Reasons for phase drift

[GMarsh]

Hi,

I am noticing phase drift in the FRF acquired on a cylindrical thin shell. When I acquire a vibration signal during machining of that structure, the time domain signal also shows a phase drift.

Please find attached snapshot of the FRF and vibration time domain signal.

Can someone suggest what might be possible causes for such phase drift and if possible how to avoid them. I made a survey on this in Google but found no useful info w.r.t my problem.

Thank you.

Regards
Geoff

P.S: Greg had a look at this FRF some weeks back and electricpete recently commented in another thread on phase drift on vibration signal. I am trying to correlate both, if possible.

 

[GregLocock]

you've got travelling waves as well as standing waves, if that data is correct.

Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Can you be a bit more descriptive of what you meant by presence of travelling and standing waves ?

So is it the nature of structure which I can't do anything. But this phase drift is giving me that dominant modes of 4870Hz and other peaks looking like its harmonics. This is a bit of concern as long as I don't understand about it.

Thank you.

Kind regards
Geoff

 

[GregLocock]

Forget about modes. The vibrations take a finite time to travel through the medium, so there is an increasing phase lag the further you get from the excitation, and the higher the frequency. The actual speed depends on the mode of transmission and the material properties.

However that phase trend can be caused by poor data processing or possibly even a bad setup or instrumentation.

How does your reciprocity look?

Cheers

Greg Locock


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

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

 

[Strong]

Testing a thin cylindrical shell may be affected by mass loading of accelerometer and damping variation by location from accelerometer mounting.

Walt

 

[GMarsh]

Greg - thank you. Reciprocity is good.

Considering that we see phase drift in both FRF and acquired vibration signal, don't you think this is something to do with structural behaviour or boundary condition non-linearity, etc. ?

Strong - Thank you for the response. The mass of thin shell is roughly 4.5kgs and the mass of accelerometer is 0.2 gms, very small accelerometer size - 2.8 X 8.6 X 4.1 mm. I don't think any mass loading effect will have this much phase drift.

Kind regards
Geoff

 

[Strong]

Geoff,

You may be OK on mass loading with the tiny accelerometer, but you really need to compare surface mass of structure to accelerometer and not the total mass. There could still ba an issue considering the high frequency at 4800 Hz. I use a microphone for mass-less load when in doubt to see if frequency changes.

Walt

 

[GregLocock]

Yes, a travelling wave is terminated by an absorber (or infinity), so the boundary conditions are vitally relevant.


Cheers

Greg Locock


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

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

 

[GMarsh]

Strong - What do you mean by 'surface mass'? Inside surface area of thin shell ( only modally active area) is 11.5 e-2 m2 and surface area of accelerometer is 24 e-6 m2.

The mass of 4.5kgs is only modally active mass. Actual mass of whole casing with bottom base is 86kg.


Greg - Can you suggest some good book which I can refer to understand this phase drift, travelling wave concepts, etc.

 

[Strong]

Surface mass = modally active mass, so your tiny accelerometer should have no mass loading effect on the structure.

Walt

 

[Strong]

Geoff,

If you Google: "phase drift from traveling wave" you will get a lot of references to your original question.

Walt

 

[GregLocock]

I have no refs for them, we first came across them when we were doing acoustic modals of cars.

Cheers

Greg Locock


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

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

 

[MikeyP]

What happens if you measure the driving point FRF? If you still see the phase drift then there is something wrong with your hardware setup and/or signal processing.

M

--
Dr Michael F Platten

 

[hacksaw]

just a follow up comment, impulse tests in distributed systems always exhibit phase shifts with propagation delay

 

[electricpete]

If you have ideal linear model of distributed system, speed of propogation of a wave depends on only material properties density and young's modulus and not on frequency. Some materials conform closer to the ideal and others vary far from the ideal. The characteristic where different frequencies travel at different speeds is called "dispersive" media.

=====================================
(2B)+(2B)' ?

 

[GregLocock]

Common dispersive mechanisms (it isn't a material property necessarily) are bending waves in steel, for example, and of course pressure waves in air.



Cheers

Greg Locock


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

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

 

[electricpete]

For longitudinal waves, there is a simple formula v=sqrt(E/rho) depending on material properties E and rho, frequency nowhere in sight.

I looked up Rao derivation of transverse resonant vibration of long-thin beam, it again relies on the same equation v = sqrt(E/rho) as shown in Rao’s “Mechanical Vibrations” 3rd Ed Equation 8.50.

If there is a frequency dependent effect not related to material properties, it does not show up in either of the above simple cases. Is it maybe related to longitudinal vibration where the beam cannot be considered long and thin?


=====================================
(2B)+(2B)' ?

 

[hacksaw]

at the risk of hijacking the original discussion you can have longitudinal waves in bodies of infinite extent, and psuedo-longitudinal waves in plates and bars, and various shear waves.

the phase velocities are frequency independent in the linear case, but as soon as you deal with cylinders, the wave number solutions are not linear and an entirely separate discussion is required

the phase velocity between the excitation and sensor give rise to phase shifts.

these are evident when impulse tests are used as time delay or in phase shifts in the frequency domain.

 

[GregLocock]

Sorry presure waves in air aren't dispersive, to any useful extent. Brain fade.


Cheers

Greg Locock


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

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

 

[GMarsh]

Hi all,

Thanks to everyone for their useful comments. Sorry for not being immediate in response as I was fully drowned in sorting out the same problem - phase drift, of course jointly with the supplier of modal kit.

Found the culprit - Ground loop !

I think MikeyP answer is SPOT ON - if drive point FRF is showing phase drift, I think we have to look at hardware.

For benefit of whosoever reads this thread in future, here is summary:

I am testing with the structure on a machine tool bed (as I showed in one of the pictures earlier). Now it is so found that our workshop ground has some leakage which I never realised. Being not that knowledgeable in reading phase of signal I continued my experiments. Even for drive point FRFs phase was scattered everywhere, coherence at resonance peak is drooping in unexplainable way, so on so forth.

When I kept that structure outside machine bed and tested, then no problem ! So again when tested on machine, as soon as hammer touches (no impact), a spike can be seen and when hammer leaves, another spike again. Also when a plastic tip is used, no problem. As my work needs high frequency range, I used metallic tip all the time.

Another important symptom is: the magnitude of force while impacting is always coming negative ! That is you hit anywhere on machine, you see -100 N, etc.

Final solution: place a small piece of insulation tape or masking tape on node that is going to be impacted, we see a positive force spike, good phase, good coherence. Only one problem is in phase we get noise kind of oscillations at non-resonance region. As I thought it is not that critical, I am continuing with it.

All this problem is because the hammer is not electrically isolated, while accelerometer does. It seems this is the way they are designed.

The supplier of modal kit said the amplifier inside hammer is getting overloaded, due to ground loop.

Many thanks to all the participants in this thread.

Kind regards
Geoff