基于改进VMD的航空发动机故障信号特征提取

从冗长、非线性、非静态的航空发动机信号中提取故障特征非常困难,使得航空发动机状态监测(CM)成为最具挑战性的任务之一。针对原有的变分模态分解(VMD)的特征提取方法,对模数K和滤波器频率带宽a均采用默认值的问题,将导致特征提取不准确,小波变换的结果不可靠。文章提出了一种基于优化VMD的精确特征提取方法。利用包络谱曲线确定最佳模态数K,利用信号能量分配比(SEDR)确定最优带宽a,实验结果表明,利用文中提出的优化策略,成功地提取了小波变换CM信号中的故障相关特,验证了改进VMD方法的有效性。

Feature Extraction of Aeroengine Fault Signal Based on Improved VMD

It is very difficult to extract fault features from redundant, non-linear and non-static aero-engine signals, which makes condition monitoring(CM) one of the most challenging tasks. In view of the original feature extraction method of Variational mode decomposition(VMD), the default values of modulus K and filter frequency bandwidth a are adopted, which will lead to inaccurate feature extraction and unreliable results of wavelet transform. This paper presents an accurate feature extraction method based on optimized VMD. The optimal mode number K is determined by the envelope spectrum curve, and the optimal bandwidth a is determined by the signal energy allocation ratio(SEDR). The experimental results show that the fault correlation characteristics of CM signals based on wavelet transform are successfully extracted by the proposed optimization strategy, and the effectiveness of the improved VMD method is verified.