3
$\begingroup$

I have to find whether there is any pattern (I mean periodicity or close to periodicity) and if there is, for one cycle i have to perform numerical integration to determine mean. In the first picture I have done moving average(every 254 data points) using excel. enter image description here Then in the next picture i have done again moving average(with 250 points) using the data got after performing first moving average.

enter image description here I have also thought about doing Fourier transform. These datas are values of cl (lift co-efficient) vs time in a unsteady solution of CFD simulation.Variation of values at initial time steps are not significant as some time is needed for simulation to give exact value. DATA file is here (drive.google.com/open?id=0B1KEir1gFkaMQW5GbmFoZlpRUVk)

$\endgroup$
2
$\begingroup$

To determine periodicity, do an autocorrelation. The autocorrelation will have peaks at time offsets consistent with the periodicity.

You mentioned using the Fourier Transform, if convenient you can do a circular autocorrelation using the FT as follows:

$$Corr = ifft(fft(x) fft(x)^*)$$

Where $fft(x)^*$ is the complex conjugate (if you signal is complex).

$\endgroup$
  • $\begingroup$ i am an mechanical engineering student-senior year.this is needed for my thesis.i have done auto correlation with excel(i don't know much,have learned it today,so don't know whether i have done it correctly)here is the excel file [link]( drive.google.com/open?id=0By4UphfVdojqbGNOZ2xmbGhONE0) @Dan Boschen here is autocorelation plot pic drive.google.com/open?id=0By4UphfVdojqa0NkN1EtUlI4bmc $\endgroup$ – Fringe Emanuel Mar 14 '17 at 15:12
  • $\begingroup$ Hmmm didn't look at the Excel but would imagine that would be a difficult path. Take the plunge and install Octave, it's free! gnu.org/software/octave/download.html $\endgroup$ – Dan Boschen Mar 14 '17 at 15:34
  • $\begingroup$ there are examples available how to load data from a csv file, fairly straight forward, but once you have your data in two variable a and b, the command is xcorr(a,b); or using the fft's as I described it would be out= ifft( fft(a) .* conj(fft(b) ). And then you can plot your result using plot(out). That might get you started if you wanted to try going down that path. $\endgroup$ – Dan Boschen Mar 14 '17 at 15:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.