Anomaly in drive point FRF 1

[GMarsh]

Hi,

I am doing modal testing on a casing (rigid base with thin wall at top) fixed on the base with 8 bolts. Roving hammer test.

When I am looking at my drive point FRF (acquired on thin wall at the top) I am finding two anomalies:

1. In the imaginary part of FRF, two modes are appearing on other side. Usually we expect all modes of FRF in imaginary plot to lie on one side. Then what could be the reason for this peculiar behaviour? Does it indicate wrong response / excitation point or something else ?

2. No anti-resonance found after first mode. Usually we expect one anti-resonance after resonance peak in drive point FRF. While I have anti-resonances after all resonances, first mode has none.

For those of you who wish to see these problems graphically, I am attaching word document with figures.

Apart from above issues, any other general comments on the FRF behaviour also would be greatly useful as I am in learning phase.

Thanks and regards
Geoff

 

[GregLocock]

The no antiresonance issue is something to do with residual damping of the lower mode holding the response clear of the origin, on the Nyquist plot. Not uncommon, your curve fitter should be able to manage this.

The first problem with the phase on the wrong side of the origin is also common, but it means that you need to improve your test technique probably (or something else). My guess would be that your hammer and accelerometer are not perfectly in line. Try using an impedance head instead of a separate load cell and acc.

How is your reciprocity?








Cheers

Greg Locock


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

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

 

[GregLocock]

Incidentally you have two close coupled modes at 1400 Hz, that often causes drift away from the origin, hence killing the antiresonance.

Note that the characteristic deep valley at antiresonance is an artefact of the bode plot, on the more sensible nyquist plot it is a relatively minor feature.



Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Thank you very much for this reply. I am counting on you while posting this!!

I will definitely check reciprocity. I have not done till now as I was baffled when I saw two peaks on the other side in Imaginary!! Anyway now I am reassured and will take care while impacting at drive point. I don't have impedance head.

I have one more question:

When I plot the modulus shown above with log frequency scale, I can see lot of spurious modes before the 1st mode which are basically disturbing the stiffness line (I think so). Considering the phase change it looks they are some modes (pseudo, at least). When I run FE analysis for the same structure, I can notice rigid body movements of casing though it is fixed with 8 bolts. Do you think those rigid body modes are hiding in these initial modes before 1st mode ? I am attaching the figure here. (Please ignore the y-axis of phase plot - software has some bug).

Many thanks and kind regards
Geoff

 

[GregLocock]

The rubbish below 200 Hz is just noisy stuff, probably low s/n ratio, but just possibly insufficient resolution. It won't upset your first proper mode at 1 kHz.

Worryingly everything above 4 kHz also looks like noisy rubbish.

plot your input and output spectra, they may contain some clues.

Also your coherence may help you to decide where else you have problems.

Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Thanks again for your reply.

The modes above 4kHz are not rubbish. They are genuine modes. Due to log nature of frequency scale, they are cluttered.

I have plotted coherence also. Please find attached the picture with coherence overlaid on modulus. Do you think it is acceptable? At present I am taking 4 averages at each point.

Many thanks.
Geoff

 

[GMarsh]

I did reciprocity check also. Please find attached the picture. Please note that green colour curve corresponds to outer axis.

While the frequencies are matching almost, the amplitude magnitude is not matching. Is it ok to proceed? Partly the reason for low magnitude in one curve is I gave less force at that accelerometer as it is a bit more sensitive than the other one and I am ending up with overload error if higher force is given.

Thank you.

Kind regards
Geoff

 

[GregLocock]

Coherence should not be less than 0.9 at a frequency of interest, and your plot looks rather odd to me. Normally the coherence is good on the shoulders of peaks, and falls off sharply at resonance, due to phase scatter, due to limited resolution, and at the antiresonances due to poor s/n. Your coherence plot looks nothing like that.

The phase in that plot is also all over the place, is it a driving point function? If not that's ok, but you are running into speed of vibration issues (the basic rising trend of phase), which is interesting.

Your transfer function should not be affected by force level, in a linear system. Reciprocity should give exactly the same transfer function for each test, both modulus and phase, in a linear system. So no, your reciprocity isn't much good.

Howver, you need to investigate the force linearity of your system. If it is not linear with force then test technique becomes vitally important, obviously.

Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Many thanks for your response.

Yes. I can now see that my data is not well. I read Ewins text on these issues and planning to do following modifications:

1. Testing with finer frequency resolution - Though this doesn't seem to be a problem as Nyquist plot is showing smooth circle for all frequencies. But loss of coherence is mainly attributed to poor frequency resolution with lightly damped structures. As my structure is a thin cylinder, being lightly damped, this is the only reason I can see as of now for poor coherence. But for this to happen there should be a drop in coherence at resonance. This is not so in my case !! Really confused on this.

2. I am also sure that my structure is behaving non-linearly. As I looked at all other data which I have. The response is changing depending on the excitation load I give. Moreover some modes are getting excited in some cases and not in other, though the excitation is being applied at same node. So according to Ewins book, a quick check is to plot real part of inverse of FRF and I should see a linear trend. If not that confirms the non-linearity. And I may have to look at solutions suggested such as response control, etc. What is your suggestion on this ?


As you rightly noticed my signal phase also is not good. Is this what you are meaning as phase scatter ? Can you slightly elaborate what this phase scatter is ?

Thanks a lot Greg. Discussion with you always is a great learning experience.

Kind regards
Geoff

 

[GregLocock]

Phase scatter is the error in phase estimation at the peaks of lightly damped resonances. Coherence is basically a measure of phase scatter. Increasing the resolution is the usual cure, but it can also be caused by force softening, if the input impedance of the exciter is too high the structure is unable to 'accept' as much force at that frequency as it should. So you might find a smaller hammer helps.

My terminology may be out of whack with Ewins, it's been a long time since I did a modal!



Cheers

Greg Locock


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

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