基于EEMD的单通道盲源分离在轴承故障诊断中的应用

Application of Single-channel Blind Source Separation Based on EEMD in Bearing Fault Diagnosis

针对一维观测矩阵的极度欠定盲分离模型,结合盲源分离和总体经验模式分解的优点,利用总体经验模式分解将单通道信号转化为固有模态矩阵,重组观测矩阵,再通过近似联合对角化实现信号的盲分离。数据仿真说明该方法能提取低信噪比下的轴承故障信息。实验中,对2种不同故障的轴承进行故障诊断,从而进一步证明了该方法的有效性。

For the problem of extremely underdetermined BSS, which had only one dimensional observing matrix, an algorithm which combining the advantages of EEMD and BSS was used. First, the single-channel was decomposed into intrinsic mode matrix by EEMD, then observing matrix was rebuilt and the estimated sources by joint approximate diagonalization of eigenmatrix was obtained fi- nally. The simulation results show that this method can extract the bearing fault information under low SNR. In experiments, two bearings with different faults were diagnosed, and further confirmed the effectivity of this method.