High frequency vibration in large HP motor

[Sparky4598]

Trying to diagnose some vibration issues here in a large 8,600hp VFD driven electric motor. Bearings are pressurized oil journal bearings. Directly coupled to another motor to apply a load, so no gear mesh frequencies or blade pass frequencies possible here. Frame is rigidly bolted to a stong floor and properly shimed. Rotating frequency is 500rpm.

The frequency analysis shows something i have never seen before. The acceleration components for 1x-~15x are very low, but there are multiple large spikes around 500Hz and 1kHz. I believe the carrier frequency for the VFD is 500Hz.
Does anyone have any insight here as to what could be causing this? Or any suggestions of things to check? The controller is shutting down the system because of the large vibrations in the higher frequencies.
Thanks in advance for any suggestions!

 

[GregLocock]

Can you do an FFT on the motor drive current?


Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Sparky4598]

@GregLocock I do not believe so. There are no current transformers installed that I am aware of. If there are, they are most likely internal to the motor controller and we would not be able to access those.
EDIT: actually maybe, I do remember seeing current on the hmi screen so if the PLC is logging it then yes I can. I'll check Monday.

 

[Strong]

Sparky,
"I do remember seeing current on the hmi screen so if the PLC is logging it then yes I can."
No you cannot do an FFT of a PLC input that is either 4-20 ma current or a DC voltage. It must be a dynamic signal for an FFT frequency analysis. Typically a current transformer (CT) is used as a sensor.

Walt

 

[Sparky4598]

Strong,
Thanks for replying! I apologize I did not word that very clearly. I meant if the PLC is logging it then there is a signal that we could connect the data acquisition system to and record the data, then do the FFT. My apologies.

 

[electricpete]

The controller is shutting down the system because of the large vibrations in the higher frequencie

Wait what? The vibration looks fairly low to me at least when viewed as an acceleration spectrum (all peaks < 0.1g). It seems like maybe the setpoint is really low, or you have some intermittent event occuring which wasn't going on when you captured this data. What is the tripe setpoint? (units? how is the unit processed for example as rms or as true peak?) Is there any trend of the parameter upon which the trip is based? Do you have an acceleration TWF to show us?

This spectrum wouldn't necessarily grab my attention for diagnostic purposes. If I did want to understand it, I would identify the frequencies and look for patterns of spacing of the peaks. You have no frequencies labeled. There can be a variety of frequencies that might show up:
[ul]
[li]The most common pattern includes sidebands spaced at "twice line frequency"... where line frequency in this case is the frequency supplied to the motor. We responding here don't know what frequency needs to be supplied to this motor to cause 500rpm speed because we don't know how many poles the motor has (what is the nameplate speed and associated nameplate supply frequency?).[/li]
[li]Of course vfd can generate other complex patterns of em vibration which are generally not of concern. [/li]
[li]Rolling bearing impacting can generate patterns in this frequency range. If this motor has rolling bearings, do you have the part numbers? (I would also want to know true peak acceleration and preferably trend to help characterize any rolling bearing defects)[/li]
[/ul]
Do you have anything to compare this motor to? (Have these peaks changed from historical or from similar motor on similar vfd?)

 

[Sparky4598]

Wait what? The vibration looks fairly low to me at least when viewed as an acceleration spectrum (all peaks < 0.1g)....

Hey electricpete, thank you very much for the detailed response! I apologize it has taken me so long to respond.

So the sensors on these motors are vibration sensors reading in in/sec from 3-1000Hz. The manufacturer recommended trip setting is 0.15in/sec RMS (which I also thought sounded low, but when it's vibrating at 0.20 in/sec for 5 seconds before it shuts down, you can really feel it in the floor). The manufacturer also recommended a trip setting of 1.0g for a 3-10,000Hz accelerometer probe if that is installed in place of the in/sec probes. The preferred probe is in/sec.

The plot I shared in the original post was from a tri-axial accelerometer reading in G's that was mounted on the same bearing cap as the vibration sensors. The in/sec sensors were only being logged at 10Hz data rate so pretty much useless for FFT. As of today, they were logged at 5,120Hz but I have not had a chance to analyze the data and create the FFT plots yet.

What is an acceleration TWF? I am not familiar with the term.

The motor is a 4pole motor. 1400rpm @ 46.7Hz @ 3,300V.

The motor does not have roller bearings, it has pressurized oil journal bearings.

I have tried, but have not been able to easily label the spikes other than doing it manually, which I will do before uploading another plot.

Thank you again for your response! Any help and guidance is very much appreciated!!

 

[electricpete]

4pole motor. 1400rpm @ 46.7Hz @ 3,300V.

ok then the supply frequency to obtain 500rpm = 8.33hz would be 46.7*(500/1400) = 4*500/120 = 16.7hz
The 2*LF spacing would be 33.3hz. It looks like 33hz is the spacing betwen some of the peaks between 400 and 600hz and again between 1000 and 1200.
If I had to guess your rotor bar frequency may be around 500hz which would correspond to 500/8.33 ~ 60 rotor bars. So it may be you have a familiar of RBPF+/- k*2LF and another family at 2*RBPF +/-k*2LF. There are some additional peaks in there that may not be fully explained by the pattern but that pattern is fairly benign in my opinion (please ignore any chart which tells you that indicates rotor problem... in some cases rotor problem can be accompanied by increasing rbpf pattern but the pattern also exists in healthy-rotor motors and can vary over time for a variety of reasons unrelated to rotor condition.).

What is an acceleration TWF? I am not familiar with the term.

Sorry, TWF stands for Time Waveform. It is a plot of acceleration samples vs time. From that one might be able to deduce true peak acceleration (although true peak can be missed if sample rate is not high enough) and whether the vibration is stable over time. The TWF may also give other clues... impacting has a distinctive appearance with sharp peaks and possible ringdowns, whereas electrical/rbpf-pattern vibration tends to be smoother (at least on non-vfd motors).

So the sensors on these motors are vibration sensors reading in in/sec from 3-1000Hz. The manufacturer recommended trip setting is 0.15in/sec RMS (which I also thought sounded low, but when it's vibrating at 0.20 in/sec for 5 seconds before it shuts down, you can really feel it in the floor)

The in/sec sensors were only being logged at 10Hz data rate so pretty much useless for FFT. As of today, they were logged at 5,120Hz but I have not had a chance to analyze the data and create the FFT plots yet.

If your setpoint is in ips and you are trying to troubleshoot the reason for the trip then you should be looking at a spectrum in velocity units (not in acceleration units) [EDIT - I apologize for not reading closer, it appears you knew that, and are already working your way in that direction]. That will give you a better idea of what is contributing to the trip assuming the trip is based on an rms of the velocity which I think is typical (true peak is traditionally reserved for acceleration). I think trying to analyse the high frequency stuff visible in acceleration is not very productive in this case, as in most cases... it tends to be harder to interpret and carries less diagnostic value (other than maybe for rolling element impacting or gear diagnosis)

If 33hz is a big contributor to velocity, that's 2*LF electrical, usually not much can be improved there (maybe check airgap and voltage balance).

If 1x is a big contributor to your velocity then maybe balancing would be in order.

I would also question why the trip needs to be so low (0.15ips). That is not a typical trip level for a motor unless there is sensitive equipment driven or nearby

Vibration felt on your feet may be a characteristic of the support structure floor. Also if you are comparing to other machines I'd say you will feel 0.15ips at 500cpm a lot more than you will feel 0.15ips at 3000cpm (our touch is more sensitive to the lower frequency). Either way, if you want to better understand the significance of the floor vibration then try to determine the frequency of the floor vibration.

 

[Strong]

OP Stated: "So the sensors on these motors are vibration sensors reading in in/sec from 3-1000Hz. The manufacturer recommended trip setting is 0.15in/sec RMS (which I also thought sounded low, but when it's vibrating at 0.20 in/sec for 5 seconds before it shuts down, you can really feel it in the floor)."

You should be doing spectrum analysis in velocity (in/sec RMS) in frequency range 3-1000 Hz to match permanent sensors to indicate frequency(s) that are present when overall level is exceeded. If the floor is shaking, then analyze the 3-200 Hz frequency range. The acceleration spectrum in the original posting is not very useful (useless) to answer/diagnose these two vibration symptoms.

Walt