基于尺度共生矩阵的滚动轴承故障诊断研究

Investigation of rolling bearing fault diagnosis based on scale-based concurrent matrix

基于滚动轴承振动信号的各种三维或二维谱图中包含的不同故障信息的客观现实,通过二维小波变换在不同尺度空间下构造的小波共生矩阵提取了谱图的纹理特征向量.利用灰关联分析表征这些纹理特征的不同发展态势从而实现滚动轴承的故障诊断.对实测滚动轴承不同状态故障数据的分析表明:该方法具有较高的故障模式分类精度;随着故障尺寸的增加,由于轴承各部件的相互影响诊断正确率会有所降低.同时研究表明对于特定的诊断方法是否进行特征向量归一化需区别对待.

Based on the fact that rolling bearing vibration signal＇s three-dimensional（3-D） graphics and two-dimensional（2-D） graphics contain different fault information,the texture feature vectors were extracted from different scale-based concurrent matrixes which were received from 2-D wavelet translation.The gray degree association was employed to express these texture feature space geometry similarity.The method was used to analyze different bearing defects＇ real test vibration data.The result shows that this means can get high pattern classification accuracy,but it will decrease with growing malfunction size because of influences of rolling bearing elements.It＇s also found that the normalization of texture feature vector is not reasonable to some diagnosis ways.