No abstract

Scenario-based earthquake simulations at regional scales hold the promise in advancing the state-of-the-art in seismic risk assessment studies. In this study, a computational workflow is presented that combines (i) a broadband Green's function-based fault-rupture and ground motion simulation-herein carried out using the "UCSB (University of California at Santa Barbara) method", (ii) a three-dimensional physics-based regional-scale wave propagation simulation that is resolved at max = 11.2 Hz, and (iii) a local soil-foundation-structure finite element analysis model. These models are interfaced with each other using the domain reduction method. The innermost local model-implemented in ABAQUS-is additionally enveloped with perfectly matched layer boundaries that absorb outbound waves scattered by the structures contained within it. The intermediate wave propagation simulation is carried out using Hercules, which is an explicit time-stepping finite element code that is developed and licensed by the CMU-QUAKE group. The devised workflow is applied to a 80 × 40 × 40 km 3 region on the European side of Istanbul, which was modeled using detailed soil stratigraphy data and realistic fault rupture properties, which are available from prior microzonation surveys and earthquake scenario studies. The innermost local model comprises a chevron-braced steel frame building supported by a shallow foundation slab, which, in turn, rests atop a three-dimensional soil domain. To demonstrate the utility of the workflow, results obtained using various simplified soil-structure interaction analysis techniques are compared with those from the detailed direct model. While the aforementioned demonstration has a limited scope, the devised workflow can be used in a multitude of ways, for example, to examine the effects of shallow-layer soil nonlinearities and surface topography, to devise site-and structure-specific seismic fragilities, and for calibrating regional loss models, to name a few.

No abstract

This paper presents a numerical scheme based on a fictitious domain framework for the numerical modeling of earthquake-induced ground motion in the presence of realistic surface topography of the Earth's crust. We show that by adopting a non-conforming octree-based meshing approach associated with a virtual representation of the surficial irregularity, we can obtain accurate representations of ground motion. From the computational point of view, our methodology proves to be also efficient, and more importantly, it allows us to preserve the salient features of multi-resolution cubic-shaped finite elements for wave propagation applications. We implemented the non-conforming meshing scheme for the treatment of realistic topographies into Hercules, the octree-based finite-element earthquake simulator developed by the Quake Group at Carnegie Mellon University. We tested the benefits of the strategy by benchmarking its results against reference examples, and by means of numerical error estimate analyses. Our qualitative and quantitative comparisons showed a close agreement between our numerical results and the reference results, and also, that the order of convergence of the displacement field is preserved in the presence of surface topography. Moreover, this performance was obtained by using the same mesh refinement techniques with cubic elements as in traditional flat free-surface simulations. functional behavior of the solution to the trial function space by means of partition of unity concepts [10]. The GFEM enriches the approximated global solution with the so-called handbook functions. These special functions are derived from analytical or highly accurate numerical solutions of simplified boundary-value problems that account for the localized features of the actual problem, for example, cracks, material discontinuities, voids (e.g., [11]). On the other hand, the XFEM employs the partition of unity only locally through special functions of a simpler nature added only to the degrees of freedom near the discontinuity. As a result, this gives the XFEM a broader use. GFEM/XFEM ideas are, however, not the most suitable when mesh approximations only emerge because of misalignment with the domain's boundaries because no field enrichment is required in such scenarios.On the other hand, fictitious domain (FD) ideas [12] essentially preserve the same concepts as in traditional finite element modeling (FEM) but dramatically reduce any mesh related geometrical constraints. The basic idea is to embed an intricate body into a simpler shaped one Q , namely Q (Figure 1(a) and (b)). The augmented body is usually chosen, in 2D, as a rectangle of zero traction on @ Q , which is further discretized via voxel-like meshes. The artificial material in the augmented region Q n is penalized by a small˛factor, seeking to reduce its contribution to the overall solution and to ensure zero traction condition throughout @ Q . Boundary conditions within exterior elements intersecting @ are mostly implemented via Lagrange multipliers, although bo...