Calculation Procedures for Installation of Suction Caissons

by

G.T. Houlsby and B.W. Byrne


Report No. OUEL 2268/04




















University of Oxford
Department of Engineering Science
Parks Road, Oxford, OX1 3PJ, U.K.

Tel. 01865 273162/283300
Fax. 01865 283301
Email Civil@eng.ox.ac.uk
http://www-civil.eng.ox.ac.uk/

1
Design procedures for installation of
suction caissons in clay and other soils
G.T. Houlsby
1
and B.W. Byrne
1


Keywords: clay, foundations, suction caissons, installation
Abstract
Suction-installed skirted foundations, often referred to as suction caissons, are being increasingly
used for a variety of offshore applications. In designing a caisson a geotechnical engineer must
consider the installation process as well as the in-place performance. The purpose of this paper is to
present calculation procedures for the installation of a caisson in clay. For clay sites, the caisson
will often be used as an anchor, with the ratio of the skirt length (L) to the diameter (D) as high as 5.
Calculation methods are presented for determining the resistance to penetration of open-ended
cylindrical caisson foundations with and without the application of suction inside the caisson.
Comparisons of between predictions and case records are made. A companion paper (Houlsby and
Byrne, 2004) describes the calculation procedure for installation in sand soils. Finally comments are
made here about installation in s variety of soils other than homogeneous deposits of clay or sand.
Introduction
A suction caisson is a large cylindrical structure, usually made of steel, open at the base and closed
at the top. It might be used either as a shallow foundation or as a short stubby pile (often called a
suction anchor). The shallow foundation option is more common at sandy soil sites (e.g. Bye et al.,
1995; Hogervost 1980; Tjelta, 1994; Tjelta, 1995; Tjelta et al., 1990) whilst the anchor/pile
application is commoner in clay or layered soils (e.g. Colliat et al., 1996; Colliat et al., 1998;
Erbrich and Hefer, 2002; Lacasse, 1999; Solhjell et al., 1998). Figure 1 shows typical diameter and
skirt depths for various projects reported in the literature (the figure is taken from Byrne (2000) and
with further data from Tjelta (2001)). More recently there is an emerging application of caissons as
the foundations for offshore wind turbines (Byrne et al., 2002; Byrne and Houlsby, 2003).

This paper addresses installation in clays and other soils whilst a companion paper (Houlsby and
Byrne, 2004) considers installation in sand. In the anchor application the caisson will be designed
so that the skirt length (L) is much greater than the diameter (D) and the ratio L/D might be as large
as 5 (as shown in Figure 1). As oil and gas exploration heads further offshore and into deeper water,
it is likely that anchor applications will become more common. There are particular advantages to
using the suction caisson over other anchoring methods (e.g. drag anchors), in that the caisson can
be accurately located, allowing complex mooring line arrangements to be accommodated. The
ability to remove a caisson (by simply reversing the installation procedure) allows mooring line
arrangements to be altered over the life of a production vessel; and removal at the end of the design
life.

After an initial penetration into the seabed caused by self weight, a suction (relative to seabed water
pressure) is applied within the caisson, which forces the remainder of the caisson to embed itself,
leaving the top flush with the seabed. The purpose of this paper is to present design calculations for
the installation of the caisson. Separate calculations are of course necessary to assess the capacity of
the caisson once installed either as a shallow foundation or as an anchor. Analyses are presented
for the magnitude of the self-weight penetration, the relationship between suction and further

1
Department of Engineering Science, Parks Road, Oxford OX1 3PJ
2
penetration, and the limits to penetration that can be achieved by suction. The analyses are
classical in the sense that they make simplifying assumptions, borrowing techniques from both
pile design and bearing capacity theory. More rigorous analyses, using for instance finite element
techniques, could be used for particular installations. The analyses presented here should, however,
provide a reasonable approximation for design purposes. Similar methods (although differing in
some details) to those described below have been published e.g. by House et al. (1999), but our
purpose here is to draw together a comprehensive design method and compare with case records
from several sources.
Installation in Clay
Figure 2 shows the key variables in the suction caisson problem, so far as the installation is
concerned. For the purposes of the installation calculation the strength of the clay is characterised
by an undrained strength, which is assumed to increases with depth linearly in the form
z s s
uo u
+ = . The methods described below can readily be adapted to more complex strength
variations.
Self-weight penetration
The resistance to penetration is calculated as the sum of adhesion on the outside and inside of the
caisson, and the end bearing on the annular rim. The adhesion terms are calculated, following usual
practice in pile design, by applying a factor to the value of the undrained strength. The end
bearing is calculated, again following standard bearing capacity analyses, as the sum of an
q
N and
an N
c
term. The result is:
( ) ( ) ( )( ) Dt N s h D s h D s h V
c u i u i o u o
+ + + =
2 1 1
(1)
adhesion on outside adhesion on inside end bearing on annulus

where 2
1
h s s
uo u
+ = is the average undrained shear strength between mudline and depth h,
h s s
uo u
+ =
2
is the undrained shear strength at depth h,
o
and
i
are adhesion factors on the
outside and inside of the caisson (as used in undrained pile design) and N
c
is an appropriate
bearing capacity factor for a deep strip footing in clay (typically a value of about 9 might be
adopted). For undrained analysis 1 =
q
N .
Suction-assisted penetration
Once the self-weight penetration phase has been completed, so that a seal is formed around the edge
of the caisson, it will be possible to commence the suction installation phase. The applied suction in
the caisson is s relative to seabed water pressure, i.e. the absolute pressure inside the caisson is
s h p
w w a
+ . There are a number of practical limits to the maximum attainable value of s.
Amongst these are (a) the absolute pressure at which the water cavitates (usually a small fraction of
atmospheric pressure), (b) the minimum absolute pressure that can be achieved by the given pump
design, (c) the minimum relative pressure that can be achieved by the pump.

The suction causes a pressure differential across the top plate of the caisson, which results
effectively in an additional vertical load equal to the suction times the plan area of the caisson. The
capacity is again calculated as the sum of the external and internal friction, and end bearing term.
Note that the overburden term is reduced in the end bearing calculation by the suction pressure,
assuming that the flow of soil under the rim occurs entirely inwards. The result is:
3
( ) ( ) ( ) ( )( ) Dt N s s h D s h D s h D s V
c u i u i o u o i
+ + + = +
2 1 1
2
4 (2)
which is readily rearranged to:
( ) ( ) ( ) ( )( ) Dt N s h D s h D s h D s V
c u i u i o u o o
+ + + = +
2 1 1
2
4 (2a)
Note that if the variation of soil strength is not simply linear, all that is necessary is to replace
1 u
s
with the average strength from mudline to depth h, and
2 u
s with the strength at depth h. Equation
(2) gives a simple relationship between suction and depth. For constant V and a linear increase of
strength with depth (so that
1 u
s and
2 u
s are linear functions of h), s is a quadratic function of h .
Limits to suction assisted penetration
In addition to the limit imposed by the maximum available suction, there is a limit to the depth of
penetration that can be achieved by the action of suction. If the difference between the vertical
stress inside and outside the caisson, at the level of the caisson tip, exceeds a certain amount, then
local plastic failure may occur, and further penetration may not be possible. The mechanism may be
thought of as a reverse bearing capacity problem, in which the soil flows into the caisson.

The average vertical stress (relative to local hydrostatic) inside the caisson at tip level is relatively
straightforward to estimate as
4
2
1
i
u i i
D
s h D
h s

+ + . The third term in this expression arises from
the downward friction inside the caisson, and here it is assumed (for simplicity) that this results in a
uniform increase of vertical stress at all radii in the caisson. Note that the assumption of a uniform
increase in vertical stress within the caisson is clearly unreasonable at small values of D h , but it
will be seen below that this calculation is only needed at D h values greater than about 2, for
which the uniform increase is a reasonable approximation.

The relevant stress outside the caisson is much harder to estimate, since the downward load from
adhesion on the outside of the caisson will enhance the stress in the vicinity of the caisson, but this
enhancement is difficult to calculate. However, making the simplifying assumption that the
downward load from the adhesion is carried by a constant stress over an annulus with inner and
outer diameters
o
D and
m
D , this stress (again relative to local hydrostatic) may be calculated as
( ) 4
2 2
1
o m
u o o
D D
s h D
h


+ . Thus the reverse bearing capacity failure would occur when
( )
2
*
2 2
1
2
1
4 4
u c
o m
u o o
i
u i i
s N
D D
s h D
h
D
s h D
h s


+ =

+ + , where
*
c
N is a bearing capacity factor
appropriate for uplift of a buried circular footing. Substituting the solution for s into equation (2)
and simplifying gives:

( )
( )
( )( ) Dt N s h
D D
D
D s h
D
s N V
c u
o m
i
o u o
i
u c
+ +

+ =

+
2
2 2
2
1
2
2
*
1
4
(3)
which can be solved for h. Note, however, that although the above equation appears linear in h, in
fact
1 u
s and
2 u
s are themselves linear functions of h, so that the solution again involves solving a
quadratic. Furthermore it would be rational to assume that
m
D increases with penetration, for
instance in the form h f D D
o o m
2 = , where
o
f is a constant loadspread factor. A further
4
development would be to allow the enhancement of the stress to vary (say linearly) from zero at
m
D to a maximum at the caisson surface.

It is worth, however, considering some approximate solutions for the maximum penetration. For
many cases the final term (the end bearing) is small. We consider also the case where the applied
load V is small, and make the approximation
i o
D D D . If we write mD D
m
= , then equation
(3) leads to the following result for this simplified case:

2
1
2
*
1
1
4
m
s
s N
D
h
u
u
o
c
(4)
The factor
o c
N 4
*
is likely to be in the region of about 3, although it could vary considerably.
The factor
1 2 u u
s s would be 1.0 for a homogeneous soil, and 2.0 for the extreme of a soil with a
strength increasing linearly with depth from a value of zero at the surface. The final factor varies
from 1.0 if m is assumed to be very large, to 0.75 if say 2 = m . The overall result is that the
calculated maximum attainable value of D h is likely to be from about 2.5 for stiff clays (with
strengths approximately uniform with depth) to 5 for soft normally consolidated clays (with
strengths approximately proportional to depth). The effect of accounting for the external load V
would be to increase these values. Equation (4), however, provides a useful estimate of the
maximum D h ratio of a suction-installed caisson that could be reliably installed in clay. If
different assumptions are made about the way the external adhesion load enhances the vertical
stress are made, the same broad conclusions arise, although the precise figures will vary.

It should be noted that some measured values of installations indicated that higher D h ratios than
implied by the above calculation may be achievable. The above may therefore be treated as a
conservative calculation.

Note also that the end bearing calculation in equations (1) and (2) does not take into account any
enhancement of the stress level inside or outside the caisson due to the frictional terms. This follows
conventional piling design calculations, in which no such correction is usually included. If this
effect was to be taken into account, the factor h in equation (1) would be replaced by whichever
is the smaller of
( ) 4
2 2
1
o m
u o o
D D
s h D
h


+ or
4
2
1
i
u i i
D
s h D
h

+ (almost invariably the former). Once
suction is started, s h in equation 2 is replaced by the smaller of
( ) 4
2 2
1
o m
u o o
D D
s h D
s h


+ or
4
2
1
i
u i i
D
s h D
h

+ (usually the latter except at very small suction). In practice these changes make
very small differences to the calculation.
The effect of internal stiffeners
Most suction caissons include some internal structure, usually consisting of either vertical plates or
annular plates, to provide strength and stiffness to the cylindrical shell, either to suppress buckling
during suction-assisted penetration, or (in the case of a caisson anchor) to reinforce the caisson at
the pad-eye connection. The analysis for the case of annular stiffeners is not considered here, but
the use of vertical stiffeners results in only a small change in the calculation.

5
In principle stiffeners could be located on the outside of the caisson, but this option does not usually
seem to be adopted. The additional resistance offered by the stiffeners can be taken into account by
an adhesion term of the form l s h
u1
, where l is the perimeter length of the stiffeners (usually
approximately twice the plate length for thin plate stiffeners), and an end bearing term of the form
( )A N s h
c u2
+ , where A is the end area of the stiffeners. The area on which the suction acts (on the
left side of Equation 2) should also be reduced by A, although this correction will usually be tiny.

Note that if the stiffeners do not extend the full depth of the caisson, appropriate corrections are
required for the value of h used in the contribution from the stiffeners, and in the appropriate
1 u
s
and
2 u
s values.

In the calculation of the maximum attainable depth using suction, note that the terms involving
adhesion on the inside of the caisson cancel, and have no overall effect on the calculation. The same
is true for terms resulting from the resistance from internal (but not external) stiffeners, so for
internal stiffeners only Equation (4) can still be used.
Example 1
Consider a suction caisson of outside diameter 12m, wall thickness 45mm and depth 5m. Such a
caisson might be considered as a foundation for an offshore structure. The caisson is stiffened by 30
plates 25mm thick and 200mm deep welded as radial fins on the inside of the caisson, and
extending for the top 4m of the caisson only. The soil profile is idealised as a layer 2m thick of
constant strength 20kPa, with below that a linear increase of strength from 25kPa at 2m at a rate
2.5kPa/m. The buoyant unit weight is taken as 6kN/m
3
. The end bearing factor
c
N is taken as 9,
and the adhesion factor as 0.6 for the outside of the caisson and 0.5 for inside and for the
stiffeners. The maximum applied vertical load (including the weight of the caisson and buoyancy
effects) is 1000kN and the water depth is 50m.

The calculations described above have been implemented in a spreadsheet-based program SCIP
(Suction Caisson Installation Prediction). Figure 3 shows the calculated loads required to install the
caisson in the absence of suction. Figure 4 shows the predictions from SCIP of the variation of
suction with depth required for installation, and in this case the maximum suction required is 49kPa.
Example 2: Predicted of installation pressures compared to centrifuge tests
House and Randolph (2001) conducted a series of tests on the centrifuge at the University of
Western Australia, investigating the installation of suction caissons in normally consolidated clay.
The experiments were carried out at 120g. The strength profile of the clay could be idealised as zero
at the surface increasing with depth at a gradient of 144kPa/m to a depth of 67mm then at 204kPa/m
(at prototype scale these represent rates of increase of 1.2kPa/m and 1.7kPa/m). The effective unit
weight of the soil (accounting for the 120g acceleration) was determined to be 792kN/m
3
. The
dimensions of the caisson were 30mm diameter, 0.5mm wall thickness and 120mm skirt length.
(Equivalent prototype dimensions 3.6m diameter, 60mm wall thickness, 14.4m skirt length). An
effective vertical load of 15.3N was applied to the caisson.

Figure 5 shows the penetration resistance for the caisson without using suction, showing that most
of the resistance is in the skirt friction. Figure 6 shows an estimated suction penetration curve,
which shows good agreement with the experimental data reported by House and Randolph (2001).
The self-weight penetration amounts to 41mm and the maximum suction pressure required is
143.9kPa. An adhesion factor of 0.5 was used for both internal and external walls.
6
Example 3: Prediction of plug failure
A series of tests were conducted by House et al. (1999) on the laboratory floor to investigate plug
failure during installation of suction caissons in normally consolidated clay. They investigated three
caissons with diameters 10.4mm, 15.9mm and 37.2mm. All caissons had a wall thickness of 0.4
mm and an L/D ratio of 8.

In the following a comparison is made for the 15.9mm diameter caisson. The soil strength profile
was estimated by House et al. (1999) to be 75kPa/m and the effective unit weight to be 5.9kN/m
3
.
The caissons were initially pushed into the clay to a penetration of approximately one diameter
before the suction was applied. Assuming a circular end bearing capacity factor of 8.5 (Houlsby and
Martin, 2003), the maximum calculated penetration by SCIP that is possible before a plug failure is
expected is 83mm or an h/D =5.2. This can be compared to the conclusions drawn by House et al.
(1999). They compare the volume of water withdrawn from the caisson cavity during installation to
the displaced volume within the caisson (assuming heave has not occurred). When more water is
evacuated than can be accounted for by the installed portion of the caisson, they infer that plug
heave has occurred. Figure 8 shows, for two installations of the 15.9mm diameter caisson, the
excess volume of water removed, plotted against normalised penetration. For the cases shown,
House et al. (1999) deduced that plug failure occurs at an L/D ratio between 4 and 5, which agrees
with the prediction given above. Again an adhesion factor of 0.5 was used. Note that although plug
failure occurred it was still possible to install the caisson further. Installation continues until all
water has been withdrawn from the internal cavity. The consequence of plug failure is that the
caisson cannot be installed to its full design depth.
Example 4: Nkossa Field Installation
This calculation involves some modification to the basic procedures described above to account for
the geometry of the caissons used in the Nkossa field off the coast of West Africa (Colliat et al.,
1996; Colliat et al., 1998). Two different anchor sizes were used depending on the loading
conditions. We will only consider the installation of the smaller of the two, defined by Colliat et al.
(1998) as a Type I anchor. The geometry of the caissons is unusual, as they have a step change in
diameter part way down the caisson. The bottom section is 4m in diameter and extends for 4.8m
whilst the top section is 4.5m in diameter and is 7.5m long. The anchor chain lug is located at the
change in caisson diameter. The wall thickness for the pipe sections was 15mm and the design
penetration was 11.8m. The larger top section was to accommodate any soil heave that occurred
during installation. Internal stiffening plates are also believed to have been used. However, these are
omitted in the calculation here, as there is insufficient information about the detailed geometry of
the stiffeners. The weight (in air) of the caisson is given as 41 tons which converts to a
submerged weight of approximately 350kN.

Colliat et al. (1998) give a summary of the soil conditions, which includes average shear strengths
as well as upper and lower bound strength envelopes. For the purpose of this calculation, the
average strength is taken, and is 5kPa at the surface increasing at 1.0kPa/m for the first 5m, below
which the gradient changes to 1.67kPa/m. The effective unit weight of the soil is taken as 6kN/m
3
.
Whilst Colliat et al. (1998) suggest an adhesion factor of 0.3 based on model scale field tests, but
the calculations here show an excellent agreement with the measurements if an adhesion factor of
0.45 (which seems quite reasonable) is used. To account for the effect of the increase in diameter of
the top section of the caisson, the internal adhesion factor was set to zero for the top section. End
bearing is also taken into account at the step between the two diameters. Figure 7 shows the suction
pressures required compared to the average and range measured during the field installation (on the
basis of data presented by Colliat et al., 1998). The slight underestimation of the required suction
may be because the stiffeners are not taken into account.

7
In the three example calculations where it is possible to compare with data, it is clear that a good
relation exists between predicted and observed behaviour, using reasonable estimates of soil
parameters. Obviously the key parameter that is required for predictions of caissons in clay is the
undrained strength profile, and an estimation of the adhesion factor .
Installation in other materials
Layered materials
Figure 1 shows that a number of installations have occurred in layered materials. We describe
briefly the issues that must be considered during the design for these sites.

Sand over Clay
The sequence of sand over clay probably would not cause problems for installation - typically the
installation would proceed through the sand (using the calculations given by Houlsby and Byrne
(2004)), and once into the clay the resistance would in most cases be lower, and could be calculated
using the same principles as for clay alone (although with a modification to the calculation of the
friction).

Clay over Sand
Clay over sand is likely to be more problematical. The caisson penetrates through sand when the
applied suction creates gradients in the sand which degrades the tip resistance to almost zero. The
pressure differential also provides a net downward force on the caisson, but this contributes less
significantly to the installation. Without the flow field in the soil it might be impossible to install the
caisson, due to the high bearing resistance of the sand (especially if it is very dense). During
installation in clay it is the net downward force caused by the pressure differential which causes the
caisson to be forced into the soil. When the installation occurs in a layered soil there are questions
as to whether the caisson will penetrate through a sand layer after it has passed through a clay layer,
as it will not be possible to develop the flow regime which degrades the skirt tip resistance to near
zero.

There are several field case studies which provide evidence that installation under these conditions
may, however, still be possible. The most notable is the large scale deepwater penetration test which
was conducted during the investigations for the Gullfaks C platform (Tjelta and Hermstad, 1986).
The soil profile consists of a number of layers of medium to dense sand and clay. The cone tip
resistances reach 20-24MPa in the denser sand layers, 4-10MPa in the medium sand layers and 1-
2MPa in the clay layers. The foundation consists of two 6.5m diameter cylinders joined by a
concrete beam, the structure being 22m in depth. A maximum suction of about 480kPa (linearly
increasing with depth) was required to install the caisson to its full depth. A water jetting system at
the caisson tip was used during the penetration of the initial sand layer, thus reducing the tip
resistance. Removal was also possible, requiring approximately 250kPa of overpressure (linearly
decreasing) at the maximum depth.

Further references to suction anchor installation in layered material can be found in Senpere and
Auvergne (1982) and Tjelta (2001). The former describe the installation in the Gorm field, where
soil plug failure occurred in all caissons. The installation was nonetheless successful as a jetting
procedure was used to remove material from within the caisson. Tjelta (2001) describes in issues
related to the Curlew, YME and Harding fields but does not give specific details.

Finely interbedded materials
There is no particular reason to suppose that finely interbedded materials would pose problems,
unless the composition of the beds differed in some extreme way. There are, however, no recorded
cases in such materials.
8
Stiff (possibly fissured) clay
There is a concern that it might not be possible to install suction caissons in stiff clays. The
principal reason is that, given that such materials are often fissured, or are prone to fissuring, it may
not be possible to form the necessary seal around the rim of the caisson for penetration to proceed.
One possibility is that fracturing may occur, with water simply flowing through the fissures. This
problem may be exacerbated by the fact that the penetration resistance in very stiff clays would be
high.

Information for this case is relatively scarce. In most cases where stiff clays have been encountered
(i.e. in the Visund, Njord and Aquila fields as discussed by Solhjell et al., 1998) the soil conditions
consisted of a layer of soft clay overlying much stiffer clay. In these cases it appears that the soft
clay layer is deep enough so that a seal could be created.

Whilst there is no evidence to support whether or not installation in stiff fissured clay is possible or
not, it should be noted that the condition where a stiff clay exists at mudline might be a rather scarce
occurrence.
Coarse materials
For obvious reasons, extremely heterogeneous materials would be likely to cause problems for
installation of a suction caisson. Materials with a significant fraction of coarse gravel or larger sizes
would almost certainly present an obstacle to installation. Certain (but not all) glacial tills would
therefore be problematical. Very open gravels, even if not particularly coarse, would present
problems in that flows during pumping would be very high.
Silts
It is difficult to do calculations for silts, because it is difficult to determine whether drained and
undrained behaviour would be appropriate, and partially drained calculations for caisson penetration
have not been formulated. However, given that penetration in clays and sands is relatively
straightforward, it would be expected that reasonably homogeneous silts would not pose difficulties.
Carbonate soils
Erbrich and Hefer (2002) present the case history of the installation of suction anchors at the
Laminaria site in the Timor Sea. Whilst the installation of the 9 anchors was successful, the suction
pressures measured were significantly lower than those predicted in the original design calculations.
Erbrich and Hefer (2002) report very low values for the adhesion factor (of the order of 0.1 0.2)
that arise from the back-analysis of the field data. It is clear that for extremely fine-grained
carbonate soils (as at the Laminaria case) the clay calculation is appropriate, while for the coarser
materials the sand calculation is appropriate. Because of the crushability of carbonate materials,
very low values of tan K would probably be appropriate in the friction calculation.
Rocks
It is unlikely that suction caissons could be installed into any but the very softest of rocks.
Special conditions
The influence of special conditions (e.g. shallow gas deposits within the depth of the caisson,
organic material etc.) is almost unknown and would have to be dealt with on an ad hoc basis.
9
Pumping requirements
The flow capacity of pumps for installation in clay needs only to be that necessary (with a suitable
margin) to remove the water from the caisson as penetration proceeds, that is v
D
q
i
4
2
= , where q
is the required flow rate and v is the vertical penetration velocity. In sands the capacity must also be
sufficient to cope with the seepage beneath the foundation. This can be assessed by conventional
seepage calculations, giving a total required flow rate of
w
i
skD
F v
D
q

+ =
4
2
, where F is a
dimensionless factor that depends on D h and k is the sand permeability (Houlsby and Byrne,
2004).
Conclusions
In this paper we present the calculation procedures that are required for suction caisson installation
in clay. Calculations include those for self-weight penetration, penetration under suction and the
limits to the suction assisted penetration. The calculation procedures are compared to case records,
showing good agreement with the measured responses. The paper concludes with discussion of
potential issues when installing suction caissons in a variety of other soils.
Acknowledgements
B.W.B. acknowledges generous support from Magdalen College, Oxford. The authors are grateful
to Dr Andrew House for provision of original data for use in examples 2 and 3.
Nomenclature
D caisson diameter
f load spread factor for vertical stress enhancement
h installed depth of caisson
h
w
height of water above mudline
K factor relating vertical stress to horizontal stress
L caisson skirt depth
l perimeter length of stiffeners within caisson
m multiple of the diameter that the vertical stress is enhanced (i.e.
o m
mD D = )
N
q
bearing capacity factor (overburden)
N
c
bearing capacity factor (cohesion)
p
a
atmospheric pressure
s suction within the caisson with respect to the ambient seabed water pressure
0 u
s shear strength at mudline
1 u
s average shear strength over depth of skirt
2 u
s shear strength at caisson skirt tip
t wall thickness
V, V vertical load, effective vertical load
z vertical coordinate below the mudline
adhesion factor
interface friction angle
, unit weight of soil, effective unit weight of soil

w
unit weight of water
rate of change of shear strength with depth
10

v
,
v
vertical stress, effective vertical stress

subscripts
i inside caisson
o outside caisson
References
Bye, A., Erbrich, C.T., Rognlien, B. and Tjelta, T.I. (1995) Geotechnical design of bucket
foundations Paper OTC 7793, Offshore Technology Conference, Houston, Texas
Byrne, B.W. (2000) Investigation of suction caissons in dense sand DPhil Thesis, Oxford
University
Byrne, B.W. and Houlsby, G.T. (2003) Foundations for offshore wind turbines Phil. Trans. Roy.
Soc. of London, Series A, 361, December, pp 2909-2930
Byrne, B.W., Houlsby, G.T., Martin, C.M. and Fish, P. (2002) Suction caisson foundations for
offshore wind turbines Wind Engineering 26, N
o
3
Colliat, J -L., Boisard, P., Gramet, J -C. and Sparrevik, P. (1996) Design and installation of suction
anchor piles at a soft clay site in the Gulf of Guinea Paper OTC 8150, Offshore Technology
Conference, Houston, Texas
Colliat, J -L., Boisard, P., Sparrevik, P. and Gramet, J -C. (1998) Design and installation of suction
anchor piles at a soft clay site Proc ASCE Jour. of Waterway, Port, Coastal and Ocean Eng.
124, No 4, pp 179188
Erbrich, C.T. and Hefer, P.A. (2002) Installation of the Laminaria suction piles a case history
Paper OTC 14240, Offshore Technology Conference, Houston, Texas
Houlsby, G.T. and Byrne, B.W. (2004) Design procedures for installation of suction caissons in
sand Submitted to Geotechnical Engineering
Houlsby, G.T and Martin, C.M. (2003) "Undrained Bearing Capacity Factors for Conical Footings
on Clay", Gotechnique, Vol. 53, No. 5, J une, pp 513-520
House, A.R., Randolph, M.F. and Borbas, M.E. (1999) Limiting aspect ratio for suction caisson
installation in clay Proc. 9
th
Int. Symp. on Offshore and Polar Eng., Brest, France
House, A.R. and Randolph, M.F. (2001) Installation and pull-out capacity of stiffened suction
caissons in cohesive sediments Proc. 11
th
Int. Symp. on Offshore and Polar Eng., Stavangar,
Norway
Hogervost, J .R. (1980) Field trials with large diameter suction piles Paper OTC 3817, Offshore
Technology Conference, Houston, Texas
Lacasse, S. (1999) Ninth OTRC Honors Lecture: Geotechnical contributions to offshore
development. Paper OTC 10822, Offshore Technology Conference, Houston, Texas
Senpere, D. and Auvergne, G.A. (1982) Suction anchor piles a proven alternative to driving or
drilling Paper OTC 4206, Offshore Technology Conference, Houston, Texas
Solhjell, E., Sparrevik, P., Haldorsen, K. and Karlsen, V. (1998) Comparison and back calculation
of penetration resistance from suction anchor installation in soft to stiff clay at the Njord and
Visund Fields in the North Sea Proc SUT Conf. on Offshore Site Investigation and Foundation
Behaviour, London, UK
Tjelta, T.I. and Hermstad, J . (1986) Large-scale penetration test at a deepwater site Paper OTC
5103, Offshore Technology Conference, Houston, Texas
Tjelta, T.I. (1994) Geotechnical aspects of bucket foundations replacing piles for the Europipe
16/11-E J acket Paper OTC 7379, Offshore Technology Conference, Houston, Texas
Tjelta, T.I. (1995) Geotechnical experience from the installation of the Europipe J acket with
bucket foundations Paper OTC 7795, Offshore Technology Conference, Houston, Texas.
Tjelta, T.I., Aas, P.M., Hermstad, J . and Andenaes, E. (1990) The skirt piled Gullfaks C Platform
installation Paper OTC 6473, Offshore Technology Conference, Houston, Texas
11
Tjelta, T.I. (2001) Suction piles: their position and application today Proc. 11
th
Int. Symp. on
Offshore and Polar Eng., Stavangar, Norway
Figures
0
5
10
15
20
25
30
35
0 5 10 15 20
Diameter (m)
D
e
p
t
h

(
m
)
Trials
Laminaria
Project - Clay
Project - Layered
Project - Sand
L/D =1
Snorre TLP
Sleipner T
Draupner E
YME J ack-up
Anchor
Foundations
Shallow Foundations
Gullfaks C
Figure 1: summary of uses of caisson foundations (from Byrne (2000) with further data from
Tjelta (2001))

z
t
D
o
D
i
h
V'
h
c
Mudline

Figure 2: outline of suction caisson

12
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 1000 2000 3000 4000 5000 6000 7000
Load without suction (kN)
D
e
p
t
h

z

(
m
)
Adhesion
End bearing
Total

Figure 3: calculated loads on caisson for Example 1 in the absence of suction



0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 5 10 15 20 25 30 35 40 45 50
Required suction s (kPa)
D
e
p
t
h

z

(
m
)

Figure 4: calculated suction for Example 1

13
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Load without suction (kN)
D
e
p
t
h

z

(
m
)
Adhesion
End bearing
Total

Figure 5: calculated loads for Example 2 in the absence of suction



0
0.02
0.04
0.06
0.08
0.1
0.12
0 20 40 60 80 100 120 140 160 180
Required suction s (kPa)
D
e
p
t
h

z

(
m
)
Calculated
Experiment

Figure 6: calculated suction for Example 2

14
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 20 40 60 80 100 120 140 160 180
Required suction s (kPa)
D
e
p
t
h

z

(
m
)
Calculated
Experimental

Figure 7: comparison of calculated and experimental suction pressures for House and
Randolph (2001) NC-IP2 experiment, Example 3



0
1
2
3
4
5
6
7
8
-2 -1 0 1 2 3 4 5 6 7
Excess Volume of Fluid Removed (ml)
N
o
r
m
a
l
i
s
e
d

p
e
n
e
t
r
a
t
i
o
n
:

h

/

D
Installation 2
Installation 4
Predicted level of plug failure
Figure 8: variation with depth of excess volume of water removed for Example 3

15
0
2
4
6
8
10
12
14
0 20 40 60 80 100 120 140 160
Required suction s (kPa)
D
e
p
t
h

z

(
m
)

Figure 9: Comparison between calculated and observed suction pressures at the Nkossa
installation, Example 4


1
Design procedures for installation of
suction caissons in sand
G.T. Houlsby
1
and B.W. Byrne
1


Keywords: sand, foundations, suction caissons, installation
Abstract
Suction installed caisson foundations are being used or considered for a wide variety of offshore
applications ranging from anchors for floating facilities to shallow foundations for offshore wind
turbines. In the design of the caissons the installation procedure must be considered as well as the
in-place performance. The scope of this paper is to consider the calculations appropriate for the
installation of caissons in sands. Calculation methods are presented for determining the resistance to
penetration of open-ended cylindrical caisson foundations both with and without the application of
suction inside the caisson. Comparisons are made with case records. A companion paper (Houlsby
and Byrne, 2004) addresses the calculation procedure for installation in clays as well as other soils.
Introduction
Suction caissons are large cylindrical structures, usually made of steel, open at the base and closed
at the top. After an initial penetration into the seabed caused by self weight, a suction (relative to
seabed water pressure) is applied within the caisson, which forces the remainder of the caisson to
embed itself, leaving the top flush with the seabed. The purpose of this paper is to present design
calculations for the installation of caissons in sand. When the suction is applied the pressure
differential on the lid of the caisson effectively increases the downward force on the foundation.
However, in sand the applied suction also generates flow within the soil. The pore pressure
gradients are beneficial to the installation process and must be accounted for in the design
calculation. Separate calculations are of course necessary to assess the capacity of the caisson once
installed whether used as a shallow foundation or anchor. Some of the issues that need to be
addressed for the in-service performance of suction caissons in sand are discussed by Byrne and
Houlsby (2002, 2003, 2004).

The first major structure installed in dense sand using suction caissons was Statoils Draupner E
riser platform (formerly Europipe 16/11 E) in the North Sea. This was installed successfully during
1994 in 70 m water depth. The caisson foundations were 12m in diameter and the skirts were 6m
long, and designed to be installed with suction. The design for the installation was based on a
combination of field testing, laboratory testing and finite element modelling (as described by Tjelta,
1994; Bye et al., 1995; Tjelta, 1995). Statoil installed a second caisson founded structure in the
North Sea in 1996 (Sleipner T). During the detailed design of this structure Erbrich and Tjelta
(1999) developed a design methodology using design charts based on finite element calculations.
The analyses presented in this paper differ from the finite element approach in that they are
classical in the sense that they employ simplifying assumptions, borrowing techniques from both
pile design and bearing capacity theory. More rigorous analyses, using for instance finite element
techniques (as for instance outlined by Erbrich and Tjelta (1999)), could be used for particular
installations. The analyses presented here should, however, provide a reasonable approximation for
design purposes.

1
Department of Engineering Science, Parks Road, Oxford OX1 3PJ
2
Analysis
We will consider a circular caisson of outside diameter
o
D and wall thickness t, so that the inside
diameter is t D D
o i
2 = . It is useful also to define the mean diameter ( ) 2
i o
D D D + = . For most
cases D t << , so that
i o
D D D . The water depth is
w
h , and the vertical coordinate, measured
as a depth below mudline, is z. The current embedment of the caisson is h and the height of the
caisson is
c
h , see Figure 1. The unit weight of water is
w
and of the soil is . The buoyant unit
weight of the soil is
w
= . Atmospheric pressure is
a
p .

It is assumed that the net downward vertical load on the caisson, when it is submerged in water, is
V . This is the weight of the caisson, less any buoyant effects and any part of the weight supported
by craneage, plus any applied downward loading e.g. from the weight of an attached structure. Note
that V will vary with the penetration of the caisson in some way that is unrelated to the following
calculations, as more of the attached structure becomes submerged, and as the load taken by a crane
is reduced.

The caisson is assumed to be a simple cylinder, although in practice a number of complicating
features are often employed such as:
(a) vertical stiffeners attached to the inside of the caisson,
(b) annular stiffeners attached to the inside of the caisson,
(c) more complex designs such as stepped caissons.
Installation Calculations for Sand
The installation process can be broken into two components; (a) self-weight penetration and (b)
suction installation. The self-weight penetration in the absence of suction is important, as a seal is
necessary at the edge of the foundation in order for the suction component to be performed
adequately. The designer will need to understand the interaction between the soil density (and
therefore peak friction angle), the skirt wall thickness and the effective vertical load (V) acting on
the foundation so that a sufficient penetration into the sand can be obtained. Once a seal can be
assured the suction phase can be completed. The designer will need to predict the required suction
as a function of depth of penetration. This information can be used to assess pump capacity, and the
rate at which suction needs to be applied. The installation contractor will need to appreciate the
implications of variations in applied suction, so that effective control of installation is achieved. If
the suction is applied too quickly, then localised piping may occur. This could prevent full
installation of the foundation. Finally the designer needs to be aware of any limitations to the
design, such as the maximum aspect ratio that can be installed with suction whilst avoiding the
possibility of liquefaction of the internal plug of soil. For the purposes calculation an idealised case
of a foundation on a homogeneous deposit of sand (assumed drained) will be considered in this
paper.
Self-weight penetration
The resistance on the caisson is calculated as the sum of friction on outside and inside, and the end
bearing on the annulus. The frictional terms are calculated in a similar way as in pile design, by
calculating the vertical effective stress adjacent to the caisson, then assuming that horizontal
effective stress is a factor K times the vertical effective stress. Assuming that the mobilised angle of
friction between the caisson wall and the soil is then we obtain the result that the shear stress
acting on the caisson is tan K
v
. Note that in the subsequent analysis the values of K and never
appear separately, but only in the combination tan K , so it is not possible to separate out the
effects of these two variables. Allowance is made, however, for the possibility of different values of
3
tan K acting on the outside and inside of the caisson. The difference in the following analysis
from conventional pile design is that the contribution of friction in enhancing the vertical stress
further down the caisson is taken into account. The end bearing is taken as the sum of
q
N and

N
terms in the conventional way, and it is assumed that solutions for a strip footing of width t are
appropriate for the caisson rim.

If (following conventional pile design practice) no account is taken of the enhancement of vertical
stress close to the pile due to the frictional forces further up the caisson, then the result for the
vertical load on the caisson for penetration to depth h, in the absence of suction, is given by:
( ) ( ) ( ) ( ) ( ) Dt N
t
hN D K
h
D K
h
V
q i i o o

+ +

=

2
tan
2
tan
2
2 2
(1)
friction on outside friction on inside end bearing on annulus

However, ignoring the enhancement of the stress in this case proves unconservative (i.e. it would
underestimate the force and suction required for full penetration), so we develop here a theory
which takes this effect into account.

Consider first the soil within the caisson. If we assume that the vertical effective stress is constant
across the section of the caisson, then the vertical equilibrium equation for a disc of soil within the
caisson (see Figure 2) leads to the equation:

( ) ( ) ( )
i
i v
i
i i v v
D
K
D
D K
dz
d
+ =

+ =
tan 4
4
tan
2
(2)
Writing ( ) ( )
i i i
Z K D = tan 4 this equation becomes =

i
v v
Z dz
d
, which has the solution
( ) ( ) 1 exp =
i i v
Z z Z for 0 =
v
at 0 = z . The total frictional terms in fact depend on the
integral of the vertical effective stress with depth, and we can also obtain
( ) ( ) ( )
i i i
h
v
Z h Z h Z dz =

1 exp
2
0
. For small
i
Z h the integral simplifies to 2
2
h .

A similar analysis follows for the stress on the outside of the caisson. If we make the assumption
that the enhanced stress is constant between diameters
o
D and
o m
mD D = , and further assume that
there is no shear stress on vertical planes at diameter
m
D , then we obtain the same results as for
inside the caisson, but with
i
Z replaced by ( ) ( ) ( )
o o o
K m D Z = tan 4 1
2
.

If the more realistic assumption z f D D
o o m
2 + = is made then
( ) ( ) ( ) ( ) ( )
o o o o o
K D z f D Z + = tan 4 1 2 1
2
, and an analytical solution to the equation
=

o
v v
Z dz
d
cannot be obtained (or at the very least is not straightforward). The equation can,
however, readily be integrated numerically to give the variation of vertical stress with depth. If this
approach is adopted then it would be consistent to assume that within the caisson at small D z the
stress is only enhanced in an annulus between
n
D and
i
D , where z f D D
i i n
2 = . This leads to
( ) ( ) ( ) ( ) ( )
i i i i i
K D z f D Z = tan 4 2 1 1
2
, and the differential equation for the vertical stress must
4
again be solved numerically. The resulting solution applies down to a value
i i
f D z 2 = , at which
0 =
n
D . Below this the original expression for
i
Z is appropriate.

In Equation (1) the end bearing term accounts for a triangular assumed stress distribution across the
tip of the caisson, see Figure 3(a). Now that the stress inside and outside the caisson may be
different, the assumed stress distribution across the tip of the caisson is as in Figure 3(b). The mean
stress on the tip is calculated as follows. First determine
vo
and
vi
. For all likely combinations
of parameters
vi vo
< ; because there is greater enhancement of the stress within the caisson
rather than outside. If
q
vo vi
N
tN

<
2
then the stress distribution is as shown in Figure 3(b) and

+ = N
t
x
t N
q vo end
2
2
where
( )

+ =
N
N
t
x
q vi vo
4 2
.

If
q
vo vi
N
tN

2
then all the flow occurs outwards, 0 = x and

+ = tN N
q vo end
.

Accounting for these effects of stress enhancement, Equation (1) becomes modified to:
( ) ( ) ( ) ( ) ( ) Dt D K dz D K dz V
end i i
h
vi o o
h
vo
+ + =

tan tan
0 0

(3)
In the special case where m is taken as a constant and uniform stress is assumed within the caisson
this can be expressed as:

( ) ( )
( ) ( ) ( ) Dt D K
Z
h
Z
h
Z
D K
Z
h
Z
h
Z V
end i i
i i
i
o o
o o
o
+

=
tan 1 exp
tan 1 exp
2
2
(4)
Suction-assisted penetration
If the pressure in the caisson is s with respect to the ambient seabed water pressure, i.e. the absolute
pressure in the caisson is s h p
w w a
+ , then it is assumed that the excess pore pressure at the tip
of the caisson is as , i.e. the absolute pressure is ( ) as h h p
w w a
+ + . There is therefore an
average downward hydraulic gradient of h as
w
on the outside of the caisson and upward
hydraulic gradient of ( ) h s a
w
1 on the inside. Because the flow is more restricted inside the
caisson than outside, a is expected to be a factor somewhat less than 0.5. Calculations for a are
presented later in this paper.

We assume that the distribution of pore pressure on the inside and outside of the caisson is linear
with depth. The solutions for the vertical stresses inside and outside the caisson are exactly as
before, except that is replaced by h as + outside the caisson and by ( ) h s a 1 inside the
caisson. We further assume that the internal vertical effective stress is reduced sufficiently so that
the failure mechanism involves movement of soil entirely inwards ( t x = in Figure 3(b)). The
capacity, accounting for the pressure differential across the top of the caisson, is again calculated as
the sum of the external and internal frictional terms, and the end bearing terms:
5

( ) ( ) ( )
( ) ( ) ( )( ) Dt tN N D K dz
D K dz D s V
q vi i i
h
vi
o o
h
vo i
+ + +
= +

tan
tan 4
0
0
2
(5)
Where in general the external and internal vertical stresses are determined by numerical integration
using the modified values of the effective unit weight on the outside and inside of the caisson.
In the special case where m is taken as a constant and uniform stress is assumed within the caisson
this can be expressed as:

( ) ( ) ( )
( )
( ) ( )
( )
( ) Dt tN N
Z
h
Z
h
s a
D K
Z
h
Z
h
Z
h
s a
D K
Z
h
Z
h
Z
h
as
D s V
q
i
i
i i
i i
i
o o
o o
o i

+ = +

1 exp
1
tan 1 exp
1
tan 1 exp 4
2
2 2
(6)
Equations (5) and (6) are each linear equation in s, and can be used to solve for the suction required
to achieve a penetration h.

Note that because of the assumption of pure inward failure, Equation (6) does not reduce to exactly
Equation (3) in the absence of suction. The difference, which is very small, can be resolved as
follows. Noting that

+ = 1 exp
o
o vo
Z
h
Z
h
as
and
( )

= 1 exp
1
i
i vi
Z
h
Z
h
s a
,
for
q
vi vo
N
tN

2
then t x = and Equation (6) applies. For
q
v vi
N
tN

2
0
then 0 = x and
pure outward flow occurs, and the final term in Equation (6) should be replaced by
( ) Dt tN N
Z
h
Z
h
as
q
o
o

+

1 exp . For intermediate cases the last term is replaced by
( ) Dt
end
, where

+ = N
t
x
t N
q vo end
2
2
and
( )

+ =
N
N
t
x
q vi vo
4 2
.

It can be verified that there are smooth transitions between each of these conditions. In each case
there is an equation that can be solved for s (for the intermediate case it is a quadratic, in the other
cases linear). It can, however, easily be verified that that for most cases the outward flow and
intermediate solutions only apply for a very small range of penetration at the beginning of the
suction process, during which there is a transition in the flow direction of the sand. For practical
purposes the transition phase can be ignored.

If the sand is not homogeneous, but consists of a number of layers with different design values of
, and tan K , then the above calculation can be adapted in a reasonably straightforward way,
although the integrals for the vertical stress solution become even more cumbersome. Of more
significance could be changes of permeability with depth, since this might affect the pore pressure
factor a. Caution should therefore be exercised in this case.
6
Limits to suction-assisted penetration
As the suction is increased, the upward hydraulic gradient on the inside of the caisson approaches
the value at which a piping failure might be induced. At this stage the vertical effective stress inside
the caisson at the caisson tip (and in fact throughout the depth of the caisson) falls to zero. It is
anticipated that if attempts were made to increase the suction further, local piping failures would be
induced, possibly with a major inflow of water into the caisson, but without significant further
penetration. This condition will occur when ( ) 0 1 = s a h , i.e. ( ) a h s = 1 . Substituting this
into the penetration equation (6) (for the simplified vertical stress distribution) and simplifying we
obtain:
( ) ( )
( ) ( ) ( )( ) Dt tN D K
Z
h
Z
h
Z
a
D
a
h
V
o o
o o
o
i
+

+

tan 1 exp
1 4 1
2
2
(7)
Solving the above equation for h leads to the maximum depth of penetration that can be achieved
using suction. Note because the equation is transcendental in h, and the factor a is a function of
D h , the equation needs to be solved iteratively.

We can observe, however, that the last term is typically small. For small applied vertical loads V ,
and taking
o i
D D D , the above leads to the simple solution
( )

=
o o
o o
Z
h
Z
h
D
Z K
h 1 exp
tan 4
2
. If we ignore the effect of stress enhancement, i.e. for
small
o
Z h , then
2
2
2
1 exp
o
o o Z
h
Z
h
Z
h

and we obtain
( )
o
K
D
h

tan 2
. This can be used
to provide an initial estimate of the maximum achievable penetration with suction. Since tan K is
often approximately 0.5, we conclude that in sand the limit on suction assisted penetration is likely
to be of similar magnitude to the diameter. Note that this limit is much smaller than for installation
in clays (Byrne and Houlsby, 2004).

In fact a slightly more stringent limit on suction-assisted penetration can be established on the same
basis as the reverse bearing capacity solution used for clays (Byrne and Houlsby, 2004). Again a
plastic failure could occur with flow of soil into the caisson and without further penetration. This
condition will occur when
vi q vo
N = . That is (for the simplified vertical stress distribution):

( )

+ 1 exp
1
1 exp
i
i q
i
i
Z
h
Z
h
s a
N
Z
h
Z
h
as
(8)
Although Equation (12) provides a more conservative estimate of the limit of suction-assisted
penetration, since
q
N is usually a large number (typically more than 30), in fact it differs very little
from the condition ( ) 0 1 = s a h discussed above.
The effect of internal stiffeners
The resistance of internal axial stiffening plates can be taken into account by adding resistance
terms accounting for the friction and the end bearing on each plate. For the simplified vertical stress
distribution the frictional terms will be of the form ( ) l K
Z
h
Z
h
Z
s
i i
i

tan 1 exp
2
, where l
is the perimeter length of the stiffeners, and the end bearing terms will be of the form
7
A N
t
N
Z
h
Z
s
q
i
i

2
1 exp , where
s
t is the stiffener thickness and A is the end area of
the stiffener. Once suction is applied the friction term is modified by replacing by
( )

h
s a 1
, and the end bearing expression becomes
( )
A N
t
N
Z
h
Z
h
s a
s
q
i
i

2
1 exp
1
. Note the factor 2
s
t in this expression rather
than t for the caisson wall: this is because the soil flows to either side of the stiffener.

Furthermore the influence of the stiffeners alters the expression for
i
Z to
( ) ( ) ( )
s i i
i
i
K l K D
D
Z
+

=
tan tan 4
2
.

If the stiffeners do not extend to the base of the caisson then again corrections need to be made to
the above expressions to account for this. It should be noted though that the calculations for the
vertical effective stress at any level in the caisson then may involve rather cumbersome integrations.

External stiffeners in sand would almost certainly cause problems during installation, as their tips
would not be in the region where the effective stress is reduced. As a result they would attract a
very large tip resistance. Annular stiffeners, which are often used for caissons in clay, would almost
certainly prevent installation in sand.
Pressure factor a and flow calculations
The factor a should be 0.5 for very shallow penetration in a soil of uniform permeability, and would
be a function of D h . As suction is applied there is the possibility that the sand within the caisson
becomes loosened and thus exhibits a higher permeability. For simplicity one can consider a
permeability
o
k for the soil outside the caisson and
o i
k k > inside the caisson. The ratio
o i f
k k k = will affect the value of a .

Finite element analyses have been used to calculate the value of a in a soil of uniform permeability
for a thin-walled caisson for values of D h up to 0.8 and for values of
f
k of 1, 2 and 5. The results
of two separate studies (using slightly different mesh details, Aldwinkle (1994), J unaideen (2004)))
are shown in Figure 4.

An approximate fit for 1 =
f
k given by:

= =
D c
h
c c a a
2
1 0 1
exp 1 (9)
with the values 45 . 0
0
= c , 36 . 0
1
= c , 48 . 0
2
= c . This equation captures the trend of the
calculations reasonably well, although for 0 = D h the value should theoretically be 0.5 and for
very large D h the factor would be expected to tend to zero.
The effects of different
f
k values can be accounted for by a simple calculation in which the head
loss within the caisson is reduced in inverse proportion to the permeability. This results in:
8

( )
f
f
ak a
k a
a
+
=
1
1
1
(10)
where
1
a is the value from Equation 9. Figure 4 shows a comparison between calculated factors
using Equations 9 and 10 and numerically calculated factors using finite element analysis.

The flow beneath the caisson due to the suction can be determined using Darcys law as:
F
Ds k
q
w
o

= (11)
where F is a dimensionless factor that depends on the ratios D h and
f
k . Calculated values of F
using finite element analysis (J unaideen, 2004) are presented in Figure 5. If it is assumed that the
excess pressure across the base of the caisson is uniform and of value ( ) a s 1 , then F can be
estimated from the expression ( ) ( ) D h k a F
f
4 1 = . The faint lines in Figure 5 show the
computed F values from this expression with a determined by equations 9 and 10. It can be seen
that at large D h / the results from the finite element analysis approach this value. This is because
the assumption of uniform pressure across the base of the caisson is reasonable in this case, whilst
at shallow depths there is a higher pressure towards the centre of the caisson, resulting in much
higher flows.

As an example of the flow calculation, for a caisson of diameter 6m penetrated 4m into a soil with a
uniform permeability of s m 10 2
4
, and with an applied suction of 58kPa the estimated flow
would be s m 006 . 0 85 . 0
10
58 6 10 2
3
4
=


. If the caisson was installed to this depth in a period
of 2 hours, then the pumping rate simply to remove the water from the caisson would be
s m 016 . 0
3600 2
4 3
3
2
=

. Such a calculation can be used to assess the relative contributions of the
two flow terms to the net required pumping capacity.

The calculations described in the preceding sections have been implemented in a spreadsheet-based
program SCIP (Suction Caisson Installation Prediction). This is used for the calculations in the
examples below. In all these examples the diameter over which the vertical stress is enhanced varies
linearly with depth (i.e. the loadspread factors are 1 = =
i o
f f ).
Example 1: trial installations at Tenby and Sandy Haven
Results are reported here of two trial installations of caissons made by Offshore Data Ltd at Tenby
and Sandy Haven, on the south coast of Wales. At Tenby (Example 1a) a 2m diameter, 2m high
caisson was installed in dense sand. Figure 6 shows a comparison between the calculated suction
and the measured data. The data used for the computations in this and subsequent examples are
given in Table 1. It can be seen that the observed suction against penetration response is fitted well,
and that the limit to suction-assisted penetration of 1.4m is also fitted: at this depth no further
penetration was observed in spite of an increase in suction.

The second case (Example 1b) is a 4m diameter caisson, 2.5m high and with a wall thickness of
20mm, which was installed at Sandy Haven. Figure 7 presents the output from the SCIP program
and compares it to the measured suction against penetration of the caisson. The suction increased
approximately linearly with depth up to the full penetration of 2.5m.

9
The results from the SCIP program have been obtained by choosing parameters to fit the data best.
The suction against depth curve (which is in fact almost a straight line) depends principally on the
values of tan K and
f
k , and very little on other quantities.

Clearly the above figures match very closely the observations at these sites, and all the figures in
Table 1 are entirely plausible. Note that for the purposes of the installation a conservative
calculation (somewhat unusually) will require higher estimates of the strength parameters than
might be used for a capacity calculation.
Example 2: Draupner E
As described in the introduction, this structure was installed by Statoil in 1994 and was the first
jacket structure to be installed using suction caissons for the foundation. The soil conditions consist
of a very dense sand, and a friction angle of 44 is estimated. As described by Tjelta (1994, 1995)
the foundations consist of 12m diameter caissons of skirt length 6m. The wall thickness is taken to
be 45mm. Tjelta (1995) describes the installation procedure and states that the self weight
penetration was achieved by flooding the jacket legs, thereby increasing the submerged weight of
the structure from 1350t to 2700t (i.e. 6622kN per foundation). Internal stiffeners have been
neglected as no information is available about the geometry.

Figure 8 shows the predictions as given by the spreadsheet program SCIP compared to the range of
field measurements as obtained on site for the four caissons and reported by Tjelta (1995). The
permeability factor was taken as 3.0 to provide a good correlation between the case record and the
calculated suction pressures.
Example 3: Sleipner T
The Sleipner T structure was the second jacket to be installed by Statoil in the North Sea. The
foundations are 15m in diameter and the skirt depth is 5m (Bye et al., 1995; Lacasse, 1999). The
wall thickness is taken to be 45mm and again the internal stiffeners are neglected as there is
insufficient information available on this detail. The soil conditions consist of a very dense sand,
and a friction angle of 45 is used in the calculations. Lacasse (1999) presents measured data from
the installation, which show that the self weight penetration is about 1.95m. It is possible to use this
information to back-calculate the effective vertical load applied to each foundation as this is not
given in the literature. An effective vertical load per foundation of 12MN gives the appropriate self-
weight penetration, and this corresponds to a weight of the jacket of about 48MN, which is
reasonable. Using the parameter values in Table 1 provides a good fit to the range of data on
observed suction as reported by Lacasse (1999), as shown in Figure 9.
Example 4: Laboratory Tests
This final example, shown in Figure 10, reports laboratory scale tests by Sanham (2003). The
caisson foundation is 150mm in diameter with a skirt length of 200mm (so that, unlike the previous
cases, the L/D ratio is greater than 1). The applied vertical load in three tests reported here was 45N,
85N and 165N giving a self-weight penetration of approximately 28mm, 49mm and 79mm. In this
particular case, due to the combination of the different variables, it is possible for the foundation to
be installed, even though the L/D ratio is greater than 1.0. Figure 10 shows the three suction against
penetration curves, compared to the theoretical calculations. The SCIP calculation captures the
trend of variation between the curves as a function of the applied load.

The examples show that, with plausible choice of input parameters, the presented method of
analysis fits case records of installation in sand well. It is appreciated that there are, of course a
number of parameters on which the calculation depends, so that to a large extent this match can be
10
achieved by careful choice of parameters. For use as a predictive tool, experience needs to be
gained on appropriate parameter selection.
Conclusions
In this paper we present a series of design calculations that can used to assess the installation of
skirted foundations installed into sand using suction. The calculations cover self-weight penetration,
suction assisted penetration and an assessment of the limitations to the suction installation process.
These calculations have been implemented in spreadsheet program to enable predictions of
installation to be made. In the final part of the paper we compare the predictions with four case
records, and a good agreement is found with the measured data.
Acknowledgements
B.W.B. acknowledges generous support from Magdalen College, Oxford. The authors are grateful
to Zeena J unaideen, who carried out the finite element calculations reported in Figures 4 and 5, and
to Rob Ellis of Offshore Data Ltd for supplying the data used in Examples 1a and 1b.
Nomenclature
a ratio of excess pore pressure at tip of caisson skirt to beneath the base
D caisson diameter
f rate of change of diameter that the vertical stress is enhanced
h installed depth of caisson
h
w
height of water above mudline
K factor relating vertical stress to horizontal stress
k
f
ratio of permeability within caisson to outside caisson (i.e. k
f
=k
i
/k
o
)
L caisson skirt depth
l perimeter length of stiffeners within caisson
m multiple of the diameter that the vertical stress is enhanced (i.e. D
m
=mD
o
)
N
q
bearing capacity factor (overburden)
N

bearing capacity factor (self-weight)

p
a
atmospheric pressure
s suction within the caisson with respect to the ambient seabed water pressure
t wall thickness
V, V vertical load, effective vertical load
z vertical coordinate below the mudline
interface friction angle
angle of friction of the soil
, unit weight of soil, effective unit weight of soil

w
unit weight of water

v
,
v
vertical stress, effective vertical stress

subscripts
i inside caisson
o outside caisson
References
Aldwinkle, C.G. (1994) The installation of offshore plated foundations for oil rig. Final Year
Project Report, Department of Engineering Science, Oxford University
Bye, A., Erbrich, C.T., Rognlien, B. and Tjelta, T.I. (1995) Geotechnical design of bucket
foundations Paper OTC 7793, Offshore Technology Conference, Houston, Texas
11
Byrne, B.W. and Houlsby, G.T. (2002) Experimental investigations of the response of suction
caissons to transient vertical loading Proc. ASCE, Journal of Geotechnical Engineering
128, N
o
11, November, pp 926-939
Byrne, B.W. and Houlsby, G.T. (2003) Foundations for offshore wind turbines Philosophical
Transactions of the Royal Society of London, Series A 361, December, pp 2909-2930
Byrne, B.W. and Houlsby, G.T. (2004) Experimental investigations of the response of suction
caissons to transient combined loading Proc. ASCE, Journal of Geotechnical and
Geoenvironmental Engineering 130, N
o
3, March, pp 240-253
Erbrich, C.T. and Tjelta, T.I. (1999) Installation of bucket foundations and suction caissons in sand
Geotechnical performance Paper OTC 10990, Offshore Technology Conference,
Houston, Texas
Houlsby, G.T. and Byrne, B.W. (2004) Design procedures for installation of suction caissons in
clay and other materials Submitted to Geotechnical Engineering
J unaideen, Z. (2004) Private communication
Lacasse, S. (1999) Ninth OTRC Honors Lecture: Geotechnical contributions to offshore
development Paper OTC 10822, Offshore Technology Conference, Houston, Texas
Sanham, S.C. (2003) Investigations into the installation of suction assisted caisson foundation.
Final Year Project Report, Department of Engineering Science, Oxford University
Tjelta, T.I. (1994) Geotechnical aspects of bucket foundations replacing piles for the Europipe
16/11-E J acket Paper OTC 7379, Offshore Technology Conference, Houston, Texas
Tjelta, T.I. (1995) Geotechnical experience from the installation of the Europipe J acket with
bucket foundations OTC 7795, Offshore Technology Conference, Houston, Texas
Table
Exampl
e
Location D
(m)
L
(m)
t
(mm)
' V
(kN)

(kN/m
3
)
tan K

o i
k k
1a Tenby 2.0 2.0 8 10 40
o
8.5 0.48 3.0
1b Sandy Haven 4.0 2.5 20 100 40
o
8.5 0.48 2.0
2 Draupner E 12.0 6.0 45 6622 44
o
8.5 0.63 3.0
3 Sleipner T 15.0 5.0 45 12000 45
o
8.5 0.8 3.0
4 Laboratory tests 0.15 0.2 1.65 variable 45
o
8.5 0.45 2.5
Table 1: Data used for SCIP calculations
12
Figures
z
t
D
o
D
i
h
V'
h
c
Mudline

Figure 1: outline of suction caisson



'
v
(Ktan)
i
D
i
dz
'
v
D
i
2
/4
('
v
+d'
v
)D
i
2
/4
'(D
i
2
/4)dz

Figure 2: equilibrium of slice of soil within caisson



'
v
'
v
'
v
N
q
2'N

1
t
(a)
'
vo
'
vi
'
vo
N
q
2'N

1
t
(b)
'
vi
N
q
x

Figure 3: stress distribution across tip of caisson

13
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 0.2 0.4 0.6 0.8 1
Pentration h/D
P
o
r
e

p
r
e
s
s
u
r
e

f
a
c
t
o
r

a
J unaideen kf =1
J unaideen kf =2
J unaideen kf =5
Aldwinkle kf =1
Aldwinkle kf =2
Aldwinkle kf =5
Equation 9
Equations 9, 10 (kf =2)
Equations 9, 10 (kf =5)

Figure 4: variation of pore pressure parameter a with D h and
f
k



0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 0.2 0.4 0.6 0.8 1
Penetration h/D
D
i
m
e
n
s
i
o
n
l
e
s
s

f
l
o
w

f
a
c
t
o
r

F
kf =1
kf =2
kf =5

Figure 5: variation of dimensionless flow parameter F

14
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 5 10 15 20 25 30
Required suction s (kPa)
D
e
p
t
h

z

(
m
)
Suction
Maximum penetration
Tenby installation

Figure 6: Calculated suction and case record for Tenby installation, Example 1a



0
0.5
1
1.5
2
2.5
3
0 10 20 30 40 50
Required suction s (kPa)
D
e
p
t
h

z

(
m
)
Predicted suction
Sandy Haven installation

Figure 7: Calculated suction and case record for Sandy Haven installation, Example 1b

15
0
1
2
3
4
5
6
0 20 40 60 80 100
Required suction s (kPa)
D
e
p
t
h

h

(
m
)
Predicted suction
Range of measured suction

Figure 8: Calculated and measured suction at Draupner E, Example 2



0
1
2
3
4
5
6
0 20 40 60 80 100
Required suction s (kPa)
D
e
p
t
h

h

(
m
)
Predicted suction
Range of measured suction

Figure 9: Calculated and measured suction at Sleipner T, Example 3

16
0
20
40
60
80
100
120
140
160
180
200
0 500 1000 1500 2000 2500 3000
Suction s (Pa)
P
e
n
t
r
a
t
i
o
n
h

(
m
m
)
Experiments
Theory

Figure10: Calculated and measured suction in laboratory, Example 4