Time-domain numerical simulations of multiple scattering to extract elastic effective wavenumbers

TL;DR

This paper introduces a numerical method for simulating elastic wave scattering in heterogeneous media to accurately determine effective wavenumbers, validating models against concrete data and various scattering theories.

Contribution

A novel time-domain numerical approach using immersed interface methods to extract effective wavenumbers in elastic media with scatterers, avoiding limitations of low concentration assumptions.

Findings

Numerical phase velocities and attenuations match analytical models.

Method effectively handles high scatterer concentrations.

Validation against multiple scattering theories confirms accuracy.

Abstract

Elastic wave propagation is studied in a heterogeneous 2-D medium consisting of an elastic matrix containing randomly distributed circular elastic inclusions. The aim of this study is to determine the effective wavenumbers when the incident wavelength is similar to the radius of the inclusions. A purely numerical methodology is presented, with which the limitations usually associated with low scatterer concentrations can be avoided. The elastodynamic equations are integrated by a fourth-order time-domain numerical scheme. An immersed interface method is used to accurately discretize the interfaces on a Cartesian grid. The effective field is extracted from the simulated data, and signal-processing tools are used to obtain the complex effective wavenumbers. The numerical reference solution thus-obtained can be used to check the validity of multiple scattering analytical models. The method is applied to the case of concrete. A parametric study is performed on longitudinal and transverse incident plane waves at various scatterers concentrations. The phase velocities and attenuations determined numerically are compared with predictions obtained with multiple scattering models, such as the Independent Scattering Approximation model, the Waterman-Truell model, and the more recent Conoir-Norris model.