Ground motion simulations: validation and application for civil engineering problems

Engineers use earthquake ground motions for a variety of reasons, including seismic hazard assessment, calibration of ground motion prediction equations (GMPEs), and input to nonlinear response history analysis. These analyses require a significant number of ground motions and for some scenarios, such as earthquakes with large magnitudes and short distances, it may be difficult to obtain a sufficient number of
ground motion recordings. When sufficient recordings do not exist, engineers modify
available recordings using scaling or spectrum matching, or they use ground motion
simulations.
Ground motion simulations have existed for decades, but recent advances in simulation
methods due to improved source characterization and wave propagation, coupled
with increased computing power, have increased potential benefits for engineers.
But before simulations can be used in engineering applications, simulations must be
accessible and consistent with natural observations. This dissertation contributes to
the latter issue, and it investigates the application of simulations to specific engineering
problems.
The Southern California Earthquake Center (SCEC) Broadband Platform (BBP)
is an open-source software distribution that enables third-party users to simulate
ground motions using research code contributed by model developers. Because the
BBP allows users to compute their own simulations with little knowledge of the underlying
implementation and it ensures that all calculations are reproducible, it is
extremely valuable for simulation validation and engineering applications. In this dissertation, the BBP is evaluated as a simulation generation tool from an engineering
perspective. Ground motions are simulated to study parameters of engineering
interest, such as high-frequency variability, near-fault ground motions, and local
site response. Though some parameters need further development, such as site response
(which is currently implemented using simple empirical amplification), the
BBP proves to be an e↵ective tool for facilitating these types of engineering studies.
This dissertation proposes a simulation validation framework based on simple and
robust proxies for the response of more complicated structures. We compile a list of
proxies with robust empirical models that are insensitive to changes in earthquake scenario
and do not rely on extrapolation for rarely observed events. Because predictions
of these proxies are reliable under a variety of earthquake events, we can confidently
compare them with simulations. The proposed proxies include correlation of " across
periods, ratio of maximum to median response across horizontal orientations, and
ratio of inelastic to elastic displacement. The validation framework is applied to example
simulations and successfully exposes some parameters that need work, such as
variability and correlation of spectral acceleration.
Finally, this dissertation investigates the application of simulations to response
history analysis and fling-step characterization. A 3D nonlinear structural model
is analyzed using recordings and simulations with similar elastic response spectra.
The structural performance and resulting design decisions are similar, indicating that
simulations are e↵ective for response history analysis subject to certain conditions. To
investigate fling-step, we extract fling pulses from a large set of simulations. Extracted
fling properties such as amplitude and period are then compared to specially-processed
recordings and relevant empirical models for surface displacement and pulse period.
Reasonably good agreement is found between simulations, recordings, and empirical
models.
In general, ground motion simulations are found to be an e↵ective alternative
or supplement to recordings in several engineering applications. Because simulation
methods are still developing, this work is not intended as an evaluation of existing
methods, but rather as a development of procedures that can be used in ongoing
work.