On Using Posterior Samples for Model Selection for Structural Identification

Abstract

In recent years, Bayesian model updating techniques based on measured response data have been applied in structural identification and health monitoring. These techniques are robust and appropriate because of their ability to characterize modeling uncertainties associated with the structural system. Another important problem is how to select the model class from a set of competing candidate model classes most plausible for the system based on data. To tackle this problem, Bayesian model class selection may be used, which provides a rigorous Bayesian updating procedure to give the probability of the different candidate classes for a system, based on data from the system. The above problems are known to be computationally challenging. A new hybrid approach for solving this challenging problem is proposed. The performance of this approach is illustrated by identification of nonlinear hysteretic models using dynamic data from the structure.

Additional details