Statistical feature extraction of pre-lockup torsional vibration signals for condition monitoring of wet friction clutches

Wet friction clutches play a critical role in vehicles equipped with automatic transmissions. It is unavoidable that they progressively degrade and eventually break down while the vehicles are under operational condition. To avoid unexpected failures, a proper condition monitoring tool extended with a lifetime prediction method should be implemented. In order to achieve this, appropriate parameters delivering any information about the failure modes and levels must be determined. This paper presents an experimental investigation on a set of statistical parameters generated from the zerothand first-order derivative of time domain pre-lockup torsional vibration signals, to determine the appropriate parameters, often called relevant features, for condition monitoring of wet friction clutches. The torsional vibrations are measured by using two Ferraris sensors installed on both input and output shaft of the clutch. The determination of the relevant features is based on an evaluation of the historical trends of the parameters. Clearness and fluctuation level of the historical trends are considered as criteria for selecting relevant features from the generated statistical parameters. The trends of the parameters are extracted from lifetime data obtained from an accelerated lifetime test on a paper-based wet clutch, using a SAE#2 test setup. This study reveals that the RMS value and Shannon entropy extracted from the first-order derivatives of the pre-lockup torsional vibration signals, and the kurtosis extracted from the zeroth-order derivatives of the pre-lockup torsional vibration signals, can be used as relevant features for condition monitoring of wet friction clutches.