Multidamage Detection of Breathing Cracks in Plate-Like Bridges: Experimental and Numerical Study

Abstract

Bridges may develop breathing cracks under excessive overloading vehicles, while conventional beam models are ineffective in analyzing the effect of spatial distribution of these cracks. This study proposes a data-driven detection model with the consideration of spatial distribution of breathing cracks that can detect the multiple damage locations and degrees of breathing cracks in plate-like bridges. Firstly, a 2D vehicle–bridge interaction model containing breathing cracks is established, and the damage indicator, contact point displacement variation (CPDV), is calculated using vehicle acceleration data. Next, a dataset with CPDV as the input feature is generated using the finite element method to train the CatBoost-based damage prediction model, which considers the random distribution of single and multiple cracks, as well as the influence of different vehicle speeds. Finally, by calculating the CPDV related to the actual bridge and feeding it into the trained model, the location and degree of the damage can be predicted. The numerical simulation results demonstrate that this approach can accurately detect complex crack information under various vehicle speeds and exhibits robustness against road roughness. A laboratory experiment further confirms the effectiveness, applicability, and feasibility of this method to multiple damage locations and degree of breathing cracks.