ndustrial (FEM-) models used for seismic vulnerability analysis contain generally an important number of DOFs and are computationally intensive since complex damage models and failure modes have to be considered. This is why the choice of an accurate metamodel is crucial for uncertainty propagation and sensitivity analysis. We will present results obtained using High dimensional model representation (HDMR). The terms of the latter decomposition are evaluated by means of state-dependent regression, as recently proposed by Ratto et al., 2007. As we will show, the HDMR decomposition provides a convenient framework to account for stochastic uncertainties introduced by seismic load. We compute fragility curves and perform sensitivity analysis. It is well known that the seismic load, an intrinsically random phenomena, is the most important contributor to output variability. However, engineers are interested in the influence of the model parameters on the model output. The latter are generally considered as rather epistemic uncertainties. This is why we will consider two study cases. The first study concerns the case where only seismic hazard (random uncertainties) has to be accounted for. We construct the HDMR metamodel and compute fragility curves. In a second time, we consider the case where seismic hazard as well as parameter uncertainties have to be considered. For this second case study, we make use of the insights gained by the first study case. This allows us to compute the HDMR model and to perform sensitivity and fragility analysis at very low cost.
Use of HDMR metamodel for seismic fragility analysis
BORGONOVO, EMANUELE;
2011
Abstract
ndustrial (FEM-) models used for seismic vulnerability analysis contain generally an important number of DOFs and are computationally intensive since complex damage models and failure modes have to be considered. This is why the choice of an accurate metamodel is crucial for uncertainty propagation and sensitivity analysis. We will present results obtained using High dimensional model representation (HDMR). The terms of the latter decomposition are evaluated by means of state-dependent regression, as recently proposed by Ratto et al., 2007. As we will show, the HDMR decomposition provides a convenient framework to account for stochastic uncertainties introduced by seismic load. We compute fragility curves and perform sensitivity analysis. It is well known that the seismic load, an intrinsically random phenomena, is the most important contributor to output variability. However, engineers are interested in the influence of the model parameters on the model output. The latter are generally considered as rather epistemic uncertainties. This is why we will consider two study cases. The first study concerns the case where only seismic hazard (random uncertainties) has to be accounted for. We construct the HDMR metamodel and compute fragility curves. In a second time, we consider the case where seismic hazard as well as parameter uncertainties have to be considered. For this second case study, we make use of the insights gained by the first study case. This allows us to compute the HDMR model and to perform sensitivity and fragility analysis at very low cost.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
