Random vibration analysis- Time domain Vs. Frequency domain 4

[elogesh]

Hi all,

I am associated with an heavy commercial vehicle automotive company.We do computer aided engineering for evaluating the stresses/strains and there by life of components during the product development stage.

The components (mounting brackets) as per the internal standard is tested for a PSD of 2.302 (m/s^2)^2/Hz for the frequency range between 5 to 40 HZ and maximum amplitude is limited to 30g and also the kurtosis limited to, as per the lab test spec.

Currently, we do this random vibration evaluation in time domain, which is equivalent to PSD value mentioned in the last paragraph with constant damping value of 2%. But after this analysis the peak stresses are compared with the allowable strength of the material.I feel there are pitfalls in time domain analysis approach and therefore planning to propose them carrying out the analysis in frequency domain for the following reasons,

1) Frequency domain is economical compared to time domain ( solution time as well as hard disk space).

2) Currently time domain, the peak stresses are compared against the allowable strength of the material, which may generally result in overdesign. Being stochastic in nature, in time domain, we need to plot the output stresss as probabilistic distribution function and then determine the mean and sigma stresses. But plotting the probability distribution function may be cumbersome.

In frequency domain, it straight away reports that stresses are below 1-sigma stresses for 68.2% of the time. Depending upon our requirement, we can multiply it by 2(95%) or by 3(99.7%) and correspondingly, we can interpret as stresses below 2-sigma and 3-sigma for 95.4% and 99.7% of time respectively.

3) Currently, in time domain, we do the analysis for a fraction of the time compared to the lab test, but still maintaining the PSD level as 2.302 (m/s^2)^2/Hz with peak amplitude as 30g.Being lightly damped, I fell like, we may be measuring the transient and not the steady state vibrations on time domain analysis.But in frequency domain analysis, it means steady state, hence the issue of transients can be resolved.

But, before proposing them, I would like to have expert opinions from this forum and also about the general practice followed in the industry.

Thanks in advance for your suggestions.

Regards,
Logesh.E

 

[Transient1]

What program are you using for analysis? I work in the space industry and we use the frequency domain for our analysis. I can run a PSD analysis and input the frequency versus (G^2)/Hz graph at the fixed supports with a specified damping (~3%-5%). I agree with you that this method is effective at characterizing the statistical nature of the deflections. The nature of the stresses is a matter of some conjecture. For ANSYS the accepted method is based upon the Segalman-Reese algorithm which according to ANSYS is conservative. A google search on this algorithm will supply plenty of information.

I agree about the possibility of measuring transient and not steady state response of the system. Although, could the transient response ever be a concern for your system?

 

[SomptingGuy]

This is Peak vs RMS, simply put. Transient (peak) stress breaks things, steady (RMS) stress wears them out.

- Steve

 

[cbrn]

Hi,
there may be a drawback, though, in making analyses in frequency domain. Afaik, no FE program is able to do that with systems which have internal non-linearities. So, from my point of view, first of all you should check if you have non-linearities (contacts? large deflections?...) and, if yes, whether you can "approximate" them with linear formulations or not.

regards

 

[GregLocock]

That's a problem in the real world as well. One of the arguments in favour of white noise excitation rather than swept sine for modal analysis or fatigue is that at least there is some chance of averaging across some non-linearities. Personally I'd rather take the clean data set from a swept sine modal, and run it at two levels of excitation.






Cheers

Greg Locock

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