A stall diagnosis method based on entropy feature identification in axial compressors

Abstract

A stall diagnosis method based on the entropy feature extraction algorithm is developed in axial compressors. The reliability of the proposed method is determined and a parametric sensitivity analysis is experimentally conducted for two different types of compressor stall diagnoses. A collection of time-resolved pressure sensors is mounted circumferentially and along the chord direction to measure the dynamic pressure on the casing. Results show that the stall and prestall precursor embedded in the dynamic pressures are identified through nonlinear feature perturbation extraction using the entropy feature extraction algorithm. Further analysis demonstrates that the prestall precursor with the peak entropy value is related to the unsteady tip leakage flow for the spike-type stall diagnosis. The modal wave inception with increasing amplitude is identified by the considerable increase of the entropy value. The flow field in the tip region indicates that the modal wave corresponds to the flow separation in the suction side of the rotor blade. The warning time is 100–300 rotor revolutions for both types of stall diagnoses, which is beneficial for stall control in different axial compressors. Moreover, a parametric study of the embedding dimension m, similar tolerance n, similar radius r, and data length N in the fuzzy entropy method is conducted to determine the optimal parameter setting for stall diagnosis. The stall warning based on the entropy feature extraction algorithm provides a new stall diagnosis approach in the axial compressor with different stall types. This stall warning can also be adopted as an online stability monitoring index when using the concept of active stall control.