|
Titel |
A mesh-free approach to numerical rock mechanics simulations |
VerfasserIn |
Gunnar Jansen, Boris Galvan, Stephen Miller |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250089917
|
Publikation (Nr.) |
EGU/EGU2014-4130.pdf |
|
|
|
Zusammenfassung |
Numerical simulation of the nucleation, growth, and coalescence of fracture networks is a
fundamental aspect of lithospheric geodynamics and engineering applications such as
enhanced geothermal systems, hydraulic fracturing and CO2 sequestration. Modeling the
underlying mechanics is challenging because of several numerical difficulties. In
particular, fracture path evolution predicted by mesh-based models can be heavily
affected by numerical resolution of the chosen discretization scheme. Additionally,
large deformations can lead to numerical errors associated with highly deformed
elements.
We are developing algorithms that simulate fracture nucleation and growth using
mesh-free methods that overcome the difficulties arising from the mesh-sensitivity of
conventional mesh-based methods. We implemented a mesh-free local Petrov-Galerkin
method (MLPG), which is based on the local weak form of the problem under consideration.
This method requires no mesh for interpolation or integration, and thus may be well-suited to
handle strain localization occurring during fracture development. Interpolation
is performed using moving least squares approximation (MLS) shape functions,
and since nodal integration is performed locally, this approach can be parallelized
efficiently.
We present a mesh-free 2D elasto-plastic model for geomaterials that includes frictional
hardening and cohesion softening using the Mohr-Coulomb failure criterion to simulate
fracture network evolution and dynamic fracture propagation. Model performance is further
enhanced through parallelization by utilising a hybrid CPU/GPU cluster using the PETSc
library. We outline the implementation of the developed code, and evaluate its performance
from a series of benchmark simulations. |
|
|
|
|
|