Modal Damping 2

[meshparts]

Hello,

I need to identify modal damping coefficients for a modal superposition analysis with FEM. I know that one method to compute the damping is the half power bandwidth, when the frequency response is known.

I have the frequency response curve of the real structure. But this structure have high modal coupling or high modal density.

I assume that the half power bandwidth method will be to inaccurate in such a case, wright?

The question is: Is there still a possibility to identify the damping?

I am grateful for any comments.

Regards,
Alex

 

[MikeyP]

Yes, but this can only be done with a full experimental modal analysis. The "modal damping" model also assumes that there is no modal damping coupling between modes. This is not always true!

M

--
Dr Michael F Platten

 

[meshparts]

Thanks MikeyP for your answer!

I will try to be more specific. The structure that I'm simulating have high modal density (1 to 2 Eigenvalues pro Hz!). By the method of half power bandwidth one mesures the bandwidth at 3 dB under the peak of the eigenfrequency and divides it by 2 times eigenfrequency to get the damping ratio:

psi=(f_2 - f_1)/(2*f_0)

This works only if one can realy measure the half power bandwidth. In the case of high modal density of my structure I can't measure it!

Is there another way (other then half bandwidth) to get the damping ratio from the frequency response curve and how?

Thanks!

 

[GregLocock]

That's an interesting problem.

Without seeing your measured results (I'm not much interested in the FEA side, for various reasons), one very unrobust measure is to look at the rate of phase change for each mode. This is reasonably well described in the original SDOF circle fitting paper by whoever that was. Kentucky?


However, the damping coefficient of close coupled modes is, in my experience, so non linear as to be useless.



Cheers

Greg Locock

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

 

[MikeyP]

I've just been thinking about how I would go about this if I was trying to MEASURE damping in a high modal density situation.

In that case, you attempt to measure an average modal damping over a frequency band (say 1/3 octave). You do this by applying a broadband random excitation over that frequency band until a steady state is achieved. Then you switch off the excitation and measure the decay response. When the envelope of this response is plotted on a log scale it will be approximately a straight line and the slope of this line will be related to the average loss factor for that frequency band.

You could then apply that average loss factor to all the modes in that band. It assumes that the modes are close enough together that the damping couples energy between them, but it is a possible approach.

M

--
Dr Michael F Platten

 

[tomirvine]

I recall that there is a method called the "complex exponential multi-degree-of-freedom method" that performs curve-fitting on the frequency response function (FRF). This method requires some engineering judgment and trial-and-error. It is described in Ewins' Modal Testing book.

There is a also a method for doing time-domain curve-fitting on the decaying signal, assuming that the structure was excited by a sharp impulse. This method can resolve the frequencies and damping ratios of modes that are closely-spaced in terms of frequencies. I have some software and papers for carrying out this method. Please contact me via my website if you are interested.

Tom Irvine

http://www.vibrationdata.com

 

[cincibcats]

I agree with Tom. Not only could you use CEA but any other form (Least Squares Complex Exponential, Rational Fraction Polynomial, Ibrahim's Method). Depending on if you have multiple inputs and outputs then other methods may work better (Eigensystem Realization Algorithm, Poly-Reference Time Domain, Poly-Reference Frequency Domain).

 

[meshparts]

Thanks you all for the professional answers! It seems to me now that the best way to compute damping coefficients for high modal density is to go in time domain.

Tom I will think about your offer, but first I will do some reasearches on my own. Thanks!

Regards,
Alex

 

[MikeyP]

Polyreference LSCE/Polyreference least-squares z-domain(PolyMax)/ITD etc would require a very high freqeuncy resolution for a modal density this high. However, if you have some FRF data it may be worth a try before you go to the trouble of doing more testing.

I have matlab code for Polyreferene LSCE and PolyMax which I could run your data through if you want.

M

--
Dr Michael F Platten

 

[GregLocock]

I like the third octave idea.

Cheers

Greg Locock

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

 

[meshparts]

All this modal parameters estimation algorithms are new to me... I've done some research on this and it seems to me, that the measured data should be of very good quality. This ist not the case for my FR curves.

Regards,
Alex

 

[GregLocock]

I'm not surprised. Getting good quality FRFs from real structures (or more typically assemblies) with high modal densities is very hard. For instance, if you do a reciprocity check you'll probably feel inclined to throw all the data away.

This is where MIMO systems help, to some extent. To be honest they tend to reduce the FRFs into more of a blur sometimes, but that may be a better representation of in-service use.



Cheers

Greg Locock

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

 

[AH2]

Hello,
I read about the matlab programs to polymax and LSCE-algorithms of Mr. Dr. Patten.
I have a measured FRF and want to approximate it with
an analytical function. So I need the modal parameters.

Is there any m-file available that can estimate modal parameters (f.e. an LSCF-algorithm)?

Thank you for a reply.

 

[cincibcats]

I believe there is a script on the MATLAB file exchange for the Eigensystem Realization Algorithm (ERA).

 

[JOO2]

Is the polymax and LSCE-algorithms is available in matlab ?

Is there any modal analysis software in open source ?