基于双谱熵的齿轮裂纹故障特征提取

Fault Feature Extraction for Gear Crack Based on Bispectral Entropy

针对裂纹故障导致齿轮振动信号非高斯性变化这一特点,提出采用双谱熵对信号非高斯成分在双频域内的分布形态进行定量描述,并据此提取故障信息,得到裂纹产生期、扩展期的特征趋势。结果表明,双谱熵不基于信号能量信息,受非故障因素影响小,而且能有效抑制高斯噪声,同时又对微弱故障十分敏感。研究结果为后续故障诊断与趋势预测提供了新的有效方法。

Due to gear crack leads to the change of non--Gaussian characteristics of vibration signals, bispectral entropy was proposed to deal with the signals and quantitatively describe the distribution morphology of non--Gaussian components in bifrequency domain. The fault information was extracted by this method, and the feature trend of gear crack in its generation and extension period was obtained. The results show that, bispectral entropy is not based on the information of signal energy, and it is less influenced by the non--fault factors. Meanwhile it can inhibit the Gaussian noise efficiently and be sensitive to the weak fault. So bispectral entropy supplies a new efficient path for the follow--up fault diagnosis and trend prediction.