Fault diagnosis of rotating machinery using a two-stage adaptive line enhancer

For a long time acoustic and vibration signals from rotating machinery have been used for fault detection. These faults commonly manifest themselves by radiating impulsive signals due to irregular impacting. However, it is very often difficult to detect these impulsive signals since they are embedded in background noise, such as narrowband signals at the harmonics of the rotation speed and broadband random processes. These background noises hinder the early detection of faults in rotational machinery. Our aim in this study is to extract these impulsive signals from the background noise. The model we consider consists of the impulsive signal associated with a fault, a narrowband harmonic signal, and broadband noise. In the study of gear fault detection, time averaging methods have been considered for the removal of harmonic noises. Unfortunately, these techniques require a signal which is exactly synchronised to the shaft rotation. We consider scenarios in which such signals are not available. (6 pages)