Monograph records
https://feeds.library.caltech.edu/people/Ross-Z-E/monograph.rss
A Caltech Library Repository Feedhttp://www.rssboard.org/rss-specificationpython-feedgenenFri, 08 Dec 2023 12:39:44 +0000Data-driven Accelerogram Synthesis using Deep Generative Models
https://resolver.caltech.edu/CaltechAUTHORS:20210111-160825629
Authors: Florez, Manuel A.; Caporale, Michaelangelo; Buabthong, Pakpoom; Ross, Zachary E.; Asimaki, Domniki; Meier, Men-Andrin
Year: 2021
DOI: 10.48550/arXiv.2011.09038
Robust estimation of ground motions generated by scenario earthquakes is critical for many engineering applications. We leverage recent advances in Generative Adversarial Networks (GANs) to develop a new framework for synthesizing earthquake acceleration time histories. Our approach extends the Wasserstein GAN formulation to allow for the generation of ground-motions conditioned on a set of continuous physical variables. Our model is trained to approximate the intrinsic probability distribution of a massive set of strong-motion recordings from Japan. We show that the trained generator model can synthesize realistic 3-Component accelerograms conditioned on magnitude, distance, and V_(s30). Our model captures the expected statistical features of the acceleration spectra and waveform envelopes. The output seismograms display clear P and S-wave arrivals with the appropriate energy content and relative onset timing. The synthesized Peak Ground Acceleration (PGA) estimates are also consistent with observations. We develop a set of metrics that allow us to assess the training process's stability and tune model hyperparameters. We further show that the trained generator network can interpolate to conditions where no earthquake ground motion recordings exist. Our approach allows the on-demand synthesis of accelerograms for engineering purposes.https://authors.library.caltech.edu/records/je98a-2c414HypoSVI: Hypocenter inversion with Stein variational inference and Physics Informed Neural Networks
https://resolver.caltech.edu/CaltechAUTHORS:20210225-132745364
Authors: Smith, Jonathan D.; Ross, Zachary E.; Azizzadenesheli, Kamyar; Muir, Jack B.
Year: 2021
DOI: 10.48550/arXiv.2101.03271
We introduce a scheme for probabilistic hypocenter inversion with Stein variational inference. Our approach uses a differentiable forward model in the form of a physics-informed neural network, which we train to solve the Eikonal equation. This allows for rapid approximation of the posterior by iteratively optimizing a collection of particles against a kernelized Stein discrepancy. We show that the method is well-equipped to handle highly non-convex posterior distributions, which are common in hypocentral inverse problems. A suite of experiments is performed to examine the influence of the various hyperparameters. Once trained, the method is valid for any network geometry within the study area without the need to build travel time tables. We show that the computational demands scale efficiently with the number of differential times, making it ideal for large-N sensing technologies like Distributed Acoustic Sensing.https://authors.library.caltech.edu/records/zy9td-0ks32Bayesian framework for inversion of second-order stress glut moments: application to the 2020 M_w 7.7 Caribbean Earthquake
https://resolver.caltech.edu/CaltechAUTHORS:20210730-182807109
Authors: Atterholt, James; Ross, Zachary E.
Year: 2021
We present a fully Bayesian inverse scheme to determine second moments of the stress glut using teleseismic earthquake seismograms. The second moments form a low-dimensional, physically-motivated representation of the rupture process that captures its spatial extent, source duration, and directivity effects. We determine an ensemble of second moment solutions by employing Hamiltonian Monte Carlo and automatic differentiation to efficiently approximate the posterior. Our method explicitly constrains the parameter space to be symmetric positive definite, ensuring the derived source properties have physically meaningful values. The framework accounts for the autocorrelation structure of the errors and incorporates hyperpriors on the uncertainty. We validate the methodology using a synthetic test and subsequently apply it to the 2020 M_w 7.7 Caribbean earthquake. The second moments determined for this event indicate the rupture was nearly unilateral and relatively compact along-strike. The solutions from this inverse framework can resolve ambiguities between slip distributions with minimal a priori assumptions on the rupture process.https://authors.library.caltech.edu/records/md31r-0eb69Seismic wave propagation and inversion with Neural Operators
https://resolver.caltech.edu/CaltechAUTHORS:20211006-164248015
Authors: Yang, Yan; Gao, Angela F.; Castellanos, Jorge C.; Ross, Zachary E.; Azizzadenesheli, Kamyar; Clayton, Robert W.
Year: 2021
DOI: 10.48550/arXiv.2108.05421
Seismic wave propagation forms the basis for most aspects of seismological research, yet solving the wave equation is a major computational burden that inhibits the progress of research. This is exaspirated by the fact that new simulations must be performed when the velocity structure or source location is perturbed. Here, we explore a prototype framework for learning general solutions using a recently developed machine learning paradigm called Neural Operator. A trained Neural Operator can compute a solution in negligible time for any velocity structure or source location. We develop a scheme to train Neural Operators on an ensemble of simulations performed with random velocity models and source locations. As Neural Operators are grid-free, it is possible to evaluate solutions on higher resolution velocity models than trained on, providing additional computational efficiency. We illustrate the method with the 2D acoustic wave equation and demonstrate the method's applicability to seismic tomography, using reverse mode automatic differentiation to compute gradients of the wavefield with respect to the velocity structure. The developed procedure is nearly an order of magnitude faster than using conventional numerical methods for full waveform inversion.https://authors.library.caltech.edu/records/xwp00-2fy53Generative Adversarial Neural Operators
https://resolver.caltech.edu/CaltechAUTHORS:20220714-212515070
Authors: Rahman, Md Ashiqur; Florez, Manuel A.; Anandkumar, Anima; Ross, Zachary E.; Azizzadenesheli, Kamyar
Year: 2022
DOI: 10.48550/arXiv.arXiv.2205.03017
We propose the generative adversarial neural operator (GANO), a generative model paradigm for learning probabilities on infinite-dimensional function spaces. The natural sciences and engineering are known to have many types of data that are sampled from infinite-dimensional function spaces, where classical finite-dimensional deep generative adversarial networks (GANs) may not be directly applicable. GANO generalizes the GAN framework and allows for the sampling of functions by learning push-forward operator maps in infinite-dimensional spaces. GANO consists of two main components, a generator neural operator and a discriminator neural functional. The inputs to the generator are samples of functions from a user-specified probability measure, e.g., Gaussian random field (GRF), and the generator outputs are synthetic data functions. The input to the discriminator is either a real or synthetic data function. In this work, we instantiate GANO using the Wasserstein criterion and show how the Wasserstein loss can be computed in infinite-dimensional spaces. We empirically study GANOs in controlled cases where both input and output functions are samples from GRFs and compare its performance to the finite-dimensional counterpart GAN. We empirically study the efficacy of GANO on real-world function data of volcanic activities and show its superior performance over GAN. Furthermore, we find that for the function-based data considered, GANOs are more stable to train than GANs and require less hyperparameter optimization.https://authors.library.caltech.edu/records/9ecxn-t8p79Finite Source Properties of Large Strike-Slip Earthquakes
https://resolver.caltech.edu/CaltechAUTHORS:20230327-962779000.5
Authors: Atterholt, James; Ross, Zachary E.
Year: 2023
DOI: 10.22541/essoar.167898487.74643181/v1
Earthquake ruptures are complex physical processes that may vary with the structure and tectonics of the region in which they occur. Characterizing the factors controlling this variability would provide fundamental constraints on the physics of earthquakes and faults. We investigate this by determining finite source properties from second moments of the stress glut for a global dataset of large strike-slip earthquakes. Our approach uses a Bayesian inverse formulation with teleseismic body and surface waves, which yields a low-dimensional probabilistic description of rupture properties including spatial extent, directivity, and duration. This technique is useful for comparing events because it makes only minor geometric constraints, avoids bias due to rupture velocity parameterization, and yields a full ensemble of possible solutions given the uncertainties of the data. We apply this framework to all great strike-slip earthquakes of the past three decades, and we use the resultant second moments to compare source quantities like directivity ratio, rectilinearity, stress drop, and depth extent. We find that most strike-slip earthquakes have a large component of unilateral directivity, and many of these earthquakes show a mixture of unilateral and bilateral behavior. We also notice that oceanic intraplate earthquakes usually rupture a much larger width of the seismogenic zone than other strike-slip earthquakes, suggesting these earthquakes consistently breach the expected thermal boundary for oceanic ruptures. We also use these second moments to resolve nodal plane ambiguity for the large oceanic intraplate earthquakes and find that the rupture orientation is usually unaligned with encompassing fossil fracture zones.https://authors.library.caltech.edu/records/kmz31-awy62