You are on page 1of 7

Pipelines, risers and subsea earthworks

Frontiers in Offshore Geotechnics III Meyer (Ed.)


2015 Taylor & Francis Group, London, ISBN: 978-1-138-02848-7

Seabed stiffness model for steel catenary risers


C.P. Aubeny & T.A. White
Civil Engineering, Texas A&M University, College Station, Texas, USA

T. Langford & V. Meyer


Norwegian Geotechnical Institute, Oslo, Norway

E.C. Clukey
BP Upstream Technology Group, Houston, Texas, USA

ABSTRACT: The stress ranges used in the fatigue analysis of Steel Catenary Risers (SCRs) are calculated
from changes in riser stresses caused by wave motions. The Touchdown Zone (TDZ) the region where the
riser comes into contact with the seabed is often the critical location from the standpoint of fatigue life of
the riser. Relatively minor variations in seabed stiffness can have a significant effect on estimated fatigue life;
therefore, the seabed stiffness model must realistically portray actual non-linear behavior. Riser interaction
with the seabed involves a number of complexities including plastic deformation during virgin penetration,
nonlinear load-displacement response, softening during cyclic loading, reconsolidation and thixotropic strength
gain during rest periods, and suction-induced tensile resistance during breakaway. Physical model testing has
been an important tool for characterizing these features of seabed behavior. This paper first presents an analysis
of single-element riser model tests that measured soil resistance for different conditions of cyclic displacement
magnitude and rest intervals. These data are then used to develop an equivalent soil spring model.

INTRODUCTION

Large, floating structures are widely used by the oil


and gas industry to facilitate the extraction of hydrocarbon resources in deep waters throughout the world.
These structures are tethered to the seafloor using
anchors and mooring lines, and steel catenary risers
(SCRs) are regularly used to transport oil and gas
between these structures and other facilities. SCRs are
attached directly to the structures (Figure 1), and thus
are affected by any movements of the structure due
to wave motions. This results in cyclic loading behavior at the touchdown zone (TDZ), the region in which
the riser contacts the seafloor. A major design concern
regards the fatigue life of the riser in the TDZ, which is
strongly influenced by the soil conditions in this region
(Hatton 2006). Potential fatigue failures are related to
bending stress in the SCR (e.g. Aubeny et al. 2006,
Clukey et al. 2005, Randolph & Quiggin 2009), which
depends primarily on the seafloor stiffness, variation
in embedment depth of the SCR throughout the TDZ,
and amplitude of movements of the SCR due to wave
and current motions.
Simplified characterization of the seabed, rigid
or linear elastic, typically lead to overestimates of
cyclic bending stresses with a consequent underestimate of fatigue life. This paper presents a refined
spring model for soil resistance considering nonlinear stiffness, suction-controlled breakaway during

Figure 1. SCR schematic.

uplift, and stiffness degradation during cyclic loading. While riser motions in the TDZ can be both
vertical and lateral, the model presented herein considers only vertical motions. Input parameters for the
model are derived from large-scale tests conducted in
a laboratory test bed.

MODEL TESTS

Physical model testing has been an important tool


to investigate seabed-SCR interaction in the TDZ, as
reported by several authors including Bridge et al.

351

Table 1.

Summary of tests.

Penetration (z/D)
Test
2-1
2-2
2-3
2-4
2-5
2-6

init.

max

min

Av.
velocity
(mm/s)

0.25

0.5

0.0

0.5

20

0.475

0.5

0.45

0.05

200

No. of
cycles

Delay
between
episodes
(days)
1
0.17
1618
1
0.17
911

Figure 2. Undrained shear strength profile (Langford &


Meyer 2010).

(2004) and Giertsen et al. (2004). Large scale testing was performed at NGI using a biaxial rig with
hydraulic actuators and 50 kN maximum load capacity (Langford & Meyer 2010). The model riser was
subjected to vertical loading only, using the vertical
actuator. The test bin has plan dimensions of 3.6 m by
1.75 m, which allowed for up to 6 riser model test pits.
The riser model had a length of 1300 mm, a diameter of
174 mm, and was coated with roughened polypropylene. During testing, vertical displacement, force, and
pore pressure measurements were taken. Details of the
testing apparatus are given by Langford & Aubeny
(2008a, b).
The soil used for the test program was a West
African clay with a plasticity index of approximately
100%. Prior to testing, the slurrified clay was fully
consolidated using a stress of 15 kPa applied with
a vacuum. Mini T-bar penetration tests were performed immediately prior to testing to generate a shear
strength profile for the test pit. The testing was performed at a rate of 2 cm/s, and the shear strength profile
was interpreted using a T-bar bearing factor of 10.5
(Stewart & Randolph 1994). Figure 2 shows the interpreted strength profile, including corrections based on
a non-linearly increased T-bar factor and a roughness
coefficient approach (Aubeny et al. 2006). The soil
strength profile was similar to in-situ profiles seen
for West African developments, with a relatively high
apparent OCR.All testing in the test bin was performed
under water.
Two sets of test were performed, as shown in Table 1,
with different episodes of cyclic vertical loading. Tests
2-1, 2-2 and 2-3 included 3 episodes of 20 cycles with a
magnitude of 0.5 times the diameter (D) of the model
riser. Tests 2-4, 2-5, and 2-6 included 3 episodes of
200 cycles at a magnitude of 0.05D. The cycles were
applied using sinusoidal waveforms. Cyclic parcels
were applied after initial penetration using a constant
rate of 0.5 mm/s. Different rest, or set-up, periods were
prescribed between each cyclic episode for each test

Figure 3. Overview of soil spring model.

riser. The first, second, and third cyclic episodes are


hereafter referred to as A, B, and C, respectively.
It is noted that although the large displacement
test involved uplift displacements of 0.5D, the riser
detached from the soil at around 0.09D. From the
standpoint of soil behavior, the maximum uplift is considered to be 0.09D, since displacements larger than
that occur under conditions of the riser pipe being
complete separated from the soil.
3

SPRING MODEL

The main objective for this soil spring model is to


generate load-displacement (P-y) curves based on
cyclic displacement magnitude and the total number
of unload-reload cycles that the riser experiences. The
model predicts soil response for the three modes of
load-displacement behavior shown in Figure 3: virgin
penetration, elastic rebound prior to suction breakaway during riser uplift, and reloading when the riser
re-contacts with the soil surface.
Solid lines depict the load path tracked by the
model, with dashed lines showing typical experimental observations. Most notably, breakaway is modeled as an abrupt detachment when the suction limit
is reached, in contrast to the gradual detachment
observed experimentally (dashed lines in Figure 3).
No stiffness degradation is assumed once the pipe

352

detaches from the soil, although hydrodynamic action


caused by the pipe approaching could affect soil stiffness. Effects of the rest and setup periods are not taken
into account in the current version of the model.
3.1

Initial penetration

The backbone curve (virgin penetration) is represented by Equation 1, where P1 is soil resistance
(force/length), su is the shear strength at the pipe invert,
y is the depth of the riser beyond the mudline, and D
is the riser diameter.
The parameters a and b define the shape of the backbone curve and vary depending on trench geometry
and pipe roughness. They can be evaluated from the
laboratory model test data from the initial monotonic
penetration stage of the test.

3.2

Uplift

Unloading behavior occurs when the riser moves up


and away from the soil mass prior to reaching the suction breakaway displacement. The unloading behavior
for the model is controlled by the hyperbolic relationship shown in Equation 2, in which the dimensionless
parameters Kmax and f define the initial (e.g., small
displacement) tangent stiffness and asymptote of the
hyperbola, respectively.

Figure 4. Degradation of Kmax versus N (ycyc = 0.09D).

Both Kmax and f degrade with cyclic loading. Initial


unload tangent stiffness Kmax was evaluated for each
load cycle N from the data set, for which the following
equation provides a reasonable fit:

Equation 3 defines an inverse exponential degradation law where Kmax at cycle N transitions from a
maximum value at load cycle N = 1 to its steady state
value Kmax SS , where K controls the rate of degradation. Figure 4 shows the curve fit for the large
displacement tests, ycyc = 0.09D.
The parameter f in Eq. 2 varies with load cycle
according to a power law relationship (Eq. 4), where af
defines the value of f for the first unload cycle (N = 1),
while bf controls the rate of degradation of f with
respect to load cycle N. Figure 5 shows an example
curve fit of Eq. 3 to measurements.

Figure 5. Degradation of f with increasing N (ycyc =


0.09D).

riser from the soil. The model tests indicated that yb
degrades with load cycle as an inverse exponential:
Breakaway characterized in terms of an uplift displacement yb that is sufficiently large to detach the

353

Figure 6. Degradation of yb with increasing N


(ycyc = 0.09D).

Equation 5 defines an inverse exponential degradation law where yb at cycle N transitions from a
maximum value at load cycle N = 1 to its steady state
value ybss , where controls the rate of degradation.
Figure 6 shows the curve fit for the large displacement
tests, ycyc = 0.09D.
3.3

Setdown

Reloading behavior occurs when the riser moves down


and is in contact with the soil. The soil resistance during reloading is described in term of a simple reduction
factor Rf applied to Eq. 3. The values of Kmax and f
in Eq. 5 are the same occurring in the unload cycle
immediately preceding reload.

The parameter Rf is defined in terms of the reversal


soil resistance PN and PN+1 (Figure 3) for the load
cycle in question. The data (Figure 7) suggest a power
law relationship defining the rate of degradation in soil
resistance versus load cycle:

where aR and bR are fitting coefficients. The reloading


stiffness reduction factor becomes as follows:

Figure 7. Degradation
(ycyc = 0.09D).
Table 2.

of

PN

with

increasing

Damage functions.

Description
Unload stiffness
Kmax
Asymptote
parameter f
Steady state
breakaway yb
Breakaway
degradation y
Reversal soil
resistance PN
Reversal soil
resistance PN

Ref.
Eq.

Equation

K = Kref [1 + cK log (y/yref )]

bf = bfref [1 + cbf log (y/yref )]

ybss = ybssref [1 + cyss log (y/yref )]

y = yref [1 + cy log (y/yref )]

aR = aRref [1 + caR log (y/yref )]

bR = bRref [1 + cbR log (y/yref )]

3.4 Degradation laws


The equations above characterize the degradation of
unload-reload stiffness for a single magnitude load
cycle. In actuality, displacement magnitude is not uniform. This requires a mechanism in the model for
predicting degradation associated with different levels
of cyclic displacement. The logarithmic damage functions in Table 2 were adopted to define how cyclic
displacement magnitude affects the various parameters controlling rates of degradation. In these equations
the term ref in the subscripts denotes the value of

354

Table 3.
tests.

Input parameters derived from laboratory model

Process

Equation

Parameter

ycyc =
0.09D

ycyc =
0.05D

Initial
Penetration
Unloading

a*
b*
af
Kmax1
Kmaxss
yb1 /D
ybss /D
aR
K
y
bf
bR

5.2
0.33
1.90
70
34
0.09
0.023
0.742
0.16
0.16
0.58
0.46

0.0037
3.54
0.08
0.023
0.51
0.76

Reloading
Degradation

4
3
3
5
5
7
3
5
4
7

*Fitted to strength profile su = 4 kPa + z 12.7 kPa/m.

Table 4.
Process

Figure 8. Comparison of model calculations to measurements ycyc /D = 0.05.

Damage function parameters.


Reference Parameter Damage Coefficient

Unloading ybss /D
Reloading
aRref
Degradation Kref
yref
bfref
bRref

0.023
0.742
0.16
0.16
0.58
0.46

cyss
caR
cK
cy
cbf
cbR

3.29
14.8
1.96
3.36
0.473
2.49

a given parameter derived from test data conducted at


an arbitrarily selected cyclic displacement magnitude.
In this study, ycyc /D = 0.09 is taken as the reference data set. As noted earlier, the test program
being interpreted included two cyclic load magnitudes
ycyc /D = 0.05 and 0.09, permitting some assessment
of the relative effect of load magnitude on degradation.
Figure 9. Comparison of model calculations to measurements ycyc /D = 0.5.

3.5

Input parameter values

Table 3 summarizes the model parameters derived


from model tests conducted at cyclic load magnitudes
ycyc = 0.05D and 0.09D. Table 4 presents the parameters for the damage functions in Table 2 that are used
to unify the data set for ycyc = 0.05D and 0.09D into
a single model.
The damage function equations in Table 2 introduce two important capabilities for simulating SCR
P-y behavior. The model provides capabilities for simulating variable cyclic displacement magnitudes in a
somewhat restricted sense. The cyclic displacement
magnitude can be varied from cycle to cycle. However, within a given load cycle the unload (uplift)
displacement must equal the reload (laydown) displacement. While the model cannot simulate P-y
behavior for fully random displacement loading, analysis of a non-uniform displacement magnitude test
series is possible.

4 VALIDATION AND PARAMETRIC ANALYSIS


As a back check on the input parameter selection,
the model parameters in Tables 3 and 4 were applied
to the model to recreate Test Series 2-1, 2-2, and 23 (ycyc /D = 0.5) and Test Series 2-4, 2-5, and 2-6
(ycyc /D = 0.05). Comparisons of model to measurement are made on the basis of secant stiffness during
unload from the reversal point PN in Figure 3, where
Ksec = P/y. The secant stiffness is expressed in normalized form Ksec /(P1 /D). Comparisons are made for
selected cycles N = 1, 10 and 50 (Figures 8 and 9).
The curves are truncated when soil-pipe separation
occurs, so they do not always extend to the target cyclic
displacements. Agreement between model and measurement is excellent for the initial load cycle N = 1
for both test series. There is some tendency for the
model to under-predict the stiffness at larger numbers

355

Figure 10. Effect of displacement magnitude and load cycle


on secant stiffness.

of load cycles in the case of ycyc /D = 0.05. At the


larger cyclic displacement magnitude ycyc /D = 0.5
there is significant scatter amongst the three tests in
the series, especially for N = 10. The parameters were
selected as a best fit to the combined data from a
given test series, so differences between model predictions and measurements from individual tests are
not unexpected. Considering the scatter in the measurements, the agreement with the model calculations
is considered good.
A primary function of the P-y model is to integrate
model test data for various load and soil conditions
into a unified model to be used as a subroutine in a
soil-riser interaction analysis. In the case the model
provides a means of unifying data from data sets
having cyclic displacement ycyc = 0.05D and 0.5D.
Figure 10 presents the results of a series of model
predictions ycyc = 0.05D and 0.09D for load cycle
repetitions N = 1, 10, and 100. It is again noted that
although the second test series involved a cyclic displacement of the riser pipe ycyc = 0.5D the model
simulations were only carried out to ycyc = 0.09D,
since complete detachment of the riser from the soil
occurred for displacements greater than that value.
Cycle degradation is seen to decrease stiffness by an
order of magnitude between the initial load cycle as the
steady state is approached. The magnitude of cyclic
displacement has a comparable level of significance
on stiffness.
5

CONCLUSIONS

interacting with a soil seabed. The model analyses only


vertical motions of the riser. Features of the model
include a power law function defining the backbone
curve, a hyperbolic function describing unload-reload
(uplift-laydown) cycles, and degradation equations
describing the decay in stiffness as cycling occurs.
The degradation equations were of an either inverse
exponential or a power law form, according to which
gave the best fit. The degradation formulation allows
for greater rates of stiffness reduction to accompany
greater magnitudes of cyclic loading. The model is
designed for providing spring stiffness coefficient
values along the length of the SCR touchdown in a soilstructure interaction program to support a fatigue life
evaluation. Agreement between model and laboratory
measurement is generally good. Cyclic displacement
magnitudes on the order of ycyc /D = 0.1 can reduce
stiffness by one order of magnitude. Similarly repeated
loading can degrade soil stiffness by another order of
magnitude.
REFERENCES
Aubeny, C.P., Biscontin, G. & Zhang, J. 2006. Seafloor Interaction with Steel Catenary Riser, Final Project Report,
OTRC Industry Consortium.
Bridge, C., Laver, K., Clukey, E. & Evans, T. 2004. Steel
catenary riser touchdown point vertical interaction model,
Offshore Technology Conference, Houston, TX. OTC
16628.
Clukey, E., Haustermans, L. & Dyvik, R. 2005. Model tests to
simulate riser-soil interaction effects in touchdown point
region, International Symposium on Frontiers in Offshore
Geotechnics.
Giertsen, E., Verley, R. & Schrder K. 2004. CARISIMA A
Catenary Riser/Soil Interaction Model for Global Riser
Analysis, Proc. Conf. on Offshore Mechanics and Arctic
Engineering (OMAE).
Hatton, S. 2006. Riser Reassurance, Offshore Engineer
Magazine, July 2006.
Langford, T.E. & Meyer, V.M. 2010. Vertical cyclic testing
of model steel catenary riser at large scale, Frontiers in
Offshore Geotechnics II, pp. 803808.
Langford, T.E. & Aubeny, C.P. 2008a. Large scale soilriser model testing on high plasticity clay, International
Symposium for Offshore and Polar Engineering, ISOPE2008-TPS-543.
Langford, T. & Aubeny, C.P. 2008b. Model tests on catenary riser in marine clay, 2008 Offshore Technology
Conference, OTC 19495, Houston.
Stewart, D.P. & Randolph, M.F. 1994. T-bar penetration in soft
clay, J. Geotechnical Engineering, ASCE, Vol. 118(12),
22302235.
Randolph, M. & Quiggin, P. 2009. Non-linear hysteretic
seabed model for catenary pipeline contact, OMAEA200979259, ASME.

This paper presents a model of soil resistance versus


displacement (P-y) behavior for a steel catenary riser

356

You might also like