Time Domain Feature Enhancement Model Based on Correlated B-spline Wavelet Base Convolutional Sparse

When a bearing fails, the interference from redundant components weakens signal fault features. It makes early warning and diagnosis of bearing faults a considerable challenge. Therefore, a convolution sparse model based on the correlated B-spline wavelet base is proposed to enhance the time domain impulse feature of the signal. First, an optimization method for B-spline wavelet base parameters is proposed. The method calculates the correlation between the B-spline wavelet bases in the bandwidth and center frequency parameters domains and the original signal. It can also adaptively construct the wavelet base atom based on the maximum correlation parameter. Second, a convolutional sparse algorithm based on the shift-invariance of the signal and the Alternating Direction Method of Multipliers (ADMM) optimization is used to enhance the main features of the signal in the time domain. Third, feature-enhanced sparse parametric envelope spectral analysis identifies the main components of the reconstructed signal in the time domain, which can be used for early warning and diagnosing bearing faults. Finally, the experimental analysis of the periodic simulated signal and fault signals proves that the model has effective feature enhancement capability for the periodic time domain features. Fast spectral kurtosis comparison experiments also verify the advantages of the model.