English

Numerical methods for multiscale inverse problems

Numerical Analysis 2016-10-26 v3

Abstract

We consider the inverse problem of determining the highly oscillatory coefficient aϵa^\epsilon in partial differential equations of the form (aϵuϵ)+buϵ=f-\nabla\cdot (a^\epsilon\nabla u^\epsilon)+bu^\epsilon = f from given measurements of the solutions. Here, ϵ\epsilon indicates the smallest characteristic wavelength in the problem (0<ϵ10<\epsilon\ll1). In addition to the general difficulty of finding an inverse, the oscillatory nature of the forward problem creates an additional challenge of multiscale modeling, which is hard even for forward computations. The inverse problem in its full generality is typically ill-posed and one common approach is to replace the original problem with an effective parameter estimation problem. We will here include microscale features directly in the inverse problem and avoid ill-posedness by assuming that the microscale can be accurately represented by a low-dimensional parametrization. The basis for our inversion will be a coupling of the parametrization to analytic homogenization or a coupling to efficient multiscale numerical methods when analytic homogenization is not available. We will analyze the reduced problem, b=0b = 0, by proving uniqueness of the inverse in certain problem classes and by numerical examples and also include numerical model examples for medical imaging, b>0b > 0, and exploration seismology, b<0b < 0.

Keywords

Cite

@article{arxiv.1401.2431,
  title  = {Numerical methods for multiscale inverse problems},
  author = {Christina Frederick and Bjorn Engquist},
  journal= {arXiv preprint arXiv:1401.2431},
  year   = {2016}
}
R2 v1 2026-06-22T02:43:06.384Z