| Abstract | Inflow distortions are the main cause of instabilities, such as stall and surge, in axial compressors. An improved understanding of prestall behavior has implications in compressor design and operation for stall warning and avoidance, and stall control. As observedby many researchers, the flow structure, traveling waves prior to stall inception, changes with time growing gradually from small amplitude waves to a fully developed stall. In a previous report the authors have successfully demonstrated the application of some non-stationary analysis techniques, short time Fourier transforms, Wigner-Ville transforms and harmonic wavelet transforms, to detect features in static pressure data, which are precursors to stall inception. Wavelet analysis techniques are being increasingly used in non-stationary signal analysis, especially in mechanical signature analysis, to detect and diagnose faults in machines. In this report the flow structures of some compressor rig data have been analyzed for the period prior to stall inception with wavelet analysis techniques. Both continuous and discrete wavelet transforms have been used in analyzing the static pressure signals. Statistical parameters describing the nature of unsteadiness in static pressures on the compressor case are extracted with the aim of identifying robust parameters correlated to stall inception and thereby increase stall-warning times to values useful for engine control. The analysis methods reveal the existence of traveling waves in the flow rotating at a wide range of speeds corresponding to 15% to 140% of shaft speed with a dominant mode near 80% of the speed. The results are in broad agreement with the earlier investigations of the authors based on short time Fourier transforms, Wigner-Ville transforms and harmonic wavelet transforms. The developing stall features can be detected using the ratio (pre-cursor: normal) of standard deviations of the wavelet transform coefficients in both discrete and continuous wavelet analysis. With continuous wavelet transforms only, the developing stall features can also be detected using the kurtosis ratio (pre-cursor: normal) of the wavelet transform coefficients. The ratio of the standard deviations of the wavelet coefficients appears to be a robust parameter in detecting features in wall pressure data, which are pre-cursors to stall inception. The techniques reveal the possibilities of predicting stall inception conditions one second before the occurrence of serious flow breakdown and surge. |
|---|
