Occaisionally I get a spectra like this from one of my AHUs:
Motor 1706 .01"@sec
Fan 1660 .01"@sec
Belt freqs = 479, 957, 1436, 1915, 2393, 2872.

no appreciable amplitudes for belts

Spectral peak 2670.1 cpm (1.608 orders), .3329" @sec with 2-3 Harmonics

I have ruled out resonance since resonance does not have Harmonics.. nor do resonance occur where there is not exitation force.

1.6x is sub bearing freq.

2670.1 is not a higher harmonic of a lower freq, none could be found even on the logarythmic scale

Obviously the pattern suggests looseness... but looseness of what?
I am considering that my basic speed assumptions may be wrong. (used a strobe on the motor)

Any other suggestions?

1706 rpm would be an oddball frequency for a motor unless there is a variable speed drive (not common with a belt drive machine).

As you know, typical motor speeds in 60hz land are just below 3600rpm, 1800 rpm, 1200 rpm etc and in 50 hz land 3000rpm, 1500 rpm, 1000 rpm etc.

If this is a 4-pole 60hz machine with 1800 syncronous speed, 1706 would be more than 5% slip below syncronous. That is more than I have ever heard of. NEMA limits slip to 5% at full load and it is typically much less.

Martin,

Of course, speeds should be verified. Is motor VFD driven?

But what if 2607.1 cpm is the 4th harmonics of the shaft whirl frequency which is, most of the time, between 0.39x and 0.4x.
Here is the arithmetics
0.395 x 1660 = 655.7
2607.1 / 655.7 = 3.98, which is almost 4x

I know even if whirl does take place, why only 4x, 8x, 12x of whirl frequency are present?

Just a wild guess...

David

When I find odd freq´s I try to look for some noise or friction that can excite a resonance, belts are friction noise generating squeeky things that could supply that, why it then would be multiples, beats me. Just a suggestion. Olov

Pete:
Believe it or not, we have a lot of belt driven AHUs run through VFD. (most a 60hz!) Please don't ask me why.. you don't have that much time...

I could not find a VFD in the vicinity, but that doesn't mean it is not tucked away in a corner of the building.

Anyway, the things you said are the very reason I am questioning the RPM.

David: The peak in question is definately no harmonic of something else.

Oli: I too was considering resonance, but I have never heard of a resonance with Harmonics. But that does not mean that no one has ever seen one.

So I ask:
Has anyone ever seen a resonance with apparent Harmonics?

I think that harmonics of resonance exist.

An example would be the chiming of a guitar string. If you chime it at one half the length you get one note and if you do it at 1/4 or 3/4 of the length you get another but an octave higher or lower.

This seems to be harmonics of resonance, but I could be wrong. Maybe someone will clear that up for us.

Ironically, as I read this post, I was sitting here listening to some recordings of Robert Johnsons' Delta Blues..

This analogy is a good one, and in fact I play guitar and frequently use what I know about music and apply it to analysis.

There is no question that harmonics mathmatically exist, and that they are available for exitation. The question is whether it actually happens.

String Musicians know about "sympathetic harmonics" that is, when you strike a string, if there is a free string that happens to be tuned to the same note but a different key, it may vibrate.

The reason it "may" vibrate is one of energy. How much energy did the original string impart to the instrument, and was it enough to excite the other string. Sometime yes, sometimes no.

Most of the time this "sympathetic Harmonic" is barely audible. I suppose it is a matter of amplitude, a matter of degree.

But considering that a stringed instrument is specifically designed to do this, I would think that this occuring in any random structure would be virtually nil.

There have been a few times in my career where I've thought that something was highly unlikely only to have it turn out to be the root cause of the problem.

If you suspect a resonance then try to confirm it with a bump test, variable speed exciter, or by changing mass or stiffness. If you confirm the resonance then you have answered your question about harmonics of resonance.

It would take a system that is virtually free of damping such as a guitar string

Keep an open mind as you investigate this. Some basic fundamentals might be helpful here.
First, if there's any reason for the system to become non-linear, the waveform will distort from pure sinusoidal, and there'll be harmonics. The most likely non-linearity would likely be an increase or decrease of the "spring stiffness" as amplitude increases.
Second, you do not need excitation directly at the natural frequency (ie a resonance) to see a natural frequency of a system. If the excitation is changing (eg frequency wobbling a bit; response large enough to get bumped a bit, etc) then there's no single steady-state response. The system will always be trying to find a new steady state. Such changes (ie transients) will always occur at system natural frequencies. So you'll see them in the response.

To Add to Martin Moore's - ours is a small new facility and the AHUs are belt driven with VFDs. When you measure and record vibrations, are you taking in consideration the possible change is speed between measurements, particularly for trending purposes, if yes by which method?. From the SCADA system I can observe that the speed of motor is not steady but oscillates a bit. cheers

Regarding Frequency Drift: This is a newly monitored system, so I don't have much data to determine frequency drift over long periods of time.

However what was said about a non-sinusoidal waveform is true in this case. Typically when viewed in dispalcement, you can see a near sinusoid at turning speed.

But in this case each peak is a different distance from the next. I am aware of the difficulties of using displacement derived from acceleration, but I have also viewed a lot of waveforms in displacement, and I am used to seeing a fairly sinusoidal wf at turning speed.