Adaptive Information Filtering Method Based on Sensitivity Analysis for Bayesian Updating

Abstract

Bayesian updating is a powerful tool for updating engineering models with observed information. As a result of structural health monitoring sensors or platforms, up-to-date information reflecting characteristics of structures and infrastructure systems is available. However, there is usually a large amount of data collected from monitoring technologies in practical engineering, which means the associated computational cost for Bayesian updating will be considerably challenging. The lack of knowledge of observed information makes it impossible to select valuable information for updating. To overcome these limitations, this paper proposes an adaptive information filtering (AIF) method based on sensitivity analysis for Bayesian updating. Specifically, observed information is classified by means of sensitivity analysis and the information valuable to the updating target is filtered out. Moreover, the dispersion of the posterior distribution is adopted as the metric for quantifying updating effectiveness. One linear algebraic example and one case study of chloride-induced concrete corrosion considering carbonation are investigated to demonstrate the computational performance of the proposed method.