ABSTRACT
Bayesian updating is increasingly used in structural engineering; it is applicable as an inverse method to identify the model of uncertainty which best matches some available experimental data. This paper introduces a novel method for the definition of the likelihood function in case the numerical model is a time dependent function. The set of time instants which best describes the experimental data is identified using the maximum entropy principle. The marginal distributions of the model response are identified as well using the maximum of entropy. The dependence between the responses of the model for the different time instants is implemented using the linear coefficients of correlation obtained after a mapping into standard normal distributions. The joint probability density function is subsequently used in the formulation of the likelihood function. The relevance of the method is demonstrated through an application example.
