基于振动图像纹理特征识别的轴承故障程度诊断方法研究

Bearing fault severity assessment based on vibration image

针对轴承故障诊断中故障分类研究多,故障程度研究少,振动图像信息丰富得不到充分利用问题,提出利用振动图像纹理特征识别技术进行轴承故障程度诊断方法。该方法先对轴承振动响应信号进行EMD-形态差值滤波处理,后将滤波后信号转换为双谱等高线图,利用灰度三角共生矩阵得到双谱图形纹理特征,应用主成份分析法从纹理特征参数中提取轴承故障程度特征参量,用支持向量机进行模式识别。实验结果表明该方法能有效区别轴承外圈、内圈及内外圈的故障严重程度,可为旋转机械故障程度诊断提供新方法。

The knowledge about bearing fault degree identification is still not much up to now, while the abundant information included in vibration image has not yet been used fully. So, a method of fault degree identification of bearing using vibration image was proposed. The original vibration signals were de-noised with EMD-morphology filter, and then converted to bispeetrum contour image. By using gray-level co-occurrence matrix and principal component analysis, character parameters for assessment of fault severity were acquired. At last the fault degree was diagnosed by support vector machine. The results of experiments show that the method can diagnose the fault degree of bearing effectively, and it provides a new diagnosis approach for the fault degree identification of rotating machinery.