OVERVIEW: Current signature analysis shows decreasing trend in the db difference between 60hz peak and estimated sideband peak... in the neighborhood 45db.  Tests are done at low load below 50% (system does not allow loading this machine any higher).   No change in vibration.

MACHINE DATA: 500hp, 3600rpm horizontal sleeve bearing motor with cast aluminum rotor, driving centrifugal chiller compressor through shim pack coupling.  Compressor has integral gearbox and rotates around 7100rpm.

SYMPTOMS

Slide 1 - Decreasing trend in "db down".

  • Note 1A - we are not able to obtain 50% load due to low heat load on chiler
  • Note 1b - Not all points have good agreement between sideband frequency and predicted fault frequency, but the last point 04/06/16 shows very good agreement and also agrees with speed deduced from vib spectrum.

Slide 2 thru 6 - Spectra used to determine db down for 12C.

Slides 8 thru 12 - Spectra for sister chillers,   all 57db or greater difference

Slide 13 - Vibration overall values remain completely stable

Slides 14/15 - low-level pole-pass frequency sidebands in vibration spectrum

Slide 16 shows expected relationships: load vs current; load vs speed

Slide 17 shows estimates of load from measured current and speed (measured at the same time) on 12C. 

  • There is not good agreement between estimates from current and from speed  However, there is no upward trend.  If rotor were significantly degraded, for a given actual load there would be increased current and increased slip.  Assuming the average actual load is roughly constant over time, the  estimated load based on these two paramters would increase over time (not evident).

 Other observations:

  • On 4/7/16, observed ammeter. It was oscillating regularly at once per four seconds. Range was about 30-30.5A.  Matches pole pass frequency based on speed measured during vibration 4/6/16.
  • On 4/18/16, observed ammeter again.  It was oscillating regularly at once per 3 seconds.  Range was 35-36A.   Inspected linkages (inlet vane and hot gas bypass valve) and no oscillation was present. Went back and inspected ammeter and still oscillating the same. Inspected ammeters for the three other running sister chiller compressors and no similar oscillation was present.    This suggests the oscillation is related to rotor problems rather than to process oscillation.
  • The current clamp-on probe in use since 2014 has thumbwheel calibration adjustment which is susceptible to being inadvertantly moved. This may cause errors in current magnitude (which affects prediction of sideband frequency) but will not cause error in db difference. This affects the 2013 reading and January 2016 readings. April 2016 reading was taken in conjunction with vibration so that speed is available for more reliable judging of sideband frequency (this will be continued)

 ACTIONS

  • Retest when available
  • Long term -Coordinate current test to occur at same time as vib test to reduce uncertainty in determining pole pass frequency. Look for alternate current probe.

 
QUESTIONS

  1. Is there reasonable confidence in diagnosis of rotor defect.
  2. What is urgency of repair?
  3. Any additional in-place tests recommended ? (at low load). 

Attachments

Files (1)
Original Post

Hi Pete.

Difficult to make a conclusive judgement based on the low load test situation. 

Here's the guidance from GE ref DB difference and associated level of damage. For your difference of approx 45dB it suggests High resistance joints or likelihood of a cracked bar. I am guessing that these limits are however based on higher load situation and at a guess your dB difference would be less (SB higher) under more load due to increased torque?

There is a lack of harmonics of 1x with PP SB which is often quoted as a sign of deterioration (still cannot get my head round harmonics of 1x for rotor bar issues though!)

I would agree that the lack of speed reduction-current increase would point toward a lack of significant deterioration. 

I dont think this is urgent at the moment however what it the risk? Given the design of the rotor is there any chance of a bar breaking loose and contacting the stator?

Sorry I cannot be of more help

 

Gary

 

Attachments

Thanks for your comments Gary - I was hoping you'd weigh in.

 [quote]I am guessing that these limits are however based on higher load situation and at a guess your dB difference would be less (SB higher) under more load due to increased torque? [/quote]

That's the way I look at it also.

[quote]I dont think this is urgent at the moment however what it the risk? Given the design of the rotor is there any chance of a bar breaking loose and contacting the stator? [/quote]

No, that scenario does not seem a big risk. But at the same time, I'm also having a hard time visualizing the degradation mode of cast rotors that could result in increasing pole pass sidebands in current (if that's what we're seeing). Does anyone know of case studies where this happened?

 [quote]There is a lack of harmonics of 1x with PP SB which is often quoted as a sign of deterioration (still cannot get my head round harmonics of 1x for rotor bar issues though!) [/quote]

The two occurences I have faced previously were fabricated rotors and both resulted in vibration with increasing 1x plus increasing harmonics of 1x (and sidebands around all that).

This one shows no change in 1x appearance and no appearance of harmonics. There are sidebands of the low level 1x/2x now, but I don't know if it's increasing… never took high resolution until recently.

 So it's a different pattern, which makes me wonder if it's real.  Another thing that initially made me wonder: If you notice in the powerpoint slides 2-7 the location (frequency) of sideband that I predicted from the current magnitude (calculation in upperp right hand corner of the slide) did not always match the location that a large sideband appeared on the spectrum. It leads to suspicion that maybe it is not a pole pass problem, just a process oscillation. But I think I rule that out based on two factors:

1 - As discussed above on 4/18/16 I saw current oscillation but confirmed there were no linkage oscillations going on at the time. The visible linkages control the only automatic controls on the refrigerant side (inlet vane and hot gas bypass). I don’t think an oscillation on the chill water side of the system would show up that strongly and with that fast a rate of oscillation (15cpm or so).

2 - Confirmed the location of the sideband using an actual speed measurement (from vibration) rather than guessing from current during the most recent measurement 4/6/16.  So the only time I actually measured speed the peak matched, but many times when I used current it didn't match. Suggests the estimate of load from current might be bad. Although I spent some time developing what I think is a good formula for relationship between current and load for this motor based on factory-recorded currents at no-load/25%/50%/75%/100% (slide 16), I have a feeling that for some reason there is an error in estimating load from current on this motor. This feeling is based on slide 17 which was developed from panel meter current readings and speed (vib) measurements taken at the same time (yes we log current every time we take vib on this particular machine). We would expect these two to agree but they don't. Why not?? I don't have a good explanation but I suspect the estimate of load from current is poor because 1 - we're at low load where the relationship is trickier and 2 - we don't account for routine voltage changes over a range of about 150 volts on the 4kv bus feeding this machine.  Of course voltage plays into the relationship used to infer load from speed also, but there is no need to know precisely the intermediate variable load when determining slip from speed, we calculate slip directly from speed. So in the future I'm going to try to always get vib at same time as current signature so I won't be left guessing load from the current to estimate slip.

Would cast rotors not respond in the same way as fabricated ones i.e. a high resistance in one particular rotor bar pair 'loop' due to cracking leading to a momentary torque reduction?

The software I used to use (Motormonitor) quite often used to 'miss' the PP sidebands in the current spectrum and it was often better to use the manual search mode. I do recall however that this system used the nameplate details to calculate slip. I no longer use this software (now using ascent) and always try to confirm speed at the time of measurement from either a vibration reading or by direct measurement.

I have never done it using current/load I would like to say it is because there are too many variables but the real reason is because I hadn't thought of it! Still giving some thought to a current/load method for speed but it does seem like there are a number of variables to account for and no direct linear relationship between rotor degradation, current and speed?

There is always the possibility that the PP SB's are due to a dynamic eccentricity (rotating variable air gap) as has been suggested previously but I have never seen this as we have previously discussed. 

I would be tempted to just watch this one progress. May be worth if you can to get a direct speed measurement (strobe/tacho etc) to give you 100% confidence in the PP calculation?

 

Gary

 

 

 

 

Epete/Gary,

Two cents

Had a one time check on a 500HP 3600 rpm motor driving ammonia compressor, cast aluminum rotor with rotor bar issue. Motor audibly modulated but yet the vane controls were rock solid. Strobe light inspection of the running shaft also showed the motor was "speeding up and slowing down". Large amperage swings especially under a heavier load with analog ammeter. Motor exhibited signs of excessive heating/reduced HP.

We loaded the machine and unloaded it several times. The PPF sidebands would change frequency as the load changed because the rotor was turning faster/slower.

RIC test of rotor showed rotor bar fault, customer changed the motor and no problems since

1xPPF sidebands were present at 1 2 3 4 x rpm and there were multiple PPF sidebands around each harmonic of run speed

Dave

Dave Reynolds posted:

Epete/Gary,

Two cents

Had a one time check on a 500HP 3600 rpm motor driving ammonia compressor, cast aluminum rotor with rotor bar issue. Motor audibly modulated but yet the vane controls were rock solid. Strobe light inspection of the running shaft also showed the motor was "speeding up and slowing down". Large amperage swings especially under a heavier load with analog ammeter. Motor exhibited signs of excessive heating/reduced HP.

We loaded the machine and unloaded it several times. The PPF sidebands would change frequency as the load changed because the rotor was turning faster/slower.

RIC test of rotor showed rotor bar fault, customer changed the motor and no problems since

1xPPF sidebands were present at 1 2 3 4 x rpm and there were multiple PPF sidebands around each harmonic of run speed

Dave

Thanks.  Do you have an idea which of the following two categories your case  falls into:
A - condition present since initial manufacture with no change in symptoms
OR
B - condition which became worse over time

The signatures do indeed look like a rotor defect. Why not put analogue ammeters around two phases & see if the needles move in sync or not? Mechanical modulation would cause them to oscillate in sync, whereas a rotor defect would cause one needle to lag behind the other. Learnt that from Tom Bishop of EASA, works quite well.

I have indeed seen actual rotor damage in a cast aluminium rotor. These were three 440 kW motors; sidebands on one were 35 dB down, the other two were 61 dB down.

Regards,

Aditya

Thanks Aditya, always glad to have your thoughts.

The ammeter might be good for our situation since I am a little uneasy that maybe I am missing process oscillation and I don't have access to pdma set for swirl effect.

I saw description on page 7 here: 

http://www.pump-magazine.com/p..._29/rotortesting.pdf

Do you happen to know if it could be done with two identical analog clamp-on ammeter vs in-line ammeters?  That would make it a heckuva lot easier  for us.

 

 

 

 

We only used clamp-on ammeters, had no problem with them.

Regarding the 440 kW motors, I have only one set of data so don't know if it was a manufacturing defect. As the plant was only a few years old, it may well have been. These were Siemens motors. The problematic motor (Motor B) didn't have high vibration:

Motor A - 3.4, 4.3 & 1.9 mm/sec, RMS (H, V, A- driving end)

Motor B - 1.5, 3.2 & 1.9 mm/sec, RMS (H, V, A - driving end)

Motor C - 3.8, 9.2 & 2.1 mm/sec, RMS (H, V, A - driving end)

However, it did have some very small PPF sidebands around 1X in vibration. These were missing on the other two motors.

Regards,

Aditya

Summary (TLDR):  Strobe test shows SLIGHT indication of pole pass frequency oscillation of the rotor position.

Details:

Machine is running ~ 3590rpm. Pole pass 20cpm ~ once per three seconds.

Getting a view of the shaft and capturing it on video was quite challenging due to:

1 - cylindrical (vs u-shaped) coupling guard. It is anchored to the compressor (I had to take videos of shaft through a small gap at the motor end… luckily there are marks on the shaft to help see its movement)

2 - Digital frame rate of the camera interacts with strobe to create image pulsations. I reoriented the camera by 90 degrees so the moving bands on the camera are perpendicular to the apparement movement of the shaft.

3 - Machine is elevated and we're not allowed to climb onto it.. so I had to shoot from a little distance.

4 - Had to focus on metal shaft, from a distance, with a strobe that can cause glare. My approach was brute force - took about 15 videos from various angles. Refocusing between each video. Luckily one or two came out where you could actually see the shaft moving.

5 - Speed is drifting maybe 3589 to 3591 over a timeframe of 30 seconds or so presumably releated to load and unrelated to pole pass oscillation.

6 - Position oscillation at pole pass frequency (assuming it exists) is very slight.

But I muddled through it.  In the first half of  video linked below, I think there is some oscillation evident superimposed on wandeirng speed oscillation. Over the first 10 seconds the observed side of the shaft is rotating downwards, but it stops about every 3 seconds and alternately pulses downward about every three seconds. Between 10 and 20 seconds it is roughly going back and forth every three seconds (changing direction). 

https://drive.google.com/open?...ibh6UTTd0SVFKZVhKb0U

EDIT - THE VIDEO QUALITY IS BETTER IF YOU DOWNLOAD TO YOUR HARDDRIVE (CLICK ON LINK AND THEN CLICK ON ARROW TO DOWNLOAD), BUT THAT DOES GIVE A SECURITY WARNING SINCE FILE IS TOO BIG FOR GOOGLE TO SCAN. IT IS SAFE AS FAR AS I'M CONCERNED (WENT FROM MY CAMERA TO MY COMPUTER TO GOOGLE DRIVE).  IT'S UP TO YOU.

I did study a sister machine for awhile and did not see the same thing.

In the old days I wouldn't have thought that much movement (a few degrees, barely noticeable amongst the noise of random speed variation) is significant but based on the other thread I think it is roughly the amount expected for our sideband magnitude based on the other thread. I'll see if I can estimate from the math in other thread when I get a chance.

I don't think strobe rules out mechanical load oscillation as a cause (might be coincidentally at the same frequency… hopefully can rule that out with two ammeter test - scheduled four weeks from now). But at least maybe it rules out "dynamic eccentricity" which has been mentioned as possible cause of false alarms in current. I don't think dynamic eccentricity would cause noticeable torque pulsations (although there is a mechanism). Probably more importantly on that front is that our 1x vibration hasn't changed a bit.

In the old days I wouldn't have thought that much movement (a few degrees, barely noticeable amongst the noise of random speed variation) is significant but based on the other thread I think it is roughly the amount expected for our sideband magnitude based on the other thread. I'll see if I can estimate from the math in other thread when I get a chance.

Here's my attempt to crunch those numbers based on my best estimates for this motor and the formula from the other thread )

33 % load and sum of sidebands is ~ 40 db below. X ~ (0.33)*(1/100)
X:= (0.33) /  (10^(40/20)) = 0.0033

Taccel ~ 2*sec; # Taccel  unloaded acceleration time assuming motor nameplate torque throughout the start

Fnp = 60/sec; # nameplate speed for 3600 rpm machine

Tpp = 3*sec; # polepass period when running at  3590RPM

Estimate:

ThetaDegrees = 18*Tpp^2*X*Fnp/Taccel = 18*3^2*0.0033*60/2 ~ 15 degrees.

I don't think I saw that much oscillation in the video but I'll look at it again.

ElectricPete

I have not had a chance to look through most of this post.  However, based on the original slides I would suspect that you may have some fractures in the cast rotor.  The PPF (Pole Pass Frequency) will vary with load (2x Slip Frequency).  If you are taking the data from the machine CTs, sometimes this will also dampen the peaks or cause the sidebands to vary a little - non-linear to load.

Rotor bars will primarily cause a torsional vibration, so may not show up significantly in vibration data, which is why MCSA/ESA is important for rotor bar detection.

The action point is usually -35dB from the peak current.  Additionally, if you have harmonics of PPF, then you should consider that you have multiple fractured/broken bars.  However, this is still unusual in cast aluminum rotors unless the rotors are overheating or the end rings are separating.

As you have gear involved, and you are seeing similar lower level signatures on the sister motor, you will also want to verify that you are not seeing the gear mesh.  Also, if the motor with the potentially poor rotor bars and the sister motor are running at the same time, and if they are on the same branch, you may be seeing a reflection of the rotor bars from the motor in question on the sister motor.  One of the reasons why ESA (voltage and current) is important when analyzing.

Based on your initial data and the signatures, I would check the gear frequencies as well as calculating the PPF based upon the actual operating speed.  If it is PPF, you need to go no further.

Sincerely,

Howard W Penrose, Ph.D., CMRP

MotorDoc.com

electricpete posted:

OVERVIEW: Current signature analysis shows decreasing trend in the db difference between 60hz peak and estimated sideband peak... in the neighborhood 45db.  Tests are done at low load below 50% (system does not allow loading this machine any higher).   No change in vibration.

MACHINE DATA: 500hp, 3600rpm horizontal sleeve bearing motor with cast aluminum rotor, driving centrifugal chiller compressor through shim pack coupling.  Compressor has integral gearbox and rotates around 7100rpm.

SYMPTOMS

Slide 1 - Decreasing trend in "db down".

  • Note 1A - we are not able to obtain 50% load due to low heat load on chiler
  • Note 1b - Not all points have good agreement between sideband frequency and predicted fault frequency, but the last point 04/06/16 shows very good agreement and also agrees with speed deduced from vib spectrum.

Slide 2 thru 6 - Spectra used to determine db down for 12C.

Slides 8 thru 12 - Spectra for sister chillers,   all 57db or greater difference

Slide 13 - Vibration overall values remain completely stable

Slides 14/15 - low-level pole-pass frequency sidebands in vibration spectrum

Slide 16 shows expected relationships: load vs current; load vs speed

Slide 17 shows estimates of load from measured current and speed (measured at the same time) on 12C. 

  • There is not good agreement between estimates from current and from speed  However, there is no upward trend.  If rotor were significantly degraded, for a given actual load there would be increased current and increased slip.  Assuming the average actual load is roughly constant over time, the  estimated load based on these two paramters would increase over time (not evident).

 Other observations:

  • On 4/7/16, observed ammeter. It was oscillating regularly at once per four seconds. Range was about 30-30.5A.  Matches pole pass frequency based on speed measured during vibration 4/6/16.
  • On 4/18/16, observed ammeter again.  It was oscillating regularly at once per 3 seconds.  Range was 35-36A.   Inspected linkages (inlet vane and hot gas bypass valve) and no oscillation was present. Went back and inspected ammeter and still oscillating the same. Inspected ammeters for the three other running sister chiller compressors and no similar oscillation was present.    This suggests the oscillation is related to rotor problems rather than to process oscillation.
  • The current clamp-on probe in use since 2014 has thumbwheel calibration adjustment which is susceptible to being inadvertantly moved. This may cause errors in current magnitude (which affects prediction of sideband frequency) but will not cause error in db difference. This affects the 2013 reading and January 2016 readings. April 2016 reading was taken in conjunction with vibration so that speed is available for more reliable judging of sideband frequency (this will be continued)

 ACTIONS

  • Retest when available
  • Long term -Coordinate current test to occur at same time as vib test to reduce uncertainty in determining pole pass frequency. Look for alternate current probe.

 
QUESTIONS

  1. Is there reasonable confidence in diagnosis of rotor defect.
  2. What is urgency of repair?
  3. Any additional in-place tests recommended ? (at low load). 

Hi Pete

I couldn't download attachment can you mail it please.

orugbosamuel@gmail.com

thanks.

Add Reply

Post
×
×
×
×