Bayesian Model Updating and Its Limitations for Detecting Local Damage of an Existing Truss Bridge

An efficient and robust Bayesian model updating is presented in this paper from the viewpoint of practical application by introducing a new objective function and a realistic parameterization of the mass and stiffness matrices. In this framework, the likelihood function for mode shapes is formulated based on the cosine of the angle between the analytical and measured mode shapes, which does not require any scaling or normalization. For the parameterization of the stiffness matrix, the framework uses element-level parameterization because it is evident from practical experience that different parts of the structure are subjected to different levels of deterioration as a result of corrosion and fatigue. The proposed updating method was validated experimentally by updating a finite-element model (FEM) of an existing steel truss bridge that utilized the vibration data obtained from a car-running test. The damage-detection capability of the proposed model-updating framework was then investigated by considering the data from a simulated damaged bridge and the experimental data from a damaged span of the same bridge with partial fractures on one of the diagonal members.