基于中值滤波-SVD和EMD的声发射信号特征提取

Feature extraction of acoustic emission signals based on median filter-singular value decomposition and empirical mode decomposition

针对随机噪声和脉冲干扰对经验模态分解(EMD)质量的影响,提出中值滤波和奇异值分解(SVD)联合降噪方法,并将其与EMD分解相结合形成一种新的声发射(AE)信号特征提取方法。首先对原始AE信号进行中值滤波,去除幅值较大的异常值;其次对去除异常值的信号序列进行相空间重构和SVD分解,并针对难以确定重构阶数这一问题,提出奇异值能量差分谱概念,利用谱峰的较大值位置来确定重构阶数,以进一步降噪;最后对降噪信号进行EMD分解,以本征模态函数(IMF)的能量占比作为表征各损伤信号的特征向量。数值仿真和5层胶合板损伤的实测数据表明,该方法不仅能够滤除噪声干扰,提高EMD分解的时效性和准确性,而且能够有效地提取出胶合板AE信号特征,对其损伤类型进行有效地识别。

In order to reduce the influence caused by random noises and pulse disturbances in acoustic emission （AE） signals on empirical mode decomposition （EMD） results and extract the features of AE signals, a new method combining median filter-singular value decomposition （SVD） and EMD is proposed. The outliers in original AE signals are suppressed with median filter at first. Then the outlier removed signal sequences are reconstructed in phase space, and the attractor track matrix is decomposed using SVD for further noise reduction, and the concept of energy difference spectrum of singular value is put forward to determine the reconstruction order of singular value according to its major peak position. Finally, the de-noised signals are decomposed with EMD and the energy ratios of intrinsic mode function （IMF） are used as the characteristic feature to identify the fault signal types. Numerical simulation re- sults and measured data of five-plywood damage show that the proposed method can eliminate noises, improve the time-liness and precision of EMD, and also can extract AE signal characteristics efficiently and identify damage types.