一种多元素多尺度形态非抽样小波分解方法

摘要

Aiming at that the fault signal of motor bearings is always hidden by heavy background noise,a kind of multi-element and multi-scale morphological undecimated wavelet decomposition method is proposed to extract fault feature of rolling bearings.The method is based on the normal framework of morphological undecimated wavelet,combining the filtering character of open-close and close-open mixed operator with the character of impulse feature extraction in gradient operator.Triangle and flat structuring element are used in the two parts of filters,and the efficiency of filter is optimized.The experiment results show that the method is able to filter harmonic wave and noise,and effectively extract impact components.A better performance is achieved compared with traditional morphological undecimated wavelet method.%针对电动机轴承故障信号常被强背景噪声淹没的问题,提出了一种多元素多尺度形态非抽样小波分解方法,并将其应用于滚动轴承故障特征提取中.该方法基于形态非抽样小波的一般框架,结合了形态开闭-闭开混合算子的滤波特性以及形态梯度算子提取信号冲击成分的特点,对该算子的两部分分别使用三角形和扁平形结构元素,使效率最优化.仿真和试验证明,该方法既可以进行谐波与噪声滤除,又可以有效地提取冲击成分,较现有的形态非抽样小波方法有更好的效果.