SCEC Award Number 21126 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Optimizing near-fault refinement of brittle damage model in rupture dynamics
Name Organization
Christine Goulet University of Southern California Yongfei Wang University of Southern California
Other Participants
SCEC Priorities 2c, 2d, 4a SCEC Groups FARM, CS, EEII
Report Due Date 03/15/2022 Date Report Submitted 05/11/2022
Project Abstract
Coseismic fault displacements can pose a failure risk to distributed infrastructures (SCEC5 Theme) sited near active faults. Relative to well-developed empirical ground-motion models (GMMs), fault displacement models are sparse and poorly constrained by empirical data. Physics-based models such as dynamic rupture simulations present a viable alternative to quantify fault displacements, provided they are appropriately validated against available information. However, there is a considerable resolution gap between observations that matter to infrastructure (> tens of centimeters) and what the models can currently resolve (~tens of meters). Dynamic rupture models at full resolution are not currently computationally tractable, even through the use of high-performance computing (HPC) resources. In order to address this issue, we optimize a highly-scalable structured-mesh rupture dynamics application (SORD) by implementing a block-wise energy-conserving adaptive refinement scheme. In this study, we develop a localized refinement domain that encloses the dynamic fault for accommodating a high-accuracy source solution. The adaptive refinement layer is zipped into a larger exterior coarse domain that carries on input and output tasks (IO). Benchmarks with uniformly fine and coarse grids show that the adaptive refinement approach achieves results consistent with those for the fine grid for on- and off-fault displacements in shorter computational time, while maintaining the similar low memory usage of the coarse grid setup. This new simulation approach enables us to study fault physics within a very fine aperture and allows a fine-enough prediction of small-scale localized fault-displacement features to be validated for hazard analysis applications.
Intellectual Merit Our previously developed brittle damage model helps to capture and deepen our understanding of inelastic rock behavior during earthquakes, supporting advances under the “Beyond elasticity” SCEC research priority, but it limited for application in the near-fault region. Our newly developed near-fault adaptive refinement in the dynamic rupture simulation tool (SORD) enables us to simulate the broader zone of observable off-fault displacements while including the brittle damage model, making it ripe for validation. This model implementation will allow us to perform more realistic simulations that quantitatively capture the near-fault ground motions and small-scale fault displacements needed to support seismic hazard research.
Broader Impacts The brittle damage model and the near-fault adaptive refinement framework provide computational capabilities that are relevant to several SCEC groups. The dynamic evolution of fault zones is one of the key elements for understanding long-term earthquake cycles in the Sequences of Earthquakes and Aseismic Slip (SEAS) Technical Activity Group (TAG). Technically, the mesh interface scheme between fine and coarse regimes developed in this adaptive fault refinement approach can be potentially used in developing new regular or irregular shaped refinement applications (dynamic rupture or wave propagation) of other structured finite difference/element methods, such as AWP and Cybershake. More broadly, the incorporation of nonlinear material models into seismic hazard analyses supports the Earthquake Engineering Implementation Interface (EEII) activities, including those from the Ground Motion Simulation Validation (GMSV) TAG.
Exemplary Figure Figure 2. Mesh refinement scheme for dynamic ruptures simulations with SORD. (a) Stencil of the block-wise discontinuous mesh near the fault, (b) Discrete operator converting a nodal function to a cell function, (c) Adjoint discrete operator converting a cell function to a nodal function, (d) Stencil for updating Type A nodal functions, (e) Stencil for updating Type B nodal functions.

Credit: Wang and Goulet (2021). SCEC Report 21126.