Vibro-Acoustic fault diagnosis of two-stage spur gearbox by using the Empirical Mode Decomposition Technique

Abstract

The two-stage spur gearbox was found to be one of the most prominently used mechanical
drives in power and motion transmission. Gears were subjected to varying service conditions, that are
accidentally encountered to abnormal working conditions, would lead to a wide range of failure of teeth
such as wear, pitting, micro pitting, scuffing, partial or complete failure of teeth, etc., These faults
influence the severity in vibration and acoustic at different operating conditions. These trigger a
localized or distributed fault on the load supporting surfaces of the gear tooth. This severity increases the
Vibro-acoustic levels, reduces the efficiency of power transmission, and may impair the entire
mechanism. Health monitoring of Mechanical drives for analyzing vibration and acoustic signals have
been a challenging task for the researchers and practitioners, to root cause the fault severity. In this
situation, the “Empirical mode decomposition (EMD)” method has introduced for the effective analysis
of Vibro-acoustic signals, which would result in reliable information for the fault diagnosis. This
machine learning approach is utilized to find the fault severity of a two-stage gearbox, by studying the
vibration and acoustic signals. The output of this EMD method is an intrinsic mode function (IMF),
where the statistical parameters of IMF such as crest factor, kurtosis, RMS velocity, and peak values are
evaluated and expressed as a measure, for the fault severity. This work also demonstrates the
effectiveness of EMD based processed signal over the unprocessed acoustic signal for reliable fault
information.