Sliced higher order time-frequency representations for detection and classification. Acoustical diagnosis application to faulty fan belts

In the case of non stationary and non Gaussian signals, higher order time-frequency representations can be efficiently used for analysis and detection purposes. Different formulations can be found in the literature. We propose in this communication to extend the application of a multitime frequency representation previously used for instantaneous frequency law estimation for multiplicative random phenomena. A statistical study was performed and corresponding simulations are presented. Finally, an experimental study was produced which consists in classifying faulty fan belts in an industrial context