Poroelasticity Benchmarking for FEM on Analytical Solutions

E. Holzbecher
Georg-August Universitt Gttingen
GZG Applied Geology, Goldschmidtstr. 3, 37077 Gttingen, GERMANY; E-mail: eholzbe@gwdg.de
Abstract: We examine the poroelastics mode,
which couples hydraulics and mechanics by
some basic benchmarks. For cases with
analytical solutions we check the accuracy for
changing
meshes
and
calculate
the
convergence rate.
Keywords: Poroelastics, porous media flow,
cavity
problem,
analytical
solution,
convergence rate

1. Introduction
In poroelastics hydraulic and mechanical
processes are coupled, which allows the
simultaneous modeling of these processes
(HM-coupling) in porous media. Applications
for this type of modeling are generally
necessary in all situations, in which flow
processes through a porous material are
accompanied by deformations. The pore
pressure as a variable is responsible for fluid
flow and movements of the solid, and also
depends on both fluid and solid states.
There are several fields of applications, in
which poroelastical modeling, as described, is
relevant. In geological systems high flow
velocities may lead to changes of the solid
material, as in the vicinity of pumping or
injecting wells. Concerning the exploitation of
oil and gas reservoirs poroelasticity may
become crucial. The technique of hydraulic
fracking utilizes the mentioned coupling.
Problems of land subsidence are often related
to changes in the subsurface flow system.
Aside from soil and rocks the coupling of
hydraulic and mechanical features becomes
relevant in artificial materials, for example for
hydro gels and for swelling, highly
pronounced in sponges. Tissue mechanics is a
branch of material sciences concerned with
such substances (Cowin & Doty 2007).
The model considers Darcy-flow within
porous medium and elastic behaviour of the
solid material. We compute deformation and
pore pressures within the entire domain. The
coupling with poroelasticity is achieved using

an additional source term in the fluid mass

conservation equation.

2. Mathematical description
The mechanical model is given by the
equations
= FV , = s
s s 0 = C : ( 0 inel ) pI
=

1
2

( ( u )

+ u

(1)

p
+ ( q ) = Q V
t
t
k
q = p

with stress tensors and s, volume force

vector FV, strain tensor , deformation vector
u, pore pressure p, elastic tensor C and Biot
constant . Subscript 0 denotes initial values,
and inel indicates the inelastic part of the
strain.
For zero initial stress and strain, elastic
material and gravity as only force the
mechanical model reduces to:
= ( + b ) g
= C pI
=

1
2

( ( u )

+ u

(2)

with porosity , fluid density and bulk

density b.
In the model mechanical and hydraulic
processes are coupled with each other. The
hydraulic model is described by:

p
S + ( q ) = Q V
t
t
(3)
k
q = p

with Darcy velocity q, permeability k, fluid

dynamic viscosity , storage parameter S, fluid
source/sink-term Q, volumetric strain V.
The Biot coefficient appears in the stressstrain relation (second equation of systems (1)
and (2)) as a weighting factor for the influence
of fluid pressure. Moreover appears as
coefficient in an additional storage term in the

flow equation (first equation of system (3)).

For the steady state the coupling is only oneway, as the time derivatives in the flow
equation vanish. For the transient case the
coupling between the mechanical and the
hydraulic is two-way: the fluid pressure
influences the stress-strain relation and thus
the deformations. Vice versa the flow equation
is changed via the volumetric strain. In general
both models have to be performed
simultaneously in order to take the mutual
effects between mechanics and hydraulic into
account.
The analytical approach with parameter
was introduced by M.A. Biot already in the
first half of the last century (Biot, 1941) and is
used until today with only minor modifications
(for example: Morency et al., 2007). Carcione
et al. (2010) present a review on the relevance
of
Biot's
theory
in
computational
poroelasticity.

3. Numerical Model Set-up

The mathematical finite element code
COMSOL Multiphysics (2012) for the
solution a systems of partial differential
equations is suited for a big variety of
couplings of different physical phenomena.
Here we focus on the interaction between
hydraulics and mechanics. However, to extend
the models to account for thermal processes
within that software environment is a straightforward task and will be reported elsewhere.
COMSOL Multiphysics allows easy operation
via a graphical user interface. Several
toolboxes allow the specific set-up of models
and input of parameters. For the studies here
we utilized the structural mechanics toolbox.
Using COMSOL Multiphysics poroelastic
models have been examined against analytical
solutions by Souley & Thoraval (2011). The
software was applied for various fluid
injection problems (Freeman et al., 2008), for
CO2 storage (Bjrnara et al. 2010) and for
tight gas reservoirs (Hagemann et al. 2012).
Spatafora & Boschetti (2010) study
permeation tests. Rmond et al. (2008) as well
as Nguyen et al. (2011) apply the software for
a micro scale poroelastic problems.
We present benchmarks in spherical and in
cylindrical geometry and compare with
analytical solutions. Meshes are shown in
Figure 1.

Figure 1. Benchmark model regions with meshes;

for spherical cavity in 3D (top) and circular cavity
in 2D (bottom)

Concerning flow a constant hydraulic

gradient is assumed (in Figure 1 from right to
the left). All other outer boundaries obey a noflow condition, i.e. it is assumed that all
boundaries are far enough away from the
pumping regime on the vertical axis through
the center of the model region in order to
remain unchanged. The borehole is cut out of
the permeable layers, and a mass flux
condition is required at the so constructed
inner surfaces, in order to achieve infiltration
and pumping. In the baserock formation the
borehole is not modeled, as the influence of
the borehole on both hydraulic and mechanical
regime in the large scale are surely not of
importance. On the other hand meshing
produces a significantly increased number of
degrees of freedom of the numerical problem,
if the borehole is considered entirely in the
model geometry.
Concerning mechanics we allow free
deformations in the interior and all boundaries,
except the base at the bottom, where we
require the spatial system to be fixed. Here we

present results for the steady state, as these

represent the final maximum response of the
geo-system to the changed constant stress
regime.
Values for poroelastic parameters are
adopted from various sources, as reported by
Isaacs et al. (2008), Cappa (2009), Reyer and

Philipp (2013). The elastic tensor is specified

in our case by the Young modulus and the
Poisson number. The reference parameters
used in the model are gathered in Table 1.
Most values were adopted from Grandhi et al.
(2002).

Table 1. Reference Parameter Settings

Parameters
Young modulus E [GPa]

52.0

Radius R [m]

Poisson ratio [-]

0.1486

Length L [m]

Permeability [m2]
Porosity [-]

110-17
0.08

Vertical stress [GPa]

Max. horizontal stress H
[GPa]
Min. hor. stress / Max.
hor. stress [-]

Density [kg/m3]

2422

Biot parameter [-]

0.25

0.311 (2D)
1.0 (3D)
8.0 (2D)
10.0 (3D)
0.0644
0.1481
1.7/2.3

4. Analytical Solutions
For the elastic problem of a circular hole in
an infinite homogeneous model region and
unidirectional stress a solution was first
published by Kirsch (1898). For direction
dependent maximum and minimum principal
stresses H and h at the outer boundaries, the
solution can be generalized to:
1
2
= ( H + h ) 1 + ( R / r )
2
(4)
1
4
( H h ) 1 + 3 ( R / r ) cos(2 )
2
(Jaeger & Cook, 1979).
The deformation u in radial coordinates for
the unidirectional problem h = 0 is given by

(
(

r 2 ( + 1) + R 2 +

H
4
ur =
2

R
8 r 2 R ( + 1) 2 cos(2 )
r

u =

In the bi-axial tension case (p=p0 at outer

boundaries) for the 2D problem the solution is:

(
)
= p (1 + ( R / r ) )

r = p0 1 ( R / r )

with radial velocity:

ur =

p0 2
( r ( 1) + 2 R 2 ) (7)
4 r

For the spherical cavity in a hydrostatic

stress field, the analytical solution for the
diagonal elements of the stress tensor are
given by

r = p0 1 ( R / r )
(5)

H 2
R4
2
R ( 1) + r + 2 sin(2 )
r
4 r

= 3 4
with
and shear stiffness
= E / 2(1 + ) (Muskhelisvili 1953).

(6)

= = p0 1 + 1 2 ( R / r )

(8)

(Poulos & Davis, 1974) for Dirichlet boundary

condition for pressure p=p0 at the outer
boundary. Off-diagonal terms vanish.

5. Results
For the circular cavity, modelled in 2D,
Figure 2 shows the distribution of Tresca stress

a result.

Figure 2: Tresca stress [Pa] of 2D reference

problem in deformed mesh

Figure 3: Stress tensor components in dependence

of distance from inner radius for circular hole (2D);
comparison with analytical solution (Length in [m],
stresses in [Pa])

Figure 4: Stress tensor components in dependence

of distance from inner radius for spherical hole
(3D); comparison with analytical solution

A comparison between analytical and

numerical results is depicted in Figure 3. The
numerical model according to visual check
performs quite well.
The spherical cavity, modelled in 3D, was
modelled first in the solid mechanics mode.
The comparison, depicted in Figure 4 also
shows a good agreement between numerical
and analytical results as in the 2D case.
A more detailed study of the error is based
on the mesh refinements. Here we changed the
mesh following the default mesh options in
COMSOL Multiphysics, i.e. using the settings:
extremely coarse, extra coarse, coarser, coarse,
normal, fine, finer. The results are shown in
Table 2. The corresponding execution times
increased from 4 s for the extremely coarse
mesh up to 62 s for the finer mesh.
Table 2: Mesh dependence of error for the
spherical cavity benchmark
DOF
Max. error
Convergence
[Pa]
rate
31179
215300
35541
35720
3.00
45867
24470
2.69
56691
18942
2.58
91317
16976
2.44
131874
9112
2.29
287544
7056
2.14
The maximum error is evaluated for the xcomponent of the stress tensor Sxx. The error
decreases mesh refinement. The convergence
rate is calculated following the derivations of
Bradji & Holzbecher (2007, 2008). Thus we
find a convergence rate decreasing from a
value of 3 for coarse mesh refinements down
to 2 for the fine meshes.
As a first parameter study for the 2D
problem we first investigated the influence of
the model extension. The analytical solutions,
given above are valid for a set-up with infinite
length.
Figure 5 depicts the angular stress along
the hole boundary (0-90) for lengths, as
indicated by the length value in the legend (in
[m]). The highest deviation from the analytical
solution is for the smallest model, and with
increasing model length the numerical results
converge against the analytical solution. For
further parameter studies we decided that
length L=8 is a sufficient and appropriate
value.

Figure 5: Angular stress along hole boundary

in dependence of model extension (in [m])

Figure 6: Angular stress along hole boundary

in dependence of stress contrast at outer
boundaries

As next step we examine the influence of

the stress contrast ( h / H ) on the
performance of the numerical models. Figure 6
depicts the results. The analytical solutions are
indicated by circle markers, while the lines
represent the outcome of the numerical
simulations.
Obviously the numerical model performs
well. It can be seen that for low contrasts the
angular stress component changes its sign, i.e.
it changes between tensional and pressurized
regimes (compare: Grandhi et al. 2002).

Concerning accuracy it turns out that

deviation between analytical and numerical
solutions is higher for lower contrasts.
For k=0 we therefore checked the accuracy
in more detail. Figure 7 shows the mesh
dependency for that case, calculated for L=8.
An initial mesh with 658 triangular elements
(refined at the hole boundary) was refined four
times. The figure shows the radial stress along
the hole boundary, as modelled. The value
corresponds with the model error, as the radial
stress is zero for the modelled situation, and
thus indicates the performance.

Figure 7: Radial stress error along hole boundary in dependence of grid refinement

For this variable we examined the mesh

dependence of the error. For a wider model
region L=64 the convergence rates were
checked, following the work outlined by Bradji
& Holzbecher (2007, 2008). For the error of
angular stress in the max. norm for successive
mesh refinement we observe a convergence to
the limit convergence rate of 2 from below.
Details are given in Table 3.
Table 3: Error evaluation for the 2 benchmark
DOF
Max. error
Convergence
[Pa]
rate
3592
1.6843e7
13949
6.0152e6
1.5179
54967
1.9848e6
1.6171
218219
6.1165e5
1.7075
869587
1.6777e5
1.8713
3471779
45316
1.8910
In contrast to the former mesh dependence
study the mesh refinement was performed
manually. The DOF (degrees of freedom) thus
roughly increases by a factor of four from one
mesh to the other. Thus the mesh sequence is a
much better choice for the determination of the
convergence rate, than using the predefined
meshes as used before for the 3D cavity. Thus
we see observe an increase of the convergence
rate up to 2.
For h = 0 we examined the deformation
itself also. Figure 8 depicts the numerical values
and analytical solutions. For radial as well as for
angular deformations, numerical and analytical
solutions coincide well.

Figure 8: Deformation in elastic uni-axial stress

problem

6. Conclusions
Cavity models have been examined in two
and three dimensions. Comparisons with
analytical solutions show that the finite element
models, set up using COMSOL Multiphysics,
perform well.
With increasing mesh refinement the
convergence rate for the 2D model becomes
almost quadratic for the 2D cavity. For the 3D
cavity it is almost cubic for coarse meshes.
However, using the mesh refinements,
predefined in COMSOL Multiphysics, the
convergence rate decreases to almost quadratic.

References
1. Biot, M.A., General theory of threedimensional consolidation, J. of Applied Physics
12, 155-164 (1941)
2. Bjrnara T.J., Aker E., Cuisiat F., Skurtveit
E., Modeling CO2 storage using coupled
reservoir-geomechanical analysis, COMSOL
Conf., Paris, Proceedings (2010)
3. Bradji A., Holzbecher E., On the
convergence order of COMSOL solutions,
COMSOL Conf., Grenoble (2007)
4. Bradji A., Holzbecher E., On the
Convergence Order in Sobolev Norms of
COMSOL Solutions, COMSOL Conf., Budapest
(2008)
5. Cappa, F., Modelling fluid transfer and slip
in a fault zone when integrating heterogeneous
hydromechanical characteristics in its internal
structure, Geophys. J. Int. 178, 13571362
(2009)
6. Carcione J.M., Morency C., Santos J.E.,
Computational poroelasticity a review,
Geophys. 75(5), 75A229-75A243 (2010)
7. Cowin S.C., Doty S.B., Tissue Mechanics,
Springer, New York (2007)
8. Freeman T.T., Chalaturnyk R.J., Bogdanov
I.I., Fully coupled thermo-hydro-mechanical
modeling by COMSOL Multiphysics, with
applications
in
reservoir
geomechanical
characterization, COMSOL Conf., Boston,
Proceedings (2008)
9. Grandi S., Rao R., Nafi Toksz M.,
Geomechanical modeling of in-situ stresses
around a borehole, MIT lab reports
(http://erl.mit.edu/lab-reports.php) (2002)
10. Hagemann B., Wegner J., Ganzer L.,
Investigation of hydraulic fracture re-orientation

effects in tight gas reservoirs, COMSOL Conf.,

Milan, Proceedings (2012)
11. Isaacs A.J., Evans J.J., Kolesar P.T., Nohara
T., Composition, microstructures, and petrophysics of the Mozumi fault, Japan: In situ
analyses of fault zone properties and structure in
sedimentary rocks from shallow crustal levels, J.
Geophys.
Res.
113,
B12408,
doi:10.1029/2007JB005314 (2008)
12. Jaeger J.C., Cook N.G., Fundamentals of
Rock Mechanics, Chapman and Hall, London
(1979)
13. Kirsch, G. (1898). Die Theorie der Elastizitat
und die Bedrfnisse der Festigkeitslehre. Veit.
Deit. Ing. 42 (28), 797-807.
14. Morency C., Huismans R.S., Beaumont C.,
Fullsack P., A numerical model for coupled fluid
flow and matrix deformation with applications to
disequilibrium compaction and delta stability, J.
Geophys. Res.112, B10407 (2007)
15. Muskhelishvili, N.I., Some Basic Problems
of the Mathematical Theory of Elasticity,
Noordhoff, Groningen (1953)
16. Nguyen V.-H., Lemaire T., Naili S.,
Influence of interstitial bone microcracks on
strain-induced fluid flow, Biomech. Model
Mechanobiol. 10, 963-972 (2011)
17. Poulos H.G., Davis E.H., Elastic Solutions
for Soil and Rock Mechanics, Wiley, New York
(1974)
18. Rmond A., Naili S., Lemaire, T., Interstitial
fluid flow in the osteon with spatial gradients of
mechanical properties: a finite element study,
Biomech. Model Mechanobiol. 7, 487-495
(2008)
19. Reyer D., Phillip S.L., Correlations of rock
mechanical and physical properties with uniaxial
compressive strengths of sedimentary rocks from
the Northwest German Basin, to appear (2013)
20. Souley M., Thoraval A., Nonlinear
mechanical and poromechanical analyses:
comparison with analytical solutions, COMSOL
Conf., Stuttgart, Proceedings (2011)
21. Spatafora G., Boschetti, F., Finite element
analysis of permeation tests on articular cartilage
under different testing conditions using
COMSOL Multiphysics, COMSOL Conf., Paris,
Proceedings (2010)

Acknowledgement
The author appreciates the support of
'Niederschsisches Ministerium fr Wissenschaft und
Kultur' and Baker Hughes within the GeBo G7
project.