Applied Ocean Research 32 (2010) 414424

Contents lists available at ScienceDirect

Applied Ocean Research

journal homepage: www.elsevier.com/locate/apor

Model testing of suction caissons in clay subjected to vertical loading

Felipe A. Villalobos a,b , Byron W. Byrne a, , Guy T. Houlsby a
a

Department of Engineering Science, Oxford University, UK

Faculty of Engineering, Universidad Catlica de la Santsima Concepcin, Chile

article

info

Article history:
Received 7 May 2010
Received in revised form
25 August 2010
Accepted 7 September 2010
Available online 2 October 2010
Keywords:
Foundations
Vertical loading
Clay
Cyclic loading
Offshore engineering
Suction caissons
Model testing
Geotechnical engineering

abstract
A wide range of new offshore applications are emerging in the energy sector. The oil and gas industry
is targeting minimum facility applications, whilst the renewable energy sector is developing offshore
wind turbines, as well as a number of wave and tidal energy devices. The design and installation of
the foundations are key considerations in the financial viability of such offshore engineering projects.
Suction caisson foundations are a potential solution for these new developments, but design guidance
is relatively sparse. This paper considers the vertical loading response of a caisson foundation in clay,
during installation and under both monotonic and cyclic vertical loading. The main contribution is the
presentation and interpretation of high quality experimental data. Vertical loading is critical for the design
of a multi-footing structure of the type that might be used for large offshore wind turbines. We first
consider the installation behaviour and compare data from pushed installations and a suction installation
with results from a theoretical calculation. We then consider cyclic vertical loading tests, focussing on
cyclic amplitudes that take the foundation into tension. Detailed displacement data and pore pressure
data are presented.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
With the diminishing supply of fossil fuels, it is likely that
energy companies will seek to exploit marginal oil reserves,
whilst governments will encourage further deployment of offshore
renewable energy devices. There is some overlap in the design
of these structures. It is likely that multi-footing structures will
be required, in which case the weight is low but the horizontal
loads applied, by comparison, are high. This leads to relatively
large overturning moments which must be countered by vertical
reactions at the foundations. Typical (very approximate) loads for
a 3.5 MW offshore wind turbine would be a weight of 6 MN and
horizontal load of 4 MN applied at 30 m from the seabed. By
spacing the foundations apart, the 120 MNm base moment can
be resisted by compression and tension in the foundations. The
key design issue is therefore how much tension is permissible for
a single loading event, and how far apart the foundations must
therefore be spaced. Of course the loading on the structure from
the waves and wind is cyclic in nature, and so the load on the
foundation will be cyclic. The issue of how much tension can be
applied under serviceability conditions is therefore also critical to
the spacing of the foundations. It is important to understand the

Corresponding address: Department of Engineering Science, Oxford University,

OX1 3PJ, UK. Tel.: +44 1865 273028; fax: +44 1865 283301.
E-mail address: Byron.Byrne@eng.ox.ac.uk (B.W. Byrne).
0141-1187/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apor.2010.09.002

effects of low levels of cyclic tension applied to the foundation,

and whether this leads to degradation of the foundation response.
It is thought that for foundations on clay the extreme tension
can be safely resisted, provided that it does not exceed the
undrained capacity of the foundation. However, long term cyclic
loading into tension, exceeding the drained friction, may well
cause unacceptable displacements. For skirted foundations on sand
Byrne and Houlsby [1] and Kelly et al. [2] recommend that tension
should be limited to the drained skirt friction, and preferably
should be avoided altogether. For multi-footing structures there is
also an issue of interactions between the foundations; however, for
the purposes of this study we assume that these can be neglected.
The response of suction caissons in clay has been studied more
than caisson response in sands. Most research has concentrated on
the anchor application, where long but thin suction installed piles
are used as anchors for floating facilities. The dominant loading is
horizontal, with some tension, although this depends on the nature
of the mooring. Some experimental work carried out has explored
the effect of cycling about high negative mean vertical loads (cyclic
pullout). Relevant information can be found in [39]. There has
been relatively little research on the use of suction caissons as
foundations rather than as anchors. The principal difference is the
skirt length-to-diameter ratio as well as the nature of the applied
loading. For foundations the skirt length-to-diameter ratio is about
one, and the loading consists of vertical and horizontal loads as well
as moments.

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

415

Fig. 1. (a) Three-degrees-of-freedom loading rig at Oxford University and (b) the sign convention for loads from [17] showing positive loads and displacements.
Table 1
Representative Speswhite kaolin clay properties (after de Santa Maria [14] and
Martin [15]).
Property

Value

Specific gravity, Gs
Average effective unit weight,
Average moisture, w
Liquid limit, LL
Plastic limit, PL
Coefficient of permeability, k (p = 200 kPa)
Coefficient of consolidation, cv (p = 200 kPa)

2.61
6.85 kN/m3
50%
65%
34%
3109 m/s
0.3 mm2 /s

There has been recent motivation for exploring caisson

foundation response in clay, driven by potential demand as
foundations for offshore wind turbines as well as other offshore
applications, such as foundations for small sub-sea developments.
Previous work on this problem has been carried out by Byrne
and Cassidy [10] and Cassidy et al. [11] who explored caisson
response in normally consolidated clay on the drum centrifuge at
the University of Western Australia. Further work has been carried
out by the University of Oxford at field scale, reported by Houlsby
et al. [12], and at laboratory scale, reported by Kelly et al. [13].
These latter two papers are important because they demonstrate
the scalability of results for tests on clay from the laboratory scale
to the field scale and are very relevant for the interpretation of the
data presented below. This paper extends the experimental data
base by presenting caisson foundation tests in overconsolidated
kaolin clay, which is relevant to sites around the UK coastline
that consist of hard or stiff clay. The paper reports results from
installation tests as well as the displacement behaviour under a
wide range of loading conditions including cyclic tensile loading,
all of which are relevant to the offshore wind application.
2. Experimental techniques
The testing was carried out on overconsolidated Speswhite
kaolin clay, and properties of the clay are given in Table 1. This
material was used because its high permeability allows rapid
consolidation of large specimens from reconstituted slurry. It has
been used in numerous previous studies at Oxford University, and
at other institutions, and so is very well characterised [14,15]. A

homogeneous kaolin slurry was obtained by mixing water with

kaolin at a moisture content of 120% in a ribbon blade mixer.
A vacuum pump attached to the mixer was used to de-air the
slurry during mixing. The slurry was pumped into cylindrical tanks
of 450 mm diameter and 900 mm height. Filter material (vyon)
was placed at the top and the bottom of the specimens, allowing
drainage to atmospheric pressure during consolidation. The slurry
was consolidated to a maximum pressure of 200 kPa. The tests
were carried out in the days following the complete unloading
of the specimen. To provide a second testing site the specimen
was inverted. The strength of the clay was estimated by using
a shear vane at depths of 25 and 125 mm. For one of the test
samples (FV7_1S) the strength estimates were updated using a
fitting of strength profiles from the other tests, combined with
a back analysis of the installation and loading data (see [16] for
further details). The sample properties are presented in Table 2. To
allow further interpretation in this paper we provide an estimate
of the strength at the surface of the specimen (su0 ) and the rate
of strength increase with depth ( ), assuming that the strength
can be fitted simply by a linear profile. This is in contrast to
the more complex fitting provided by Villalobos [16]. Despite
the consistent preparation procedure, the strength of the clays
varied significantly. All normalisations presented in this paper use
the undrained strength value measured at the reference depth of
125 mm.
The tests were carried out by using a complex computercontrolled loading rig, shown in Fig. 1(a), which was designed by
Martin [15] and modified by Byrne [18]. The rig is capable of applying independently controlled displacement or load paths for
each of the three degrees of freedom (vertical, horizontal and rotation) to the foundation. The software allows the application of
both monotonic and cyclic loading. The caissons were first loaded
vertically, and feedback control was used to keep the horizontal
load and moments on the foundation to negligible values. Installation by suction assistance was performed with a vacuum pump
connected to a regulating system, so that the suction could be controlled. For the cyclic loading tests it was possible to follow a detailed time history of loading. The loads and displacements are
measured at the foundation level and are specified according to
the sign convention set out by Butterfield et al. [17] and shown in
Fig. 1(b).

416

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

Table 2
Results from site investigation and caisson installation tests.
Test

su @ 25 mm (kPa)

su0 (estimate) (kPa)

(estimate) (kPa/m)

Vc (N)

hc (mm)

Vc (estimate) (N)

FV1_1a

4.6

7.3

3.9

27

248

148

282

500

FV2_1
FV2_12
FV5_1a

5.1
4.6
5.7

8.1
7.3
6.1

4.4
3.9
5.6

30
27
4

316
260
271

146
147
145

307
280
268

998
1089
301

FV3_1
FV6_1a

6.7
4.5

10.2
8.4

5.8
3.5

35
39

417
321

146
145

393
300

478
417

FV4_1
FV7_1c, a

8.8
6.3b

13.2
9.9b

7.7
5.4

44
36

682
353

147
145

515
374

1082
353

a
b
c

su @ 125 mm (kPa)

Vmax (N)

Specimen inverted.
Updated strength estimates.
Installation by suction.

Fig. 2. Data from the installation of caissons showing (a) load-penetration curves and (b) normalised load-penetration curves.

The caisson used was of diameter (2R) 150 mm, skirt length (L)
150 mm and wall thickness (t) 1 mm. A water pressure sensor was
located in the base plate of the caisson, so that the water pressure
beneath the base plate could be measured. An assumption about
the pressure distribution beneath the base plate is necessary for
interpreting the information from this sensor and as the loading
is vertical the assumption of uniform water pressure beneath the
base plate is made.

Houlsby and Byrne [19] show that a simple expression for the
suction s required to install a caisson to the depth h is given by:

3. Installation and monotonic loading

3.2. Pushing and suction installation

3.1. Penetration resistance

Fig. 2 shows all the load-penetration curves from the pushing

installation tests, as well as the total load results (V + |S |) from
test FV7_1S, which was installed with suction assistance (after
30 mm of pushing penetration). The data presented in Fig. 2(a)
are the actual load values as well as those normalised by the
base area of the caisson. An alternative interpretation is shown
in Fig. 2(b) where the data are presented in dimensionless form,
normalised by the undrained strength of the sample at a reference
depth of 125 mm. The total load from the suction test follows
very broadly the trends from the other tests, the main difference
being due to the slightly different strength profiles. This indicates
that caisson penetration resistance is likely to be independent of
the installation method in heavily overconsolidated kaolin clay, as
previously found in normally consolidated kaolin clay and in high
aspect ratio caissons by House [7], Rauch et al. [21] and Chen and
Randolph [20].
From the loaddisplacement curves shown in Fig. 2(a) the
values of contact vertical load Vc and contact penetration hc are
obtained when the base plate makes contact with the clay. These

The vertical load required (without suction) to penetrate a

caisson into a purely cohesive soil can be obtained from the
equilibrium of the forces acting on the caisson. The frictional
resistance inside and outside the caisson are calculated using
adhesion factors i and o applied to the undrained shear strength.
The end bearing is calculated using the bearing capacity coefficient
Nc for a deep strip footing. Houlsby and Byrne [19] express the
submerged vertical load V needed to penetrate a caisson to the
depth h as:
V = o su 2 Ro h + i su 2 Ri h + 2 Rt h + su Nc ,

(1)

where su is the average undrained shear strength over the depth of

the skirt, su is the undrained shear strength at the caisson tip, Ro , R
and Ri are the outside, mean and inside caisson radii, and t is the
skirt wall thickness. Whilst House [7] and Chen and Randolph [20]
have adopted a value of Nc of 7.5 the more usual value of 9 is used
in this paper.

s =

R2o

[o su 2 Ro h + i su 2 Ri h + 2 Rt ( h + su Nc ) V ], (2)

where V in this equation refers to the net force applied to the

caisson (i.e. after accounting for any buoyancy forces). For the
remainder of the paper the load V will refer to this submerged load.

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

417

Fig. 3. Measured and calculated (a) load-penetration curve, and (b) suction-penetration curve (where = dsu /dz).

are summarised in Table 2. The difference between the values of hc

and the skirt length (150 mm) of about 4 mm is due to internal
soil plug upheaval, which appears to be caused by the inward
displacement of soil by the skirt penetration. The upheaval in the
suction test was similar to that in the pushing tests. Assuming only
inward movement of the soil at the skirt tip would lead to a heave
of 4 mm, suggesting that the volume occupied by the penetrating
skirts is fully displaced inwards, and therefore indicating the
existence of a non-symmetrical shear failure mechanism at the
skirt tip. The plug heave in this instance is much less than that
for other reported tests, such as those by House [7], who reported
heave as large as 30% of the skirt length. However House [7] was
using caissons with a skirt depth-to-diameter ratio of greater than
8, in contrast to 1 in this study, and the study was carried out on
normally consolidated kaolin.
Eq. (1) was used to fit the load-penetration curves, assuming
that was constant for all tests and equal inside and outside of
the caisson skirt. A bearing capacity factor (Nc ) of 9 was used. The
results of this calculation at the depth of penetration of hc are
shown in Table 2 for comparison with the recorded values of Vc .
The value of was varied until the best fit was obtained with
the data using a least squares method. The value of adopted
for the calculations was 0.3. To assess the closeness of the fit for
the pushed installation, Fig. 3(a) shows a comparison between the
FV2 installation and the results of Eq. (1). Most of the resistance
is from the adhesion of the clay to the skirt walls, and a minor
contribution is made by the bearing resistance at the skirt tip, and
so the calculation is relatively insensitive to the chosen value for
Nc . The match of the calculated total load with the experimental
data is very close for this experiment for most of the penetration.
The calculation does not provide a good estimate for test FV4_1
where it appears that either a much higher value for is needed or
the undrained shear strength of the soil has been underestimated
by the shear vane tests. If is allowed to vary for each test then the
range is 0.260.32 except for FV4_1 which requires an value of
0.41 for a match to be obtained.
Eq. (2) is used to predict the suction required for the suction installation phase of test FV7, shown in Fig. 3(b). The calculation includes the value of vertical load applied to the caisson and recorded
during the test. To account for the variation of strength with depth
in the same plot the suction appears normalised as s/2R , where
the shear strength gradient = dsu /dz was assumed constant and

equal to 36 kPa/m. This is a very high value compared with values

around 1 kPa/m reported for NC clay by Chen and Randolph [20].
The critical component of Eqs. (1) and (2) is the assessment of
an appropriate value for . Andersen and Jostad [6,22] suggest that
the reduction of su acting on the skirt, due to penetration, is caused
by the remoulding of the clay and this might be estimated from
the inverse of the sensitivity of the clay. This is expressed as =
Ct /St (=Ct su,remoulded /su,peak ), where Ct is the thixotropic strength
ratio. For kaolin clay, as used in this study, Ct is close to 1, but for
other clays Skempton and Northey [23] or Andersen and Jostad [22]
provide advice. Fig. 4 shows curves obtained from shear
vane tests on samples of clay used in the experiments. The majority
of these tests were stopped at 60 of rotation after peak torque
was reached. However, three vane tests, carried out in sample 6,
were continued for more than 360 of rotation until a remoulded
condition (assumed as a residual strength) was reached. These
results indicate that a value of of about 0.4 is obtained, this being
slightly higher than the value of 0.3 used in the above calculations.
Of course the estimate of as 1/St provides neither an upper
bound nor a lower bound on . It represents simply the concept
that the strains adjacent to the caisson might be sufficient to
reduce the strength locally to the residual value. However it is
also possible that (i) a more diffuse shear mechanism occurs and
the strength does not reduce fully to residual (so 1/St would be
an underestimate of ), or more likely (ii) there are more critical
metal/soil slip mechanisms which mean that the full residual
strength is not mobilised on the caisson surface (so 1/St would
be an overestimate of ). The results obtained here lend some
credibility to the second possibility.
3.3. Compressive and tensile capacity
After installation, the vertical loading capacity of the caisson is
clearly of importance. This paper concentrates on the cyclic loading
response of the caisson; however, Fig. 5 shows the post-installation
loading response from two tests. In test FV1 the caisson was
installed and loaded until a vertical bearing capacity occurred. The
maximum load is 1545 N, and this can be compared to the results
of Martin [24] by calculating a bearing capacity factor according to:
Nc =

V
Asu

(3)

418

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

extraction the pressure measured beneath the lid was 80 kPa

and this compared to the pressure measured at the skirt tip of
100 kPa, indicating that cavitation had occurred in the region
beneath the skirt tip. In Fig. 6(a) the difference between the total
load and the water pressure load is the load arising from friction
on the skirt.
The data from the pullout test has been interpreted using
Eq. (3) to give Fig. 6(b) and a maximum Nc factor for pullout of
10.1 is very close to that obtained from the compression test FV1.
The responses shown here agree with those described by Byrne and
Cassidy [10] for a caisson in normally consolidated clay.
4. Cyclic vertical loading

Fig. 4. Adhesion factor ( ) as the inverse of the sensitivity = su,remoulded /

su,peak = 1/St versus vane rotation .

Fig. 5. Installation, compression and pullout responses for two different tests.

where A is the plan area of the foundation and su is the undrained

strength at the skirt tip. This calculation gives a value of Nc
of 10.15 for test FV1. This compares with Martins solutions of
9.26 (lower bound) to 10.82 (upper bound) for a smooth skirted
caisson. The caisson is then pulled out of the clay and the
maximum tensile force is 1015 N which is substantially lower
(66%) than the vertical compression capacity. The stiffness of
the response reduces gradually as the loads on the caisson cross
from compressive to tensile, and larger overall movements are
required to mobilise the maximum tensile force as compared to
the maximum compressive.
In test FV3 the caisson is loaded in tension very shortly
after installation, and without any significant vertical compressive
preloading. The response is very stiff, and a maximum tension of
1791 N is recorded after a relative displacement comparable to
that required to mobilise the compressive capacity. Indeed the
response is very stiff even under substantial tensions. The water
pressure under the lid was measured, and is shown in Fig. 6
compared with the vertical load converted to a stress by dividing
by the plan area of the foundation. The maximum water pressure
under the lid is close to 80 kPa, indicating that cavitation may
have occurred. This result is very similar to the work described by
Houlsby et al. [12] of a field test of a caisson in clay. Under rapid

In this study, three sets of tests were carried out, each with
sequences of increasing load intensity of 10 cycles per load packet.
The first series involved applying the cyclic load around V =
Vc , which is the load reached by the caisson after the pushing
installation. In the second set, the cyclic loading was applied about
V = 0 N, after unloading from the load reached during installation.
The third set of experiments involved the same load sequence as
the second, but with the difference being the method of installation
(suction instead of pushing installation). An assessment of the
effect of installation method on the caisson cyclic response was
made by evaluating stiffness degradation, displacement and pore
water pressure behaviour.
Load-controlled tests were conducted by means of feedback
control using an input cyclic loading history on the vertical
load. The other load components (moment and horizontal load)
were kept to negligible values by feedback control. Examples of
sinusoidal loading history inputs with a period of 12 s (0.08 Hz)
are shown in Fig. 7 with V /A plotted against time for two separate
test sequences. The first sequence (FV5_4) is a cyclic load about
a positive mean load, such that the load never goes into tension.
The second sequence (FV6_6) is a cyclic load about a mean load
of zero, so that half of the loading on the caisson is tensile. Note
that in this case the maximum tension applied is much lower
than the static pullout capacities indicated on Fig. 5. The period of
loading matches that found offshore, although Byrne [18] found
that, on dense oil-saturated sand, there was little influence of
loading period on caisson response. In contrast, El-Gharbawy [5]
found that on normally consolidated clay there was some effect of
cyclic frequency on response, although his tests were at loads close
to the static pullout capacity. During the initial cycles shown on
Fig. 7 the load applied failed to reach the target, so an adjustment
was made to the control algorithm (by adjusting the gain) so that
the target loads were more closely followed. Minor adjustments to
the gain were made throughout the testing sequence.
Included on Fig. 7 are some definitions of parameters used
later in the paper. For example Fig. 7(a) shows the displacement
variation 1w in each cycle for test FV5_4 and the total or net
vertical movement wt of the caisson at the end of each cyclic
event. As this is a test where only a compressive load is applied,
it is expected that the caisson would penetrate into the clay.
Fig. 7(b) shows that under the application of a tension the amount
of uplift during each cycle increases, indicating a softening of the
loading response. Maximum and minimum excess pore pressure
values (umax , umin ) during a cycling event were used as measures
of the range of variation of u. The variation of u in test FV5_4
occurs mainly above the initial value ui , whereas in test FV6_6 u
varies around ui . Values of umax and umin are able to capture these
variations.
4.1. Results of cyclic loading around Vm = 250 N
In the first series the caisson was loaded, immediately after a
pushed installation, by 8 cyclic loading packets of 10 cycles each.

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

419

Fig. 6. Loaddisplacement pullout behaviour for FV3 and that interpreted for bearing capacity factor.

are normalised by the diameter of the footing; the normalisations

follow the procedures set out by Kelly et al. [13].
Fig. 8 also shows the variation, with cyclic loading, of the excess
pore pressure and the normalised excess pore load (i.e. U = uA ,
where A is the plan area of the caisson). These figures allow only
the inspection of the last three events, where extremely large
displacements occurred such that the caisson had settled by half of
the skirt length. In this particular case there is no sign of settlement
reduction and the pore water pressure accounts for almost half of
the applied stress.
Finer detail of the response for the smaller amplitude cycles
is shown in Fig. 9, which shows the behaviour for the first two
loading packets. Each cyclic loading packet induced irrecoverable
settlements, but little build-up of pore water pressure. The rate
of settlement reduces after each cycle indicating that continued
cycling at this level might lead to a cyclic shakedown. This type
of loading corresponds to the serviceability condition, and the
stiffness of the response will be important in determining how
the remainder of the structure responds (particularly important for
fatigue type calculations).
More detail is shown in Fig. 10 which compares tests FV5_4
and FV5_5, where a substantial increase in settlement is found. The
excess pore water pressure variations become significant, reaching
values that account for approximately half of the load.
4.2. Results of cyclic loading around Vm = 0 N

Fig. 7. Loading history applied (as average pressure over the lid area V /A),
displacement w , and excess pore pressure u response, showing characteristic
parameters used in the analysis.

The cyclic loading amplitudes increased from 47 to 560 N,

and were applied about a mean vertical load Vm = 250 N. The
mean load was the maximum load applied in order to install the
caisson. This is denoted as the maximum preload experienced
by the caisson prior to the cycling (Vo ). The loaddisplacement
curves of the whole sequence of cyclic events are presented in
Fig. 8, normalised by strength (su ) at the reference point (depth of
125 mm) or by the preload Vo (lower scale). The displacements

The second and third series of cyclic tests were carried out
to investigate the response of cyclic loading around a zero mean
load. The difference between the two tests was that FV6 was
installed by pushing and FV7 was installed by suction. Fig. 11
shows the loaddisplacement curves for the whole sequence of
cyclic tests, where the large displacements that occurred during
the final loading sequence dominate the figures. A different
loaddisplacement response was obtained, compared with the
response in Figs. 810 (cycling around Vm
= Vc ). There are also
differences between the response shown for FV6 and FV7,
indicating that the installation method has an effect on the
short term cyclic response. On the one hand, a gradual increase
of displacement with cyclic loading occurred in the pushing
installation case (increasing degradation due to softer response),
leading to failure at 1V = 505 N. On the other hand, the suction
installation induced a much stiffer response for values of 1V
beyond 505 N until sudden large displacements occurred for
1V = 744 N.
Fig. 11 also shows different excess pore water pressure
variations for these tests compared with the cycling about a
positive mean load. When cycling occurs about 0 N, the excess pore

420

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

Fig. 8. Series of cyclic vertical loading events FV5 under Vm = 250 N.

Fig. 9. Tests FV5_2 and FV5_3 under 1V = 50 N and 100 N, respectively.

Fig. 10. Normalised curves for a cyclic test about a mean load of 250 N.

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

421

Fig. 11. Sequence of vertical loading events FV6 (a, b) and FV7 (c, d) under Vm = 0 N showing: (a), (c) loaddisplacement response and (b), (d) pore water pressure
displacement response.

pressures are predominantly negative whilst the cycling under a

net compressive force leads mostly to positive pore pressures. This
suggests that the development of excess pore water pressure is
directly related to the caisson vertical movement. For example, for
cycling around 250 N permanent settlement generated positive
excess pore pressure, and during cycling around 0 N the mainly
upward movement generated exclusively suction. The uplift was,
however, not permanent, and the caisson moves up and down,
passing through the initial reference position in each cycle.
The first two series of loading tests exhibited very small
displacements under the application of the nominal values of
1V = 50 and 100 N and these are not visible on Fig. 11. As
mentioned before, the range of small displacements is particularly
important since they represent the expected foundation serviceability condition, with large displacements encountered only in extreme loading events. Fig. 9 shows that for the small nominal load
amplitudes (1V = 50 and 100 N) irrecoverable settlements
occurred when cycling about Vm = 250 N. Fig. 12 shows that a permanent uplift occurs when these cyclic load amplitudes are applied
around Vm = 0 N. The figures highlight that the displacements of

the caisson installed by pushing were higher than displacements of

the suction installed caisson. One explanation is that there is some
residual suction from the installation, such that the suctions measured during the loading phase are higher than for the pushed installed case. For example Fig. 12 shows that the initial suction in
FV7 is 11 kPa; a value that represents a large percentage of the
maximum suction applied during the installation (s 16 kPa). The
initial suction shown in Fig. 10 for FV6 is 3 kPa.
4.3. Vertical displacements and excess pore water pressure
The data presented here have been summarised by determining
values for 1w, 1u and stiffness during each of the cyclic packets.
Table 3 provides a summary of this information.
Fig. 13 shows clearly the increase of vertical displacement with
vertical load in the semi-log plot. For displacements only the data
of the first full cycle and the last cycle are plotted for each cyclic
packet. For the series of tests FV5 there was very little variation
of 1w within each packet of cycles (Fig. 7(a) shows an example)
and for that reason the points of the first cycle are merged with

422

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

Fig. 12. Small displacement response showing curves of normalised loaddisplacement and pore fluid response for tests FV6 (a, b) and FV7 (c, d).

Fig. 13. (a) Normalised load versus displacement variation per cycle and (b) normalised range of pore pressure variation versus average displacement variation.

the points of the last cycle. However, for series FV6 and FV7 this
was not the case, and there are some instances where the last
cycle 1w was larger than for the first cycle as shown in Fig. 7(b).
The data points from series FV5 follow a straight line in the semilog plot, and this is also the trend for series FV7 (except for the

last packet of cycles). The results for series FV6 shows significantly
more displacement for the same loads than for the other two tests.
The difference between the maximum and minimum excess
pore pressure normalised by su is shown in Fig. 13(b) as a function
of 1w/2R (plotted on a log scale), but for more clarity, 1w is

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

423

Table 3
Parameters of the series of cyclic vertical loading tests.
Test

1V (N)

Displacements, w

Pore pressure, u

Kv i (N/mm)

Kv f (N/mm)

Kv i (N/mm)

Kv f (N/mm)

1w a (mm)

wtotal (mm)

Initial

Min (kPa)

Max

Secant

Unloading

FV5_2
FV5_3
FV5_4
FV5_5
FV5_6
FV5_7
FV5_8
FV5_9

47
96
187
280
380
460
530
560

20 591
18 453
5 295
2 794
1 362
563
284
198

16 249
12 316
3 841
1 815
646
344
209
140

17 953
16 025
11 490
7 862
4 629
3 648
2 937
963

13 898
14 347
5 874
5 377
3 266
3 133
1 786
1 458

0.015
0.032
0.14
0.37
1.45
3.3
7.6
14.36

0.01
0.06
0.4
0.9
3.4
8
17.4
45.5

2.25
1.9
3
5.3
6.3
5.9
5
4.8

2.25
1.9
2.2
3
0
2.1
4.3
5.6

3
4.5
9.4
15.3
21.7
26
27.8
28

FV6_2
FV6_3
FV6_4
FV6_5
FV6_6
FV6_7
FV6_8
FV6_9

45
90
180
270
360
450
505
480

9 409
9 100
5 491
3 749
2 684
611
229
87

16 722
8 596
4 248
2 178
653
198
83
58

19 135
14 469
15 310
13 137
18 284
5 301
2 090
1 508

22 615
17 162
16 280
11 307
7 296
5 738
1 934
1 379

0.009
0.023
0.070.09
0.10.23
0.21.06
1.33.7
3.911.7
10.314.5

0.02
0.02
0.07
0.09
0.66
2.3
5.4
3.57

2.8
3.7
4.25
5.2
5
7.5
8.7
11

3.6
4.3
5.8
6.17
8.9
11.8
14.1
12.5

2.7
3.3
3.4
2.3
0.7

FV7_2
FV7_3
FV7_4
FV7_5
FV7_6
FV7_7
FV7_8
FV7_9
FV7_10
FV7_11

46
92
186
275
385
480
580
660
740
744

21 190
13 402
10 926
8 145
6 312
3 729
2 057
1 008
501
160

26 690
18 854
9 685
6 231
4 554
2 468
1 154
539
294
115

29 636
22 413
20 272
18 211
16 707
8 539
6 757
5 588
4 133
2 584

25 140
24 666
22 437
17 725
16 062
9 746
6 106
4 627
3 328
1 689

0.01
0.016
0.042
0.09
0.140.17
0.280.39
0.60.99
1.43
3.265.31
9.7833

0.01
0.01
0.02
0.01
0.02
0.08
0.37
0.96
1.6
2.5

11
11
11.2
11
10.6
9
9
11.9
15
10.6

11
11
11.2
11
10.8
10.7
14
18.2
20
22.3

11.1
11.1
10.8
10.4
8.6
7.1
8.2
10.5
12
7.9

0.3

0.85
0.7

Initial and final range of vertical displacement per cycle.

Fig. 14. Normalised initial and final vertical stiffnesses plotted against normalised displacement variation per cycle.

presented as the average of the first and last cycles. The maximum
difference was found in the series FV5 where the mean load was
250 N, whereas the lowest differences corresponded to the suction
installed caisson (mean load of 0 N).
4.4. Vertical stiffness of the caisson foundation
Normalised secant and unloading stiffness, determined as
illustrated in Fig. 12(c), are shown in Fig. 14. In these figures the
initial and final stiffness (Kv i , Kv f ) are included to reveal whether or
not degradation occurs during the cycles at the same constant load
amplitude 1V . It can be observed that, for all tests, the secant
stiffness shows a clear decrease with absolute displacement (or
load amplitude). Series FV5 also shows a stiffness decrease within
each 10 cycles of loading, whilst FV6 and FV7 show the reverse
in the first set of cycles (1V = 46 N). This only reveals that a
flexible initial response occurred, which is recovered immediately
at the second cycle (as shown in Fig. 12). Fig. 14(b) shows that the
normalised unloading stiffness does not show a strong reduction at

small displacements as in the secant stiffness, but a more regular

decrease along the whole range of 1w .
5. Conclusions
The results of an experimental study of vertical loading of
suction caisson foundations in clay have been described. A suction
caisson model with an aspect ratio of 1 was tested in heavily
overconsolidated kaolin clay. This study considered three stages:
installation, monotonic vertical loading and cyclic vertical loading.
The calculation procedure proposed by Houlsby and Byrne [19]
was used to estimate installation loads successfully. Values for the
sample undrained strength were estimated from shear vane tests
whilst values for and Nc were assumed as 0.3 and 9, respectively.
The value of used was slightly lower than the approach of
Andersen and Jostad [22] where is related to the inverse of
the clay sensitivity. The calculation procedure also provided a
good estimate of the suctions necessary to install a caisson into
clay. The results of the installation tests indicate no substantial
difference between pushed installation and suction. Monotonic

424

F.A. Villalobos et al. / Applied Ocean Research 32 (2010) 414424

compression and tension tests were carried out with the results
broadly confirming the conclusions of Byrne and Cassidy [10].
The ultimate tensile and compressive capacity was equivalent
when converted to a bearing capacity factor. This also compared
favourably to the results of Martin [24]. The tensile loading
response was softened substantially if the caisson had been loaded
into compressive failure prior to tensile loading.
The results of the cyclic vertical loading tests are very relevant
for applications of multiple caisson foundations. The cyclic vertical
loading around a mean vertical load equal to the maximum installation load induced permanent settlement of the caisson, whereas
the cycling around a mean vertical load of zero induced permanent
uplift of the caisson, although for large load amplitudes temporary
settlements were observed during compressive loading.
It was found that, in the short term, substantial difference
occurred in the vertical cyclic loading response between a caisson
installed by pushing and a caisson installed by suction. The
large magnitude of non-dissipated pore water pressure generated
during the suction installation influences the loaddisplacement
response, diminishing substantially the amount of caisson uplift.
Acknowledgements
This work was part of a major industryuniversity project to
study novel foundations for offshore wind turbines. The authors
would like to thank the financial support of the Department of
Trade and Industry, EPSRC and the industry sponsors.
References
[1] Byrne BW, Houlsby GT. Foundation for offshore wind turbines. Philosophical
Transactions of the Royal Society of London, Series A 2003;361:290930.
[2] Kelly RB, Houlsby GT, Byrne BW. Transient vertical loading of model suction
caissons in a pressure chamber. Gotechnique 2006;56(10):66575.
[3] Andersen K, Dyvik R, Schrder K, Hansteen O, Bysveen S. Field tests of
anchors in clay II: predictions and interpretation. Journal of the Geotechnical
Engineering Division ASCE 1993;119(10):153249.
[4] Clukey EC, Morrison MJ, Garnier J, Cort JF. The response of suction caissons
in normally consolidated clays to cyclic TLP loading conditions. In: Offshore
technology conference. Paper 7796. 1995.

[5] El-Gharbawy SL. The pullout capacity of suction caisson foundations for
tension leg platforms. Ph.D. thesis. University of Texas at Austin; 1998.
[6] Andersen K, Jostad HP. Foundation design of skirted foundations and anchors
in clay. In: Offshore technology conference. Paper 10824. 1999.
[7] House A. Suction caisson foundations for buoyant offshore facilities. Ph.D.
thesis. University of Western Australia; 2002.
[8] Colliat JL, Dendani H. Girassol: geotechnical design analyses and installation of
the suction anchors. In: Proceedings of the society for underwater technology
SUT conference. 2002. p. 10719.
[9] Andersen KH, Murff JD, Randolph MF, Clukey EC, Erbrich CT, Jostad HP. et al.
Suction anchors for deepwater applications. In: International symposium on
frontiers in offshore geotechnics. ISFOG. Perth. 2005. p. 330.
[10] Byrne BW, Cassidy MJ. Investigating the response of offshore foundations in
soft clay soils. In: Proc. 21st International conference on offshore mechanics
and arctic engineering OMAE02. Paper OMAE2002-28057. 2002.
[11] Cassidy MJ, Byrne BW, Randolph MF. A comparison of the combined load
behaviour of spudcan and caisson foundations on soft normally consolidated
clay. Gotechnique 2004;54(2):91106.
[12] Houlsby GT, Kelly RB, Huxtable J, Byrne BW. Field trials of suction caissons
in clay for offshore wind turbine foundations. Gotechnique 2005;55(4):
287296.
[13] Kelly RB, Houlsby GT, Byrne BW. A comparison of field and laboratory tests of
caisson foundations in sand and clay. Gotechnique 2006;56(9):61726.
[14] de Santa Maria PEL. Behaviour of footings for offshore structures under
combined loads. D.Phil. thesis. University of Oxford; 1988.
[15] Martin CM. Physical and numerical modelling of offshore foundations under
combined loads. D.Phil. thesis. University of Oxford; 1994.
[16] Villalobos FA. Model testing of foundations for offshore wind turbines. D.Phil.
thesis. University of Oxford; 2006.
[17] Butterfield R, Houlsby GT, Gottardi G. Standardised sign conventions and
notation for generally loaded foundations. Gotechnique 1997;47:10514.
[18] Byrne BW. Investigations of suction caissons in dense sand. D.Phil. thesis.
University of Oxford; 2000.
[19] Houlsby GT, Byrne BW. Design procedures for installation of suction caissons
in clay and other materials. Proceedings of the ICE, Geotechnical Engineering
2005;158(2):7582.
[20] Chen W, Randolph M. Radial stress changes around caissons installed in clay
by jacking and by suction. In: International offshore and polar engineering
conference. ISOPE. 2004. p. 4939.
[21] Rauch AF, Olson RE, Luke AM, Mecham EC. Measured response during
laboratory installation of suction caissons. In: International offshore and polar
engineering conference. ISOPE. 2003, p. 7807.
[22] Andersen K, Jostad HP. Shear strength along outside wall of suction anchors
in clay after installation. In: Offshore and polar engineering conference. ISOPE.
2002. p. 78594.
[23] Skempton AW, Northey RD. The sensitivity of clays. Gotechnique 1952;3(1):
3053.
[24] Martin CM. Vertical bearing capacity of skirted circular foundations on Tresca
soil. In: Proc. 15th int. conf. on soil mechanics and geotechnical engineering,
vol. 1. 2001. p. 7436.