DC motor risers break 2

[Skogsgurra]

Two 500 V 1500 kW DC motors for a reversing rolling mill have had catastrophic failures in that the risers have broken and (as it is assumed) caused a collector round-flash. In both cases it happened after around three months of operation and there is no sign of arcing or other electrical damage to the breaks. They look like fatigue breaks.

There are some speculations as to the reason for the breaks. I do not disclose what the thoughts are, because we would like some unbiased and fresh thoughts. What are the more common reasons for a case like this?

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Are we really seeing that riser flapping around?!?! That's a lot!

Keith Cress
kcress -

http://www.flaminsystems.com

 

[controlsdude]

I was curious why the risers were not made with multiple thin flat pieces of copper welded together for better flexing?

 

[edison123]

I am not seeing much lateral movement of the risers with or without the fan. If you suspect a greater movement, an additional lashing around the bend might help?

controldude - the risers I have seen are always of two single pieces so that they than wrap around the armature conductors for soldering/brazing.

Muthu

http://www.edison.co.in

 

[MikeHalloran]

From here, it looks like they're moving....


Mike Halloran
Pembroke Pines, FL, USA

 

[Skogsgurra]

Need to add that the fan is on in all the pictures in the 2 Feb 16 20:43 post.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Woodyjumps]

You gentlemen are way beyond me in the DC motor world but I may have some usefull feedback.
1) Whoa yeah that looks like a lot of flexing in the images for copper, which annealled or not work hardens quickly. Visually issolating just a single laminated pair based on the indications on the edge... you can see that the apex of the curve is moving radially in and out, and the gap between the laminations is not remaining consistent, so those puppies are moving.
2) My experience is with auto shredders, 2500hp to 9000hp. Cardan Shaft Drivelines for these chaos machines can last for a decade, or they can last for 10 days, even given the same maximum torque rating. It is all about the natural frequency of the ASSEMBLY in this case, not just the individual component.
3) What you are seeing on a bench mounted motor may be dramatically amplified in the installed state due to the assembly natural frequency.
4) We use wound rotor motors of various designs up to 9khp at 450rpm (16 pole, 60zh)and EMD and GE DC motors 2500hp down to 1000hp, and periodically piggybacked EMD 2500hp motors for 4000 to 5000 hp applications. In all cases, the most dramatic failures of mechanical parts of the motors involve inadequate attention to the torsional vibration of the assembly.
5) I'm suggesting, for giggles, a bump test of the installed motor.

 

[Woodyjumps]

As an asside to that... the method for calculating the spring rate of the shaft of the motor is "fuzzy" at best, particularly when dealing with large motors which have broad variations in means of construction.
In our case, attached to the other end of the drive shaft is a machine, the rotor of which is a stacked up assembly of plates, shrunk to the shaft, and then through bolted with 3" dia tie rods with 300 tons of tension on each. It is a complex assembly. Determining the spring rate for this creature is complex as well and using "rules of thumb" is insufficient.
The method of coupling, the degree of interference between components of differing diameters, all quite complex and easy to have very broad errors due to the ranges of tollerances of manufacture.
Bump test on the other hand shows the real world, not the stacked up errors of hopes and dreams.
Changing from one motor design to another... different inertias, different shaft spring rates, these things obviously will affect spring rate of the assembly.
We've (our motor supplier) have done strain gage analysis on problem assemblies to save our asses when motor problems occurred after the customer changed drive shafts, where we've found dramatic torque reversals due to torsional natural frequency convergences. Very nasty stuff there.

Talk is cheap. I pulled off company refferences to protect the innocent and attached an example.

 

[Woodyjumps]

On the subject of motor construction, couplings, and spring rate:

 

[Woodyjumps]

A big name VFD and motor supplier from Taiwan has had success and experience replacing large DC motors in rolling mills with VFD driven heavy duty squirrel cage motors.
They've been approaching us for our abusive application, and I've spoken with their refference, a mill maintenance manager claiming very good performance and reliability. If you want to know more contact me directly and I can connect you to the right guy.
If they are doing it, likely others are as well.

 

[electricpete]

When I use a strobatach, it's not such a demanding task as you have. I tune the strobe to machine speed. Then I detune slightly which would show what appears to be slowly moving rotor. If there were something attached to the rotor that was vibrating at a multiple of running speed, then if I could detune very very slightly from running speed it would also appear to vibrate slowly.

You have a more challenging task for a variety of reasons. I guess variable frequency strobe was not bright enough or precise enough.

I'm not sure exactly what results are expected by non-sync strobing while varying speed. Away from resonant you don't expect to see much. The only points of interest are near resonance and since the flashes are not sync'd with anything, the images obtained will vary somewhat randomly within their range of motion. If you only get a few snaps passing through resonance, you may not catch it near an extreme range of its motion.

If stuck with single speed non-sync'd strobe, I think I would run the machine at a speed that I think excites a resonant frequency and take lots of shots at that single speed in hopes of catching it at various extremes of motion. I guess you could figure that speed out as the speed where the wandering peaks in audio spectrum or highest. Or else estimate riser resonant frequency and run at a speed which is a fraction 1/Nrisers of that frequency.

Just thinking out loud....

=====================================
(2B)+(2B)' ?

 

[Skogsgurra]

Pete, the main reason to use a synchronized strobe light is to be sure to see one and the same part of the risers every time. If you look at different risers from frame to frame, you cannot rule out that there is a difference in shape and position from riser to riser. The reason why I used a pair of extra strong LEDs is that it wasn't possible to bring a standard strobe close enough to the risers while running. Moving objects, high voltage and such problems. Those LEDs also gave a much better light intensity than the Xenon tube. Partly because I could vary pulse width, which isn't possible with standard stroboscopes.

Not that such a difference is very big, but the possibility that there is a difference can lead to unnecessary discussions and by using a synchronized light with an estimated maximum uncertainty of 0.3 to 0.5 mm (ten to twenty mils), one can say for sure that we are looking at the same risers in all the snaps shown. And, since the variation between the pictures is, as Mike, Keith and WJ point out, quite big the reason for the breaks is flexure in the risers themselves and not any thermal or shear/stretching forces from resonances. Torsional and flexional resonances have been proposed, but there is no possibility that a complex body like a rotor with plates, insulation and a low Q can resonate in such a manner that forces concentrate in the bend of the risers. Not even if running constantly at a speed where resonances occur. (Refresher: it is a reversing mill where speed is all but constant and risers broke in two machines after around three months of operation)

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Woodyjumps]

To ignore torsional resonance is to ignore the fact that each of those "spokes" has mass, and to ignore the slenderness ratio of them. Their length compared to their stiffness is substatial. Torsional vibration will cause them to move a lot, and since they are "other than" straight, they will have a much lower individual natural frequency than would a straight spoke.
I think however that the voice of experience is where to start first, and the most relevant shared experiences so far involve the "siren" effect.
Alternating vortex building and vortex shedding create tremendous oscilating forces in every endeavor in which they occur.
Still, due dilligence should be performed. When the consequences are high it is best to consider all possible contributing factors.

 

[waross]

I understand that the torsional resonance of the rotor as a whole is what has been ignored. Any resonant frequency of the rotor as a whole is unlikely to coincide with the resonant frequency of the risers.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[Woodyjumps]

Bill, we are absolutely in agreement there. The rotor enmasse shouldn't be anywhere near the NF of the risers.

I mention the resonance of the entire assembly - of the motor connected to the rest of the machine, because it appeared to be missing in the conversation.
The curved elements that are failing have, no doubt, a substantially lower natural frequency than the rotor.

When the inertia of this rotor however is connected, commonly in rolling mills by long cardan shafts and often at relatively extreme operating angles, the motor rotor mass on one end and the mill roll on the other.... things can get pretty springy in a bad way.

And if someone accidentally puts the drive shaft together wrong... well things get exiting quickly.

Yep...one more thing to check. If Cardan Shafts are used, be sure that they are correctly aligned.
When the splines are slid together it is critical that they align properly.
I've seen so much intelectual equity dying and retiring that nothing surprised me, and nothing can be assumed to be correctly assembled.

 

[davidbeach]

well things get exiting quickly

I'm pretty sure you actually meant exciting rather than exiting, but I can sure see how when it gets exciting one would be wanting to be exiting as quickly as possible. ;-)

 

[Skogsgurra]

re: resonance in rotor

The pictures were taken when running in a winder shop test bench with motor mechanically disconnected and with no load. The cardan installation and other possible mechanical problems have been ruled out in an early stage of the investigation that took place several months ago.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Tmoose]

Do you have any close up pictures of the fracture surfaces?

Was A weighting applied to the sound spectrum?

Are the "brush holder rings" continuous complete circles, and stationary?

Are there any risers with cracks/indications that have not progressed to complete breaks?

If a concern is riser-pass causing a siren effect, there may be ways to "soften" the pulse.
- Perhaps a non conductive disk could be added to the row of risers. Like the spoke guard on a Derailleur bicycle.

https://i.ytimg.com/vi/o9oVJUplaQI/maxresdefault.jpg

Pumps are sometimes modified to reduce the strength of blade pass pulsations that are causing noise or vibration.

http://www.mcnallyinstitute.com/12-html/12-06.html

Machining a full size impeller OD even a few percent to increase the gap to the cutwater can help a lot.
Sometimes just the vane OD is reduced. Or, the vanes can be cut on a taper.

Can the brush holder be shifted to increase the spacing?.
Perhaps the profile of the leading edge of the brush holder can modified to create a "softer" pulse. Or maybe adding epoxy blocks to create a rounded, or angled profile.
Like the upper edges of these motorcycle engine exhaust ports on t5he attached file.

 

[edison123]

Gunnar

Did the OEM weigh in yet? After all, they are still under warranty?

Muthu

http://www.edison.co.in

 

[Skogsgurra]

Yes, Muthu. We have an ongoing discussion with the manufacturer. Or, rather, had. I am not directly involved in those discussions. Sorry that I cannot say much more. For obvious reasons. And, yes, still under warranty.

Tmoose. The machines are no longer available. The papers you refer to are quite interesting and show that seemingly non-relevant details can have an effect. The air streams are, in comparison to the derailleur, enormous. It is actually difficult to stand on one's feet in front of the exhaust from the commutator side when the cover has been removed.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[edison123]

Gunnar

Understood. It will be interesting to see how this problem is fixed.


Muthu

http://www.edison.co.in