Transfer Function

[SCHMIDT]

I have acceleration vs. time data for two points in a mechanical system. Is it possible to develop a transfer function (in the frequency domain) between the two points using this data? I would like to use this transfer function to predict the response at one point given an input at the other.

 

[tomirvine]

A transfer function relates acceleration to force in the frequency domain.

A transmissibilty function relates one acceleration point to another acceleration point in the frequency domain.

Thus, you can calculate a transmissibility function for your data.

Sincerely, Tom Irvine

http://www.vibrationdata.com

 

[WIM32]

Schmidt,

Random acceleration-time data for two points will not suffice for a transfer funtion, as is also pointed out by TomIrvine. What you probably want is a Frequency Response Function (FRF) for your system. This is a system transfer function evaluated along the jw axis. For this you need to do a modal analysis, i.e. measuring force signal and corresponding acceleration for several points in your system.
Good luck

 

[VibrationSpecialist]

Dear SCHMIDT,
The Transfer function is not limited to relate Acceleration and Force. For example, you can relate force to velocity or force to displacement.

When you want to relate two signals that are of the same parameter then you should be after the amplitude ratio, the phase relation and the coherence.

So the answer to your question is Yes.

Take Care. ;-)

 

[1112]

Hi:

I am looking for a transfer function of the change of frequencies in different medium, for example, i know the foundamental frequency of an object, but i need to know if it will be changed when the wave enters into the air, what's the relation between them? Can it be calculated using energy method?

Thanks.

 

[electricpete]

You can certainly compute a transfer function. How close it corresponds to what is physically going on remains another story.

For instance... if the two measured signals are originating from two different unrelated forcing functions in two separate systems, then the ratio is fairly meaningless. Perhaps some correlation functions can help to sort this out.

If one meausred signal does represent a transform of the other signal, the transformation may not be linear. In that case again the ratio of Fourier transforms has limited physical meaning.

 

[GregLocock]

1112-In conventional linear theory the frequency won't change as the wave moves from one ether to another. The wavelength changes to accomodate the change in c.

If you take an object and immerse it in a different fluid then its resonant frequency may change, depending on the mass loading of the object by the fluid, but I don't think that was what you were getting at.

In a non linear system it is possible to generate differnet frequencies from the source frequency - I would suggest a time domain model will provide the easiest way to understand that effect.
Cheers

Greg Locock

 

[GregLocock]

Sorry, I meant to answer both the questions at once

Schmidt - IF your acceleration data includes information at all the frequencies of interest then you CAN take the frequency spectrum of each, and divide one by the other to give a valid transfer function. In an ideal linear system the response at a given frequency is at a constant ratio for two points, always, if there is only one forcing function.

Proof: at a frequency x1/F3 =k1, x2/F3=k2, hence x1/x2=k1/k2, which is a constant

where k's are complex constants in a linear system.

The names often given to this in modal software is running modes, or operational deflection shapes.


Cheers

Greg Locock