OTC 20158

Re-Assessment of P-Y Curves for Soft Clays from Centrifuge Testing and
Finite Element Modeling
P. Jeanjean, BP America Inc.

Copyright 2009, Offshore Technology Conference

This paper was prepared for presentation at the 2009 Offshore Technology Conference held in Houston, Texas, USA, 4–7 May 2009.

This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any position of the Offshore Technology Conference, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Offshore Technology Conference is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.

Abstract
The paper focuses on the results of centrifuge testing on a laterally loaded conductor. The conductor was installed to
simulate jetted conditions and lateral transfer curves (so called P-Y curves) were generated for monotonic and cyclic
conditions. The influence of a previous set of cyclic loads on the subsequent behavior of the pile when subjected to
additional cyclic loads was investigated. The paper focuses on the determination of P-Y curves tangent and secant moduli
since there are the key input parameters for structural software in determining the lateral displacements of the conductor and
therefore the cyclic fatigue-generating stresses in the conductor. Results show that the P-Y curves as recommended by API
RP2A (2000) can be conservative, with moduli and ultimate unit pressures less than what test data and theoretical modeling
suggest. The centrifuge tests validated theoretical curves derived via Finite Element Analyses (FEA) and described by
Templeton (2009). Equations are proposed to generate the backbone P-Y curves and to calculate the secant modulus to the
P-Y curves to analyze the cyclic loading of interest to this paper.

Introduction

Roadmap for the paper

The paper is divided into the following parts:

• It first describes the problem of interest and defines the jetted conductor loading conditions that were studied.
• It then describes the theoretical Finite Element Analyses (FEA) that were performed and the obtained P-Y curves,
• After which the centrifuge tests performed to calibrate the FEA derived P-Y curves are detailed.
• An extensive comparison of the P-Y curves from the FEA analyses and the P-Y curves is presented,
• And new equations to derived backbone P-Y curves for soft clays are proposed.
• Last, conclusions from centrifuge pile head load displacement curves are listed, along with implications for pile
design.

Loads on jetted conductor for top tension risers

The study work reported herein focuses on the evaluation of lateral small-displacement soil-structure interaction for jetted
conductor fatigue analysis. The conductor is assumed to be connected to a floating structure via a top-tensioned production
riser. As the riser responds to the platform motions and to the sea currents, it imparts loads on the well head and the
conductor. These loads are being applied simultaneously but, as a simplification, these loads are assumed to be applied
independently, when calculating fatigue life.. A large fatigue safety factor covers this non-conservative assumption. The
loads, which are assumed to be collinear, that induce fatigue damage in the conductor can be divided into three categories:
Vessel Motions (VM), Hull Vortex Induced Hull Motion (VIM), Riser Vortex Induced Vibration (VIV) (Fig. 1):

• Vessel Motions: The platform, or vessel, motions are separated in motions caused by the wind and waves first.
These are the "vessel motions or VM". A finite element model of the hull, the mooring the risers and well
foundations are subjected to various sea and wind conditions for 3 hours. From these 3 hours simulations, the
fatigue rate of the connector is calculated. The total VM fatigue rate is the metocean condition frequency weighted
sum of these fatigue rates.
2 OTC 20158

• Hull VIM: The vessel hull will oscillate in certain current conditions and produce "hull Vortex Induced Motions” or
“hull VIM". Physical model of the hull and its mooring have been towed at various speed and orientations to
determine the anticipated amplitude and period of these oscillations. These oscillations for the various expected
current speeds and directions have been used to derive a vessel finite element model and the fatigue rate have been
combined with the occurrence frequencies to generate fatigue rates.

• Riser (VIV): The riser, when subjected to current, will VIV in the non-strake section. The normal modes of the riser
and its foundation where determined with a finite element model of the riser and its foundation. (Beynet, 2008)

For the study case, the vessel motions (VM), Hull VIM, and the riser VIV caused 20%, 32%, 48% of the total fatigue damage
respectively.

Wind and waves forces

cause vessel motions

Some current conditions

Hull cause hull VIM

Riser

Wellhead
Current forces
cause riser VIV

Conductor
Drawing not to scale
Fig. 1: Schematic of types of loads on conductor for a top tensioned riser.

Pile cyclic stresses and fatigue damage

The key point of the soil-structure analysis of the conductor is to determine the cyclic stresses in the conductor caused by the
above loads. It is the cyclic stress range that is of interest and not the absolute stress level. Indeed, fatigue failures in metals
and other materials are mechanistically driven by the accumulation of damage to the materials as characterized by the
Palmegrin-Miner Damage Rule. This rule relates the number of cycles at a specific stress range and associated plastic strain
range to the total life at this same stress and strain range. Damage, as measured by the amount of unrecoverable string or
plastic work on the materials creates voids or "free surfaces" internally that grow to a crack (crack initiation) under reversed
plastic strain. Once the crack is large and follows continuum mechanics theory (not submicrostructural) the crack now grown
by the continued cycle driving force of the stress or strain field to failure. Loading of the materials to a maximum load only
will not create this failure mechanism. Sometimes failures of static loading are misnomer as "static fatigue" which refer to
other failure modes such as creep or environmentally assisted crack (Burk, 2007).

Most software used for riser analysis are unfortunately limited in the manner with which they can model complex soil
structure interaction. Most of them are limited to linear and non-linear elastic springs. These springs are usually derived
through the method recommended by API RP2A (2000) which is based on the work of Matlock (1962). The recommended
method in API 2000 has remained unchanged since the 3rd Edition of RP2A in 1973.

There exists however sufficient data (i.e. Reese and Cox, 1971, Stevens and Audibert, 1979, Murff and Hamilton, 1993,
Randolph and Houlsby, 1984) to suggest that the API (2000) method for soft normally consolidated clays, will produce
springs that underestimate the ultimate unit pressure that can be generated at the face of the pile. It was also suspected that
the API spring were also too soft.
OTC 20158 3

Because stiffer P-Y curves will produce, for a given load range, less cyclic lateral displacement of the pile than predicted by
API, the cyclic stresses in the pile will be reduced and the fatigue life will therefore be increased. A Finite Element Analysis
(FEA) of a laterally loaded pile was performed for Gulf of Mexico soil and indeed generated P-Y curves stiffer than those of
API and with higher ultimate unit pressure. Details of the FEA analyses can be found in Templeton (2009).

The FEA curves were to be used, somewhat arbitrarily, as follows by the structural software to calculate the cyclic soil
displacements (and therefore the cyclic stresses in the pile): for a given load cycle (P1 to P2 in Fig. 2), the soil-structure
interaction was modeled as a linear spring that had a stiffness equal to the tangent modulus of the P-Y curve taken at the
mean load of the load cycle (see Fig. 2).

Modulus used to calculate cyclic

displacement FEA derived
PY curve

P2
(P1+P2)/2
P1
Pressure, P

API derived
PY curve

Cyclic
Displacement Cyclic Displacement
From FEA From API Curve
Curve

Lateral Displacement, Y
Fig. 2: Calculation of soil-structure spring stiffness in riser analysis program

The above methodology gave stresses in the conductor that were lower than if standard API curves had been used, and
therefore increased the fatigue life of the conductor. However, because these curves had not been calibrated, a series of
centrifuge tests was performed to verify the theoretical curves with experimental data. In a Class A type of prediction, the
FEA was performed again, but with soil properties closely matching those of the clay in the centrifuge rather than Gulf of
Mexico clays and P-Y curves for kaolin were generated.

This paper focuses on the comparison of the FEA and centrifuge curves for kaolin, because the comparison validates the
method used to generate the P-Y curves for Gulf of Mexico clays. The paper also confirms that the method implemented in
the structural analysis program to calculate cyclic stresses (see Fig. 2) is conservative for the cases of interest herein.
Conclusions on cyclic modulus degradation are also included.

FEA modeling of a laterally loaded conductor

Geometry and soil model

A Finite Element Analysis (FEA) of the lateral performance of the conductor was performed to generate P-Y curves for a
kaolin clay. Details of the analysis are given in Templeton (2009) and salient points are summarized herein. A free-head
0.91m (3ft) diameter, 36.5m (120ft) long, 50.8mm (2 in.) wall thickness well conductor in clay soil was modeled with the soil
properties closely matching those of the centrifuge tests. A 3D non linear total stress analysis with the ABAQUS program
was performed using an elastic-plastic, work hardening model with Mises yield for the soil behavior. Soil properties were
based on direct simple shear and resonant column test data for kaolin. The static loading applied to the conductor was a
lateral load with no moment restraint at the load application point. This applied lateral load was increased until very large
lateral displacements were achieved.

The basic geometry of the model used is shown on Fig. 3. Symmetry conditions permitted use of a half-space model (180
degrees about the conductor axis). The model included finite elements out to a diameter of 36.6m (120 ft) or 40 times the
conductor diameter and infinite elements beyond that diameter. The solid continuum elements used to represent the soil were
8-node brick elements. The conductor was modeled with 4-node shell elements, 50.8mm (2 in.) thick, using an elastic-
perfectly plastic behavior and properties of 414 MPa (60 ksi) steel. The conductor was modeled down to a depth of 61m (200
ft), at which depth its displacement was fixed.
4 OTC 20158

The soil was modeled as an elastic-plastic, work hardening material with Mises yield. The shear strength, as referenced to a
Direct Simple Shear (DSS) tests was taken as the profile measured in the centrifuge tests (see Fig. 8). This shear strength
profile was corrected for field conditions using the following factor:
⎛ T × 0.1 ⎞
f Su = 1 + 0.1× Log10 ⎜⎜ DSS ⎟⎟ ..................................................................................................................................... (1)
⎝ TRC ⎠
Where
• TDSS is the time period for peak shear strain, or 279 minutes for the Kaolin data
• TRC is the time period for peak shear strain for an assumed resonance test, or 3 seconds for the data set herein.

Fig. 3: Finite Element Mesh For the modeling of the centrifuge tests.

⎛ TDSS × 0.1 ⎞
⎜⎜ ⎟⎟
The factor 0.1 in ⎝ ⎠ in Eqn. 1 corresponds to the approximate fraction (10%) of the strength estimated to be
T RC

mobilized under small-displacement load cycles that cause most of the fatigue damage. The second 0.1 factor in front of the
Log10 term in Eqn. 1 accounts for a rate of increase of shear strength of 10% per log cycle of strain rate.

Therefore, for each soil layer the input rate-adjusted shear strength, Su, was calculated as:

Su = Su DSS ⋅ f Su ................................................................................................................................................................(1a)

With:
• Su DSS : shear strength profile obtained in the centrifuge and shown in Fig. 8.
• f Su : calculated as per Eqn. (1) is equal to 1.27, for the case studied herein.

So the shear strength, as measured in a DSS test is increased by 27% to provide the input soil profile to the FEM analysis.

The maximum shear modulus Gmax was taken as 550 times the shear strength, Su. The stress strain behavior was taken as
elastic with slope given by this Gmax, up to the initial yield point, which was taken as 10 percent of the ultimate strength.
Beyond the initial yield point, the stress strain behavior was given at each stress by additive combination of the elastic strain
and a work hardening plastic strain based on the particular ultimate strength and the plastic part of the normalized stress
strain curve of Fig. 4.

P-Y curves were then extracted at 8 depths and verified with centrifuge testing.
OTC 20158 5

1.4

1.2

τ /th/thmax
Stress, τmax
0.8

stress,
shear Shear
DSS test
1_44b_NF
0.6 elastic comp
Normalized
Elastic Component
linear piece fit to data
Linear fit to data
Normalized

0.4

0.2

0
-1.00 1.00 3.00 5.00 7.00 9.00 11.00 13.00 15.00

strain, γ (%)
Shear Strain, γ, %
Shear
Fig. 4: Comparison of DSS test result and elasto-plastic strain hardening model used in FEA.

Centrifuge Tests

Model dimensions and test set-up

The free-head conductor analyzed had a prototype outside diameter of 0.91m (3ft), and a 50.8mm (2 in.) wall thickness. In
the centrifuge the length of the model conductor was limited by the depth of the sample container strongboxes. The total
embedded model conductor length was 421mm, which was the maximum that could be accommodated within the existing
500mm deep test container (maximum depth of clay 450mm). The conductor tip was simply resting on the clay tub bottom.
Four tests were run in a single tub container.

A scale factor of 1:48 was achieved by carrying out these tests at 48 gravities (g) in C-CORE’s large geotechnical centrifuge.
The embedded prototype length of the conductor was 20.2m (66ft). The model conductor had a resulting diameter of
19.05mm and a wall thickness of 1.22 mm of steel and an additional 0.35 mm from heat shrink tubing added to the exterior of
the conductor to waterproof the strain gauges. The section modulus (EI) for the model was chosen as representative of the
prototype conductor. Fig. 5 shows the model conductor with the string of strain gauges.

Load application point

Top strain gauge Bottom strain gauge

Fig. 5: Model conductor used with its instrumentation

In prototype units, the conductor was to be cyclically loaded in the 0.1 to 1 Hz range. In the model, a loading frequency of 1
Hz was used. Ideally, at 1:48 scale, the cycling should have been conducted between 4.8 and 48 Hz, but this would have
introduced inertial effects into the model tests which are not present in the prototype.

Instrumentation:

Each conductor model was instrumented with 13 sets of strain gauges (Wheatstone bridge configuration) to measure the
strains at discrete points along the conductor as well as instrumentation to measure the load, inclination, and lateral
movement of the wellhead. The strain gauges were spread out along most of the length of the conductor with a higher
concentration of gauges between about 6.1m and 12.2m (20 and 40 feet) below the mudline where the moments were
expected to be the greatest.
6 OTC 20158

Druck PDCR 81 differential miniature pore pressure transducers (PPTs) were installed in the soil sample through ports in the
sidewall of the strongbox to monitor pore pressure dissipation in the soil tub. In addition to the PPTs installed in the soil,
consolidation was also monitored at the soil surface.

The strain gauges used on the conductor models were also full Wheatstone bridge bending pairs. The strain gauge locations
were chosen after consideration of the shear force profile, bending moment profiles and those from previous testing
programs. Thirteen 13 gauge levels were installed with the layout shown on Fig. 6. In test 1, 2 gauges failed during soil
consolidation and 4 failed in test 2. Fortunately the location of these non-functioning gauges was such that they did not
impact the resolution of the bending moment in the depth of interest (top 10m). A blown fuse in Test 3 prevented all gauges
from acquiring bending moments data. All gauges functioned for test 4.

Conductor Installation:

The centrifuge tests aimed at developing an installation method for the model pile that would be appropriate for the study of
jetted conductors. It was believed that pushing the pile into virgin soil could potentially produce lateral stresses that would
be higher than those along a jetted conductor. The conductor model was pushed (closed ended) into slightly undersized
(15.87 mm diameter hole vs 19.05 mm conductor diameter) pre-augered hole prior to each test. The conductor was pushed
into the soil such that the point where the lateral load was applied was approximately 91 mm above the mudline, or for the
prototype about 4.3m (14 feet) above the mudline as shown in Figure 7.
Test 1 Test 2 Test 3 Test 4
-5 -5 -5 -5
Pile Head

x
x
0 0 0 0 Mudline
x
Depth below mudline (m)

x
5 x 5 5 x 5
x
x Location of
x strain gauge
x
10 10 10 10 indicates non-working
x x
strain gauge
x x x

15 15 x 15 x 15

x x

20 20 20 20 Bottom of
Conductor

Fig. 6: Overview of strain gauges layout

Fig. 7: View of pre-drilled undersized hole and (b) conductor after installation.
OTC 20158 7

Shear Strength

The soil used in this test series was fine Alwhite kaolin processed by English China Clays, Lovering Pochrin & Co. Ltd.
(ECLP) of Cornwall, England, and purchased through their US distributor Hamill & Gilespie. Alwhite was previously known
as speswhite kaolin. The clay tested was designed to be lightly over consolidated to depth. Properties of the Alwhite kaolin
are summarized in Table. 1 (from C-CORE, 2005)

Table 1: Summary of Alwhite Kaolin Properties (from C-CORE, 2005)

The construction method for the kaolin cake was in two lifts, separated by a 5 mm thick sand drainage layer to accelerate
consolidation of the kaolin cake. This drainage layer was approximately 215 mm below the final clay surface. Clearance
holes in the sand layer were placed at predesigned locations to accommodate conductor and PCPT tests. The clay was
reconstituted from slurry and mixed at approximately 120% water content (twice the liquid limit) for a minimum of 3 hours
under a vacuum of approximately one-half an atmosphere. In order to reduce consolidation time in the centrifuge, and
because of the large starting slurry height, the clay sample was pre-consolidated to about 95% of its effective vertical stress
profile prior to the centrifuge test. The pre-consolidation was conducted in two stages: (1) consolidation under a uniform
effective stress, and (2) consolidation under a downward hydraulic gradient (DHG).

Because of the need to keep the pile head above the water level to preserve the integrity of the strain gauges connection, little
water could be added above the soil bed. The PCPT cones had to be kept saturated by being partially inserted into the
sediments. A PCPT test was performed before each test and the interpreted shear strength profiles are shown on Fig. 8. The
shear strength measurements near the surface show scatter and the interpreted profile was deduced from SHANSEP
calculations. The PCPT profiles also show a sharp increase around the depth of the intermediate sand layer. It is believed
that the PCPT measurements were affected by the presence of the sand drainage layer. The results and P-Y curves presented
herein are limited to the depths above the sand drainage layer and therefore were not affected by the shear strength “bump” at
11m below surface.

Tests Objectives:

Four tests were performed in a single tub. Each test was designed to meet a specific set of loading conditions that included
extreme events as well as events that cause the most fatigue damage to the conductor. The loading sequence for Tests 2, 3,
and 4 is shown on Figures 9, 10, and 11. The purpose and loading sequence for each test is provided below:
• Test 1: The first test was a reference test. The pile was monotonically loaded until the pile head moved just over 1
conductor diameter (914mm = 36 inches). Results from the remaining 3 tests were compared to the results of this
test.
8 OTC 20158

0
test 1
−2 test 2
test 3

Prototype Depth below Mudline (m)

−4 test 4
−6 Interpreted Su for FEA input and data analysis
depth of strain gauges
−8

− 10

− 12

− 14

− 16

− 18

− 20
0 10 20 30 40

Undrained Shear Strength, Su (kPa)

Fig. 8: PCPT records and interpreted shear strength

• Test 2: The objective of this test was to understand the impact of large vessel motions on soil response. The
loading sequence was as follows:
o Ten cycles of loading with an amplitude corresponding to wellhead displacements from the greatest vessel
motions VM
o Soil consolidation for about three months (prototype time)
o Fifty cycles of loading with amplitudes corresponding to wellhead displacements from the combined
effects of a) the greatest probability of occurring for vessel motion and hull VIM displacements and b) the
maximum displacements from riser VIV.
o Soil consolidation for about three months (prototype time)
o Monotonic push to failure.
300
Prototype Head Displacement (mm)

Monotonic
Test 2 Push to
Max Vessel Motion Failure
200 (10 cycles)
Mean Vessel Motion
+ Mean Hull VIM
+ Max Riser VIV
100 (50 cycles)
(a) (c)
(b)

0
1300 1400 1500

Arbitrary Elapsed Time (Prototype Units) (days)

Fig. 9: Pile head load vs time for Test 2

• Test 3. The objective was to assess the impact of large numbers of cycles with displacements corresponding to the
most damaging VIV motions. The loading sequence was as follows.
o One thousand cycles of loading with amplitudes corresponding to wellhead displacements equal to the most
damaging riser VIV case.
o Soil consolidation for the equivalent of three months.
o One thousand cycles of loading with amplitudes corresponding to wellhead displacements equal to the most
damaging riser VIV case.
o Soil consolidation for the equivalent of three months.
o Monotonic push to failure.
OTC 20158 9

150

Prototype head displacement (mm)

Test 3 Monotonic Push
to Failure
Most Damaging Most Damaging
100 Riser VIV Riser VIV
(1000 cycles) (1000 cycles)

50

0
1500 1600 1700 1800 1900

Arbitrary Elapsed Time (Prototype Units) (days)

Fig. 10: Pile head load vs time for Test 3

• Test 4. The objective was to understand the effect displacements corresponding to the maximum hull VIV will have
on the lateral response of the soil. The cyclic loadings performed in Test 3 were repeated here do to a data
acquisition problems with Test 3. The loading sequence for Test 4 was as follows.
o One thousand cycles of loading with amplitude corresponding to wellhead displacements equal to the
maximum hull VIV case.
o Soil consolidation for the equivalent of three months.
o One thousand cycles of loading with amplitude corresponding to wellhead displacements equal to the most
damaging riser VIV case.
o Soil consolidation for the equivalent of three months.
o One thousand cycles of loading with amplitude corresponding to wellhead displacements equal to the
maximum hull VIV case.
o One thousand cycles of loading with amplitude corresponding to wellhead displacements equal to the most
damaging riser VIV case.
o Soil consolidation for the equivalent of three months.
o One thousand cycles of loading with amplitude corresponding to wellhead displacements equal to the
maximum hull VIV case.
o Monotonically pushed to failure immediately after the cyclic loading.

100 Most Damaging

Most Damaging Monotonic Push
Test 4 Riser VIV Riser VIV to Failure
(1000 cycles) (1000 cycles)
Prototype head displacement (mm)

50 (b) (d)

(c)
0
(e)
Max Hull Max Hull Max Hull
VIM VIV VIM
− 50 (1000 cycles) (1000 cycles) (1000 cycles)
(a)

− 100
500 600 700 800 900 1000

Arbitrary Elapsed Time (Prototype Units), (Days)

Fig. 11: Pile head load vs time for Test 4
10 OTC 20158

Test Results:

Data Processing:
At a given depth, x, the relationships between the variables of interest are, with y being the lateral deflection of the pile:
d4y
P: soil resistance: P = EI
dx 4
d3y
V: Shear force in conductor: V = EI
dx 3
d2y
M: Bending moment in conductor: M = EI ................................................................................................... (2)
dx 2
dy
S: slope of pile deflection: S =
dx
Where:
E = conductor Young Modulus
I = conductor moment of inertia

The pressure, P, is therefore obtained by a double differentiation of the bending moments and the deflection, Y, is obtained
by a double integration of the bending moment profiles, which must be obtained from the discrete measurements along the
pile. A key step in the processing of the data is therefore the interpolation of the bending moments data recorded at discrete
depths. The recorded data was processed with the Mathcad® software using the various interpolation functions to derive a
continuous bending moment profile along the pile length:
• 6th order polynomial: a single 6th order polynomial curve was used to fit the entire bending moment profile
along the conductor. The interpolated curve does not necessarily passes through the actual data points.
• Cubic spline function: a cubic spline interpolation passes a curve through a set of points in such a way that
the first and second derivatives of the curves are continuous across each point. The curve is assembled by
taking three adjacent points and constructing a cubic polynomial passing through those points. It should be
noted that the generated moment diagram will pass through all the actual data points and is therefore most
sensitive to uncertainties in the data acquired.
• Loess function: the loess function interpolates the data by generating a second order polynomial at a given
depth by examining the data in a small user-specified neighborhood around the depth of interest. The larger
the user-specified neighborhood, the smoother the interpolated curve will be. The generated curve does not
necessarily passes through the actual data. (Mathcad 2002).

The above curve fitting was performed on the moment diagram and on the obtained shear diagram. The above functions are
precoded in Mathcad® and can be implemented in a spreadsheet with minimum effort. P-Y curves were generated for each
of the three above method of generating bending moments profiles. All the P-Y curves presented herein are the average of
the three P-Y curves obtained by interpolating the recorded bending moments with a 6th order polynomial, a series a cubic
splines, a series of loess functions.

Monotonic Backbone P-Y curves:

Monotonic backbone P-Y curves were obtained from the reference test Test 1, at 8 depths. Figure 12 shows plots of the
lateral unit pressure, P, divided by the shear strength, Su, as a function of the lateral displacement, Y, over the conductor
diameter, D.

The agreement between the FEA curves and the centrifuge curves is considered very good, both in terms of initial modulus
and ultimate pressure at large displacements. The centrifuge curves are stiffer than the API RP2A (2000) curves and the
ultimate pressure also exceeds the value of 9 times Su given by API. The average value of the ultimate unit pressure P is
12.7 times Su for the eight FEA curves considered and 13.4 for the centrifuge curves. The value of the ultimate pressure P is
plotted as a function of normalized lateral displacement in Fig. 13.
OTC 20158 11

16 16

Unit Pressure / Shear Strength, P/SuUnit Pressure / Shear Strength, P/Su Unit Pressure / Shear Strength, P/Su Unit Pressure / Shear Strength, P/Su
14 Depth = 1.4m = 1.5 D
14 Depth = 3.7m = 4 D

12 12
10 10
8 8
6 6
4 4
2 2
0
16 160
0 0.1 0.2 0 0.1 0.2
14 14
12 12
Depth = 5.5 m = 6 D Depth = 6.4 m = 7 D
10 10
8 8
6 6
4 4
2 2
160 160
0 0.1 0.2 0 0.1 0.2
14 14
12 12
Depth = 7.3 m = 8 D Depth = 8.2 m = 9 D
10 10
8 8
6 6
4 4
2 @ 2
20
0 20
0
180 0.1 0.2 180 0.1 0.2
16 16
14 14
12 Depth = 9.1 m = 10 D 12 Depth = 10.5 m = 11.5 D
10 10
8 8
6 6
4 4
2 2
0 0
0 0.1 0.2 0 0.1 0.2
Lateral Displacement / Diameter, Y/D Lateral Displacement / Diameter, Y/D

API WSD 21st FEA

API WSD 21st Errata Centrifuge
st
Fig. 12: Comparison of monotonic P-Y curves from centrifuge Test 1, FEA analysis, and API RP2A 21 Ed. (2000) method.
12 OTC 20158

The framework proposed by Murff and Hamilton (1993) for a linearly increasing shear strength profile was used, albeit with
a modified expression for the calculation of Np, to calculate the ultimate lateral unit pressure. It is therefore proposed to
calculate the ultimate unit pressure, Pmax, as follows, for shear strength profiles approximately linearly increasing with depth:

Pmax = N p ⋅ Su ........................................................................................................................................... (3)

⎛
⎜ − ξ ⋅ z ⎞⎟
⎜ ⎟
⎜ D ⎟⎠
With: N p = 12 − 4 ⋅ e ⎝
.............................................................................................. (4)
ξ = 0.25 + 0.05 ⋅ λ for λ < 6
ξ = 0.55 for λ ≥ 6
S
λ = u0
S u1 ⋅ D
Su 0 : shear strength intercept at seafloor
Su1 : rate of increase of shear strength with depth
D: pile diameter
Z: depth of interest

The value of Np at depth has been limited to 12 for use in actual design, despite the fact that higher values have been
measured in the centrifuge and calculated in the FEA analysis, in order to be consistent with the exact plasticity solutions
proposed by Randolph and Houlsby (1984), for a rough pile.

In Fig. 13, in order to compute the ultimate pressure by Eqn. 3, the above equation, the shear strength profile of Fig. 8 was
approximated to a best-fit linearly increasing profile in the zone of interest. The values of ultimate pressures for the FEA
analysis and the centrifuge tests were measured at a normalized displacement y/D of 0.2 because this is the displacement at
which the pressure would reach a maximum according to the Matlock (1962) criteria.

18
Normalized Lateral Ultimate Unit Pressure,

16
14
12
10
P/Su

8
6
Proposed Equation, modified from Murff and Hamilton (1993)
4 Centrifuge
FEA Kaolin
2 FEA GoM soil

0
0 5 10 15 20
Depth along pile / Diameter, z/D
Fig. 13: Ultimate Normalized Unit Pressure from FEA analysis, Centrifuge test 1, and Equations 3 and 4.

Despite the large monotonic displacement (i.e. up to 1 diameter) experienced by the conductor in the 4 tests, no gapping on
the back side of the conductors was observed in any of the tests, except in the top 1 diameter (Fig. 14). Surface cracks in the
active failure zone were clearly visible around the pile and indicate full adhesion of the soil to the pile for depths greater than
one diameter.

Overall, Test 1 constitutes a strong validation of the FEA analysis for the derivation of monotonic backbone P-Y curves.
OTC 20158 13

Direction of
loading

Fig. 14: Pile after large monotonic lateral push at the end of Test 4. Cracks in active failure zone clearly visible. Soil depression on
active (back) side less than one pile diameter.

The FEA derived curves are shown in Figure 15, where they have been normalized. The resistance P has been divided by the
value of the resistance P at a normalized displacement of 0.2D. The curves from Matlock (1962) are also presented, with the
P value normalized by the maximum P value which occurred at a normalized displacement of 0.2. Fig. 16 shows the
centrifuge P-Y curves from Test 1, normalized the same way. The shape of the FEA-generated backbone curve has also been
fitted with the following empirical equation, inspired from the equation proposed by O’Neill et al (1990) for P-Y curves in
stiff clays:

P ⎡ Gmax ⎛ y ⎞0.5 ⎤
= tanh ⎢ ⋅ ⎜ ⎟ ⎥ ............................................................................................................ (5)
Pmax ⎢⎣100 ⋅ Su ⎝ D ⎠ ⎥⎦

With P max = N p ⋅ Su as per Eqn. 3 and 4

Gmax : maximum shear modulus
Su : shear strength
y
: lateral displacement, y, over pile diameter, D
D
Fig. 16 shows good agreement between the FEA curves, the curves derived from the centrifuge tests, and the curves derived
according to Eqn.5. The Matlock curves shown have been derived as per Matlock (1962):
1/ 3
P ⎛ y⎞
= 0.5 ⋅ ⎜⎜ ⎟⎟ and yc = 2.5 ⋅ ε 50 ⋅ D , with ε 50 equal to 1%.
Pmax ⎝ yc ⎠
with all variables as previously defined.
14 OTC 20158

1.2

Normalized Unit Pressure, P / (P at Y/D=0.2)

1.1

0.9

Normalized Resistance, P / (P at Y/D =0.2)

0.8

0.7
API 21st Ed - e50 = 1%
0.6
Matlock (1962) curve
0.5 FEA - Depth = 1.5D
API 21st Ed - e50 = 1%
FEA - Depth = 4D
Matlock (1962) curve
0.4 FEA - Depth = 6D= 1.5D
FEA - Depth
FEA - Depth
FEA - Depth = 7D= 4D
0.3 FEA - Depth = 6D
FEA - Depth = 8D
FEA - Depth = 7D
FEA - Depth = 9D= 8D
FEA - Depth
0.2 FEA - Depth = 9D
FEA - Depth = 10D
FEA - Depth = 10D
0.1 FEA - Depth = 11.5D
FEA - Depth = 11.5D
ProposedProposed
curvecurve
- Gmax/Su
- Gmax/Su==550
550
0
0 0.1 0.2 0.3

LateralLateral
Displacement / Diameter,
Displacement / Diameter, Y/D Y/D
Fig. 15: Normalized P-Y curves from Finite Element Analyses and Eqn. 5. The resistances P are normalized by the values of the
Resistance P at a displacement Y/D of 0.2

Fig. 17 shows examples of P-Y curve calculated with Eqn. 5 for large embedment depths (i.e. Np ~ 12) for a range of
Gmax/Su ratios, along with the Matlock 1962 curve.

Curves of ultimate pressure over shear strength. P/Su, for shallow depths are given in Fig.18 with Np calculated as per Eqn.
4, along with the FEA and centrifuge curves and demonstrates the good fit. At depth where the value of Np is close to 12, a
similar comparison is shown in Fig. 19 where the proposed curve gives a lower bound of the measured curves.

1.2
Unit Pressure, P / (P at Y/D=0.2)

1.1
1
Normalized Resistance, P / (P at Y/D =0.2)

0.9
API 21st Ed - e50 = 1%
0.8
Matlock parabola
0.7 FEA - Depth = 1.5D
Centrifuge - Depth = 1.5D
0.6 FEA - Depth = 4D
Centrifuge - Depth = 4D
0.5 FEA - Depth = 6D
Centrifuge - Depth = 6D
0.4 FEA - Depth = 7D
Centrifuge - Depth = 7D
Normalized

0.3 FEA - Depth = 8D

Centrifuge - Depth = 8D
0.2 FEA - Depth = 9D
Centrifuge - Depth = 9D
0.1 Proposed curve - Gmax/Su = 550
0
0 0.1 0.2 0.3
Lateral Lateral
Displacement / Diameter,
Displacement / Diameter, Y/D Y/D
Fig. 16: Normalized P-Y curves from Finite Element Analyses, Centrifuge Test 1, and Eqn. 5. The resistances P are normalized by
the values of the Resistance P at a displacement Y/D of 0.2
OTC 20158 15

12

11

Unit Pressure / Shear Strength, P / Su

10

Resistance / Shear Strength, P/Su

8

6 Proposed PY curve:
⎡ G ⎛y⎞ ⎤
0.5
5 P = N p ⋅ Su ⋅ tanh ⎢ max ⋅ ⎜ ⎟ ⎥
100 ⋅ Su ⎝ D ⎠ ⎥
4 ⎣⎢ ⎦

3 API RP2A 21st Ed - e50 = 1%

2 Matlock (1962) curve
API RP2A 21st Ed - e50 = 1%
Matlock
Proposed curve - (1962) curve
Gmax/Su=550
1 Proposed curve - Gmax/Su=550
Proposed curve - Gmax/Su=400
Proposed curve - Gmax/Su=400
0
0 0.1 0.2 0.3

Lateral Lateral Displacement / Diameter,

Displacement Y/D
/ Diameter, Y/D
Fig. 17: P-Y curves as per Eqn.5, compared with API and Matlock curves for large embedment depths. The ultimate value of P is 12
Su instead of 9 Su for the API curve.
16
15
Unit Pressure / Shear Strength, P / Su

14
13
12
11
10 API 21st Ed - e50 = 1% - Depth = 1.5 D
Matlock parabola - Depth = 1.5 D
9
FEA - Depth = 1.5D
8 Centrifuge - Depth = 1.5D
7 FEA - Depth = 4D
Centrifuge - Depth = 4D
6
Proposed curve - Depth = 1.5D
5 Proposed curve - Depth = 4D
4
3
2
1
0
0 0.1 0.2 0.3
Lateral Displacement / Diameter, Y/D
Fig. 18: Normalized resistance curves, P/Su from FEA, centrifuge Test 1, and Eqn. 5 for shallow depths.
16
P / Su

15
14
13
P / (P Strength,

12
at Y/D =0.2)

11
10
Shear

9
Normalized /Resistance,

8
7 API 21st Ed - e50 = 1%
Matlock parabola
Unit Pressure

6
FEA - Depth = 6D
5 Centrifuge - Depth = 6D
FEA - Depth = 7D
4 Centrifuge - Depth = 7D
3 FEA - Depth = 8D
Centrifuge - Depth = 8D
2 FEA - Depth = 9D
Centrifuge - Depth = 9D
1 Proposed curve - Np = 12 - Gmax/Su = 550
0
0 0.1 0.2 0.3
Lateral Displacement / Diameter, Y/D
Lateral Displacement / Diameter, Y/D
Fig. 19: Normalized resistance curves, P/Su from FEA, centrifuge Test 1, and Eqn. 5 for depth where Np ~ 12
16 OTC 20158

Because the tangent modulus of the P-Y curve is of particular interest to calculate the cyclic displacements, Fig. 20 shows the
comparison of the tangent modulus calculated from the proposed method and from the Matlock (1963) parabola.

12 3

11

Tangent Modulus Proposed Curve

10

Tangent Modulus Matlock Curve

Resistance / Shear Strength, P / Su
9

8 2

4 1

0 0
0 0.05 0.1 0.15 0.2
Lateral Displacement / Diameter, Y/D
Matlock (1962) curve - e50=1%
Proposed curve - Gmax/Su=550
Tangent Modulus Proposed curve / Tangent Modulus Matlock curve

Fig. 20: Ratio of Moduli from Matlock (1962) equation and Eqn.5

Cyclic Testing

Test 2:
The cyclic P-Y curves obtained for Test 2 are shown on Figure 21, along with the monotonic P-Y curves obtained for the last
monotonic test at the end of the test series. The backbone P-Y curves, except for those at a depth of 8D, generally agree well
with the backbone curves of Test 1.

Test 4:
P-Y curves were also derived for Test 4. However, because the imposed displacements at the conductor head were so small,
reliable curves could only be obtained at a depth of 1 diameter below the soil surface. Results are shown in Fig. 22.

Secant Modulus Degradation

The cyclic tests were then analyzed to look at the degradation of the secant modulus as a function of the number of load
cycles. The secant modulus of cyclic load cycles was extracted from Fig.21 and 22. Ratios of the secant modulus of each
unload –reload cycle over the tangent modulus on backbone curves at mid-range load are shown on Fig. 23. The numbers
greater than 1.0 confirm that the assumptions in the structural riser software (see Fig. 2) are conservative for the load cases
analyzed (i.e. they give modulus lower than measured in the centrifuge).

The secant modulus of each load cycle is then normalized by the secant modulus of the first cycle. Curves are shown on Fig.
24. It is proposed that the degradation of the secant modulus can be estimated conservatively by the following lower bound:

Mn 0.9
= ................................................................................................................... (6)
M 1 0.9 + 2.5 ⋅ tanh(0.7 ⋅ log(n))
With:
M n : secant modulus of nth cycle
M 1 : secant modulus of 1st cycle
n : number of cycle
OTC 20158 17

166 16

Unit Pressure / Shear Strength, P/Su

144 Depth = 1.4m = 1.5 D 14 Depth = 3.7m = 4 D
122 12
100 10
88 8
66 6
4 4
API API
2 2
FEA FEA
0 Test 1 API 0 Test 1 API
FEA Test 2a - FEA
10 cycles
−2 Test 1cycles − 2
Test 2a - 10 Test 1

Test 2b - Test
502acycles
- 10 cycles Test 2b - Test
50 2acycles
- 10 cycles
−4 Test 2b - 50 cycles− 4 Test 2b - 50 cycles
@ Test 2c - monotonic
Test 2c - Monotonic Test 2c - Test
monotonic
2c - monotonic

−16
6
6 16
−6 0
0 0.1 0.2 0.1 0.2
14
Unit Pressure / Shear Strength, P/Su

14
12 12
10 10
8 Depth = 5.5 m = 6 D Depth = 6.4 m = 7 D
8
6 6
4 4
2 API 2 API
FEA FEA
0 API 0
Test 1 FEA Test 1API
FEA
−2 Test 2aTest
- 101 cycles − 2
Test 2a - 10 cycles
Test 2a
Test- 110 cycles
Test 2bTest
- 50 cycles Test 2a - 10 cycles
Test 2b - 50 cycles
−4 2b - 50 cycles
−4 Test 2b - 50 cycles
@ Test 2cTest
- monotonic
2c - monotonic Test 2c
Test- 2c
monotonic
- monotonic
−6
16 16
−6
0 0.1 0. 0 0.1 0.2
Unit Pressure / Shear Strength, P/Su

14 14
12 12
10 10
8 8
6 6
4 Depth = 7.3 m = 8 D 4 Depth = 8.2 m = 9 D
2 2
API API
0 FEA API 0 FEAAPI
FEA
Test 1
Test 1
TestFEA
1
−2 Test 2a - 10 −2 Test 1
Test 2a -cycles
10 cycles TestTest
2a 2a
- 10 cycles
- 10 cycles
−4 Test 2b - 50
Test 2b -cycles
50 cycles
−4
TestTest
2b 2b
- 50
- 50cycles
cycles
Test
@ Test 2c 2c
- monotonic TestTest
2c 2c - monotonic
- monotonic
−6
20 20
−6
18 0 0.1
180.20 0.1 0.
Unit Pressure / Shear Strength, P/Su

16 16
14 14
12 12
Depth = 10.5 m = 11.5 D
10 Depth = 9.1 m = 10 D 10
8 8
6 6
API API
4 4
FEA API FEA
API
2 2
Test 1FEA FEA 1
Test
0 Test 2a - 10 cycles 0
Test 1 Test 1
Test 2a - 10 cycles
−2 Test 2a - 10 cycles Test 2a - 10 cycles
Test 2bTest
- 50 cycles− 2
2b - 50 cycles
Test
Test 2b
2b --20
50cycles
cycles
−4 Test 2cTest
- monotonic − 4
2c - monotonic Test
Test 2c
2c --monotonic
monotonic
−6 − 6
0 0.05 0.1 0.15 0.20 0.1 0.2
Lateral Displacement / Diameter, Y/D Lateral Displacement / Diameter, Y/D
Fig. 21: Normalized P-Y curves, P/Su vs y/D, from FEA, from centrifuge Test 1, Test 2a, 2b, and 2c.
18 OTC 20158

Unit Pressure / Shear Strength, P / Su

10

0 API
FEA
Test 1 - monotonic
API
Test 2a - first FEA
loading - 1st load cycle
Test 1 - monotonic
Test 2b - second loading - 1st load cycle
Test 2a - first loading - 1st load cycle
Test 2c monotonic
Test 2b -after
secondcyclic
loading -loading
1st load cycle
Test 2c monotonic after cyclic loading
Test 4a - cycles 1-10-50-100-250-500-700-1000
Test 4a - cycles 1-10-50-100-250-500-700-1000
− 10
Test 4b - cyclesTest1-10-50-100-250-500-700-1000
4b - cycles 1-10-50-100-250-500-700-1000

0 0.05 0.1

Lateral Displacement / Diameter, Y/D

Fig. 22: Normalized resistance curves P/Su from FEA, from centrifuge Test 1, Test 2a,b,c and Tests 4a, and 4b at a depth of one (1)
diameter.

Tangent Modulus at mid-range load

100
Test 2a - Depth = 1.5D Test 2a - Depth = 4D P2
FEA derived
Test 2a - Depth = 6D Test 2a - Depth = 7D (P1+P2)/2 backbone PY
Tangent Modulus on backbone curve at mid-range load

Test 2a - Depth = 8D Test 2a - Depth = 9D P1 curve

Test 2a - Depth = 10D Test 2a - Depth = 11.5 D

Pressure, P

test 2b - Depth = 1.5 D Test 4a - Depth = 1.5 D

Secant Modulus of unload-reload cycle

Secant modulus of
Test 4a - Depth = 4D unload re-load cycle

Lateral Displacement, Y

10

1
1 10 100 1000
Load Cycle number
Fig. 23 Ratio of Secant Modulus of unload-reload cycle over tangent modulus on backbone curve at mid-range load. Numbers
above one (1) confirm that the structural software assumptions are conservative for the problem of interest herein.
OTC 20158 19

Tangent Modulus at mid-range load

Test 2a - Depth = 1.5D Test 2a - Depth = 4D

Test 2a - Depth = 6D Test 2a - Depth = 7D P2
FEA derived
(P1+P2)/2 backbone PY
Test 2a - Depth = 9D Test 2a - Depth = 10D
P1 curve
1.4 Test 2a - Depth = 11.5 D test 2b - Depth = 1.5 D
Secant modulus of 1st

Pressure, P
Test 4a - Depth = 1.5 D Test 4a - Depth = 4D unload re-load cycle
Proposed lower bound
Secant modulus of nth
1.2 unload re-load cycle
Secant Modulus of unload-reload cycle

Secant Modulus of FIRST unload cycle

Lateral Displacement, Y
1

0.8

0.6

0.4

0.2

0
1 10 100 1000
Load Cycle Number

Fig. 24 Ratio of the secant modulus of unload-reload cycles over the secant modulus of the first unload-reload cycle. Proposed
lower bound, as per Eqn. 6 is also shown

Conductor Head Load-Displacement curves.

Next the overall conductor head-displacement curves were analyzed because, thanks to their higher resolution, they can
provide additional insight into soil behavior.

First, the load-displacement curve for Test 2 (a,b,c) are shown on Fig. 25. It can be seen that the ultimate monotonic push
provides a curve that is stiffer than the reference test 1. This confirms that one-way small-displacement cycles have little
effect on subsequent large displacement behavior, if enough time is allowed for soil consolidation.

The same data is shown for Test 3 on Figs. 26 and 27 and the same conclusions can be drawn. Additionally, Fig. 27 suggests
that little global stiffness reduction is experienced beyond 200 cycles, which is consistent with the P-Y curve secant modulus
degradation curve of Fig. 24.

1000
Prototype head load (kN)

800

600

400

200 Monotonic push - test 1

test 2a - 10 cycles
Monotonic push - test 1
0 test 2b - 50 cycles
test 2a - 10 cycles
test 2b - 50 cycles
test 2c - final monotonic push to
test 2c - final monotonic pushfailure
to failure
− 200
0 200 400 600 800 1000 1200
Prototype conductor head displacement (mm)
Fig. 25 Conductor pile head load-displacement curves. Tests 1 and 2.
20 OTC 20158

backbone - test 1
test 3a - cycles 1 to 1000

Prototype head load (kN)

test 3b - cycles 1 to 1000
400
test 3c - final monotonic push

150

Prototype head displacement (mm)

200 Monotonic Push
to Failure
Most Damaging Most Damaging
100 Riser VIV Riser VIV
(1000 cycles) (1000 cycles)
(3c)
50 (3a) (3b)

0
1500 1600 1700 1800 1900
0 Arbitrary Elapsed Time (Prototype Units) (days)

0 100 200
Prototype conductor head displacement (mm)
Fig. 26 Conductor pile head load-displacement curves. Tests 1 and 3.

300
load (kN)

200
head load, (kN)
Prototype head
Prototype

100 backbone - test 1

cycles 1 to 10
cycles 10 to 50
cycles 50 to 200
cycles 200 to 500
0 cycles 500 to 750
cycles 750 to 1000

0 20 40 60 80

Prototype conductor head displacement (mm)

Prototype head displacement (mm)

Fig. 27: Conductor pile head load-displacement backbone curves. Tests 1 and 3a with cycle batches highlighted.

The global conductor head load-displacement curve for Test 4a is shown on Fig. 28. Because the loading for this test is two-
way around the origin, the global response does soften but quickly reaches equilibrium after about 50 cycles.

backbone curve - test 1

test 4 - cycles 1 to 10
Prototype head load (kN)

100 test 4 - cycles 10 to 50

test 4 - cycles 50 to 200
test 4 - cycles 200 to 500
test 4 - cycles 500 to 750
test 4 - cycles 750 to 1000
0

− 100

− 20 0 20

Prototype conductor head displacement (mm)

Fig. 28: Conductor pile head load-displacement curves. Tests 1 and 4a with cycle batches highlighted.
OTC 20158 21

Finally, all the conductor head-displacement backbone curves for all tests are plotted on Fig. 29. It is interesting to note on
Fig. 29 that the response for the last monotonic push is stiffer than that of the reference test, even for Test 4 where soil
consolidation was not allowed after the last batch of two-way cyclic loading. This confirms that small amplitude load cycles
do not seem to have any negative influence on subsequent large amplitude loads displacement behavior. Fig. 29
demonstrates that all the small displacement backbone curves follow closely the same loading path. Again, small amplitude
load cycles do not seem to have any negative influence on subsequent small amplitude loads displacement behavior if
reasonable soil consolidation is allowed. This is an important confirmation that each fatigue-inducing load event can be
treated independently from the other, if they occur sufficiently apart from one another.

1000
Prototype head load (kN)

backbone curve - test 1

test 2 - 1st cycle of 1st set of 10 cycles
test 2 - 1st cycle of 2nd set of 50 cycles
500
test 2 - last push to failure after cyclic loads
test 3 - 1st cycle of 1st set of 1000 cycles
test 3 - 1st cycle of 2nd set of 1000 cycles
test 3 - last monotonic push to failure after cyclic loads
test 4 - 1st cycle of 1st set
test 4 - 1st cycle of 2nd set
test 4 - 1st cycle of 3rd set
test 4 - 1st cycle of 4th set
test 4 - 1st cycle of 5th set
test 4 - last monotonic push to failure after cyclic loads

0
0 500 1000 1500

Prototype conductor head displacement (mm)

400
Prototype head load (kN)

300

backbone curve - test 1

test 2 - 1st cycle of 1st set of 10 cycles
200
test 2 - 1st cycle of 2nd set of 50 cycles
test 2 - last push to failure after cyclic loads
test 3 - 1st cycle of 1st set of 1000 cycles
test 3 - 1st cycle of 2nd set of 1000 cycles
test 3 - last monotonic push to failure after cyclic loads
test 4 - 1st cycle of 1st set
100
test 4 - 1st cycle of 2nd set
test 4 - 1st cycle of 3rd set
test 4 - 1st cycle of 4th set
test 4 - 1st cycle of 5th set
test 4 - last monotonic push to failure after cyclic loads

0
0 50 100 150

Prototype conductor head displacement (mm)

Fig. 29: Conductor pile head load-displacement backbone curves for all tests.
22 OTC 20158

Another important observation is that , at large displacements, the monotonic curves after cyclic loads in Tests 2, 3, and 4
give higher ultimate pile head resistances than the reference monotonic curve of Test 1. The difference in resistances at large
displacements is interpreted to be due to an increase in soil shear strength through hardening because of the small-amplitude
loads. Test 4, with the most small amplitude load cycles (and therefore the most potential hardening), does show the most
increase in ultimate pile head load on Fig. 29. The consequences of soil hardening under cyclic loads could be of critical
importance when re-assessing the foundation ultimate capacity of offshore structures that have been in service for many
years.

Conclusions:
The paper investigated the lateral soil-structure interaction for loads transferred from a top tensioned riser to a jetted
conductor. Comparisons were made between P-Y curves generated from Finite Element Analyses and centrifuge testing.
Insight into cyclic loading soil-structure interaction was also provided. The work herein:
• Validates the FEA methodology used by Templeton (2009) to generate P-Y curves for Gulf of Mexico clays (Fig.
12).
• demonstrates that, for the loads of interest herein, the lateral P-Y curves provided by API RP2A (2000) are too soft
and underestimate the ultimate unit pressure acting on the conductor. A general equation (Eqn. 5) has been
proposed to generate P-Y curves for soft clays. It should be noted that the shear strength used in the derivation of
the P-Y curves should account for rate effects.
• Confirms that the methodology used by the structural riser software (and described in Fig. 2) to derive the unload-
reload cyclic stiffness from the monotonic FEA P-Y curves is conservative for the problem of interest herein(Fig.
23), except for two-way loading such as in Test 4 (Fig. 28) which represented Hull VIM. Fortunately, this type of
loading only caused 20% of the overall fatigue damage at the depth of interest along the conductor.
• Quantified the cyclic P-Y curve secant modulus degradation as a function of load cycles (Eqn. 6). Little modulus
degradation is seen to occur after 200 load cycles.
• Confirms that small amplitude load cycles do not seem to have any negative influence on subsequent large
amplitude loads displacement behavior, even if no soil consolidation is allowed (Fig. 29).
• Suggests that small amplitude load cycles do not seem to have any negative influence on subsequent small
amplitude loads displacement behavior if reasonable soil consolidation is allowed (Fig. 29).. This is an important
confirmation that each fatigue-inducing load event can be treated independently from the other, is sufficiently
spaced in time.

The paper also suggests that, when re-assessing the lateral behavior of piled offshore structures that have been in service for
many years, the pile lateral load-deflection behavior will benefit from the following modification to API RP2A recommended
practice:
• increasing the ultimate unit pressure from 9 times the shear strength to 12 times the shear strength
• adjusting the input shear strength for rate effects, as appropriate for the loading of interest
• accounting for potential soil hardening and increase in shear strength due to many previous cycles of small
amplitude loads.

Last, it should be recognized that, although the use of API P-Y curves are conservative to determine the stresses in structural
members below the mudline, the use of API P-Y curves will be unconservative to estimate stresses in structural members
above the mudline (i.e. wellhead connectors, Blow Out Preventers, etc…).

Acknowledgement

The author acknowledges the owners of the data herein, BP America Inc. and Shell Exploration & Production Co., and is
grateful for their permission to publish.

The author would also like to acknowledge the contributions and assistance of the following people. Ed Clukey was the
centrifuge program initiator and, along with Eric Liedtke, developed the original testing program. Ed and Eric also reviewed
a draft of this paper and provided valuable feedback. The centrifuge tests were performed at the C-CORE facility in St
John’s, Newfoundland, under the direction of Ryan Phillips. Laboratory tests on kaolin were performed at the Fugro South
laboratory in Houston under the direction of Bill DeGroff. The author is also grateful to Ed Clukey and J. S. (Jack)
Templeton for insightful conversations and guidance on the topic of laterally loaded piles and advanced soil constitutive
modeling.
OTC 20158 23

References

API RP2A (2000): “Recommended Practice for Planning, Designing and Constructing Fixed Offshore Platforms – Working Stresses
Design”, 21st Edition, December 2000.

API RP2A-WSD Errata (2007): “Errata and Supplement 3 to API Recommended Practice 2A-WSD, Recommended Practice for Planning,
Designing and Constructing Fixed Offshore Platforms – Working Stresses Design”, 21st Edition, December 2000.”, October 2007.

Beynet, P. (2008): personal communication

Burk, J. (2007): personal communication

C-CORE (2005):”Centrifuge Modeling of Cyclically loaded Conductors In Clay: Tests 1-4”, Report R-04-100-293, March

Mathcad (2002): User manual for the Mathcad software, Mathsoft Engineering and Education, Inc.

Matlock, H. (1962): “Correlations for design of laterally loaded piles in soft clay”. Report to Shell Development Company, September

Murff, J.D., and Hamilton, J.M. (1993): P-Ultimate for undrained analysis of laterally loaded piles”, Journal of Geotechnical Engineering,
Vol. 119, No.1, January.

O’Neill, M.W., Reese L.C., and Cox, W.R. (1990): “Soil Behavior for Piles Under Lateral Loading”. Proceedings, Offshore Technology
Conference, Houston, TX, Paper number 6377

Randolph, M.F., and Houlsby, G.T. (1984): “The limiting pressure on a circular pile loaded laterally in cohesive soil.” Geotechnique,
London, England, 34(4), 613-623

Reese, L., and Cox, W.R. (1975): “Field Testing and Analysis of Laterally Loaded Piles in Stiff Clays”, Proceedings, Offshore Technology
Conference, Houston, TX, Paper number 2312

Stevens, J. B. and Audibert J.M.E, (1979): “Re-examination of P-Y curve formulations”, Proceedings, Offshore Technology Conference,
Houston, TX, Paper number 3402

Templeton, J.S. (2009): “Finite element analysis of conductor-seafloor interaction”, Proceedings, Offshore Technology Conference,
Houston, TX, Paper number 20197.