Open Access Open Access  Restricted Access Subscription or Fee Access

Efficient Bayesian Model Selection for Identifying Locally Nonlinear Systems Incorporating Dynamic Measurements

S. DE, E.A. JOHNSON, S.F. WOJTKIEWICZ, P.T. BREWICK

Abstract


In Bayesian model selection, suitable mathematical models are selected among a set of possible models using Bayes’ theorem. To simplify the analysis, linear structural models are often used, though they are not always adequate to accurately compute the structural response. Nonlinear models, which may be more accurate, significantly increase the required computation time for Bayesian model selection. A method is proposed in this paper to reduce the computational cost by incorporating into the model selection process the authors’ previously developed efficient dynamic response algorithm for response of locally nonlinear systems. The efficacy of the approach is demonstrated, using the numerical example of a base-isolated building on hysteretic lead rubber bearing (LRB) isolators, with different linear and nonlinear model classes as candidates to represent the LRBs.

doi: 10.12783/SHM2015/288


Full Text:

PDF