抄録
This paper presents the Bayesian inference based model calibration strategy for constructing validated numerical models of the existing structures. There exist uncertain changes in the model parameters, such as material properties and boundary conditions, from the nominal condition due to deteriorations or possible damages in the existing structures. The target in this study was the dynamic analysis model of an existing bridge, and the model calibration procedure was applied by using measured resonant frequencies as the comparative feature. It was then shown that the meaningful posterior distributions cannot be obtained without the appropriate prior distribution setting based on the engineering judgments. The numerical modelwas then successfully calibrated by using themeaningful posterior distributions.
