Frequncy response vs time response analysis 2

[elogesh]

Hai,
I know that analysing the parts in frequency domain is the most desired one, because it can be related to its natural characteristics by their modal frequncy.
In one of my structure under operating conditions, I noted down the inside gas pressure versus time data. I converted the same into frequency domain, in the form of pressure versus frequency using FFT functinality. This input data, I am using it as a loading condition for numerical simulation of the displacement response using harmonic response analysis(i.e in frequency domain).Then the displacement response data will be converted into time domain using inverse FFT functionality.

I had confusion in the above problem if it is solved using transient analysis, what difference it will make?
Whether I will be getting the same results as the above one?

I am using ANSYS6.1 with SGI,octane 512MB ram and 40GB HDD

With Thanks and regards,
Loganathan.E

 

[jasonblanik]

hi Mr.Loganathan,
Its a very interesting approach that I never considered using. Do you do the fft in ansys or someother math s/w. If in ansys, please let me know how to do it.
Thanks,
jason

 

[NicolasJobert]

elogesh,

using direct transient analysis or using the more sophisticated approach ...

using FFT time => frequency
using harmonic analysis solve for response
using IFFT frequency => time

... should yield exactly the same result !

(FFT approach may be much more economical on some occasions.)

However, if your are interested in computing long transients and not waste hundreds of CPU time or GB of disk space, I suggest that you use the following path (this one I have used sucessfully, speeding up some computations 2000 times).


=> compute using ANSYS the response of your structure to a "unitary load". (This is conceptually equivalent to computing the response to a DIRAC excitation).

=> read the "unitary response(s)" with a third party software and convolve the unitary response with the actual excitation, to get the real response(s) (displacements, strain, etc.).

Like the FFT techniques, this only works for a linear structure.

This second approach will be efficient only if your transient excitation is much, much longer than you structure natural period.

Hope this helps.

Nicolas

 

[elogesh]

Hai Jason & Nicolas,

Jason:
I did FFT in sysnoise package.It can be done in excel spread sheet also.

Nicolas

Thanks for the suggestion.

Your suggestion of "read the "unitary response(s)" with a third party software and convolve the unitary response with the actual excitation, to get the real response(s) (displacements, strain, etc.)" Seems to be simple and logical. I will explore that one. I think this has to be done in frequency domain.


With Thanks and Regards,
elogesh