Detection of Compressor-Rotating-Stall-Inception Signals Based on a Framelet-diffused Algorithm

Analyzing the dynamic pressure signals generated by compressors is extremely difficult because the clearance vortex, separating vortex, and wake inside them interfere and mix with one another generate noise in the frequency bands of the signals, especially when the compressor is near an unsteady state. To solve this problem, we developed an improved framelet-diffused algorithm based on the Ron–Shen framelet filter bank in this study for signals denosing. The method is tested on six analog signals and demonstrates good noise suppression, even for a poor signal-to-noise ratio. The robustness and reliability of our algorithm is reviewed by processing the experimental dynamic pressure signals acquired from a transonic compressor, the results realize that it can capture compressor rotating stall inception signals before the compressor stalls when conventional wavelet method could not get useful information, which is partial-surge signal at 98% design rotational speed, consistent with real-world physical properties of the tested compressor.