Testing for Broken Rotor Bars

[mbailey]

I am working on a project to test for broken rotor bars by use of a current clamp and vibration data while the motor is under load. Has anyone done this type of analysis or con you direct me toward information on this subject? This is isolated to integral horsepower motors.

Thanks,
Mike

 

[gordonl]

GE Multilin is testing a broken rotor bar detection system on their SR469 motor protection relay. They monitor harmonics of the current for the harmonics produced by a broken rotor bar. I've also seen a gizmo that stuck to the side of a motor and tried to detect broken rotor bars by monitoring the flux at the "gizmo", I don't remember the company name.

 

[electricpete]

Pole pass frequency Fp=p*Fslip where p is number of poles.

Basically in vibration spectrum you will look for pole pass sidebands around 1x, 2x, 3x etc. I have never heard an explanation for this phenomenon. But I believe that the source of these sidebands originates in the fact that the non-uniform rotor will generate nonuniform (pulsating) torque which pulses every time the defective bar passes a pole (at pole pass frequency), which leads to a speed that changes at pole pass frequency. If you'd like I can demonstrate mathematically for you that frequency modulation at Fp gives Fp sidebands around 1X running speed (If you're an EE student... try to derive it yourself). The more commonly known fact is that amplitude modulation at Fp also gives Fp sidebands around 1X... but it is tough for me to see how amplitude modulation is at work for a rotor bar problem.

In the case of current monitoring you are looking primarily for Fp sidebands around Line Frequency.

CSI's article about using current is here:

http://www.compsys.com/drknow/aplpa...1&Count=50&Expand=8</a></p><ul> <li><a href=

Siemens article about motor vibration (I'm not sure if it addresses rotor bar problems) is here:

http://www.sea.siemens.com/motorsbu/product/White Papers/VIBRATION PROBLEMS.pdf

One thing to be careful of using both technologies is false alarms. I believe that eccentricity can give similar symptoms.

 

[electricpete]

You can find some discussion on flux monitoring at the CSI link that I provided above under &quot;flux technology overview&quot;.

There is another paper on motor current monitoring at:

http://www.entekird.com/eact/eactpapers/eact99/Enteract99.htm

As I mentioned speed oscillation at FP would result in Fp sidebands around 1x in vibration spectrum. Here's my proof:

Let Wr be the radian frequency of the rotor (2*pi*1Xrunning speed) and Wp be the radian frequency associated with pole pass (2*pi*Fp). Assume we have torque oscillations causing speed oscillations so that position (angle) of rotor is given by theta(t) = cos(Wr*t+m*sin(Wp*t))
where m (known as modulation index in electrical circles) represents magnitude of frequency oscillation as a fraction of Wr. We can use trigonometry to rewrite it as theta(t)=cos(Wrt)*cos(msinWpt)-sin(Wrt)*sin(msinWpt). If m is small (<<1) then we approximate cos(msinWpt)~1 and we approximate sin(msinWpt)~msin(Wpt). Substituting into above equation for theta we get theta(t)~cos(Wrt)-msin(Wpt)sin(Wrt).
The term sin(Wpt)*sin(Wrt) is the amplitude-modulated form so well known in vibration circles, which gives rise to pole pass sidebands around rotor speed. Here's proof of that well-known fact:
sin(Wpt)*sin(Wrt) =½[cos(<Wr-Wp>*t)-cos(<Wr+Wp>*t) which has frequency components at Fp above and below running speed.

IF speed oscillation is really the source of the Fp sidebands around 1x, then we should be able to see oscillation of the shaft with a strobe. I have viewed the shaft with strobe in case where Fp sidebands were at least 10% of 1x running speed an saw no oscillations. In that case there was no broken rotor bars... presumably eccentricity. In any event it proves there are other ways for FP sidebands to show up... even though I don't understand them... don't see why amplitude modulation would come into play. Bottom line of all this rambling... can anyone explain why Fp sidebands around 1X are expected in the presense of broken rotor bar or eccentricity?

 

[mbailey]

elecricpete, thanks for you replies. I am using a CSI 2120 for vibration and rotor bar detection, also have a flux coil for stator condition testing. I have reviewed the information on the web as well as some info directly from CSI. This testing is new to me and I have conducted a trial run on two motors runing on a dynometer, one good and one suspect for bad bars. Data collected for the &quot;Rotor Bar Test Program&quot; on the 2120 showed dB levels in the same range. I am having two 5000 frame verticals built that have visible signs of burnt rotor bars on the rotor OD. These motors ran fine and passed a heat run test but on disassembly we found the burnt spots. Plan is to re-build with bad rotors and see if we can detect the broken bars. Information I have says up to 4-6 broken bars is probably OK. Do you know of any criteria for acceptable number of broken bars and is current analysis the most reliable method of determining this defect?

Mike

 

[electricpete]

I believe that most folks consider current analysis to be a better confirmation of rotor bar problems than vibration analysis. Personally I would think they would both be susceptible to false alarms. After all...how can the stator tell the difference between a rotor field assymetry caused by broken rotor bar and a rotor field assymetry caused by an eccentric rotor? I don't see how but maybe someone can explain it to me. Is rotating eccentricity supposed to show up at a different frequency than rotor bar problems in current spectrum?

Of course off-line inspection and testing is probably the best confirmation. But a very common problem I have heard is that early developing rotor bar problems seen in the field cannot be detected in the shop. Presumably temperature and centrifugal forces affect the intermittent contact.

As far as how many broken rotor bars are ok... for critical machines I would think that the answer is zero. When one rotor bar is broken it puts more stress (due to higher current) on the adjacent rotor bars... if there were preexisting design flaws and stresses sufficient to break the first bar... those same factors will surely attack the adjacent bars under increased current (due to the first broken bar). How long that takes is unknown to me. If it is necessary to keep the machine in service than continued monitoring and trending can give some ability to guesstimate how stable the machine is.... also of course in such a situation you would minimize the number of starts. There was a pretty good description of just such a scenario at the vibration forum at reliability-magazing.com very recently.

The most common location for rotor bars to become broken is at the braze joint. Your discussion of bar burn marks (apparently not at the ends) leaves me a little skeptical as to whether this truly represents a rotor bar problem. Have you confirmed by other methods? One good method might be a loop test of the rotor while viewing with infrared. Or perhaps even better would be to pass current direct through rotor by hooking onto each ring (with the rotor supported on nonconducting wood blocks) and again monitor with infrared. Or close inspection using die penetrant, mag particle etc. Perhaps this has already been done but if overheating marks are your only indication I would be skeptical.

 

[electricpete]

Here is an abstract of an Iris Conference paper. If you email them using the link provided, they will send you a copy of the paper free of charge.

http://www.irispower.com/techpapers/tp00_02.html

 

[mbailey]

electricpete,

These rotors are produced using spin cast technology in which centrifical force causes the alumimum to be spun to the outside of the rotor slots during the pouring process. These are ducted rotors, so the actual rotor bar is not visible on the completed rotor. The end rings are also formed during this process. The burn marks that I referred to are located about midway of the rotor and at the air duct spacer. We know that the shape of the air ducts can cause stress risers due to a change in cross sectional area from the air duct to the lam slot shape and size and eventually lead to a failure. An attempt was made to size the air duct spacers to the slot shape in an effort to eliminate this problem. I am assuming that the hot spots on the rotor are a symptom of a broken rotor bar and since the rotor bar is buried in the rotor, the only way I have of confirming this is to cut the rotor open. Mag-particle or infra-red will not allow me to see cracks in the bar itself. Thanks for the response, these are all good pointers and the links have been helpful also.

 

[mbailey]

I performed a test on a 5000 frame vertical 700 hp, 6 pole motor with 58 rotor bars that was suspected as having three broken rotor bars. The test showed sidebands at 2*slip frequency around line frequency with a delta dB of 42.3 dB. Using the criteria from CSI, this indicates moderate damage (<43dB, more than one broken bar) to the rotor. Calculations estimated 0.85 broken bars. Visual inspections show three areas on the OD of the rotor at two adjacent air ducts as hot spots. Actually melted the rotor bars in these locations. Question: Does anyone know how the calculation of est. broken bars compares to actual defects? The equation that I used is: n=(2R)/(10^(delta dB/20)+np) where
n=# of broken bars
R=# of rotor bars
delta dB= difference in amplitude of line freq. to next highest sideband at 2*slip freq.
np=# of stator poles
I was expecting the estimate to in the neighborhood of 2-4 broken bars.