Adaptive Sampling-Based Bayesian Model Updating for Bridges Considering Substructure Approach

Abstract

During long-term bridge monitoring, model updating is necessary because it provides the basis for accurate condition assessment and damage detection. In this study, an adaptive sampling-based Bayesian model updating method for bridges is developed considering a substructure approach. First, the substructuring method is considered to solve eigenvalue problems. By reducing the size of the characteristic equations, the substructure approach overcomes poor algorithm performance, nonconvergence of results, and inefficient model updating caused by the large number of updated parameters when updating a large-scale system. Then Bayesian model updating is applied to quantify the uncertainty existing in bridge model updating and to obtain the posterior probability density function (PDF) of updating parameters that can be further used in different fields of engineering. By introducing the affine-invariant ensemble sampler (AIES) to replace the traditional Metropolis-Hastings (MH) sampler, an adaptive transitional Markov chain Monte Carlo algorithm is proposed to obtain the posterior probability of parameters with high efficiency. Application to a bridge structure demonstrates that the proposed method is efficient and useful in engineering problems.