]

Great work. I love how you always keep pushing forward and keep raising the bar. Repeatability and consistencies where each motor gets tested to the max in a much shorter time. The higher resolution and more data points was great. Would higher amount of statorade give a different graph? Or would the the drag just become too high so that output would suffer? Has anyone quantified the effect of added drag? Will the drag it self have a greater impact on net power then the possibility of lower temperature and lower resistance by adding higher volume of statorade?

Even at statorades lowest efficiencies in the 200-360 RPM range thermal conductivity is still 25% higher then the very best thermal conductivity for non-statorade motors. Also one need to take into account that even when temp raises in a statorade motor we now would know what motor RPM to use to for quickest heat removal after a hard ride say a long and slow climb or repeated max accelerations.

What I would love to see is tests of the most popular motors in stock form, with statorade and with statorade and hubsink. A hubsink would probably help even in the low spot on the graph because of the added thermal mass, and aluminums heat shedding ability. As the motor core heats up more in the low spot on graph delta T between motor core/shell and heat sink would increase and should result in more heat getting sucked away faster.

Does this means we will see new data on motors you already got your hands on? Will you be able to test motors that you still do not have data for? We tried to start a pool for getting more motors tested a couple of years back, now I am thinking maybe we should give it one more go if you guys got the time to test motors shipped to you.

Oh man I would love to see the thermal test combined with dyno to get real data for saturation, and how higher temperature influence everything from increased resistance to max max torque and power to the rear wheel.

Fantastic work J man!

 

[Cowardlyduck]

Love your work Justin, and great to see you back in this thread!

What I would love to see, is that same test conducted with a lot more FF added. If it's the centrifugal forces flinging it out, would adding 15ml, or even 20ml reduce or even eliminate the disconnect at higher rpms?

Cheers

 

[justin_le]

What I would love to see, is that same test conducted with a lot more FF added. If it's the centrifugal forces flinging it out, would adding 15ml, or even 20ml reduce or even eliminate the disconnect at higher rpms?

You guys are like mind readers, that was more or less exactly where I was going to be going with this. Not just repeating at increased the volumes but also experimenting with some different formulations of FF that have a higher magnetic loading. This would be for people who aren't really concerned about motor drag but just want to maximize the cooling effect. In the current Statorade formulation the aim was really to have a solution that cooled the motor with the leased amount of additional drag for those who use a pedal assist ebike, but clearly a lot of the more avid FF users are running more like dirtbikes where this is inconsequential and better cooling at all rpm's would be preferred over reduced drag if that's an option.

Right now I've got the wind tunnel setup doing a bunch of tests on small skateboard hub motors and also have the test results from using hubsinks to analyze and present here, but will get to this FF quantity fill within the next week.

Has anyone quantified the effect of added drag?

Yes I did that at some length characterizing the drag torque with the mL of statorade added, there are numerous posts on the topic back around this time here:
https://endless-sphere.com/forums/viewtopic.php?p=1106508#p1106508
https://endless-sphere.com/forums/viewtopic.php?p=1106995#p1106995

Does this means we will see new data on motors you already got your hands on? Will you be able to test motors that you still do not have data for?

Yes and yes. Previously I was reluctant to engage in testing and characterizing a whole bunch of various 3rd party hub motors in this manner because of the huge time and energy commitment that that entailed, but now with this much faster and more automated test sequence we'll greatly reduce the barrier to getting new motors tested, modeled, and put up on the Trip Simulator web app. I'm probably going to first repeat all the tests on the Crystalyte, 9C, and MXUS motors that I had done previously to establish a new baseline and then open it up to many more DD motors in the mix (with and without statorade + hubsinks) once the characterization procedure is totally dialed in.

 

[macribs]

When the rest of us are thinking it you are already doing it. Way to go with the hub sink. Man I look forward to read this thread the next coming months. U da man J.

 

[John Bozi]

Awesome Justin.

This now adds a bit of a spanner in the works!

Especially for those playing with wheel sizes for their optimal speed. A small wheel needing higher rpm now starts to lose some value if FF is included.

Obviously small still has the advantage at lower speeds....

Now this brings me back to my video test I posted here ages ago, where I recorded just cooling by not moving the motor vs keeping it going at low speeds. With the FF included this would be completely different.

Imagine now knowing this 360 "ghost" info we have. You are at the top of a mountain road and have a 100 degrees C motor that you want to cool the best possible way on the descent. Do you let fly down at maximum speed or use your brakes and go down slowly with disc brakes to make sure FF is fully engaged!

Keeping the bike rolling at about 150rpm I guess roughly 20 kph looks ideal to cool.

 

[litespeed]

+1 to the great work you have done and ALWAYS thinking out side the box. Because of your pioneering in Ferrofluid you've turned my bicycle into a motorcycle......which is what I wanted in the first place and hub sinks made sure it stays a motorcycle.

Super happy with electric power these days!

Tom

 

[Offroader]

If they are going to come out with stronger FF then I will have to put my FF test on hold. I just got my resized spokes today and started to lace my wheel.

I don't think it will be a good idea to use the old stuff then try and remove the old stuff for the stronger magnetic FF. At least I'll have to wait to see if stronger magnetic FF makes any difference.

 

[macribs]

In my mind that would be like putting off getting a computer because you know in a year or two there will be faster ones available for less money. Do today what is available and enjoy it while it last or work. Then your next choice will be based on what is available in the future while you still get to enjoy what works well today.

Personally I really think you should just go ahead and do your test as of now without waiting, because you can always just drain and wipe clean the motor if you later on should wish to change to another formula of statorade. If the stronger statorade works like I picture it in my head it would mean that they have added more magnetic mass into the statorade so it will cling firmer to the magnets.

As much as we like to think our electric bikes are not in the need of service due to their electric propulsion sadly that is not the case. Bearings wear out, hall sensors die for no apparent reason, rust has formed, water ingressen or wires in need of replacement. Over time there will lots of reasons to why you might need to open any electrical motor, so even if the new mix of statorade should show significant better values your motor will not be voided if you already used statorade in it. Simply drain, wipe clean and refill.

 

[sketchism]

This photo gets me excited haha, been a long time coming!

Really appreciate everything you put into the industry man! Honored to have the sinks in the tunnel!, will be great to see how close they perform in real life to the original thermodynamic modeling!

 

[justin_le]

Awesome Justin.

This now adds a bit of a spanner in the works!
Especially for those playing with wheel sizes for their optimal speed. A small wheel needing higher rpm now starts to lose some value if FF is included.

Yup, this is true, but for sure the gains in reduced heat generation with the smaller wheel needing less torque for the same thrust will more than offset any slight reduction in heat transfer from the stator as a result of the higher RPM causing the FF to loose effectiveness. The smaller wheel diameter would still win the day for staying cool. For instance a 20% reduction in wheel diameter means a 20% reduction in phase current, which means a ~36% reduction in the copper losses and heat generated inside the motor. Even if you loose 10-20% of the thermal conductivity of FF from centripetal forces, you'll still be running cooler overall than with the big wheel.

Now this brings me back to my video test I posted here ages ago, where I recorded just cooling by not moving the motor vs keeping it going at low speeds. With the FF included this would be completely different.

At low speeds the addition of ferrofluid increases the core to shell heat conductivity dramatically, by a factor of ~3. I haven't tested the thermal conductivity from the core to the shell with FF when the motor is completely still though. That would be an easy enough test to do. I think we'd all suspect that with the motor not moving at all the ferrofluid wouldn't have much effect since it's then just a static conductive transfer medium in just a few small spots inside the airgap, rather doing liquid convection transfer wiping over the entire stator surface as happens when the motor is turning. .

Imagine now knowing this 360 "ghost" info we have. You are at the top of a mountain road and have a 100 degrees C motor that you want to cool the best possible way on the descent. Do you let fly down at maximum speed or use your brakes and go down slowly with disc brakes to make sure FF is fully engaged!

Note that the graph I posted there was showing only the motor core to motor shell conductivity, it wasn't showing the total conductivity to ambient which includes heat flow the shell to the surrounding air, which always increases at higher rpms/speeds. If I redo the graphs showing the TOTAL conductivity from the motor core to ambient, then the increasing shell to air conduction more or less offsets the decreasing effect of Statorade. It's pretty much flat at ~4 watt/degree from 200 to 400 rpm, and then above 400 rpm it starts to tick up again.

Keeping the bike rolling at about 150rpm I guess roughly 20 kph looks ideal to cool.

From this data it seems closer to 200 rpm would be most ideal, but that's splitting hairs

 

[justin_le]

This photo gets me excited haha, been a long time coming!

Indeed and I'm glad you didn't give up on me I always tell people that sooner or later I do the things I say I'll do, but it definitely winds up on the later side than the sooner side.

Anyways the development of Hubsinks partly in response to the studies on ferrofluid motor cooling is one of my more favorite outcomes of this thread and I applaud Sketch & crew for making such a well thought out and easy to install product in such a quick timeframe. It seemed somewhat obvious that additional fins to the motor shell would help with shedding the heat to ambient, but just how much? Here's the same graph I posted previously but with the addition of the test with 8mL of statorade combined with Hubsinks.



At low speeds up to about 100rpm the addition of the hubsinks improves the net thermal conductivity from the motor core by almost 25%, while the improvement is more like 15-20% in the 200-400 rpm range. That's quite a substantial boost, although since at higher speed the main bottleneck in the heat flow is still from the core to the shell, the effects of amazing heatflow from the shell to ambient don't help things out quite as much at this end. Here is the raw data table:



If you look not just at the shell to ambient heat conductivity, rather than the total conductivity through the ferrofluid and airgap, then the difference is even more pronounced. Across all speeds the thermal conductivity with Hubsinks is about 50% higher than without. Note that the 'x' axis in all of these graphs is also showing the wind speed, so the 200 rpm point corresponds to 20 kph, 300 rpm = 30 kph etc.



Now I am really curious sketch, how does this measured result stack up against the expectations you had based on computer modeling??

 

[John Bozi]

EDIT: I didn't see the new data before I posted my response.

This new graph is pretty much a show stopper. Go faster boys!!!!!!!!!!!!!!!!!!!

 

[madin88]

Yup, this is true, but for sure the gains in reduced heat generation with the smaller wheel needing less torque for the same thrust will more than offset any slight reduction in heat transfer from the stator as a result of the higher RPM causing the FF to loose effectiveness. The smaller wheel diameter would still win the day for staying cool. For instance a 20% reduction in wheel diameter means a 20% reduction in phase current, which means a ~36% reduction in the copper losses and heat generated inside the motor. Even if you loose 10-20% of the thermal conductivity of FF from centripetal forces, you'll still be running cooler overall than with the big wheel.

Thank you for those numbers and calulations about the wheel size.
It makes so much sense to not use large wheels in terms of heat genertion, but i believe above a given RPM, the iron losses could reduce the benefits to nothing.
During acceleration, or when going up a hill, the smaller wheel will be always better (or 99%), but when riding on a flat raod the larger wheel could be ahead in some cases.

 

[Offroader]

Does anyone know how much statorade should be added to a cromotor? This is a 50mm wide stator? I'm sure this was discussed before but I can't find the info easily. Or should I do the no load current and keep adding?


Does it also matter if you drill the hole to fill it towards the center of the hub motor? I assume it doesn't matter where in the motor it is squirted in from because it finds its way to the magnets?

When sealing the side covers, I read where it was recommended to use thermal paste instead of silicone. Do you think it is a good idea to use thermal paste and the thermal paste will be sufficient to stop the leaks?

Thanks.

 

[justin_le]

Does anyone know how much statorade should be added to a cromotor? This is a 50mm wide stator? I'm sure this was discussed before but I can't find the info easily. Or should I do the no load current and keep adding?

It's been discussed at some length, I haven't done statorade tests on a cromotor directly yet but you can easily extrapolate from this information on MXUS motor with 45mm wide stators:
https://endless-sphere.com/forums/viewtopic.php?p=1106508#p1106508
and the H35XX motor with 35mm stators:
https://endless-sphere.com/forums/viewtopic.php?p=1113292#p1113292

There's no harm in adding too much if you don't care about the no-load drag but 8mL should be sufficient even in a larger hub like that.

Does it also matter if you drill the hole to fill it towards the center of the hub motor? I assume it doesn't matter where in the motor it is squirted in from because it finds its way to the magnets?

Correct, it really doesn't matter at all where you put the hole as the Statraode will find its way to the magnets. You just want to avoid drilling into the windings if you are making the hole with the side plate still attached.

When sealing the side covers, I read where it was recommended to use thermal paste instead of silicone. Do you think it is a good idea to use thermal paste and the thermal paste will be sufficient to stop the leaks?

At one point I was using thermal paste on the side covers thinking that this would help with conducting the heat sideways from the rotor ring to the plates, but even though it initially seemed promising for making a seal we found that the the statorade eventually started to leak out. So no I don't recommend using thermal paste for this. All of our experience with silicone indicates that it seals perfectly with a seal that holds up fine, even if you apply it just on the outside of the seam after the motor has already been assembled.

 

[justin_le]

What I would love to see, is that same test conducted with a lot more FF added. If it's the centrifugal forces flinging it out, would adding 15ml, or even 20ml reduce or even eliminate the disconnect at higher rpms?

That test is still coming, but in the meantime I already had a nine continent 2707 motor setup in the tunnel in order to do some repeating of the statorade QTY vs conductivity graphs with this hub, and at the 6mL statorade fill level I decided to run exactly the same 14 point conductivity vs. RPM sweep as I did on the Cyrsalyte H3540 motor to see if it also showed the same decreasing conductivity from 200 to 366 rpm before curbing up again.

Now both the H3540 motor and Nine Continent are running the same nominal 205mm rotor diameter and basic motor core shape, only the Crystalyte has a 35mm wide statorand magnets vs 27mm for the 9C (end hence the 8mL v 6mL statorade fill quantity). I was therefor expecting to see something very similar, and while the nine continent motor did show a slight dip core to shell conductivity from 300 to 400 rpm, the amount dip was very slight compared to the pronounced drop with the Crysatlyte H motor. Have a look



I'll be looking at the geometry next in terms of the air gap thickness and how much space there is beside the magnets for the Statorade to fling into at the higher RPMs when centripital forces come into play. It could be that the Crysatlyte hub has a larger gap between the side plate and magnets where the Statorade can pool, while the 9C+ motor is fairly tight here so even at the fast speeds most of the fluid still stays in the gap. I'm not quite sure, but it's interesting to see how two nominally quite similar motors can have such a unique conductivity signature with the Ferrofluid, and understanding why could help better optimize motor design to take advantage of it.

 

[macribs]

That is very interesting. Could a possible solution be to build up a seem/bead of silicon to help prevent the gap from being large enough to allow for statorade to be flung out from the gap between magnets and rotor?

 

[Punx0r]

It's interesting to see the thermal conductivity plot flat-line or dip and then increase again with increasing motor speed.

Is it possible that the ferrofluid is being drawn back into the air gap at higher speeds due to a venturi effect? Or that the increased air turbulence is causing the ferrofluid to foam and re-enter the air gap?

 

[justin_le]

That is very interesting. Could a possible solution be to build up a seem/bead of silicon to help prevent the gap from being large enough to allow for statorade to be flung out from the gap between magnets and rotor?

Yes that's exactly what I'm thinking assuming that the hypothesis is right, and it will be pretty easy to test and confirm once I get the H3540 motor back in the chamber again. In the meantime though, I completed a full full round of tests with the Nine Continent motor while that was set up, going from 0 to 18mL of statorade in 2mL increments at 100, 200, 300, and 400 rpm, and also doing much higher datapoint tests at 0mL, 6mL, and 18mL of fluid. And then at the end I also added Hubsinks to the mix so that we could see the effect there.

So getting to this question here

What I would love to see, is that same test conducted with a lot more FF added. If it's the centrifugal forces flinging it out, would adding 15ml, or even 20ml reduce or even eliminate the disconnect at higher rpms?

Here's what we see with the Nine Continent motor at least, which as you saw from my last posts had less of this centripetal flinging effect than the Crystalyte H35XX hub. With 6mL you see this decrease in conductivity at 300 rpm before it increases again at about 450 rpm. But with a full 18mL Statorade added, there is no such effect. At the lower <300 rpm speeds there is basically no improvement in thermal performance to the much higher statorade fill.


Here is another perspective showing the core to shell conductivity at at 0mL, 2mL, 4mL... to 18mL of Statorade.

It's impressive just how much of an effect even 2mL has at the low 100 rpm point, but then this clearly ceases to do much as the motor spins up faster. Both the 4mL and 6mL fills sortof plateau around 300-400 rpm, while with >8mL the conductivity always increases with motor speed. What is really interesting is that it seems to max out around 10-12mL, and then the thermal conductivity actually decreases a bit with the 14,16 ,and 18mL fill levels. You can see this really clearly in the same data here but plotted with mL Statorade on the 'X' axis, and each of the 100, 200, 300, and 400 rpm data sets. I've got no idea at all why this would be and can't visualize a scenario where higher amounts of ferrofluid would have the effect of reducing the thermal heat transfer. It's not big enough to worry about overfilling, but it's a curiosity that I"m curious to see if will be replicated with other motor series.



The higher Statorade fill does have an effect on the motor drag as we'd expect and have seen previously. In these tests we set the wind tunnel script to automatically measure the no-load power to spin the hub with no field weakening after each steady state temperature condition, and so the drag in every case was measured when the statorade was at exactly 60 degree celcius (see this postto see why temp is important). The extra drag losses are pretty minimal up to about 4-6mL, but going from 8 to 12mL you really see the no load power increase especially at the higher speed range, an extra 10 watts in fluid friction to overcome at 400 rpm at 60 degrees, which would be more like 20-25 watts extra when the motor is at room temperature.


On the balance of all this I'd say 6mL is still the sweet spot for Statorade fill on a nine continent motor for most people, going to 10mL if you don't mind some extra motor drag. But injecting much more than 10mL in this motor not only doesn't improve anything, it actually makes things a bit worse, for reasons that aren't obvious to me. You'll not only have the extra drag, but you'll have a less cooling effect than if you'd stayed in the ~10mL realm.

 

[Cowardlyduck]

Awesome stuff Justin, and thanks for so directly and thoroughly testing and answering that question about volumes of Ferro Fluid.

Seems you might want to offer a 20ml syringe of the stuff for those of us with larger hub motors. Makes me wish I bough the 50ml bottle when I last ordered from the Grin store.

Another totally unrelated test that might be worth doing if you can is a longevity test, to test how long we should expect the FF to last in our motors before needing a top up. Would it be possible to setup a motor that runs through heating/cooling cycles and leave it running in the wind tunnel for a number of weeks or months to simulate 10's of thousands of km's of usage?

I mainly ask because I found my FF disappeared in my vented hubs a lot faster than I expected. Not that using it in vented hubs is a normal use case by any means, but it would correlate in some form to the longevity in a sealed hub.
I imagine it would also be useful information for the vendors who are now selling motors (including Grin) pre-filled with statorade, and those of us DIYers who make and sell the occasional hot-rod E-bike with Ferro Fluid added.

Cheers

 

[macribs]

1+ J man! Keep up the great work. This is taking things to a whole new level. Maybe we now finally can say cooked hubs are so yesteryear.

 

[Raisedeyebrows]

My own unscientific test-9C clone and 52V battery, I pumped about half the syringe of Statoraid into it, installed the Grin heat sinks. Ran it up about a mile of real steep hills wide open on a 102 degree day for a test. Stopped once I reached the top and felt the hub, it was barely warm, no where near warm enough to want to take my hand away quickly.

Stuff must work, I'm impressed and thanks to the contributors here who told me to use it, totally worth the investment for increased cooling and piece of mind.

 

[justin_le]

It's interesting to see the thermal conductivity plot flat-line or dip and then increase again with increasing motor speed.
Is it possible that the ferrofluid is being drawn back into the air gap at higher speeds due to a venturi effect? Or that the increased air turbulence is causing the ferrofluid to foam and re-enter the air gap?

I'm glad you picked up on this oddity too because it also struck me as unexpected with no obvious explanation. Have a look at the more recent test data with the 9C motor and just 2mL of fluid. At 100 rpm it really does manage to conductively bridge the air gap pretty well, almost doubling the heat flow from 2.3 to 4.4 watts/degree. Not surprisingly at at the faster speeds of 200 and 300 rpm the fluid is squished flatter against the magnets and has decreasing effect, with just a 0.5 w/degree boots at 300 rpm. It looks like this trend should continue and that at 400 rpm the 2mL of Statorade would barely touch the stator at all and the conductivity should match that of no fluid fill, but this isn't at all what happens. It then picks up again with a 1 watt/degree boost at the 400 rpm point.



The idea of it frothing/foaming a bit with higher rpms and thus occupying a higher effective volume again is an interesting theory. I think we'll continue to see this effect across the different motor series and I'll dive in with more thorough data collection to see if that can shed light.

Another totally unrelated test that might be worth doing if you can is a longevity test, to test how long we should expect the FF to last in our motors before needing a top up. Would it be possible to setup a motor that runs through heating/cooling cycles and leave it running in the wind tunnel for a number of weeks or months to simulate 10's of thousands of km's of usage?

I can't tie up the wind tunnel for months or years doing this, but I am planning at some point to do this kind of experiment just running on the roof of our building so that it's also exposed to varying outdoors conditions too, and then after so many hundreds of hours take it back down to the lab and see if there's been any change. I'll probably run 2 or 3 motors otherwise identical but with different max temperatures, say 80oC, 100oC, and 120oC, and then we'd be able to learn where the thermal rollback should be set from a Statorade longevity perspective.

This kind of test takes a lot of time to set up and patience to run, so don't expect it any time soon but maybe by fall. We discussed it a couple years ago too. With vented holes in the side cover it definitely won't last for too long. The performance results for this (vents + statorade) is really good ( see here for those who missed that https://endless-sphere.com/forums/viewtopic.php?p=1130620#p1130620) but that would be more for like single day race events, rather than a long term strategy.