Method of fault diagnosis of rolling bearing based on adaptive mathematical morphology

DOI：10.7511/dllgxb201803003

 作者 单位 齐咏生,张双龙,高胜利,李永亭,王林

作为一种非线性信号处理方法，数学形态学法对信号的特征提取完全在时域中进行，与其他非线性非平稳的信号处理方法相比，它具有幅值不偏移、不衰减等显著优点．基于此，提出一种自适应的数学形态学和谱相关分析相结合的轴承故障诊断方法．该方法通过基于信号的三角结构元素和非单一形态学开闭运算对已知故障信号加以训练，自适应得到各故障类型的结构元素高和最优开闭运算加权因子，构建形态学模型；之后将测试信号通过形态学模型进行特征提取，并与训练信号进行频域内相关性分析；最终根据相关系数大小识别故障．以西储大学轴承故障数据为例，表明了该方法不仅能识别出不同类型的故障，而且还能识别不同损伤等级的故障，相比传统的方法识别率和可靠性有所提高．

As a nonlinear signal processing method, the mathematical morphological method extracts the signal completely in the time domain. Compared with other nonlinear and non-stationary signal processing methods, it has the advantages of non-deviation and non-attenuation in amplitude. A method of bearing fault diagnosis based on adaptive mathematical morphology and spectral correlation analysis is proposed. This method uses the signal-based triangular structuring elements and non-single morphological opening and closing operation to train the known fault signal to get the average of the triangular structuring elements′ heights and optimal opening and closing operation weighting factors of each fault type. Then, the two kinds of means are applied to the unknown signal by similar morphological processing, and the gotten signal and the trained one do a correlation analysis in frequency domain. Finally, the fault type is confirmed by the size of correlation coefficient. The bearing data from Case Western Reserve University (CWRU) is taken as an example, it is shown that the method can not only accurately identify different types of faults, but also accurately identify different damage levels, and compared with the traditional method, the recognition rate and reliability are improved.