← Back to Sessions

Full-Waveform Inversion of Seismic Input Motions in a Near-Surface Domain Truncated by Wave-Absorbing Boundary Conditions

Session: Numerical Modeling of Earthquake Motion, Rupture Dynamics, Seismic Noise, Wave Propagation and Inverse Problems II

Type: Oral

Date: 4/23/2021

Presentation Time: 02:15 PM Pacific

Description: 

There is a need to estimate complex seismic input motions in a near-surface domain, without resorting to the hypocenter, from restricted seismic measurement data. Thus, engineers can replicate responses within structures and soils after an earthquake occurrence by using the estimated seismic inputs and evaluate the impact of an earthquake on the built environment.

We present a partial differential equation (PDE)-constrained optimization method for solving the full waveform inverse problem of determining complex, incoherent SH wave input motions, which are modeled as traction on a wave-absorbing boundary condition (WABC), in a 2D domain that is truncated by a WABC. The method includes the discretize-then-optimize (DTO) approach, the finite element method (FEM), which is applied to solve state and adjoint problems, and the conjugate-gradient scheme, determining the desired search path throughout a minimization process.

The numerical results show that incident, inclined plane waves, cannot be fully reconstructed if using only the top surface sensors because some portions of the incident waves do not reach the top sensors in a truncated domain. In order to improve the inversion performance, a vertical array of sensors on the side boundary of a domain should be included. Second, a sufficiently large number of sensors must be employed to improve the algorithm's inversion performance. Third, the minimizer suffers more from solution multiplicity when it identifies incident waves of a higher dominant frequency. Fourth, the larger value of the inversion error in the reconstructed traction does not necessarily translate to an error of the same magnitude in the corresponding reconstructed wave responses in the computational domain due to the intrinsic low-pass filtering of the FEM wave solver. Lastly, our presented inversion algorithm's accuracy is not compromised by the material complexity of a background domain.

Presenting Author: Bruno Peruqui Guidio

Student Presenter: No

Authors