Confused about damping 1

[mkoijn9]

I am a little confused about damping. To try and understand it I took a bar of aluminium stock and cut two lengths, one twice as long as the other. I then suspended them on elastics and attached an accelorometer and excited the bars in turn with an impact hammer.

After processing and curve fitting, the results I get are 0.0099% of critical @ 4056Hz for the short bar and 0.016% of critical @ 957Hz for the longer bar, Both results are for the first free/free bending mode.

somehow I was expecting the damping to be the same as the bars are exactly the same except for length. Where am I going wrong? If I want to compare damping ratios of different materials does this mean I have to have samples exactly the same dimensions?

Thanks

Albert

 

[electricpete]

http://books.google.com/books?id=Ef...=X&oi=book_result&resnum=3&ct=result#PPA78,M1

On page 78/79 he describes the resonant beam test. A disadvantage cited is that it gives you the damping only at the resonant frequency (apparently damping is a property that can change with frequency). Do you have two different resonant frequencies.

Later on page 80 he states that material damping depends on type of material, frequency, temperature and stress level. On page 81 he goes on to say that stress level is the most important factor. This would of course means damping can be non-linear. In that case if you strike with different force, or the mode creates different stress, the damping would be different.

By the way, this is just the result of a google search. About the only thing I am sure of when it comes to evaluation of material damping properties is that I know very little.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[mkoijn9]

Thanks ElectricPete, I did do a google search but didn't really find anything that explained it terms that I could understand! I will do some more searching. If anyone has a simple explanation tho. I would be very happy.

 

[hydtools]

How were the bars suspended? Where were the elastics attached? Node or anti-node? That may be the damping influence.

Ted

 

[hydtools]

Here is a 1965 NASA paper that may be of interest.

http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19650021096_1965021096.pdf

Ted

 

[Co]

What is the purpose of suspending them with elastics? Why not fixed-fixed, or fixed-free. ect?

 

[MikeyP]

Coppola, trying to constrain the system adds damping usually.

mkoijn9, what method did you use to estimate the damping? log dec? How many averages did you take? How many cycles did you use for the calc (if log dec)?

M

--
Dr Michael F Platten

 

[GregLocock]

In addition to the wisdom above

For all practical engineering purposes, unless you are designing musical instruments or similar, .010%==.016%

The reason is that as soon as you mount it to a real system the interface will add a great deal of damping - typically even a welded aluminium structure will have a damping 10x what you have seen, and if you have a bolted connection then it could easily be 30x greater.

As it is your test is not 'perfect' - as ep says the frequencies were different, so the precise proportion of air damping, vs damping in the suspension, vs internal damping, may be different.

Unless this result is a crucial part of your work, I'd just learn the lesson that measuring the damping of lightly damped structures is extremely difficult - you have done well to get a figure as low as that.

The purpose of the elastic suspension is to approximate a free-free constraint. In the world of modal testing 'fixed' usually isn't fixed, that is the compliance of the foundation modifies your resonant frequencies. To get good results with a fixed constraint we use 1 tonne surface plates to test 1 kg parts at 500 Hz, and the mounting brackets are a hideous bodge of improvements made over the years.




Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[mkoijn9]

Thank you all for your replys, The bars were first excited under a loudspeaker and sprinkled with glitter to find the approximate position of the node lines of the first free/free bending mode, then the bars were suspended on the elastics at those positions. I excited the bars with a small impact hammer and measured the response with an accelorometer( 10 hits). After calculating the frf's I curve fit them using the polymax algorithm.

Thanks Hydtools for the link to the paper.

I'm primarily intrested in measuring the damping of musical instrument wood, as the blocks always come in slightly different sizes, I was wondering If I can directly compare the damping estimates of the raw blocks, as the resonant frequency of the first free/free bending mode tends to differ among the blocks( usually between 300Hz and 600Hz) Should I size all the blocks to the same size? (This is a lot of work). Or should I size the blocks so that they all have the same first bending frequency? ( even more work)

I chose the aluminium bar for the experiment as I thought the material should be fairly homogenous.

Many thanks again

Albert

 

[GregLocock]

Have you considered that polymax may not be ideal for these low damping figures, and that the ratio of the frequency resolution to the resonant frequency probably affects the result in practice? perhaps LMS could give you some advice?

Polyref's strength is for highly damped, high modal density systems, is it not?

10 hits is a funny number to use, I'd often prefer 3 'perfect' hits to a larger number of less perfect hits.

Good idea with the speaker, seems like you have a good handle on this stuff.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[mkoijn9]

Thanks GregLocock, You may be right about Polymax, Lms aren't going to give me any help however as I wrote the software myself! I might just rerun the test to see if I get different results, This time I'll try 3 good hits instead of 10!

Regards

Albert

 

[MikeyP]

For the record, my opinion is that time domain curve fits (polyref) give better results than freq domain fits for light damping.

The reason is the number of points used in the least squares fit. If you are fitting to a lightly damped mode on an frf, you probably only have a few spectral lines associated with that mode, but you could have thousands of samples in the time domain (provided you use them!).

Of course polymax is neither Arthur nor Martha. Its in the z-domain which is a mix of time and freq but the work that I did a while ago on damping mesurements in lightly damped structures gave the same results for polymax and polyref, but the freq domain fits were much more variable.

I seem to remember with polymax you start from the impulse response functions (assuming you are following the LMS published paper). In which case, use as much of the IRF as possible before transforming to z-domain (Plot the IRF on a log scale and use all the samples up to the point where they disappear into the noise floor).

I think the trick with the averaging is to repeat the whole experiment several times and look at the confidence intervals on your damping measurements (rather than doing loads of averages on your frfs)

Interestingly the reason I was doing these tests was to assess the use of loudspeakers as exciters for modal damping measurements. If you not bothered about identifying modal mass and mode shapes, then you can just use the current signal thru the speaker instead of a force signal. This gave the same results as the hammer and shaker tests (for damping and natural freq). Your speaker does need to be small compared with the "wavelength" of the modes you are identifying though.

Ta

M

--
Dr Michael F Platten

 

[mkoijn9]

So I did rerun the tests and took GregLocks advice and only hit the bar 3 times, I did several measurements for each bar and took the best. It is surprisingly difficult to get good clean hits. The resulting frfs look very clean and coherence is also good. The new results are lower than before:
Short bar 0.0040% of critical @ 4056Hz
Long Bar 0.0044% of critical @ 957Hz

So indeed very similar damping values.

Many thanks again

Albert

 

[mkoijn9]

MikeyP, Thanks. In one of the papers from the developers of the polymax algorithm they prove that the idea that time domain is for light damped modes and frequency domain identification is for highly damped modes is not true. I don't think the polymax algorithm starts with impulse responses. All the papers I have about it start from frfs.

These papers are very interesting, I wrote my fortran code from these papers.

http://mech.vub.ac.be/avrg/publications/ModalAnalysis.pdf

http://mech.vub.ac.be/avrg/PhD/PhDBCauberghe.pdf

http://mech.vub.ac.be/avrg/PhD/PhDBCauberghe.pdf

Regards

Albert

 

[MikeyP]

Good links, thanks. I'll have a look through.

M

--
Dr Michael F Platten

 

[mkoijn9]

Hi MikeyP,

Heres the paper, where the author disputes the "time domain for low damping , freq domain for high damping" idea. At the end of page 92 (113 on the pdf page counter).

http://mech.vub.ac.be/avrg/PhD/thesis_PV_web.pdf

Also very good explaination of the lscf algorithm.

Regards

Albert