Finite element analysis of model piles axially

loaded in sands
Vincenzo De Gennaro,* Roger Frank,* Imen Said**

1. Introduction between a granular soil and a rigid structure within

the framework of hardening plasticity in two-di-
When considering the interaction of two media mensional or axisymmetric conditions, will be
in contact with highly-distinct deformability charac- briefly described. The role of the surrounding soil
teristics failure is often accompanied by the forma- mass will also be investigated; elasticity, ideal plas-
tion within the more deformable medium of a ticity and strain hardening plasticity will be used to
rather thin zone oriented in the direction of the model the sand mass.
contact surface. This zone, called the soil-structure in- Finally, comparative analyses of pile tests using
terface, or simply interface, experiences intense strain FEM will be provided by means of the CESAR-
localization and plays the role of a kinematic discon- LCPC finite element code [HUMBERT, 1989]. Note
tinuity characterized by extremely high strain gradi- that one important issue in the case of piles is the
ents. definition of the initial state, following installation
Quite a large number of civil engineering struc- and prior to loading, both in the soil and at the in-
tures lie in contact with soils. Such is the case, for ex- terface. In the majority of the numerical applica-
ample, in soil-retaining walls, soil-anchorage rods, tions piles are wished in place, assuming that in-
soil-piles or micropiles, or soil-reinforcements (e.g. stallation effects, if any, have only a limited impact
terre arme, nailed soils). Failure in these struc- on their mechanical behaviour. Given the obstacles
tures occurs mainly due to progressive shearing and in generating accurate simulations of pile installa-
is often observed at the interface, in the softer me- tion via FEM (e.g. simulation of driving, boring,
dium (i.e. the soil mass), where stresses and strains etc.) this problem is still far from being resolved. A
are transmitted. The description of the mechanical new numerical strategy is outlined to tackle this is-
behaviour, mainly in terms of mobilized friction be- sue.
tween the structural element and the soil, must con- The proposed numerical investigations will be
sequently entail constitutive modelling of this heav- validated against the experimental results obtained
ily loaded region. during model pile loading tests using a calibration
In this work emphasis is given on the behaviour chamber [D E G ENNARO , 1999; D E G ENNARO and
of deep foundations and, more particularly, of the FRANK, 2005] and on a real site [CHOW, 1997; JARDINE
contact between a granular soil and a pile. In this et al., 1998 and 2005].
typical soil-structure interaction problem available
analyses of the mechanical behaviour of single piles
submitted to axial loads have shown that the soil- 2. Axisymmetric FEM modelling of soil-pile
pile interface exerts significant influence in defin- interface
ing structural stability conditions.
The paper will focus first on the general frame- When solving problems with boundary condi-
work of the soil-pile interaction modelling using the tions that integrate interfaces, the use of FEM for
finite element method (FEM); a description will be modelling contact with standard elements is often
provided of how contact problems have been tack- cumbersome, since the relative movement of solids
led using FEM. The constitutive modelling of the in contact is not, at first inspection, compatible with
interface and the soils mass will be then presented. the displacement continuity condition required at
The main features of the interface model MEPI-2D the nodes of adjacent elements (e.g. P OTTS and
[DE GENNARO and FRANK, 2002], formulated on pur- ZDRAVKOVIC, 2001).
pose for describing the behaviour of the interface The soil-pile interaction provides a typical
example of an interface problem. So-called
"conventional methods" for piles computing are
based on empirical correlations that lead to deter-
*
Universit Paris-Est, Navier (ENPC- CERMES), Paris, France mining the limit load-bearing capacity. Somewhat
**
Universit Paris-Est, Navier (ENPC- CERMES), Paris, France, less attention is usually paid to determining the
ENIT, Tunis, formerly URIG deformation characteristics corresponding with the

RIVISTA ITALIANA DI GEOTECNICA 2/2008

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 45

service state of a pile, although design guidelines ensure numerical stability of calculations [PANDE and
often suggest the computation of settlements and SHARMA, 1979; HOHBERG and SCHWEIGER, 1992].
horizontal displacements, since excessive values in In this study eight-node isoparametric elements
these parameters may affect structural functions have been employed for modelling the soil-pile con-
(e.g. FRANK et al., 2004). Among the set of analytical tact. In order to successfully adapt two-dimensional
or numerical methods that enable assessing pile isoparametric finite elements to the description of
behaviour and that incorporate soil strain behav- interface behaviour, two conditions are necessary:
iour, three methods have undergone major devel- first, the element must enable simple shearing
opments: (i) the finite element method, (ii) the mechanism by solely considering the normal and
boundary element method and (iii) the load transfer shear stresses acting upon the interface; second, the
function method (t-z or p-y). This work is concerned constitutive law associated with the element must al-
with the finite element method. low connecting these stresses at the interface with
the associated relative displacements.
The first condition can be easily fulfilled consid-
Thin layer element ering the specific geometry of the thin layer element
(Fig. 1) and defining an optimal aspect ratio for the
The basic principles of the finite element same element (i.e. the ratio L/t). For concision's sake
method and its application to solving problems such the theoretical developments are not presented in
as pile behaviour numerical analyses have been set this paper. One can refer to DE GENNARO and FRANK
forth by a number of authors over the past thirty [2005] or HOHBERG and SCHWEIGER [1992] for further
years (e.g. MAQTADIR and DESAI, 1986; MABSOUT et al., details. With standard rectangular finite elements it
1995; BRANSBY and SPRINGMAN, 1996; MOHAMEDZEIN is typically advised to consider L/t aspect ratios lower
et al., 1999; WAKAI et al., 1999; DE G ENNARO and than 10, whereas for a "thin layer" element an
FRANK, 2005; COMODROMOS et al., 2003; WEHNERT and accurate description of the interface kinematics is
VERMEER, 2004). Several methods for modelling dis- obtained with aspect ratio varying from 10 to 100,
continuous behaviour at the soil-structure interface depending on the constitutive law adopted at the
have been proposed: interface. This condition implies a specific mapping
- the use of "zero-thickness" special elements of stresses and strains over the element. Consider
with two nodes, where only the connection between the representation of the interface by 8-nodes
opposite nodes is considered active (e.g. GOODMAN et quadrangular elements (Fig. 1). In the local (m, n)
al., 1968; DAY and POTTS, 1994); coordinate system, the largest element dimension L
- the use of two-dimensional or, rarely, three-di- is parallel to the direction of axis m, while its
mensional limited thickness standard finite ele- smallest dimension (thickness) is equal to t. The
ments, also called "thin-layer" elements, by ascribing element axis m forms an angle with respect to the
the constitutive laws derived from elastoplasticity direction of x1 axis in the global coordinate system.
theory to these elements (e.g. DESAI et al. 1984); Local stresses n (normal) and (tangential) exerted
- the definition of "hybrid" methods, in which on a facet parallel to the direction of axis m may be
soil and structure are modelled separately and then computed based on knowledge of the stress
related by different types of compatibility equations components expressed within the global coordinate
for forces and displacements (e.g. KATONA, 1983;
SHENG et al., 2007).
With "zero-thickness" elements, displacement
continuity at the interface is ensured by either appli-
cation of a penalization method or adaptation of the
stiffness. Typically the relative sliding condition is
controlled by the Mohr-Coulomb criterion.
In the case of 2D "thin layer" elements the con-
tact problem is modelled by means of rectangular
isoparametric finite elements characterized by a rel-
atively high aspect ratio (ratio of the largest element
dimension to its smallest dimension) and constitu-
tive laws adapted to the mechanical behaviour of the
contact zone. The advantage of this method is that
standard (rectangular) finite elements are used, so
that no further element implementation is needed.
However, a parametric study of the influence of the Fig. 1 Thin layer element.
aspect ratio on the numerical results is necessary to Fig. 1 Elemento tipo thin layer.

APRILE - GIUGNO 2008

46 DE GENNARO - FRANK - SAID

system. Let {11, 22, 12}T be the pseudo-vector thickness is decreased (t 0), the terms of matrix B
containing the stress tensor components of stresses in (3), which contain the t/L ratio (i.e. the inverse of
in the global coordinate system and {n, }T the the element aspect ratio), become negligible. It can
pseudo-vector of stresses in the local coordinate be easily demonstrated that for the element shown
system, the relation between local and global in Figure 1, assuming = 0 for the sake of simplifi-
stresses can then be given as follows: cation, the average strains in the element are writ-
ten as follows:

(1)

(4)
Analogously, if {11, 22, 12}T represents the
pseudo-vector of strains in the global coordinate
system and {n, }T the pseudo-vector of strains in
the local coordinate system, the relation between lo-
+
cal and global strains is given by: where u and u- are the displacements of the ele-
ment nodes on the opposite sides along direction
and [un] and [um] are the relative displacements in
(2) both the normal (n) and tangential (m) directions of
the opposite sides of the interface element along di-
rection . In other words the relative displacements
Relations (1) and (2) can be used for describing of the interface element depend upon the set of el-
the behaviour of a medium with an "oriented plas- ement displacements computed on the two sides
ticity criterion", as suggested by FRANK et al., [1982]. perpendicular to direction (Fig. 1). Consequently,
Without any specific hypothesis on the ratio L/t of for elements with a high aspect ratio value, the
the element (i.e. for standard L/t ratios) this crite- strains computed at element nodes indicate the
rion corresponds more with the representation of a strain mode in simple shear characteristic of the in-
soil medium with a high density of discontinuities terface. The presence of parameter t in Equation (4)
equally oriented and requires an adaptation in or- is essential to relate interface stresses with their cor-
der to model an isolated thin soil layer such as the responding relative displacements, in accordance
interface. with the kinematic hypotheses introduced in order
Following the standard finite element proce- to describe the mechanical behaviour of the soil-
dures for the element shown in Figure 1, strains may structure interface. The behaviour of the interface is
be expressed as a function of nodal displacements in then described by considering the relative tangen-
the following matrix form: tial displacement [um] and relative normal displace-
ment [un] as state variables and the shear stress
and normal stress n mobilized at the interface as as-
sociated variables. The local behaviour (whether
elastic or elastoplastic) may thus be written as fol-
(3)
lows:

(5)
where L is the derivation operator, N contains the
shape (or interpolation) functions Ni = Ni(,) at
the element nodes, and ui and vi (for i = 1,8) repre- where elements Kij are local stiffness. Note that the
sent nodal displacements of the element, in direc- local constitutive law for the interface element can
tions x1 and x2, respectively. Since the Ni functions still be expressed by means of a stress-strain rela-
are defined in the coordinate system (, ) of the ref- tionship (matrix K L in Eq. (5)). As already men-
erence element, the derivation matrix B in (3) is ob- tioned, the approach proposed by F RANK et al.
tained introducing the Jacobian matrix J of the geo- [1982] introduced in CESAR-LCPC code (called
metrical transformation between the actual element "oriented criterion") uses linear elasticity and the
and the reference rectangular element of length L Mohr-Coulomb plasticity criterion with either an as-
and thickness t (see Fig. 1). After making substitu- sociated or non-associated flow rule in order to re-
tions in relation (3), element strains will depend up- produce contact. However, based on Equation (5) it
on nodal displacements, Ni interpolation functions is clear that in the absence of any hypothesis on the
and element dimensions (t and L). If the element interface element dimension (L/t ratio) and its con-

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 47

stitutive law, the element may be considered as ca-

pable of reproducing the behaviour of a medium stant slope , being D the dilatancy). The
exhibiting a significant number of discontinuities
threshold from contraction to dilatancy defines the
equally oriented. The interface zone is therefore
phase transformation state, at this state = c and
portrayed more like a remoulded zone. The basis of
the oriented criterion formulation has been reem- , hence D = 0. Finally, at large relative tan-
ployed here and modified in order to adapt to inter- gential displacements, = r (r stands for residual)
face problems involving an elastoplastic behaviour. and no volume changes are observed (D = 0). The
latter situation reflects the well established findings
on the critical state of soils. Note that in interface
Basics for the constitutive modelling of the soil-structure tests at constant volume or constant normal stiff-
interface ness, contraction is manifested by a decrease in nor-
mal stress whereas dilatancy at large relative tan-
The striking analogy existing between the
gential displacements will be reflected by an in-
rough interface behaviour and that of granular soils
crease in normal stress (constrained dilatancy).
tested in triaxial conditions has been for many au-
thors the tenet of an accurate constitutive modelling It is worth noting that two different states of the
of interfaces (e.g. BOULON and NOVA, 1990). Typical interface are characterised by the condition D = 0
results of a constant normal stress interface test con- (i.e. zero volume changes), namely phase transfor-
sidering the interaction of a granular medium and mation and critical or, more properly, ultimate
a rough interface are presented in Figure 2. Results state. Recent experimental results have shown that
both phase transformation state and ultimate state
are expressed in terms of mobilized friction for interfaces are encountered at the same mobi-
and relative normal displacement [un] at the inter- lised friction (in other words c = r), suggesting
face vs. relative tangential displacement [ut] (being that dilatancy D is a unique function of the stress
the shear stress mobilized at the interface and n the level (e.g. LINGS and DIETZ, 2005). Other experi-
normal stress at the interface). mental findings seem to corroborate the idea that
As it can be observed mobilized friction in- transformation state and critical state are distin-
creases until reaching the failure condition p. This guishable (e.g. DE GENNARO, 1999; DE GENNARO and
occurs progressively in loose interfaces and the mo- FRANK, 2002; GHIONNA and MORTARA, 2002). The
bilized friction at failure coincides with the residual uniqueness of the condition D = 0 as a function of
value r. Conversely, peak shear strength is usually , which means a solely stress-dependent dilatancy
observed at high densities or at low normal stress for interfaces, is then still debated, although it is
levels, in both cases associated with interface dila- generally recognized that dilatancy in sand tested in
tancy. In this latter situation strain softening on the triaxial conditions show not only stress-dependency
mobilised friction curve is also observed. During in- but also state-dependency (e.g. LI and D AFALIAS,
terface tests at constant normal stress, the relative 2000). Based on the results of L INGS and D IETZ
normal displacement initially decreases (contrac- [2005] and DIETZ and LINGS [2006] it might be ar-
tion phase) and then increases at rather constant gued that apparatuses compliances could be at the
rate corresponding to the dilatancy phase (i.e. con- origin of these discrepancies.

Fig. 2 Characteristic behaviour of a granular soil-structure interface.

Fig. 2 Comportamento tipico dellinterfaccia terreno granulare-struttura.

APRILE - GIUGNO 2008

48 DE GENNARO - FRANK - SAID

The MEPI-2D model

Moving from these preliminary considerations (9)

about interface behaviour, the elastoplastic inter-
face model MEPI-2D [DE GENNARO and FRANK, 2002]
is now briefly presented and its capability in repro- Again, p is the friction coefficient at failure,
ducing experimental data of interface tests is veri- 0=tan0 is the initial friction coefficient delimiting
fied. This model will be used to describe the contact the elastic domain, t is the interface layer thickness,
between the piles and the soils during the FEM cal- a (dimensionless) is a model parameter which con-
culations presented in this paper. trols the shape of the hardening law (9), ni is the in-
itial normal stress acting on the interface layer and
Let d = {dn, d}T the vector of stress incre-
p0 = 1 kPa is a reference pressure.
ments and d{ue} = {d[uen], d[uet]}T the vector of rel- Equations (8) and (9) define a mobilized friction
ative elastic displacements increments. The elastic hardening mechanism of the yield surface F that en-
behaviour of the interface is given by the following larges the initial elastic domain until reaching the
linear relation: peak (failure) condition given by the Mohr-Cou-
lomb failure criterion, as defined by the friction an-
(6) gle p=tan-1(p) (see Fig. 2b). The softening behav-
iour, often observed during interface tests on dense
sands (Fig. 2a), may be introduced by replacing the
Matrix Ke is composed of the elastic interface relation in (9) and adopting the following softening
stiffness in both the normal (Kn) and tangential (Kt) law:
directions. Elastic behaviour does not take into ac-
count any coupling between the normal and tangen-
tial phenomena occurring at the interface. In order (10)
to express the evolution of the stiffness vs. the nor-
mal stress and density state of the material, the fol-
lowing definition is provided: where is the hardening parameter value at fail-
ure, which enables defining the position of the cur-
(7) rent peak shear strength, as an alternative to more
rigorous approaches (e.g. MUIR WOOD et al., 1994;
where kn, kt and N are three model parameters and DE GENNARO et al., 2002). Parameter f is the current
ni is the initial normal stress. Both Kn and Kt in- friction coefficient at failure (f < p), r is the resid-
crease when the initial normal stress increases fol- ual friction coefficient and b is a model parameter
lowing a power law. that controls the softening rate. An increase of b val-
During shear, the coefficient rises gradually ues accelerates softening and stabilization at the re-
and at failure reaches the maximum value p (peak sidual friction value r.
value for dense sands, threshold value r for loose The closure condition for MEPI-2D model for-
sands; see Fig. 2a). This is reflected in the plane mulation is the definition of the plastic flow ena-
(n, ) by the rotation of the line = n (Fig. 2b), bling to relate plastic relative normal and tangential
assuming that the linear failure envelope of Mohr- displacements with the corresponding stress com-
Coulomb keeps unchanged. Experimental results ponents n and . Dilatancy ratio has been
also indicate that for a rough surface the failure con-
dition is still well approximated by the Mohr-Cou- used to define the plastic flow rule of the model. As
lomb criterion. The yield function of MEPI-2D discussed in the previous section there are still not
model is a generalization of the Mohr-Coulomb fail- converging opinions about the type of flow rule as-
ure criterion neglecting cohesion and is written: sociated to the stress-dilatancy interface behaviour.
As already discussed earlier it is still a matter of dis-
(8) cussion if the flow rule for interface is simple (i.e. di-
latancy is a unique function of ) or double. One
main argument valid for sands (e.g. LI and DAFALIAS,
where is the hardening function and the rel- 2000) potentially supports double flow rule for the
ative tangential plastic displacement [upt] the corre- interface: assuming a simple flow rule the stress ra-
sponding hardening variable. Based on experimen- tio c at phase transformation state would coincide
tal evidences the hardening function is assumed hy- with the value of r at ultimate state. Being the latter
perbolical and is given by the following relation: intrinsic (although evolving with the mean effective

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 49

stress), c would be intrinsic too. This is not verified is given by:

in sands where c is a variable quantity of relative
density.
Contraction at the interface corresponds to the (12)
condition: c (D= / >0), whereas dila-
tancy is generated if: c(D= / <0). The A typical evolution of function in (12)
phase transformation state is obtained when = c,
is showed in Figure 3. Parameter c defines the rate
for this condition D = 0. The latter situation also oc-
of dilatancy stabilization at the interface. When c in-
curs at large tangential displacements at the ulti-
creases dilatancy stabilization at the zero value (i.e.
mate state. The phenomenological interpretation of
ultimate state) is reached faster. Note that although
this trend given in DE GENNARO [1999] and DE GEN-
NARO and FRANK [2002] was that yield of dense inter- function is not directly correlated with
face layers due to shearing causes plastic dilation, any state parameter of the sand, plastic strain de-
resulting in an increase of the void ratio (shrinkage pendency in (11) allows to account for the evolution
of the dilation region). On the other hand, shearing of the compaction domain and hence the phenom-
on loose interface layers causes an opposite effect, enon of dilatancy stabilization at larger relative dis-
leading to an overall compaction of the interface placements. NOVA [1977] and afterwards CUBRINOSKI
and a reduction of the void ratio (shrinkage of the and HISHIHARA [1998] proposed a similar approach
compaction region). Whatever compaction or dila- for modelling phase transition in sands. Note that if
tancy, the void ratio will tend towards the final value in Equation (11) the single flow rule of the original
defined at ultimate state and the initial value of c Cam clay model is recovered [SCHOFIELD and WROTH,
will then tend during shear towards its ultimate 1968].
value given by the residual friction coefficient r. MEPI-2D model has 11 parameters: kn, kt, N, o,
Void ratio, like in critical state theory [SCHOFIELD p, c0, r, a, b, c, that can be reduced to 8 if the in-
and WROTH , 1968], may serve as an appropriate fluence of the initial normal stress on the elastic
state variable for representing volumetric behaviour stiffness and softening behaviour are not consid-
of the interface (e.g. DE GENNARO, 1999; DE GENNARO ered. Parameters t, ni and p0 are defined from ini-
and FRANK 2003). However one can simplify the for- tial conditions and the physical properties of the
mulation, while maintaining the same basic philos- sand (t 10 D50). Model reliability has been verified
ophy, assuming that dilatancy evolution law is a by simulating various soil-structure shear tests fol-
function of the hardening parameter and adopting lowing various loading paths: constant normal
the following stress dilatancy relationship: stress, constant volume and constant normal stiff-
ness tests (e.g. DE GENNARO and FRANK 2002). Results
of simulations, not presented here for the sake of
(11)
brevity, are available in DE GENNARO [1999], DE GEN-
NARO and FRANK [2002] and SAID [2006].

where c0 is the stress ratio at phase transformation

state, is the mobilised friction coefficient and
3. Presentation of axially loaded model piles in
sands
In the following section we describe the two
cases of axially loaded models piles in sands ana-
lysed using the FEM code CESAR-LCPC. The first
example refers to the loading tests of a model pile
whished in place in a calibration chamber [DE GEN-
NARO, 1999]. The second one concerns the pile test
conducted in Dunkerque (France) using the Impe-
rial College model pile [CHOW 1997].

Type 1 pile test: model pile in calibration chamber

Fig. 3 Typical evolution of the dilatancy function of ME- The calibration chamber test set-up enables re-
PI-2D model. constituting a cylindrically-shaped soil mass (diam-
Fig. 3 Evoluzione tipica della dilatanza del modello MEPI- eter: 524 mm, height: 680 mm). DE GENNARO et al.
2D.
[1999] used Fontainebleau sand, the reconstitution

APRILE - GIUGNO 2008

50 DE GENNARO - FRANK - SAID

Fig. 4 Geometry of calibration chamber and the model Fig. 5 The IC pile installed in Dunquerke [after CHOW,
pile prior to axial loading. 1997].
Fig. 4 Geometria della camera di calibrazione e del palo Fig. 5 Il palo IC installato a Dunkerque [CHOW, 1997].
modello della prova di carico.

of the soil mass was performed by means of sand rial College [BOND, 1989] were used. Results from
pluviation at a given density (Density Index one specific test (pile DK1) among those of the site
ID) 0.47). The model pile was preliminarily in- of Dunkerque will be considered.
stalled, prior to sand pluviation. The soil mass was The site profile at Dunkerque consists of a top
thus reconstituted around the model pile that can layer of hydraulic fill, from ground level to 4 m
be considered as wished in place. This procedure depth, overlying the Flandrian sand formation. This
served to overcome the effects caused by installation sand deposit is almost homogeneous; Flandrian
(driving or jacking). The configuration of the cali- sand is a siliceous poorly graded sand having an av-
bration chamber once the pile is installed and be- erage diameter D50 of 0.25 mm. At 7.6 m depth the
fore loading is shown in Figure 4. Radial and verti- sand horizon is interrupted by an organic layer 0.6 m
cal total stresses are applied simultaneously and in- thick. The water table is found at a depth of 4 m.
dependently on the sand sample, allowing for the Densities were assessed by means of CPT. Tests
simulation of isotropic or anisotropic confinement indicated dense sand (ID = 0.75), with a net reduc-
conditions of the soil mass. tion in density in the organic layer. The hydraulic
The instrumented pile has a diameter B=20mm fill has ID = 1, the corresponding bulk densities are
and a total height of 830mm. The pile tip is close = 17.1 kN/m3 above the water table for the hydrau-
ended. The set-up allows conducting measurements
lic fill, and =19.9 kN/m 3 below the water table,
of total load at the top, in addition to decoupled
within the Flandrian sand formation.
measurements of tip load and lateral friction
The steel pile used during the test has a diame-
(Fig. 4). The results from one test performed on a
ter 2R = 0.102 m, and a length Lp = 7.4 m. The pile
model pile with slenderness ratio B/D=25 (where D
is the pile penetration depth), were selected for our tip is close ended with a 60 conical shape. The in-
purposes. The test was held using a medium-loose strumentation is concentrated in four clusters (lead-
dry Fontainebleau sand sample with density index ing, following, trailing, lagging; Fig. 5) spaced
ID 0.46 (e 0.75) submitted to an isotropic state of about 1 meter along the shaft, the first cluster being
0.2 m distant from the pile tip. Each cluster con-
stress (vo = ho = 100 kPa).
tains: (i) an axial load cell measuring the axial load
transmitted through the pile and the average shear
Type 2 pile test: model pile test in Dunkerque site (France) stress between two clusters (fs), and (ii) a surface
stress transducer, allowing the measurement of the
The second type of pile test considered is that total radial stress (r) and the local shear stress (rz).
conducted by CHOW [1997] on displacement model Further details on ICP instrumentation can be
piles in sand. Instrumented piles designed at Impe- found in BOND [1989] and CHOW [1997].

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 51

The installation of the pile was performed using

a hydraulic jack, with consecutives jacking strokes al-
lowing for a constant pile head displacement rate of
600 mm/min. At the end of the installation an equali-
sation period of about 15 hours was considered,
based on the time necessary to stabilise the values of
the measured radial effective stress. An overall in-
crease of the radial effective stress rc was found in
the four clusters at the end of the equalisation pe-
riod. The following empirical formula [CHOW, 1997;
JARDINE et al., 1998], gives the final value after equal-
isation of radial stress along the pile shaft :

(13)

where h is the vertical distance from the pile tip, R

the pile radius, vo the effective vertical stress, pa the
atmospheric pressure, rc the radial effective stress
at the end of installation and qp the net cone resist-
ance from CPT. The profile of r obtained in Dun-
kerque site is shown in Figure 6a.
It is observed that the use of Equation (13) is for-
mally equivalent to the application of a lateral (ra-
dial) pressure r = K v on the pile shaft with K val-
ues defined as a function of net cone resistance,
state of stress and geometry. In particular, relation
(13) allows to define explicitly the condition of lat-
eral pressure following installation via a direct cor-
relation with the CPT profile as given by the evolu-
tion of the net cone resistance qp.
The four clusters also allowed measurements of
the residual shear stress on the pile shaft and the
residual tip load after installation. The distribution
of the residual shear stress res available from these
clusters (Fig.6b) was found in good agreement with
the theoretical distribution proposed by ALAWNEH Fig. 6 Distribution of the radial (a) and shear stress (b)
et MALKAWI [2000] for a long flexible pile. The re- on the shaft of the IC pile after installation and equaliza-
sidual tip load Qp-res is obtained from the measure- tion [after CHOW, 1997].
ment of the axial load cell placed on the pile tip Fig. 6 Distribuzione della tensione radiale (a) ed della tensione
di taglio (b) sulla superficie laterale del palo IC alla fine
(Leading cluster) after 15 hours of equalization, it
dellinstallazione e del periodo di equalizzazione [CHOW, 1997].
was found Qp(res) 51 kN. Both res and qp-res = Qp(res)/
Sp , where Sp is the model pile tip surface, were ap-
plied numerically together with the distribution of
r given in (13). 4. Analysis of Type 1 pile test: model pile in
Compression loads on the model pile were ap- calibration chamber
plied following a variant of the LCPC procedure
The problem considered and the finite element
(e.g. BUSTAMANTE, 1982). Loads were applied by suc-
mesh is shown in Figure 7. Mesh consists of 230 ax-
cessive increments of 5% to 10% of the expected
isymmetric isoparametric 8-node solid elements
maximum load, and maintained for 10 to 20 min- (621 nodes), including 20 pile elements and 10 in-
utes. Failure was reached when the rate of pile terface thin-layer elements with an aspect ratio L/t =
movement increased significantly with time. The 25. This value was found appropriate in reproduc-
failure axial load, was 309 kN, which corresponds to ing the interface behaviour and avoiding numerical
a pile head displacement of 4.2 mm.

APRILE - GIUGNO 2008

52 DE GENNARO - FRANK - SAID

Tab. I Constitutive parameters used for Type 1 test simulations: calibration chamber.
Tab. I Parametri constitutivi utilizzati per le simulazioni della prova Type 1: camera di calibrazione.

Interface Kn (kPam-1) Kt (kPam-1) f=r co o a c ni (kPa) t (mm)

MEPI-2D 1.08x106 4.7x105 0.7 0.46 0.27 0.008 40 100 2

Interface Oriented E (MPa) n f y

Criterion 200 0.3 27 10

Sand E (MPa) n
Elastic 34 0.3
Sand E (MPa) n f y c
Mohr-Coulomb 34 0.3 36,5 12 0

Sand E (MPa) n Bo l D M m m pco (kPa)

Novas model 34 0.3 0.0016 0,0021 0,3 1,16 1,76 0,08 100

Pile E (GPa) n
Elastic 210 0.28

The model pile has been considered wished in

place and installation effects neglected. The valid-
ity of this hypothesis on the radial and shear
stresses on the pile shaft was verified by DE GEN-
NARO et al. [2004]. The authors found that radial
and shear stress changes due to sand pluviation
(progressive increase of the weight of the sand
mass) in the interface were negligible compared
to the level of isotropic confining stress applied
during initial consolidation (stress ratio Ko = vo /
ho = 1, vo = ho = 100 kPa). In other words, at
the end of the isotropic consolidation (100 kPa),
the effect of the gravity field was overridden and
simulations with or without gravity led to similar
results. These simulations have been done assum-
ing frictional contact (oriented Mohr-Coulomb
failure criterion) or adhesion and results were in-
dependent of the choice of the type of soil-pile
contact.

Constitutive models for pile, sand and interface

Fig. 7 Mesh used for the analysis of the Type 1 pile test
in calibration chamber.
Analyses of the consolidation phase and pile test
Fig. 7 Discretizzazione agli elementi finiti utilizzata per
lanalisi della prova di carico Type 1 in camera di have been conducted assuming elastic response of
calibrazione. the pile. Elasticity, perfect plasticity and elastoplas-
tic behaviour was assumed for Fontainebleau sand.
A standard oriented Mohr-Coulomb failure crite-
rion and non associated plasticity as well as MEPI-
ill-conditioning [PANDE and SHARMA, 1979; HOHBERG
2D model have been considered for the interface
and SCHWEIGER, 1992, DAY and POTTS, 1994; DE GEN-
layer. The constitutive parameters of the models are
NARO, 1999; De GENNARO and FRANK, 2005].
given in Table I.
Calculations have been performed following the Perfect plasticity for Fontainebleau sand was as-
two experimental phases: (1) the consolidation sumed considering the Mohr-Coulomb failure crite-
phase, during which the sand deposit with the em- rion and non-associated flow rule. Elastic-plastic be-
bedded model pile is submitted to an initial iso- haviour of the soil mass was introduced modelling
tropic stress (vo = ro = 100 kPa); (2) the pile test, the sand behaviour with the strain hardening plastic
during which a displacement wto is imposed at the model proposed by NOVA [1982], an extension of
head of the model. the Cam clay model for sands, available in the FEM

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 53

code CESAR-LCPC. The details of Nova model for- For the soil-pile interface simulation thin layer
mulation are here omitted for the sake of brevity. elements were used with a Mohr-Coulomb failure
Note that in CESAR-LCPC code linear elasticity is criterion oriented in the interface direction
considered instead of non linear elastic behaviour as [FRANK et al. 1982] and the elastoplastic MEPI-2D
given in the original formulation by Nova. In addi- model [DE GENNARO and FRANK, 2002] without con-
tion to the quoted reference the description of the sidering softening behaviour. The set of constitutive
implemented model, including procedures for pa- parameters selected for the interface in the numer-
rameter determination are available in DE GENNARO ical analysis of the calibration chamber test are also
[1999]. All the parameters have been obtained from given in Table I. They were determined from the re-
the results of a series of CD triaxial compression sults of interface direct shear tests between Fon-
tests [DUPLA and CANOU, 1994; DE GENNARO, 1999], tainebleau sand and a rough metal plate [DE GEN-
some of these tests were performed on specimens NARO , 1999; D E G ENNARO and F RANK , 2002]. The

having the same density of the sand mass reconsti- roughness of the metal plate was the same as that of
tuted in the calibration chamber. In Figure 8 Novas the external surface in the model pile
model predictions are satisfactorily compared with (Rmax = 0.2 mm and Rn = Rmax/D50 = 1).
experimental results.
The value of the dilatancy angle of the sand Numerical results: consolidation phase
( =12) might look somewhat high. In fact, a dif-
ference of about 30 is usually assumed between the During the consolidation phase the pile head
friction angle and the dilatancy angle. The latter is was free to move downward or upward. As already
a good approximation as long as plane strain condi- mentioned the effect of gravity has been neglected
tions are considered (i.e. biaxial tests), but it often in this phase. The results of the first step of the nu-
fails in the case of triaxial conditions (see for in- merical simulations are presented in Figure 9 in
stance: SCHANZ and VERMEER, 1996). It is worth men- terms of displacement field and evolution of the co-
tioning that the proposed value is in good agree- efficient of lateral pressure K = r / v.
ment with the experimental findings of VAID and SA- At the end of the consolidation phase the distri-
SITHARAN [1992] on Erksak sand, a quartz sand hav- bution of displacements within the soil mass is
ing physical characteristics very similar to Fon- rather heterogeneous. In the example shown in
tainebleau sand. Note finally that both values of Figure 9a Nova model was considered to simulate
friction and dilatancy angles of Fontainebleau sand the soil mass behaviour. Assuming elastic behaviour
are consistent with the plastic parameters corre- or perfect plasticity for the sand resulted in similar
sponding to the failure condition in Nova model. displacement patterns and only minor differences

Fig. 8 Simulation of drained triaxial tests on Fontainebleau sand using Nova model (continuous line): (a) deviator stress
vs axial strain and (b) volumetric strain vs axial strain.
Fig. 8 Simulazioni delle prove triassiali drenate su sabbia di Fontainebleau con il modello di Nova (curva continua): (a) deviatore
delle tensioni in funzione della deformazione assiale e (b) deforrmazione di volume in funzione della deformazione assiale.

APRILE - GIUGNO 2008

54 DE GENNARO - FRANK - SAID

Fig. 10 Plastic points and plastic zones in the interface

and the soil mass after consolidation.
Fig. 10 Punti plastici e zone di plasticizzazione allinterfaccia
e nel terreno alla fine della consolidazione.

Fig. 9b). The vertical displacements v decrease

moving from the bottom to the top of the sand de-
posit; it is worth noting that at the end of the consol-
idation phase the model pile has been lifted by
about 0.16 mm assuming a sand obeying to Nova
model. This causes the mobilisation of an initial fric-
tion at the interface between the pile and the sand
and the mobilisation of a residual shear stress of
about 16 kPa along the pile shaft in the zone of the
instrumented sleeve. The non-uniformity of the
stress field within the sand deposit is well repre-
sented by the evolution of the coefficient of lateral
pressure K (Fig. 9b). Results are presented assuming
the soil mass elastoplastic (Nova model). Coefficient
K is given for the three specific sections of the cali-
bration chamber (A, B, C). On the pile shaft the
value of K is greater than one, as expected. It is
worth noting that at the end of the consolidation
step plastic zones appear at the extremities of the
Fig. 9 Predicted distributions of the radial and vertical dis-
placements in the sand mass after installation of the model pile interface when the soil is elastic (Fig. 10a).
pile (a) and evolution of the coefficient of lateral pressure K These zones extend in the vicinity of the soil mass
along three sections of the calibration chamber (b). assuming Mohr-Coulomb failure criterion for the
Fig. 9 Distribuzione ottenuta via analisi numerica agli sand (Fig. 10b) and spread within the soil mass up
elementi finiti degli spostamenti radiali e verticali nel terreno to the model boundaries assuming elastoplastic be-
alla fine dellinstallazione del palo modello (a), evoluzione haviour for Fontainebleau sand using Nova model
radiale del coefficiente di spinta laterale K in tre sezioni della
(Fig. 10c).
camera di calibrazione (b).

Numerical results: pile loading test

in terms of vertical displacements values. The radial A series of numerical simulations have been per-
displacement u decreases progressively moving formed in order to assess the influence of various
from the external boundary to the surface of the factors on the overall behaviour of the model pile
model pile. The pattern shows a pronounced sym- during tests in calibration chamber. We will focus or
metry with respect to the middle horizontal plane of attention on two specific points: the influence of the
the calibration chamber (sec. B at z = 400 mm, constitutive laws of sand and interface.

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 55

Fig. 12 Influence of the interface constitutive model on

the mobilization of the average shear and normal stress.
Fig. 12 Influenza del modello constitutivo dellinterfaccia
Fig. 11 (a) Comparisons between the average value of
sulla mobilizzazione della tensione media di taglio e della
the shear resistance measured and calculated and (b) evo-
tensione normale.
lution of the average normal stress.
Fig. 11 (a) Confronto fra il valore medio della sforzo di taglio
misurato e calcolato via analisi numerica agli elementi finiti e (b)
evoluzione della tensione normale media.
observed in Figure 11a calculations and measure-
ments are in good agreement. The evolution of the
The first series of simulations has been per- average normal stress for the elastic case (E) in
formed in order to assess the effect of the sand be- Figure 11b shows a stabilisation at large tangential
haviour on the numerical results (Fig. 11). Assuming displacement (ultimate state) and leads to a final pla-
a linear elastic sand mass the non-linearity of the teau of mobilised friction. When the soil mass is sim-
pile-soil system is concentrated at the interface be- ulated assuming perfect plasticity and Mohr-Cou-
tween the two media. The interface behaviour is de- lomb failure criterion (MC), shearing of the soil mass
scribed using MEPI-2D model. Comparisons are close to the pile shaft induces high dilatancy of the
made between the average value of the shear resist- sand and prevents the stabilisation of the normal
ance measured with the instrumented sleeve (Fig. 4) stress at the interface. As a consequence, the average
and the corresponding average value obtained with shear stress keeps increasing at the ultimate state
the finite element analysis considering the nodes (Fig. 11a). This phenomenon of coupling between
having the same location of the sleeve. As it can be the behaviour of the interface and the surrounding

APRILE - GIUGNO 2008

56 DE GENNARO - FRANK - SAID

sand mass has been seldom analysed during FEM fect of the model pile during sand pluviation reflect-
analyses of pile tests. In contrast to the high con- ing a drop in density over the zone beneath the tip.
strained dilatancy imposed by perfectly plastic sand, Note that for all calculations the total shaft resistance
a more realistic response during loading is obtained mobilization curve Qls obtained integrating the stress
assuming elastoplastic behaviour for the sand using distribution along the lateral surface of the model
Nova model (N). In this case the stabilization of dila- pile lies close to the experimental total load curve Qt.
tancy imposed by the interface model MEPI-2D is As a consequence, the finite element computation
less affected by the surrounding soil behaviour, al- prediction of total shaft friction Qls is in close agree-
though some fluctuations of the normal stress are ment with the measured values whereas the load at
still observed (Fig. 11b). As already observed plastic pile tip Qps has been overestimated.
zones appears during the consolidation phase using
Nova model (Fig. 10c). These plastic zones are not
associated with failure, like in the case of perfect
5. Analysis of Type 2 pile test: model pile test in
plasticity and Mohr-Coulomb failure criterion, but
only with plastic hardening. This corresponds to a
Dunkerque site (france)
local volumetric compaction of the sand. Conse-
The full scale test on an instrumented model
quently the value of the normal stress on the pile
pile jacked in sand and subjected to axial compres-
shaft at the end of consolidation using Nova model
sion load [CHOW, 1997] is now analysed. The prob-
is also slightly lower than that obtained assuming
lem has been presented in section 2. The finite ele-
elastic and elastic perfectly plastic behaviour. It is
ment mesh used for the simulations of the pile test
worth noting that since on the pile shaft at the end
is shown in Figure 14. Axisymmetric conditions are
of consolidation no plastic zones are generated using
considered (Z being the axis of symmetry, coincid-
Mohr-Coulomb failure criterion for the sand mass
ing with the pile axis). The limits of the domain in-
the value of n is the same obtained assuming elastic
vestigated are fixed at 30 m in the vertical direction
behaviour of the sand.
(i.e. more than 3 times the pile length, Lp) and 3
The influence of the interface model is consid-
ered comparing the responses of the two dilatant in- times the pile length in the radial (horizontal) direc-
terface layers when the sand mass behaves as an tion. The horizontal (radial) displacements are set
elastic medium (Fig. 12). This is obtained using al- equal to zero on the vertical boundary and the ver-
ternatively the Mohr-Coulomb oriented failure cri- tical displacements are not allowed on the bottom
terion and MEPI-2D to model the interface behav- boundary. The mesh consists of 1402 axisymmetric
iour. The essential feature of the behaviour of the isoparametric 8-node solid elements, including 100
interface is related to the restrained dilatancy devel- pile elements and 50 interface elements. The inter-
oping during shear, allowing for an increase of the face layer is t = 3 mm in width (12 times D50 of the
shaft resistance. Results in Figure 12 show clearly sand) and its behaviour is simulated by means of
that an overestimation of the dilatancy of the inter- eight-node thin-layer elements. The aspect ratio L/t
face, as in the case of the Mohr-Coulomb oriented of the interface elements (L/t=48) was chosen again
failure criterion leads to an unrealistic prediction of in order to better reproduce the interface behaviour
the evolution of the normal stress at the interface and avoid numerical ill-conditioning [PANDE and
and an overestimation of the mobilised shear on the SHARMA, 1979; HOHBERG and SCHWEIGER, 1992; DE
pile shaft. The essential role played by an adequate GENNARO, 1999].
constitutive modelling from initial yield to the ulti- During numerical analyses conducted on the
mate state condition is again emphasised. pile test using the calibration chamber the model
Finally, the comparison with total experimental pile has been considered wished in place as a con-
load values at the pile head (Qt) and tip (Qp) is shown sequence of the reconstitution procedure of the soil
in Figure 13. Results refer to calculation performed mass. This situation is quite unusual in real practice
using MEPI-2D for the interface and assuming elas- and is no longer valid for the pile test of Dunkerque
tic, elastic perfectly plastic and elastoplastic behav- presented in this section. In the case of piles or piles
iour of the sand mass (Figs. 13a, b, c). Although the group installation effects have a major impact on
problem of pile failure under the tip could be better the overall behaviour after construction. This is an
simulated assuming a large strain formulation, simu- important issue which is still seldom considered in
lation results in terms of total load Qps including plas- finite element analyses.
tic behaviour in the soil mass indicate qualitatively a Indeed the installation of a single pile remoulds
good agreement with the experimental data (curve the soil and thus plays an important role on its be-
Qp). The deviation between measured and computed haviour. Remoulding occurs in the immediate vicin-
values is probably due to the pile installation proce- ity of the pile shaft (influence on the interface zone)
dure, at the origin of the underestimation of the pile and within the soil mass. In both cases it cannot be
tip resistance Qp. This is probably due to a screen ef- assumed anymore that the soil characteristics are in-

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 57

Fig. 13 Comparisons between measured and predicted total loads during Type 1 pile test in calibration chamber: (a)
elastic soil mass, (b) elastic perfectly plastic soil mass (Mohr-Coulomb) and (c) elastoplastic soil mass (Nova model).
Fig. 13 Confronto fra i valori dei carichi misurati durante la prova di carico Type 1 in camera di calibrazione e calcolati via analisi
numerica agli elementi finiti: (a) terreno elastico, (b) terreno elastico perfettamente plastico (Mohr-Coulomb) e (c) terreno elastoplastico
(modello di Nova).

Fig. 14 Mesh adopted for the numerical analysis of type

2 pile test in Dunkerque.
Fig. 14 Discretizzazione agli elementi finiti utilizzata per Fig. 15 Measured and predicted distribution of the ra-
lanalisi della prova di carico Type 2a Dunkerque. dial stress on the shaft of the IC pile assuming elastic per-
fectly plastic soil mass (Mohr-Coulomb). Experimental
data after CHOW [1997].
tact, as often assumed in numerical modelling of soil Fig. 15 Confronto fra i valori della tensioni radiali misurati
masses. Note that this remoulding can sometimes sulla superficie laterale del palo IC e calcolati via analisi agli
be beneficial. Driven piles in loose sands, for in- elementi finiti assumendo il comportameto del terreno elastico
stance, lead to a densification of the neighbouring perfettamente plastico (Mohr-Coulomb). Dati sperimentali di
soil and an increase of the bearing capacity of the CHOW [1997].
foundation.
The various installation techniques for piles can
be cut down to two broad methods, as far as numeri- differently in terms of numerical modelling. Moreo-
cal modelling is concerned. These methods are ei- ver, since the bearing capacity depends on both shaft
ther driving a prefabricated pile, or by boring a hole and tip resistances, modifications of both due to the
in the ground and by pouring concrete inside it. installation effects need to be carefully assessed prior
Each of these two methods should clearly be treated to any finite element analysis of the mechanical be-

APRILE - GIUGNO 2008

58 DE GENNARO - FRANK - SAID

Tab. II Constitutive parameters used for Type 2 test simulations: Dunkerque pile test.
Tab. II Parametri constitutivi utilizzati per le simulazioni della prova Type 2: prova di carico su palo a Dunkerque.

Interface Kn (kPam-1) Kt (kPam-1) f=r co o a c ni (kPa) t (mm)

MEPI-2D
(during loading) 5x105 2.5x105 0.51 0.25 0.017 0.0008 0.065 100 3

Interface Oriented E (MPa) n f y c

Criterion
(during initialisation) 200 0.3 27 10 0

E (MPa) n f y c
Sand
Mohr-Coulomb
234 0.3 37 12 0

E (GPa) n c
Pile
Elastic
195 0.28

haviour under service loading and/or at failure. Mod- During the initialisation of the state of stress in
elling of driven piles needs either the use of sophisti- soil (geostatic stresses and installation effect), the in-
cated numerical techniques (e.g. large strains analy- terface layer is simulated assuming Mohr-Coulomb
sis, conform mapping, meshless techniques, others) oriented failure criterion and non-associated per-
or the adaptation of available experimental data and/ fect plasticity [FRANK et al. 1982]. Following FRANK et
or correlations enabling to initialize the state of stress al. [1982], elastic parameters of the interface (Ei, i)
within the soils prior to further loading. The initiali- are the same of the sand mass, whereas plastic pa-
zation of the normal stress and the residual shear
rameters (ci, i and i) have been obtained from the
stress acting on the pile shaft and the estimation of
data by KUWANO [1996], quoted by CHOW [1997].
the residual tip load after driving are likely to be the
crucial issues associated with this phase. Modelling of During the simulation of the compression test
bored piles looks easier, though some features still MEPI-2D model is considered at the interface. All
need serious investigation (unloading of soils, for in- the constitutive parameters are summarised in Ta-
stance). Both methods may require advanced consti- ble II.
tutive equations for the ground which are not com-
monly used in standard practice.
Numerical results: installation effects
The calculations presented herein aim at simu-
lating the mechanical behaviour of the pile consid- To reproduce the initial conditions in terms of
ering explicitly (i) the soil and the interface geotech-
effective stresses in the soil mass before loading,
nical characterisation, and (ii) the effect of installa-
sand weights above and below the water-table (at
tion in terms of normal stress and shear stress
4 m depth) were first activated and Ko condition ap-
changes on the interface layer and residual tip load.
This is done following the standard FEM approach plied (geostatic state). Afterwards, during the sec-
and considering available experimental data and ond step calculation, the residual radial and shear
known correlations enabling to initialize the state of stresses distributions along the pile shaft, Equation
stress within the soil mass prior to further loading. (13) and Figure 6, and the residual tip load were im-
posed numerically as an external loading on the soil
mass. The aim of this second calculation step was to
Constitutive models for pile, sand and interface recover the residual effective stress values measured
at the end of the installation in order to account for
The mechanical parameters have been obtained
the overall effect of the jacking procedure on the
from data by K UWANO [1996], quoted by C HOW
model pile and within the soil mass.
[1997], derived from triaxial tests and interface di-
rect shear tests on Flandrian sand from Dunkerque. The distribution of radial stress against depth
In the lack of detailed tests results to be analysed obtained during the numerical simulations is com-
only the elastic parameters (E, ) and the plastic pa- pared with the experimental values in Figure 15. At
rameters (c, , ) given by the author were consid- the end of this calculation it was verified that the to-
ered and during simulations Flandrian sand has tal tip resistance was equilibrated by the sum of the
been assumed elastic perfectly plastic, obeying the average total shear stress applied on the pile shaft
Mohr-Coulomb failure criterion. and the weight of the model pile.

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 59

Numerical results: pile loading test terface, given by the initial slope of the shear curves
in Figure 16a, are slightly overestimated and com-
Moving from the initial conditions obtained pare less satisfactorily with the experimental data.
during the previous calculation step, pile test has Since the constitutive parameters of the interface
been simulated imposing a total pile head displace- and the sand models are defined from the results of
ment wto=8 mm. This has been achieved applying laboratory tests, the observed scatter in Figure 16a
incrementally 1.25% of the final vertical displace- may reflect possible scaling effects related to the
ment on the pile head in 80 increments. passage from the laboratory conditions to the real
A comparison between the mobilization of the in-situ conditions. Possibly Poissons ratio effect
shear stress along the pile shaft predicted by the nu- (s=0.3 in the calculations, see Tab. II) could be also
merical model and obtained during the pile test is at the origin of an excessive increase of the radial
shown in Figure 16. Measurements refer to the stress in numerical calculations. Nevertheless, inter-
shear stress rz measured locally, in each cluster lo- face tests revealed strong dilatant behaviour [CHOW,
cated on the ICP shaft. The shape of the simulated 1997], in agreement with the high density of the
shear stress curves appears in good agreement with sand deposit (ID=0.75). These data are confirmed
the measured ones. In particular the failure condi- by the results of pile load test giving an increase of
tion is well reproduced in FEM calculations. The the measured radial stress r of about 60 kPa. In-
predicted initial shear stiffnesses at the sand-pile in- creasing radial stress values of the same order have

Fig. 16 Comparison between the local values of the shear resistance measured and calculated (a) and predicted evolution of
the local normal stress (b) assuming elastic perfectly plastic soil mass (Mohr-Coulomb). Experimental data after CHOW [1997].
Fig. 16 Confronto fra i valori delle tensioni di taglio misurati sulla superficie laterale del palo IC e calcolati via analisi agli elementi
finiti (a), evoluzione della tensione normale locale calcolata via analisi agli elementi finiti assumendo il comportameto del terreno elastico
perfettamente plastico (Mohr-Coulomb). Dati sperimentali di CHOW [1997].

APRILE - GIUGNO 2008

60 DE GENNARO - FRANK - SAID

tions. Beyond this point a constant value of Qt-exp =

Tab. III Dunkerque pile test: comparison of predicted
and measured values of loads at failure. 309 kN for increasing values of the imposed pile
Tab. III Prova di carico su palo a Dunkerque: confronto fra i head displacement can be considered.
valori dei carichi di rottura misurati e calcolati via analisi The values of Qt include the tip load and the av-
numerica agli elementi finiti. erage shear stress mobilized along the pile shaft in-
tegrated all over the lateral surface of the model pile.
Qs (kN) Qp(kN) Qt (kN) It can be observed a fairly good agreement between
the predictions of the FEM simulation and the meas-
Test 217 92 309 ured values of Qs , Qp and Qt at failure. It seems licit
to infer that the slight underestimation of the total
Simulations 211 71 282
load at failure is mainly related to the low value ob-
tained for the tip load, 71 kN compared to the 92 kN
measured during the test (Tab. III). This depends
been found during calculations, as shown in very likely on the change in base capacity due to the
Figure 16b, and dilatancy of the interface, associ- densification and the increased soil stiffness below
ated to the increase of the radial (normal) stress, is the pile tip induced by the installation process.
well captured by the numerical simulations. r in These factors are not accounted in the numerical
the Leading and Following clusters is of about 60 analysis as neither the simulation of the pile penetra-
kPa and is close to 80 kPa in the Trailing and Lag- tion process assuming large strains analysis nor a
ging clusters (Fig. 16b). constitutive relationship for the sand mass capable to
The prediction of the axial capacity of the pile include changes in soil stiffness as a function of the
during the test is shown in Figure 17 in terms of total current state of density have been considered.
load (Qt) and tip load (Qp) vs pile head displacement
and summarized in Table III comparing the values
of Qs , Qp and Qt at failure, being Qs the average shear 6. Conclusions
stress mobilized along the pile shaft (experimental
data after CHOW 1997). Note that Qt-exp stops abruptly In this paper the results of the numerical analy-
sis of two compression tests on instrumented model
at failure (309 kN), while this is still increasing. This
piles installed in sand have been presented.
point corresponds to the loss of control of the pile
The importance of the constitutive modelling of
loading test performed under load controlled condi-
the interface layer has been clearly illustrated,
mostly in relation with the reproduction of the vol-
umetric behaviour at the interface in terms of dila-
tancy and/or contraction. It has been shown that un-
realistic dilatancy at the interface can be at the ori-
gin of unreasonable overestimation of the pile shaft
friction and of total bearing capacity in numerical
calculations.
Perfect plasticity and elastoplastic behaviour
have been considered for the sand surrounding the
pile and the interface. It has been found that a cou-
pling between the volumetric behaviour of the in-
terface and the surrounding soil might be possible
when the soil is considered elastoplastic. This cou-
pling does not modify radically the overall behav-
iour of the interface layer but changes the values of
normal stress on the pile shaft. The influence of
the soil model is by far more important if the be-
haviour of the soil below the model pile tip is con-
sidered.
Fig. 17 Comparisons between measured and predicted
For the analysis of the pile test in Dunkerque in-
total loads during Type 2 pile test in Dunkerque assuming
elastic perfectly plastic soil mass (Mohr-Coulomb). Exper- stallation effects in numerical simulations via finite
imental data after CHOW [1997]. elements have been simulated using empirical cor-
Fig. 17 Prova di carico su palo Type 2 a Dunkerque, relations based on field data. These have been used
confronto fra i valori dei carichi misurati e calcolati via analisi in order to account for the change in radial stress,
numerica agli elementi finiti assumendo il comportameto del shaft friction and base resistance. This strategy has
terreno elastico perfettamente plastico (Mohr-Coulomb). Dati been proved adequate in reproducing in a simple
sperimentali di CHOW [1997]. way the changes in the state of stress produced by

RIVISTA ITALIANA DI GEOTECNICA

FINITE ELEMENT ANALYSIS OF MODEL PILES AXIALLY LOADED IN SANDS 61

the installation procedure. Despite the good agree- terface. In Novel approaches in civil engineering,
ment between experimental data and numerical Frmond and Maceri (Eds.), Springer Verlag, pp.
predictions some aspect (e.g. densification of the 133-158.
sand below the pile tip ad increased soil stiffness) DE GENNARO V., SAID I., FRANK R. & MESTAT P.H.
still require a refined analysis able to reproduce the (2004) Numerical simulations of installation effects
real installation process. on embedded model piles in calibration chamber. Proc.
9th Int. Symp. on Num. Models in Geom. (NU-
MOG IX), Ottawa (Canada), pp. 197-203.
References DE GENNARO V., FRANK R. (2005) Modlisation de
linteraction sol-pieu par la mthode des lments finis.
ALAWNEH A.S., MALKAWI A.I.H. (2000) Estimation of Bull. de Liaison des Laboratoire des Ponts et
post-driving residual stresses along driven piles in Chausses, pp. 256-257, pp. 107-133.
sand. Geotechnical Testing Journal, vol. XXIII, n. DESAI C.S., ZAMAN M.M., LIGHTNER J.G., SIRIRWAR-
3, pp. 313-326. DANE H.J. (1984) Thin-layer element for interfaces
BRANSBY M.F., SPRINGMAN S.M. (1996) 3D finite ele- and joints. Int. J. Numer. Anal. Methods Geo-
ment modelling of piles groups adjacent to surcharge mech. 8, pp. 19-43.
loads. Computers and Geotechnics, 19, pp. 301- DIETZ M.S., LINGS M.L. (2006) Post peak Strength of
324. Interfaces in a Stress-Dilatancy Framework. Journal
BOND A.J. (1989) Behaviour of diplacment piles in of Geotechnical and Geoenvironmental Engi-
overconsolidated clays. PhD Thesis, Imperial Col- neering, ASCE, vol. cXXXII, n. 11.
lege, London. DUPLA J.C., CANOU J. (1994) Caractrisation mca-
BOULON M., NOVA R. (1990) Modelling of soil-struc- nique du sable de Fontainbleau partir dessais triaxi-
ture interface behaviour, a comparison between elasto- aux de compression et dextension. Rapport de Syn-
plastic and rate type laws. Computers and Geotech- these Projet National Clouterre II, CERMES/
nics, 9, pp. 21-46. ENPC, Paris, France.
BUSTAMANTE M. (1982) The pile loading test. Founda- FRANK R., GUENOT A., HUMBERT P. (1982) Numerical
tion Engineering, vol. I, G. Pilot (Ed.), Presse de analysis of contacts in geomechnics. Proc. 4th Int.
lcole Nationale des Ponts et Chausses, pp. Conf. on Num. Meth. in Geomechanics, Edmon-
263-273. ton, May 31-June 4, pp. 37-45.
CHOW F.C. (1997) Investigations in the behaviour of FRANK R., B AUDUIN C., DRISCOLL R., KAVVADAS M.,
displacement piles for offshore foundations. PhD The- KREBS OVESEN N., ORR T., SCHUPPENER B. (2004)
sis, Imperial College, London. Designers guide to EN 1997-1 Eurocode 7: Geotech-
COMODROMOS E.M., ANAGNOSTOPOULOS C.T., GEOR- nical design - General rules. Thomas Telford, Lon-
GIADIS M.K. (2003) Numerical assessment of axial don, 216 pp.
group response based on load test. Computers and GHIONNA V.N., MORTARA G. (2002) An elastoplastic
Geotechnics, 30, pp. 505-515. model for sand-structure interface behaviour. Go-
CUBRINOVSKI M., ISHIHARA K. (1998) Modelling of technique, 52, n. 1, pp. 41-50.
sand behaviour based on state concept. Soils and GOODMAN R.E., TAYLOR R.L., BREKKE T.L. (1968) A
Foundations, 38, 3, pp. 115-127. model for the mechanics of jointed rocks. J. of Soils
DAY R.A., POTTS D.M. (1994) Zero thickness interface Mech. and Found. Div., ASCE, 94, (SM3), pp.
elements. Numerical stability and application. Int. 637-659.
Journ. Num. Anal. Meth. Geomech., 18, pp. 689- HOHBERG J-M., SCHWEIGER H. (1992) On the penalty
708. behaviour of thin-layer elements. Proc. 4th Int.
DE GENNARO V. (1999) Etude du comportement mca- Symp. on Num. Models in Geom. (NUMOG IV),
nique des interfaces sol-structure. Application lanalyse Swansea, UK, pp. 241-248.
du comportement des pieux. PhD Thesis, Ecole Na- HUMBERT P. (1989) CESAR-LCPC, un code gnral
tionale des Ponts et Chausses, Paris, France, pp. de calcul par lments finis. Bull. Liaison Labo.
427. Ponts et Chausses, 160, fvr-mars, pp. 112-116.
DE GENNARO V., PANDE G.N., LERAT P. (2002) Stability JARDINE R.J., OVERY R.F., CHOW F.C. (1998) Axial
problems in soil-structure interfaces: experimental obser- capacity of offshore piles in dense marine sand. Journal
vations and numerical study. International Journal of of Geotech. and Geoenvir. Engrg., ASCE, 124 (2),
Geomechanics, vol. II, n. 2, pp. 175-203. pp. 171-178.
DE GENNARO V,. FRANK F. (2002) Elasto-plastic anal- JARDINE R., CHOW F.C., OVERY R., STANDING J. (2005)
ysis of the interface behaviour between granular media ICP design methods for driven piles in sands and
and structure. Computers and Geotechnics, 29, clays. Thomas Telford, pp. 112.
pp. 547-572. KATONA M.G., (1983) A simple contact-friction inter-
DE GENNARO V., FRANK F. (2003) Elastoplastic model- face element with application to buried culverts. Int.
ling of the behaviour of the granular soil-structure in- Journ. Num. Anal. Geomech., 7, pp. 371-384.

APRILE - GIUGNO 2008

62 DE GENNARO - FRANK - SAID

KUWANO R. (1996) Laboratory tests on Dunkirk sand. VAID Y.P., SASITHARAN S. (1992) The strength and di-
Imperial College Internal Report. latancy of sand. Can. Geotech. J., 29, pp. 522-526.
LI X.S., DAFALIAS Y.F. (2000) Dilatancy for cohesion- WAKAI A., GOSE S., UGAI K. (1999) 3-D elasto-plastic
less soils. Gtechnique, 50, n. 4, pp. 449-460. finite element analyses of pile foundations subjected to
LINGS M.L., DIETZ M.S. (2004) An improved direct lateral loading. Soils and Foundations 39, (1), pp.
shear apparatus for sand. Gotechnique, 54, n. 4, 97-111.
pp. 245256. WEHNERT M., VERMEER P.A. (2004) Numerical analy-
M ABSOUT M.E., REESE L.C. T ASSOULAS L. (1995) ses of load tests on bored piles. Proc. Num. Models in
Study of pile driving by finite element method. Journal Geomechanics, (NUMOG IX), pp. 505-511.
of Geotechnical Engineering, ASCE, 121, 7, pp.
535-543.
MAQTADIR A., DESAI C.S. (1986) Three-dimensional
analysis of a pile-group foundation. Int. J. Numer.
Anal. Methods Geomech. 10, pp. 41-58.
MAYNE P.M., KULHAWY F.H. (1982) Ko-OCR relation-
ships in soils. Journal of Geotechnical Engineering.
Analisi agli elementi finiti di pali
ASCE, 108 (GT6), pp. 851-872. modello sottoposti a carico assiale in
M OHAMEDZEIN Y.E.A., MOHAMED M.G., EL SHARIEF sabbie
A.M. (1999) Finite element analysis of short piles in
expansive soils. Computers and Geotechnics, 24,
pp. 231-243. Sommario
MUIR WOOD D., BELKHEIR K., LIU D.F. (1994) Strain Nel corso dellinterazione fra due materiali con caratteristiche
softening and state parameter for sand modelling. di deformabilit molto diverse la rottura spesso associata alla
Gotechnique, 44, n. 2, pp. 335-339. formazione allinterno del mezzo pi deformable di una zona di
N OVA R. (1977) On the hardening of soils. Arch. spessore ridotto, orientata parallelamente alla superficie di
Mech. Stosow, 29, (3), pp. 435458. contatto. In questa zona, generalmente identificata con il termine
NOVA R. (1982) A model of soil behaviour in plastic and di interfaccia, le deformazioni si localizzano creando una
hysteric ranges. Part I: Monotonic loading. Results of discontinuit cinematica caratterizzata da forti gradienti di
the International Workshop on Constitutive Re- tensioni. La maggior parte delle strutture in ingegneria civile in
contatto con i terreni. In tali circostanze, e per delle condizioni di
lations for Soils, Grenoble, 6-8 September, pp.
sforzo critiche, la rottura spesso osservata allinterfaccia, nel
289-309.
mezzo pi deformabile: il terreno. La descrizione del
PANDE G.N., SHARMA K.G. (1979) On joint/interface comportamento meccanico del sistema terreno-struttura,
elements and associated problems of numerical ill-con- pricipalmente in termini di attrito mobilizzato fra lelemento
ditioning. Int. J. Numer. Anal. Methods Geo- strutturale e il terreno, deve conseguentemente richiedere una
mech., 3, pp. 293-300. modellazione precisa del comportamento dellinterfaccia.
POTTS D.M., ZDRAVKOVIC L. (2001) Finite element In questo lavoro lattenzione sar concentrata sul
analysis in geotechnical engineering. Thomas Tel- comportamento dei pali di fondazione in sabbie sottoposti a carico
ford, London. assiale, un esempio tipico dinterazione terreno-struttura fortemente
SAID I. (2006) Comportement des interfaces et modlisa- influenzato dal comportamento dinterfaccia. Il problema sar
tion des pieux sous charge axiale. PhD Thesis, Ecole analizzato utilizzando il metodo degli elementi finiti (FEM). Si
Nationale des Ponts et Chausses, Paris, France. fornir una descrizione preliminare della modellazione numerica
SCHANZ T., VERMEER P.A. (1996) Angles of friction and FEM del problema del contatto. Ci si interesser successivamente,
dilatancy of sand. Gotechnique, 46, n. 1, pp. 145- in dettaglio, al comportamento dinterfaccia e alla sua descrizione
matematica nel quadro della teoria dellelastoplasticit. Si
151.
forniranno anche dei richiami sulla modellazione matematica del
SHENG D., WRIGGERS P., SLOAN S.W. (2007) Applica-
comportamento delle sabbie. Al fine di risolvere gli ostacoli numerici
tion of frictional contactin geotechnical engineering. associati alla simulazione degli effetti dellinstallazione del palo
International Journal of Geomechanics, vol. VII, sar proposta una strategia numerica alternativa per attenuare la
n. 3, pp. 176-185. relativa complessit dei metodi numerici utilizati per simulare i
SCHOFIELD A.N., WROTH P. (1968) Critical state soil problemi di penetrazione. Saranno infine proposte, come
mechanics. McGraw-Hill, European Civil Engi- validazione, delle analisi comparative dei risultati di prove di carico
neering series, London. su pali modello e delle relative analisi FEM.

RIVISTA ITALIANA DI GEOTECNICA