Marine Structures 50 (2016) 260e275

Contents lists available at ScienceDirect

Marine Structures
journal homepage: www.elsevier.com/locate/marstruc

Ultimate bearing capacity of laterally loaded piles

in clay e Some practical considerations
Youhu Zhang a, *, Knut H. Andersen a, Giacomo Tedesco b
a
Norwegian Geotechnical Institute, Sognsveien 72, 0855 Oslo, Norway
b
Department of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Via Terracini 28, 40131 Bologna, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The paper re-visits the topic of ultimate bearing capacity of laterally loaded piles in clay.
Received 17 April 2016 The paper first presents a review of various recommendations made by design guidelines
Received in revised form 14 September 2016 and industry practitioners which illustrates inconsistency and need for further work. A
Accepted 22 September 2016
literature study is therefore performed and a generalised recommendation of the ultimate
lateral capacity of piles in clay that is self-consistent and flexible for a wide range of
conditions is made. The paper further investigates two practical considerations often
Keywords:
encountered in design, namely the effect of axial loading and the effect of soil strength
Pile
Clay
anisotropy by means of finite element analyses. Practical methods to account for these
Lateral loading effects are proposed.
Bearing capacity © 2016 Elsevier Ltd. All rights reserved.
Anisotropy
Suction

1. Introduction

Piles, particularly offshore piles, often have to be designed to resist lateral loading. Load transfer curves (commonly known
as p-y springs) which express the lateral soil resistance versus deflection are usually used in design and the overall pile
response is solved by a beam-column analysis. This paper deals with the ultimate lateral bearing capacity for piles in clay, i.e.,
the peak value for the p-y springs. It is well established that for a laterally loaded slender pile in clay, two soil failure
mechanisms can be involved, as illustrated in Fig. 1. In the upper part, soil fails in a conical wedge mechanism that extends to
the soil surface. If suction is not available (i.e. allow separation), only a passive wedge is mobilised on the front side (Fig. 1a),
with a gap potentially opening up on the rear side. Whereas if suction is available, an active wedge is also mobilised on the
rear side of the loading direction. At a certain depth, soil fails in a localised flow around mechanism (Fig. 1b), as the soil
resistance encountered in this mechanism becomes smaller (i.e. preferential) than the wedge mechanism.
There have been extensive studies in the literature on the limiting lateral bearing capacity of pile foundations in clay. To
name only a few, these include, for example, experimental studies: [8,14,19,22]; analytical: [13,15,18,20,21]; and numerical:
[4,5,16,20,21]. However, in the industry, a common agreement seems not reached on what is the limiting pile capacity in clay
and its evolution with depth. Different recommendations are seen in various sources. Design engineers therefore may face the
question on which recommendation to follow.

* Corresponding author.
E-mail addresses: youhu.zhang@ngi.no (Y. Zhang), knut.h.andersen@ngi.no (K.H. Andersen), giacomo.tedesco@unibo.it (G. Tedesco).

http://dx.doi.org/10.1016/j.marstruc.2016.09.002
0951-8339/© 2016 Elsevier Ltd. All rights reserved.
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 261

Fig. 1. Illustration of (a) conical wedge failure mechanism (after Murff and Hamilton [15], and (b) flow around mechanism (after Randolph and Houlsby [18].

A pile loaded laterally can often be subjected to axial loading at the same time. A further question often faced by design
engineers is how to account for the effect of axial loading on the lateral pile capacity. In the current industry practice, the
interaction between axial and lateral loading are usually not explicitly considered. Instead, piles are designed with separate
verifications for axial and lateral loading.
It is noted that most analytical and numerical studies in the literature have almost exclusively considered isotropic ma-
terial behaviour, with a single measure of the undrained shear strength su, based on which solutions are developed. However,
it is well known that clay exhibits different undrained shear strengths, depending on the shearing mode/direction with
respect to deposition plane and major principal stress direction. Different testing methods normally measure different un-
drained shear strengths. In the current offshore industry practice, anisotropically consolidated undrained triaxial compres-
sion (CAUC), anisotropically consolidated triaxial extension (CAUE) and direct simple shear (DSS) strengths are typically
measured in the laboratory. Unconsolidated and undrained compression (UU) tests are also often performed offshore and
onshore, which measure soil strengths that are known as sUU u . In addition, soil resistance of in situ CPT or T-bar tests are often
correlated to undrained shear strength by applying empirical/theoretical correlations. The question faced by design engineers
is how to account for shear strength anisotropy, or in other words, what shear strength should be used to compute the ul-
timate lateral bearing capacity while using the solutions derived for isotropic material?
This paper re-visits the topic of the ultimate lateral bearing capacity of pile foundations in clay in an attempt to tackle the
three important questions faced by design engineers, namely: 1) what is the limiting lateral bearing pressure in clay and its
evolution with depth in ideally isotropic material; 2) what is the effect of axial loading on lateral capacity; 3) what is the effect
of strength anisotropy? Practical guidance is provided on these topics.

2. Methodology

In this study, a combined methodology of literature study and finite element analyses (FEA) is adopted. For the first
question, literature study is the main method, complemented by FEA results. For the second and third questions (effect of
axial loading and strength anisotropy), FEA is mainly relied on. This section provides a brief description of the finite element
model.

2.1. Finite element model

The FEA are performed with commercial finite element program Plaxis 3D [17]. The finite element model is shown in Fig. 2.
The pile is 2.14 m in diameter (D), and 40 m embedded in soil. Due to symmetry condition, only half of the pile is modelled.
The pile is modelled as a linearly elastic solid object, assigned with bending stiffness equivalent to a steel hollow pile 60 mm
in wall thickness. For ease of extraction results from analyses, a beam with a small bending stiffness, 1/1000 relative to the
actual pile, is attached to the pile at the pile centre, as illustrated in Fig. 2. The dummy beam is enforced to deform together
with the pile. The cross-section forces (shear force and bending moment) of the pile are simply 1000 times of those of the
beam, which are direct outputs from the finite element program.
The finite element model is discretised with 10-node tetrahedral elements with biased mesh densities. A refined zone with
a radius of three pile diameters from the centre line of the pile is used, in which typical element dimension is about 0.3D.
Outside the refined zone, the element size increases gradually with distance from the pile. The finite element mesh is cali-
brated against existing solutions, as discussed in Section 2.3.
262 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

Fig. 2. Finite element model.

2.2. Soil and contact properties

In order to investigate the ultimate lateral bearing capacity of pile foundations in isotropic material and study the
interaction between axial and lateral loading, the soil was modelled as a linear elastic perfect plastic material with a Tresca
yield criterion, with the yield stress defined as the undrained shear strength (su).
To further investigate the effect of strength anisotropy, the NGI-ADP model [6], which allows for definition of anisotropic
strength, was used. The model is total stress based, with a Tresca type yield criterion and takes directly the anisotropy ratios
measured in laboratory tests as input. The model is available in Plaxis 3D as a built-in soil model.
Contact elements available in Plaxis 3D were used to simulate the pile-soil interface. Either fully smooth or fully rough
interface was simulated in the analyses. For a smooth interface, a shear strength of 0.1 kPa was assigned to the interface
material. For a rough interface, the intact soil strength was assigned to the interface material.
Note that the paper is mainly concerned with offshore pile applications, i.e. the soil is always under free water. Therefore
effective unit weight is used in all analyses. The tension cut-off feature with the contact elements allows for simulation of full
suction or no suction conditions. When tension cut-off is ticked on, no suction condition is simulated and vice versa.

2.3. Benchmarking of finite element mesh

In order to benchmark the finite element mesh, homogenous soil strength (no variation with depth), and full suction (i.e.
no separation between pile and soil) was considered; therefore, both passive and active wedges are mobilised at shallow
depth. In this case, the soil weight does not contribute to the ultimate lateral bearing pressure pu. Fig. 3 presents the calculated
ultimate lateral bearing capacity factor (Np), defined as pu/su, along the pile for both smooth and rough interface conditions. It
is shown that, for both roughness conditions, the Np factor increases steadily from the soil surface down to 2.0e2.5D and then
becomes stable. This transition is coupled with the change of soil failure mechanism from a conical wedge mechanism to a
localised flow around mechanism, as illustrated in Fig. 1. For the flow around mechanism, the Np factor calculated by Plaxis 3D
averages around 9.86 and 12.2 for the smooth and rough interface conditions respectively. Compared with the theoretical
solutions provided by Refs. [13,18] that Np ¼ 9.14e9.20 (lower and upper bounds) and 11.94 (exact) for smooth and rough
interface respectively, the current numerical results represent a numerical error of 7% and 2% respectively. It is unclear why
the numerical model yields slightly higher overshoot for the smooth interface than for the rough interface. In the rest of this
paper, the numerical results presented are not corrected for the numerical overshooting.
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 263

Fig. 3. Calculated Np factor in soil with homogenous shear strength (suction assumed).

3. Ultimate lateral bearing capacity in isotropic soil

3.1. Selected recommendations

3.1.1. API recommendation

In the current industry guidelines widely adopted for offshore pile design, such as [1]; the method for calculating the
ultimate lateral bearing pressure (pu) is mainly based on the back analyses of some limited field pile tests performed in the
1950s [14]. The recommendation is:

pu ¼ ð3 þ Jz=DÞsu þ g’z (1)

where su is the undrained shear strength of the soil at the depth of consideration z, J is a dimensionless empirical factor. D is
the diameter of the pile and g0 is the effective unit weight of the soil.
The capacity calculated by Eq. (1) is limited by the flow around capacity of 9su. The onset depth of flow around mechanism
is the depth where pu calculated by Eq. (1) reaches 9su.
In Eq. (1), the term (3 þ Jz/D)su represents the contribution to pile capacity due to soil strength, and the g'z term represents
the contribution of soil weight to lateral bearing capacity before a localised flow around mechanism is formed. Eq. (1)
suggests a linear increase of contribution to capacity by soil strength with normalised depth (z/D) by an empirical factor
of J, which has a value of 0.25e0.5. For Golf of Mexico clay, J is recommended to be 0.5 [1]. At z ¼ 0, Eq. (1) suggests Np ¼ 3.
Furthermore, the inclusion of a soil weight term to pu implies assumption of a single wedge mechanism (i.e. no suction
condition). The limiting bearing capacity of 9su implies a fully smooth interface assumption, based on the theoretical study of
[13,18].

3.1.2. Jeanjean [8] recommendation

Based on centrifuge experiments and finite element analyses in lightly over-consolidated clay [8], Jeanjean proposed the
following expression to calculate the lateral bearing capacity factor Np, defined as pu/su, following the framework proposed by
Ref. [15]:


xz
D
Np ¼ 12  4e (2)

The parameter x is calculated as

264 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

x ¼ 0:25 þ 0:05l where l < 6

(3)
¼ 0:55 where l > 6

where l ¼ sum/kD, which represents a normalised soil strength gradient in a linearly increasing strength profile that can be
described by:

su ¼ sum þ kz (4)

where sum is the shear strength at soil surface, k is the shear strength gradient with depth.
It should be noted that in Eq. (2), the contribution to bearing capacity by the soil weight is absent. This is because in their
centrifuge experiments, a gap on the rear side of the pile is not observed, in which case, the soil weight does not contribute to
the capacity. Instead, an active soil wedge is mobilised.
Eq. (2) has two important implications. At depth, it gives a limiting Np factor of 12. This corresponds to an assumption of a
fully rough soil-pile interface. This is perhaps a reasonable assumption for the soft clay tested in the centrifuge, which, ac-
cording to methods proposed for axial skin friction calculations for example [1,9], would typically suggest an interface factor a
close to 1. At the soil surface, Eq. (2) suggests an Np ¼ 8. As will be discussed in Section 3.4, a surface Np ¼ 8 is too high, even for
a fully rough soil-pile interface with suction.

3.1.3. Nichols et al. [16] recommendation

Based on a combined literature and finite element study [16], Nichols et al. proposed the following equations to calculate
the ultimate lateral bearing pressure:

pu ¼ Np0su þ g0z (5)

Np0 ¼ 4 þ 2ðz=DÞ0:6 (6)

It should be noted that the capacity calculated by Eq. (5) is proposed to be limited to 11.9su, corresponding to capacity for a
flow around mechanism with a fully rough interface.
In Eq. (5), the total capacity is calculated as a sum of contributions from soil strength and soil weight, which implies an
assumption of no suction condition, similar to the API recommendation. Note that a subscript '00 is added to Np in Eq. (6) since
the contribution of soil weight to pile capacity is isolated in its definition. However, as it will be shown later, the factor
Np0 ¼ 4 at the soil surface is too high for a single wedge mechanism even for a fully rough interface.
Perhaps the most important contribution from Ref. [16] is the inclusion of soil strength anisotropy since su to be used in Eq.
(5) is recommended to be an average value between direct simple shear strength (sDSS E
u ) and triaxial extension strength (su).
This will be further discussed in Section 6 of this paper.

3.1.4. Some observations from review of selected recommendations

In the above sections, a widely used industry guideline and two proposals from industry practitioners on the ultimate
lateral bearing capacity of pile foundations in clay are presented, which do not suggest a common agreement among each
other. A design engineer may therefore feel confused on what to choose to use. As discussed above, the differences among the
various recommendations seem to arise from the following aspects:

 assumption of pile-soil interface roughness

 assumption of suction on rear side of the pile (therefore inclusion/exclusion of soil weight contribution)
 the effect of increasing shear strength profile with depth
 the variation of lateral bearing capacity factor with depth
 the effect of strength anisotropy

In the next section, a literature study will be presented, complemented by FEA results in order to shed light on the above
mentioned various aspects (except the strength anisotropy, which will be discussed in Section 6). A recommendation that is
intended to be general for a range of conditions and self-consistent with various assumptions is attempted.

3.2. Literature study e ultimate pile capacity in weightless, isotropic soil

In depth, where flow around failure mechanism is formed, [13,18] provide theoretical solutions based on limit analyses for
piles with different interface roughness. The ultimate lateral bearing capacity factor is found to vary in an approximately
linear manner from 9.14 to 9.20 (lower and upper bounds) for a fully smooth interface to 11.94 (exact) for a fully rough
interface. These solutions for the flow around mechanism are widely accepted in the industry. However, considerable
confusion still exist in at what depth the flow around mechanism is formed and what is the limiting capacity before such
mechanism is formed.
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 265

Murff and Hamilton [15] presents a fundamental study on the ultimate bearing capacity mobilised in a passive wedge
failure mechanism in clay by means of upper bound limit analysis. No suction was assumed and the soil was idealised as
weightless in order to isolate the contribution of soil weight. Soil profiles with uniform or linearly increasing undrained shear
strength with depth were studied. The following expression is summarized to calculate the ultimate lateral bearing capacity
factor in idealised weightless soil (Np0 ¼ pu/su) at depth z:


xz
D
Np0 ¼ N1  N2 e (7)

where N1 is the limiting Np0 factor at depth corresponding to a flow around mechanism, N1-N2 is the Np0 value at the soil
surface. For a smooth pile-soil interface, N1 ¼ 9 and N2 ¼ 7. x is taken to be a function of normalised strength gradient l and
calculated according to Eq. (3).
Based on similar principles, Yu et al. [20] presents an upper-bound limit analysis for the same problem, utilizing an
improved failure mechanism by a) adopting of a curved wedge failure surface; b) including the energy dissipation of the
transition interface between wedge failure and plane strain flow around failure. As a result, lower Np0 values are found than
those presented in Ref. [15]. The transition depths for flow-around mechanism predicted by Ref. [20] are consequently deeper.
In Ref. [20]; the upper bound solutions are also verified with finite element analyses (FEA). Fig. 4 illustrates a comparison of
upper-bound solutions between the two studies for homogeneous clay, which are for a no suction pile-soil interface.
Based on their results, Yu et al. [20] provides an alternative equation to describe the Np0 factor in homogeneous clay, which
has the form:
"  #1:35


z=D 0:6
Np0 ¼ N1  ðN1  N2 Þ 1   ð1  aÞ (8)
14:5

where N1 ¼ 11.94 is the limiting value at depth for a fully rough pile, corresponding to a flow around mechanism. N2 ¼ 3.22,
which is the Np0 factor at soil surface for a fully rough pile. a is the interface roughness factor. The term (1-a) represents a
correction term for interface roughness. Note that the Np0 value calculated by Eq. (8) is limited by the flow-around bearing
capacity factor, which is 11.94 and 9.20 for rough and smooth interfaces respectively. The normalised depth where soil
mechanism transits from wedge to flow around mechanism in an idealised weightless soil, as implied by Eq. (8), varies in a
linearly manner from 14.5D to 8.0D for a fully rough (a ¼ 1) to a fully smooth (a ¼ 0) pile soil interface.
Yu et al. [21] further present upper bound results for soil profiles with linearly increasing undrained shear strength. Fig. 5
re-presents the data for a smooth soil-pile interface with different normalised strength gradient (l ¼ sum/kD). The results are
for the idealised weightless soil, with no suction on pile-soil interface. Over a wide normalised strength gradient, the nor-
malised strength gradient is shown to have only modest impact on the bearing capacity factors. As the soil becomes
increasingly heterogeneous (l reduces), the bearing capacity factor reduces for the same normalised depth (z/D). As a result,
the transition depth to flow around mechanism increases. To capture this trend, a minor modification to Eq. (8) is proposed in
this paper:

Fig. 4. Comparison of upper bound solutions in idealised weightless homogeneous clay (data points digitised from original papers).
266 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

Fig. 5. Effect of normalised shear strength gradient on distribution of Np0 factor (smooth soil pile interface).

"
 #1:35
z=D 0:6
Np0 ¼ N1  ðN1  N2 Þ 1   ð1  aÞ (9)
d

where d stands for the normalised depth at which the mechanism transits from wedge failure to flow around failure for a fully
rough soil-pile interface in idealised weightless soil. The value of d is related to the normalised strength gradient (l) by

d ¼ 16:8  2:3 log10 ðlÞ  14:5 (10)

The proposed modification is shown to capture the trend well, as illustrated in Fig. 5, where the theoretical solutions
presented by Yu et al. [21] for various different l values are compared against prediction of Eq. (9).

3.3. Effect of soil weight and onset depth of flow around mechanism

In real soil, the weight of the soil also contributes to the lateral bearing capacity if suction is not available. Denoting the
ultimate lateral bearing capacity factor in soil with unit weight of g0 by Np, the following relation between Np and Np0 can be
assumed for a shallow mechanism when there is a gap on the rear side:

g0z
Np ¼ Np0 þ  Npd (11)
su

where Npd is the bearing capacity factor for a flow around mechanism.
The depth at which Np ¼ Npd is therefore the onset depth for flow around failure mechanism. The effect of soil weight is
that the transition of mechanism occurs at shallower depth compared to idealised weightless soil condition. Replacing
su ¼ sum þ kz in Eq. (11), the following is obtained:

g0z ðsum =g0DÞ1

Np ¼ Np0 þ ¼ Np0 þ  Npd (12)
sum þ kz ðsum =kDÞ1 þ ðz=DÞ1

It can be seen that the transition depth is related to the normalised soil weight (sum/g0 D) and normalised strength gradient
l (i.e. sum/kD). In a uniform soil profile, the transition depth is a function of su/g0 D only. To demonstrate this, three finite
element analyses are performed, representing three su/g0 D ratios from 0.47 to 2.35. Fully rough interface and no suction
conditions are considered. The variation of Np factor with depth from FEA is illustrated in Fig. 6. It should be noted that
although the numerical overshoot is not corrected, the transition depth of soil failure mechanism is expected to be not
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 267

Fig. 6. Effect of normalised soil weight (su/g0 D) on the transition depth.

Table 1
Summary of analyses with different su/g0 D values.

Case 1 2 3
su (kPa) 10 10 10
D (m) 2.14 2.14 2.14
g0 (kN/m3) 2 4 10
su/g0 D 2.34 1.17 0.47
Observed onset depth (z/D) in FEA 6 4.5 2.5
Predicted by Ref. [15] 3.3 2.6 1.7
Predicted by Ref. [20] 6.1 4.3 2.4

influenced. As su/g0 D decreases (soil becomes heavier relative to its strength), transition to flow around mechanism occurs at
shallower depths (marked by solid dots in Fig. 6). The transition depths are also compared with the predictions of Refs. [15,20]
models, as tabulated in Table 1. It can be seen that the Murff and Hamilton model predicts generally shallower transition
depths than observed in FEA. The Yu et al. model predicts excellently. The under-prediction of the transition depth by the
Murff and Hamilton model is due to the over-estimation of Np0, as illustrated in Fig. 4.

3.4. Effect of suction

The above has mainly dealt with the ultimate lateral bearing capacity factor and transitional depth when a gap is free to
develop on the rear side of the pile. However, in certain situations, for example, in fast wave loading, or when the free
drainage path is blocked around the pile by for instance a mudmat, it may become appropriate to assume suction on the rear
side.
If suction is available on the rear side of the pile (i.e. no gap), a passive and an active wedges will be mobilised in the
shallow part. This results in higher Np factor and shallower transitional depth. Assuming isotropic soil strength, Np in the
shallow part is simply twice the Np0 factor calculated for the single (passive) wedge mechanism. It should also be noted that
when suction is assumed, the soil weight does not contribute to the lateral bearing capacity. Furthermore, it should be
checked that the mobilised suction does not exceed the cavitation pressure. For offshore piles, the cavitation pressure is equal
to the atmospheric pressure plus the hydrostatic pressure at the depth of consideration. Np can therefore be calculated this
way:

Np ¼ 2Np0  Npd (13)

where Np0 and Npd are as defined earlier and should be calculated in the same way as for no suction condition.
268 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

The transitional depth from conical wedge mechanism to localised flow around mechanism is the depth that satisfies
Np ¼ Npd.
The question of whether suction should be considered in design (or whether a gap will develop on the rear side of the pile)
is complicated. Various factors, for instance the soil strength at mudline, the loading rate, the pile installation method, the
presence of a pile head mudmat, will influence the possibility of a gap development. It is however not the intention of the
paper to address this question. The designer should perform an evaluation on the gap development based on project specific
conditions.
Referring back to the discussions in Section 3.1, in Jeanjean's recommendation, assuming suction, an Np ¼ 8 is recom-
mended at z ¼ 0, which implies Np0 ¼ 4 at z ¼ 0 for a single wedge mechanism. This is also suggested by the Nichols et al.
model. This is too high compared to Np0 z 3.5 suggested by upper bound solutions [15,20].

3.5. Effect of relative pile stiffness

In Refs. [20,21]; the reported lateral bearing capacity factor Np was calculated for a horizontally translating pile. In reality,
this is rarely the case. Depending on the pile stiffness and pile head constraint, the pile can deform in different manners. In
this section, the effect of pile stiffness is discussed. It should be noted that the pile stiffness should be discussed in the relative
sense because it depends on the soil stiffness, the pile length and the cross-sectional properties. The effect of pile rotation was
investigated by Ref. [15]. In their upper bound analyses, they studied the soil reaction encountered by a horizontally trans-
lating pile and a pile that is kinematically constrained to rotate at the pile tip. They found that the inferred ultimate lateral soil
reaction for conical wedge failure for both conditions is similar. This is also consistent with the findings by Ref. [14].
The pile stiffness influences the pile response by changing the soil mobilisation along the pile. For example, for two similar
piles (same pile length and diameter) with two different cross-sectional properties (e.g. wall thickness) and under the same
lateral load, the stiffer pile will mobilise the soil to a greater depth, while the softer pile will mobilise the soil to a shallower
depth, but to a greater mobilisation level. This is schematically illustrated by Fig. 7. However, this phenomenon can be
automatically captured when the pile is analysed in a beam-column model, where soil reaction is represented by distributed
p-y springs along the pile length. The purpose of this study is to estimate the peak values of the p-y springs to be used in such
analyses.
In the case when gapping between soil-pile interface can develop, the stiffness of the pile may have an impact of the depth
of the gap. This is because that the depth of gap development must be limited to the depth of the rotation point, as defined in
Fig. 7. Below the rotation point and on the rear side of loading direction, the pile is displacing against the soil. This seals up the
gap and allows suction to be developed on the front side of the pile because the pile is also displacing against the soil on the
front side above the rotation point. Due to the suction, the transition depth from the wedge failure to the flow around failure
could be potentially reduced compared to that determined by Eq. (11) for a flexible pile. In practical design, this can be
considered by running the beam-column analysis with the Np factors established for a horizontally translating pile (i.e.) such
as those recommended by Refs. [20,21]), which will predict a rotation point. The analysis can then be repeated by using p-y
springs with Np factor accounting for suction below the rotation point. Such iterative process can be repeated until the
assumed rotation point and calculated rotation point is close enough.

Fig. 7. Schematic illustration of effect of pile stiffness.

 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 269

When the horizontal load is applied at a certain eccentricity, it creates a rotational moment at the mudline level. The
height of the load application point is expected to have similar effects on the soil reaction to those by the relative pile stiffness.
A pile with a higher load application point is similar to a pile that is relatively softer.

3.6. A generalised recommendation

Based on the above discussions, a generalised recommendation that is self-consistent and flexible for various scenario is
attempted. The first step is to establish Np0, which is the bearing capacity factor in idealised weightless soil, with no suction.
"
 #1:35
z=D 0:6
Np0 ¼ N1  ðN1  N2 Þ 1   ð1  aÞ  Npd
d

where.

N1 ¼ 11.94
N2 ¼ 3.22
d ¼ 16.8e2.3 log10 (l)  14.5
Npd ¼ 9.14 þ 2.8a

The recommendation of Np0 is mainly based on the work presented [20,21]; with introduction of the variable d in this
paper to cover the effect of normalised strength gradient (l). The equation for the flow around capacity factor Npd is based on
the linear approximation of the results presented by Ref. [18].
For real soil with weight, if a gap is considered, the bearing capacity factor is:

g0 z
Np ¼ Np0 þ  Npd
sum þ kz

For real soil with weight, if no gap is considered, the bearing capacity factor is:

Np ¼ 2Np0  Npd

Transition from the wedge mechanism to the flow around mechanism occurs at depth that satisfies Np ¼ Npd. For the case
when gapping between pile-soil interface is assumed, it should be checked whether reduced transition depth is possible due
to pile rotation, as discussed in Section 3.5.

4. Comparison of recommendations

In order to illustrate the similarities/differences of the currently proposed approach with respect to other recommended
approaches in the literature and industry guideline, the predicted Np profile of a slender pile in two soil profiles are compared:
1) a lightly over-consolidated clay profile, with su ¼ 5.1 þ 2.4z (kPa); 2) a heavily over-consolidated clay profile, with a
constant shear strength of 150 kPa. For both cases, the pile diameter is 2.14 m and g0 ¼ 8 kN/m3. Note that in the comparisons
shown below, the API J factor is 0.5 and 0.25 for the first and second profile respectively. Note that in the comparisons below,
the potential pile stiffness effect on the transition depth from the wedge failure to the flow around failure as discussed in
Section 3.5 is not considered.

4.1. Comparison in lightly over-consolidated clay

Comparison of calculated Np profiles using [1,8,15,16] and currently proposed methods are shown in Fig. 8 for the lightly
over-consolidated soil profile. Both without/with suction conditions are compared. For the no suction condition (i.e. gapping
on the rear side of the pile), the soil weight contribution is included in the calculated Np. A fully rough interface (i.e. a ¼ 1) is
considered for comparison as methods proposed for axial skin friction calculation will typically suggest an a value close to 1
for the soil condition examined here.
It can be seen that API predicts the lowest Np factor at all depths than the other methods, since a limiting Np factor of 9 is
adopted, regardless of interface roughness. For the no suction condition, the other three methods, namely Murff and Ham-
ilton, Nichols et al. and the currently proposed method, all have a limiting Np about 12. The Murff and Hamilton predicts the
shallowest transition depth to flow around mechanism, whereas the currently proposed method predicts the deepest
transition depth. Coincidently, the API predicts similar transition depth to the currently proposed method for the case
considered here. Since Jeanjean's model assumes full suction (i.e. no gapping), it is therefore not appropriate to include it in
the comparison for no suction condition. Similarly the Nichols et al. method is not included in the comparison for with suction
condition.
270 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

Fig. 8. Comparison in lightly over-consolidated soil profile.

When suction is assumed (i.e. no gapping), due to inclusion of an active wedge on the rear side, the transition from wedge
failure to localised flow around failure occurs at a much shallower depth than when no suction is assumed. The currently
proposed method and Murff and Hamilton [15] method suggest a steeper increase of Np with depth, and therefore a shallower
onset depth for flow around mechanism than the Jeanjean method. It should be noted that the API recommendation is in-
dependent of the assumption on suction condition.

4.2. Comparison in heavily over-consolidated clay

Fig. 9 presents similar comparisons in the heavily over-consolidated soil profile. Only the no suction condition is
compared. For a smooth interface, Murff and Hamilton; API and the currently proposed methods all have a similar limiting Np
factor about 9. However, the API methods predicts the deepest onset depth for flow around mechanism, therefore the lowest
Np profile prior to the onset depth. The currently proposed method predicts a deeper onset depth than the Murff and
Hamilton method, as expected. For a rough interface, the Murff and Hamilton and the currently proposed methods have
similar limiting Np factor of 12 whereas the API method adopts a limiting factor of 9. Again, the API method predicts the
deepest onset depth for flow around mechanism. In reality, in a heavily consolidated soil profile, the interface roughness is
likely to be neither fully rough nor fully smooth. Note that the Jeanjean method, and the Nichols et al. method are proposed
for normally consolidated or lightly over-consolidated clays; therefore, they are not included in Fig. 9.

5. Effect of axial loading on lateral resistance

5.1. Current practice

A pile loaded laterally can often also be subjected to axial loading. In the industry design practice, the interaction between
axial and lateral loading are usually not explicitly considered. Instead, piles are designed with separate verifications for axial
and lateral loading. A rational approach to account for such interaction is therefore needed.

5.2. A simple concept for considering the interaction

A simple and practical approach to consider the interaction in design might be utilizing the interface roughness mobi-
lisation concept illustrated in Fig. 10. The resultant of interface skin friction (or roughness) mobilisation for axial and lateral
loading is limited by the maximum available interface shear stress (or roughness), which forms the circular envelope.
Therefore, if the interface roughness mobilisation needed for axial loading is known, the allowable interface roughness for
lateral loading can be determined. This roughness can then be used to estimate the limiting lateral bearing capacity, by
interpolating between fully smooth and fully rough conditions established above.

5.3. Verification

In order to demonstrate the concept, two finite element analyses in Plaxis 3D are performed. The soil is homogeneous in
strength and suction is assumed. Prior to the lateral loading phase, an axial load is applied. In both cases, fully rough interface
is used. In Case 1, the applied axial tension loading corresponds to 100% of the theoretical axial capacity (i.e. axial loading fully
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 271

Fig. 9. Comparison in heavily over-consolidated profile (assume no suction).

Fig. 10. Axial and lateral interaction envelope.

Fig. 11. Load-displacement responses from Case 1 and Case 2.

utilises the available interface roughness). In Case 2, the applied axial tension loading corresponds to 50% of theoretical axial
capacity (i.e. axial loading utilises half of the available interface roughness). The pile head lateral load-displacement from
these two analyses are illustrated in Fig. 11. According to the assumption illustrated in Fig. 10, the available roughness for
lateral loading is zero (i.e. fully smooth) in Case 1. Indeed, the pile load-displacement curve from this analysis does match the
272 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

response from a fully smooth pile without axial loading. For Case 2, the available roughness for lateral loading is estimated to
be 0.87. The load-displacement response of this Case is compared with the response of a full rough pile without axial loading,
which illustrates only marginal difference on the capacity due to the small difference in available roughness for lateral
loading.
As discussed in Section 3.1 that the API guidelines imply assumption of a fully smooth interface for lateral capacity
calculation (with a limiting Np factor 9), regardless of actual pile-soil roughness. This might be a too conservative provision for
accounting for the effect of axial mobilisation of interface roughness. However, the approach illustrated in Fig. 9 does offer a
rational method for more detailed considerations. As illustrated by Case 2, there is still considerable portion of interface
roughness available for lateral loading even under moderate level of axial mobilisation (50% in the example).
It is worth noting that the method to determine the maximum available interface roughness is not within the scope of this
paper. However, there are abundant references in the literature covering this, including the effect of set-up, aging, etc.
[1,7,9e11].

6. Effect of the soil strength anisotropy on ultimate lateral bearing capacity

6.1. General

It should be noted that all the discussions in previous sections of the paper are limited to the assumption of isotropic
material. However, it is well known that clay exhibits different undrained shear strengths, depending on the shearing mode/
direction. While undrained shear strengths measured in different shear modes and therefore information on soil strength
anisotropy are typically available in soil reports for design of offshore piles nowadays, methodology to utilise the information
in actual design is not readily available. The question faced by design engineers is how to account for shear strength
anisotropy, or in other words, what shear strength should be used to compute the ultimate lateral bearing capacity while
using the solutions derived for isotropic material? In this section, finite element analyses have been performed using a soil
model that is capable of describing the strength anisotropy of the soil. Two sets of typical anisotropic ratios are considered.
The purpose is to examine the effect of strength anisotropy and to provide guidance on how the strength anisotropy could be
considered in practical design.

6.2. Parametric range

Two sets of shear strength anisotropy ratios have been studied, as summarized in Table 2. The two sets of anisotropy ratios
considered here are according to Andersen et al. [2]. The first set represents statistical mean values measured on typical
offshore quality samples of soft, slightly overconsolidated clays. The second set, which has lower anisotropy strength ratios, is
the average value obtained from high quality block/piston samples for Norwegian onshore soft clays. Because sample
disturbance has a greater impact on sCu than sDSS
u and sEu [12], lower anisotropy ratios are obtained from soil samples of higher
quality. Both set of anisotropy ratios imply average shear strength (sAV C E DSS
u , defined as an average of su, su and su ) similar to DSS
DSS
shear strength (su ). For simplicity, homogenous soil strength, i.e., no variation with depth, is considered. The soil is idealised
as weightless. For each set of anisotropy ratios, two analyses are performed, with one assuming no suction and the other
assuming suction.
To simulate the soil anisotropic strength, the NGI-ADP soil model [6] is used. The input to the soil model is undrained shear
strength measured in triaxial compression mode (sCu) as well as the anisotropic ratios (sEu/sCuand sDSS C
u /su). A Tresca type yield
criteria is followed. For comparison purpose, parallel analyses are also performed using isotropic Tresca soil model.

6.3. Effect of strength anisotropy - with suction

When suction is assumed, active and passive wedges are mobilised in the shallow part and localised flow around
mechanism is formed at a small normalised depth. In Fig. 12, the bearing capacity factor Np (defined as p/su) calculated from
isotropic soil (Tresca model) and anisotropic soil (NGI-ADP model) are compared. For the anisotropic soil, sDSS u is used for
normalisation to derive Np. It should be noted that there is no need to differentiate Np0 and Np when suction is assumed since
soil weight does not contribute to the lateral capacity.
The results shown in Fig. 12 are for Set 1 of anisotropy ratios listed in Table 2. It can be seen that by normalising the lateral
pressure with sDSS
u , almost identical Np profiles with depth are obtained from two comparative analyses. Similarly good match

Table 2
Summary of anisotropy ratios considered.

sEu/sCu sDSS C
u /su sAv C
u /su

Set 1 0.62 0.80 0.806

Set 2 0.46 0.72 0.727
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 273

Fig. 12. Effect of strength anisotropy on lateral bearing capacity factor (assuming suction).

is obtained for Set 2 of anisotropic ratios, and is therefore not presented. Note that the results presented in Fig. 12 have not
been corrected for the numerical overshooting. The paragraphs below attempt to explain the phenomenon.
As implied by the Np profile with depth in Fig. 12, as well as revealed by soil displacement pattern, localised flow around
mechanism is formed from a depth around 2D, for both smooth and rough interface conditions. In a flow around mechanism,
there is no soil elements that undergo triaxial compression/extension shearing. Instead, the failure pressure is controlled by
the shear stress along the slip surface. The shearing mode is similar to the DSS mode, though the shearing plane is vertical,
instead of horizontal, which is the case in conventional laboratory DSS tests. This explains why a good match is obtained when
the lateral pressure is normalised by sDSS
u .
Due to assumption of suction, both active and passive conical wedges develop in the shallow part. It is a mix of triaxial
extension and DSS shearing modes in the passive wedge whilst a mix of triaxial compression and DSS shearing modes in the
active wedge. It can therefore be expected that an average shear strength of sCu, sEu and sDSSu may be a good choice for nor-
malisation. For the two sets of soil anisotropy ratios considered in this study, the average shear strength sAv DSS
u is close to su .
Indeed, it is shown in Fig. 12 that an excellent match is obtained in the shallow part by normalising the bearing pressure with
sDSS
u .

6.4. Effect of strength anisotropy - without suction

As the soil is assumed to be weightless and no suction condition is assumed, flow around mechanism is not obtained for
the cases investigated herein. Only the passive wedge is mobilised. It is therefore expected that some sort of average strength
between sDSSu and sEu could be used for normalisation. Previous study by Ref. [16] recommended to use su ¼ 0.5(sDSS u þ sEu). The
Plaxis results presented below provide further basis for recommendation.
The Np0 factor calculated in weightless, anisotropic soil is compared with those in weightless, isotropic soil for the con-
dition of no suction in Fig. 13. The comparison is for Set 2 of anisotropy ratios, but similar results are obtained for Set 1. In
Fig. 13, the results in anisotropic soil are presented for two normalisations. The first one, as suggested by Ref. [16]; uses the
mean value of sDSSu and sEu, i.e., su ¼ 0.5sDSS
u þ 0.5sEu. As shown in Fig. 13, the resulting Np0 factor calculated this way is higher
than those obtained in isotropic soil. The second one uses a weighted average of sDSS u and sEu, with su ¼ 0.75sDSS E
u þ 0.25su. It can
be seen that excellent match with the Np0 profile in isotropic soil is then obtained for both smooth and rough interface
conditions.

6.5. Summary on effect of strength anisotropy

In order to apply the bearing capacity factors obtained in idealised isotropic material to calculate the ultimate lateral
bearing pressure in anisotropic soils in reality, appropriate choices of shear strengths should be made. Based on results of this
study, the following is recommended:
If suction is assumed or evaluated to sustain, in the shallow part where double conical wedge failure mechanism governs,
an average strength of sCu, sEu and sDSS
u should be used. Since sDSS
u is normally a good measure of average shear strength, sDSS
u
could be used for simplicity. However, if suction is not assumed or evaluated not to sustain, in the shallow part where single
(passive) wedge failure mechanism governs, a weighted average su calculated as 0.75 sDSS u þ 0.25sEu should be used. After
DSS
transition from wedge to localised flow around mechanism, su strength should be used, regardless of assumption on
suction.
274 Y. Zhang et al. / Marine Structures 50 (2016) 260e275

Fig. 13. Effect of strength anisotropy on lateral bearing capacity factor (assuming no suction).

7. Conclusions

This paper re-visits the topic of the ultimate lateral bearing capacity of piles in clay. Three important questions that are
faced by design engineers, namely: 1) what is the limiting lateral bearing pressure and its variation with depth, 2) what is the
effect of axial loading on lateral capacity and 3) how to account for strength anisotropy effect, are investigated and an attempt
is made to provide guidance on these aspects. To summarise this paper, the recommendations are presented in Table 3.
It should be noted that this paper has only examined a single clay layer with constant or linearly increasing undrained
shear strength. Other studies, for example by Refs. [15,20,21]; show that soil layering can have a significant impact on the
variation of pu with depth, particularly when the contrast between the soil layers is large. The recommendation presented in
Table 3 should therefore be treated with caution when applied to layered soil profiles. Site specific solutions, such as by means
of finite element analyses, may become necessary depending on the magnitude of layering and significance of lateral capacity
in relation to the overall design.
It should be further noted that this study is performed with the application in mind for typical offshore slender piles with
large length to diameter ratios, such as those used for supporting offshore jacket platforms and for anchoring offshore floating
facilities. The recommendations, as summarized in Table 3, are therefore strictly only applicable to the slender piles. Large
diameter monopiles, which are popular foundation solutions for offshore wind turbines structures, have diameters in the
order of 5e10 m and length to diameter ratio as low as 3 to 6. To capture the response of the monopiles fully, additional
modelling aspects, such as the tip resistance (both in shear and rotation) and side shear skin friction along the pile length due
to pile rotation, are important and need to be included, as demonstrated in Byrne et al. [3]. For slender piles, these effects are
negligible and are therefore typically not considered.

Table 3
Summary of recommendations.

Np0 
0:6 1:35

(Idealised weightless soil, no Np0 ¼ N1  ðN1  N2 Þ 1  z=D
d
 ð1  aÞ  Npd
suction) N1 ¼ 11.94
N2 ¼ 3.22
d ¼ 16.8e2.3 log10 (l)  14.5
l ¼ sum/kD
Npd ¼ 9.14 þ 2.8a
Np Np ¼ Np0 þ s gþkz
0z
 Npd
um
(real soil with g0 , with gap)
Np Np ¼ 2Np0  Npd
(real soil with g0 , without gap)
Onset depth for flow around Transition from wedge mechanism to flow around mechanism occurs at depth that satisfies Np ¼ Npd. When gapping is
mechanism considered, it should be checked whether reduced transition depth is possible due to pile rotation, as discussed in
Section 3.5.
Ultimate lateral bearing pu ¼ Npsu
pressure su ¼ sum þ kz
Effect of axial loading Interface roughness mobilisation concept
Effect of strength anisotropy Single wedge (no suction): su ¼ 0.75sDSS
u þ 0.25sEu
Double wedge (with suction): su ¼ sDSS
u ,
assuming sDSS
u ¼ sAv
u
Flow around: su ¼ sDSS u
 Y. Zhang et al. / Marine Structures 50 (2016) 260e275 275

Acknowledgements

The authors wish to gratefully acknowledge the financial support received through a NGI internal research project, which
is partially funded by the Norwegian Research Council.

References

[1] American Petroleum Institute. API Recommended Practice 2A-WSD Planning, designing, and constructing fixed offshore platforms-working stress
design, API recommended practice 2A-WSD. twenty-second ed. 2014.
[2] Andersen KH, Lunne T, Kvalstad TJ, Forsberg CF. Deep water geotechnical engineering. In: Proc., XXIV Nat. Conf. of the mexican Soc. of soil mechanics.
Mexico: Aguascalientes; 2008.
[3] Byrne BW, McAdam R, Burd HJ, Houlsby GT, Martin CM, Zdravkovíc L, et al. New design methods for large diameter piles under lateral loading for
offshore wind applications. In: Meyer V, editor. 3rd internaltion symposium on frontiers in offshore geotechnics (ISFOG). Oslo, Norway: Taylor &
Francis Group; 2015. p. 705e10.
[4] Georgiadis K, Georgiadis M. Undrianed lateral pile response in sloping ground. J Geotechnical Geoenvironmental Eng 2010;136(11):1489e500.
[5] Georgiadis K. Variation of limiting lateral soil pressure with depth for pile rows in clay. Comput Geotechnics 2014;62(2014):164e74.
[6] Grimstad G, Andresen L, Jostad HP. NGI-ADP: anisotropic shear strength model for clay. Int J Numer Anal Methods Geomechnics 2012;36(4):483e97.
[7] Jardine RJ, Chow FC. New design methods for offshore piles. The Marine Technology Directorate Ltd; 1996.
[8] Jeanjean P. Re-assessment of p-y curves for soft clays from centrifuge testing and fintie element modeling. Houston, Texas, OTC 20158: Offshore
Technology Conference; 2009.
[9] Karlsrud K, Clausen CJF, Aas PM. Bearing capacity of driven piles in clay, the NGI approach. In: Gourvenec S, Cassidy MJ, editors. 1st internation
symposium on frontiers in offshore geotechnics (ISFOG). London, Perth, Australia: Taylor & Francis Group; 2005. p. 775e82.
[10] Karlsrud K. Prediction of load-displacement behaviour and capacity of axially loaded piles in clay based on analyses and interpretation of pile load test
results. PhD thesis. Norwegian University of Sicence and Technology; 2012.
[11] Kolk HJ, der Velde E. A reliable method to determine friction capacity of piles driven into clays. Houston, Texas, OTC 7993: Offshore Technology
Conference; 1996.
[12] Lunne T, Berre T, Andersen KH, Strandvik S, Sjursen M. Effects of sample disturbance and consolidation procedures on measured shear strength of soft
marine Norwegian clays. Can Geotech Jouranl 2006;43(7):726e50.
[13] Martin C, Randolph MF. Upper-bound analysis of lateral pile capacity in cohesive soil. Geotechnique 2006;56(2):141e5.
[14] Matlock H. Correlations for design of laterally loaded piles in soft clay. Houston, Texas, OTC 1204: Offshore Technology Conference; 1970.
[15] Murff JD, Hamilton JM. P-ultimate for undrained analysis of laterally loaded piles. J Geotechnical Eng 1993;119(1):91e107.
[16] Nichols NW, Rohani MJ, Mukherjee K, Ayob MSB, Clausen CJF, Lunne T. Effect of lateral soil strength and stiffness on jacket foundation integrity and
design for south China sea sites. Kuala Lumpur, Malaysia, OTC-24842-MS: Offshore Technology Conference Asia; 2014.
[17] Plaxis. Plaxis 3D reference manual 2013. 2013.
[18] Randolph MF, Houlsby GT. The limiting pressure on a circular pile loaded laterally in cohesive soil. Geotechnique 1984;34(4):613e23.
[19] Reese LC, Cox WE, Koop FD. Field testing and analysis of laterally loaded piles in stiff clay. Houston, Texas, OTC 2312: Offshore Technology Conference;
1975.
[20] Yu J, Huang M, Zhang C. Three-dimensional upper-bound analysis for ultimate bearing capacity of laterally loaded rigid pile in undrained clay. Can
Geotechnical J 2015a;52(11):1775e90.
[21] Yu J, Huang M, Zhang C. Ultimate lateral resistance of laterally loaded piles in undrained clay. In: Meyer V, editor. 3rd international symposium on
frontiers in offshore geotechnics (ISFOG). Oslo, Norway: Taylor & Francis Group; 2015b. p. 661e6.
[22] Zhang C, White D, Randolph MF. Centrifuge modeling of the cyclic lateral response of a rigid pile in soft clay. J Geotechnical Geoenvironmental Eng
2011;137(7):717e29.