Mode shapes in experimental modal analysis 2

[GMarsh]

Hi,

I am doing experimental modal analysis of a cylindrical casing. I have 64 FRFs. I have curve fitted dominant modes and obtained a synthesised FRF. Now when I see the animation of mode shapes corresponding to each of the frequencies in FRF, I am finding a not-so-significant (amplitude wise) mode in FRF is having the highest vibration amplitude on relative scale.

To explain in detail a bit, I have nearly 65 modes in frequency range 1200 - 15000 Hz. As expected the initial bending modes are showing greater magnitude in FRF. But when I see mode shapes, a small mode at 4900 Hz is vibrating heavily. This is validated through experiment as well. In operational condition, 4900 Hz and its harmonics clearly stand out in FFT of vibration signal.

How to identify the strength of a mode from FRF ? And is it possible that a low amplitude mode in FRF has heavy vibrating amplitude in mode shape.

I am attaching the file with FRFs and mode shape.

Thank you in advance for your time and help.

Regards
Geoff

 

[GregLocock]

can you display the fft of points 64 and 45 in radial coordinates ?

I suspect you are just seeing an autoscaling artefact from your post processor

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 responding to my query. In fact when I posted this in another group, they referred me to this group and said if I am LUCKY enough, you will respond. Now I can thank my Luck and you too !!

I am attaching the word document with the FRFs you requested and also FRF at point 5 (which also is shown to have max amplitude in the mode shape I have shown earlier). Honestly, I also did not understand why these points are showing such unusual behaviour. When I searched literature I read that either improper measurement or poor curve fitting may cause this. But I am not sure where I did mistake. Can you kindly comment ?

Coming to the main question - Earlier I showed you the mode shape which was taken on a relative scale basis. That is, the ones with maximum & minimum amplitude are taken as reference for scaling all other modes (only for displaying). Unfortunately I cannot post such a huge video showing all modes. But initial modes which show very high amplitude in FRF - their mode shape amplitude is not at all significant when compared to few modes such as this 4910 Hz and 12400 Hz.

Generally does the mode having high amplitude in curve-fitted FRF shows high amplitude in mode shape also ? If not how to find this ?

Actually this is mode shape of a structure which will be machined using milling operation. When I took FFT of vibration signal measured during machining, the mode which shows high amplitude in mode shape is significantly standing out. We can see its harmonics as well. For your reference, I attached FFT of that vibration signal also where you can see peaks around 4900 Hz and its harmonics.

Sorry for writing such long mail. But just want to explain in detail. I request you indicate if I have interpreted something wrong.

Many thanks for your help.

Kind regards
Geoff

 

[GregLocock]

I'll have a think.

However first of all I'd say you will always get odd results when you do a modal on thin axisymmetric shells at high frequencies. The reason is that the tiny restraint offered at the driving point by the stinger, and the response point by the accelerometer, upsets the mode shapes. When looking at the kHz region I've had more luck with non contacting methods (lasers), and sand, than accelerometers.

You can see from your FRFs at the points of maximum displacement in the mode shape plot that the 4900 mode is not especially strong, so it looks like you have a curve fit problem or that your perception that "The boxed mode, 4900 Hz, is showing highest amplitude in mode shape relative to all other modes. Below I am giving the mode shape picture." is wrong in some way, perhaps you meant that the mode shown is strong in the operating condition, from looking at your waterfall plot.

So, if I ignore the mode shape animation, and assume that the FRF data and the operational data is all good, what seems to be happening is that a small peak in the FRF is the resonance that shows up most strongly in operation.

That's not terribly unusual, it just means that the excitation is stronger at that frequency than at others.

Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Thank you for your detailed response.

The operational condition basically involves an impact and then some free time. So the results in waterfall plot show the response of structure to an impact and subsequent idle period. Now there is nothing substantial to justify that the frequency of excitation impact is near to 4900 Hz.

Then I started looking at mode shapes closely on Relative scale. During animation of all modes, suddenly 4900 Hz mode starts vibrating greatly. And after that again all modes are normal. So I concluded that this is the reason for showing up of 4900 Hz mode during operational condition.

When I looked at curve-fitted FRF or individual FRF there is no way to tell that this (4900 Hz mode) could dominate all other modes. So I am just thinking whether amplitude of FRF is an indication of mode shape dominance or not?

I will also look into my curve-fitting procedure. If you get some logic behind this, please comment.

Thanks a lot for your response.

Kind regards
Geoff

 

[GregLocock]

I'm not at all familiar with that approach to analysis, I'm afraid I cannot understand what you think you have seen.

I agree an impact doesn't have a frequency, in which case your waterfall is yet another puzzle.

Cheers

Greg Locock


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

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

 

[GMarsh]

Hi Greg,

Sorry if I have confused you with my previous mail. Please forget the operational response. I just showed it to reinforce the fact that 4900Hz is dominant.

I just wanted to know one thing - Is there a direct correspondence between magnitude of a mode in FRF and mode shape amplitude (on a relative scale of all modes) ?

Intuitively it should be yes, but I just want to confirm.

Thank you very much for your support.

Kind regards
Geoff

 

[GregLocock]

No. typically an animated mode shape's deflection is autoscaled in some way by the postprocessor, hence what you see when you look at two different modes animated can be misleading.


Cheers

Greg Locock


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

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

 

[electricpete]

I'll add two cents that may or may not be obvious/relevant.

A mode shape (characteristic of the system) does not inherently have any magnitude, either absolute or on relative basis compared to other modes. We have many choices in normalization which may create appearance of relative magnitudes, but these are characteristics of the normalization choice, not the system.

An operating deflection shape (characteristic of system plus excitation) will typically include contributions from several modes. Which modes are excited tend to be a function of:
1 - how close is the excitation to the frequency of the mode (already discussed)
2 - to what extend does the "shape" of the excitation match the shape of the mode. For example applying excitation at a single location which is the the node of a modeshape, will not excite that modeshape. The most effective location to apply excitation to obtain maximum response of a modeshape is at the anti-node (peak) of the modeshape.


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

 

[GMarsh]

Hi,

Thanks to both of you.

The 2nd point of electricpete makes some meaning and stirred up some ideas for my further analysis. I am now thinking of looking at my excitation like point on the cylinder at which the impact was made and whether this point is close to antinode of 4900 Hz and also the natural frequency of tool which made this impact.

One last doubt - do you think the energy of excitation has a role to pick which mode gets excited ? I mean, if I give impact with lower magnitude, it will excite low frequencies and when I give impact with higher magnitude it will excite higher modes. Do you think this can be possible?

I ask this because I think lower frequencies are easily excited than the higher ones. Not quite sure though.

Thanks a lot for this useful discussion.

Kind regards
Geoff

 

[m2fm2]

My two cents,

The plots that you attached. Are they for a single axis? If so, then perhaps the mode dominates in another axis (then again, since it's cylindrical you probably performed the analysis in those coords rather than cartesian).

Also, do you have the mass fractions/modal mass for each mode? Sometimes a look at those can help evaluate the dominate modes.

cheers,
Ken

 

[GMarsh]

Hi Ken,

Thank you very much for your reply. All those two cents are adding up to a decent total!!

As you rightly predicted I did excitation and measurement only in radial axis.

Your suggestion of mass fraction is very much valid and helpful. When I do modal analysis in FE software, I use this to determine which mode might be strong and also to decide whether the number of modes I extracted are a sufficient representation of dynamics of structure.

But I don't know how to do the same for Experimental Modal Analysis. I am using ME'scope software in which I could not find any option of getting, Mass and Stiffness matrices. After curve fitting we get only Frequency, damping and residues.

A search in Google gives me some publications where the algorithm to extract mass and stiffness data for each mode is given. But as of now it looks difficult for me to code it. Any help / suggestion in this regard ?

Thank you,

Kind regards
Geoff

 

[m2fm2]

hi Geoff,

Glad to be of some help. But at this point I will have to bow out of the discussion. I am new to this lab and we do use ME'scope. However, I am not trained on the software yet so cannot comment on that. My background is(was) with NASTRAN.

Good luck!
Ken