Non-Linear Rotor-Rubbing Vibration Analysis Through the Wavelet Transform

Abstract

Vibration analysis is the basis for an early detection of faults in rotating machinery. It is the main element for any maintenance program. Due to the basics of the FFT (Fast Fourier Transform), vibration prediction systems based on this analysis tool are unable to identify unsteady and non-linear vibrations. They provide only the spectrum content of a signal, but due to its nature, the time interval and the instant when the vibration occurs are eliminated. Therefore, new modern signal analysis tools have been developed. Among them, wavelets have been extensively applied to non-linear vibration and transient analysis. In the case of a rotor rubbing the housing or the bearing, wavelets allow the identification of the “wave shape” through a predetermined time-frequency function. The challenge of this method is the definition of the function that represents the actual phenomenon. Results show that rubbing is a non-linear phenomenon, this was verified through a ramp-up and ramp-down test procedure. It was found that the most appropriate procedure for its identification is through wavelet analysis. The key factor is the definition of the best-fit wavelet.