SCEC Award Number 13095 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Multiscale Dynamic Rupture Simulations using Adaptive Mesh Refinement
Investigator(s)
Name Organization
Jeremy Kozdon Naval Postgraduate School
Other Participants
SCEC Priorities 3, 6, 4 SCEC Groups CS, FARM, DRCV
Report Due Date 03/15/2015 Date Report Submitted N/A
Project Abstract
 In this ongoing project, we are developing a set of computational tools to enable dynamic rupture simulation using laboratory measured friction parameters. To accomplish this we believe that three key components are required: scalability, adaptivity, and high-order numerical methods. To this end, we have been developing a new adaptive mesh refinement tool for dynamic rupture simulations based on high-order accurate, discontinuous Galerkin finite element methods using the p4est mesh refinement library. With this support we have extended the physics capabilities of the code to handle off-fault plasticity, unstructured block connectivity, non-planar fault interfaces, and variable material properties. We have also completed most of the back-end development required to allow for dynamic hp-adaptivity, that is adaptivity in element size and order. Under separate support, the back-end of the code is being extended to support GPU computing which we will be able to leverage for rupture dynamics.
Intellectual Merit Dynamic rupture seems to be an ideal application for adaptive mesh refinement. The adaptivity in both element order and size that is enabled by this work holds out the promise to answer whether this is true. The tools developed will allow the exploration of the questions related to earthquake energy balance and enable simulations with large inner-outer scales.
Broader Impacts During this project the PI mentored a summer student from Hartnell Community College in Salinas. The student was exposed to both scientific computing and earthquake science for the first time. Additionally, the developed software is public domain and available online via the github project page https://github.com/bfam
Exemplary Figure Figure 3:
(top) Example of the block structured AMR mesh used for the branching fault test problem TPV15. The black lines represent the ``coarse'' mesh and the finer lines the refined mesh. The green line indicates the fault. (bottom) comparison between beard (green line) and SeisSol (dashed red line) for two on-fault stations. Both SeisSol and beard are modeling the fault junction directly (no gap between the "main" fault and "branch"), though both codes are capable of handing the gap case.