A mechanical fault detection strategy based on the doubly iterative empirical mode decomposition

Abstract

Empirical mode decomposition (EMD) has been widely used in the fault detection of rotating machineries. However, the deviation of the local extrema can decline the decomposition performance of the conventional EMD. This paper proposes a fault detection strategy based on the doubly iterative empirical mode decomposition (DIEMD), where the local extrema are fixed during the sifting process. This strategy combines the DIEMD to obtain decomposed intrinsic mode functions (IMFs), correlation analysis to select the representative IMF and envelope spectrum analysis to extract the fault feature frequency. The decomposition performance of the DIEMD was validated through the dual signal analysis. A numerical simulation and experimental investigations on four fault cases (the bearing with the inner race and outer race fault, the gear with broken teeth and the piston pump with cylinder block fault) of rotating machineries were carried out. The amplitude enhancement index (AEI) was defined as an indicator of the performance improvement. Results showed that the proposed strategy is applicable and effective for the fault detection of rotating machineries with AEI being 17.8%, 8.6%, 23.0%, 19.0% and 4.1%, respectively.