Cyclostationarity and real order derivatives in roller bearing fault detection

Abstract

Abstract Various methods for recognizing cyclostationarity in signals are relatively often used in the field of machine diagnostics. Enveloping techniques are commonly applied for this purpose. Spectral correlation and spectral coherence have been studied rather widely as well, but they are not often applied to condition monitoring in industrial applications. The squared envelope is technically very similar to the conventional envelope, but is less often applied in this field of engineering.
In this paper, we present results of our experiments combining real order derivatives with other methods. The results suggest that adjusting the order of derivative can enhance results to some extent. It even seems possible to replace spectral correlation and spectral coherence with real order derivatives and squared envelope, with no significant difference in the results.
Because this way the results can be obtained with less computational power, it can be a useful result when utilizing inexpensive hardware is desired.