机械振动信号自适应多尺度非线性动力学特征提取方法研究

针对机械振动信号的故障特征提取问题,提出了基于独立变分模态分解与多尺度非线性动力学参数的特征提取方法。①提出频谱循环相干系数选取匹配波形对机械振动信号进行端点延拓后再进行VMD分解得到不同频率尺度的IMF分量;②根据互相关准则选取有效的IMF分量进行核独立成分分析,分离出相互独立的有效故障特征频带分量;③计算各独立分量的复合多尺度模糊熵偏均值,并利用正交变换将独立分量正交化后构造多维超体,进而利用多维超体体积定义并计算信号的双测度分形维数,从而获得多尺度非线性动力学特征参数,实现机械故障诊断。仿真和实验结果表明:所提方法可有效抑制VMD分解的端点效应和模态混叠,信号分解效果好,特征参数分类精度高,极大地提高了机械故障诊断准确率。

Adaptive multi-scale method for the non-linear dynamic feature extraction of mechanical vibration signals

Aiming at the fault feature extraction of mechanical vibration signals, a feature extraction method based on the independent variational mode decomposition(VMD) and multi-scale nonlinear dynamic parameters was put forward.The spectral cyclic coherence coefficient was proposed to select the matching waveform which was used to complete the endpoint extension for the mechanical vibration signal.The extended signal was decomposed into some intrinsic mode functions (IMFs) in different frequency scales by using the VMD.The effective IMFs were selected according to the cross-correlation criterion and the independent components with effective frequency band were separated from the effective IMFs by using the kernel independent component analysis.The composite multi-scale fuzzy entropy partial mean of each independent IMF was calculated.The orthogonal transform was used to orthogonalize independent IMFs to construct a multi-dimensional hyperbody, and its volume was used to define and calculate the dual measure fractal dimension of the vibration signal.Thereby,the multi-scale nonlinear dynamic parameters were obtained to achieve mechanical fault diagnosis.The simulation and experimental results show that the proposed method can effectively suppress the end effect and mode mixing in the VMD,which improves the effect of signal decomposition; the feature parameters have higher classification accuracy, which greatly improves the accuracy of mechanical fault diagnosis.