EEMD与SVD的齿轮故障诊断技术研究

Study on Rolling Bearings Fault Feature Extraction Based on Improved EEMD and SVD

对齿轮信号奇异值分布规律进行研究,提出一种EEMD-SVD差分谱组合模式。对原始信号做集合经验模态分解得到一系列固有模态分量,对其进行有效的筛选并且重构,对重构的信号构造Hankel矩阵,再通过SVD对矩阵做正交分解,利用奇异值差分谱来选择奇异值进行SVD重构,由此实现对弱故障特征信息提取。从齿轮信号的处理结果看出,该方法对复杂信号中的弱故障特征信息具有优良的提取效果。

Based on distribution law of singular values for gear signal, a new method of EEMD-SVD difference spectrum is proposed. In this method, an original signal is decomposed by EEMD and a group of intrinsic mode functions are obtained, which is selected and reconstructed effectively. Reconstructed signals are used to create a Hankel matrix, and then SVD operation of each matrix is made to obtain its orthogonal decomposition results. Furthermore, the singular values are selected and then by dint of different spectrum of singular value the SVD reconstruction is achieved. The faint feature information can be extracted through these procedures. According to the result of disposing gear signal, the proposed method demonstrates the excellent effect on the extraction of faint fault feature information from the complicated signal.