Hammer Test - Frequency Response Analysis

[nvhuser]

Hi, I have some acceleration curves measured by a test-engineer (the data available consists into Magnitude and Phase).
Currently, I would like to have a list of the normal modes (eigenfrequencies). So, the question is: how to extract the normal modes using FRF-experimental data?

Since I have a MDOF system, the extraction of the normal modes is not directly inferred from the resonance peaks. Thus, I am wondering which is the best method to obtain this information. Anyone has an idea about the methods that can be used?

Thank you

 

[JoshPlumSE]

The only time I've ever done anything like this, I relied entirely on software to do the heavy lifting. The same software that measured the FRF's (for excitation and response at each discrete point) then converted the response into model shapes and natural frequencies for the structure. It was a lot of testing and measuring, to come up with the response. A mostly academic exercise for real world structures here on earth.... But, the intended application was for a space station. Where you could easily detect damage based on the FRF values.

I believe the program we used was "IDEAS". I don't know that much about it because that project (circa 1995) was the only time I ever used it.

There should be some published research papers written on the project, though I don't have them. Only some rough drafts and thesis work. They would be written by the following authors.
J.B. Kosmatka and K.L. Napolitano

Possibly titled "Damage Detection of Highly Damped Structures Using Direct Frequency Response Measurements and Residual Force Vectors".

That paper should have the theoretical basis (i.e. equations and derivation of concept). Hopefully this helps!

 

[JoshPlumSE]

FYI: I looked them up for old times sake. Didn't find that specific paper. But, Kevin Napolitano is at ATA engineering in San Diego. And, John Kosmatka is still an engineering professor at UC San Diego.

 

[JoshPlumSE]

Found it.... That paper was Published in SPIE Proceedings Vol. 2720: Smart Structures and Materials 1996: Passive Damping and Isolation

 

[nvhuser]

Hi JoshPlum, thank you for all your excitement during this quest!
It is good to see people committed in FRF analysis

Sure, this is really an academic problem, IDEAS is very popular to do this kind of stuff, but sadly I don't have a license available.
Could you send me the link where I can download the paper? Or could you upload the paper here?
That would be very helpful! Thank you

 

[GregLocock]

You are basically talking about analysing an experimental modal analysis. The best introductory text on that is Ewins. There are many gotchas both in taking the data and analysing it, to be honest I hate working with other people's test data unless I have worked with them.

I dare say Tom Irvine has some scripts for FRF fitting

http://www.vibrationdata.com/

, depending on the data quality and the complexity of the model you might get away with anything from circle fitting up to a full MDOF model.



Cheers

Greg Locock


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

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

 

[nvhuser]

I read the book from Ewins (Imperial College London) - Modal Testing, and he describes there the circle fitting method. However, Nyquist plot for a MDOF model unfortunately presents "poor circularity". I think it is difficult the modes there.
Also the peak picking method to determine the normal modes is quite inaccurate.

One property of accelerance/inertance for SDOF models is that the real component is zero for a frequency equal to the eigenmode. I don't know if it can be true for MDOF systems, but this could be a way to find the list of modes.

The most famous method uses the "Complex Mode Indicator Function".

Link

If anybody is experienced on that or has any clue how this works, please I am looking forward your feedback. Thank you

 

[JoshPlumSE]

I don't have an electronic copy of that paper. Heck, I'm not even sure if my copy is the final paper or not. However, here is a link to a website where it can be downloaded / purchased. Only $18.00 if I read the site correctly.

Link

 

[nvhuser]

Hi JoshPlum, thank you for your link, I will ask my manager to buy it for me. Thank you for your suggestion!

 

[GregLocock]

Peak picking and circle fits are as you imply not suitable for FRFs with highly damped modes or high modal density, but then unless you did a MIMO modal analyss then I suspect you'll struggle in those two cases anyway.



Cheers

Greg Locock


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

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

 

[JoshPlumSE]

Greg / Nvhuser -

Good discussion (though I'm not sure anyone else is really following). Based on your comments I find it ironic that all the lab work / testing I did in my college days was for structures with so little damping. Looking back, I think there were probably better (or simpler) methods available for doing this.

Now I understand why the professor (and his PhD student) were so focused on what would happen after we replaced one of the truss members with a "damaged" member. I remember them trying figure out what kind of member to use to give them the desired behavior. It was lost on me at the time, but now I understand that the whole point was that the "damaged member" needed to provide high damping values overall. Or, non-proportional damping.

Funny how something like this can click into place in my mind.... 20 years after my initial involvement in the project! Thanks to both of you.

 

[nvhuser]

Hi JoshPlum,

you are welcome. I think this is the purpose of this forum: to learn with the community and have fun