Gramian angular field fusion with dual-branch ConvNeXt and channel-aware transformer for vibration sensor-based fault diagnosis

To address insufficient fault feature representation in vibration sensing signals of rotating machinery, the limited ability of a single image-based representation to characterize complex dynamic information, and the restricted capability of deep models in selecting key features, a fault diagnosis method integrating Gramian Angular Field, dual-branch ConvNeXt, and a channel-aware Transformer is proposed. First, one-dimensional vibration signals are mapped into two types of two-dimensional image representations, namely Gramian Angular Summation Field (GASF) and Gramian Angular Difference Field (GADF), to enhance the structured representation of raw sensing signals from the perspectives of global structural correlation and local dynamic variation. Then, a dual-branch ConvNeXt feature extraction network is constructed to separately model GASF and GADF images, thereby fully exploiting the complementary fault information embedded in the two image representations. On this basis, a channel-aware Transformer module is designed to adaptively enhance fault-sensitive features and suppress redundant information through a channel recalibration mechanism. Meanwhile, the self-attention mechanism is used to capture global dependencies among the fused features. Finally, experiments were conducted on the Paderborn University bearing dataset and the University of Connecticut gear dataset. The results showed that the proposed method achieved accuracies of 97.08% and 98.15% on the two datasets, respectively. It also outperformed the comparison methods in terms of precision, recall, and F1-score, verifying its effectiveness and generalization capability for fault diagnosis of vibration sensing signals.