Abstract:

For the problem of rolling bearing fault type and damage degree with manual intervention, a new fault diagnosis method based on Self-Adaptive Manifold Learning（SAML） is proposed. This algorithm extracts fault features of vibration signal by means of Ensemble Empirical Mode Decomposition（EEMD） and bispectrum analysis, constructs texture feature matrix of fault information by texture analysis method, and reduces the dimension of high-dimensional texture feature matrix by adaptive manifold learning method. The whole process can remove noise well, select parameters adaptively, and have good clustering performance and complex signal processing ability. The experimental results show that this method can distinguish different fault types well, and have a good classification performance in judging the fault degradation degree of inner ring fault, outer ring fault and rolling element fault.

Key words: self-adaptive, manifold learning, ensemble empirical mode decomposition, bispectrum analysis, texture feature construction, fault diagnosis

摘要：

针对人工干预的旋转轴承故障类型及损坏程度诊断问题，提出了一种基于自适应流形学习的故障诊断新方法。该算法借助集合经验模态分解和双谱分析提取振动信号的故障特征，用纹理分析法构建故障信息的纹理特征矩阵，通过自适应流形学习的方法对高维纹理特征矩阵进行降维。整个过程能够很好地去除噪声，同时自适应选择参数，具有很好的聚类性能和复杂信号处理能力。实验结果表明该方法能够很好地区分不同的故障类型，同时在区分内圈故障、外圈故障、滚动元素故障退化程度方面也有着较好的性能。

关键词: 自适应, 流形学习, 集合经验模式分解, 双谱分析, 纹理特征构造, 故障诊断