An official website of the United States government
SUMMARY Geodetic observations of post-seismic deformation due to afterslip and viscoelastic relaxation can be used to infer fault and lithosphere rheologies by combining the observations with mechanical models of post-seismic processes. However, estimating the spatial distributions of rheological parameters remains challenging because it requires solving a nonlinear inverse problem with a high-dimensional parameter space and potentially computationally expensive forward model. Here we introduce an inversion method to estimate spatially varying fault and lithospheric rheological parameters in a mechanical model of post-seismic deformation using geodetic time series. The forward model combines afterslip and viscoelastic relaxation governed by a velocity-strengthening frictional rheology and a power-law Burgers rheology, respectively, and incorporates the mechanical coupling between coseismic slip, afterslip and viscoelastic relaxation. The inversion method estimates spatially varying fault frictional parameters, viscoelastic constitutive parameters and coseismic stress change. We formulate the inverse problem in a Bayesian framework to quantify the uncertainties of the estimated parameters. To solve this problem with reasonable computational costs, we develop an algorithm to estimate the mean and covariance matrix of the posterior probability distribution based on an ensemble Kalman filter. We validate our method through numerical tests using a 2-D forward model and synthetic post-seismic GNSS time-series. The test results suggest that our method can estimate the spatially varying rheological parameters and their uncertainties reasonably well with tolerable computational costs. Our method can also recover spatially and temporally varying afterslip, viscous strain and effective viscosities and can distinguish the contributions of afterslip and viscoelastic relaxation to observed post-seismic deformation.
more »
« less
A three‐dimensional hybrid finite element — spectral boundary integral method for modeling earthquakes in complex unbounded domains
Abstract We present a 3D hybrid method which combines the finite element method (FEM) and the spectral boundary integral method (SBIM) to model nonlinear problems in unbounded domains. The flexibility of FEM is used to model the complex, heterogeneous, and nonlinear part— such as the dynamic rupture along a fault with near fault plasticity—and the high accuracy and computational efficiency of SBIM is used to simulate the exterior half spaces perfectly truncating all incident waves. The exact truncation allows us to greatly reduce the domain of spatial discretization compared to a traditional FEM approach, leading to considerable savings in computational time and memory requirements. The coupling of FEM and SBIM is achieved by the exchange of traction and displacement boundary conditions at the computationally defined boundary. The method is suited to implementation on massively parallel computers. We validate the developed method by means of a benchmark problem. Three more complex examples with a low velocity fault zone, low velocity off‐fault inclusion, and interaction of multiple faults, respectively, demonstrate the capability of the hybrid scheme in solving problems of very large sizes. Finally, we discuss potential applications of the hybrid method for problems in geophysics and engineering.
more »
« less
- Award ID(s):
- 1753249
- PAR ID:
- 10445391
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- International Journal for Numerical Methods in Engineering
- Volume:
- 122
- Issue:
- 23
- ISSN:
- 0029-5981
- Page Range / eLocation ID:
- p. 6905-6923
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The reaction-diffusion system is naturally used in chemistry to represent substances reacting and diffusing over the spatial domain. Its solution illustrates the underlying process of a chemical reaction and displays diverse spatial patterns of the substances. Numerical methods like finite element method (FEM) are widely used to derive the approximate solution for the reaction-diffusion system. However, these methods require long computation time and huge computation resources when the system becomes complex. In this paper, we study the physics of a two-dimensional one-component reaction-diffusion system by using machine learning. An encoder-decoder based convolutional neural network (CNN) is designed and trained to directly predict the concentration distribution, bypassing the expensive FEM calculation process. Different simulation parameters, boundary conditions, geometry configurations and time are considered as the input features of the proposed learning model. In particular, the trained CNN model manages to learn the time-dependent behaviour of the reaction-diffusion system through the input time feature. Thus, the model is capable of providing concentration prediction at certain time directly with high test accuracy (mean relative error <3.04%) and 300 times faster than the traditional FEM. Our CNN-based learning model provides a rapid and accurate tool for predicting the concentration distribution of the reaction-diffusion system.more » « less
-
Abstract Immersed boundary methods are high-order accurate computational tools used to model geometrically complex problems in computational mechanics. While traditional finite element methods require the construction of high-quality boundary-fitted meshes, immersed boundary methods instead embed the computational domain in a structured background grid. Interpolation-based immersed boundary methods augment existing finite element software to non-invasively implement immersed boundary capabilities through extraction. Extraction interpolates the structured background basis as a linear combination of Lagrange polynomials defined on a foreground mesh, creating an interpolated basis that can be easily integrated by existing methods. This work extends the interpolation-based immersed isogeometric method to multi-material and multi-physics problems. Beginning from level-set descriptions of domain geometries, Heaviside enrichment is implemented to accommodate discontinuities in state variable fields across material interfaces. Adaptive refinement with truncated hierarchically refined B-splines (THB-splines) is used to both improve interface geometry representations and to resolve large solution gradients near interfaces. Multi-physics problems typically involve coupled fields where each field has unique discretization requirements. This work presents a novel discretization method for coupled problems through the application of extraction, using a single foreground mesh for all fields. Numerical examples illustrate optimal convergence rates for this method in both 2D and 3D, for partial differential equations representing heat conduction, linear elasticity, and a coupled thermo-mechanical problem. The utility of this method is demonstrated through image-based analysis of a composite sample, where in addition to circumventing typical meshing difficulties, this method reduces the required degrees of freedom when compared to classical boundary-fitted finite element methods.more » « less
-
Abstract As insects fly, their wings generate complex wake structures that play a crucial role in their aerodynamic force production. This study focuses on utilizing reduced order modeling techniques to gain valuable insights into the fluid dynamic principles underlying insect flight. Specifically, we used an immersed-boundary-method-based computational fluid dynamics (CFD) solver to simulate a hovering hawkmoth’s wake, and then identified the most energetic modes of the wake using proper orthogonal decomposition (POD). Furthermore, we employed a sparse identification of nonlinear dynamics (SINDy) approach to find a simple reduced order model that relates the most energetic POD modes. Through this process, we formulated multiple different models incorporating varying numbers of POD modes. To compare the accuracy of these models, we utilized a force survey method to estimate the aerodynamic forces produced by the hawkmoth’s wings. This force survey method uses an impulse-based approach to calculate the aerodynamic lift and drag based solely on the velocity and vorticity information provided by the model. By comparing the estimated aerodynamic force with the actual force production calculated by the CFD solver, we were able to find the simplest model that still provides an accurate representation of the complex wake produced by the hovering hawkmoth wings. We also evaluated the stability and accuracy of this model as the number of flapping cycles increases with time. The reduced order modeling of insect flight has important implications for the design and control of bio-inspired micro-aerial vehicles. In addition, it holds the potential to reduce the computational cost associated with high-fidelity CFD simulations of complex flows.more » « less
-
Abstract Melt pool dynamics in metal additive manufacturing (AM) is critical to process stability, microstructure formation, and final properties of the printed materials. Physics-based simulation, including computational fluid dynamics (CFD), is the dominant approach to predict melt pool dynamics. However, the physics-based simulation approaches suffer from the inherent issue of very high computational cost. This paper provides a physics-informed machine learning method by integrating the conventional neural networks with the governing physical laws to predict the melt pool dynamics, such as temperature, velocity, and pressure, without using any training data on velocity and pressure. This approach avoids solving the nonlinear Navier–Stokes equation numerically, which significantly reduces the computational cost (if including the cost of velocity data generation). The difficult-to-determine parameters' values of the governing equations can also be inferred through data-driven discovery. In addition, the physics-informed neural network (PINN) architecture has been optimized for efficient model training. The data-efficient PINN model is attributed to the extra penalty by incorporating governing PDEs, initial conditions, and boundary conditions in the PINN model.more » « less
