A joint time-invariant wavelet transform and kurtosis approach to the improvement of in-line oil debris sensor capability

In-line oil debris sensors are important devices for the detection of machinery failures. However, two key issues remain to be addressed to more effectively make use of the existing oil debris sensors: the responsiveness to early machine failures and false alarms. The responsiveness level depends on the size of the debris that can be detected by an oil debris sensor. The detectable particle size in turn is mainly limited by the background noise. The false alarms are often caused by spurious impulses such as vibration-like signals. The challenge of improving the responsiveness and reducing false alarms lies in the very weak particle signals and their similarity to spurious signals. In this paper, a joint time-invariant wavelet transform and kurtosis analysis method is proposed to address the two issues simultaneously. The proposed method has been tested by extracting signatures of ultra-small metal particles from background noise and a wide range of simulated vibration-like and real vibration signals. Our test results have demonstrated that the proposed method can effectively detect very weak particle signals buried in strong background noise and eliminate vibration-like spurious signals. The implementation of the proposed method will substantially enhance many existing oil debris sensors