Study on Data-Driven Identification Method of Hinge Joint Damage under Moving Vehicle Excitation

The hinge joint is an important and fragile component of assembled hollow-slab bridges. Therefore, it is necessary to regularly identify hinge joint damage for guaranteeing the safety of assembled hollow-slab bridges. However, conventional hinge joint damage identification methods are time consuming and expensive. Therefore, this study proposes a data-driven hinge joint damage identification method under moving vehicle excitation to quantitatively identify hinge joint damage conveniently. First, we established a refined finite-element model of a hollow-slab bridge with damaged hinge joints and analyze the dynamic response of the bridge under vehicle loads. The Pearson correlation coefficient between the acceleration time history of the adjacent slabs was proposed as the damage index. Further, an ensemble learning algorithm called gradient boosted regression trees (GBRT) was employed to develop a model for identifying hinged joint damage. Finally, the performance of the model was thoroughly compared with commonly utilized machine-learning algorithms and the auto-encoder-based method. The results show that the proposed model exhibits the highest accuracy. Under different signal-to-noise ratio conditions, the model's coefficient of determination (R2) is always above 0.85, the mean absolute error (MAE) is below 4.40 cm, and the root mean squared error (RMSE) is below 7.91 cm. This confirms the feasibility of the model for quantitative and convenient identification of the damage height of hinged joints.