科研之友
微信
新浪微博
Facebook
分享链接
摘要
Minimum entropy deconvolution (MED) has been widely applied to extract the repetitive transients. Its effectiveness for bearing fault diagnosis, however, might be undermined by the drawbacks: 1) it prefers a solution of a single impulse under the harsh working conditions, and 2) it is sensitive to the length of inverse filters. To overcome such limitations, a novel l(0)-norm embedded MED method is proposed for early bearing fault diagnosis in this paper. The performances of l(0)-norm regularized solution are first detailedly discussed with different smooth functions. Subsequently, the approximate hyperbolic tangent AHTF) with a better steepness is selected as the smooth function to estimate the l(0)-norm. The l(0)-norm regularized solution is then incorporated into the iterative de-convolving procedure of MED as it can suppress interferences of singular impulses and wipe the noise component off the raw signal. As a result, the reliabilities of iterative de-convolving procedure of MED are enhanced via feeding the l(0)-norm regularized solution. Simulated studies and experimental verifications are performed to demonstrate the superiorities of the proposed method by comparing with the conventional MED, SL0-MED and spectral kurtosis (SK) methods. The comparison results illustrate that the proposed method is more effective for weak feature extraction using different lengths of inverse filters. The fault characteristic frequency can be easily identified on the envelope spectra of the signal processed by the proposed method with comparison of the qualitative and quantitative analysis results.
- 出版日期2018-10
- 单位苏州大学
