基于自适应多尺度形态梯度与非负矩阵分解的轴承故障诊断

Bearing fault diagnosis based on adaptive multi-scale morphological gradient and non-negative matrix factorization

轴承故障诊断的关键步骤是信号处理与特征参数提取。提出采用自适应多尺度形态梯度算法对轴承振动信号进行处理,综合利用小尺度下能保留信号细节和大尺度下抑制噪声的优点,可有效地提取振动信号中反映轴承状态的冲击分量;在此基础上提出采用非负矩阵分解技术对信号进行压缩,计算用于轴承故障诊断的特征参量。采用轴承在七种状态下的振动信号对所提出的信号处理和特征参数提取方法进行验证,结果表明:与传统的信号处理与特征参量提取方法相比,本文提出的方法具有更高的轴承故障分类精度,为准确判断轴承工作状态提供了一种行之有效的新方法。

Signal processing and feature extraction are two of the most significant steps for bearing fault diagnosis.Here,the adaptive multi-scale morphological gradient( AMMG) algorithm was employed to extract impulsive components hiding in vibration signals of bearings,so as to keep the detail of a signal under small scales and depress noise under large scales. Furthermore,the non-negative matrix factorization technology was utilized to calculate the features of the signal processed with AMMG for bearings fault diagnosis. The vibration signals acquired from bearings under 7 states were employed to validate the proposed signal processing and feature extraction methods. Test results demonstrated that the proposed methods are superior to the traditional signal processing and feature extraction ones.