Você está na página 1de 7

Dynamic behavior of unsaturated soils: the full formulation

and its finite element solution

N. Ravichandran & K.K. Muraleetharan


School of Civil Engineering and Environmental Science, The University of Oklahoma, Norman, Oklahoma, USA

ABSTRACT: Unsaturated soils are three-phase porous media consisting of a solid skeleton, pore liquid and
pore gas. The overall behavior of unsaturated soils is influenced not only by these three bulk phases, but also
by the interfaces between them. Therefore, a rigorous solution for the behavior of unsaturated soils requires the
consideration of the interactions between the bulk phases and the interfaces at the governing equation level.
The coupled equations governing the dynamic behavior of unsaturated soils can be derived based on the theory
of mixtures with interfaces (TMI). In a typical finite element solution of these governing equations, the solid
skeleton displacement, u, pore liquid pressure, pl , and pore gas pressure, pg , are used as the nodal unknowns.
In order to use the above-mentioned nodal unknowns, it is necessary to make certain assumptions about the
relative accelerations of the fluids. For certain high frequency problems, such as the blast loading of unsaturated
soils, however, it is necessary to consider the relative acceleration terms. This in turn leads to the full finite
element formulation, where solid displacement, u, pore liquid displacement, Ul , pore gas displacement, Ug ,
pore liquid pressure, pl , and pore gas pressure, pg , all have to be considered as the nodal unknowns. The full
formulation increases the number simultaneous equations that have to be solved in a finite element analysis
and requires a larger computational capacity. A framework-based approach can be used to develop an efficient
parallel finite element code for solving such large problems. The full formulation for the unsaturated soils and
the finite element framework are briefly described. Simulation of a fully saturated clay embankment subjected
to earthquake loading is presented to show the capabilities of the framework-based computer code developed as
a part of this research.

1 INTRODUCTION
of the bulk phases and interfaces at the governing
equation level is important for correctly modeling the
1.1 Unsaturated soils
behavior of unsaturated soils.
Unsaturated soils are encountered near the earth sur- In a typical solution procedure of the coupled gov-
face where most of the engineering structures are erning equations using finite element technique, solid
ultimately supported. Even though geotechnical engi- displacement, u, pore liquid pressure, pl , and pore gas
neering projects encounter saturated, dry, and unsatu- pressure, pg , are used as nodal unknowns (Schrefler
rated soils, most of the past studies have been done et al. 1990; Muraleetharan & Wei, 1999; Wei 2001).
only on saturated and dry soils. Saturated and dry The degree of saturation can also be introduced in
soils can become unsaturated due to seasonal varia- the formulation and considered as a nodal unknown
tions. Unlike saturated or dry soils, an unsaturated soil (Xikui & Zienkiewicz, 1992; Schrefler et al. 1990). In
has three phases, solid, liquid and gas, which make all of the above mentioned studies, both the convec-
the study of unsaturated soils more complicated. In tive and the relative acceleration terms of the liquid
addition to these three bulk phases there exist three and gas phases were neglected in the momentum
interfaces: solid-liquid interface, liquid-gas interface balance equations. The convective acceleration terms
and solid-gas interface. pose numerical difficulties in the finite element imple-
Among the three interfaces mentioned above, the mentation of the governing equations. The relative
liquid-gas interface, characterized by surface tension, acceleration terms, however, can be taken into consid-
plays an important roll in the behavior of unsaturated eration in the full formulation. In the full formulation,
soil system. Wei (2001) and Wei & Muraleetharan solid displacement, u, liquid displacement, Ul , gas dis-
(2002a,b) have shown that the solid-gas and solid- placement, Ug , pore liquid pressure, pl , and pore gas
liquid interfaces can also affect the overall behavior of pressure, pg , are all considered as nodal unknowns.The
the unsaturated soil system. Coupling the interaction only difficulty in solving the full formulation using

489
Copyright 2005 Taylor & Francis Group plc, London, UK
the finite element technique is that a larger computa- = ns s + nl l + ng g ; v = relative velocity of
tional capacity is typically needed. In this paper a new phase defined by v = v vs ; and = l, g.
approach, framework-based finite element technique, The linear momentum balance equations for the
is used to overcome this difficulty. liquid and gas phases are basically the Darcys flow
A framework-based finite element approach is one equation for these phases. The major resistance to
of the powerful and efficient methods for developing the flow of liquid and gas is the drag force from the
extensible finite element applications. A framework solid skeleton. The main driving force is the pressure
represents a collection of common software compo- gradient existing across the flow path.
nents for building different finite element codes. The Linear momentum balance for the liquid and gas
basic premise behind the use of a framework is the phases can be written in the following form.
recognition of a common set of tasks that must be
accomplished in writing any finite element application
code. These tasks can be factored out of the appli-
cation codes and collected into a single tool kit. The
goal is to separate the physics aspects from the com- where p = pressure of phase; and = viscosity of
puter science aspects of writing a computer code and phase.
thereby making the code development more efficient. Taking the material time derivative with respect to
The framework also makes it easy to develop codes the solid phase, we get:
to take advantage of modern computational advances
such as parallel and distributed computing.
Following brief descriptions of the fully coupled
governing equations and the finite element framework,
we present the results from the analysis of a saturated
clay embankment subjected to base shaking obtained
using TeraScale_Dysac (Muraleetharan et al. 2003),
a framework-based soil analysis code developed as a
part of this research. where k = intrinsic permeability defined by

2 GOVERNING EQUATIONS

2.1 Momentum balance equations The convective acceleration terms in the above equa-
The momentum balance equation for a bulk phase is tions are neglected because of numerical difficulties,
given as but in contrast to previous studies, relative acceleration
terms are retained. The momentum balance equations
for the mixture, liquid and gas can be written in indicial
notation as follows.
where = s, l, g (s = solid, l = liquid, g = gas);
v = absolute velocity of phase; t = stress tensor of
phase; n = volume fraction of phase; = density
of phase; and b = gravitational acceleration.
The motion of the fluids should be described with
respect to the solid phase. Therefore, the material time
derivatives of liquid and gas velocities should also be
taken with respect to the solid phase motion. Introduc-
ing the relative velocities of the liquid and gas phases,

the linear momentum balance for the unsaturated soil where w i = Darcys
 velocity of phase defined by
mixture is given as
w i = n vi vis ; and = l, g.
In the full formulation, solid displacement, liquid
displacement and gas displacement are all part of the
nodal variables. Therefore, the relative Darcys veloc-
ities in the above equations can be substituted with
absolute velocities as follows.

where = total stress tensor; = total density of the


mixture defined by

490
Copyright 2005 Taylor & Francis Group plc, London, UK
2.2 Mass balance equations l g
u y , Uy , U y
The mass balance equations are written for each indi-
vidual phase and then combined to obtain independent
l g
equations. u x, U x, U x
l g u
Mass balance for the solid phase: p, p
Ul
g
x= U
pl
By taking the material time derivative with respect pg
to the solid phase and assuming that the solid grains
are incompressible compared to other two phases, the
Figure 1. Nodal unknowns for a 2-D problem.
above equation reduces to:
Mass balance for the solid and gas phase is
expressed as:

Mass balance for the liquid and gas phases:

where the coefficients are defined as:

2.3 Constitutive equation for the volume


fraction of the liquid phase
The volume fraction of the liquid phase is expressed
in terms of the volumetric strain and matric suction
(Wei, 2001).
The finite element equations corresponding to the gov-
erning equations (6), (7), (13) and (14) can be written
in the following compact form:
where s = matric suction defined by s = pg pl ;
v = volumetric strain of the solid skeleton. From
Equation (11)
where M = mass matrix; C = damping matrix;
K = pore fluid stiffness matrix; p = internal load vec-
tor; f = external load vector; and x = nodal unknown
vector.
In the case of reduced formulation (u pl pg ),
Substituting the above equation in the mass balance there will be only four unknowns per node for a
equations, the mass balance for the solid and liquid 2-dimensional problem. In the case of the full for-
phase is expressed as: mulation (u Ul Ug pl pg ), there will be eight
unknowns for a 2-dimensional problem (Fig. 1).There-
fore the full formulation requires a larger computa-
tional capacity, especially when solving 3-dimensional
problems (11 unknowns per node). A framework-
where the coefficients are defined as: based finite element approach can be used to develop
an efficient finite element code to solve such large
problems. The capabilities typically provided by a
framework are described in the next section.

and K l = bulk modulus of liquid phase defined by


3 FINITE ELEMENT FRAMEWORK

Developing an efficient finite element computer code


for analyzing any problem requires a lot of computer

491
Copyright 2005 Taylor & Francis Group plc, London, UK
science knowledge. Historically, the finite element 6. Flexible Data Model: a generic container object
application developer has been an engineer trained (Data Model) for holding meta data. The data
in some specific field of mechanics. Typically they model has an XML representation, which allows
spend an inordinate portion of their software devel- the framework to read and write the data across the
opment time dealing with computer science aspects internet.
rather than focusing on algorithms and mechanics. A 7. Surface Physics Infrastructure: an abstraction of
finite element framework insulates the engineer from a surface and a set of interfaces for performing
the computer science aspects and lets the engineer con- physics operations on the surface. Support for scal-
centrate on the computational mechanics aspects of the able parallel surface physics algorithms is the key
application. to the design of this service.
A finite element framework represents a collection
The TeraScale_Dysac is structured as shown in
of software components for building finite element
Fig. 2. Any finite element procedure can be accom-
applications. By collecting these common calculations
modated within the six major classes shown in Fig. 2.
into a single toolkit, the framework enables the appli-
Main is the overall global construct and holds a
cation developer to use these components in many
number of Procedures. It stores global information
different applications. Consequently, the amount of
central to the application. Procedure holds a collec-
work and code required for developing and maintain-
tion of Physics objects. The time marching algorithm,
ing an application is greatly reduced. In addition to
advancing the states of the fields and reading and
the computer science details, the framework may also
writing the results are implemented in this class. The
provide the tools for parallel coding. The major coding
Physics layer of the framework maps cleanly onto the
required for running an analysis on multiple processors
traditional notion of a finite element code. Physics
is hidden behind the framework. The major advantage
object holds a set of Element Blocks, which have dif-
of using a framework is, that the physics of a prob-
ferent element formulations. Element Blocks can hold
lem is completely separated from the computer science
any number of Sections. Section is the physical rep-
aspect of solving that problem.
resentation of the element at each integration station.
The finite element framework developed by
Each Section will hold only a single Material object.
the TeraScale, LLC (Cedar Crest, New Mexico,
The Material object is designed in such a way that any
U.S.A.) is used to develop a soil analysis code called
existing stress-strain relationship code segment (linear
TeraScale_Dysac (Muraleetharan et al. 2003). The
elastic, nonlinear elastic, or elastoplastic) can be easily
TeraScale framework provides the following services.
used with a few lines of modification.
1. Finite Element Procedural Abstractions: a set of
algorithmic abstractions that is common to all
4 EXAMPLE SIMULATION
finite element applications (e.g. physics algorithm,
element based algorithm, material algorithm, etc.).
The results from the simulation of a saturated
2. Finite Element Infrastructure: the basic core
clay embankment subjected to base shaking using
libraries of finite element applications necessary
TeraScale_Dysac are presented here. The fully cou-
to support general-purpose physics applications.
pled finite element equations for the saturated soil have
This consists of an element library, which holds
all the discrete calculus methods necessary for the Main
application.
3. Multi-physics Infrastructure: the capability to
rapidly couple single physics applications into a Procedure P1 Procedure P2
multi-physics package based on solution transfer
operator. These operators are parallel in nature and
highly scalable. Physics 101 Physics 201
4. Scalable Linear Equation Solver: a common inter-
face to linear solver services for scalable solution
of sparse systems of equations on distributed and
Element Block A Element Block B
shared memory parallel architectures. This abstrac-
tion allows the interchange or selection of differ-
ent linear algebra solvers and/or preconditioners
without modifying the physics implementation. Section S1 Section S2
5. Parallel Mesh Object (PMO): a high-level mesh
object for the finite element application. The PMO
is fundamentally structured to support parallel Material M1 Material M2
computation in Single Program Multiple Data man-
ner (SPMD). Figure 2. Organization of framework classes.

492
Copyright 2005 Taylor & Francis Group plc, London, UK
been implemented into TeraScale_Dysac. The sum- bounding surface concept is that plastic deformations
mary of these equations can be found in Muraleetharan can occur for stress points inside the bounding surface.
et al. (1994). The unsaturated governing equations In classical plasticity theory, no plastic deformations
presented earlier are currently being implemented. are allowed inside the yield surface. The classical yield
surface formulation is transformed into a bounding
4.1 Stress-strain behavior of the solid skeleton surface formulation based on the concept that for any
stress point inside the surface, a unique image point
If we want to predict the behavior of structures made can be defined on the surface by means of a radial
of soils accurately under complex loading conditions, mapping rule (Fig. 3). The value of the plastic modu-
it is important to model the stress-strain behavior of lus depends on the distance between the actual stress
the soil skeleton realistically. Elastic material behavior point and its image on the bounding surface. The
assumption is used in many analyses. This is mainly gradient of the bounding surface is used to define the
because of the computational efficiency. However, direction of the plastic loading for the actual stress
elastic behavior assumption is a very poor assumption point. The utilization of an elastoplastic constitutive
for soils under most loading conditions. Here, we use model requires additional computational capacity due
an elastoplastic phenomenological constitutive model to the local iterations involved within the constitutive
based on the bounding surface concept to represent equation code segment.
the stress-strain behavior of the soil skeleton (Dafalias The finite element mesh for the saturated clay
& Herrmann, 1986). The prominent feature of the (Speswhite Kaolin) embankment is shown in Fig. 4.
The initial stresses, required for the dynamic anal-
J ysis were calculated using a static analysis. Dur-
L

ing the dynamic analysis the El Centro Earthquake


CS

Hyperbola
acceleration-time history shown in Fig. 5 was applied
F at the base of the embankment. The properties of the
F Speswhite Kaolin are listed in Table 1.
Ellipse1
I, J
Ellipse2 Table 1. Material properties of Speswhite Kaolin.
I, J
I Model parameter Value

Projection Center Elastic Nuclei Initial state parameters


Initial void ratio 1.40
Figure 3. Schematic of the bounding surface. (I and J are Traditional model parameters
stress invariants, CSL = Critical State Line) Slope of virgin consolidation line () 0.25
Slope of swelling line () 0.05
Slope of critical state line in compression (Mc ) 0.88
N21 EL 184 Ratio of extension to compression for slope of 1.00
critical state line (Me /Mc )
N12
N12 Poissons ratio 0.30
40
40 m Limit pressure (Pl ) 30.40
Atmospheric pressure (Pa ) 101.40
186 m Surface configuration parameters
Bounding surface shape parameter in 2.40
Figure 4. Finite element mesh. compression (Rc )
Bounding surface shape parameter in 0.01
compression (Ac )
0.4 Bounding surface shape parameter (T) 0.01
Ratio of extension to compression value 0.92
Acceleration (g)

0.2 of R (Re /Rc )


Ratio of extension to compression value 1.20
0.0 of A, (Ae /Ac )
Projection center parameter (C) 0.00
Elastic nucleus parameter (S) 1.00
-0.2
Hardening parameters
Hardening parameter (m) 0.02
-0.4
Hardening parameter in compression (Hc ) 3.00
0 5 10 15 20 25 30
Ratio of extension to compression value for 1.00
Time (sec) hardening parameter (He /Hc )
Hardening parameter on the I-Axis (Ho ) 2.00
Figure 5. El Centro earthquake acceleration-time history.

493
Copyright 2005 Taylor & Francis Group plc, London, UK
120 3.8
Pore water pressure (kPa)

100
80
60
98.3 kPa
40
20
0 Figure 9. Excess pore water pressure contours super-
0 5 10 15 20 25 30 imposed on the deformed mesh at the end of the analysis
Time (sec) (t = 30 sec, displacements magnified by 4).

Figure 6. Dynamic pore water pressure-time history for 5 SUMMARY AND CONCLUSION
EL 184.
The full formulation for the unsaturated soils con-
0.4 sidering solid displacement, liquid displacement, gas
displacement, liquid pressure and gas pressure as the
Vertical Disp. (m)

0.0 nodal unknowns is derived. The relative acceleration


terms can be taken into account in the full formula-
-0.4 tion. The implementation of the finite element equa-
tions using the TeraScale framework is presented. The
-0.8 example simulation of a saturated clay embankment
subjected to earthquake loading shows the impor-
-1.2 tance of using an elastoplastic constitutive model to
0 5 10 15 20 25 30 represent the stress-strain behavior of the soil skeleton.
Time (sec)

Figure 7. Vertical displacement-time history at Node 21. ACKNOWLEDGMENTS

1.0 This research was supported by the U.S. National Sci-


ence Foundation (Grant No. CMS-0112950) and this
Horizontal Disp. (m)

0.5 support is acknowledged.


0.0

-0.5 REFERENCES
-1.0 Dafalias, Y.F., & Herrmann, L.R. 1986. Bounding Surface
Plasticity II:Application to Isotropic Cohesive Soils. Jour-
-1.5
0 5 10 15 20 25 30 nal of the Engineering Mechanics Division, ASCE, v 112,
n 12: 12631291.
Time (sec)
Muraleetharan, K.K., Mish, K.D., & Arulanandan, K. 1994.
A Fully Coupled Nonlinear Dynamic Analysis Procedure
Figure 8. Horizontal displacement-time history at Node 12. and its Verification Using Centrifuge Test Results. Inter-
national Journal for Numerical and Analytical Methods in
Geomechanics, v 18: 305325.
The pore water pressure within Element 184 Muraleetharan, K.K., Ravichandran, N., &Taylor, L.M. 2003.
increases at the beginning and then stabilizes (Fig. 6). TeraScale_Dysac:TeraScale Dynamic SoilAnalysis Code.
A permanent settlement of about 1 m is observed at Computer Code, School of Civil Engineering and Envi-
the top of the embankment (Fig. 7). The horizontal ronmental Science, University of Oklahoma, Norman,
displacement at Node 12, shown in (Fig. 8), also indi- Oklahoma, October.
cates permanent movement towards the left. It would Muraleetharan, K.K., & Wei, C.-F. 1999. U_DYSAC2: Unsat-
not have been possible to predict these permanent urated Dynamic Soil Analysis Code for 2-dimensional
displacements without using an elastoplastic consti- Problems. Computer Code, School of Civil Engineer-
ing and Environmental Science, University of Oklahoma,
tutive model to describe the stress-strain behavior of Norman, Oklahoma, October.
the solid skeleton. The excess pore water pressure Schrefler, B.A., Simoni, L., Li Xikui, & Zienkiwicz, O.C.
contours at t = 30 seconds are super imposed on the 1990. Mechanics of Partially Saturated Porous Media. In
deformed mesh and shown in (Fig. 9). High pore pres- Desai, C. S. and Gioda, G. (eds), Numerical Methods and
sures are concentrated around bottom-center of the Constitutive Modelling In Geomechanics, CISM Lecture
embankment. Notes, Springer-Verlag, Wein: 169209.

494
Copyright 2005 Taylor & Francis Group plc, London, UK
Wei, C.-F. 2001. Static and Dynamic Behavior of Multi- Wei, C.-F., & Muraleetharan, K.K. 2002b. A continuum the-
Phase Porous Media: Governing Equations and Finite ory of porous media saturated by multiphase fluids: II.
Element Implementation. PhD Dissertation, University of Lagrangian description and variational structure. Interna-
Oklahoma, Norman. tional Journal of Engineering Science, v 40: 18351854.
Wei, C.-F., & Muraleetharan, K.K. 2002a. A continuum Xikui, L., & Zienkiwicz, O.C. 1992. Multiphase Flow in
theory of porous media saturated by multiphase fluids: Deforming Porous Media and Finite Element Solution,
I. Linear poroelasticity. International Journal of Engineer- Computers and Structures, v 45, n 2: 211227.
ing Science, v 40: 18071833.

495
Copyright 2005 Taylor & Francis Group plc, London, UK

Você também pode gostar