改进的局域均值分解在机械故障诊断中的应用

Improved Local Mean Decomposition and its Application in Mechanical Fault Diagnosis

为了准确地提取出故障特征,把局域均值分解算法应用于机械故障振动信号的特征提取中。然而在对信号进行局域均值分解时,由于端点的趋势无法预知,在分解时会污染到整个信号序列,而且滑动平均造成了信号的过平滑处理,导致故障特征不能准确地提取。采用波形匹配解决端点处信号的走势,之后再利用线性插值求得信号的局域均值函数与包络函数,通过在Matlab环境下仿真调幅信号验证了该算法在端点处能很好的保持原有信号的特征;并采用实际的机械振动信号进行实验验证,引入端点效应评价指标,根据分解后的生产函数与原信号的相关系数验证了改进的局域均值分解更能准确地提取故障特征。

In order to accurately extract the fault characteristics, local mean decomposition is applied to mechanical fail- ure vibration signal feature extraction. However, in the local mean decomposition of the signal, the trend of the endpoint can not be predicted that cause contaminate the entire signal sequence, the original moving average of the signal used over-smoot- hing treatment, resulting in failure characteristics can not accurately extract. Waveform matching is introduced to solve the original features of signals at the endpoints, using linear interpolation to get local mean and envelope function, it is verified by simulation amplitude modulated signal in Matlab and has good effect on keeping the original features of signals at the end points; and the introduction of end effect evaluation index, based on the production function decomposition of the correlation coefficient with the original signal, it verifies that the improved algorithm has more effectively extract fault in the mechanical vibration signal.