Roving Accelerometer Vs Roving Hammer 1

[phanikumar]

I was trying to show one of my colleagues by experimentation how we can use both this methods to generate the same modal parameters. As I was looking at the data of Imaginary FRF'S obtained by this two methods (at different points) and overlaying them on each other, I observed a variation in the magnitude of the peaks by considerable amount at one of the resonant frequencies though it matched up pretty well for other resonant frequencies. Since both the experimentation methods were used consecutively with the same tri-axial accelerometer and hammer (hence no change in sensitivity) and with no change in the setup and the structure being fairly linear why didn't the reciprocity rule hold true for all the frequency ranges. What's the factor I am missing?

 

[SomptingGuy]

Is this perhaps a coherence thing? Are you effectively computing the upper estimator with one method and the lower estimator with the other?

 

[GregLocock]

Try increasing the frequency resolution. If that improves things then you'll figure out the cause.

Definitely look at the coherence.

Look at the input power spectrum for each case.

How much magnitude error are you talking about?

Obvious question - how good are you with a hammer?







Cheers

Greg Locock

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

 

[MikeyP]

In addition to the above

1) Are either the accelerometer (in the roving hammer case) or hammer positions (in the roving accelerometer case) close to a nodal line for that mode?

2) For the roving accelerometer case, you change the mass distribution of the structure very slightly each time you move the sensor. This does not happen when you do the roving hammer test.

3) Remember that reciprocity only generally holds for linear systems. That one mode may well be behaving non-linearly.

4) Is the anomalous result repeatable?

M

--
Dr Michael F Platten

 

[kaiserman]

Phanikumar:

I have been performing modal anlysis on body-in-white automotive structures and other structures for nearly 20 years now. I would like to comment.

Theoretically, as you attempted to point out to your colleague one should obtain the same modal parameters by roving either the accel or the hammer. I have only found 6 general cases when consistancy in the theory didn't hold true.
(1) A complete or full row or a full column of the modal space matrix was not obtained. Be sure that 1 full row or 1 full column of FRF's are being obtained in the data set. by not obtaining this one could skip some information.
(2) A non linear system was being investigated. Perform reciprocity checks at many locations to confirm linearity.
(3) Accelerometer mass errors. Be sure the accelerometer is not providing mass loading.
(4) In consistent hammer hitting. Be sure, extremely sure, that all hammer hits are orthogonal, in the same range of levels and that no double hits have occured. Doulbe hits are murder! Also consider averaging more hits together. Better yet consider using a controlled minishaker with a stinger and burst random excitation.
(5) Was data being taken at or near a nodal point or line. This can cause great havoc in data.
(6) The coherence was bad. Be sure that the coherence is great for each hammer hit. If it isn't (0.9 or better for our purposes) at the mode then reject the hit and attempta better hit. Hammer hitting is an acomplished art.

I hope this helps.

Kaiserman