ABSTRACT
ABSTRACT Bayesian updating is a versatile method to update models and the parameters with observation data and enable quantitative reliability analysis for civil infrastructures. Although the particle filter (PF) is one of the promising method for Bayesian updating, the PF often confront a serious problem called “filter degeneracy” where all but one of the weights are very close to zero. We present a new algorithm for Bayesian updating called iterative particle filter with Gaussian mixture models (IPFGMM). The idea behind IPFGMM is to apply Gaussian mixture model (GMM) as the proposal density and to introduce an iterative algorithm to avoid filter degeneracy. The proposed method is demonstrated by application to parameter identification in two-degree-of-freedom (2DoF) building model and reliability analysis of a geotechnical structure using elasto-plastic finite element analysis.
