Computational homogenization and reduced order modeling of dif- fusion processes in fluid-saturated porous media
Book, 2016
Fluid-saturated porous media are well-known for their manifold attenuation mechanisms
caused by wave-induced fluid flow between microscopic or mesoscopic heterogeneities. If
elastic waves propagate through such a medium, the pore space is heterogeneously com-
pacted resulting in local pressure gradients. These pressure gradients are equilibrated by
a redistribution of the viscous pore fluid (water, gas, oil, etc.), which causes pore pressure
diffusion. Hereby, part of the wave energy is lost, and the wave is attenuated. In this
contribution, we concentrate on the upscaling behavior of fluid-saturated rocks featuring
double porosity, patchy saturation or networks of fluid conduits. Based on Biot’s quasi-
static equations of consolidation and an appropriate hybrid-dimensional description of
the fracture networks, we establish a computational homogenization framework. To this
end, we assume the pressure diffusion to occur within mesoscopic volume elements much
smaller than the macroscopic wave length. Hence, diffusion takes place on a length scale
much smaller than the observer scale and is considered, from the macroscopic viewpoint,
as a local process. The heterogeneous poroelastic medium, therefore, is substituted by
a homogeneous macroscopic Cauchy medium with apparent viscoelastic properties. The
material properties of the substitute model are derived making use of a novel order reduc-
tion technique allowing for a numerically efficient treatment of the scale-transition. The
macroscopic internal variables representing the overall viscoelasticity are interpreted as
parameters controlling the activity of mesoscopic pressure modes identified by the Proper
Orthogonal Decomposition (POD) technique. We show that the resulting decoupled sys-
tem of evolution equations is equivalent to that of a generalized Maxwell-Zener model.
In the second part of this contribution we weaken the locality constraint and allow for
macroscopically observable seepage of the pore fluid. Hence, the character of the ho-
mogenization problem changes and the macroscopic substitute medium is considered as a
poroviscoelastic material. We derive a consistent computational homogenization scheme
and enrich the proposed order reduction concept by the required additional features. The
contribution is completed by numerous numerical investigations including validation tests
and extensive discussions of our findings.