Bayesian Nonparametric Modeling of Structural Health Indicators under Severe Typhoons and Its Application to Modeling Modal Frequency

Abstract

Structural health indicators, such as modal frequencies, have been commonly utilized to interpret the health condition of monitored structures. This study modeled the relationship between structural health indicators and ambient conditions under severe typhoons. For this purpose, a two-stage Bayesian probabilistic procedure was established. In the first stage, the Bayesian spectral density approach (BSDA) is applied to identify the structural health indicators, namely the modal frequencies in this study, using the measured structural response. In the second stage, the Bayesian nonparametric general regression (BNGR) is introduced to model the relationship between the identified structural health indicators and some selected typhoon-induced ambient conditions. By using Bayesian model selection in conjunction with general regression, BNGR is able to select the most appropriate set of influencing/input variables for the prediction of the structural health indicators without prescribing any functional form. Full-scale measurements of a 22-story reinforced concrete (RC) building were used to demonstrate the efficacy of the procedure. The measurements consisted of over 280 h of structural response and the corresponding ambient conditions captured under the five most severe tropical cyclones that affected the region from 2011 to 2013. This study provides a promising framework for reliable interpretation of the variation of structural health indicators. Although the modal frequencies were considered in this study, the proposed two-stage procedure is applicable for other structural health indicators.