Effects of model uncertainty in nonlinear structural finite element model updating by numerical simulation of building structures

• Published in: Structural control and health monitoring Vol. 26; no. 3; pp. e2297 - n/a
• Main Authors: Astroza, Rodrigo, Alessandri, Andrés
• Format: Journal Article
• Language: English
• Published: Pavia Wiley Subscription Services, Inc 01.03.2019
• Subjects: Aseismic buildings; Building codes; Building design; Computer simulation; Concrete; Concrete construction; Damage detection; damage identification; Data processing; Earthquake damage; Earthquakes; Economic models; Estimating techniques; Excitation; Finite element method; finite element model updating; Iron; Kalman filters; Mathematical models; model uncertainty; Model updating; Noise measurement; nonlinear finite element model; Parameter estimation; Parameter uncertainty; Reinforced concrete; Reinforcing steels; Seismic activity; Seismic engineering; Seismic response; Steel; Steel structures; unscented Kalman filter
• ISSN: 1545-2255; 1545-2263
• DOI: 10.1002/stc.2297

Summary Uncertainties in finite element (FE) model updating arise from two main sources: measurement noise and modeling errors. The latter includes model parameter uncertainty and model uncertainty itself. Among these sources of uncertainty, model uncertainty has been proven to be the most influential source of error in FE model updating, which is particularly important when using the updated model for damage identification (DID) purposes. This paper investigates the effects of model uncertainty when updating mechanics‐based nonlinear FE models of building structures subjected to earthquake excitation. To solve the parameter estimation problem, the unscented Kalman filter is used as parameter estimation tool. Numerically simulated response data of two state‐of‐the‐art nonlinear FE models of building structures designed according to modern design codes are used as application examples. A two‐dimensional steel building and a three‐dimensional reinforced concrete building, both subjected to seismic base excitation, are analyzed for different types and levels of model uncertainty. The results show that model uncertainty may have significant detrimental effects when using the updated FE model for DID, chiefly in the case of large modeling uncertainty. Although a good match between the measured (observed) and FE predicted responses is usually achieved, unobserved responses at global and local levels often show significant errors.