1

I am using Empirical Mode Decomposition (EMD) algorithm to decompose a vibration signal generated by outer race bearing defect using case western réserve University data.

I am facing a problem: unlike noise-free faulty signal results, noisy signal results (I add noise to this signal with snr=3 or snr=-3) in only the first Intrinsic Mode Function (IMF) selected and giving the fault information.

Is it possible that only the first IMF be sensitive to the fault?

Royi
  • 19,608
  • 4
  • 197
  • 238
RIMA
  • 43
  • 6

1 Answers1

4

As its name suggests, Empirical Mode Decomposition is empirical. Therefore, its behavior is very unstable and hard to predict. Moreover, it is highly computationally demanding. For bearing faults' detection, I strongly suggest cyclostationary analysis. This is the state of the art and it is covered throughly. For this task, EMD was popular for a short time but I strongly suggest avoiding it.

Gideon Genadi Kogan
  • 1,156
  • 5
  • 17
  • I'm really grateful for your response, I'll take your suggestion into consideration – RIMA Nov 15 '22 at 15:28