Development of a practical algorithm for Bayesian model updating of a coupled slab system utilizing field test data

The Bayesian statistical system identification framework is capable to fully exploit available information from the measurement. However, most Bayesian based model updating works were verified or demonstrated using very simple structural systems (e.g., single-DOF or planar truss with less than 10 DOFs) due to its computational demanding characteristic. This paper reports the rigorous derivations of the Bayesian probabilistic structural model updating methodology and its application to full scale civil engineering structure. The three dimensional (3D) finite element (FE) model of the coupled floor slab system of the Tin Shui Wai Indoor Recreation Center in Hong Kong first was established. This paper reports the step-by-step modification of the class of FE models and the selection of a representative model class based on the set of ambient vibration data obtained from multi-setup field tests. A newly developed pragmatic data-exchange algorithm is employed to integrate the Bayesian structural model updating method in MATLAB with the finite element analysis results in ANSYS. Discussions on the relative accuracy of identified natural frequency and mode shape and their relative significance in the model updating process have been made via processing the vibration data through the Fast Bayesian FFT modal identification method. This work provides valuable information and experience for engineers in modeling and updating similar types of civil engineering structures in the future. © 2014 Elsevier Ltd.