Acoustic and vibration signals contain distinctive patterns useful for detecting defects in machinery. A number of signal processing strategies have been proposed for detecting and quantifying incipient defects in gears. Nonintrusive measurement techniques like those based on laser Doppler vibrometry (LDV) and microphones need to be studied in order to help users in selecting the appropriate transducer according to the application. This article presents an experimental design to determine, as a novel contribution, the minimum detectable crack size in a geared system comparing three different transducers: a microphone, an accelerometer, and an LDV. The comparison intends to discriminate the three transducers with respect to its early detection capability. Experiments were conducted in a closed loop torque test rig for several torque levels and speeds. Experimental factorial design was used to determine the main effects and their interactions in the detection process. The well-known Hilbert and wavelet transforms have been used as signal processing technique. Their advantages in the detection of incipient defects are highlighted in this article. The results indicate that the acoustic signal stands out as the method that first detects an incipient progressive crack in gears (it detected 1.3-mm-long cracks), although, as a drawback, the results obtained using the microphone signal are more sensitive to speed and torque. The second place was for LDV with 1.8-mm crack detection, and the third place for the accelerometer with 2.3-mm crack detection.
Read full abstract