A double impulsiveness measurement indices-bilaterally driven empirical wavelet transform and its application to wheelset-bearing-system compound fault detection

A wheelset-bearing system is one of the most critical subsystems in a high-speed train, and the detection and diagnosis of its compound faults have important sense to the service safety of the high-speed train. To address three related limitations of the single impulsiveness measurement index (SIMI)-unidirectionally driven (UD) empirical wavelet transform (SIMIUDEWT) in detecting compound faults of wheelset-bearing system, (1) a single SIMI cannot simultaneously locate the centric positions of those sidebands with low and high-density impulsive series, (2) unidirectionally driven EWT generates error segment boundaries of EWT, and (3) the scale brake, a novel compound-fault detection method known as double impulsiveness measurement indices (DIMI)-bilaterally driven (BD) empirical wavelet transform (DIMIBDEWT), is proposed. First, the centric positions of the sidebands around the resonant frequencies that carry low and high-density impulse series are preliminarily estimated using the sparsity index and envelope spectra kurtosis (ESKT), respectively. Second, the lower-upper boundaries of the sidebands with low and high-density impulse series are determined by executing a left-right bidirectional search under the respective drives of the sparsity index and ESKT of the signals contained in the different decomposition bandwidths (ESKTDB). Third, the segment boundaries of EWT are determined by merging the sideband lower-upper boundaries. Finally, the signals contained in the sideband lower-upper boundary pairs of EWT are further demodulated to detect compound faults in the wheelset-bearing system. The proposed DIMIBDEWT is validated by simulation and bench and running tests.