Time-Frequency Slice-Sparsity Extracting Transform With Application to Rotor Fault Diagnosis

The composite nonstationary signals that include both harmonic and impulsive features usually indicate the occurrence of abnormal mechanical faults, especially rotor rub-impact fault. However, it is still challenging for the existing methods to clearly describe feature information contained in these two modes at the same time. Therefore, this article proposes a time-frequency slice-sparsity extracting transform (TFS2ET) technique, which achieves bilateral extracting operation by fusing sparsity of the time and frequency slices and synchronous representation of harmonic and impulsive parts. This work can not only adaptively improve the overall energy concentration and accuracy but maintain signal reconstruction ability to extract the critical components indicating specific faults, which are quantified through indicators, such as Rényi entropy (RE), Earth mover’s distance (EMD), and output signal-to-noise ratio (SNR). The effectiveness of the proposed method is validated by numerical simulation and experimental analysis.