Beating / modulating noise and vibration 2

[jeyaselvan]

I have a drive train comprising of a motor (4 Pole 3 phase induction motor, 50Hz), resilient coupling and a compressor (screw compressor- 4 male lobes / 5 female lobe combination). These compressors make a strong modulating / beating noise, which gives a perceptional feel that something is wrong with the machine. I tried measuring noise as well as vibration data on the machine. Both the data indicate strong sidebands of 1.6Hz, which we hear as beating frequency. I have attached the measured vibration data for reference.

Since this machine is a direct drive machine (no gears), the rotational inertia on the driven side (compressor) is around 30% lesser compared to the regular machines. This is the only difference I could see it from the design point of view. Or is the absence of gears in this machine is by itself does not have the flywheel kind of effect, which I am not sure.

The sidebands are there in most of the response frequencies. The response frequencies are all multiples of the male lobe frequencies, as expected in a typical positive displacement compressor.

I also did a motor current signature analysis (MCSA) measurements and observed that the data is fine, with only a 30dB of 250Hz (5th harmonic) component in addition to the line frequency of 50Hz. I suspected whether I could get some side bands related to pole passing frequency, but I didn't get it.

Kindly looking for your valuable suggestions & views in this problem.

Regards & Thanks in Advance
Jeyaselvan

 

[electricpete]

Here is the attachment that should have accompanied my previous post.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[jeyaselvan]

Thanks electripete for your reply
The torsional inertia of the complete drive train as below
J1=0.93/4; % Motor mass inertia in kgm^2
J2=0.012/2; % Coupling motor side(drive side)
J3=0.012/2; % Coupling airend side(driven side)
J4=3.69/9.81/4; % kgm^2, Airend mass inertia
k1=18240*9.81; % Nm/rad,MOTOR'S torsional STIFFNESS
k2=11913; % Nm/rad, Coupling torsional stiffness
k3=288245;% Nm/rad, Compressor shaft torsional stiffness

Thanks for your reference on the sidebands arising out of soft foot for the blower. I did carried out some bump tests, but could find the first natural frequency at 16Hz.

Apart from soft foot, " is there any other dynamics that could induce low modulation frequencies"?

Kindly for your suggestions.

Regards
Jey

 

[electricpete]

Thanks. Now I have some homework to do...

One other thing to check: is there evidence of oscillation at this frequency in the output pressure or flow indications?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[jeyaselvan]

I did measure the pressure to check for process oscillations, but I do not see the modulating pattern with the same. Hence eliminating one more possible cause.

My suspect is now on the rotor rub. Since when I try to freeze TWF for two revolutions, I could see four impacts per revolution corresponding to the lobe numbers ( 4 male lobes / 5 female flutes).

HOW DOES RUBBING RESULT IN LOW FREQUENCY MODULATION? This is what I am trying to answer this. I am looking for mathematical models / papers. Kindly for your suggestions and reference.

Thanks in advance

Jeyaselvan

 

[TPL]

As Greg L pointed out, beating is a result of 2 components of 2 components of very close - the beat frequency is given by f1-f2.

Induction motor noise in the form of beating, commonly arises from the interaction between mechanical vibration frequencies and line frequency. For a 4-pole motor, this interaction is between line frequency and twice actual running speed frequency.

4 pole 50Hz motor
Nominal Speed = 1500rpm = 25Hz
Actual speed is load dependant

Quote:For your query on slip frequency, this is 0.6333Hz. The speed is at 1462 rpm against synchronus speed of 1500rpm ( 50 Hz 4 Pole). I am not able to find where the 1.6Hz in coming from Unquote

Actual speed = 1462rpm – 2X = 48.73Hz - dS =1.3Hz

Where dS = difference between 2X running speed and line frequency

Quote:I have some new information from my latest test. When I couple the compressor to a slightly larger motor (larger GD^2, 30% higher), the modulation frequency gets dropped from 1.6Hz to 1.0Hz and audible irritation of modulating noise has also come down(not fully, but tolerable). I am not able to explain this. Unquote: –How about bigger motor, working less hard, higher actual running speed and lower slip, lower dS

I would suggest that before going off chasing esoteric hypotheses, you obtain some high resolution data (I further suggest a spectrum with a span of 200Hz and 3200 lines): note that the time taken to collect one spectrum of this nature will be quite long, so you need patience. It would be useful if you could collect several successive spectra and plot them in a waterfall format allowing you to see which components are present.
Beating is usually more of a nuisance than an actual problem, since operators complain about it. You need to quantify the vibration and compare to a standard in order to assess the severity of the problem in terms of potential to shorten machine life.

You can spend a lot of money and achieve little in these situations

 

[electricpete]

I agree TPL’s comments about resolution, double-checking pole pass frequency, and considering effect of alternate motor nameplate speed upon slip speed... in fact have made all those comments myself (as have others).

But TPL’s calculation of the frequency of interest (1.3hz) does not seem relevant to me. 2x 1462rpm running speed would be 1.3hz away from 50hz line frequency.... but who cares.... line frequency is not present in vibration spectrum of an induction motor.

What is usually present in vibration is not 1*line frequency, but 2*line frequency. The interval of interest for adjacent peaks is pole pass frequency:
FP = (1500-1462)*4 = 152 cpm = 2.5hz.

This FP = 2.5hz would be the spacing between 2*LF vibration peak and 4*Running speed peak on a 4-pole motor.

Additionally, in presence of rotor electromagnetic assymetry (dynamic eccentricity or rotor bar defect0, the pole pass frequency 2.5hz is also the frequency of sidebands around running speed in vibration and around line frequency in current. The 1462 rpm speed measurement is to small extent corroborated by mentioned 2.5 hz sidebands in current (14 Jul 10 9:31) which is very normal at a level around 45 or more db below fundamental since there is always some assymetry.

However, I was just noticing you mentioned 35 db difference (14 Jul 10 9:31), which is very much on the high side for these current sidebands (high sideband magnitude, low db down from main peak). Other places you mentioned 60db. Can you clarify or post current spectrum?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I agree TPL's comments about resolution, double-checking pole pass frequency, and considering effect of alternate motor nameplate speed upon slip speed... in fact have made all those comments myself (as have others). [emphasis added]

I apologize if the bolded part sounded like I was diminishing the value of TPL’s contribution. In a thread with as broad a discussion as this one, it is useful for anyone to frame the discussion of facts and possibilities in a way that better focuses on the aspects they feel are important. TPL’s opinion is certainly as valuable as anyone else here.


A few more random thoughts:
1 – are there any of these 2.5 hz sideband around running speed? If so, it would be interesting to look at the shaft under a strobe tuned to running speed.
2 – Is this machine fed from vfd by any chance?


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[jeyaselvan]

Thanks TPL & Electripete for your feedback.

I will post the MCSA & vibration data & waterfall plots after making high resolution measurements by the end of this week.

TPL : I understand beating is more of a perception issue than the real issue, but this is critical for me, since if not addressed, this problem has a significant influence on the business and hence I am trying to address this. I want to make one point more clearer, the problem is MODULATION (with sidebands on either side of response frequencies) technically and not BEATING.(ofcourse, the perception & TWF for both the cases are the same, but not the spectrum. Kindly ignore this, if this is clarified apriori).

This modulation problem is there with most of our machines which have this particular type of compressor (am an OEM). Just to clarify that the problem is not for one particular machine, it is there for the whole family which uses this type of compressor.

However, on checking with our another family of compressors I do not see the kind the rubbing which I have posted on 3Aug 10 1:08. Hence my approach tends to converge more towards the compressor than the motor (Also, all the tests are done with new motors). The absolute vibration levels are 3 times higher than our regular machines (4.5g against 1.5g).

Thanks electripete : 1) I had plans to strobe the machine. Since I don't have one in hand, I will try this thorough external means. With 0.16Hz resolution for most of my data, I see the side band at 1.6Hz from vibn data. For MCSA, I used an Agilent scope in addition to recording through DAQ/Labview (I need to check the dynamic range of my DAQ card).

2) This machine does not have a VFD. I have plans to connect this through one to understand the sensitivity to operating speed (+/-20%).

Will post on the developments.

Thanks & Regards
Jeyaselvan

 

[TPL]

The modulation or beat is not obvious from your data - I think you have used some artistic licence to enhance that line of thought.

The waveform plot suggests that your quote of 4.5g is in peak terms - although you say this is higher than typical, 4.5g peak isn't really high - I wouldn't normally have any concerns until levels exceeded 6g.

The acceleration data has value, but can you collect the same data in velocity format?

Rubs generally don't last for very long - they either clear themselves by wearing between the contact surfaces or cause a bit of a bang. The nature of a rub means that its vibration characteristics are not very stable, showing significant changes with time. What are you basing the idea of a rub on? The waveform data doesn't really look like a rub, appearing to be more of an impulsive nature, such as that which might result from a pressure pulsation - can you get a pressure transducer into the suction and discharge line?. Any odd valves or piping arrangements that might explain your concerns? Is this compressor fitted with a slide valve that might be misbehaving?

I would suggest that if your concerns are related to your customer being bothered by noise, that you address this issue directly and follow the advice given by TMoose and carry out analysis of the noise.

You have lots of suspicions and concerns, all of which appear to be supported by assumptions rather than data. No harm in carrying out a precision alignment (and checking/eliminating softfoot) - check for piping strain as well.

Sorry to sound negative, but I do think you've missed something here. This type of compressor is inherently noisy in terms of vibration and sound.

 

[jeyaselvan]

Thanks TPL for your valuable comments.

I will post the vibration velocity data after measurements. Will you kindly look into the attachment in my original post, where I have posted both the TWF and spectrum and some zoomed portions of spectrum with sidebands to make my point on modulation over beating. Please correct me, if my representation of the problem as modulation over beating is incorrect.

I was only comparing the acceleration values with machines, which do not have this modulation problem, since I suspect the modulation problem with all machines using this family of airends (terminology used to represent the compressor elements-for those not in compressor industry).

Wish to inform you kindly, all the assumptions (atleast most of them) are validated with measurements and the tests are all carried out to support my observations based on the attached PDPC chart, which I normally use for specific design & development related problems. I have made supporting measurements for nullifying each and every hypothesis in the chart that possibly could result in the problem I am working on.

I have also enclosed the pressure measurements made on the discharge of the compressor. I am aware that this sensor has a dynamic response only until 250Hz, however used to see whether any kindly of low freq modulation is induced by the downstream. This compressor does not have a slide valve.

Thanks again. I am determined to see to it, probably to the end of it.

Regards
Jeyaselvan

 

[TPL]

From that 1st waveform plot, you have 14 cycles of vibration in 9 seconds - 1 cycle takes 0.64 seconds corresponding to 38 cycles per minute, which is the slip.

You put a bigger motor in place and the problem reduces - do you have data similar to that shown to suggest that the slip changes?

Is it possible that the 1.6 Hz sidebands are irrelevant to this?

If it is the noise that is the concern, are you sure that the making vibration measurements is the right way to go? I would have thought that characterising the noise issues would have preceded, and driven the vibration measurements.

 

[jeyaselvan]

Thanks TPL for your comments.

Is it not the other way round for beat freq calculation?
1 cycle takes 0.64secs. Hence beating cycle frequency 1/0.64=1.6Hz.

You may kindly find attached the data with the larger size motor, with which the modulation changed to 1.1Hz (from 1.6Hz with orginal motor).

In fact all my analysis started with microphone measurements (because the complaint was primarily on noise & to ensure relaibility). The spectrum is a lot distorted, probably by the radiation effcienices of the housings.(Nevertheless the modulation is also observed from TWF of mic data as well, which i have attached in the last slide). When similar modulation is observed in vibration data, then all further analysis are carried out with vibration data. (I do have a mic meas at every test I carried out).

Regards
Jeyaselvan

 

[TPL]

Apologies for that- I was very sloppy. Still, just goes to show that you should always get someone to check your sums.

From your latest post QuoteNo modulation as seen in TWF, but spectrum shows 1.1Hz sideband ( may be low modulation index, hence not seen in TWF. Not audible as well)Unquote

Not audible as well????? doesn't all that just mean that the 'problem'just isn't there?

In the absence of anything other than a noise that your customer complains about, isn't the answer simply to use the larger motor?

What was the actual speed of the larger motor? Is it possible (with a slip if 38rpm) that the orginal smaller motor was simply a little undersized and that the noise is just its way of protesting?

 

[jeyaselvan]

Thanks TPL for your comments & sorry for pointing the typo in your calculations.In fact, for a moment, I was stunned whether I was wrong from the day 1. Thanks again for confirming my calculations.

Since this is for a OEM product, I may not have the option to use a higher motor size, since these are manufactured in larger nos. Your point on undersized motor could be one possible cause, because quite often these motors (quite common to load upto 1.1SF on a 1.15SF motor)are loaded to their throat! The larger motor runs at 1483RPM (since we are not loading the motor to its rated load, being a larger size motor)against the nameplate rated speed of 1471rpm.

Having spent quite a good amount of time on this, I am more interested to know about the source of modulation. The areas I could think of with the larger motor could be
i) the effect of inertia ratios of the drive and driven side of a drive train: but why would this influence the modulation ?
ii) In the event of rub (usually these compressors contact at the pitch point, specifically pitch line, but if not? ), the larger motor has more accelerating torque and hence can quickly (1.1Hz) pass across the rub while with the normal sized motor this is quite hard for the motor to provide torque to overcome the rub (1.6Hz). I will look for this on tear down of this machine after my measurements.

Still, I will reserve the "inertia effect" for the last option, if I am not able to make progress in my diagnosis. Atleast, I will have a containment for the problem until I unserstand the cause.

Regards
Jeyaselvan

 

[Strong]

Regarding possible torsional vibration:
a) Motor current spectrum analysis is a good method for detecting mechanical torsional vibrations
b) Have you tried to stop oil injection momemtarily to measure if modulation stops?
c) How about adding a trial flywheel mass to cooling fan end of motor to simulate the equivalant inertia of the larger motor that does not have modulation?

Walt

 

[jeyaselvan]

To update the forum, the cause for beating was found to be with the twice line frequency(99.1Hz) and the excitation frequency (being 4 lobed compressor 1462/60*4=97.5). This was established through operations at multiple speeds with a VFD. By shifting the speed by 200 RPM, the beating vanished away. Thanks to all of your who have contributed to this thread.

Regards
Jeyaselvan