Mining rolling bearing fault diagnosis method based on using Superlet Transform and ODConvNeXt-ELA

To solve the problem of limited feature extraction capabilities and poor generalization in conventional convolutional neural networks (CNNs) under mining bearings, we utilized a novel approach for rolling bearing fault diagnosis. This approach combines Super Wavelet Transform (SLT) with an advanced neural network architecture, ODConvNeXt-ELA. Recognizing the inadequacies in feature adaptation and the difficulty of focusing on critical fault features in standard CNNs, we developed ODConvNeXt-ELA by integrating Full-Dimensional Dynamic Convolution (ODConv), Batch Normalization (BN), and Efficient Local Attention (ELA) within the ConvNeXt framework. To fully utilize deep learning's potential in picture processing, we transformed one-dimensional vibration signals into two-dimensional time-frequency images using SLT. These images were then utilized to train the ODConvNeXt-ELA model.The proposed method demonstrates high accuracy and robust generalization capabilities in diagnosing faults across different bearings and operating conditions. These findings indicate that the method is effective for cross-bearing and cross-condition fault diagnosis, offering a reliable solution for complex diagnostic tasks.A series of experiments were conducted to validate the effectiveness of the proposed method. Initially, a dataset from Case Western Reserve University, representing a single working condition, was used to separately evaluate the superiority of the SLT and the effectiveness of the ODConvNeXt-ELA model. Subsequently, the method’s capability for cross-condition fault diagnosis was assessed using a variable-condition dataset from the same university. Finally, to further evaluate the method’s ability to diagnose faults across different bearings and conditions, a dataset containing diverse bearings from the University of Paderborn was utilized for comprehensive testing. The results show that the proposed method has the advantages of high accuracy and strong generalization ability, and can accurately identify the fault type.