Vibration analysis of a vehicle cab with FFT

[claudiodeviaje]

Hi everyone.
I'm working in a proyect for the university where I have to create a computer model in Simulink to simulate the behaviour of the cab of a vehicle with a semi active damping system.
The model is complete and running. The results look quite good. Now the problem is that I need to find its natural frequencies.
My model has the street (its surface and lumps) as disturbance, and es result I get the postition, velocity and acceleration.

In order to calculate the natural frequencies I use the hammer test by raising ans lowering the cab in 0.001 seconds (which should be like hitting it with a hammer, since I dont have forces or accelerations, just changes in the height of the road, as disturbance for my model). Then I calculate the Fast Fourire Transformation for the Acceleration. The results I get are a littlebit too low.

The question. Whats the differenece from calculating the FFT for the vertical acceleration, velocity an position?

To get the natural frequency I have to calculate de FFt for the acceleration, right?

Thanks,

Claudio

 

[Rob45]

No.
If you have a natural frequency, the characteristic frequency is the same regardless of whether you are measuring displacement, velocity (time rate of change of displacement), or acceleration (time rate of change of velocity).

There will of course be differences in phase.

 

[claudiodeviaje]

Thanks Rob for your answer!
I've calculated the FFT for the velocity and acceleration and as you say, the results are the same. Theres just one small difference between the graphs I get.
The results of the FFT for the acceleration are really clean. I mean there are just a few peaks for the natural frequency and some smalles in the multiples of the NF.
On the other hand, the FFT for the velocity looks quite dirty. Theres clearly a higher peak for the natular frequency, but theres also lots of noise. At the end the graph look more like a "hill" with peaks and not as single peaks.
For the FFT I'm using a triangle windows in order to reduce the affects of the first and last results. I also use a power of 2 for the data. I calculate with 16384 samples. The NF for the cab is about 0.7 Hz. Which I think its a little low, since according to the ISO, with 0,5 the driver might get motion sickness!

Where does this difference comes from?
Is it normal such a difference?
How could I "clean" the results for the FFT for the velocity, other than using a triangle window?

Do you know a good resource in the internet where I could learn more of the usage of FFT for vibration analysis?.... That way I won't have to bother you that much with my questions

Thanks!

Claudio

 

[Rob45]

I can't readily give you good answers to most of your questions, but I can to the last one:
her are two good links:

http://www.steema.com/FFTProp/DFT Interpretation/DFTInterpretation.htm

(you may have to cut/paste this link)

and especially this one, "FFT Properties:"

http://www.steema.com/FFTProp/FFTProperties/FFTProperties.htm

Best regards,
Rob

 

[GregLocock]

I'd just add that a triangle window is an eccentric choice, you'd probably be better off with Hamming or Hanning or Blackman-Harris.

Your velocity signal is just the acceleration divided by jw, so it's signal to noise ratio will fall off proportional to frequency. This may explain your high frequency problem.


0.7 hz is on the low side but with active control it may work. 1.2 Hz would be better


Cheers

Greg Locock

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

 

[Vibralogger]

Maybe FFT is not the best tool for this analysis since it is best for time-continuous signals. Short term events are best done with STFT (Short time frequency Transform) which I understand is like an FFT done on short bursts of time signals. The result is presented by colour coding the output magnitude, and plotting frequency v/s time (x). Thus each interval on the (x) time axis is a small finite strip of time, y axis the frequency, and colours represent the magnitude within this tiny strip (reds: high; greens: low) Thus you get a spectrogram. On this, patches of red can be read off on y axis for frequencies. Advantage: no need to window results. You get a picture of pre-impact and post-impact spectra.

 

[MikeyP]

STFT or spectragram is just the same as nay DFT process. You still need to window to minimise leakage and maximise resolution.

M



--
Dr Michael F Platten

 

[Rob45]

I'm guessing that by this time Claudio has either solved his problem or moved on to other things!

 

[electricpete]

I agree he's probably moved on. On the off chance he stops back, my few comments

If the test is a simulated impact test, you can avoid leakage by using the "rectangular" window (i.e. no window) and ensuring the duration of your FFT input is much longer than the settling time of your response (dies close to 0). (unless calculation time becomes a problem).

The difference between acceleration, velocity, and displacement spectra is just a scaling factor proportional to frequency (acceration raises high-freq peaks relative to low-freq peaks, while displacement raises low-freq peaks relative to high frequency peaks).

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

 

[electricpete]

Just to clarify for components at a frequency f
|a|=2*pi*f|v|
|v|=2*pi*f|d|
additional unit conversion factors may apply.

This is nothing new to the other folks in this forum, but I just wanted to clarify before someone might take exception with my poorly worded prevoius comment.

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

 

[Rob45]

And lastly, I'd guess that the 0.7 Hz natural frequency is much too low, so i'd question those results.