Marine Diesel Engine Fault Diagnosis by Using an Improved Hilbert Spectrum

Hongkun Li,Peilin Zhou,Xiaojiang Ma
DOI: https://doi.org/10.5957/jsr.2006.50.4.378
2006-01-01
Journal of Ship Research
Abstract:Vibration signal analysis is a useful method for recognizing the pattern of a machine's working condition. However, it is difficult to recognize nonstationary and nonlinear vibration signal patterns satisfactorily with the traditional Fourier spectrum method. This paper introduces a novel time-frequency distribution method: an improved Hilbert spectrum (HS) for nonstationary vibration signal analysis, that is, applying a wavelet packet de-noise method as a preprocess of the HS. The HS is developed according to instantaneous frequency analysis by using a new kind of signal analysis method called empirical mode decomposition (EMD), which is highly accurate in analyzing various nonstationary and nonlinear signals. Due to a self-adaptive decomposition process, noise has a great effect on the accuracy of the EMD process and the corresponding HS. This has limited the application of the HS on real vibration analysis. In this study, a wavelet packet de-noising technique is employed as a preprocessing to improve the signal-to-noise ratio and the accuracy of the HS. Experimental data of a marine diesel fuel injection system are used to evaluate the improved methodology for system pattern recognition and fault diagnosis.
What problem does this paper attempt to address?