You are on page 1of 6

Numerical Methods in Geotechnical Engineering – Schweiger (ed.

)
© 2006 Taylor & Francis Group, London, ISBN 0-415-40822-9

Modelling piles under multi-directional loading conditions

N.H. Levy
Centre for Offshore Foundation Systems, University of Western Australia, Perth, Australia

I. Einav
School of Civil Engineering, University of Sydney, Sydney, Australia

ABSTRACT: The behaviour of pile foundations subjected to single directional lateral and axial loading has
been substantially investigated in the past. However, the response of a pile subjected to a change in lateral load-
ing direction or inclined loading (i.e. combined axial and lateral loading) has not been treated adequately. There
is an increasing demand for a design method that considers the interaction between the responses of the pile-
soil system when loads are applied in more than one direction, particularly for pile foundations that are installed
offshore, which incur substantial lateral and torsional loads. This paper presents a multi-directional Winkler
system that is modeled using an energy based variational (EBV) method, considering interaction at each depth
along the pile using a series of local yield surfaces.

1 INTRODUCTION benefit from the more realistic damping capabilities


of multi-surface plasticity when compared to single
The response of piles subjected to lateral or axial surface plasticity. It is shown that the new model pre-
loading has been investigated in detail and established dicts a stronger effect of recent load history when
methods exist for determining this response (Fleming smaller pile displacements are applied, but equivalent
et al. 1992, API 1993). These methods generally results in large displacements.
model the soil as non-linear elastic, which limits their
use to monotonic loading. In recent years the soil has
been represented as an elasto-plastic medium with a 2 THE MODEL
yield or ultimate soil pressure defined as a function of
depth (Murff 1975, Guo & Randolph 1997, Guo & The EBV approach allows all information regarding
Lee 2001, Hsiung 2003, Einav 2005). The concept of the geometry and constitutive assumptions of the pile-
using a continuous field of yield surfaces to create a soil interaction problem to be encapsulated within two
smooth non-linear curve was introduced by Einav potential functionals (which are defined explicitly). For
(2005) for both axial and lateral loading. rate independent isothermal deformations the energy
The current paper presents an energy based vari- and dissipation potentials must satisfy the following
ational (EBV) approach to solve the pile-soil system equation:
using the method presented by Einav (2005) for dissi-
pative soil. A general model is presented for multi- (1)
directional loading based on the yield surface model
presented by Levy et al. (2005). This model allows where ␦W is the change in virtual work, ␦⌿ is the
inclined loading to be modeled considering the inter- change in internal energy and ⌽ is the change in
action between the lateral and axial pile deformations. mechanical dissipation (which is strictly non-negative).
This type of model is also useful for the loading of The change is virtual work is defined as:
piles in unaligned lateral directions, where the resid-
ual displacements from loading in previous lateral
directions (“recent load history”) are taken into (2)
account for subsequent loading (Levy et al. 2006).
In this paper we extend our pile model of recent
load history to a model with infinite yield surfaces at where r is the unbalanced force vector and ␦u is the
each point along the length of the pile. In this way we change in displacement vector, each having a length n

563
defined by the number of perpendicular directions The components of the unbalanced force can now
being considered (i.e. including x, y and/or z where x be defined as (for derivation see Einav 2005):
and y are perpendicular lateral directions and z
represents the axial direction). The symbol L repre-
sents the length of the pile and z is the depth below (6)
the pile head. The unbalanced force vector is generally
defined as zero along the length of the pile and as the
imposed external forces at the pile head. If required a The soil energy density function ␺Si in each direction
non-zero force could be applied along the pile length i is defined as:
to represent lateral soil movement.
The internal energy is defined as the sum of the (7)
contributions from the pile and soil as follows:
(3)
The soil energy potential ⌿S is then defined as:
where the subscripts P and S refer to the pile and soil
respectively.
(8)
The pile is assumed to deform in an elastic man-
ner according to Euler-Bernoulli beam theory and is
defined for a circular pile as (Reddy 2002, Einav 2005):
The change in energy potential can be defined as:

(9)
(4)

where EP is the Young’s modulus of the pile (assumed The yielding behaviour of a soil subjected to lateral
constant along the pile for simplicity), IP is the moment loading can be defined using the Randolph-Houlsby
of inertia of the pile about the axis of bending (for a mechanism (Randolph & Houlsby 1984, Einav &
circular pile Ixx ⫽ Iyy ⫽ IP), AP is the cross-sectional Randolph 2005). As the limiting reaction will be the
area of the pile. The x, y and z terms are included same in all lateral directions, a circular yield surface
when required according to the loading conditions is suitable to define this behaviour.
and the functions ux, uy and uz represent the x, y and z It is less obvious how the soil will behave when
components of the total displacement vector u, which subjected to combined lateral and axial loading. Levy
has a length n. et al. (2005) presented an elliptical yield surface for
this behaviour but suggested that further research was
2.1 Single yield surface soil model required to refine this shape.
The general form of the dissipation function, assum-
The soil is modeled at each depth along the pile using ing an elliptical relationship between the lateral and
independent Winkler springs, one each in the perpen- axial directions, will be as follows:
dicular directions, and a multi-dimensional yield
surface.
In this model the soil resistance force (fi) in each (10)
direction i is the same as the spring force (␹i), which
is defined by a linear force-displacement relationship
with the elastic displacement and has spring stiffness
Esi. The spring stiffness can also be referred to as the where fui is the ultimate soil resistance force in the
modulus of subgrade reaction and can be defined as a i direction.
function of the pile and soil properties. The soil resist- A set of yield functions can be obtained from the
ance force fi can be defined by: dissipation potential using a set of degenerate Legendre
transformations. The local elastic behaviour is bounded
by a three dimensional ellipsoidal yield function (y)
(5) defined by the following:

where ␣i is the plastic displacement in the i direction


and the elastic displacement is defined as the total (11)
displacement minus the plastic displacement.

564
Yield surface The total soil resistance force at depth z in the i direc-
tion is defined as the integral of all the spring forces
fu(z) fy(z) over the internal variable ␩:
fx(z)

L
z

(13)

Figure 1. Schematic representation of a two dimensional


Therefore the soil free energy function can be
single yield surface interaction model. defined as:

(14)
An example of this type of model is presented in
Figure 1 for two dimensional lateral loading, where
fu ⫽ fux ⫽ fuy. This can be extended to three dimen-
sions and an ellipsoidal yield surface, as expressed in The dissipation function can be defined as:
Equation 11.

2.2 Soil model 2 – continuous yield surfaces (15)


An elasto-plastic non-linear force versus displace-
ment relationship can be represented using the EBV
approach. Rather than having a single spring-slider The yield surfaces are defined as:
element in a given direction Einav (2005) presented a
continuous Masing-Iwan model. This model consists
of a group of spring-slider elements arranged in series (16)
with each element having a different spring stiffness
and slider threshold depending on a local internal vari-
able ␩ that signifies the amount of yielding at each
point along the length of the pile. Initially, before the
occurrence of any yielding, ␩ is 0, but as yielding 3 EXAMPLES
in a soil layer progresses, ␩ tends to 1. In this one-
directional model, each value of ␩ marked a single The multi-dimensional EBV model provides a versatile
spring and a single one-directional slider. approach to solving the cases of both simultaneous
In this paper we extend this model to two dimen- and staged loading in different directions. An example
sions by considering an additional spring per internal of each of these cases will be presented here.
coordinate ␩ and representing the yielding by a two-
dimensional yield surface rather than the slider. 3.1 Simultaneous lateral and axial loading
Distribution functions are defined for the spring
stiffness (H) and yield threshold (k) in each direction i The case of simultaneous lateral and axial loading was
and these must satisfy the following condition: presented by Levy et al. (2005) using the differential
equations defined for each loading in each direction
and a two-dimensional elliptical yield surface to couple
the soil forces. These differential equations can be
derived from the energy potentials (Einav 2005) and
therefore this situation can be defined using the single
yield surface model. Sensitivity analyses were com-
The spring forces are defined for each internal vari- pleted on the influence of pile flexibility and the inclin-
able ␩ in a direction i as: ation of the imposed displacement. For the case of
a 20 m length, 2 m diameter solid circular steel pile
(12) embedded in normally consolidated clay the applied

565
9000 12
θ = 90° L = 5m
8000
θ = 60° 10 L = 10m
axial pile head force (kN)

7000 L = 15m

% increase in u0
6000 L = 20m
8
5000
θ = 45° θ = 30° 6
4000
3000 4
2000
2
1000
0 0
0 0.01 0.02 0.03
0 20 40 60 80 100
axial pile head displacment (m)
% of Pult
Figure 2. Axial pile head force versus displacement for
inclined loading (from Levy et al. (2005)). Figure 3. Percentage increase in resultant displacement rela-
tive to monotonic loading conditions (␪ ⫽ 90°) (from Levy
et al. 2006).

axial force versus the pile head displacement is shown


in Figure 2 for various displacement inclinations from applied load and the angle between loading stages on
the horizontal (␪). the pile behaviour. An example of a steel pile with a
The results from this study suggest that the inclin- diameter of 2 m and 20 mm wall thickness embed-
ation of the applied load will influence the axial pile ded in normally consolidated clay is presented below.
behaviour. A greater lateral load will cause the pile to Figure 3 shows the influence of the magnitude of
displace further in the vertical direction. A sensitivity loading during initial loading (as a percentage of the
analysis on the pile flexibility shows that a more flex- pile capacity, Pult) on the resultant pile head displace-
ible pile will be influenced more in the axial direction ment (u0), as a percentage increase from the displace-
by an inclined load. ment achieved for monotonic lateral loading. The
This influence was shown to be less prominent for change in loading direction (␪) for this case was 90°
the lateral pile behaviour, with an applied axial load and four different pile lengths were considered.
having minimal effect. This plot suggests that the more a pile is loaded
This study was presented as a starting point for this initially the greater the increase in the final pile head
type of yield surface interaction model. The results show displacement. It also shows that a longer, (i.e. more
trends that agree with previous studies on inclined load- flexible) pile will be influenced more by previous
ing (Shahrour & Meimon 1991, Anagnostopoulos & loading. This is because more stored energy is frozen
Georgiadis 1993). Currently work is being undertaken along the length of the pile as a result of the yielding
to refine the yield surface shape to provide a more real- such that after unloading a more significant plastic
istic representation of the interaction mechanism. displacement remains further from the zero position.
An even more astonishing factor is presented when
examining the resultant load-displacement response.
3.2 Staged two dimensional lateral loading A typical load-unload-reload curve is shown in
A pile subjected to monotonic lateral loading will Figure 4 where Stage I is initially loading to a given
behave the same in any direction using the EBV force (P/Pult ⫽ 0.75) in the x direction, Stage II is
model. This model can be applied when a pile is unloading to zero force and Stage III is loading at 90°
loaded in a lateral direction and then in a different (in the y direction) to the same P/Pult as Stage I. The
direction on the horizontal plane. Levy et al. (2006) shape of the curve of Stage III is markedly different
coined the term “recent load history” to describe the than that of Stage I, ending with additional displace-
soil-pile response in these conditions. As this term ment as summarised in Figure 3.
suggests loading history influences the behaviour of a
pile during subsequent loading.
3.3 Extension to continuous yield
Levy et al. (2006) presented a numerical example
surface model
of a pile loaded in a given direction, unloaded to zero
and then loaded in a new direction. The study explores In this section we extend the above model to the case
the influence of the pile flexibility, magnitude of the of continuous yield surfaces. This allows a non-linear

566
0.8 has reached the yield value over a section of the pile
some plastic deformation will remain after unloading.
Initial Loading Curve
0.7
(Stage I) In contrast, for the case where P/Pult ⫽ 0.125 the sin-
0.6 gle yield surface model does not reach yield at any
point along the pile and therefore the pile will return
0.5 to its original position after unloading. Due to the
F/Pult

gradual yielding the multiple yield surface model


0.4 Stage II does exhibit a plastic displacement. Since it is well
0.3 established that the fundamental constitutive stress-
strain behaviour of clays presents small strain non-
0.2 Stage III linearity the multiple yield surface model is more
representative, accounting for the additional damping
0.1
in the soil. It is shown that the new model predicts a
0 stronger effect of recent load history when smaller
0 0.02 0.04 0.06 0.08 0.1 0.12 pile displacements are applied, but similar results in
u0/D larger displacements.

Figure 4. F/Pult versus u0/D at the pile head for P/Pult ⫽ 0.75.
4 CONCLUSIONS

1.2 A model is presented in the paper that enables an


Multiple Surface analysis of the behaviour of a pile loaded in multiple
directions. This model allows different soil conditions
1 Single Surface
to be considered as well as many loading combin-
ations. This type of interaction model has enormous
0.8 potential to consider the interaction between loading
P/Pult = 0.75 directions and design piles more efficiently.
uy0/umono

0.6
REFERENCES
0.4
P/Pult = 0.125 Anagnostopoulos, C., & Georgiadis, M. (1993). Interaction
of axial and lateral pile responses. Journal of Geotechnical
0.2 Engineering, 119(4), 793–798.
API. (1993). Recommended practice for planning, design-
ing and constructing fixed offshore platforms – Working
0 stress design, 20th edition. API-RP-2A.
-0.2 0 0.2 0.4 0.6 0.8 1 Einav, I. (2005). Energy and variational principles for piles
ux0/umono in dissipative soil. Geotechnique, 55(7), 515–525.
Einav, I., & Randolph, M. F. (2005). Combining upper
Figure 5. Evolution of pile head displacement determined bound and strain path methods for evaluating penetration
using a single and multiple yield surface models. resistance. International Journal for Numerical Methods
in Engineering, 63(14), 1991–2016.
Fleming, W. G., Weltman, A. J., Randolph, M. F., &
Elson, W. K. (1992). Piling Engineering, E & FN Spon,
force displacement profile to be considered, which will New York & London.
represent the soil in a way that more closely resembles Guo, W. D., & Lee, F. H. (2001). Load transfer approach for
the curves defined from physical testing results laterally loaded piles. International Journal for
(API 1993). Numerical and Analytical Methods in Geomechanics,
The continuous yield surface model allows the 25(11), 1101–1129.
behaviour of the soil during unloading of a pile to be Guo, W. D., & Randolph, M. F. (1997). Vertically loaded
better represented. The analysis discussed in the pre- piles in non-homogeneous media. International Journal
for Numerical and Analytical Methods in Geomechanics,
vious section for staged lateral loading was repeated
21(8), 507–532.
using the continuous yield surface model. A compari- Hsiung, Y. -M. (2003). Theoretical elastic-plastic solution
son between the pile head location as suggested by for laterally loaded piles. Journal of Geotechnical and
the two alternative models is shown in Figure 5. Geoenvironmental Engineering, 129(5), 475–480.
At large loads the behaviour of the pile does not Levy, N. H., Einav, I., & Randolph, M. F. (2005). Modelling
vary significantly between the two models. As the soil combined loading of piles with local interacting yield

567
surfaces. International Symposium on Frontiers in Randolph, M. F., & Houlsby, G. T. (1984). Limiting Pressure
Offshore Geotechnics, Perth, Australia, 873–879. on a Circular Pile Loaded Laterally in Cohesive Soil.
Levy, N. H., Einav, I., & Randolph, M. F. (2006). Effect Geotechnique, 34(4), 613–623.
of Recent Load History on Laterally Loaded Piles in Reddy, J. N. (2002). Energy Principles and Variational
Normally Consolidated Clay. The International Journal Methods in Applied Mechanics, John Wiley & Sons, Inc.,
of Geomechanics, (submitted). Hoboken, New Jersey.
Murff, J. D. (1975). Response of axially loaded piles. Shahrour, I., & Meimon, Y. (1991). Analysis of the behav-
Journal of Geotechnical Engineering Division, ASCE, iour of offshore piles under inclined loads. International
101(3), 356–360. Conference on Deep Foundations, 277–284.

568

You might also like