Professional Documents
Culture Documents
Abstract: Offshore structures for oil and gas exploitation are designed to accommodate severe environments with large cyclic loads. These
structures are either founded directly on the seabed, or they are moored to anchors installed in the seabed soil. The permanent and cyclic
loading, the foundation or anchor geometry, and the nonlinear soil behavior may be very complex, and many interrelated aspects must be
Downloaded from ascelibrary.org by Auburn University on 09/10/13. Copyright ASCE. For personal use only; all rights reserved.
considered in the geotechnical design of the foundations. Finite-element analyses (FEAs) are used increasingly to deal with these complex-
ities and offer the potential to increase accuracy, efficiency, and reliability and reduce the uncertainty of the design process. This paper
presents the major geotechnical aspects in the design of foundations for offshore structures and examples from finite-element analyses carried
out at the Norwegian Geotechnical Institute (NGI) to deal with these aspects. A brief review of the procedure used at NGI to obtain soil
stress-strain-strength relationships from cyclic laboratory tests is included. It is demonstrated that FEA offers several benefits over classical
methods, such as limiting equilibrium calculations. DOI: 10.1061/(ASCE)GM.1943-5622.0000020. © 2011 American Society of Civil
Engineers.
CE Database subject headings: Finite element method; Offshore structures; Foundations; Design; Cyclic Loads.
Author keywords: Finite-element method; Offshore structures; Foundation design; Cyclic loads.
Fig. 2. Major geotechnical aspects considered in design of foundations for offshore structures
Fig. 3. Loading and soil response for an offshore gravity base structure: (a) plan view of foundation and loading; (b) cross section with simplified soil
stress conditions; (c) design load parcels; (d) definition of average F a and cyclic F cy load components
following the cyclic load history with varying load intensity. For
clays, i.e., undrained conditions, N eqv may be determined by keep-
ing track of the cyclic shear strain during the cyclic load history by
the “strain accumulation” procedure (Andersen et al. 1992). For
sands or conditions with drainage, N eqv can be determined by keep-
ing track of the permanent pore pressure accumulated during the
cyclic load history (Jostad et al. 1997; Andersen et al. 1994). The
reason for using the pore pressure accumulation procedure for sand
is that some drainage is likely to occur during the load history in
sand. It is assumed, however, that drainage does not have time to
occur within each cycle. To account for drainage, it is necessary to
keep track of the accumulated pore pressure.
The accumulation procedures use strain or pore pressure con-
tour diagrams of the type presented in Fig. 7 and storm load
compositions of the type presented in Fig. 3(c). The diagrams in
Fig. 7 were established based on the same laboratory tests as the
ones used to establish the diagrams in Fig. 5.
In principle, a constitutive model could be formulated that fol-
lows each individual cycle and used in a time domain FE analysis
with the complete load history. At the present, NGI does not have
such a model for that purpose that is sufficiently robust and fits
laboratory data accurately enough. Instead, the relationships given
in Figs. 5–7 are used in the constitutive models that are imple-
mented into the FE programs used for offshore foundation design
at NGI. Because the average shear stress and the cyclic strength are
functions of the cyclic shear stress, it is necessary to know the
cyclic shear stress when entering the diagrams in Figs. 5–7. This
is done by calculating the shear stresses in the soil as functions of
the cyclic loads F cy . In FEA, iterations are performed to update
τ a ¼ τ a ðτ cy Þ as the τ cy distribution changes (Jostad and Andresen Fig. 4. Shear stress, shear strain, and pore pressure during cyclic
loading
2009).
30% of the maximum waves were used because a lower load was
found to have no degrading effect on the strength. This resulted
in N eqv ¼ 10 and the shear strength profiles that are shown in
Fig. 8(b). The figure shows the cyclic shear strength for τ a ¼ 0
for the clay layer (layer II) and the sand layer (layer III) below
the platform and outside the platform. The wave loading is two-
way symmetrical with F a ¼ 0, which results in τ a ¼ 0 on the Fig. 8. FE calculations for bearing capacity of a gravity base platform:
horizontal plane that constitutes nearly 90% of the failure mecha- (a) element discretization and material layering below and outside of
nism. The strength below the platform is higher than outside the foundation; (b) cyclic shear strength τ f ;cy profiles for triaxial compres-
platform footprint because of the increased effective stresses from sion (TXC), direct simple shear (DSS), and triaxial extension (TXE)
the platform weight. The effect of the platform weight load spread
(distribution of Δσ0v below the platform) is accounted for by an
approximated simple 1∶3 rule, as shown on Fig. 8(a). but with different meshes with varying numbers of elements. The
The resulting failure mechanism is shown in Fig. 9 and consists results are shown in Fig. 10. It is seen that, by increasing the
mainly of sliding in the y-direction along the top of the weaker layer number of elements [reducing the average element size (AES)],
at 5 m depth. Passive and active earth pressure resistance develop the calculated capacity is reduced. It is possible to fit a curve to
along the 200-m-wide and 5-m-deep windward and leeward planes. the results that can be extrapolated to AES ¼ 0 (infinite number
Side shear develops along the 100-m-wide and 5-m-deep side of elements), where the discretization error vanishes.
planes. This characterizes the geometry of the main failure mecha- The example from the section on the bearing capacity of a
nism; however, torsion, overturning moments, and the uneven gravity base platform demonstrates that 3D FEAs are suited for
distribution of seabed pressure also affect the mechanism. Fig. 9(a) calculating bearing capacities of offshore foundations with multi-
shows that the torsion causes some rotation of the foundation. directional loading and complex geometries, such as skirted bases.
However, one should be aware that there may be a need for a
Discretization Error very fine mesh discretization to avoid overshooting. Furthermore,
care should be taken because little experience and only limited
Capacity calculations carried out by FEA will always contain dis-
validation examples exist for the application of 3D FEAs to such
cretization errors, i.e., there will be an overshoot in the calculated
problems.
ultimate capacity. This error is a result of the simplification done in
FEA, where the continuous displacement field is represented by a Holding Capacity of Suction Anchors
finite number of element interpolations. The discretization error can
be reduced by using a finer mesh, with the penalty of an increased An industry sponsored study on the design and analysis of
computational cost. It is not possible to quantify the error from the deepwater anchors in soft clay was completed in 2003, in which
result of one analysis with only one mesh realization, and often one NGI participated with the Offshore Technology Research Center
must then rely on experience from similar problems to judge if the (OTRC) in the United States and the Center for Offshore Founda-
discretization error is acceptable. tion Systems (COFS) in Australia. The study is documented in
For the problem presented in the section on the bearing capacity Andersen et al. (2005). Independent 3D finite element analyses
of a gravity base platform, the discretization error was quantified for several hypothetical cases were performed by NGI, COFS,
by calculating the horizontal load capacity with the same FE model and OTRC.
It can be seen that the anchor is not rotating and that the mechanism
involves some of the soil underneath the skirt tip level.
The overall conclusion from this study was that the 3D FEA
results were in good agreement. The difference in capacity calcu-
lated by the different groups was generally less than 3%, and the
capacities were approximately 10% higher than the results NGI ob-
tained by using limiting equilibrium methods. Hence, it was dem-
Fig. 10. The effect of mesh refinement in FEA: Horizontal load capa- onstrated that by 3D FEA, reliable results were obtained that were
city versus average element size (number of elements); dotted curve less conservative than the results obtained by limiting equilibrium
represents proposed extrapolation towards AES ¼ 0 calculation. From this and other studies, it has been found that it is
important to model the anisotropy of the undrained shear strength
and the reduced undrained shear strength along the outside skirt
The case, C2, consisted of an open cylindrical anchor with wall adequately. It has also proven to be very efficient to use special
weight W 0 ¼ 300 kN, diameter B ¼ 5 m, and depth D ¼ 7:5 m, zero-thickness interface elements along the skirt outside and under-
giving a depth-to-diameter ratio of D=B ¼ 1:5. The soil cyclic shear neath the skirt tip. If such elements are not used, it may be neces-
strength (τ f cy ¼ su ) data was: sDSS
u ¼ 1:25 × z, sCu ¼ 1:2 × sDSS
u , sary to use an extremely fine mesh discretization in these areas to
su ¼ 0:8 × su
E DSS and with su along the outside skirt wall = allow for a possible full slip between the anchor and the soil.
0:65 × sDSS
u . Fig. 11(a) shows holding capacities for this case calcu-
lated by the three groups (NGI, UWA, and OTRC) for various load Capacity of Skirted Foundations on a Steel Jacket
Platform
angles from pure horizontal to pure vertical loading. In each case the
load is attached to the anchor at the optimal attachment point such The Draupner-E platform is a steel jacket located 160 km off-
that the anchor is prevented from rotating. shore from Norway at a water depth of 70 m. The structure and
Fig. 11(b) shows the failure mechanism for a load angle of 45° the experience of installation of the jacket are documented by
calculated by NGI using the program BIFURC 3D (NGI 1999). Tjelta (1995). Unlike most jacket platforms, which are supported
Fig. 13. Idealized cyclic vertical load history and calculated vertical
displacements of leeward leg of a jacket during a 6-h peak storm period
Fig. 12. (a) Draupner jacket being lifted onto a barge (Image courtesy By using this constitutive model together with a coupled stress
of Leif Berge, Statoil); (b) uplift capacity of skirted foundation versus equilibrium and pore water flow (consolidation) finite-element for-
vertical displacements for different loading rates calculated by fully mulation, it is possible to analyze pore pressure accumulation and
coupled FEA dissipation problems. The procedure is described in more detail in
Jostad et al. (1997). As a validation of the procedure, the response
of a bucket foundation resembling the ones for the Draupner plat-
by piles, Draupner is founded on steel bucket foundations (see form was calculated. The average vertical load on each foundation
Fig. 12(a)). prior to the storm loading was 10 MN. Fig. 13 shows the idealized
The four cylindrical foundations, each 12 m in diameter, are cyclic load history, with increasing average V a and cyclic V cy ver-
equipped with 40-mm-thick steel skirts that penetrate 6 m into tical load, and the calculated vertical displacements (maximum,
the seabed by applied suction. The soil conditions consist of very minimum, and permanent) during a 6-h peak design storm period.
dense sand with a relative density Dr in the range 90–100% in the The results are for the skirted foundation that experiences increased
top 23 m. average vertical load during the storm (leeward leg). The horizontal
Whereas clay generally is undrained during a typical storm his- load component is assumed to be taken by the less mobilized foun-
tory, sand may respond fully drained during changes in the average dations. It can be seen that the failure mode is the development of
load and, at the same time, respond undrained to the short-term large vertical settlements.
loading from single wave cycles. The long-term drained vertical
uplift capacity of these foundations is, therefore, quite low and con-
sists only of drained inside and outside wall friction. However, the Installation
capacity for the short-duration (T p ∼ 11 s) wave load with a typical
Suction anchors and skirted gravity base foundations have steel or
rate of 0:2 MN=s is much higher, as illustrated in Fig. 12(b).
concrete skirts that protrude into the soil during installation. The
NGI has implemented a special constitutive model for the analy-
skirts penetrate down into the soil as a result of the weight of the
sis of such jackets with skirted foundations in sand in the in-house
structure or a combination of the self-weight and an applied under-
finite element program BIFURC (NGI 1999). The main parameter
pressure under the base. The installation method and process affect
in this model is the accumulated pore pressure ua as a function
important aspects of the design such as the penetration resistance,
of the cyclic shear-stress amplitude τ cy and number of cycles N. the distribution of contact stresses between the foundation structure
The soil response for the average load F a is calculated using the and the soil, and the shear strength along the skirts (friction capac-
mobilized friction model (Nordal et al. 1989). The accumulated ity of the skirts).
pore pressure is calculated based on the cyclic shear-stress ampli- Numerical analysis of installation processes such as the penetra-
tude τ cy , calculated from the cyclic loads F cy in a separate analysis, tion of a steel skirt into the seabed is extremely challenging. Ideally,
with input of the actual equivalent number of cycles in each inte- the analysis should account for a large deformation, a continuously
gration point N eq . The τ cy may be calculated by a nonlinear elastic changing contact area, and remolding of the soil. Methods that
constitutive model expressing the relationship between τ cy and γcy , handle large deformations such as the updated Lagrangian (UL) or
shown in Fig. 5. The equivalent number of cycles is found from the material point method (MPM) (Coetzee et al. 2005; Beuth et al.
pore pressure contour diagrams of the type shown in Fig. 7(b). 2007) are promising but still under development and not used
Fig. 16. Troll-A platform with its concrete gravity base structure and
Abaqus model for calculation of the rotational stiffness of foundation
Fig. 15. Results from suction anchor skirt penetration: (a) build-up of
horizontal stresses outside tapered skirt tip during self-weight penetra-
tion; (b) mean stress distribution in horizontal cross section at 13.7 m
depth for situation where skirt is being penetrated to its final depth
of 16 m
Fig. 20. (a) Cam-clay type of stress-strain relationship; (b) 1st year For the structural design of the foundations, the distribution of
loading history and vertical settlements for gravity base platform reaction stresses on the foundation for the different loading condi-
(GBS) and nearby seabed, calculated by FEA tions must be known. Aspects that are of particular interest may be
the distribution of contact stress underneath the base and how much
of the load is carried by the skirt wall and skirt tip compared to the
base. Normally, soil reactions are provided to the structural engi-
seabed in the vicinity of the platform. The settlement of the seabed neers in the form of a number of possible distribution diagrams for
was used as input in the design of a pipeline connection to the a set of unit load cases. There are aspects such as uneven seabed,
platform. installation effects, and redistribution with time that make it very
The input data for the analysis are the geometry, soil layering, difficult to accurately calculate reliable distributions. The reactions
soil permeability and stiffness, drainage conditions, and the load are, therefore, in most cases, based on engineering judgment and
history. The permeability and stiffness are both stress-dependent. conservative estimates to provide a robust structural design. FEA
Cam-clay-type models like the one shown in Fig. 20(a) are well- may, however, provide valuable insight into the mechanisms of load
suited for representing stress-dependent stiffness and have memory transfer between the foundation and the soil. For skirted gravity
that accounts for preconsolidation stress. base foundations such as the one shown in Fig. 21(a), the main
Fig. 20(b) shows the first year load history and calculated set- interest is to assess the fraction of the submerged weight carried
tlements for the gravity base platform and for points on the seabed by base contact stresses, skirt friction, and skirt tip resistance,
respectively. A reasonable estimate may be provided by applying
5, 10, 15, and 20 m off the platform. The fully coupled pore pres-
the submerged weight W 0 to a FE model such as the one shown
sure dissipation and equilibrium analyses were performed using the
in Fig. 21(a) for the subsoil and the foundation with base and
Plaxis FE program with a cam-clay-type material model. The load
skirts.
history reflects the gradually increasing ballast weight during the
first 90 days after installation. One cycle of on- and off-loading Redistribution during Cyclic Loading
represents the situation where the platform is filled to maximum
The loads will, however, redistribute with time. During cyclic load-
weight with liquid natural gas and then off-loaded to the average ing, there will be a tendency for redistributing weight from the
weight. It can be seen that the platform settles ~80 cm during skirts to the base. This is a result of the degradation of strength and
the first year, whereas the surrounding seabed generally settles stiffness for combinations of average stress (caused by the weight)
10–20 cm. The pipeline connection was then designed for a relative and cyclic stress. A constitutive model for cyclic loading has been
settlement of 70 cm over the nearest 20 m from the platform. Other developed at NGI in which the input is diagrams of the type shown
important components contributing to the total settlement of off- in Fig. 6. This model has been used in FEA of this load transfer
shore foundations, like immediate settlements, creep, and effects mechanism from skirt to base. The mechanism is illustrated in
of cyclic loading, are not considered here. Fig. 21(b). The soil below the skirt tip is highly mobilized as a
Soil-Structure Interaction
Acknowledgments
References
Andersen, K. H., and Lauritzsen, R. (1988). “Bearing capacity for founda-
tion with cyclic loads.” J. Geotech. Eng., 114(5), 540–555.
Andersen, K. H., and Jostad, H. P. (2002). “Shear strength along outside
wall of suction anchors in clay after installation.” Proc., 12th Int. Off-
shore and Polar Engineering Conf., International Society of Offshore
and Polar Engineers (ISOPE), Cupertino, CA, 26–31.
Andersen, K. H., Dyvik, R., Lauritzen, R., Heien, D., Hårvik, L., and
Amundsen, T. (1989). “Model tests on gravity platforms. II: Interpre-
tation.” J. Geotech. Engrg., 115(11), 1550–1568.
Andersen, K. H., Dyvik, R., Kikuchi, Y., and Skomedal, E. (1992). “Clay
behaviour under irregular cyclic loading.” Proc., Int. Conf. on Behavior
of Offshore Structures, Vol. 2, 937–950.
Andersen, K. H., Dyvik, R., Schrøder, K., Hansteen, O. E., and Bysveen, S.
(1993). “Field tests of anchors in clay II: Predictions and interpretation.”
J. Geotech. Engrg., 119(10), 1532–1549.
Andersen, K. H., Allard, M. A., and Hermstad, J. (1994). “Centrifuge
model tests of a gravity platform on very dense sand; II: Interpretation.”
Proc., 7th Int. Conf. on Behavior of Offshore Structure, Vol. 1, MIT,
Cambridge, MA.
Andersen, K. H., Andresen, L., Jostad, H. P., and Clukey, E. C. (2004).
“Effect of skirt-tip geometry on set-up outside suction anchors in soft
clay.” Proc., 23rd Int. Conf. Offshore Mech. Artic Eng., Vol. 1, ASME,
New York, 20–25.
Andersen, K. H., et al. (2005). “Suction anchors for deepwater applica-
tions.” Proc., Int. Symp. on Frontiers in Offshore Geotechnics, Taylor
Fig. 23. (a) Calculated leg-moment at lower guides versus load
& Francis, Oxford, UK.
factor for spud-can with and without skirts in stiff clay; (b) cyclic hor- Andresen, L., Andersen, K. H., Jostad, H. P. J., and Rahim, A. (2007).
izontal footing load versus total vertical footing load for spud-can “Bearing capacity of offshore gravity platforms by 3D FEM.” Proc.,
with skirts in clay when loaded in direction that gives one single 10th Int. Symp. on Numerical Methods in Geomechanics, Taylor &
leeward leg Francis, Oxford, UK, 509–515.
Beuth, L., Coetzee, C. J., Bonnier, P., and van den Berg, P. (2007).
“Formulation and validation of a quasi-static material point method.”
Proc., 10th Int. Symp. on Numerical Methods in Geomechanics, Taylor
Summary and Conclusions
& Francis, New York.
Coetzee, C. J., Vermeer, P. A., and Basson, A. H. (2005). “The modeling of
In this paper, various aspects of the design of foundations and an- anchors using the material point method.” Int. J. Numer. Anal. Meth.
chors for offshore structures used in oil and gas exploitation have Geomech., 29(9), 879–895.
been presented. The main difference between onshore and offshore Cundall, P. A., and Strack, O. D. L. (1979). “A discrete numerical model for
foundation design is that the offshore foundations are always sub- granular assemblies.” Geotechnique, 29(1), 47–65.
jected to cyclic loading which may cause soil strength and stiffness Hansen, B., Nowacki, F., Skomedal, E., and Hermstad, J. (1992).
degradation. In offshore geotechnics, we are also mostly dealing “Foundation design, Troll platform.” Proc., Int. Conf. on Behavior
of Offshore Structures, Vol. 2, 921–936.
with saturated soil.
Jostad, H. P., and Andersen, K. H. (2006). “Potential benefits of using
A framework for accounting for the cyclic load history when skirted foundations for jackup platforms.” Proc., Offshore Technology
determining the static and cyclic soil stress-strain-strength relation- Conf., Houston, Paper No. 18016.
ship has been developed at NGI and has been briefly presented Jostad, H. P., and Andresen, L. (2009). “A FE procedure for calculation of
in the section of this paper regarding cyclic loading and soil displacements and capacity of foundations subjected to cyclic loading.”