Comparison of Various Numerical Assumptions for 1-D Nonlinear Site Response Analysis on Real Sites.

Abstract eng:
In the framework of the PRENOLIN international benchmark, aiming at testing multiple numerical simulation codes capable of predicting non-linear seismic site response with various constitutive models, the validation phase has consisted in comparing the predictions of numerical estimations with actual strong motion recordings obtained at well-known Japanese sites, selected within the Japanese KiK-net and PARI (Port and Airport Research Institute) accelerometric networks, and being as close as possible to a 1D geometry (horizontal layers), with complete field and laboratory measurements. In this paper, we present the results obtained with the CyberQuake computation code [1] for the deep Kushiro (KiK-net KSRH10) and shallow Sendai (PARI) sites. For each site, different input (field and lab) data were provided by the organizers: weak and strong input motions; degradation curves (shear modulus and damping ratio evolution with respect to the cyclic shear strain distortion), either selected in Darendeli’s curves or derived from available cyclic tri-axial laboratory tests or considering an hyperbolic constitutive model. Here, a comparison is shown between actual recordings available on these sites (including those from Tohoku earthquake 2011) with various 1D computing assumptions and methods, assuming vertical incidence for propagating waves: (1) the socalled "equivalent linear model", which consists in a linear iterative viscoelastic approach performed in the frequency domain; (2) the nonlinear transient dynamic approach, based on finite-element time-domain computations, considering an elastoplastic cyclic constitutive model [2] and either effective-stress approach (called “preferred soil model” as chosen freely by each PRENOLIN participant from the available geotechnical data) and total-stress approach (called “imposed soil model”, as set by PRENOLIN organizers). For the shallow site, the nonlinear effective-stress approach gives better results for all the selected input motions (waveform, amplitude and frequency content), whereas with the total-stress approaches (equivalent linear and nonlinear), the soil response is overestimated and strains are underestimated for strong motions. For the deep site, the soil response is globally overestimated and frequency content is not well captured, even for weak input motions. This case has raised questions about the 1D assumption and the vertical incidence of input motions. Further investigations and computations are still needed, but they could not be performed during the PRENOLIN benchmark duration.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket