Morphological filtering has been extensively applied to rotating machinery diagnostics, whereas traditional morphological operators cannot effectively extract fault-triggered transient impulse components from noisy mechanical vibration signal. In this paper, a framework of generalized compound morphological operator (GCMO) is presented to enhance the extraction ability of impulsive fault features. Further, several new compound morphological operators are developed for transient impulse extraction by introducing the product, convolution, and cross-correlation operations into the GCMO framework. In addition, a novel strategy for selecting the structural element length is proposed to optimize the repetitive impulse feature extraction of the compound morphological operators. The fault feature extraction performance of the developed compound morphological operators is investigated and validated on the simulation signals and measured railway bearing vibration signals, and compared with the combined morphological operators and five existing feature extraction methods. The results demonstrate that the morphological cross-correlation operators are more efficient in repetitive fault impulse feature extraction and bearing fault diagnosis than the combined morphological operators and the comparison methods.

形态滤波已广泛应用于旋转机械诊断，而传统的形态算子无法有效地从嘈杂的机械振动信号中提取故障触发的瞬态脉冲分量。本文提出了一种广义复合形态算子（GCMO）框架，以提高脉冲断层特征的提取能力。此外，通过将乘积、卷积和互相关运算引入 GCMO 框架，开发了几种新的复合形态算子用于瞬态脉冲提取。此外，提出了一种选择结构元素长度的新策略，以优化复合形态算子的重复脉冲特征提取。在仿真信号和实测铁路轴承振动信号上对所开发的复合形态算子的故障特征提取性能进行了研究和验证，并与组合形态算子和现有的五种特征提取方法进行了比较。结果表明，形态互相关算子在重复故障脉冲特征提取和轴承故障诊断方面比组合形态算子和比较方法更有效。