Experimental investigation on axial compressor stall phenomena using aeroacoustics measurements via empirical mode and proper orthogonal decomposition methods

In this work, we conduct experimental investigations on a single-stage axial compressor to shed lights on the dynamic stall phenomena via aeroacoustic measurements. For this, 8 acoustic pressure sensors are installed equally around the circumference of the compressor intake. The acoustic pressure data are simultaneously logged in real-time. Classical and conventional Fourier-transform based methods reveal that the compressor stall will occur beyond a critical pressure ratio or a flow coefficient as illustrated on a compressor map. However, there is no warning precursor obtained from the conventional signal analysis methods. Further investigations are conducted using a number of advanced signal processing techniques, such as empirical mode decomposition (EMD) and proper orthogonal decomposition (POD) and continuous wavelet (CW) methods. EMD analysis reveals that the compressor stall is corresponding to a low-frequency intrinsic mode function (IMF). It is growing rapidly from negligible amplitude random disturbances to limit cycle oscillations. POD shows that the number of the dominant acoustic modes contributing to more than 98.5% of the total fluctuation energy is changed from 3 to 7, when the stall occurs. This is insightful for low-order modelingof the compressor stall phenomena. Finally, applying CW transform could provide a warning of 0.05 s precursor on the tested axial compressor. In general, the present work opens up an alternative approach to study the dynamic physics of a compressor stall by applying an array of acoustic sensors with proper advanced data-processing methods implemented. (c) 2021 Elsevier Masson SAS. All rights reserved.

Automated summary

Experimental investigations on dynamic stall phenomena in a single-stage axial compressor were conducted using aeroacoustic measurements, revealing that the compressor stall corresponds to a low-frequency intrinsic mode and an increase in the dominant acoustic modes during stall. Traditional signal analysis methods did not provide any warning precursors, while advanced signal processing techniques such as EMD and POD analysis showed precursors before stall. CW transform could give a 0.05 s warning precursor.