改进时域盲解卷积算法在轴承故障诊断中的应用

IMPROVED TIME DOMAIN BLIND DECONVOLUTION ALGORITHM IN BEARING FAULT DIAGNOSIS

针对实际工业现场强背景噪声、干扰源多、旋转机械故障盲提取算法的不足,为了有效提取并分离出轴承的故障特征,提出一种基于广义形态滤波和改进KL距离相结合的改进时域盲解卷积故障特征提取算法。首先利用广义形态滤波提取信号中重要特征频率;然后利用正交匹配追踪算法去除滤波后信号的周期成分;最后,使用改进KL距离计算各分量的距离,通过模糊C均值聚类获得分离信号。实验仿真和故障滚动轴承声信号及振动信号的分析结果表明,该方法能够有效提取滚动轴承故障特征。

In order to extract fault feature of signal. An improved blind deconvolution algorithm which based on generalized morphological filtering and improved KL distance clustering methods was proposed to deal with industrial field noise,multi interference sources and disadvantage of blind extraction algorithm. First,the generalized morphological filter was used to extract the characteristic signal of observation signal. Then,the orthogonal matching pursuit algorithm was used to remove the period component of signal after being filtered. Finally,the improved KL distance was used to calculate distance of each component and obtain the separated signal by fuzzy C cluster. The results of computer simulation and real rolling bearing signals analysis show that this proposed method is quite effective.