I want to perform Spectral Kurtosis (Kurtogram) on 10-second vibration signal of 25.6 kHz sample rate, where the shaft is varying in speed.

Will a preprocessing step of resampling the signal be necessary (upsamling; so that there is an equal amount of samples within each shaft rotation), or will the SK handle the non-stationary behaviour?

I'm using Matlabs implementation:

"The fast kurtogram algorithm uses bandpass filtering along with a simplified computation to approximate the spectral kurtosis for each window size and frequency ..


(The kurtogram looks much better on the resampled signal, but I am not sure if this is the correct approach.) enter image description here

The objective is to diagnose the health of a bearing by first applying an optimal band-pass filter based on SK. After that, I wish to find the enveloped signal. Then, I want to decompose the filtered signal using EEMD to look for fault characteristics. The bearing is on an operating wind turbine.

Much like this paper: Faulty bearing signal recovery from large noise using a hybrid method on spectral kurtosis and ensemble empirical mode decomposition

  • $\begingroup$ It all depends on what the objective for using kurtosis is. Is it possible to provide some more information around the problem? $\endgroup$
    – A_A
    Mar 19 '20 at 9:45
  • $\begingroup$ @A_A added some information now. $\endgroup$
    – meerkat
    Mar 19 '20 at 11:33

Your Answer

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

Browse other questions tagged or ask your own question.