| Résumé | Aerodynamic instabilities, such as rotating stall and surge, have long been important considerations in the design and operation of compressors, especially in gas turbine engines. Inflow distortions are the main cause of instabilities in axial compressors. An improved understanding of prestall behavior has implications in compressor design and operation for stall warning and avoidance, and stall control. Fourier analysis including power spectral methods is a useful technique for stationary signals whose statistical properties do not change with time. As observed by many researchers, the flow structure, with traveling waves prior to stall inception, changes with time growing gradually from small amplitude waves to a fully developed stall. Non-stationary signal analysis techniques appear to be attractive to analyze prestall behavior because of the unsteady nature of the flow structure prior to stall. In this report an attempt is made to examine the flow structure of some compressor rig data for the period prior to stall inception by using non-stationary signal analysis techniques. Inflow structures with an axial compressor operating under normal and surge conditions were examined using Short-time Fourier transform, Wigner-Ville transform and Harmonic wavelet analyses. Statistical parameters describing the nature of unsteadiness are extracted with the aim of identifying stall inception and increasing warning times to values useful for engine control. Non-stationary time frequency transforms show promising results in detecting features, in wall static pressure data, which may be pre-cursors to stall inception. The analysis techniques reveal the possible existence of traveling waves with components rotating at various speeds, ranging from 0.5 to 1.5 times the shaft speed. All three transforms examined show that the standard deviations of the coefficients of the transforms can be used to identify significant signal characteristics approximately 0.5 to 1.0 seconds before stall inception. Using STFT, the standard deviation ratio at two frequencies corresponding to 80% and 150% of the shaft speeds can be used in predicting the stall inception. The Wigner-Ville transforms indicate consistent changes only at the frequency corresponding to 80% of the shaft speed. Harmonic wavelet analysis gives useful results at 50% and 80% of the shaft speed. Recommendations are presented for further studies with these and other methods and also other types of stall initiation. |
|---|
