基于自适应Laplace小波周期字典的滚动轴承弱故障特征提取方法OA

The weak fault features induced by early-stage damage in rolling bearings are often deeply buried under random interference and strong background noise.The traditional sparse representation meth-ods suffer from limited capability to perceive signal periodicity and a lack of parameter adaptivity,resulting in unstable extraction accuracy.To address these issues,a sparse representation method based on an adap-tive Laplace wavelet periodic dictionary combined with segmented clustering regularized orthogonal matc-hing pursuit(SCROMP)is proposed.The damping oscillatory attenuation characteristics of Laplace atoms are exploited to match impulsive responses,while the enhanced harmonic product spectrum(EHPS)is em-ployed to estimate the signal period and construct a periodic atom library.Periodicity-weighted kurtosis(PWK)is introduced as an evaluation metric,and a periodicity weighted spectral kurtosis(PWSK),togeth-er with an improved particle swarm optimization algorithm guided by PWK(PWK-PSO),is proposed to adaptively select the input parameters of the dictionary.Finally,SCROMP is adopted to improve the effi-ciency of sparse coding.Both simulation and experimental investigations verify the robustness and adapta-bility of the proposed method.