Reliability analysis for stall warning methods in an axial flow compressor

Abstract

Abstract The reliability of widely studied four stall warning methods, namely, auto- and cross-correlation, Root-Mean-Square (RMS), and fast wavelet analysis, were experimentally studied in a low-speed axial flow compressor. Unsteady pressures used for stall warning were measured by time-resolved sensors mounted on the casing along axial and circumferential spatial resolution. Three influence factors of rotational speed fluctuation, circumferential and axial location of pressure sensors for the stall warning reliability were analyzed and discussed. The rotational speed fluctuation that inevitably exists in the actual aero-engine would lead to the phase shift of phase-locked pressure data and thus deteriorates the stall warning reliability using the auto-correlation and RMS analysis methods. The circumferential location of pressure sensors would affect the reliability of proposed stall warning methods due to the uneven tip clearance distribution. Auto-correlation, RMS, and fast wavelet methods are more effective at the large tip clearance location than at the small tip clearance location. For the cross-correlation method that needs two symmetrically-arranged pressure sensors, the negative effect caused by the circumferential unevenness of tip clearance size can be avoided. The ideal axial location of the sensor installation for stall warning is near the rotor leading edge where the strong tip leakage flow fluctuation exists. In addition, considering the actual complex inflow condition, the stall warning methods reliability with inlet circumferential distortion was studied. In the distorted region, the adverse effect of distortion on stability can be detected by above four methods. In the undistorted region, only cross-correlation coefficient correspondingly decreased to reveal the stability margin deterioration caused by circumferential distortion. Considering the random circumferential position of distortion, cross-correlation method was proved to have a promising application compared to others.