Fictitious Domain Methods for Fracture Models in Elasticity.

Monday, 15 December 2014
Sebastien Court1,2, Olivier Bodart2, Valerie Cayol3 and Jonas Koko2, (1)Laboratoire Magmas et Volcans, Clermont-Ferrand Cedex, France, (2)University Blaise Pascal Clermont-Ferrand II, Clermont-Ferrand, France, (3)University Jean Monnet Saint-Etienne, Saint-Etienne Cedex 02, France
As surface displacements depend non linearly on sources location and shape, simplifying assumptions are generally required to reduce computation time when inverting geodetic data. We present a generic Finite Element Method designed for pressurized or sheared cracks inside a linear elastic medium. A fictitious domain method is used to take the crack into account independently of the mesh. Besides the possibility of considering heterogeneous media, the approach permits the evolution of the crack through time or more generally through iterations: The goal is to change the less things we need when the crack geometry is modified; In particular no re-meshing is required (the boundary conditions at the level of the crack are imposed by Lagrange multipliers), leading to a gain of computation time and resources with respect to classic finite element methods.

This method is also robust with respect to the geometry, since we expect to observe the same behavior whatever the shape and the position of the crack. We present numerical experiments which highlight the accuracy of our method (using convergence curves), the optimality of errors, and the robustness with respect to the geometry (with computation of errors on some quantities for all kind of geometric configurations). We will also provide 2D benchmark tests. The method is then applied to Piton de la Fournaise volcano, considering a pressurized crack - inside a 3-dimensional domain - and the corresponding computation time and accuracy are compared with results from a mixed Boundary element method.

In order to determine the crack geometrical characteristics, and pressure, inversions are performed combining fictitious domain computations with a near neighborhood algorithm. Performances are compared with those obtained combining a mixed boundary element method with the same inversion algorithm.