EQSIM—A multidisciplinary framework for fault-to-structure earthquake simulations on exascale computers part I: Computational models and workflow

Abstract

Computational simulations have become central to the seismic analysis and design of major infrastructure over the past several decades. Most major structures are now “proof tested” virtually through representative simulations of earthquake-induced response. More recently, with the advancement of high-performance computing (HPC) platforms and the associated massively parallel computational ecosystems, simulation is beginning to play a role in increased understanding and prediction of ground motions for earthquake hazard assessments. However, the computational requirements for regional-scale geophysics-based ground motion simulations are extreme, which has restricted the frequency resolution of direct simulations and limited the ability to perform the large number of simulations required to numerically explore the problem parametric space. In this article, recent developments toward an integrated, multidisciplinary earth science-engineering computational framework for the regional-scale simulation of both ground motions and resulting structural response are described with a particular emphasis on advancing simulations to frequencies relevant to engineered systems. This multidisciplinary computational development is being carried out as part of the US Department of Energy (DOE) Exascale Computing Project with the goal of achieving a computational framework poised to exploit emerging DOE exaflop computer platforms scheduled for the 2022–2023 timeframe.

Validation of (not‐historical) large‐event near‐fault ground‐motion simulations for use in civil engineering applications

“…Overall, for both the FN and FP components the median PID never exceeds the value of 1%, which marks the onset of the nonlinear behavior of the 40‐story building. However, as shown in a separate study conducted by the authors, there is a high site‐specific variability of the demand posed to the structures residing in the vicinity of the fault rupture for all three simulated scenarios, with PID values that go as high as 3.5% for stations located at 1 km from the fault, 33,34 which corresponds to major nonlinearities in the structural systems and short of collapse. Therefore, for studies conducted on single simulated records or a subset of simulated records, the relationship between spectral correlations parameters and observed structural response should focus on the spectral shape at periods longer than 3.76 s.…”

“…In this study, the ground motions generated from three M7 strike‐slip earthquake scenarios developed based on the rupture model of Graves and Pitarka 35 and Pitarka et al 2 are used to demonstrate the proposed validation procedure. The three scenarios differ from each other for the location of the hypocenter only, 33,34 have a geologic structure characterized by an average shallow shear‐wave velocity of Vs 30 = 381 m/s and are resolved at frequencies up to 5 Hz. For each earthquake simulation, ground‐motion time histories were stored at 1 km spacing intervals resulting in a total of 2490 near‐field (within 10 km of the fault) simulated records.…”

“…With this aim, this work proposes a general methodology to evaluate the reliability of simulated ground motions from unknown (not‐historical) earthquake events. The procedure is demonstrated for deterministic simulations up to 5 Hz for three M7 strike‐slip earthquake scenarios that differ from each other for the location of the hypocenter only and are generated from the Pitarka et al 32 kinematic rupture model 33,34 . The validation process includes four steps: Selection of a population of real records collected from multiple events consistent with the simulated scenarios. Comparison of the distribution of ground‐motion IMs from the simulated records, real records, and GMPEs Comparison of the distribution of simple proxies for building response. Comparison of the distribution of EDPs for a realistic model of a structure. …”

“…The procedure is demonstrated for deterministic simulations up to 5 Hz for three M7 strike-slip earthquake scenarios that differ from each other for the location of the hypocenter only and are generated from the Pitarka et al 32 kinematic rupture model. 33,34 The validation process includes four steps:…”

EQSIM—A multidisciplinary framework for fault-to-structure earthquake simulations on exascale computers, part II: Regional simulations of building response

“…A three-dimensional computational earth model was constructed for the regional domain representing the shallow sedimentary basin with a maximum depth of 600 m. As shown in Figure 1, the shear wave velocity in the basin gradually increases with depth from 320 m/s at the free surface to 800 m/s at the base. The seismic velocity model is automatically mapped into a computational grid by internal subroutines utilized by the SW4 program, Nilsson et al (2007), Sjogreen and Petersson (2012), and Petersson and Sjogreen (2012), as described in a companion paper, McCallen et al (in press). The earth model included 8.88 billion grid points to resolve ground motion frequencies up to 5 Hz, and the Cartesian grid employed mesh refinement with depth to maintain numerical accuracy for 5 Hz simulations with grid spacings of 8, 16, and 32 m in the depth ranges 0–600 m, 600–1200 m, and below 1200 m, respectively.…”