Description

A variational deduction of second gradient poroelasticity I: general theory

Categories

Published

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Share

Transcript

A VARIATIONAL DEDUCTION OF SECOND GRADIENT POROELASTICITYPART I: GENERAL THEORY
G
IULIO
S
CIARRA
,
F
RANCESCO DELL
’I
SOLA
,
N
ICOLETTA
I
ANIRO
AND
A
NGELA
M
ADEO
Second gradient theories have to be used to capture how local micro heterogeneities macroscopicallyaffect the behavior of a continuum. In this paper a conﬁgurational space for a solid matrix ﬁlled by an
unknown amount of ﬂuid is introduced. The Euler–Lagrange equations valid for second gradient porome-
chanics, generalizing those due to Biot, are deduced by means of a Lagrangian variational formulation.
Starting from a generalized Clausius–Duhem inequality, valid in the framework of second gradient the-
ories, the existence of a macroscopic solid skeleton Lagrangian deformation energy, depending on the
solid strain and the Lagrangian ﬂuid mass density as well as on their Lagrangian gradients, is proven.
1. Introduction
Poroelasticity stems from Biot’s pioneering contributions on consolidating ﬂuid saturated porous mate-
rials [Biot 1941] and now spans a lot of different interrelated topics, from geo- to biomechanics, wave
propagation, transport, unsaturated media, etc. Many of these topics are related to modeling coupled
phenomena (for example, chemomechanical swelling of shales [Dormieux et al. 2003; Coussy 2004], or
biomechanical models of cartilaginous tissues), and nonstandard constitutive features (for instance, infreezing materials [Coussy 2005]). In all these cases, complexity generally remains in rendering how
heterogeneities affect the macroscopic mechanical behavior of the overall material.
It is well known from the literature how microscopically heterogeneous materials can be described in
the framework of statistically homogeneous media [Torquato 2002] considering suitable generalizations
of the dilute approximation due to Eshelby [Nemat-Nasser and Hori 1993; Dormieux et al. 2006]; how-
ever, some lack in the general description of the homogenization procedure arises when dealing withheterogeneous materials, the characteristic length of which can be compared with the thickness of theregion where high deformation gradients occur. This could be due, for example, to external periodic
loading, the wavelength of which is comparable with the characteristic length of the material, or to phase
transition, etc.
From the macroscopic point of view the quoted modeling difﬁculties, arising when high gradientsoccur, are discussed in the framework of so called high gradient theories [Germain 1973], where theassumption of locality in the characterization of the material response is relaxed. In these theories,the momentum balance equation reads in a more complex way than the classical one used for Cauchycontinua. As a matter of fact, it is the divergence of the difference between the stress tensor and thedivergence of so-called hyperstresses that balance the external bulk forces. Stress and hyperstress are
introduced by a straightforward application of the principle of virtual power, as those quantities working
on the gradient of velocity and the second gradient of velocity, respectively [Casal 1972; Casal and
Keywords:
poromechanics, second gradient materials, lagrangian variational principle.
507
508 GIULIO SCIARRA, FRANCESCO DELL’ISOLA, NICOLETTA IANIRO
AND
ANGELA MADEO
Gouin 1988]. Even the classical Cauchy theorem is, in this context, revised by introducing dependence
of tractions not only on the outward normal unit vector but also on the local curvature of the boundary[dell’Isola and Seppecher 1997]; moreover symmetric and skew-symmetric couples (the actions called
“double-forces” by Germain) must be prescribed on the boundary in terms of the hyperstress tensor
together with contact edge forces along the lines where discontinuities of the normal vector occur.
Following the early papers on ﬂuid capillarity [Casal 1972; Casal and Gouin 1988], the second gradient
model can indeed be introduced by means of a variational formulation where the considered Helmholtzfree energy depends both on the strain and the strain gradient tensors.
In the case of ﬂuids, second gradient theories are typically applied for modeling phase transition
phenomena [de Gennes 1985] or for modeling wetting phenomena [de Gennes 1985], when a character-
istic length, say the thickness of a liquid ﬁlm on a wall, becomes comparable with the thickness of the
liquid/vapor interface [Seppecher 1993], annihilation (nucleation) of spherical droplets, when the radius
of curvature is of the same order of the thickness of the interface [dell’Isola et al. 1996], or topological
transition [Lowengrub and Truskinovsky 1998].
In the case of solids, second gradient theories are applied, for instance, when modeling the failure
process associated with strain localization [Elhers 1992; Vardoulakis and Aifantis 1995; Chambon et al.
2004]. To the best of our knowledge, second gradient theories are very seldom applied in the mechanics
of porous materials [dell’Isola et al. 2003] and no second gradient poromechanical model, consistent
with the classical Biot theory, is available except the one presented in [Sciarra et al. 2007]. As gradient
ﬂuid models, second gradient poromechanics will be capable of providing signiﬁcant corrections tothe classical Biot model when considering porous media with characteristic length comparable to the
thickness of the region where high ﬂuid density (deformation) gradients occur. We refer, for instance, to
crack/pore opening phenomena triggered by strain gradients or ﬂuid percolation, the characteristic length
being in this case the average length of the space between grains (pores).
Several authors have focused their attention on the development of homogenization procedures capable
of rendering the heterogeneous response of the material at the microlevel by means of a second gradi-
ent macroscopic constitutive relation [Pideri and Seppecher 1997; Camar-Eddine and Seppecher 2003];
however, very few contributions seem to address this problem in the framework of averaging techniques
[Drugan and Willis 1996; Gologanu and Leblond 1997; Koutzetzova et al. 2002]. The present work does
not investigate the microscopic interpretation of second gradient poromechanics, but directly discusses
its macroscopic formulation. It is divided into two papers: in the ﬁrst paper the basics of kinematics,
Section 2; the physical principles, Section 3; the thermodynamical restrictions, Section 4; and in Section
5 the variational deduction of the governing equations for a second gradient ﬂuid ﬁlled porous material
are presented.
In particular, in Section 2 a purely macroscopic Lagrangian description of motion is addressed byintroducing two placement maps in
χ
s
and
φ
f
(Equation (1)). We do not explicitly distinguish which
part of the current conﬁguration of the ﬂuid ﬁlled porous material is occupied at any time
t
by the solid
and ﬂuid constituents, this information being partially included by the solid and ﬂuid apparent density
ﬁelds, which provide the density of solid/ﬂuid mass with respect to the volume of the porous system
(Equation (5)).
A VARIATIONAL DEDUCTION OF SECOND GRADIENT POROELASTICITY I 509
The deformation power, or stress working (Equation (12)), following Truesdell [1977] is deduced in
Section 3 starting from the second gradient expression of power of external forces (Equation (9)) Cauchy
theorem (Equation (10)) and balance of global momentum (see (11)).
In the spirit of Coussy et al. [1998] and Coussy [2004] thermodynamical restrictions on admissible
constitutive relations are stated in Section 4, ﬁnding out a suitable overall potential, deﬁned on thereference conﬁguration of the solid skeleton. This last depends on the skeleton strain tensor and the
ﬂuid mass content, measured in the reference conﬁguration of the solid, as well as on their Lagrangian
gradients, in Equation (18).
Finally a deduction of the governing equations is presented in Section 5, based on the principle of
virtual works, by requiring the variation of the internal energy to be equal to the virtual work of external
and dissipative forces (see (19)). A second gradient extension of the two classical Biot equations of
motion [Coussy 2004; Sciarra et al. 2007], endowed with the corresponding transversality conditions on
the boundary, is therefore formulated (see Equations (30)–(33)). Generalizing the treatment developed,
for example, by Baek and Srinivasa [2004] for ﬁrst gradient theories, one of the equations of motionfound by means of a variational principle is interpreted as the balance law for total momentum, when
suitable deﬁnitions of the global stress and hyperstress tensors are introduced (see (34)).
In a subsequent paper (Part II, to be published in a forthcoming issue of this journal), an application of
the second gradient model to the classical consolidation problem will be discussed. Our aim is to show
how the present model enriches the description of a well-known phenomenon, typical of geomechanics,
curing some of the weaknesses of the classical Terzaghi equation [von Terzaghi 1943]. In particular
we will ﬁgure out the behavior of the ﬂuid pressure during the consolidation process when varying the
initial pressures of the solid skeleton and/or the saturating ﬂuid. From the mathematical point of view,
the initial boundary value problem will be discussed according with the theory of linear pencils.
2. Kinematics of ﬂuid ﬁlled porous media and mass balances
The behavior of a ﬂuid ﬁlled porous material is described, in the framework of a macroscopic model,adopting a Lagrangian description of motion with respect to the reference conﬁguration of the solidskeleton. At any current time
t
the conﬁguration of the system is determined by the maps
χ
s
and
φ
f
,
deﬁned as
χ
s
:
s
×
→
, φ
f
:
s
×
→
f
,
(1)
where
α
(α
=
s
,
f
)
is the reference conﬁguration of the
α
-th constituent, while
is the Euclidean place
manifold, and
indicates a time interval. The map
χ
s
(
·
,
t
)
prescribes the current (time
t
) placement
x
of the skeleton material particle
X
s
in
s
. The map
φ
f
(
·
,
t
)
, on the other hand, identiﬁes the ﬂuidmaterial particle
X
f
in
f
which, at time
t
, occupies the same current place
x
as the solid particle
X
s
.Therefore the set of ﬂuid material particles ﬁlling the solid skeleton is unknown, to be determined bymeans of evolution equations. Both these maps are assumed to be at least diffeomorphisms on
. The
current conﬁguration
t
of the porous material is the image of
s
under
χ
s
(
·
,
t
)
. In accordance with
the properties of
χ
s
and
φ
f
it is straightforward to introduce the ﬂuid placement map as
χ
f
:
f
×
→
,
such that
χ
f
(
·
,
t
)
=
χ
s
(
·
,
t
)
◦
φ
f
(
·
,
t
)
−
1
,
510 GIULIO SCIARRA, FRANCESCO DELL’ISOLA, NICOLETTA IANIRO
AND
ANGELA MADEO
!
t
x
!
s
!
s
"
f
!
t
*
X
s
x
*
!
#
s
"
#
f
!
f
X
f
X
*
f
!
f
*
!
f -1
!
f *-1
Figure 1.
Lagrangian variations of the placement maps
χ
s
,
φ
f
, and
χ
f
.where
χ
f
(
·
,
t
)
is still a diffeomorphism on
. Figure 1 shows how the introduced maps operate on the
skeleton particle
X
s
∈
s
; admissible variations of the two maps
χ
s
(
·
,
t
)
and
φ
f
(
·
,
t
)
are also depicted,
in Section 5. In this way the space of conﬁgurations we will use has been introduced.Independently of
t
∈
, the Lagrangian gradients of
χ
s
and
φ
f
are introduced as
F
s
(
·
,
t
)
:
s
→
Lin
(
V
),
f
(
·
,
t
)
:
s
→
Lin
(
V
),
X
s
→ ∇
s
χ
s
(
X
s
,
t
),
X
s
→ ∇
s
φ
f
(
X
s
,
t
),
(2)
with
V
being the space of translations associated to the Euclidean place manifold. In Equation (2)
∇
s
indicates the Lagrangian gradient in the reference conﬁguration of the solid skeleton; analogously, the
gradient of
χ
f
is given by
F
f
X
f
,
t
=
F
s
(
X
s
,
t
).
f
(
X
s
,
t
)
−
1
, where
X
f
=
φ
f
(
X
s
,
t
)
.
1
In the following the ﬂuid Lagrangian gradient of
χ
f
will be indicated both by
F
f
or
∇
f
χ
f
whenconfusion can arise. Moreover, the time derivatives of
χ
s
and
χ
f
, say the Lagrangian velocities of the
solid skeleton and the ﬂuid, can be introduced asfor all
X
α
∈
α
,
α
(
X
α
,
·
)
:
→
V
,
t
→
d
χ
α
dt
(
X
α
,
t
)
.
We also introduce the Eulerian velocities
v
α
as the push-forward of
α
into the current domain
v
α
(
·
,
t
)
=
α
(
·
,
t
)
◦
χ
α
(
·
,
t
)
−
1
.
In the following we do not explicitly distinguish the map
χ
s
from its section
χ
s
(
·
,
t
)
if no ambiguity can
arise. Moreover we will distinguish between the Lagrangian gradient (
∇
s
) in the reference conﬁguration
of the solid skeleton and the Eulerian gradient (
∇
) with respect to the current position
x
. Analogously,
the solid Lagrangian and the Eulerian divergence operations will be noted by
div
s
and
div
, respectively.
All the classical transport formulas can be derived both for the solid and the ﬂuid quantities; in particular,
1
From now on we will indicate single, double and triple contraction between two tensors with
.
,
:
, and
...
respectively.
A VARIATIONAL DEDUCTION OF SECOND GRADIENT POROELASTICITY I 511
those ones for an image volume and oriented surface element turn to be
d
t
=
J
α
d
α
,
n
dS
t
=
J
α
F
−
T
α
.
n
α
dS
α
,
where
d
t
and
dS
t
represent the current elementary volume and elementary oriented surface corre-sponding to
d
α
and
dS
α
, respectively, where
J
α
=
det
F
α
, and where
n
and
n
α
are the outward unit
normal vectors to
dS
t
and
dS
α
. As far as only the solid constituent is concerned, we can understand that
deformation induces changes in both the lengths of the material vectors and the angles between them.
As it is well known, the Green–Lagrange strain tensor
ε
measures these changes, and is deﬁned as
ε
:=
12
F
T s
.
F
s
−
I
,
(3)where
I
clearly represents the second order identity tensor.The balance of mass both for the solid and the ﬂuid constituent are introduced as
α
=
t
ρ
α
d
t
=
const
=
α
ρ
0
α
d
α
, (α
=
s
,
f
),
(4)
where
α
is the total mass of the
α
-th constituent,
ρ
α
is the current apparent density of mass of the
α
-th
constituent per unit volume of the porous material, while
ρ
0
α
is the corresponding density in the reference
conﬁguration of the
α
-th constituent. When localizing, Equation (4) reads
ρ
α
J
α
=
ρ
0
α
, (α
=
s
,
f
),
or, in differential form,
d
α
ρ
α
dt
+
ρ
α
div
(
v
α
)
=
0
, (α
=
s
,
f
),
(5)
where
d
α
ρ
α
/
dt
represents the material time derivative relative to the motion of the
α
-th constituent. In
other words,
d
α
dt
:=
d dt
X
α
=
const
.
The macroscopic conservation laws could also be deduced in the framework of micromechanics[Dormieux and Ulm 2005; Dormieux et al. 2006] starting from a reﬁned model, where the solid and
the ﬂuid material particles occupy two disjoint subsets of the current conﬁguration, and considering an
average of the solid and ﬂuid microscopic mass balances. The macroscopic laws do involve the so called
apparent density of the constituents and suitable macroscopic velocity ﬁelds. For a detailed description
of the procedure which leads to averaged conservation laws we refer to the literature [Coussy 2004].
2.1.
Pull back of continuity equations.
It is clear that Equation (5) consists of Eulerian equations, mean-ing that they are deﬁned on the current conﬁguration of the porous medium. Following Wilmanski [1996]
and Coussy [2004] we want to write both these equations in the reference conﬁguration of the solidskeleton. With this purpose in mind let us deﬁne the relative ﬂuid mass ﬂow
w
as
w
:=
ρ
f
v
f
−
v
s
.
The use of this deﬁnition allows us to rearrange the ﬂuid continuity (5) in the form
d
s
ρ
f
dt
+
ρ
f
div
v
s
+
div
w
=
0
.
(6)

We Need Your Support

Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...Sign Now!

We are very appreciated for your Prompt Action!

x