Multiparameter Identification of Bridge Cables Using XGBoost Algorithm

Accurately identifying tension force on cables is of great significance for construction control and the operational status assessment of a bridge during its lifetime. Unlike the conventional vibration methods that encounter problems in the inaccurate identification of short cables and difficulties when identifying multiparameters simultaneously, when solving the vibration differential equation inversely, a novel strategy was proposed that was based on an intelligent algorithm for cable parameter monitoring onsite. The Extreme Gradient Boosting (XGBoost) model was employed to establish the mapping relationship between the natural frequencies of the cable and its tension, bending stiffness, and boundary conditions through data mining. The results revealed that when the measured natural frequencies of a cable were fed into the XGBoost model, the previously mentioned multiparameters could be identified simultaneously with a relative error of <5%. Meanwhile, the proposed intelligent method with the XGBoost algorithm produced a more accurate identification of the cable parameters than the extreme learning machine (ELM) and conventional vibration methods. The current intelligent strategy might provide efficient tools for the simultaneous identification of multiple parameters in cables and, therefore, might facilitate policy decisions for the structural maintenance of cable-supported bridges.