Conditioned Simulation of Ground Motion Time Series Using Gaussian Process Regression With Application to Ridgecrest Ground Motions
Session: Numerical Modeling of Earthquake Motion, Rupture Dynamics, Seismic Noise, Wave Propagation and Inverse Problems II
Type: Oral
Date: 4/23/2021
Presentation Time: 03:00 PM Pacific
Description:
Ground motion time series are critical elements of earthquake engineering for performance analysis of seismic regions. The number of available instruments to record free-field ground motions in the U.S. is currently too sparse. Thus, ground motion simulation methods are developed to obtain input motion estimates at locations where there is no available instrumentation. We have constructed ground motion time series using a Gaussian Process Regression method, which models the real and imaginary parts of the Fourier spectrum as random Gaussian variables. The proposed models’ training and validation are carried out using physics-based simulated ground motions of the 1906 M7.9 San Francisco Earthquake. The evaluation of the model’s performance is also carried out using the physics-based simulated M7.0 Hayward fault earthquake and observed M7.1 2019 Ridgecrest earthquake ground motions recorded by the Community Seismic Network in Los Angeles. Most of the evaluations illustrate that the trained Gaussian Process Regression model is able to estimate the ground motion time series properly. The trained Gaussian Process Regression model has a decent performance in the prediction of the long-period content of the ground motions, including pulses due to directivity. Furthermore, the response spectra of the estimated ground motions are compatible with the corresponding observed ground motions' response spectra at the same locations. The results also illustrate that the prediction for locations either at the boundaries or in regions with fewer observations might be less accurate, and the predicted short-period content of the estimated ground motion time series is less reliable than the long-period content.
Presenting Author: Aidin Tamhidi
Student Presenter: Yes
Authors
Aidin Tamhidi Presenting Author Corresponding Author aidintamhidi@ucla.edu University of California, Los Angeles |
Nicolas Kuehn kuehn@ucla.edu University of California, Los Angeles, J. Garrick Institute for the Risk Science |
S. Farid Ghahari ghahari@seas.ucla.edu University of California, Los Angeles |
Arthur Rodgers rodgers7@llnl.gov Lawrence Livermore National Laboratory |
Monica Kohler kohler@caltech.edu California Institute of Technology |
Ertugrul Taciroglu etacir@ucla.edu University of California, Los Angeles |
Yousef Bozorgnia yousef.bozorgnia@ucla.edu University of California, Los Angeles |
|
|
Conditioned Simulation of Ground Motion Time Series Using Gaussian Process Regression With Application to Ridgecrest Ground Motions
Category
General Session