100 kRPM Brushless Motor Design 4

[sdk_imported]

Hello,

I would like to see some discussion about high-speed brushless motor & controller design(60-120kRPM, 1-5kW). Our biggest challenge is preventing rotor heating and maintaining stable speed control. We currently are using two-pole segmented magnets, segmented magnet sleeves, high resistivity materials, high speed switching, and external inductors ... but, we are still having issues. Will any of the following help me?

Slotless configuration
Non-PWM based speed control
Laminated rotor yoke
Something else I'm not thinking of..

If anyone has successfully designed and tested a motor in this speed and power class I would very interested in hearing from you. Finally, do you know any companies that make motors like this in large quantities?

Thanks.
Scott

 

[Skogsgurra]

sdk,

I have had the same problem with a motor for air for fuel cells. No other solution than using pure sinewave. Check NFO Drives. They are doing just this.

 

[ozmosis]

You could try Precise Corp on

http://www.precise-corp.com

I've used them in past lives and they are very good and reliable capable of going upto 160k rpm but it depends on the application. These are used in high speed machining applications. I used them in the semi-conductor industry.

 

[jbartos]

Suggestion: It appears that magnetically levitated bearings would eliminate lubrication problems and increase Mean Time To Repair (MTTR).

 

[sdk_imported]

skogsgurra,

Did you have to use external inductors with the sine drive? Did you design your motor to create a sine back EMF also? I checked NFO's website and it appears that they make drives for AC induction motors. I have a DC input to the drive. I sent them a message for more info. Thanks for the suggestion and the lead!

Scott

 

[Skogsgurra]

sdk,

Actually, I did not have the problem myself. It was a company that I did some consulting for. They use PM motors with a very precisely calculated field distribution that makes the EMF of the motor sine-shaped. NFO is currently developing special inverters for them since the rotor is very small and cannot stand the heat that losses from eddy currents and hysterisis cause in the rotor. There is also some concern with the bearings being destroyed by PWM inverters as well as having the CAN-bus disturbed by the PWM HF pollution. There are no inductors or filters between motor and inverter.

 

[Highspeed]

We have a running design up to 90kRpm but not at that power requirement, only a hundred watts.

Tips :
>> Slotless/toothless "large airgap" stator (toroidal)
>> keep magnetization around 1.8 Tesla using N40H 2 poles ring magnets

>> Sensorless operation (ST72141) using back-emf or Field Oriented Control (not sure that DSP can handle that speed)

>> 0.05mm thick laminations of high grade silicon steel

>> Keep bearings as far as possible from magnet & stator

>> non-metallic shaft is a must (but expensive as you'll need stiffness)

>> The stiffer-the better

>> Shoot for a balancing nightmare (we have a 5µgram resolution in our application to keep it quiet)
We ended up balancing using bonding UV flashed
(see

http://www.loctite.co.uk/PDF/Brochures/MagnetBonders.pdf).

>> driver side : if you can keep voltage below 40-60V then you'll have very low rds-on mosfets making things a lot easier also the wire insulation will be kept reasonnable.
>> Bearings selection : you'll need to have the Mfg involved in the design : up to 30-50 krpm standard bearings can do the job, above that, special lubrication formulations are required.
>> miscellaneous
-In the stiffness equation, bearing preload is an important parameter as well as play .
-H5 and K5 tolerances for shaft and bearing housing is what we have.
-FEM-BEM softwares can be of great help.
Good Luck
HS

 

[jbartos]

Suggestion: The manufacturing accuracies or very small tolerances will reduce the motor vibration and noise. Also, any mechanical and magnetic eccentricities should be negligible.

 

[sdk_imported]

So far all of your comments have been very helpful.

Sounds like sinewave drive, sensorless, two pole rotor, with a sinusoidal back emf current wave shape will help.

Cost is a big concern for me.Can a slotless motor be made low cost in high volume? If I go with a slotted stator, what are some things I can do generate a sinusoidal emf waveform?

Highspeed: How does a non-metalic shaft help me (stiffness or losses)? What material would you recommend looking at? Also, do you have any recommendation for high speed lube on ball bearings?

 

[yurim]

A nice didactic example of a complex electrical / mechanical / thermal etc. engineering problem and professional insight from Highspeed.

Let me put my couple of cents.

1. For high speed, most electromagnetical (eddy current) loss takes place in stator yoke due to fundamental flux, hence thick high grade laminations. PWM core loss rather counts for moderate speed / direct drives (

http://www.drbrushless.com).

2. Good point is to keep bearings far away from PWM influence. PWM is not a major evil from losses perspective - chokes will not help, rather thieve AC voltage.

3. Laminated rotor yoke does not make much sense for PM machine due to large airgap (no fundamental frequency rotor losses). Slotless rotor structure helps to reduce second order PWM loss in rotor.

4. Field Oriented control helps to keep copper loss minimal(sensorless is orthogonal).

5. To prevent magnets flying off consider appropriate glue and magnets wraping.

 

[Skogsgurra]

yurim,

Yes, chokes do steal voltage. They steal room and weight as well.

Yes, most of the losses are in the stator. But that is usually not a problem with water cooled machines. The problem is to get the heat out of the rotor - so it is crucial to keep the rotor losses very small. And this is where sinewaves help.

No, it does not help to "keep bearings far away" - the problem is the capacitively coupled voltage from stator winding to rotor and distance does not help much. EDM seems to be a bigger problem with these small, high speed machines than anyone would think. Sine does not cause that kind of problem.

 

[UKpete]

Some time ago in a previous company I worked on 5kW 120krpm PM motor/alternators, both slotted and slotless. Performance of the slotted machine was superior, we used the Micro Linear 4426 driving a separate igbt module on both types, and no external inductors. The slotless machine is potentially cheaper though, the very large effective airgap means that the stator core (which is a plain cylinder) works down at something like 0.3T, so you can use a cheap lamination material or even powder iron core. I tried both helical and axial conductor slotless, the former has virtually no end-windings but the latter has better copper utilization in the airgap – with helical you are putting a skewed conductor in the gap, effectively more copper and lower voltage than purely axial conductors. Because the gap flux passes through the winding, it has to be thin stranded (like Litz wire) to limit eddy currents in the copper.

The back emf for the slotless machines did have a very good sinusoidal shape, this would certainly suit the sinusoidal drive. We did talk to NFO at the time. With the Microlinear controller, the rotor temperature was significantly lower if controlling speed by varying the dc link voltage i.e. not allowing it to go into PWM. According to my old notebooks we could get 5.75kW @ 120krpm without PWM compared with only 2.5kW at the same speed with PWM, on the slotted motor for the same rotor temperature (140°C).

Skogs is also right about the rotor losses, they are very difficult to remove. The rotor eddy current problem is due to both space harmonics (due to slotting) and the time harmonics (due to harmonics in the current waveform). In a slotted machine, the back emf isn’t too sinusoidal, I don’t know if you would get the full benefit of a sinusoidal drive, and I don’t know the cost.

There are several influencing factors that determine the rotor losses. A metal can doesn’t help, carbon-fibre is better though it can act as a bit of a thermal blanket and it’s not so rugged mechanically – and it will fail at too high a temperature! The slot opening/ slot pitch ratio should be minimum i.e. make the slot openings as narrow as possible and use more slots (yes, it puts the cost up). A large airgap will also dramatically reduce rotor losses, with some reduction in performance. This is where the carbon-fibre sleeve helps. Cooling by air through the gap, or oil-spray at the rotor ends may help although allegedly the oil may increase the windage losses.

The losses occur in the surface of the magnets and some in the surface of the shaft where the magnets are fixed. There is probably limited benefit from segmenting the magnets and any metal can, and even less from laminating the rotor core.

Using ceramic ball bearings will eliminate any problem with bearing currents.

One possible way forward might be to do a collaborative project with a suitable University dept, I only know the UK ones so I can’t really help there. This is because there is a lot of complex calcs involved in predicting rotor losses. Testing alone is difficult, it’s hard enough just measuring the rotor temperature reliably.

There are a few companies around who work in this field:

http://www.calnetix.com

http://www.levitec.de

http://www.s2m.fr

 

[Highspeed]

Still based on our 100W design:
Keeping bearings far away in high induction high speed motor not to put them in the flux path (1/100th of the flux through magnets is still a lot).
Now what is far away: an initial prototype with 1.5 mm airgap toroidal stator, we had bearings 10mm away
Eddy currents in bearings resulted in a 2-5% efficiency loss at nominal speed/load (I let you evaluate the over heating ....).
Things were fixed when moving them at 25mm away (empirically determined).
When using toroidal stators, air gap is, by construction, big simply because windings are in between magnet and stator core. But with a toroidal stator it's straight forward to get a sine shaped back-emf.

Another tip :
Ring magnets : Zn coated...sourced from china they are really cheap and you can get high tolerances within lots.
our supplier :

http://www.millennium.com.hk

due to federal regulations and patents issues if you are in the US, dealing magnets with china is a bit more complex

While centering them on shaft we have a <0.01mm excentricity but we used a centering tool rather than trying to tight fit them on shaft.
We adjust the centering tool for each lot we receive.

For the non metallic shaft, initially we had a standard 314S metallic shaft and had some derived flux through it meaning it magnetized (dunno why) going to a ceramic shaft resolved the issue but raised another: the cost so we went back to metallic shaft with a spacer between it and magnet (was good enough and under high pressure to get done with the design, didn't go further)

Back to driving mode, we are using the 6 steps constant torque approach using STMicro back-emf topology with a 40kHz PWM, reducing commutation losses. This is probably not the best fit for your application as precise speed control could be tough.

Grease/Oil formulation : I don't have any access to it, we have our reference at NSK and we had a 3 month process to tune and validate it.
ToDo: call them, give CRS & quantities, they are very reactive.
Cheers
HS

 

[sdk_imported]

UKpete,
Thanks for sharing from your experience. I wish I would have posted this question sooner as I have spent a lot of time wrestling with this problem.

I don't understand how increasing the airgap size reduces rotor heating. Is it because it allows more cooling flow, or because of the additinal reluctance in the magnetic circuit? My airgap is currently ~1mm.

There seems to be something to segmenting the magnets ... some say it will help, others say it won't. The testing I've done shows that there is a benefit. I wire EDM'ed radial rings into the magnet (1mm thick with .1mm gaps between) and saw significant reduction in the rotor temperature.

 

[UKpete]

sdk, I worked on it for a quite a long time, I think the project got shelved eventually (how does that make me feel!) I'm happy to help where I can.

In answer to your question, it's the latter. It puts a large reluctance between the non-synchronous fluxes eminating from the stator, and any rotating conductor. The 1mm you quote is fairly typical, but if you have a non-conducting rotor sleeve (carbon-fibre is best) say 2.5mm thick, thats a 3.5mm airgap that the non-desireable fluxes have to cross. If you've got a metal rotor sleeve (what are you using?), it is very close to the stator so even though the gap reluctance is much the same, the sleeve catches all those fringing fluxes around the slot openings etc.

I don't think the airgap has much effect on the cooling, and I don't think it affects the windage losses very much. Do you actively cool the rotor? I'm also interested to know how you measure the temperature, we used infra-red but it wasn't easy (or cheap).

I can understand now why the magnet segmenting helps, I was assuming they were several mm thick. I guess that's pretty labourious and expensive. Are you using NdFeB? it has slightly higher resistivity than SmCo, every little bit helps.

One thing I didn't try was a backfilled winding, i.e. the windings are loaded from the outside of the stator so the slots are effectively closed on the stator bore. That should reduce the rotor loss.
There is a European supplier that has done some work on this:

http://www.cogent-power.com/

(Surahammars Bruks is the Swedish supplier of speciality lamination steel). I can probably get more info if you need it.

The problem with all these mods you can try is that it is time-consuming and expensive to evaluate it from tests alone, that is why I think it is helpful to get someone to do some accurate analysis of the rotor losses. It's too complex for me, but there are people out there who can but mostly in Universities.

 

[jbartos]

Comment: The original posting does not indicate any duty cycle for that motor. The ball bearings do produce noticeable heat unless lubricated and cooled simultaneously. The larger motor 100kW has magnetic bearing on this site:

http://www.calnetix.com/Applications/MPD100.tml

Any heat from those magnetic bearings is easier to remove since the bearing is essentially in the motor stator.
To cool the motor rotor with 1mm gap and overcome heat sources at the ends of the motor will need an effective cooling system.

 

[UKpete]

jbartos the use of magnetic bearings in high-speed machines is an interesting point.

One problem I have noticed with them is that they are quite bulky, but there is a strong requirement to reduce the length of the machine for dynamic reasons. It is also debateable whether they need to be used with touch-down bearings for those occasions when things may go wrong.

But on the positive side, if they are working correctly, at least they don't need changing every 10,000hrs or so, and no oil system is required (assuming there is no other need for one). They also allow a degree of dynamic balancing and fine adjustment of rotor position.

As regards cooling, unfortunately there is a relatively large pressure drop through the airgap, possibly too high for a regular blower. In other words, it may need a pump rather than a blower if the airgap is small.

 

[jbartos]

Comment: There is more going on for magnetic bearings, namely, they provide galvanic isolation between the motor stator and rotor; therefore, there is no path for bearing currents or damages done by bearing currents, if VFDs are being applied.
Aircraft engines are often using them. They do not appear very bulky there. Visit

http://search.netscape.com/ns/boomf...ft+engine%22&amp;clickedItemURN=http%3A%2F%2F

http://www.pr.afrl.af.mil%2Fsstorie...e_url=http://www.pr.afrl.af.mil/sstories.html

etc. for more info

 

[Skogsgurra]

So far, this has been a very interesting and educational discussion.

In high-volume applications magnetic bearings have some disadvantages. The need for a toch-down system is one. The relativly high cost is another (an active system needs sensors and coils built into the end bells).

Alternatives are air bearings, but tests and calculations have shown that air gets very &quot;lossy&quot; when speed goes up and the heat will be very high. So air is not an option in the dives that I have seen.

Passive magnetic bearings (no external electronics) also need touch-down and the heat generated in the induction winding gets very high if the system is to work over a wide speed range.

Regarding life: A bearing that gives 10 000 hours life (L90 is probably what UKpete is talking about) is more than adequate in an automotive system. It would mean 500 000 - 1000 000 kilometers and that is an interval that usually exceeds the life of the car the system is built into.

My personal thinking is that hybrid bearings will be used in these motors. The ceramic balls are much lighter than steel balls and this is very important since the centrifugal forces that press the balls against the outer race will be much lower and hence increase life of the bearings. Or allow for smaller and cheaper bearings.

 

[jbartos]

Suggestion: Visit

http://dmt

http://www.epfl.ch/~jsandtne/GPMB/project1.htm

for: magnetic suspension possibly eliminating touch down bearings

http://www.bardenbearings.co.uk/media/adobe/seminar_amb_-_March_2001.pdf

for: magnetic and touch-down bearings in one system
etc. for more info