This is an I-R screw air compressor (150 hp motor, 1785 rpm). Compressor data (# of lobes, coupling ratio, timing gear ratio) is unfortunally unknown.

The plot is attached and looking at the pattern, amplitudes, and frequencies ( possibly 8 male lobes, 1.5:1 coupling ratio) it is possibly flow/pressure pulsation frequencies.

If this is correct, what kind of flow related issues would cause it? Pressure is normal, no flow readings or pressure vs. flow curve are available...

Thanks,
David

Air_compr3_flow.doc (166 Kb, 73 downloads)

I have quite a few screw compressors that I look at. I am not sure that what you are seeing is flow related. Most of them have a pressure pulse around 4 times turning speed, but I have not seen that with the distinct sidebands that you have.
It would be nice if you could get a little better idea of the speed of the screws. If you take high resolution data for electrical defects you may be able to pull out the speed of the screws with that. Look for a peak with harmonics above motor turning speed.
Is it possible that this compressor is on an inverter and that the sidebands you are showing could be turning speed?

David,

This is an answer that I posted to a previous thread. It may help you to identify the lobe pass frequency, but figure you've probably already gone down this route:

"First identify the largest amplitude harmonics in your velocity spectra. Chances are that these will be multiples of your lobe pass (or meshing) frequency. All the compressors that I've worked on have 4 lobes on the male rotor and six on the female rotor. So divide the lobe pass frequency by these numbers to get the shaft speeds. If the calculated frequencies match actual peaks in your spectra, then you're probably on to a winner. This method works on the compressors that I check and has been verified by calculating out the known data from the manufacturer. Of course if your compressors have a different design, it may not work. I've even been able to identify the gearmesh frequency on some compressors by looking for a high frequency peak with known sidebands."

If the sidebands in your spectrum turn out to be RPM of one of the screws, then you may have a screw contacting the casing.

Al, I'm not quite clear on the math ("divide lobe pass frequency by these numbers to get the shaft speeds"). I've attached a spectrum for an ammonia compressor. It has 4 male and 6 female lobes. The largest peak is @ 8x. Would you run thru your calculations on this particular spectrum so I can see what you're talking about?
I'm assuming this can be calculated with compressors other than air compressors. Please correct me if I'm wrong.....

Thanks,

Doc2.doc (32 Kb, 18 downloads)

Sharkey,

I hope I’ve got the math right on this. I’m a cpm guy and Hz just confuse me, but that’s another thread which has been visited many times before.

If we take your spectrum as an example, the single biggest peak (the one with the cursor on it) is at 476.76Hz. This looks to me to be the 1st harmonic (2x) of 238.38Hz, which I would assume is the lobe pass frequency. If we divide this by 4 we get 59.595Hz or 3575.7cpm, which should be the male rotor speed. If we divide by 6 we get 39.73Hz or 2383.8cpm which should be the female rotor speed. Do any of the lower frequency peaks in your spectrum match these frequencies?

The compressors that I work on have a set of gears between the motor and the male rotor. If the number of teeth is unknown, then sometimes it is also possible to work out the gearmesh frequency from your spectra. Just look for a peak usually beyond the lobe frequency harmonics, which has sidebands of either the I/P (motor) or O/P (male rotor) speed. This peak might be very small compared to the lobe frequency harmonics and difficult to identify.

I once had a problem with a screw compressor, but according to the data provided by the manufacturer, none of the theoretical frequencies matched what I was seeing. I used the method above to work out the lobe frequency, the rotor speeds and then have a stab at the gearmesh. I then phoned them back and asked if the # gear teeth could be different to what they had stated and even gave them the # teeth that I suspected. They said that they do supply a compressor with teeth at that ratio, but insisted that this compressor had a different ratio. When the compressor was pulled out of service, I counted the gear teeth myself and found that I was right.

Hope this makes sense.

Al,
It makes sense to me, however, I don't think I'm capturing the data I need. How do you set up your points (fmax, etc.) on your compressors?Thanks in advance....

motor_speed.doc (58 Kb, 17 downloads)

Let me rephrase my last post. I believe I am capturing the right amount of data. My question is ... on screw ammonia compressors (Frick), when should you be concerned of a gas pulse amplitude? And are high gp amplitudes necessarily an indication of other problems with the compressor? Because of the "noisy" nature of these machines, I'm having trouble honing in on whether the motor,compressor or separator is the source of the problem. I have not had any failures , at this point, in which to draw experience from. Amplitudes vary according to load, weather conditions,etc., so it's difficult to accurately trend these machines. One month , the compressor looks like it's gonna blow up, the next month, it looks like it was just taken out of the box.Is anyone out there familiar with these that can help enlighten me??

Sorry, can't help you with Frick ammonia compressors. I'm used to Ingersoll Rand and Atlas Copco air compressors. I've no idea how much these may differ. Normally I collect velocity spectra at a frequency high enough to capture the gearmesh, but also at a resolution capable of picking up sub-synch harmonics of the rotors. I'm told that 1/3 rpm harmonics are a problem to watch out for. I also take demodulated acceleration spectra on each brg.

When I view the data I have an overall trend, current spectrum and waterfall plot (last 12 months)on the screen simultaneously and I'm looking for anything that indicates a change in normal running. One compressor can run differently to the next, so I just look for change in each individual. I try to collect data on a weekly basis, because I've seen these fail fast. Using this method I've made several good calls on compressors.