See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254509216

Numerical Analysis of the Behavior of Suction Caissons In Clay

Article  in  International Journal of Offshore and Polar Engineering · January 2002

CITATIONS READS
43 1,471

5 authors, including:

Jianchun Cao Ryan Phillips

SBM Offshore C-CORE
12 PUBLICATIONS   108 CITATIONS    114 PUBLICATIONS   2,006 CITATIONS   

SEE PROFILE SEE PROFILE

Radu Popescu Basim yousif Al-khafaji

California Polytechnic State University, San Luis Obispo University of Thi-Qar
66 PUBLICATIONS   1,663 CITATIONS    7 PUBLICATIONS   107 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

DOE Applications View project

All content following this page was uploaded by Radu Popescu on 02 December 2014.

The user has requested enhancement of the downloaded file.

Proceedings of The Twelfth (2002) International Offshore and Polar Engineering Conference
Kitakyushu, Japan, May 26 –31, 2002
Copyright © 2002 by The International Society of Offshore and Polar Engineers
ISBN 1-880653-58-3 (Set); ISSN 1098-6189 (Set)

Numerical Analysis of the Behavior of Suction Caissons in Clay

J. Cao*, R. Phillips**, R. Popescu**, J.M.E. Audibert* and Z. Al-Khafaji*
*Fugro-McClelland Marine Geosciences, Inc., Houston, USA
** C-CORE, Memorial University of Newfoundland, St. John’s, NF, Canada

ABSTRACT
This paper presents the development of a finite element model for
simulating the behavior of suction caissons subjected to vertical
loading, and the validation of the numerical model based on centrifuge
experimental results. The pull out load versus displacement (Pu-d)
curve obtained from finite element analyses (FEA), using the passive
suction recorded in the centrifuge tests, was in close agreement with
that obtained using the centrifuge pull out tests. Numerical simulation
of passive suction is still a challenging problem, which is not fully
solved at the present time. A new method, in which the water inside the
caisson is simulated by a very soft poro-elastic material, was used to
simulate the development of passive suction. Although the passive
suction versus pull out displacement curve obtained using the FEA was
slightly different from that obtained from the centrifuge tests, the
magnitude of the maximum suction was almost the same, and the pull
out force with displacement curve obtained using the FEA was in close
agreement with that obtained from the centrifuge tests.
KEY WORDS: Suction Caisson; Passive Suction; Finite Element
Analysis; Modified Cam-Clay Model; Soil/Caisson Interaction; Contact
Surface.
INTRODUCTION
As more exploration and development are carried out in deep waters,
floating platforms are increasingly being used for both drilling and
production. Recently, suction caissons have become the preferred
mooring system for those offshore structures since they have several
advantages over other mooring systems. Consequently, an
understanding of suction caisson’s behavior is necessary for both
designers and operators. In the present investigation, a series of finite
element analyses (FEA) of the behavior of suction caissons under
vertical pull out loads was performed using the finite element code
ABAQUS/Standard.
Suction caissons were first introduced in the 1950s. When suction
caissons were first used as foundation elements, the suction was only
used for caisson installation and was not accounted for in the pull out
capacity. With time, it was realized that the contribution of the passive
suction, generated during caisson’s pull out, to the total pullout
capacity is significant. To date, various studies (laboratory tests, field
tests, centrifuge tests, and numerical modeling) have been carried out to
investigate the behavior of suction caissons. FEA analyses simulating
pull out tests were performed by many researchers. In these analyses,
the passive suction was simulated using different approaches, (Aas and
Andersen, 1992, Deng and Carter, 2000, Zdravkovic et al, 1998). FEA
analyses were also used to investigate the behavior of suction caissons
subjected to static vertical or inclined loading (Handayanu and et al,
2000, Zdravkovic et al, 1998), as well as during installation in sand and
clay deposits under undrained, partially drained, and drained conditions
(Clukey and Morrison, 1993, Deng and Carter, 2000). However, the
numerical simulation of passive suction in the FEA is still a
challenging problem that is not fully solved at the present time.
This paper presents the development of a numerical model to
investigate the behavior of suction caisson using the Abaqus software.
The model was developed to simulate the conditions of centrifuge tests
carried out at C-CORE (Cao et al, 2001, Cao et al, 2002). Contact
surfaces between soil and caisson wall, as well as soil surfaces outside
and inside the caisson wall, were used. The interactions between these
surfaces were validated using the results of centrifuge tests. A soft
poro-elastic material was used to simulate the water inside the caisson
top, and the development of passive suction.
DESCRIPTION OF THE CENTRIFUGE TESTS
Centrifuge tests were carried out with a scale factor (model to
prototype) for linear dimensions of 1/100. The outside diameter and
wall thickness of the model caisson were 51.7 mm (2 in.) and 0.65 mm
(1/40 in.), respectively, yielding a D/t ratio slightly over 79. The
maximum penetration depth for all tests was about 180 mm (7 in.),
yielding an L/D ratio of about 3.5. Fine speswhite kaolin clay was used
to construct the testbed. A normally consolidated (NC) clay testbed was
constructed from a slurry having a water content of 120% and
consolidated under both laboratory (1-g) and centrifuge (100-g)
conditions.

795
The target undrained strength (su) profile for the clay testbed The Modified Cam-Clay model variant describes a yield surface shown
was su=0 at the surface increasing with depth at a rate of in Figure 2 and Figure 3 (Hibbitt, Karlsson & Sorensen, Inc., 1998). It
approximately 1.14 kPa/m (7.26 psf/ft). The diameter and the can be expressed as
thickness of the testbeds were 904 mm (~ 3 ft) and 320 mm (~ 1 1 p t 2
ft), respectively. The caisson was installed in-flight by self-
( − 1) 2 + ( ) −1 = 0 (1)
β a0 Ma 0
weight, followed by suction. The model caisson was
where:
subsequently pulled out at a velocity of 10 mm/second (~ 0.4
in./second) by applying a vertical load at the top of the caisson. 1
p = − traceσ is the main effective stress;
3
FINITE ELEMENT MODELING
1 1 r
t = q[1 + − (1 − )( ) 3 ] is a measure of the
Finite Element Mesh and Boundaries k k q
deviatoric stress at failure;
One of the centrifuge models (Cao et al., 2002) has been numerically
simulated, at the prototype scale. The axisymmetric assumption was 2
used in this study. The soil mass was modeled using 8-node q= S : S is the Mises equivalent stress;
axisymmetric quadrilateral biquadratic displacement, bilinear pore 3
pressure elements and reduced integration (CAX8RP). The caisson was 9
modeled using 8-node biquadratic axisymmetric quadrilateral elements r 3 = S : S : S is the third stress invariant, and S is the
(CAX8R). The mesh consisted of 1111 nodes, 576 CAX8RP elements 2
for the soil material, and 39 CAX8R elements for the caisson. All deviatoric stress tensor;
resulting stresses were effective stresses. M is the critical state ratio defined as the ratio of the shear
stress, q, to the effective pressure, p, at critical state. The ratio M can
The finite element mesh and boundary conditions are shown in Figure be estimated using the stress values in triaxial compression tests, when
1. The side boundaries of the FEA mesh were assumed to be the material is at the critical state;
frictionless surfaces constrained from moving in the horizontal β is the wet cap parameter, which controls the shape of
direction. The base was fixed in both vertical and horizontal directions. the yield surface on the wet side of the critical state;
The caisson was constrained from moving in the horizontal direction. a0 is the initial overconsolidation parameter, which can be
estimated using the e-ln p plots along with the initial effective stress;
and
Soft Material K is the third stress invariant parameter, which introduces
a smooth approximation of the Mohr-Coulomb surface.

Other parameters in the Modified Cam-Clay model in Abaqus standard

implementation include:
υ Poisson ratio;
κ logarithmic elastic bulk modulus; and
λ logarithmic hardening modulus.

The parameters κ and λ can be estimated from the results of the

hydrostatic compression test (slope of the rebound line) or from an
oedometer test (slope of the virgin compression line) in an e-ln p plot.
Other parameters, such as the initial void ratio e0, the initial effective
Caisson Tip
stress σ v′ 0 and the coefficient of earth pressure at rest k0 , which
correspond to the initial effective stress state, are also needed.

Figure 1: Finite Element Mesh and Boundary Conditions

Soil Model
In this study, the behavior of suction caissons in normally consolidated
clay was investigated. The Modified Cam-Clay model was used to
model the stress-strain behavior of the porous soil material. The soil
was assumed fully saturated with the flow of the pore fluid through its
voids governed by Darcy’s law.
Fig. 2: Modified Cam-Clay yield Fig. 3: Modified Cam-Clay yield
surfaces in the p-t plane surface sections in the B plane
The parameters used in the modified Cam-Clay model were estimated
using the results of laboratory consolidation tests, direct shear tests,
mini-cone penetration tests and mini-vane shear tests, as well as C-
CORE’s in-house data and other published data in the literature. A
summary of the parameters used in the model is presented in Table 1.

796
 Table 1: Soil Parameters used for the Modified Cam-Clay Model f s = ksu (3)
Soil Parameter Value Used
where su is the undrained shear strength and k is a stress transfer
Poisson’s ratio, ν 0.3 coefficient. According to the centrifuge test results (Cao et al. 2001,
Logarithmic elastic bulk modulus, κ 0.02 Cao et al. 2002) and literature (API, 1993, and NGI, 1999), the average
Logarithmic hardening modulus, λ 0.26 value of k is about 0.85 with a consideration of internal and external
Critical state ratio, M 0.9 friction as well as the effect of self-weight and active suction
penetrations.
Initial overconsolidation parameter*, a0 4.56 kPa/m
Wet cap parameter, β 1.0 The undrained shear strength of clay, su can be estimated by
Third stress invariant parameter, K 1.0
Initial void ratio*, e0 1.1~1.48
s u = m(OCR ) n σ v′ (4)

Coefficient of earth pressure at rest, k0 0.64 where m and n are empirical coefficients equal to 0.19 and 0.59
respectively for speswhite kaolin clay, OCR is the overconsolidation
Friction coefficient at soil/caisson interface, α 0.25
* variable with depth, 4.56 kPa/m = 29 lbf/ft2/ft. ratio equal to 1 for normally consolidated clay, and σ v′ is the effective
vertical stress.
Soil/Caisson Interaction
Eq. 4 can be written as
Two approaches could be used in Abaqus for specifying finite sliding
interaction between deformable bodies: the contact element approach σ h′
s u = m(OCR ) n (5)
and the contact surface approach. The contact element approach is used k0
more frequently since it is applicable in more situations. In this study,
however, the contact surface approach was used based on the modeling where σ h′ is the effective horizontal stress (equivalent to p in Eq. 2),
and computational efforts of a study conducted to identify the most
appropriate technique for simulating pipe/soil interaction (Popescu, and k0 is the coefficient of earth pressure at rest for normally
1999). In the contact surface approach, Abaqus automatically consolidated clay and is equal to 0.64.
generates the appropriate contact elements.
Substituting Eq 5 into Eq. 3, then
Two pairs of surfaces were used to simulate the soil/caisson interaction:
km(OCR ) n
1) the caisson wall surface and the soil surface in contact with the fs = σ h′ (6)
caisson, and 2) the soil surface inside the caisson with the soil surface k0
outside the caisson. The first pair (Figure 4a) depended the soil/caisson
interaction. The second pair (Figure 4b) prevented inter-penetration of Comparing Eq. 2 and Eq. 6, where σ h′ is equivalent to p, the friction
the soils outside and inside the caisson during and after caisson pull
coefficient, µ, results:
out.
km(OCR ) n
Caisson Wall
µ= = 0.25 (7)
k0
Therefore, the friction coefficient at soil/caisson interface, α =µ = 0.25
was used in the FEA.
Inside Soil Outside Soil Surface Inside Soil Outside Soil Surface
Surface Surface Simulation of Passive Suction
Caisson
Surface One advantage of suction caissons is that passive suction is generated
during uplift loading resulting in an increase in the foundation pull out
capacity. The passive suction contributes to the overall pull out
capacity through developing an end bearing resistance at the base of the
(a) Caisson\Soil Surfaces (b) Soil Surfaces Inside and Outside caisson and also through increasing the effective stresses inside the
Caisson caisson, leading to higher skin friction along the inside caisson’s wall.
Consequently, the correct simulation of the passive suction plays a key
Figure 4: Defined Surface Pairs role in the successful modeling of the behavior of suction caissons in
the numerical analyses. However, this simulation is still a challenging
Coulomb friction contact was used. The shear stress between the problem.
surfaces in contact, i.e. unit wall skin friction (fs), is limited by a critical
stress: The passive suction was simulated here by the development of negative
τ critical = µp (2) pore water pressures in a poro-elastic material subjected to tensile
strains. As shown in Figure 5, three aspects were considered in the
where µ is a user-defined friction coefficient (e.g. α in Table 1) and p is simulation of the passive suction in this study:
the normal contact pressure equivalent to horizontal effective stress
σ h′ . 1. The water inside the caisson top was replaced by a porous
soft material through which the suction was applied at the top
The unit wall skin friction was calculated by

797
 of the caisson. At the contact with soil, the poro-elastic obtained from the FEA runs are in close agreement with those
elements shared the same nodes with the soil. measured in the centrifuge test. Consequently, it was concluded that the
2. Perfect contact between the soft material and caisson cap was numerical model could adequately simulate the soil/caisson interaction
assumed. The top surface of the soft material did not have as modeled in the centrifuge tests.
any relative movement with respect to the caisson cap during
pull out. FEA Results and Comparison

FEA runs were conducted following the validation of soil/caisson

3. Both the poro-elastic material and the soil surface inside the
interaction. The water inside the caisson top was simulated by a soft
caisson were constrained from horizontal movement,
resulting in the elimination of cavity development between poro-elastic material. The soil outside the caisson experienced very
water, soil and the caisson’s inside wall. little movement during the pullout of the caisson, resulting in the
development of large relative displacement between the caisson wall
and the outside soil. The failure occurred at the interaction surface of
Soft Material the caisson wall and the soil. The soil inside the caisson had almost the
No Gap
same uplift displacement as the caisson during pullout. In other words,
the soil plug moved up at the same velocity as the caisson during the
pullout. This failure mode was also confirmed by centrifuge test results
(Cao, 2002).
Clay Figure 7 presents a comparison of the total pullout force obtained from
the FEA and those measured during the centrifuge test SAT06-3. Even
Inside Soil Surface

though the experimental curve was not monotonically increasing and

then flattening out at low displacement due to the resolution and
“scatter” in the displacement measurements in the centrifuge model test
(at 100 g), generally, as it can be seen, the FEA results indicated that
the pullout force developed very rapidly as the caisson’s vertical
displacement increased. The peak pull out force value was reached after
a vertical displacement equivalent to about 2% of the caisson diameter.
The agreement between the total pullout force versus pullout
displacement curve obtained from the FEA and that measured in the
centrifuge test is very encouraging.
Figure 5: Simulation of Passive Suction

0.8
FEA RESULTS AND COMPARISON
Total Pullout Force (kN, x 0.225 kips)

0.7
Validation of Soil/Caisson Interaction 0.6
SAT06-3
In order to validate the soil/caisson interaction, the caisson top was 0.5 FEA
separated from the mudline by a gap. The poro-elastic elements
described in the previous section were not used in this first round of 0.4
calculations. The passive suction measured during the centrifuge test 0.3
was applied on both the mudline nodes as negative pore pressures and
in the inside top surface of the caisson as downward pressures. 0.2

0.1
SAT06-1 FEA and Experimental Results
0.7
Test result 0.0
0.6 FEA result 0.0 0.1 0.2 0.3 0.4 0.5
Total Pullout Force (kN, x 0.225 kips)

Pullout Distance (m, x 3.3 ft )

0.5

0.4
Figure 7: Comparison of Total Pullout Force Development

0.3 Figure 8 presents a comparison between the passive suction

development observed in FEA and that measured in the centrifuge test.
0.2 As can be seen, the FEA results indicated a faster development of the
passive suction as compared to that measured in the centrifuge test.
0.1
Also, the passive suction obtained from the FEA started to drop off
0 after it reached the peak. The peak/maximum passive suction measured
0 0.5 1.0 1.5 2.0 2.5
Pullout Distance (m, x 3.3 ft)
in the centrifuge test, on the other hand, remained constant for a
relatively large vertical displacement.
Figure 6: Validation of Soil/Caisson Interaction
As shown in Figure 8, the passive suction development measured in the
Figure 6 presents the calibration of the FEA results with those centrifuge test is somehow different from that predicted by the FEA
measured in the centrifuge test. As can be seen, the total pullout forces due to the limitations of the numerical model. These limitations include

798
the simulation of the discontinuous soil materials using a continuum Aas P.M. and Andersen K.H. “Skirted Foundation for Offshore
medium and Cam-Clay model in extension. Moreover, the intention in Structures”, 9th Offshore South East Asia Conference, 1992.
the FEA was to use the soil properties of the clay testbeds used in the American Petroleum Institute (API), Recommended Practice for
centrifuge tests, and not to fit the curves. Planning, Designing, and Constructing Fixed Offshore Platforms-
Load and Resistance Factor Design, API Recommended Practice
2A-LRFD (RP 2A-LRFD). 1st Ed., API, Washington, D.C, 1993.
Comparison of Centrifuge and FEA Results Cao J. “Centrifuge Modelling and Numerical Analysis of the Behaviour
180 of Suction Caissons in Clay”. Ph. D thesis, Memorial University
of Newfoundland, St, John’s, NF, Canada, 2002.
Passive Suction, kPa (1 kPa = 0.145 psi)

160 Cao J., Phillips R. and Popescu R. “Physical and Numerical Modelling
SAT06-3 on Suction Caissons in Clay”, 18th Canadian Congress of Applied
140 Mechanics CANCAM 2001 Memorial University of
FEA
Newfoundland June 3-7, 2001, pp.217-218.
120
Cao J., Phillips R., Popescu R., Al-Khafaji Z. and Audibert J.M.E.
100
“Penetration Resistance of Suction Caissons in Clay”, the Twelfth
International Offshore and Polar Engineering Conference,
80 Kitayushu, Japan, May 26-31, 2002.
Clukey E.C. and Morrison M.J., “A Centrifuge and Analytical Study to
60 Evaluate Suction Caissons for TLP Applications in Gulf of
Mexico”, Design and Performance of Deep Foundations: Piles and
40 Piers in Soil and Soft Rock, Proceedings, Dallas, Texas,
Geotechnical Publication No 38, 1993.
20 Deng W. and Carter J.P., “Inclined Uplift Capacity of Suction Caissons
in Sand”. 2000’ OTC 12196, Houston, Texas, May 1-4 2000.
0 Deng W. and Carter J.P., “Uplift Capacity of Suction Caissons in
0.0 0.1 0.2 0.3 0.4 0.5 Uniform Soil”, Personal Communication, 2000.
Pullout Distance, m (1 m = 3.28 ft) Handayanu, Swamidas A.S.J. and Booton M., “Ultimate Strength of
Offshore Tension Foundations under Vertical and Inclined
Loads”, Proceedings of OMAE 2000, 19th International
Figure 8 Comparison of Passive Suction Development Conference on Offshore Mechanics and Artic Engineering,
February 14-17, 2000, New Orleans, Lpuisiana, USA.
Hibbitt, Karlsson & Sorensen, Inc., “ABAQUS/Standard User’s
CONCLUSIONS
Manual, Volume I, Version 5.8, 1998.
Norwegian Geotechnical Institute (NGI), “Set-up Effects Outside Skirt
An attempt to simulate the passive suction that develops at the top of a
Wall – Skirted Foundations and Anchors in Clay”, Joint Industry
suction caisson using the FE method was presented in this paper. The
Sponsored, May 11, 1999.
water inside the caisson was replaced by a soft porous material. The
Popescu R., “Finite Element Analysis of Pipe/Soil Interaction Phase 1 –
passive suction was simulated by negative pore pressures generated by
Two-dimensional Plane Strain Analyses”, Contract Project Report
tensile strains, and the suction was transferred to the caisson top by the
for the Geological Survey of Canada. C-CORE Publication 99-
soft porous material. Contact surface elements were used to model the
C23, June 1999.
behavior of the soil/caisson interaction. Considering the resolution and
Zdravkovic L., Potts D.M. and Jardine R.J., “Pull-out Capacity of
“scatter” in the displacement measurements in the centrifuge model
Bucket Foundations in Soft Clay”, Offshore Site Investigation and
test, the agreement between the total pullout force versus pullout
Foundation Behaviour ’98, 1998.
displacement curve obtained from the FEA and that measured in the
centrifuge test is very encouraging.

Although the passive suction curve generated by the FEA was slightly
different from that recorded in the centrifuge tests due to the limitations
of the numerical model, the maximum values of the suction were very
close. As this is the first time that this model was used to simulate the
passive suction, its improvement is the subject of future research.

Other FEA results, such as the development and distribution of

effective stress in the soil with the corresponding centrifuge test results,
will be presented at the conference and future publications.

ACKNOWLEGDEMENT

The first author would gratefully appreciate the computational facilities

by the CCAE, Faculty of Engineering and Applied Science, Memorial
University of Newfoundland, St. John’s, NF, Canada.

REFERENCES

799

View publication stats