基于时频周期窗的旋转机械故障诊断方法

周期平稳信号的Wigner-Ville分布是由时频域上若干个离散分量组成，某些分量对故障较为敏感，这些特征分量的检测及提取，对故障诊断十分重要。该文利用旋转机械故障信号及其高阶矩的周期性，通过傅立叶级数对周期平稳信号的分解，认为分解各项在U 0, 2p]区间内两两正交，提出了提取故障特征的时频周期窗观察方法，可以使用时频域局部开周期窗的方法，与故障特征频率项进行周期内相关运算，取出周期自相关函数在故障特征频率分量（调制频率）上的正交投影，从而获得某一故障特征信息，与通常的功率谱分析相比，排除了周期窗外噪声对分析结果的影响，使得特征提取更加精确。通过仿真和实例进行了有效性验证。

A Fault Diagnosis Approach of Rotation Machinery Based on Time-frequency Period Window

The WVD of a cyclostation signal is composed of serial scatter components , and some components named feature components are sensitive to fault. It is very important for fault diagnosis to inspect these feature components. Based on the cyclostation property of high-order moment of rotating machinery fault signal, each Fourier progression element of cyclostation signal is thought as orthogonal at period region U 0, 2p]. Therefore time-frequency period window approach is presented in order to extracting fault feature, and this method exploits the each Fourier progression translation element correlating with fault feature component element within period region to obtain feature component orthogonal projection information. Compared with usual power spectrum method, the analysis result eliminates the noise affect of time-frequency period window outer, and fault feature is extracted more precisely. Finally, the simulation and fault example are demonstrated to show the validity of this approach.