Alternative Methods for Bayesian Updating in Modal Analysis

Abstract

Structural health monitoring and damage detection methods often rely on numerical models to interpret recorded data. These models, however, are frequently subject to inaccuracies and require validation using empirical data that is inherently uncertain. To address this challenge, researchers and practitioners commonly employ Bayesian model updating techniques, utilizing Markov Chain Monte Carlo methods to sample from the posterior distribution. These approaches are valued for their robustness and flexibility.

Recent advancements in model reduction and surrogate modeling have further enhanced the efficiency and accuracy of Bayesian updating methods. In this contribution, we present two such approaches: a Kalman filter-based method and a transport map-based method, both incorporating the generalized Polynomial Chaos Expansion. The performance of these methods is demonstrated through the updating of model parameters for a wooden frame structure based on measured natural frequencies.