Rolling Bearing Fault Diagnosis Method Based on Wavelet Time–Frequency Map and Improved ConvNext

The purpose of this article is to address the issue of traditional fault diagnosis methods for rolling bearings, which are unable to accurately diagnose the fault signals of non-linear and non-stationary rolling bearings. To achieve this goal, this article proposes a rolling bearing fault diagnosis method based on wavelet time–frequency maps and an improved ConvNext model. This method first leverages the advantages of wavelet time–frequency analysis to transform the original signal into a two-dimensional wavelet time–frequency map. Subsequently, a new generation of convolutional neural networks, ConvNext, is utilized as the foundational network for rolling bearing fault diagnosis models. The network is enhanced by introducing lightweight attention modules, such as the Convolutional Block Attention Module (CBAM) and the Efficient Channel Attention Network (ECA), to improve the Block modules in the basic network. The wavelet time–frequency map is then used as a feature map input to the improved ConvNext network for training, aiming to achieve rolling bearing fault diagnosis recognition. Finally, experimental results demonstrate that the proposed method exhibits a high level of fault recognition accuracy and stability in laboratory data. It also demonstrates a certain degree of robustness, offering a new solution for rolling bearing fault diagnosis.