振动信号短时分析方法及在机械故障诊断中的应用

研究了低信噪比情况下提取周期性冲击故障信号特征的短时分析方法。定量分析了短时能量方法的抑噪特征，指出信号的局部特征越明显则短时能量处理越能显著的提高信噪比，短时能量函数中实际有用信号能量与实际噪声信号能量的比值可作为一特征量使用。建立了信号的物理含义相同，反映的只是信号功率谱重心的变化，与信号能量的变化无关，故不适宜于分析受强噪声干扰的时变故障信号。最后以滚动轴承故障信号分析实例验证了短时能量分析

Short-Time Processing of Vibration Signal and Its Application in Machine Fault Diagnosis

For diagnosing localized defects in the components of rotating machinery, the utility of short-time processing techniques to analyze the time-development signal of periodic impulses with background noise is studied. From the point of view of signal detection, the anti-noise character of short-time energy processing technique is qualitatively analyzed, and it is shown that this technique can improve the SNR due to its enhancement on the local feature of time-dependent signal. A specific parameter, namely the ratio of the energy of useful component to that of noisy componen in the short-time energy function sequence, is proposed to monitor the formation and development of the defects. The qualitative relationships among the barycenter of the power spectrum, the auto-correlation coefficients and the average zero crossing rate are established. It is shown that both the short-time auto-correlation coefficient at one lag and the short-time average zero crossing rate only reflect the time-variation of the frequency contents of a signal, so they are poor in revealing the time-dependent feature of the band-filtered contaminated signal. The effectiveness of the short-time energy processing technique for machine fault diagnosis is verified by experimental data.