Fault Diagnosis Method of Rolling Bearing Based on LDA-Shapelet Transform

To improve composite fault feature extraction and diagnosis accuracy for rolling bearings, and to address the over-reliance on signal processing and poor interpretability of existing methods, a new feature extraction method combining shapelet transformation and linear discriminant analysis (LDA) is proposed. This method uses the Pelican algorithm to optimize SVM for fault diagnosis. First, the shapelet algorithm maps vibration signals into shapelet space, creating a high-dimensional feature matrix based on distances. This matrix is then reduced using LDA for visualization. Next, feature samples are classified using the Pelican-optimized SVM model. Tests show the proposed method that improves accuracy by over 10%, achieving more than 99%.