A Time-Frequency Post-Processing Technique for Characterization of the Unsteady Phenomena in the Turbomachines

This paper presented a data processing technique that combines different signal analyses for the spectral characterization of the unsteady phenomena developing in the turbomachines. Tools of the classical Fourier analysis, such as second order spectra, were combined with tools of the time-frequency analysis (wavelets, cross-wavelet, coherence wavelet functions) to identify unsteady phenomena, to define their spectral structure, to analyse their evolution in time and to determine their possible propagation direction and velocity. Because of the non-linearity of several unsteady phenomena developing in turbomachines, high-order moment functions were applied to discriminate between non-linearly coupled pulsations and self-excited pulsations and to determine the fraction of the power of each pulsation that was due to the non-linear interaction of unsteady phenomena.