DamageYOLO: An In Situ Damage Detection Framework for Rubber Bearings Using the Active‐Sensing Method and an Attention‐Enhanced YOLO Network

During long‐term service, the mechanical properties of laminated rubber bearings are altered by earthquakes, sustained loads, temperature variations, and other factors, thereby reducing their seismic isolation performance. To address the limitation that existing methods cannot achieve in situ, high‐precision detection of bearing mechanical properties, this study proposes an innovative detection framework, DamageYOLO. This framework integrates active sensing, the Continuous Wavelet Transform (CWT), and a pretrained YOLOv5s model enhanced with attention mechanisms. The pretrained YOLOv5s model is fine‐tuned to adapt to the bearing damage detection task, and the Squeeze‐and‐Excitation (SE) and Multihead Attention (MHA) modules are introduced to enhance the model’s feature representation capability. To validate the effectiveness of the DamageYOLO framework, a database containing 2880 samples was established through accelerated aging and active‐sensing experiments. The corresponding detection signals were converted into two‐dimensional wavelet scalograms using CWT and used as input features for the model. The results show that, on the test set, the developed model, DamageYOLO–shear, achieved a coefficient of determination ( R 2 ) of 0.9996 and a mean absolute error (MAE) of 0.8 N/mm. Furthermore, the proposed model exhibits superior predictive performance compared with deep learning models developed using time‐domain images and wavelet packet energy spectra. This suggests that the DamageYOLO framework provides a new and effective approach for damage detection in laminated rubber bearings.