小波脊线解调与两次EMD分解相结合的故障识别方法及应用研究

Research on wavelet ridge demodulation and twice EMD-based fault identification method and its application

提出了一种基于小波脊线解调与两次经验模态分解(EMD)相结合的故障识别方法,用于轴向柱塞泵的故障诊断。结合EMD与小波脊线法在处理非平稳信号方面的优势,首先对故障原始信号进行了EMD分解,利用边际谱发现故障发生时的共振频带范围并据此找出对故障敏感的固有模态函数(IMF)分量,然后对该敏感IMF分量分别进行小波脊线解调和Hilbert解调,最后通过比较两种解调方法解调后敏感分量的时频谱和三维谱图发现,小波脊线解调比Hilbert解调具有更高的时频定位精度和抗干扰能力。此后,分别对小波脊线解调与Hilbert解调后的敏感分量进行EMD再分解,利用所得的各阶二次IMF分量的归一化特征能量来构造特征向量,得到液压泵5种典型状态的样本集,结合K均值聚类算法对故障状态进行识别。研究结果表明,与采用Hilbert解调处理方法相比,利用2次EMD分解与小波脊线解调相结合的故障特征向量提取方法显著提高了故障识别准确率,故障确诊率可高达92%。

A novel fault identification method based on wavelet ridge demodulation and twice empirical mode decom- position （EMD） is proposed, and applied to the fault diagnosis of the axial plunger pump. The method integrates the advantages of EMD and wavelet ridge demodulation in the non-stationary signal processing. Firstly, the original fault signal is decomposed by means of EMD, and the marginal spectrum is used to determine the resonance frequency band range when the fault occurs. According to the marginal spectrum, the fault sensitive intrinsic mode function （IMF） components are found. Then, the sensitive IMF components are demodulated by means of wavelet ridge de- modulation and Hilbert demodulation. Finally, through comparing the time-frequency spectrums and three-dimensional spectrums of the demodulated sensitive IMF components obtained using the two kinds of demodulation methods, it is found that the wavelet ridge demodulation has higher time-frequency locating accuracy and anti-noise ability than Hil- bert demodulation. Furthermore, the sensitive IMF components demodulated with the wavelet ridge demodulation and Hilbert demodulation are decomposed again usintang the EMD method,the normalized feature energies of the second- ary IMF components are used to construct the feature vectors. Then the sample set of the five typical states of the hy-draulic pump is acquired. The K means clustering algorithm is used to recognize the pump fault states. The study re- suits show that compared with the fault identification method adopting Hilbert demodulation, the proposed feature vec- tor extraction method combining twice empirical mode decomposition and wavelet ridge demodulation obviously im- proves the fault correct recognition rate, and the fault correct recognition rate reaches as high as 92%.