基于模态波理论的压气机失速先兆识别方法

针对压气机气动失速预警存在无法兼顾对渐进型和突尖型失速先兆有效识别的不足，以及提取失速团幅值需要多路传感器信号的局限。分析了压气机失速发展过程中的信号特征；根据压气机失速信号的旋转特性，提出基于单路传感器信号重构周向多路失速先兆信号的方法，讨论了信号重构步长的取值；基于模态波理论，通过对重构信号进行空间傅里叶变换，得到失速扰动信号各阶模态幅值，根据低阶模态的幅值变化实现对渐进型与突尖型失速先兆的识别。以某压气机为例进行仿真，结果表明:当信号重构的步长小于突尖波时间尺度的一半时，重构的多路信号能够复现压气机周向位置失速发展过程，且步长越小，越有利于失速先兆的识别；由模态分解得到的1阶模态幅值的变化能够有效反映失速发展过程；通过合理设置模态幅值的阀值，可以在失速发展前期对失速先兆进行准确识别，从而，基于单路传感器失速信号即可实现对渐进型失速和突尖型失速均能进行预警。 

Identification method of compressor stall inception based on modal wave theory

Targeting the shortages of the compressor aerodynamic stall warning that the existing algorithm cannot identify progressive type stall and spike type stall inception effectively and multi sensor channels are required for obtaining the amplitude of stall cell, the characteristics of signal during compressor stall development were analyzed. And according to the rotation characteristics of compressor stall signal, the method of circumferential multipath stall inception signal reconstruction was proposed based on single channel sensor signal by parallel moving algorithm. The value of signal reconstruction step was discussed. Furthermore, based on the modal wave theory, the stall signal amplitudes of each mode disturbance were obtained by the application of the spatial Fourier transformation on the reconstructed signal. Progressive type and spike type stall precursor were identified according to the amplitude variation of low order modes, thus predicting the occurrence of stall. Taking certain compressor as an example for simulation. Results show that as long as the signal reconstruction steps is less than half the time scale of the spike wave, the reconstructed multipath signal can reproduce the development process of circumferential position compressor stall well, the smellier step means more conducive to identifying stall inception. Changes in the amplitude of first order modal acquired from modal decomposition can effectively reflect the stall development. The stall inception can be accurately identified in the pre stall development by setting reasonable modal amplitude threshold. Therefore, both progressive type and spike type stall enable warning based on a single sensor signal.