The autocorrelation function is used to reveal periodicities in waveform data. • The FFT is also used to reveal periodicities, but autocorrelation keeps the data in time waveform format • We will also see that autocorrelation is effective at revealing signals with low duty cycles
The autocorrelation function can be considered “a measure of the similarity a signal has with a time shifted version of itself*”
Autocorrelation can be calculated by performing the inverse Fourier transform of the power spectrum.
The remaining waveform contains the correlated signals. Uncorrelated (random) data is removed. • Synchronous and non-synchronous periodic data survives, unlike time synchronous averaged data
It is therefore useful for rolling element bearing analysis and other applications where the periodic patterns are difficult to see in the waveform. It is very useful for identifying faults at a low periodic rate relative to analysis bandwidth* • For example, faults in the cage of a bearing • Signals with higher periodic rates and higher duty cycles will be revealed via the FFT
