Difference equation based empirical mode decomposition with application to separation enhancement of multi-fault vibration signals

Empirical mode decomposition (EMD) has been applied to various applications in signal processing. However, EMD is susceptible to close mode characteristic frequencies and noise, resulting in the problem of mode mixing. The performance of multi-fault detection in gearboxes will be significantly degraded due to mode mixing in the vibration analysis. Hence, this paper presents a new method to address the mode mixing problem in EMD based gearbox multi-fault diagnosis. In this new method, the differential operation is introduced into the decomposition of the intrinsic mode functions. The decomposition ability of close frequency bands can be improved by the differential operation, and hence, the differential EMD can better identify the modes with close characteristic frequencies than its non-differential counterpart. In addition, time synchronous averaging (TSA) is combined with the differential EMD to address the noise issue. Thus, the proposed TAS and differential EMD based method (TDEMD) can solve the mode mixing problem to provide effective multi-fault detection for gearboxes. The TDEMD has been tested experimentally using vibration data collected from a gearbox with concurrent defects on two different gears. Results showed effective detection of gear multiple faults.