This is the authors' abstract. We don't add key points for this paper.
Identifying accurate models of large structural systems conditional on experimental data is critical for understanding structural behaviors and predicting future structural conditions for life-cycle management of structures. Experimental modal parameters are usually used as data for structural identification due to the lack of excitations, so the eigenvalue problem needs to be repeatedly solved in the identification process. This is very time consuming for full-scale structures with large models. The computational time will even escalate to a level that prohibits practical applications if uncertainties are to be calculated. Considering the above problems, an efficient Bayesian method is developed. It has the following contributions. First, the eigenvalue problem equation is embedded in the posterior probability density function (PDF) of uncertain parameters so that experimental modes can be directly used without spending extra time solving for many calculated modes for mode matching. Second, the parameterization based on sub-structures is used to introduce a linear relationship between structural and modal parameters. A reduced-order sub-structure method is proposed to keep only a few modes for un-measured locations while keeping all the modes for measured locations for each sub-structure. This considerably reduces the dimension of structural system matrices. Third, a new sampling method combined with an enhanced iterative algorithm is developed to explore the parameter space for constructing the posterior PDF and quantifying uncertainties. Iteration is done for each parameter sample to locate a local most probable value (MPV) of uncertain parameters. Leveraging on the proposed linear relationship and the reduced sub-structures, the analytical formula of the optimum of each parameter conditional on others is obtained to facilitate efficient iteration. The posterior PDF is then constructed by a weighted sum of Gaussian PDFs centered at the possible local MPVs using Laplace’s method of asymptotic approximation. The proposed method is verified numerically and experimentally with two shear buildings and a full-scale building, respectively.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Peng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: