论文部分内容阅读
本文叙述一种利用振动噪声来判别机器故障的方法。首先,利用傅里叶变换将机器的振动噪声从时间域变换到频率域。它的对数功率谱为N(f)然后定义谱相关系数为 在相关处理以前将N(f)和N_yy(f)用它们各自的平均值归一化,并减去“值流分量”。 于是,正常机器自己的谱相关系数比正常状态和故障之间的谱相关系数大。机器不同故障的谱相关系数比同样故障的谱相关系数小,但比正常状态和故障之间的相关系数大。如果选择某个阈(谱相关阈),故障将被判别出来。
This article describes a method of using vibration noise to identify machine faults. First, the machine vibration noise is transformed from the time domain to the frequency domain using a Fourier transform. Its logarithmic power spectrum is N (f) and the spectral correlation coefficient is then defined as N (f) and N_yy (f) normalized with their respective averages and subtracted from the “stream component” prior to the correlation process. Thus, the normal machine’s own spectral correlation coefficient is larger than that between the normal state and the fault. The spectral correlation coefficient of different faults of the machine is smaller than the spectral correlation coefficient of the same fault, but larger than the correlation coefficient between the normal state and the fault. If you select a threshold (spectral correlation threshold), the fault will be identified.