If speed can accurately be determined using using stroboscope or FFT vibration analyser, a rotor bar problem may show up as higher than expected slip for existing load. For example if motor FLA is 100A, nameplate speed 1780 (60hz world), actual measured current is 80A, then we expect slip to be approx 0.8*20rpm = 16 rpm and expect speed 1784. Higher slip/lower speed may be one indication of rotor bar problem.
In current, rotor bar problem may cause oscillating current indication. That oscillation is more readily detected using FFT analysis of current waveform. Look for pole pass sidebands around 60hz. When they get in the neighborhood of 0.5% of the main 60hz peak we start to suspect rotor bar problem.
Similarly in vibration we look for pole pass sidebands around 1x and harmonics of 1x. This is an indication but of rotor bar problems but not conclusive.
Rotor bar problems are often thermal sensitive and often vibration and oscillation gets worse over a period of hours after start from cold. Also often accompanied by a growling sound.
An relatively non-intrusive off-line test is the single-phase test. With rotor removed there are additional inspections available.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
Most people tend to think that motor current FFT is the way to go. But there are a couple of other ways, as you indicated.
In the good old days, the analoge ammeter did quite a good job in filtering away all frequencies above five or so Hz. So it was quite easy to see the small low-frequency current variations that a broken rotor bar results in. It is very similar to the frequency analysis that you get from an FFT process. Look for small periodic variations in motor current. I use a clamp-on ammeter with soft iron instrument. Works well and is not affected by PWM inverters. Hard to get these days, though.
Measuring slip and comparing to calculated slip is also a useful technique. As an example, assume that your motor has 1450 RPM rated speed at full load (50 Hz grid). The slip will be close to zero at no load and 50 RPM at full load. Now, you have to know the shaft load. So you will need to know the kW input and then multiply by efficiency AT THAT LOAD, and that is where it gets tricky. But if you arrive at, say, 70 % load, then your slip should be 35 RPM. That is, the speed shall be 1465 RPM. If you measure a much lower speed, like 1440 RPM or so, you can be sure that one or more rotor bars are broken.
A broken bar often/always shows as excess rotor temperature and if someone tells you that the shaft or the bearings are unusually hot compared to what it used to be, you should start thinking rotor bars.
The number one method (not counting FFT) is to look for slow periodic motor current variations, though.
I agree with electricpete. The best way is to look for the sidebands around 60Hz in the current spectrum.
WARNING! There is a chance to identify something as rotor bar problems, yet it really was an oscillation of the load. The rotor bar signature is on one very definite spot (as a function of slip) in the current vs. frequency spectrum. Here is how you make sure that it is a broken bar and not an oscillation of the load: You can calculate the particular location of the broken bar signature in your spectrum to be at the "fr" frequency:
fr = ffund * (1-2*s)
where "ffund" is the fundamental (50 or 60 Hz for line operated, or whatever frequency your VFD is giving out).
"s" is the slip, which is equal to
s = (synch. speed - operating speed) / (synch. speed)
The fr frequency should come out to up to 1 or 2 Hz below the fundamental. If there is not enough "separation" from the fundamental, then either increase your time of data acquisition, or your load.
Make sure that you have no load change during your data acquisition, because that shifts the sideband around, and you end missing a severe rotor bar problem.
Other effects of broken bars are: longer start-up time (even causing a trip), higher stator currents than a healthy cage, hotter motor, and the stator current going up and down with a consistent beat of less than 1 time per second (looking at it with a Fluke or at the display of your MCC).
There are also tools available on the market, that make this whole thing automatically for you.
By the way, there are several devices made that do this. I can't think of the names right now, but I have info on them somewhere that I will post, probably at 3AM when they occur to me.
"Venditori de oleum-vipera non vigere excordis populi"
We use our Entek vibration data collector with clamp-on current probe instead of accerometer. The data is processed, stored, and displayed in the same manner as our vib data.
PDMA also makes an on-line and off-line box with pretty big array of capabilities. I've never used it but heard some good things.
One test possible with pdma off-line tester:
Rotor influence test - Similar to single-phase test except it uses low power signal and relies on rotor residual magnetism.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
If you do not have analysis equipment, just use an analog current meter to measure the current in one phase. If at reative light load, you see the needle swinging at a low frequency and a reasonably high swing, and when the load is increased, the frequency of the swing of the needle increases, then this can be a good indication of a broken bar or two. i.e. you can see an oscillation in the current that is proportional to the rotor slip. This is a poor man on the job out in the sticks approach, but it is a good indicator. I have witnessed indicated current oscillations of 20 - 30% under these conditions.
I would suggest syncronizing the data collention with
the rotational speed : Establish in software N bins
and sample the current of each phase N times per revolution and add up the values in each bins.
You can repeat this process for a number of revolution
which is selected to represent close to an integer cycles of the line freqency so it is integrated out ( the integral of a sine for 2*pi interval == 0 !)
