Abstract:

A traditional approach to the analysis of vibration and noise in condition based maintenance of machinery is to assume stationarity of the measured signals. However, in the case of machinery working cyclically (e.g., gearboxes, engines, turbines) the generated vibration and noise are basically nonstationary (cyclostationary). This feature, in the traditional approach partially suppressed by continuous time averaging, can be exploited to give more detailed information on the condition of the monitored machinery. To achieve this, periodic time averaging is used. In this paper, the second-order statistical characteristics of the cyclostationary signals are considered. These are double correlation functions, double autospectral densities, and instantaneous autospectra. The disadvantage of these statistical characteristics is their greater complexity. However, they yield a more detailed description of measured signals. Another advantage is a greater immunity to interfering signals. Algorithms for computation of the characteristics will be presented. Theoretical conclusions will be demonstrated on a concrete example of vibration data measured on a passenger car gearbox. [Work supported by the Grant Agency of the Czech Republic.] [See NOISE-CON Proceedings for full paper.]