Model Updating Using Bayesian Estimation

Variability in real structures, which could arise from manufacturing processes, and the modelling assumptions and limitations require the creation of a statistical model of the relationship between experimental and model predictions and the quantification of the uncertainty of this estimate. In this paper Markov-Chain Monte Carlo theory (MCMC) is discussed and applied to model updating in the case of multiple sets of experimental results by using frequency responses functions. The MCMC method allows the solution of complex problems in a unifying framework, by integrating over high dimensional probability distributions in order to make inferences about the model parameters. A simulated three degree-of-freedom system is used to illustrate some aspects of the method, allowing for practical assumptions to be tested on a simple example within the WINBUGS environment (Bayesian inference Using Gibbs Sampling).