PSD and Sine Sweep tests - Which is more appropriate

[jamesbewley]

Hello

I am currently doing some FE work on a part which undergos vibration during its life time.
These vibrations are causing failure in the component so my customer wants to do some FE work to validate a new design.

I have been provided with a sine sweep curve, which details double amplitude displacement V frequency.

I have also be provided with PSD graph, specifically acceleration PSD v frequency.

I have been asked to simulate both inputs, however from my limited knowledge (or lack of it!)I cannot unserstand why you would want to do a sine sweep if you have PSD data.

So I have the following questions.

What is the purpose of a sine sweep test?
Is a Sine Sweep representative of actual life?
What data should I be interested from a sine sweep test?

From PSD data is is possible to determine a peak amplitude of acceleration, which would relate back to a sine sweep for comparisson pruposes.


Cant think of anymore right now. Appologies for the length of thread and many thanks in advance for your help.

James

 

[CESSNA1]

JAMES: A sine sweep test is called for in some test specs. or it can be used to search for resonances in a fiture and/or test item. In general a random test is more representative of actual life. Low level random is also a good way to find resonances. In a sine sweep you would look for resonances and possibly damage to the unit under test. The PSD data will show the amplitude(in G^2rms/Hz) at each frequency. An approximation of the amplitude at a certain frequency can be obtained by taking the G^2rms/Hz value of the amplitude at a specific frequency and taking the square root of it to get the approximate Grms value of the amplitude at that frequency. Multiply by 1.414 to get the peak amplitude.

Part of the concern with a sine sweep is that it takes time for the structure to respond, If you sweep too fast you may miss a resonance.

The advantage to random vibration is that all the frequencies are present at the same time. In a sine sweep you hit them one at a time.

Hope this helps
Regards
Dave

I suspect that your customer really does not understand what they want. Usually FEA is good for loading a structure with a known displacement vs. time function and observing the transient reaponse.

 

[CESSNA1]

JAMES: A sine sweep test is called for in some test specs. or it can be used to search for resonances in a fiture and/or test item. In general a random test is more representative of actual life. Low level random is also a good way to find resonances. In a sine sweep you would look for resonances and possibly damage to the unit under test. The PSD data will show the amplitude(in G^2rms/Hz) at each frequency. An approximation of the amplitude at a certain frequency can be obtained by taking the G^2rms/Hz value of the amplitude at a specific frequency and taking the square root of it to get the approximate Grms value of the amplitude at that frequency. Multiply by 1.414 to get the peak amplitude.

Part of the concern with a sine sweep is that it takes time for the structure to respond, If you sweep too fast you may miss a resonance.

The advantage to random vibration is that all the frequencies are present at the same time. In a sine sweep you hit them one at a time.

Hope this helps
Regards
Dave

I suspect that your customer really does not understand what they want. Usually FEA is good for loading a structure with a known displacement vs. time function and observing the transient reaponse.

 

[jamesbewley]

Dave

Many thanks for your comments.
I certainly belive that my customer has been asked to do this without considering why.

So in summary the main purpose of a sine sweep is to identify resonant frequencies. This can obviously be done much more simply in FEA by conducting a modal analysis, therefore rendering a sine sweep analysis pointless.

Using a modal analysis I have already confirmed that the resonant frequencies are well outside the frequency range of the sine sweep.

Would this explain why stresses determined from FE analysis of the sine sweep analysis are well below any problem values, whle the PSD analysis shows failure (corresponding to real life).

With regards to the random vibration, can I confirm the method of obtaining the peak acceleration would be as follows.

Multiply the G^2rms/Hz value by its corresponding frequency.
Take the square root of this value and then multiply by 1.414.
This would leave a value with units G.

Many thanks again

James

 

[BobM3]

It's been awhile, but I don't believe that's the correct way of geting the average amplitude at a frequency from a PSD. First,

1) Most PSDs show only the positive frequencies. Half the power is in the unshown negative frequencies.

2) You can't technically get the answer at a specific frequency because of the bandwidth of the filter (or delta F resolution, if digital) used to generate the curve.

Given that, if you know the frequency resolution used to calculate the Fourier transforms then:

1) Take the value at the frequency of interest,
2) Multiply it by the frequency resolution,
3) Take the square root of that number
3) Double that number to get the total power.

That gives you the peak value of the sine wave.

So for example, if your PSD says 100 g2/hz at 10 hz and you know the resolution was .2 hz, then the power of the signal at 10 hertz would be (nearly):

square root((100 * .2)) x 2 = 8.94 g's peak

But it has been awhile. You may want others to share their knowledge.

 

[CESSNA1]

JAMES: When you estimate the amplitude level do not multiply by the frequency. What you essentially see is an edge on view of a sine wave. So by taking the square root of the RMS amplitude you get the Grms level AT THAT frequency. Sorry, I really cannot comment on your questions because I have not seen the data or the analysis.

BOBM3: The method described is an approximation only. When you see the amplitude of G^2rms/Hz what you really see is the average of all the sine amplitudes in the bin. Say, for examply the center frequency is 1000 Hz and the bin is 900 to 1100 Hz. All the frequencies in the bin are averaged to get the center frequency amplitude at 1000 Hz.

If you add up all the amplitudes under the PSD curve and take the square root you will get the total energy in Grms of the entire profile. Note that usually the PSD curve is presented as a log-log plot and has straight lines. If you display the PSD curve linearly the lines are curved, so you just cannot add up the areas, there are specific equations that must be used.

If you would like the equations for the PSD profile please respond with your E-mail address and I will send them to you. Mine is dphall@sbcglobal.net

Hope this helps
Regards
Dave

 

[tomirvine]

As an aside, there are some special cases where a sine sweep actually represents an expected vibration environment.

Certain solid rocket motors, for example, have cavity pressure oscillations that produce sinusoidal excitation that sweeps downward in frequency with time. Avionics must be designed and tested accordingly.

I can also image that the start-up and shutdown transients of, say, a helicopter rotor generate sine sweep vibration.

Again, these are special cases. Sine sweep is mainly used nowadays for natural frequency identification, as has already been mentioned.

NAVMAT P-9492 has a good, historical discussion about random versus sine sweep.

Tom Irvine

http://www.vibrationdata.com

 

[IRstuff]

I would dispute "This can obviously be done much more simply in FEA by conducting a modal analysis, therefore rendering a sine sweep analysis pointless."

An FEA is only as good as the detail put into the model. I've been on two projects where FEA was used and both times, the model did not accurately reflect the actual sine-sweep results.

TTFN

Eng-Tips Policies FAQ731-376