几种非平稳随机信号分析方法的比较与应用

Comparison and Application of Three Methods for Analyzing Random Non-stationary Signals

比较了短时傅里叶变换、小波变换、Cohen类时频分布等三种典型非平稳随机信号分析方法的性能,从理论上讨论了三种典型方法的优缺点,并分析了其他特性.针对非平稳随机信号中难以检测却又最常见的高阻接地信号,通过仿真对比,展示了三种方法的应用效果.结果表明:Cohen类时频分布能够从时间和频率两个维度更充分地揭示高阻接地故障类非平稳随机信号的特性;尤其是Choi-Williams分布,其较好的时频聚集性和抑制交叉项的效果有利于特征的进一步提取分析,能为故障诊断提供更为丰富的早期数据.

The performance of the three typical non-stationary random signal analysis methods of short time Fourier transform, wavelet transform, and Cohen time-frequency distribution is compared and the theoretical advantages and disadvantages of the three methods and their other properties are discussed and analyzed. In light of the high resistance ground signals which are difficult to detect and also very common, the application effect of the three methods are shown through simulation comparison. The results show that the Cohen time-frequency distribution method can more fully reveal the characteristics of high resistance grounding fault types of non-stationary random signals from the two dimensions of time and frequency, especially the Choi-Williams distribution whose good time-frequency resolution and cross term suppression effect can facilitate further feature extraction for analyses and provide rich early-time data for fault diagnosis.