Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Experimental Evaluation of Bayesian Finite Element Model Updating Using Combined Normal and Lognormal Distributions
In the present work, a finite element (FE) model updating approach in Bayesian framework is presented based on maximizing the posterior probability. Model updating is performed targeting modal measurements like measured natural frequencies and measured mode shapes. A typical FE updating in Bayesian framework utilizes Gaussian/normal distribution for describing the probability density function of uncertain parameters, in spite of statistical issues associated with Gaussian distribution for strictly positive parameters. In order to deal with these issues, lognormal distribution is employed for such parameters, while normal distribution is used for the remaining parameters. Associated formulations including the uncertainty estimation and probabilistic damage detection are concisely presented. The proposed approach is experimentally evaluated using a four story building structure primarily consisting of steel members with multiple damage cases and a steel cantilever beam. FE models of both these structures are updated from modal testing measurements obtained using impact hammer, accelerometers and data acquisition system. Performances in structural identification are evaluated for both the laboratory structures and subsequent probabilistic estimation of damages for the building structure. Further comparison is performed between the proposed and the typical updating approaches, where almost similar level of performances both in terms of accuracy and computation-time are observed in both the cases.
Experimental Evaluation of Bayesian Finite Element Model Updating Using Combined Normal and Lognormal Distributions
In the present work, a finite element (FE) model updating approach in Bayesian framework is presented based on maximizing the posterior probability. Model updating is performed targeting modal measurements like measured natural frequencies and measured mode shapes. A typical FE updating in Bayesian framework utilizes Gaussian/normal distribution for describing the probability density function of uncertain parameters, in spite of statistical issues associated with Gaussian distribution for strictly positive parameters. In order to deal with these issues, lognormal distribution is employed for such parameters, while normal distribution is used for the remaining parameters. Associated formulations including the uncertainty estimation and probabilistic damage detection are concisely presented. The proposed approach is experimentally evaluated using a four story building structure primarily consisting of steel members with multiple damage cases and a steel cantilever beam. FE models of both these structures are updated from modal testing measurements obtained using impact hammer, accelerometers and data acquisition system. Performances in structural identification are evaluated for both the laboratory structures and subsequent probabilistic estimation of damages for the building structure. Further comparison is performed between the proposed and the typical updating approaches, where almost similar level of performances both in terms of accuracy and computation-time are observed in both the cases.
Experimental Evaluation of Bayesian Finite Element Model Updating Using Combined Normal and Lognormal Distributions
Structural Integrity
Fonseca de Oliveira Correia, José António (Herausgeber:in) / Choudhury, Satyabrata (Herausgeber:in) / Dutta, Subhrajit (Herausgeber:in) / Das, Ayan (Autor:in) / Debnath, Nirmalendu (Autor:in)
International Conference on Advances in Structural Mechanics and Applications ; 2021 ; Silchar, India
Advances in Structural Mechanics and Applications ; Kapitel: 30 ; 447-463
Structural Integrity ; 19
08.06.2022
17 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Invertible Alternatives to Normal and Lognormal Distributions
| Online Contents | 2007
Invertible Alternatives to Normal and Lognormal Distributions
| British Library Online Contents | 2007
| Wiley | 2018
Shifted Mixture Model Using Weibull, Lognormal, and Gamma Distributions
| Springer Verlag | 2023