Finite Difference wave simulations in homogenized geomodels Homogenization theory

in: 2017 Ring Meeting, pages 1--8, ASGA

Abstract

When dealing with the simulation of wave propagation in complex models, accounting for the effect of small-scale features is a major issue, because i) the generation of a complex mesh is usually required and ii) the computation cost drastically increases with the resolution of the model. Small heterogeneities indeed imply both small elements and a small time-step. Although numerical capacity and speed significantly grew in the last decades, taking small scales into account properly is still a challenging task. Homogenization, by upscaling the model, allows the correct account of the effect of small heterogeneities in wave simulations while highly decreasing computation costs and meshing efforts. We here implement this method in C++ and we output the upscaled parameters on a regular grid. Then, we perform 3D finite difference wave simulations using these parameters to test the accuracy and the efficiency of the homogenization. We first present the theory of the technique and then we analyse the results obtained in a finely-layered model. To exemplify the interest of our technique, we also present the homogenization of a more complex model with a fold and a fault.

Download / Links

BibTeX Reference

@INPROCEEDINGS{Schuh-senlis2017,
    author = { Schuh-senlis, Melchior and Cupillard, Paul and Mazuyer, Antoine and Irakarama, Modeste },
     title = { Finite Difference wave simulations in homogenized geomodels Homogenization theory },
 booktitle = { 2017 Ring Meeting },
      year = { 2017 },
     pages = { 1--8 },
 publisher = { ASGA },
  abstract = { When dealing with the simulation of wave propagation in complex models, accounting for the effect of small-scale features is a major issue, because i) the generation of a complex mesh is usually required and ii) the computation cost drastically increases with the resolution of the model. Small heterogeneities indeed imply both small elements and a small time-step. Although numerical capacity and speed significantly grew in the last decades, taking small scales into account properly is still a challenging task. Homogenization, by upscaling the model, allows the correct account of the effect of small heterogeneities in wave simulations while highly decreasing computation costs and meshing efforts. We here implement this method in C++ and we output the upscaled parameters on a regular grid. Then, we perform 3D finite difference wave simulations using these parameters to test the accuracy and the efficiency of the homogenization. We first present the theory of the technique and then we analyse the results obtained in a finely-layered model. To exemplify the interest of our technique, we also present the homogenization of a more complex model with a fold and a fault. }
}