Experimental mode decomposition investigation on 3-stage axial flow compressor stall phenomena using aeroacoustics measurements

Abstract

Safely operating compressors is quite critical. Rotating stall is highly undesirable and affects negatively the performance and stability of aero-engine compressors. It is important to identify a precursor of such rotating stalling phenomena observed on the compressor. In this work, we conduct experimental measurements on a 3-stage axial flow compressor by using 8 acoustic pressure sensors. Emphasis is being placed on applying different mode decomposition methods such as Proper Orthogonal Decomposition (POD) and Empirical Mode Decomposition (EMD) on these acoustics data to shed lights on the stall phenomena. Comparison is then made between these methods and conventional means like continuous wavelet (CWA), and PSD (Power Spectrum Density) analyses to achieve a timely detection of stall inception and the energy distribution before and after stall occurred. It is found from POD that the dominant acoustic modes contributing to more than 90% of the total fluctuation energy are changed from 3 to 4, when the stall occurred. The EMD investigation shows that the compressor stall is associated with a few low-frequency IMFs (intrinsic mode function). Time evolutions of such IMFs are observed to grow from negligible amplitude disturbances into limit cycles. Additionally, applying CWA reveals that the stall cell is originated from the first stage and then propagates circumferentially and axially to the second and third stage. Finally, PSD analysis shows that the stall frequency related to the spike-type stall cell exists only on the first stage. This is different from the blade passing frequency as observed on all stages of the compressor. In summary, the present work offers alternative acoustic measurements to examine the dynamic instability of a compressor stall with advanced signal-processing techniques applied.