Abstract: Unstart is an abnormal state, and special care is devoted to the phenomenon for a hypersonic inlet. In this article, the process of a two-dimensional hypersonic inlet from start to unstart is experimentally demonstrated by recording the corresponding time history of the wall static pressure. Then the pressure signals are analyzed by means of such time-frequency signal-processing methods as short-time Fourier transform (STFT) and wavelet transform (WT). Furthermore, two change-detection algorithms, i.e. the cumulative sum (CUSUM) and generalized likelihood ratio (GLR), are utilized to predict the buzz phenomenon. Results indicate that when a buzz starts, considerable fluctuation of the pressure signals accompanied by clustering power spectrum density (PSD) can be observed and the base frequency increases with the throttling ratio (TR), ranging from 200 Hz to 340 Hz. In addition, ahead of the buzz, due to the regular variation and movement of the large-scale separation bubble existing near the bottom surface of the duct entrance, substantial fluctuation of the wall static pressure can be noted, suggesting the probability of predicting the onset of an inlet unstart. According to the experimental data, the alarm time calculated by CUSUM and GLR are both approximately 220 ms ahead of the big buzz, which verifies the feasibility of the methods adopted.

Key words: hypersonic inlet, unstart, buzz, time-frequency analysis, Fourier transform, wavelet transform