A novel bearing fault diagnosis on basis of empirical mode decomposition (EMD) and improved Chebyshev distance is presented. After normalization, each group of sample data is divided into 10 equal parts on average. EMD is used to decompose the equal part signal into several eigenmode functions, and the first five intrinsic mode function (IMF) components are retained and transformed into symmetrical in polar coordinates by symmetrized dot pattern (SDP) method, each SDP image is processed via binarization and localization, then the local SDP images are averaged to obtained the mean image as benchmark. The maximum eigenvalue of the average matrix after denoising is computed, in this way the improved Chebyshev distance is constructed and bridges the gap between the local matrix of each IMF component and the average matrix. Using the improved Chebyshev distance of IMF1 as feature, this method can effectively diagnose the faults of rolling bearing. Finally, test experiments are carried out to verify the accuracy and robustness of present approach.

提出了一种基于经验模态分解（EMD）和改进的切比雪夫距离的轴承故障诊断方法。归一化后，每组样本数据平均分为10个相等的部分。EMD用于将等份信号分解为几个本征模函数，并通过对称点阵图（SDP）方法将前五个本征模函数（IMF）分量保留并转换为极坐标对称，通过二值化处理每个SDP图像定位后，将本地SDP图像平均以得到平均图像作为基准。计算去噪后平均矩阵的最大特征值，以此方式构造改进的切比雪夫距离，并弥合每个IMF分量的局部矩阵与平均矩阵之间的间隙。该方法以改进的IMF1 Chebyshev距离为特征，可以有效地诊断滚动轴承的故障。最后，进行测试实验以验证本方法的准确性和鲁棒性。