Wavelet Entropy Applied in gearbox fault diagnosis

The vibration signal of the gearbox usually has non-stationary and non-linear characteristics, including quite weak fault signals. Therefore, the feature extraction of faulty gearboxes is usually very difficult, which has attracted the attention of many scholars. In this article, a gearbox fault diagnosis method based on wavelet entropy and information fusion is proposed. This paper extracts four different wavelet entropy fault features and calculates wavelet root mean square (RMS) entropy, wavelet crest factor (PF) entropy, wavelet singularity (WS) entropy, and wavelet time-frequency (WTF) entropy. In this wavelet entropy, wavelet RMS entropy and wavelet PF entropy can be classified as wavelet time-domain feature entropy. In this way, a fault feature vector containing four elements can be obtained. As an innovation of this article, principal component analysis (PCA) is used to fuse these wavelet entropies. Through orthogonal transformation, the fault feature vector can be got under the new coordinates. This paper the first three components is chose as the final fault feature vector. Another innovation of this article is the feature parameter part. In this article, the matrix norm is calculated as the feature to be extracted, which is an innovation in fault diagnosis. Finally, this paper conducted two case studies on this method. In case study 1, this paper uses single failure mode gearbox failure data to validate the method. Experimental results show that PCA-based information fusion has a great contribution to feature extraction. In practice, gearbox failures usually include mixed failure modes. Therefore, in order to verity the fault diagnosis ability of this method in practice, this paper uses the mixed failure mode gearbox fault data to test the method in case study 2. The test results show that the method based on wavelet entropy and information fusion has good diagnostic performance. Practice of gearbox fault diagnosis.