NOVEL APPLICATION OF MODEL UPDATING FOR DAMAGE DETECTION OF UHPC XUAN DUC BRIDGE

This article introduces an innovative approach to assess the structural health of bridges based on dynamic and static load test data from the Xuan Duc bridge, a bridge constructed using Ultra-High-Performance Concrete (UHPC) in Tuyen Quang Province, Vietnam.The measured deflection values of all girders and the natural frequency of the superstructure, along with the PSO (Particle Swarm Optimization) algorithm, were employed to update a finite element model developed in SAP2000 (a commercial structural analysis and design software).This updating process resulted in a significant reduction in error from 5.81% to 0.22% for deflection values and from 2.48% to 0.02% for natural frequencies when compared with the measured data.It is shown that the updated numerical model accurately reflects the operational condition of the bridge during load testing, facilitating the determination of the elastic modulus values of UHPC material.Additionally, this paper explores the feasibility of the approach in identifying the location and degree of damage in superstructures by conducting two numerical case studies with high accuracy.Furthermore, the effect of noise in load testing on the updating process was also considered.With a maximum noise level of 3%, the method maintains accuracy in locating damaged zones, yielding damage level values of 24.70% and 36.47%compared to respective 20% and 30% without noise.Results from this paper confirm the effectiveness of applying machine learning in advanced structural health monitoring.