轴承故障特征自动提取的诊断方法

Diagnosis method for automatic extraction of bearing fault features

为提供一种在噪声环境下准确诊断轴承故障的方法,提出了Hankel非凸鲁棒主成分分析的智能轴承故障诊断方法,重排原始振动信号得到Hankel训练矩阵,采用非凸鲁棒主成分分析和支持向量机方法方法自动提取故障特征与识别故障,利用该方法直接训练含有噪声干扰的滚动轴承和行星齿轮箱故障原始数据集,验证该算法的有效性。结果表明,在高噪声情况下,该方法仍然能够保持在95%以上的故障诊断精度,相比于其他几种对比方法具有明显的优势,其对滚动轴承性能评估是有效的,在噪声环境下具有较高的故障诊断精度。

This paper proposes a method of intelligent fault diagnosis based on Hankel non-convex robust PCA method by which bear faults can be diagnosed accurately in noisy environment.The study includes regrouping the original vibration signals to obtain the Hankel training matrix and extracting and identifying the fault features automatically with non-convex principal component analysis(NCRPCA)method and support vector machine(SVM)method.The effectivity of the study can be verified by training the rolling bearings and the original date set of planetary gearboxes under noise environment.The experimental result shows that high accuracy of fault diagnosis under heigh noise environment keeps above 95 percent,the advantages compared with other experiments is valid to evaluate its performance of the rolling bearings.