Volume 129, Number 2, January 2020
|Number of page(s)||7|
|Section||Electromagnetism, Optics, Acoustics, Heat Transfer, Classical Mechanics, and Fluid Dynamics|
|Published online||19 February 2020|
Bayesian inference of material properties in disordered media using sound characteristics
1 Indian Institute of Technology Bombay, Department of Earth Sciences - Mumbai, 400076, India
2 Shell Technology Center Bangalore, Seismic Analytics - Bengaluru, Karnataka, 562149, India
Received: 24 June 2019
Accepted: 27 January 2020
A simulation study of sound wave propagation is used to characterise material properties. The synthetic model consisting of a one-dimensional chain of particles with different mass distributions is used to generate space-time responses. The standard deviation of this mass distribution is chosen as the disorder parameter ξ, which eventually gives information about the longitudinal wave propagation (P-wave). The space-time responses obtained from the synthetic model are used by a Bayesian inference technique to infer the most probable values of the parameter ξ, which would have caused the responses. While it is found that, during the sampling process, the sampler could effectively choose a region in the neighborhood of the true causative ξ, results show that the estimation in a low disordered setting are acceptable when . A frequency domain analysis is also carried out which shows superior performance in identifying the true region, suggesting that it is a more suitable approach for inference in random media.
PACS: 46.40.Cd – Mechanical wave propagation (including diffraction, scattering, and dispersion) / 46.65.+g – Random phenomena and media / 43.60.Cg – Statistical properties of signals and noise
© EPLA, 2020
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.