Time-frequency analysis and detection of supersonic inlet buzz

Supersonic inlet buzz and, in particular, its onset were studied through the analysis of pressure records. These records are issued from an experimental study that was presented in a previous paper. The onset of buzz is a nonstationary phenomenon, and so the pressure signals could not be analyzed using classical signal-processing tools such as the Fourier transform, which does not take into account the possible evolutions in time. That is why time-frequency methods (spectrogram, wavelet transform, and Wigner-Ville distribution) were used. The evolution in time of the energy levels corresponding to buzz frequencies were studied. This analysis shows the existence of precursor phenomena that can appear several tenths of seconds before the onset of buzz. The detection of these precursors could, with the appropriate activation of a control device, allow the avoidance of buzz. Two change-detection algorithms (cumulative sum and generalized likelihood ratio) were tested on experimental pressure signals and proved their ability to successfully detect these precursors.