There were lots of postings about the rotor bar problem lately. But I still want to add another one. This one was an exceptionally easy call. Motor 1400 hp, 1786 rpm running at 6000 rpm pump through a gearbox. There were all signs of a serious problem. Just listening to the woo-hoo sound was almost enough to shut the motor down for good.
I am posting it just to demonstrate how the vibration analysis and the current analysis go hand in hand.
Look at the attached spectrum from the MOH (Motor Outboard Horizontal). The splitting of the vibration peaks is remarkable. The waveform tells you clearly why there are so many sidebands to the 1x, 2x, 3x… and so on peaks. The vibration signature was taken with Fmax = 400Hz.
The ultimate confirmation was the current analysis. The current signature shows several sidebands. From the speed measured by strobe light I calculated the position of the first sideband (Fc=59.16 Hz). And indeed the rotor bar sideband was exactly there. The decibel ratio of 38 dB made the call a breeze.
The motor was taken apart the next day (yesterday) and 3 bars were broken just at the point of attachment to the ring.
What is remarkable, that 3 motors out of about 15 in a row of similar pumps had broken bars! One would not expect anything like that with low inertia pumps that start in a few seconds.
jank

1400_hp_with_broken_bars.ppt (486 Kb, 174 downloads) corelation of current analysis and the vibration analysis

Thanks for sharing that case.

A DOL start lasting a few seconds is a tough start in comparison to large pump motors at our plant which start in less than one second (with a few exceptions like RCP motors driving large flywheel). But then again we have a very stiff power system and conservative design which may not be typical of other plants. (also we don't have frequent starting).

As you probably know, the 3.3:1 speed increaser gearbox makes the pump inertia appear to the motor about 11 times larger than it really is.

This message has been edited. Last edited by: electricpete,

Alright... Here's a 1000HP with 4 Poles (40 Rotor Bars) that we pulled last week. Broken/Cracked Rotor Bars is a common problem with this group of machines, so this too was an exceptionally easy call...

I haven't heard yet what was found in the shop, but based on past experience with these machines I'm setting the Over/Under on # of Broken Bars at 13.

_2_3rd_Grind_Mill_Data.pdf (79 Kb, 111 downloads)

Thanks for posting that Michel.

I notice that you have a LSB (lower sideband), but not much USB (upper sideband) .

In a previous thread, we have discussed the fact that some theoretical papers predict this occurs for a high inertia load. I also posted in that thread an example of RCP motors with very large flywheel inertia that act this way (I have never seen any other motors that act that way).
http://maintenanceforums.com/eve/forums/a/tpc/f/7161085...912/m/1011028692/p/1

Your signature has this same pattern... almost no USB. Is yours a high inertia load?

This motor drives a machine that definitely has a high inertia load… The attached pic is nearly identical to my machine.

Thanks, that's another good data point. Seems to confirm what that author said. Apparently speed modulation shifts portion of the LSB into the USB.... can't happen with high inertia. I wonder if we should allow a little less stringent limit if there is only LSB since apparently it is higher magnitude (lower DB difference) than it would have been on a lower inertia load.

ElectricPete,

I just received the pics that were taken of the rotor in question (see attachment). 3 bars were broken at the connection on both ends.

Your comment about allowing a “little less stringent limit” with only ‘LSB’ rings true in this case. With a –31 dB in the current spectrum I suspected much more rotor damage than what was actually found!

1000HP_Rotor_Pics.doc (806 Kb, 73 downloads)

Michel,
The pictures of your failed rotor could be substituted for mine (that I did not post). The rotor is of identical design, probably the same manufacturer. (But who wants to talk about manufacturer on a failed motor?)
The bar failure looked the same in my case. A hairline crack, barely visible, just at the point where the bar sinks into the pool of silver solder in the grove of the ring.
As in your case 3 bars were broken, one of them on both ends (?????). Why on both ends?!!!
Yet our current signatures were a world apart.
This seems to be a good opportunity to look at the differences. I will describe my settings and other details:

a) The motor: 1400hp, 1800rpm, 4160 Volts
b) Running a 6000 rpm centrifugal pump through a gearbox.
c) Easy start across the line in about 2 seconds.
d) Infrequent starts, unless there is a problem, and then nobody counts the starts; it simply has to run no matter what.
e) Current signature was taken with upper frequency set to 80 Hz. (CSI 2120).
f) 3200 lines of resolution.
g) Current clamp was Fluke 400i - AC current clamp 1000:1 ratio. The current is fed into 5 ohm resistor. The voltage developed across the resistor is fed into the analyzer input. No other processing as far as I know.

The point of this exercise is to find out if the inertia is the main difference (as pointed out by Electricpete) or if there is something else. A rare opportunity it would be interesting to take advantage off.
jank

This message has been edited. Last edited by: jank,

Sounds like a good opportunity to make use of measuring current inrush, either with PDMA Emax or a Dranitz. BEWARE!!! Many OEM's will KNOWINGLY undersize the motor, just to keep the cost down. I had a Cooper Turbocompressor sales engineer tell a customer, in front of me, that if he didn't operate the unit all the way to the top of the motor service factor, he would not get the published air capacity!!!!!
I looked at him and asked why his company didn't put a bigger motor on the unit? The customer told us to take our argument outside!!! The customer, in my mind, was being screwed and didn't even want to know about it.
Broken rotor bars only occur for three reasons,
1. Poor workmanship
2. Incorrectly sized motor for application
3. Abuse by end user (too many starts in a given interval).
I understand that 'it has to run', but if you punch in a motor start multiple times in a short span on a high inertia load application, you have doomed the motor.

I am interested in whether the Velocity sidebands around harmonics of shaft speed has the same difference frequency as the current signature sidebands. Can you tell me the delta frequency please.
Your spectrum only shows harmonics.

Your current signature sidebands appear to be at 4x the difference between sychronous speed and actual speed. Is this correct ?
What formulae do you use to calculate the delta frequency.

Urgent repley will be apprecitated.

Thanks

John

I am attaching a detailed view of the separation of the “mechanical” sidebands. They are indeed separated by 4 times slip frequency for 4 poles. The broken bar causes a disturbance under each pole regardless of the polarity. And in this case the motor was a 4-pole.
The electrical separation is always 2x slip frequency for any number of poles. More exactly like this:
slip= (Ns-N)/Ns
Ns=synchronous speed
N=running speed
frequency of the lower sideband:
Fls=Lf(1-2s)
Fls=Frequency of lower sideband
Lf=Line frequency (50 or 60Hz)
s=slip

jank

This message has been edited. Last edited by: jank,

separation.pdf (198 Kb, 38 downloads) mecanical sidebands around harmonics of running speed

Thanks very much for the zoom which shows mechanical sidebands at the same frequency as the current sidebands.
I've always thought that because it's an electro magnetic force the frequencies should be the same.
One more question, the time domain cycle goes on for about 35 seconds best guess which is about 0.029 Hz.
What do you think causes this modulation frequency to be generated so strongly.

Cheers

John

UPDATE ON 1400 HP MOTOR.
The motor has been put back into service several weeks ago. Finally I had a chance to take a base reading after the repairs. The highlights are in the attached ppt file.
First slide is a new current analysis. The load was little lower but still the decibel ratio grew from 38 dB to over 62 dB.
The comparison spectrum from the second slide clearly shows that the large, split peaks (the bottom spectrum) have disappeared (the tag shows July 17, the new spectrum is from Oct 4). The vibration dropped to virtual zero. There are only 3 peaks visible on the new spectrum. A tiny 1x, little larger 2x and then a peak transferred from the pump (~3.3x).
The waveform from the slide #3 is much more revealing. The bottom waveform from July 17 indicates that the spectrum peaks had to have sidebands due to amplitude modulation. The waveform from Oct 4 is basically constant amplitude, mostly due to the vibration transferred from the pump.
There is an identical motor next to this one. Again the vibration is a virtual nil.

jank

1400hp_REPAIRED.ppt (519 Kb, 23 downloads)

quote:

.
What do you think causes this modulation frequency to be generated so strongly.

Cheers

John

There are 38 bars on the rotor in this particular motor. Since it is a 4-pole there are at least 4 bars where the current is zero or nearly zero. I would go even farther: There are at least something like 3 bars per pole that contribute to the torque very little or nothing. It leaves 38 – 12= 26 torque producing bars. Now consider that 3 out of the 26 bars do not function because they are broken. The 3 bars constitute a huge asymmetry of the rotor. On the other hand when the broken bars are in the position where they would not carry any current anyways (in a particular moment), the motor appears perfect.
Any asymmetry of the rotor (electric or magnetic) will generate sidebands around the harmonics of the running speed. The broken bars produce an asymmetry that sometimes create noticeable sidebands, hence they can be used as an indicator of the motor problem. The broken bars produced noticeable sidebands in the case of my 1400 hp.

jank

Hi Jank,

Thanks for the reply.

I understand the 4x slip speed modulation ok an its modulation period will be the inverse of the 4x slip speed frequency.

But my second question is about the much longer period modulation in the original waveform which has a period of about 35 seconds best guess, which is about 0.029 Hz.

Any ideas ?

Cheers

John

John,
The time axis in the waveform originally posted is actually in revolutions. Converted to the time it is 1000 milliseconds. Hence the large amplitude excursions of the waveform explain the splitting of the peaks pretty good.
I took great care to take the vibration spectrum on the repaired motor under the same conditions. The upper frequency was 400 Hz, just the signature on the repaired motor was taken with 6400 lines, and the original is with 3200 lines of resolution.

jank

That RB signature looks like the one I had on my SKF current tester. I'm not familiar with RMS though. We always use ips & db. I'm still learning all these things though.

What kind of data collector did you use to get that spectrum? What is the slip freq on your 1800 rpm motor about 18 to 23cpm. PPF should be around 60-80cpm for a 4-pole motor right...I can't see that on your spectrum. There is some other kind of measurement there. In amps...And the scale on the right I'm used to reading in DB's...

quote:

Originally posted by jank:
There were lots of postings about the rotor bar problem lately. But I still want to add another one. This one was an exceptionally easy call. Motor 1400 hp, 1786 rpm running at 6000 rpm pump through a gearbox. There were all signs of a serious problem. Just listening to the woo-hoo sound was almost enough to shut the motor down for good.
I am posting it just to demonstrate how the vibration analysis and the current analysis go hand in hand.
Look at the attached spectrum from the MOH (Motor Outboard Horizontal). The splitting of the vibration peaks is remarkable. The waveform tells you clearly why there are so many sidebands to the 1x, 2x, 3x… and so on peaks. The vibration signature was taken with Fmax = 400Hz.
The ultimate confirmation was the current analysis. The current signature shows several sidebands. From the speed measured by strobe light I calculated the position of the first sideband (Fc=59.16 Hz). And indeed the rotor bar sideband was exactly there. The decibel ratio of 38 dB made the call a breeze.
The motor was taken apart the next day (yesterday) and 3 bars were broken just at the point of attachment to the ring.
What is remarkable, that 3 motors out of about 15 in a row of similar pumps had broken bars! One would not expect anything like that with low inertia pumps that start in a few seconds.
jank

Rod,
I use CSI 2120 analyzer, but I do not use any program for current analysis. I simply let the current from the current clamp flow through a 5-ohm resistor. The voltage drop is then fed into the analyzer on the voltage input. The voltage is proportional to the current.
So unlike your current signature in decibels, my current signature comes out in Amps.
However converting the Amps to decibels (dB) is a simple task: Just look what the amplitude of the line frequency current peak is. Also look up what the amplitude of the first sideband is and do the following calculation:

dB= 20 * log (AMP/ASP)

AMP = amplitude of the main peak
ASP = amplitude of the sideband peak
dB= decibels

In other words, since you have chosen not to do this simple calculation, you have to pay the big bucks in maintenance fees of the program. On the other hand you may be lucky enough to have the money and do not have to be as cheep as me.

The speed of the motor is below the spectrum: 1787.4 RPM.

jank

This message has been edited. Last edited by: jank,

We do it very similarly. We use an Entek data collector and put the data into our Emonitor vibration data base along side our vibration data. By clicking on the vertical scale you can switch between linear/log/decibel so you don't even need the formula.