基于改进连续小波变换增强非凸正则项稀疏分解的滚动轴承变转速故障冲击特征提取方法

To extract the nonperiodic fault transients of rolling bearings under varying speed with high fidelity, a method termed nonconvex sparse representation enhanced by modified continuous Morlet wavelet transform is proposed. A waveform adjusting factor is introduced which enables the modified Morlet wavelet to well match the fault impulses with different oscillating properties, and an index termed angular envelope harmonic to noise ratio is developed based on which the waveform adjusting factor and threshold are optimized. The modified continuous Morlet wavelet transform enjoys higher coefficients sparisity in decomposing the vibration signal compared with discrete wavelet transforms. The sparse representation model is then fabricated via combining the modified continuous Morlet wavelet transform with the nonconvex penalty function, and nonperiodic fault transients are extracted via further solving the sparse model. The effectiveness of the proposed method is validated via analysing both simulation and experimental data, as well as compared with traditional thresholding denoising, frequency band filtering, and tunable Q-factor wavelet transform enhanced sparse representation methods. The results show that the proposed method could effectively extract the nonperiodic fault impulses of rolling bearings under time-varying speed with high fidelity.