MCSA - Motor Current Signature Analysis 4

[JP.Duranti]

I would like a mathematical example of how to calculate MCSA frequencies in faults like broken rotor bars and air gap eccentricities and bearings. I trying to make a python code to automatically analyse it. Any tip will be welcome!

 

[rc10]

I have a spreadsheet to do this and superimpose the frequencies over our MCSA data. The points of interest I can look at are: Stator slot passing sidebands, stator line frequency sidebands, broken rotor bar and swirl effect sidebands, and eccentricity.
You will need the following motor info: line frequency, number of poles, running speed, and number of stator slots and rotor bars.

Stator Sidebands
Center frequency (Hz) = # stator slots x Running Speed (Hz)
Sidebands are at: Center Frequency (Hz) +/- the Running Speed (Hz)

Notes: Stator winding problems are found when sidebands appear around the stator slot passing frequencies center frequency (# slots x running speed).
Stator mechanical (ie loose coils, stator core movement, etc): line frequency sidebands
Stator shorts: line frequency sidebands with running speed sidebands
NOTE: If viewing a demod plot, these sidebands will be around the running speed (line frequency is removed)

Broken Rotor Bars
Upper Rotor Pole Pass Sideband 1: Line frequency (Hz) +/- (Synchronous Speed - Running Speed (Hz))*# poles
Upper Rotor Pole Pass Sideband 2: Line frequency (Hz) +/- 2*(Synchronous Speed - Running Speed (Hz))*# poles

Notes: Broken rotor bars can be found looking at the slip frequency sidebands of the line frequency. Values are negative relative to line frequency FFT. (<-50dB is rule of thumb.)
>-60dB Excellent
55-60dB Good
49-54dB Moderate (Trend)
43-48dB Rotor fracture or high resistance joint (Increase test interval)
37-42dB Two or more bars cracked or broken (Confirm w/ motor circuit analysis)
31-36dB Multiple cracked or broken bars and end ring problems (overhaul!)
<30dB Multiple broken rotor bars and other severe rotor problems (overhaul!)
Swirl Effect: Swirl effect occurs at the 5th harmonic of line frequency (300 Hz on a 60 Hz line frequency). Swirl peaks are a confirming tool for the sidebands around line frequency.

Eccentricity
Eccentricity Frequency: Running Speed (Hz) * # Rotor Bars
Eccentricity Peak 1: Eccentricity Frequency +/- Running Speed (Hz)
Eccentricity Peak 2: Eccentricity Frequency +/- 2 * Running Speed (Hz)

Notes: Static eccentricity can be found in line frequency sidebands of the rotor bar x running speed center frequency. Dynamic eccentricity is the same as static eccentricity but also includes running speed sidebands around the static eccentricity sidebands.

 

[rc10]

Attached for my sheet

 

[JP.Duranti]

@rc10, thank you so much for providing the sheet.
Additionally, do you happen to have an example that utilizes these frequencies to analyze a machine with one of the mentioned faults? I'm looking to visualize how it appears in the data.

 

[electricpete]

Attached is a case study of slowly declining trend in current signature results related to pole pass sidebands around line frequency (expressed in db), representing a degrading rotor condition. I think starting in slide 2 you see several examples of calculated db from the spectrum. The unit 12C was the problem unit, the other ones were sister units.

We swapped out the rotor and it fixed the db trend. We did some testing / inspection of the removed rotor and there was nothing obviously wrong in single phase test or thermography test. We did see some anomalies near an end ring using green paper but I'm not convinced it was a big problem.

We only use current signature for rotor problems. I'm not connvinced there is a useful stator anomaly to be found on current signature analysis. We did have a scenario when loose stator coils showed themselves in vibration on a 4000hp motor (2*LF and harmonics of 2*LF... pattern went away when the loose coils were wedged which proved the cause/effect... but the inspection was a lot more definitive than the vibration and there is danger of false alarm if anyone uses this symptom alone to diagnose loose coils).

 

[electricpete]

I notice the terminoilogy related to the rotor related sidebands discussed above is different among vib analysts and electrical folks, which can lead to a lot of confusion.

Vib folks tend to talk about sidebands spaced at "pole pass frequency" Fp, where Fp is the number of poles times the difference between sync speed and actual speedd.
[ul]
[li]Fp = p * (SyncSpeed-ActualSpeed)[/li]
[li]If I have a 2 pole motor at 60hz supply running at 3590rpm, then the pole pass frequency is 2*(3600-3590=2 x 10 = 20 rpm[/li]
[li]If I have a 4 pole motor at 60hz supply running at 1790rpm, then the pole pass frequency is 4*(1800-1790)=4 x 10 = 40 rpm[/li]
[/ul]


Electrical folks talk about sidebands spaced at "twice slip frequency" by which they mean 2* s * FL . That's bad terminology imo (*) but setting aside the terminology, 2*s*LF works out to be the same thing as Fp:
[ul]
[li]Fp = p * (SyncSpeed-ActualSpeed) = p * SyncSpeed* [1-(1-s)] = p * SyncSpeed* s[/li]
[li]Substitte SyncSpedd = 2*LF / p[/li]
[li]Fp = p * SyncSpeed* s = p * (2*LF / p) * s[/li]
[li]Fp = 2*LF * s (this is what is intended by the phrase "twice slip frequency")[/li]
[/ul]

(*) Where does the terminology "twice slip frequency" even come from? I think "tiwce slip" refers to 2*s. And they tack on "frequency" to mean LF.
But there's a big danger in this terminology is if you interpret the phase "twice slip frequency" to mean twice (SyncSpeed-ActualSpeed).
It's an easy interprtation to reach since "slip speed" often means (SyncSpeed-ActualSpeed).
But that interpretation will get you in trouble. IF we apply that interpretation to the previous examples of 2 and 4 pole motors where SyncSpeed-ActualSpeed=10rpm, we come up with 2*10=20rpm for both. That is the correct pole pass frequency for the 2-pole motor (which may lead to a false sense of understandinmg) but it is the wrong answer for the 4 pole motor.

 

[rc10]

@JP.Duranti: attached is a combo of my personal reference for identifying these points of interest using our new online MCSA monitoring tool and a report I had made using our field tester. Although the tester software can automatically find these points of interest for me, I want to be able to come at the results with an idea of where the ROI should be.

@electricpete: great explanation! Looking back at my formulas, am I in error? I've used my sheet on 4-, 6-, and 8- pole motors but could be missing something.

 

[JP.Duranti]

@rc10 and @electricpete, thank you for providing the examples. The remaining doubt that I have pertains to the amplitude of the peaks of eccentricities and bearing faults. While there is a fixed table for identifying broken rotor bars, I'm uncertain about how to determine the threshold or notice these faults based on their amplitude. Could you please provide insights on this matter?

 

[electricpete]

rc10 - your formulas look absolutely correct to me. I was not responding to your post, I was just pointing out a common mistake which I have seen people make.

JP Duranti - I'm not sure I understand your question. If your talking about broken bars, rc10 provided db limits and I provided example. I can provide more details on those calculations if that's what you're asking...

But I think you want more details about other faults. Personally I don't use current analysis for anything else and don't know of any other limits.

It's interesting you mentioned "bearing faults". From my standpoint, there is zero reason do use MCSA to look for bearing faults if you have access to vibration (because measuring vibration is a much more direct and established means of monitoriong bearings). But if the motor is difficult to access for vibration, then I can see it would be tremendous benefit if we could analyse current to detect/charcterize bearing faults. I have seen case studies where frequencies associated with bearings were shown in the current spectrum, but none of the case studies I saw convinced me that current signature can give reliable detection / discrimination of rolling bearing faults. Then again I'm not particularly up on the latest developments in condition monitoring... if anyone has link to something like that, I'm all ears.

 

[JP.Duranti]

Yes, the broken rotor bar analysis seems to be the more often used. But my job is try to make this work for other faults too.
I have some articles about it. But I could't find a established amplitude for determining if there is problem or not.

 

[electricpete]

Thanks for the article. I vote you a little purple star for posting that.

My first impression of the paper after a few minutes is not favorable. Namely the defect they are detecting is shown in figure 2. It occurs by drilling all the way through the cage (and then contacting the ball). It is not a realistic defect. It is also a large defect (in terms of amount of metal removed) compared to the types of small defects that we easily detect and characterize with vibration on the rolling elements and races. Then again the cage has a much different role than rolling elements and races so it's hard to compare. The size of the ball defect caused by drilling through the cage was really never shown... they should have disassembled the bearing afterwards to inspect and photograph the ball defects. For that matter I would have preferred to create a defect on accessible portion of rolling elements or races using a small punch (that would seem like a more realistic defect to me). In the end I am left wondering about the sensitivity (is it only drilling into the bearing that can be reliably detected, or can we detect much smaller defects like we do with vibration).

But I may have some confirmation bias, or else a tendency to want to prove what I already said on the forum, so I'll definitely give it a closer look and see what I can learn from it.

I did find another interesting reference which is available for free if your employer is an EPRI member (otherwise not available)
Electrical Signature Analysis (ESA) for On-Line Equipment Condition Monitoring: Fault Detection Testing

I'm lucky enough to have access to that so I'll make some time to look at it along with your paper. I plan to do that this weekend.

 

[electricpete]

I looked a little further at your posted article. Some more thoughts.

I choose to focus on the ball freuqency sideband and not the cage frequency sideband. Here's why:
[ul]
[li]I can understand what a ball defect is and how it relates to ball defect frequency (also called ball spin frequency). Ball defect frequency is how long it takes the ball to spin around once. If there is a defect on the ball then it will impact a given race at ball defect frequency. Sometimes in vibration we see twice ball defect frequency appear because during a single spin of the ball defect, it contacts two races (inner an outer).[/li]
[li]I don't understand what a cage defect is practically (cages don't generally get discrete defects, they typically wear) nor how such defect would produce cage frequency vibration. Kinematically, the cage frequency is the rate that the cage moves with respect to the stationary race... if you took a photo of the rotating cage at t=0 and t=1/Fcage, the cage would appear in the same position in both photos. There is no impacting expected to occur at cage frequency under almost any circumstance (a defect in the cage may be continuously in contact with the ball as ball surface slides by, but there is no impacting associated with the cage defect like there is for ball defect or race defect). Cage frequency is of interest in the vibration world primarily as a modulating frequency for ball pass frequency. Let me explain what I mean by that. We place our accelerometer at at a given location adjacent to the outer ring. Then a given defective ball (which travels around the bearing at same speed as the cage) passes by that sensor location at a frequency of cage frequency. So the ball defect defect pattern measured at that location is impacting at ball defect frequency (or twice ball defect frequency) where the impact magnitudes increase and decrease at a rate of cage frequency (as the impacting occurs nearer and further from the sensor). The associated spectrum is harmonics of ball defect frequency (or harmonics of twice ball defect frequency) with cage frequency sidebands. Knowing cage frequency helps us recognize the pattern in vibration. It's not clear what the role of cage frequency might play in modulating a ball defect signal in current, since the aspect of the defect moving past the sensor at cage frequency no longer applies in current like it did in vibration.[/li]
[li]All of which is to say, I don't even know what exactly it is that is represented by this cage frequency sideband. They never explained it and there is no obvious physical connection.[/li]
[/ul]
... so I am inclined to focus on the ball defect sideband. Inspecting tables III and V for the ball defect sidebands, I see that 5 out of 8 ball defect measurements went up, while 3 out of 8 ball defect measurements went down! So it is barely more significant than the 50/50 split of 4 up / 4 down than we might expect from random variation... and that's the take home message for me.

The vibration does not suffer the same problem. All 8 of the ball frequency measurements went up in table VII and IX. And on top of that the changes from normal to damage level 1 and from damage level 1 to damage level 2 were much higher and more distinctive.

I know no-one here in the paper is claiming current signature is better than vibration signature for analysing bearings, but from this little bit of information, my first impression is that analysing by current appears waaaay less definitive than analysing by vibration for this particular defect. And speaking of "this particular defect", we should return to the point of my previous post which is that we don't even know the nature of this particular ball defect. We only know that it was created by drilling all the way through the cage. I would say it may have been quite severe damage to the ball after the drill penetrated the cage. So if the indication on current is marginal / questionable for this potentially severe ball defect, I don't have much confidence it would show for a more typical / less severe ball defect.

 

[electricpete]

I did skim the EPRI document bearing section. It looks like they did have a pretty thorough test program. They introduced 3 different faults (loss of bearing lubricant, particle infiltration with fine particle sand, and bearing corrosion by submerging in acid). And yes, those sound more realistic to me although I still didn't get a good understanding of the severity of the faults (except "loss" of lubrication which might mean none). They tested systems from 5 different vendors (All-Test Pro, Baker-Megger, Bentley Nevada, Motor Doc LLC, and PdMA). Two vendors detected all 3 faults, one vendor detected none, and the other two vendors detected one or two of the three defects.

The report is terse without a lot of data. There are a few figures but not enough explanation to go with them to understand WTF is going on imo. It's not the kind of proof I'd like to see, but the tone of words in the report suggest the author thought it would be useful for detecting bearing defects. And no doubt, he would know a lot better than me.

The report also looked at quite a few other fault types besides bearings and the other faults we talked about already.

But to go all the way back to your question... no, there are no limits listed in the EPRI report.

I think my tangent is done now...

 

[JP.Duranti]

Perfect, Thank you very much for the research!