1
$\begingroup$

I have a vibration signal (25.6 kHz) which has been through the following filters:

1: highpass (2000 Hz) -> (I'm not interesten in the harmonics of the system. I then perform spectral kurtosis to find the optimal bandpass bandwidth.) 2: Bandpass (8750 Hz,9250 Hz) -> 3: Rectification (abs(xi)^2) -> 4.Lowpass (2000 Hz)

The fft of the final signal looks like this: enter image description here

The Bandpass filter bandwidth and center frequency was selected based on the kurtogram of signal after initial lowpass.

I am no expert in signal processing- does the FFT look "correct"? It seems odd that the amplitudes rarely go below 0.05...

Updates

The original signal (25.6 kHz): enter image description here

Zoomed in in original signal (25.6 kHz): enter image description here

Final signal in time domain: enter image description here

FFT of original signal (before step 1) enter image description here

Objective of project:

  • To look for any fault development in the early stages of a gearbox fault on a wind turbine. I have signal data 412 days, in the form of 412 10-second intervals.
  • Want to extract features from time and freq. domain and cluster the 412 intervals, to see if there is a pattern.
Improve this question
$\endgroup$
9
  • 1
    $\begingroup$ Any chance you can attach the signal plot as well? $\endgroup$
    – DSP Novice
    Mar 25, 2020 at 10:55
  • $\begingroup$ Hm, why would you first in 1: cut off anything above 2000 Hz, then in 2: cut off anything below 8750 Hz? That very much sounds like a design mistake. If your low-pass filters are good, you don't have any signal after that. $\endgroup$
    – Marcus Müller
    Mar 25, 2020 at 10:58
  • $\begingroup$ The 25.6Hz in your question is a typo, correct? It should be 25.6kHz based on your previous questions.. $\endgroup$
    – jithin
    Mar 25, 2020 at 11:44
  • $\begingroup$ @DSPNovice added some more plots. Let me know if you would like to see more. $\endgroup$
    – meerkat
    Mar 25, 2020 at 14:38
  • $\begingroup$ @MarcusMüller I updated the process. The optimal bandwidth is around 9000 Hz from the spectral kurtosis. $\endgroup$
    – meerkat
    Mar 25, 2020 at 14:40

2 Answers 2

Reset to default
1
$\begingroup$

If your description is correct, you are seeing noise.

After step 1, everything above 2 kHz should be gone. After your band-bass everything below 8750 Hz is gone, so you basically end up with a null set. Since your filters are not infinitely steep there is still something non-zero left over but it's mostly going to be noise and very poorly defined.

Rectification is also a highly non-linear operation, so chances are you also going to get some aliasing in there as well.

Improve this answer
$\endgroup$
3
  • $\begingroup$ True, there seems to be a bin at around 25 Hz and another at 50 Hz probably caused by the rectification process. $\endgroup$
    – Ben
    Mar 25, 2020 at 13:34
  • $\begingroup$ Wouldn't the rectification step (I am doing the abs value and then raising to the power of 2) move the high frequency resonances to lower frequencies abnds? I'm looking to perform an enveloping process in my case. What is an alternative to the rectification step? (I remember you mentioned the issue with just performing the abs value in another answer). $\endgroup$
    – meerkat
    Mar 25, 2020 at 14:49
  • $\begingroup$ I had a major typo, I meant to write step 1: Highpass (2000 Hz) $\endgroup$
    – meerkat
    Mar 25, 2020 at 14:55
0
$\begingroup$

Maybe you should average your FFT over time to enhance you signal's SNR, you may then be able to see a spectrum that fits your expectations. It also seems that your signal is filtered around 420 Hz. Unless this is your initial signal's bandwidth, your filtering may not have been applied correctly.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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