Statistical flaw characterization through Bayesian shape inversion from scattered wave observations
| Autor: | Amanda K. Criner, Jerry A. McMahan |
|---|---|
| Rok vydání: | 2016 |
| Předmět: | |
| Zdroj: | AIP Conference Proceedings. |
| ISSN: | 0094-243X |
| DOI: | 10.1063/1.4940608 |
| Popis: | A method is discussed to characterize the shape of a flaw from noisy far-field measurements of a scattered wave. The scattering model employed is a two-dimensional Helmholtz equation which quantifies scattering due to interrogating signals from various physical phenomena such as acoustics or electromagnetics. The well-known inherent ill-posedness of the inverse scattering problem is addressed via Bayesian regularization. The method is loosely related to the approach described in [1] which uses the framework of [2] to prove the well-posedness of the infinite-dimensional problem and derive estimates of the error for a particular discretization approach. The method computes the posterior probability density for the flaw shape from the scattered field observations, taking into account prior assumptions which are used to describe any a priori knowledge of the flaw. We describe the computational approach to the forward problem as well as the Markov chain Monte Carlo (MCMC) based approach to approximating the po... |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|