Probabilistic seismic analysis for liquefiable embankment through multi-fidelity codes approach
| Autor: | Fernando Lopez-Caballero |
|---|---|
| Přispěvatelé: | Laboratoire de mécanique des sols, structures et matériaux (MSSMat), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), ANR-17-CE22-0009,ISOLATE,Caractériser et prévenir la liquéfaction des sols de fondation des ouvrages(2017) |
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Computer science
media_common.quotation_subject 0211 other engineering and technologies Soil Science Fidelity 020101 civil engineering 02 engineering and technology 0201 civil engineering Seismic analysis Surrogate model [SPI.GCIV.RISQ]Engineering Sciences [physics]/Civil Engineering/Risques ComputingMilieux_MISCELLANEOUS 021101 geological & geomatics engineering Civil and Structural Engineering media_common business.industry Settlement (structural) [SPI.GCIV.GEOTECH]Engineering Sciences [physics]/Civil Engineering/Géotechnique Probabilistic logic Liquefaction Structural engineering Geotechnical Engineering and Engineering Geology Finite element method Seismic hazard 13. Climate action business |
| Zdroj: | Soil Dynamics and Earthquake Engineering Soil Dynamics and Earthquake Engineering, Elsevier, 2021, 149 (1), pp.106849. ⟨10.1016/j.soildyn.2021.106849⟩ |
| ISSN: | 0267-7261 |
| Popis: | The study focuses on both the effects of soil foundation liquefaction on the induced damage of an embankment and on the generation of a probabilistic seismic demand model that relates few specific characteristics of seismic hazard at a site with the liquefaction-induced settlement. To this aim, several non-linear dynamic Finite Element (FE) analyses were performed. However, given the costly evaluation of the numerical model simulations for a large number of earthquake records, a surrogate model based on multi-fidelity approach is used to represent the output of the expensive FE model. In this approach, the training database is composed by computational low-fidelity data together with limited high-fidelity one. Thus, the proposed methodology is used to generate the curves describing the annual rate of exceeding different values of damage levels. A comparison with the FE reference results suggests that the accuracy of the prediction of the proposed surrogate model is comparable to those from direct numerical FE analysis. Findings also illustrate clearly the importance and the advantages of an adequate definition of the input parameters to build the surrogate model. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect