Combining Continuous Wavelet Transform and Sparse Coding Algorithm in Network for Mechanical Fault Diagnosis

The fault diagnosis of rotating machinery is a critical issue to ensure industrial equipment’s safe and efficient operation. Although various methods have been developed and applied in this field, traditional approaches heavily rely on manual intervention and deep learning methods lack interpretability, limiting their respective applications. This paper proposes a method called Continuous Wavelet Transform-based Nested Iterative Soft Thresholding Algorithm (CWT-NISTA), which combines continuous wavelet transform with an interpretable unfolding network. This method intelligently extracts shallow features of signals using continuous wavelet transform at multiple scales and then uncovers deep features embedded within these shallow features using the unfolding network for fault classification. The effectiveness of this method is validated on a publicly available bearing fault dataset.