Study On Extraction Of Fluctuation Pressure In Compressor

Stall inception identify of axial compressor is a difficulty in aviation sector.The aim of the studyis not just how to identify the stall style,more than to understand the mechanism it generated anddeveloped, and then be prevented efficiently. The previous signal processing methods had made greatcontributions in recognizing the stall characteristic, but they are not that enough for stall in theoccurrence and development mechanism.Considering the previous experiment data,the paper use timedomain and frequency domain methods to analyze the stall characteristic,the result show that it isparticularly important for the study. The details of this paper are listed as follows:1. Based on short time Fourier transform,the paper apply feature methods of the maincharacteristic frequency and the main amplitude to analyze the time-frequency domain character ofdifferent stall style and sum up signal frequency and amplitude features in the different states ofstationary, transitional and instability.2. Considering the spike stall,the paper proposes the cross-correlation thinking,it compares thepressure time histories of one blade pitch with that of the identical blade one revolution before andcomputes their correlation coefficient and based on the correlation degradation to provide stallwarning sufficiently.The result shows the coefficient indices are fluctuate in a range in stationarystate,slightly lower nearly stall state,and down to the lowest range below the stationary stage. Thismethod is simple and the best situation0.1seconds earlier can detect the inception.Moreover,thepaper discuss the reason of the coefficient range from0.3to0.5,which find that the actualcross-correlation coefficient values are the blade passing frequency values and the lower300Hzfrequency more than the stall charactic frequency can affect the mutual trend.3. A software about feature analysis of fluctuation pressure on the compressor is developed basedon the MATLAB GUI. This software has lots of practical functions like time-domain feature analysis、feature extraction、cross-correlation coefficient calculation、filter and the display and output ofanalysis results and so on. These functions can meet presupposed requirements. The interface of thesoftware is simple but generous which is convenient for operation. and it raises analyzing efficiencyextremity. Applied to computer simulation and testing data, the results show that the method and theprogram are valid and effective.