Content area
Abstract
This dissertation advances physics-based earthquake simulations through three interconnected research thrusts: the development and application of a site-specific seismic response analysis workflow in Istanbul, significant computational innovations embodied in the Hercules 2.0 earthquake simulator, and regional-scale ground motion modeling and validation in Southern California. The overarching goal is to enhance the accuracy, reliability, and accessibility of simulation tools for improved seismic hazard assessment and engineering practice. The first part addresses site-specific seismic response by developing a modular workflow for soil-structure interaction (SSI) analyses that integrates regional ground motion simulations from Hercules with local structural models in Abaqus, using the domain reduction method (DRM) and perfectly matched layers (PML). A case study of a four-story office building in Istanbul at six different locations subjected to a hypothetical Mw 6.81 earthquake revealed that the accuracy of simplified input methods, specifically foundation input motion (FIM) and free-field motion (FFM), is highly site-dependent. The results demonstrated that full SSI simulation outcomes could not always be accurately replicated by these simplified approaches, likely due to complex wave scattering effects in heterogeneous media, thus highlighting the critical importance of conducting complete SSI analyses. A suite of semi-automated Python scripts was developed to streamline the complex DRM/PML implementation. The second thrust details substantial computational innovations in Hercules 2.0, an octree-based finite element earthquake simulator. Key improvements include comprehensive documentation, support for optional parameters to simplify input, integration of the PROJ library for accurate coordinate transformations, and support for 3D velocity models. A new Hercules Toolbox, comprising Python-based post-processing scripts, facilitates efficient data management, animation of wave propagation, and visualization of station data and velocity profiles. The enhanced capabilities and scalability of Hercules 2.0 were demonstrated through a simulation of the 2020 Mw 7.0 Samos earthquake. The third component of the research focuses on high-frequency (up to 5 Hz) deterministic earthquake simulations using SW4 and the SCEC community velocity model CVM-S4.26.M01 for the 2008 Mw 5.4 Chino Hills, 2024 Mw 4.4 Highland Park, and 2021 Mw 4.3 Carson earthquakes in Southern California. Simulations demonstrated good agreement with recorded ground motions up to 4 Hz. For the Highland Park earthquake, discrepancies were observed above 2 Hz at near-field stations, attributed to complex geological features not fully captured by the velocity model, whereas the Carson earthquake, with its epicenter within the Los Angeles Basin, showed better agreement at short source distances. Parametric sensitivity analyses highlighted the influence of hypocenter depth, corner frequency selection (3.5 Hz for Highland Park, 4 Hz for Carson proving suitable), and the choice of a computational minimum shear-wave velocity (VS,min = 425 m/s) for balancing accuracy and efficiency. These findings highlight the capabilities of high-frequency physics-based earthquake simulations in complex geological settings and underline the importance of careful model parameter selection. Collectively, this research provides insights into the capabilities and limitations of current regional simulation practices, offers a more accessible workflow for detailed site-specific SSI analyses, and delivers an enhanced open-source simulation tool. These advancements contribute to more reliable seismic hazard assessments, inform safer engineering design, and promote collaborative, physics-informed approaches to earthquake risk reduction.
