TY - GEN
T1 - Model validation of a four-story base isolated building using measurements from seismic shake-table experiments
AU - De, S.
AU - Yu, T.
AU - Johnson, E. A.
AU - Wojtkiewicz, S. F.
N1 - Publisher Copyright:
© NCEE 2018.All rights reserved.
PY - 2018
Y1 - 2018
N2 - The presence of uncertainties in real-life structures results in many mathematical models proposed to describe a structure's behavior. However, structural control design, lifetime prognosis, hazard analysis and so forth all require a model (or a set of models) that accurately predicts the response. Model falsification can be used to eliminate some of the incorrect models based on the philosophy that measurement data can only be used to reject a model instead of supporting one; however, it cannot judge the relative merits of retained (unfalsified) models. In contrast, Bayesian model class selection can be used to provide such relative judgments; however, it can be computationally intensive for complex models and can, in the end, choose as best a model that is itself invalid. This paper investigates a synergy of these two approaches to mitigate their respective shortcomings. The proposed model validation approach is illustrated using a four-story base-isolated reinforced concrete (RC) building that was tested in 2013 at Japan's E-Defense earthquake engineering research center by subjecting it to different earthquake excitations. The structure is an RC moment frame with structural walls in one corner to resist lateral loading. The isolation layer consists of several kinds of passive devices. The data from the 8 August 2013 tests are used to validate different model classes of the superstructure. The results demonstrate that the Occam's razor built in this validation approach prefers models with fewer parameters.
AB - The presence of uncertainties in real-life structures results in many mathematical models proposed to describe a structure's behavior. However, structural control design, lifetime prognosis, hazard analysis and so forth all require a model (or a set of models) that accurately predicts the response. Model falsification can be used to eliminate some of the incorrect models based on the philosophy that measurement data can only be used to reject a model instead of supporting one; however, it cannot judge the relative merits of retained (unfalsified) models. In contrast, Bayesian model class selection can be used to provide such relative judgments; however, it can be computationally intensive for complex models and can, in the end, choose as best a model that is itself invalid. This paper investigates a synergy of these two approaches to mitigate their respective shortcomings. The proposed model validation approach is illustrated using a four-story base-isolated reinforced concrete (RC) building that was tested in 2013 at Japan's E-Defense earthquake engineering research center by subjecting it to different earthquake excitations. The structure is an RC moment frame with structural walls in one corner to resist lateral loading. The isolation layer consists of several kinds of passive devices. The data from the 8 August 2013 tests are used to validate different model classes of the superstructure. The results demonstrate that the Occam's razor built in this validation approach prefers models with fewer parameters.
UR - http://www.scopus.com/inward/record.url?scp=85085591875&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85085591875&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85085591875
T3 - 11th National Conference on Earthquake Engineering 2018, NCEE 2018: Integrating Science, Engineering, and Policy
SP - 1789
EP - 1793
BT - 11th National Conference on Earthquake Engineering 2018, NCEE 2018
PB - Earthquake Engineering Research Institute
T2 - 11th National Conference on Earthquake Engineering 2018: Integrating Science, Engineering, and Policy, NCEE 2018
Y2 - 25 June 2018 through 29 June 2018
ER -