Fault diagnosis of rotating machinery based on noise reduction using empirical mode decomposition and singular value decomposition

Vibration signals collected from a faulty rotating machine include in general impulse information reflecting fault types, irrelevant vibration components caused by other normal mechanical parts, and other environmental noise. Cleaning the obtained vibration signals can prove practical significance for the fault diagnosis of rotating machinery. To address this issue, this paper proposes a new fault diagnosis method based on noise reduction technology using empirical mode decomposition (EMD) and singular value decomposition (SVD). In this approach, EMD is first applied to decompose the collected vibration signal into a set of intrinsic mode functions (IMFs) and residual signal. Then the first several IMFs including bearing characteristic damage frequencies (CDFs) and higher frequency components are selected to do further noise reduction by SVD for features, and the other remaining decomposition components of EMD are abandoned as noise. Finally, the fault diagnosis of rotating machinery is realized by these obtained features using a support vector machine (SVM) model. Experimental results testify that the proposed method is effective for mechanical fault diagnosis.