Domain Reduction Method

During this allocation we implemented the domain reduction method (DRM) in Hercules. The DRM is a modular two-step, finite-element methodology for modeling earthquake ground motion in highly heterogeneous localized regions with large contrasts in wavelengths. The DRM solves the problem of multiple physical scales by subdividing the original problem into two simpler ones. In the first problem, the earthquake source and propagation path effects are simulated in a model that encompasses the source and the background structure from which the localized features (i.e., the buildings) have been removed. This first step entails saving the displacement field over a (one-element) layer composed by two adjacent surfaces that define the interface between the exterior and interior regions. The second step solves the problem in the interior region containing the features. This second model takes as input a set of equivalent seismic or effective forces, which are calculated from the displacements obtained in the first step and the properties of elements in the interface. The DRM has significant implications in terms of computational resources usage and performance. Table 1 shows the simulation domain characteristics and properties of a case study for the Northridge earthquake, and the simulation performance for the DRM steps I and II for two soil conditions. It illustrates the computational gain obtained from using DRM for parametric studies in the Step II sub-region. These simulations were run in Kraken at NICS.

Table 1. Simulation domain characteristics and properties of a case study for the Northridge earthquake, and the simulation performance for the DRM steps I and II for two soil conditions.