Definitive Tests on the Heating and Cooling of Hub Motors

[Kodin]

I'd assume just like any abrasive suspended in fluid it will have some effect, but if anything, it will simply keep the surfaces it touches polished unless you're talking thousands upon thousands of hours running time. ...Assuming the particles really are as tiny as I think they are.... (1-10 micron maybe?)

 

[flathill]

The main benefit of potting the stator is it reduces hot spots on the windings, which means the insulation is much less like to fail.

The downside to FF is now the maximum temp of the motor is limited by the FF

Normally china hub motor magnets are rated to 150c but u can spec much higher but it gets $$$

220c mag wire is readily available

Typically I have found ferrofluid turns to thick sludge after being run for a week continuous at 100c and the fluid was rated much higher (200c transient, 125c "extended")

Note the max temp of of FF is only spec'd for short duration. What you need is the gel time data.

Real world this may or not be an issue depending on your duty cycle and which grade of FF you have access to

 

[justin_le]

Can someone explain what exactly potting a hub motor does?

Offers the potential to improve cooling performance (and hence improve both burst power and continuous power handling), as well as weatherproofing reliability.
Potting also doesn't leak or weep or drip unlike many fluid based solutions (with FF being an exception due to clinging into the gaps between magnets.

And the hope is that the combination of potting and a FF can provide one of the most direct paths possible for heat removal from the copper windings in a stator to the motor shell, where it is quickly shed to ambient via the passing airflow. The ferrofluid by itself very effectively moves heat from the iron stator out to the casing, but it doesn't come into any direct contact with the copper wire which is where most of the heat is being generated in the first place.

The copper wires are wound around the stator and you'd think would be in pretty good thermal contact, but each of the winding slots is lined with a thick special insulating paper to prevent the copper from shorting to the core, and there are similarly fiberglass plates at the two ends of the lamination stack. These are intended for electrical isolation but surely cause some degree of thermal isolation too. By potting the stator with a thermally conductive resin we'd expect that the heat generated from the copper will be better coupled to the steel core.

We even removed the fiberglass strips that are placed over tops of the windings to hold the folded paper down so that there was a much resin fill inside the slots as possible


But as I said, I haven't yet done a test where I have separately measured both the copper winding temperatures compared to the stator lamination temps, as the core temperature probe in all the previous experiments was glued to the stator right at the copper/core junction so would show a hybrid value between the two. If it turns out that the DeltaT between the copper wire and the steel isn't very high, like 5oC or so, then there would be little point in potting or additional attempts at thermally cooling the copper directly since there aren't many gains to be had by doing this. Just pull the heat out of the stator with FF and you're good.

What brand and type potting liquid did you use ? and what temperatures and how long did you bake it for ?

This first pour was a UL listed thermal epoxy from epoxies etc.
http://www.epoxies.com/_resources/common/userfiles/file/50-3151NCFR.pdf
Mixed viscosity of 5000cps, and it was put in a chamber at 55oC for about 6hrs to accelerate the cure, just because we were impatient, it's not really necessary.

 

[justin_le]

The downside to FF is now the maximum temp of the motor is limited by the FF
Normally china hub motor magnets are rated to 150c but u can spec much higher but it gets $$$
220c mag wire is readily available

This is all true, but even if your motor components can be pushed up to 150 or 200c, there is a huge penalty in efficiency that you pay for doing so as a result of increased copper winding resistance. At 200c on the windings you have 70% more resistance than at room temp, which means 70% more wasted watts of heat for every unit of torque output, and a corresponding rise in the inefficiency of the motor (eg. if a given motor is at a 75% efficiency point at room temp, it will be only ~63% efficient at 200 degrees). So even if nothing fails at these elevated temps, you generally want to avoid ever going there for best motor performance, and active cooling methods can go a long way towards that.

Typically I have found ferrofluid turns to thick sludge after being run for a week continuous at 100c and the fluid was rated much higher (200c transient, 125c "extended")

This is quite interesting, and I'll be sure to run some extended tests of the fluid on hot plates or in the oven to get some more empirical data. Was this in a motor application or something else related to voicecoils or other? We were advised by the chemists that the temperatures I gave as examples in the motors were considered "not high temperature" as far as the ferrofluids were concerned when I raised this.

 

[Bluefang]

Really keen to see how you go with the potted motor!! Freaking awesome that you are showing the basics of how you are going about all the tests, thanks also for the info on what potting compound you used.

 

[justin_le]

Interesting points from a manufacturer of ferrofluid:

The thermal stability of a ferrofluid is related to particle density. The particles appear to behave like a catalyst and produce free radicals, which lead to cross linking of molecular chains and eventual congealing of the fluid. Catalytic activity is higher at elevated temperatures and, therefore, ferrofluids congeal more rapidly at these temperatures.

Seems to be definitely inline with Flathill's firsthand experience.

High magnetization ferrofluids are of interest as they produce volumetric efficiencies of magnetic circuit designs leading to lightweight and lower cost products. They can also be used to reduce reluctance of magnetic circuits and fringing field thus increasing useful flux density in the air gap. The domain magnetization of magnetite ultimately limits the maximum magnetization value that can be realized in a ferrofluid.

This is why I think it's a bit wishful to hope that FF in the air gap are going to help improve the motor kV or have other magical effects on the motor parameters. In the realm of the strong ~1 Telsa magnetic fields present in a hub motor air gap, very few materials can offer any increase in the magnetization. Iron is pretty much the only magnetic material that isn't totally saturated. Nickel and Cobalt both saturate at ~0.5T, while magnetite from what I can tell is like 0.2T. If you have an old crappy motor with ferrite magnets and a large air gap then there's a good chance that the FF will reduce the reluctance and increase the flux through the air gap, but I don't think there would be any such effect with the fields present in our hub motors with rare earth magnets and such.

Thermal conductivity of a ferrofluid depends linearly on the solid loading. Fluorocarbon based ferrofluids have the lowest thermal conductivity of all commercial ferrofluids, therefore they are the least desirable materials for heat transfer applications.

Interesting, wouldn't have guessed that the thermal conductivity would be so directly related to the solids fill of magnetic particles. But I would think that in the motor application with so much active agitation and motion that you'll have great convective heat transfer regardless of the actual thermal conductivity of the fluid in a static sense. So if fluorocarbon based fluids have the best stability or other property, I'd say they'd be the better bet even if the static thermal conductivity is listed as poor.

Anyways interesting and relevant tidbits from the source Fechter, so thanks for sharing.

 

[speedmd]

Dow has a big line of synthetic heat transfer oils and claim to greatly reduce the breakdowns of the standard type oils. Wonder if any are used in FF apps.

Syltherm silicone based. http://www.dow.com/heattrans/products/synthetic/syltherm.htm
http://www.dow.com/heattrans/products/synthetic/dowtherm.htm

 

[flathill]

Typically I have found ferrofluid turns to thick sludge after being run for a week continuous at 100c and the fluid was rated much higher (200c transient, 125c "extended")

This is quite interesting, and I'll be sure to run some extended tests of the fluid on hot plates or in the oven to get some more empirical data. Was this in a motor application or something else related to voicecoils or other? We were advised by the chemists that the temperatures I gave as examples in the motors were considered "not high temperature" as far as the ferrofluids were concerned when I raised this.

The rated gel time of the fluid (by the mfg) was ~160 hours at 170C but I don't know exactly how the test was done and I don't know precisely what "gel" is defined as, but it is useful for comparison purposes.

My own test was done with movement and heat in the application, not just heat as in the gel test, which may change evaporation rate.

 

[speedmd]

Most gels form by mild cross linking. Evaporation is certainly a big issue in oils degradation. Agree, motion/mixing may be a big factor also. One of the high temp dow Silicones would be interesting oils to try if it could be mixed with a compatible magnetic base material. They are claiming ten year life at high temps. Considering the results Justin has shown so far, it may be worth some further research and development of something custom for hub motors.

Looks like someone has worked on this.

http://www.sciencedirect.com/science/article/pii/S0169433211011706

http://www.google.com/patents/US4356098

 

[madin88]

THANKS for the tests Justin.

I believe the potted core will lead to much lower temperature of winding so better core-to shell conductivity. What a fantastic idea!

The thing about congealing of ferrofluid makes me a bit sceptical about using it. also almost double of the drag when the motor is cold i do not like.
I'm very looking forward to the test with oil fill. would be nice if you also make a milliliter vs. thermal resistance chart like you did with the ferro fluid.

 

[John in CR]

I've got a pair of WE36 motors destined for a beach and rain 2wd ebike, and a pair of older larger hubbies destined for a 2wd emoto, so I'll offer them up as sacrificial lambs for one with and one without comparisons running identical controller settings are higher power and higher rpm use than Justin's control tests.

My high efficiency hubmotors are more difficult to open, so I prefer to only do it once. I'm having great success air cooling, so the risk of FF sludging up leaves me wary of trying it for now, especially since I often run them well above 1000rpm... even 2krpm as I move to mid-drive use for offroad.

Except maybe to protect hall legs, I don't see any upside to potting the stator. At least with my motors the stock thermal connection of the copper to stator steel isn't lacking. Otherwise I'd see the temp rise significantly if I stop immediately after pressing the motor hard at high current. I install my temp sensors 1cm away from the copper to avoid the "noise" in temp readings of the short spikes. Instead I want to know the temp of the motor as a whole, and measuring at the copper doesn't tell you that. Sure there's a delta T between the copper and steel, but I doubt epoxy has a high enough thermal conductivity to matter much. The bigger strike against potting is the decrease in surface area for air cooling.

It's great to see those reduced temperatures with the FF, but that's just passing 160W (actually less since the heat still has other pathways). That's well and good for modest power, but how will it scale in terms of power? ie How is 1kw or more going to so easily pass through a few ml of FF? For audio, FF is used to dissipate heat in tweeters and midrange drivers, but I don't think it's used with big subwoofer drivers with greater amounts of heat to dissipate. Are the same limitations likely to apply with our motors, and the guys wanting high power are mistaken to believe FF is the cure-all?

John

 

[macribs]

Very good question John. How will FF scale up? We don't know yet. What we do know is that people who have used oil cooling (ATF, motor oil and even coconut oil) are seeing great heat reduction even when overvolting. But they have the oil cooling downside. Leakage. Oil dripping everywhere.

FF might not be a "fix all solution" for high powered hubs, but so far I think it is the cleanest solution to use as a base. Fill them hubs with FF and start pumping loads of current into hub motor. If motor gets too hot too quick in real world testing, add more FF, or add more thermal masses.

I am sure Justin will crank up the current later on in the testing so we might even get to see real test data from a motor pushing way beyond the manufacturers max peak power. FF might not be able to serve all needs, but it sure looks like it will be the best and cleanest way to remove heat from hubs for most people. I am sure by next summer we will see several high powered builds using Ferro Fluid to aid in the heat removal department. Then we will know more about any limitations for FF cooling.

If people are happy with their way of cooling their hub as of today I say stick with it until we know more about long term effect. But if you worry about the internal of your hub motors because of the vented side covers it might be time to look into Ferro Fluid. If not for anything else just to test it out to see how Ferro Fluid performs with your line of high powered monster hubs.

 

[John in CR]

What we do know is that people who have used oil cooling (ATF, motor oil and even coconut oil) are seeing great heat reduction even when overvolting. But they have the oil cooling downside.

Really? I don't know of anyone running high power with oil in their hubmotors. Those who have run moderate power ended up seeing temps start rising, just somewhat later than before. I expect the same here using FF with the only way to get rid of more heat fast enough to keep things cool are:
1. Direct better airflow at the motor shell for a higher coefficient of convective heat transfer.
2. Increase the exterior surface area of the shell to transfer more heat at the same temperature differential between the shell and ambient.

Otherwise the shell must have a higher temperature at it's surface to dissipate more heat. It's not the bottleneck at low power, but the shell becomes the bottleneck at higher power. Otherwise we'd never see significant temps at the shell. Hopefully the FF proves stable in our use, so changes at the shell can really pay off without the mess and added losses of even a partial oil fill.

Looking at heat in electrical terms..."resistance" can be useful when looking at conductive heat transfer, but once it gets to our motor shells it's heat dissipation via forced convection where electrical terms aren't readily applicable.

 

[fechter]

I'm not sure what kind of oil they use to make ferrofluid, but it seems like it should be possible to do with silicone oil. Silicone oil seems to have the best properties for heat transfer fluid.
If the stuff gels up over time, it would require taking the motor apart to clean it off. Not terribly hard on some motors and a real pain on others (like by A2B).

 

[justin_le]

It's not the bottleneck at low power, but the shell becomes the bottleneck at higher power. Otherwise we'd never see significant temps at the shell.

Well it depends what you mean by bottleneck. But as you can see in this earlier result I posted here:

The thermal conductivity of an unmodified MXUS motor shell to ambient is about 20 Watts/degree at 60ph. So with 1000 watts of waste heat being dumped into the motor core the shell temperature would stabilize at around 50oC above ambient, which is pretty tolerable. I can't imagine that the copper->air heat thermal conductivity in an air cooled motor gets this high unless you had some very aggressive forced airflow over the windings.

But yes if 20 W/deg is considered a "bottleneck", then any kind of additional fins on the motor rotor or side plates ought to improve that a lot, since the heat dissipation should be roughly proportional to the area exposed to the airflow.

All of this does assume that the hub motor is exposed to air at near vehicle velocity. If there is a a shroud or fairing around the vehicle that traps a pocket of relatively still around the wheel it would be prone to getting much hotter. And even the presence of fat tires and rims could create a noticeable draft for the motor body. This latter bit I'll be able to quantify in the tunnel.

 

[justin_le]

also almost double of the drag when the motor is cold i do not like.

Perhaps you missed the bit that at the minimum fill level required to achieve full conductivity there is no measurable increase in the motor drag? You only have extra resistance when you put in so much as to completely fill the gap with FF, but there is no point in doing since it doesn't improve the heat flow. That was more or less the crux of these results, have a closer careful read of these two posts:
https://endless-sphere.com/forums/viewtopic.php?p=1095879#p1095879
https://endless-sphere.com/forums/viewtopic.php?p=1106508#p1106508

I'm very looking forward to the test with oil fill. would be nice if you also make a milliliter vs. thermal resistance chart like you did with the ferro fluid.

I already finished and posted the oil fill result here:
https://endless-sphere.com/forums/viewtopic.php?p=1106995#p1106995

I agree that a similar plot of thermal resistance vs. oil fill would be really interesting, since it doesn't seem like anyone's made much of a systematic effort at finding out how much is truly required. But unlike FF, oils like ATF are cheap cheap cheap, so there's less of an incentive to try and find the minimum amount needed. I suppose there's a chance that if it turns out to be a smallish amount like 20-30mL, then the problems of leakage could be less pronounced with less fluid overall splashing around. Worthy of some tests for sure.

 

[Kodin]

Biggest issue I see is with oil fills the oil isn't propelled around by anything other than centripedal force. That's where ferrofluid shines, as the particles cling to the magnets in a somewhat uniform manner, and keep the oil only where it needs to be. I'd imagine with purely ATF you'd get a heat gradient on the stator where the bottom is coolest and top is hottest not because heat rises, but because the bulk of the ATF returns to the bottom, where as not so much will reach the top.

 

[bowlofsalad]

It's not the bottleneck at low power, but the shell becomes the bottleneck at higher power. Otherwise we'd never see significant temps at the shell.

The thermal conductivity of an unmodified MXUS motor shell to ambient is about 20 Watts/degree at 60ph. So with 1000 watts of waste heat being dumped into the motor core the shell temperature would stabilize at around 50oC above ambient, which is pretty tolerable. I can't imagine that the copper->air heat thermal conductivity in an air cooled motor gets this high unless you had some very aggressive forced airflow over the windings.

But yes if 20 W/deg is considered a "bottleneck", then any kind of additional fins on the motor rotor or side plates ought to improve that a lot, since the heat dissipation should be roughly proportional to the area exposed to the airflow

This is a question I've been asking myself for some a while concerning the oil versus vented notion. Often the conclusion I come to, but not always, is that a vented hub motor is likely to come out cheaper than alternatives relating to a fairly to extremely well sealed hub motor with expended surface area.

All of this does assume that the hub motor is exposed to air at near vehicle velocity. If there is a a shroud or fairing around the vehicle that traps a pocket of relatively still around the wheel it would be prone to getting much hotter. And even the presence of fat tires and rims could create a noticeable draft for the motor body. This latter bit I'll be able to quantify in the tunnel.

True in any case, but I imagine that due to the suction created naturally in appropriately drilled hub motor covers that the vented motor might win (your tests suggests as such at higher speeds, but I am not sure the vents were ideal) in this comparison. However, I think a finned hub motor has a pretty incredible potential combined with ferro fluid (FF) for shedding heat. The way I look at this is how we might look at the velocity of the tip of a wind turbine. The tips of fins on the exterior of a hub motor will likely be able do a lot of it's cooling even at 0kph or very slow speeds http://www.windpowerengineering.com/design/mechanical/blades/calculate-blade-tip-speed/. What I am trying to say is that the fins one the surface of the hub motor would not only be very capable of breaking surface tension (likely wrong phrase) through high levels of turbulence and pressure but the they also will be moving at 100s of KPH at slow speeds and 1000s of KPH at higher speeds. Pair all this with air scoops on top of fins that are between the hub flanges and it sounds like a brutal amount of thermal capacity can be had. But if were going to compare apples to apples, there is plenty of room to expand to performance of a vented hub motor, I am sure. So if you are considering the value of a finned hub motor in an argument such as this, it is likely at least worth while to consider the value of a vented hub motor with expanded air flow.

Of course, I am sure you are in a radically better position than I am to form an opinion on this, but looking at it in a power density per $, bang for your buck sort of view is likely an extremely important idea to consider. Personally, what is boils down to for myself and customers is a hub motor that is of great value while achieving excellent reliability. I feel fairly confident in the idea of the reliability of a vented hub motor, but we still have a large number of very interesting unanswered questions for FF.

Another angle to look at is asynchronous motors and which area would be more appealing to field experience in with the potential advent of that idea. No magnets in an asynchronous motor means FF is useless. I love the idea of exploring FF though, but gaining a deeper understanding and potentially better development of a vented motor might be of more value in the future. I wouldn't be surprised if you and others have already strongly considered a lot of these ideas, but it's fun to talk about them.

 

[markz]

Can anyone tell me why on ebikes.ca simulator, the vented mxus motor does not change at all in comparison to same unvented motor?

 

[justin_le]

Can anyone tell me why on ebikes.ca simulator, the vented mxus motor does not change at all in comparison to same unvented motor?

Oops, the drilled side cover one isn't supposed to show up on the motor simulator. I'll take that off right now, sorry.

But while it was there, the reason it looked the same is because absolutely none of the instantaneous or static parameters change by adding cooling modifications. The bench tested "performance" of a hub motor will by identical whether you drill holes in it or not, it's only after you run it for a long time and allow the core to get hot enough that the winding resistance has changed by an appreciable amount that there would be any change in the shape of the dyno curve of cooled vs uncooled.

Even then, it would be fairly slight. You can use the simulator to illustrate this by adjusting the controller resistance, which has the same effect of increased motor phase winding resistance. For instance, the MXUS 4505 hub has room temperature resistance of 0.156 ohms. At 75oC, this would be about 0.187 ohms, and at 100oC it's about 0.203 ohms. So if you want to compare the motor at 100oC vs 75oC, add 0.047 ohms to the controller resistance in the simulator for former case and 0.031 ohms for the latter case, and see what you get.

A tool to properly illustrate the effects of different cooling options needs to be a time dependent simulation the development of which I touched upon here:
https://endless-sphere.com/forums/viewtopic.php?p=1075569#p1075569
and more interestingly if you want to see the test/dev version of the tool
https://endless-sphere.com/forums/viewtopic.php?p=1095582#p1095582

 

[John in CR]

Bowlofsalad,
While just making holes in a motor helps as Justin has already measured, it requires meaningful flow to move a lot of heat. Quantifying that flow gets disappointing in a hurry when you calculate how much heat air can carry with it. You really have to get beyond "holes in a pizza pan" if you want to take a hubbie to what it's really capable of.

Justin,
I'm talking about more than 1kw, but let's look at the just 5X heat increase. There's some wind blockage as you mentioned, but look at the delta T between the exterior surface and the core. With 100° covers that's putting the core at what 120°C, and that's assuming the 1kw heat moves just as well thru the FF and the same delta T is maintained. 120° is way too hot for my taste, and I run my hubbie at 27kw peak input with a stator temp that has never exceeded 105°C. The 120° also doesn't consider ambient temps so near the road surface, and it also doesn't consider the increase in heat from increased resistance.

I realize you don't even consider the kind of power I'm talking about, and I agree that even at a few kw of input power on the street the heat transfer from the shell isn't going to be a big limitation unless you get bogged down on a hill. Those needing the most help are the offroad guys, who are keeping it slow under heavy loads of hills and frequent accelerations, and they don't have to run extreme power for serious heat problems. They need to increase the heat transfer to the environment by more than just an increase in the motor shell temp, so increasing only the interior heat transfer of the motor simply isn't enough. They already get the motor shells hot, and it isn't moving heat fast enough.

Some of the scooter hubbie manufacturers are starting to "get it" and make changes, but smooth side covers (even at typical ebike power levels) is downright horrible from a thermodynamics perspective, and to me it's inexcusable since the motors are so thermally limited. Better heat management means lighter more efficient lower cost hubbies are within easy reach.

I'm not knocking your testing at all. Your efforts are awesome, and I understand that high power is outside your area of interest. I simply disagree with making motor changes related to heat transfer without making any changes to the exterior. It's not nearly as easy as squirting in a few ml of FF, but useful changes are certainly much easier than many of the mods some people do.

 

[madin88]

about drag with FF:
i was talking about the graph below, but now i see it was with 16ml



at 4-5ml the curve would be almost flat, right?

thank you for the info about the oil test Justin. i missed that sorry

About tests with high power or 1000W of heat:
shouldn't thermal resistance not stay the same between about 200W used for the tests and 1000W?

 

[macribs]

@John

I am with you on the outer shell parts. I would love to see manufacturers making an effort to allow for more thermal masses on the outside parts.
I keep thinking of my old Suzuki GSX 1100. All them fins really made heat removal easy at speed. Those motors was not really advanced like todays liquid cooled motors. But due to the large outside surface you could ride forever WOT without overheating, accelerate over and over again without any over heat problems.

If covers, or the part between the flanges had an even larger surface area more heat could be removed faster. As you said such a solution would even cater for the off road e-bikers, and it would be interesting to see how a modified large surfaced motor shell would perform in conjunction with Ferro Fluid. But for street riders it seems Ferro Fluid would be all that is needed for hub to stay cool.


@Justin Thx for all the testing you are doing. It is amazing to see real data on cooling, before all cooling threads have been based on peoples experiences or even thesis rather then on cold hard test data, I fell what your are doing here with all the testing is a quantum leap forward for more stable motors, longevity and possible also longevity for high powered motors. Kudos for your hard work Justin. Much obliged.

 

[gensem]

A hubmotor with integrated aluminium rim should transfer alot more heat with FF.

 

[flathill]

Good point about a mag rim heat sink. If only we could get some dedicated mag rim hub motors for e-bicycles that look cool and are strong and light (will not make it look like a scooter). Maybe real magnesium.

About the FF and gelling. This is not a concern for 99% of the members on here unless you do a long mountain climb every day and have a huge battery pack. Most member don't have a large enough battery for FF gelling to be a concern. So what if you need new fluid after 5 years or whatever. By then hub motors will by a 1/2 the weight with with the same torque

FF is the great addition for the everyday ebike that goes up and down hills and rides in the rain and dirt (sealed motor). For a race bike liquid cooling or forced air venting is still going to be obviously superior. Potting the windings will help in any case but don't expect huge gains in heat sinking, the big gain with potting is increased insulation life on the magnet wire