Professional Documents
Culture Documents
www.isope.org
Design of Piles for Offshore Wind Energy Foundations with Respect to Horizontal Loading
Martin Achmus and Khalid Abdel-Rahman
Institute for Geotechnical Engineering, Leibniz University of Hannover
Hannover, Germany
ABSTRACT
Monopiles and tripod or jacket structures are favorized foundation
solutions for offshore wind energy converters. The piles of these
structures are subject to intensive cyclic loading by wind and wave
actions. Degradation of pile capacity as well as pile stiffness and
accumulation of deflections can occur and has to be considered in the
design according to German guidelines. The state of knowledge
regarding the behavior of piles under cyclic horizontal loading is
discussed and methods to account for cyclic load effects in the design
are presented.
INTRODUCTION
In the North Sea and the Baltic Sea in Europe a vast number of offshore
wind farms are being planned and several have already been installed in
recent years. In the beginning, in most cases wind farms were erected in
moderate water depths (less than 20m) and monopile foundations have
been built as support structures for the wind tower and the turbine.
Diameters of up to 5m have been realized recently, and monopiles with
diameters of 6m or even more are under planning.
Design methods and experience with offshore piles exist mainly from
structures built by the oil and gas industry. However, the loading
conditions for offshore wind mill foundations are different. The vertical
loads are much smaller than for usual oil or gas platforms, and thus the
horizontal loads at least for monopiles are of similar magnitude
compared to the vertical loads. This means that the extremely cyclic
nature of wind and wave forces is much more important than for very
heavy structures. Due to that, consideration of cyclic load effects is
extremely important. According to German guidelines, the effects of
cyclic loading on the pile behavior have to be considered explicitly in
the design. In that, a 35h-design storm has to be taken into account with
respect to degradation of capacity. For the estimation of deflection,
cyclic load spectra representing all loads over the structures lifetime
have to be considered. An accurate prediction of permanent deformations is of particular importance for monopile foundations, since
wind energy converters are relatively sensitive to permanent tilting.
143
p y 1 / 3
u
p= 2 y
c
pu
p-y Method
The usual design procedure for foundations of offshore wind energy
converters in Germany is given in the Germanische Lloyd (GL) rules
and regulations (GL, 2005). In these regulations, concerning the
behavior of piles under horizontal loading reference is made to the
regulation code of the American Petroleum Institute (API, 2000). In the
API code the p-y method is recommended for the design of horizontally
loaded piles. Also the Scandinavian guideline DNV (2007) proposes
the p-y method for the design of horizontally loaded piles.
In principle, the p-y method is a subgrade modulus method with nonlinear and depth-dependent load-deformation (p-y) characteristics of
the soil springs. API (2000) and DNV (2007) describe the construction
of p-y-curves for soft and stiff clay as well as for sandy soils.
(1a)
pud = c3 D ' z
(1b)
for y 8 y c
(4)
for y > 8 y c
pu y
2 y c
y 3yc
p = 0.72 p u 1 (1 z / z R )
12 y c
0.72 p u
for y 3 y c
for 3 y c < y 15 y c
(5)
for y > 15 y c
The first mentioned equation applies to small depths (pus) and the
second equation to larger depths (pud), the smaller of both values is
to be considered. The influence of the internal friction angle is
described by the factors c1, c2 and c3, which are given in API (2000)
dependent on the angle of internal friction of the sand.
The p-y-curve is described by the following equation:
kz
p = A p u tanh
y
A pu
(2)
Fig. 2. p-y curves for cohesive soil according to DNV (2007)
with A = 3.0 0.8 z/D for static loading and A = 0.9 for cyclic
loading.
Here p is the soil resistance per unit length of the pile and y is the
horizontal deflection. The parameter k also given in API (2000)
describes the initial modulus of subgrade reaction and is dependent
on the relative density ID and with that on the angle of internal
friction.
The approach was proposed by Matlock (1970) based on load tests with
piles with a diameter of ca. 0.3m in lightly overconsolidated clays.
The p-y method was used successfully over decades in the design of
offshore structures. However, no experience exists with the application
to monopiles with very large diameter. Most applications and also the
pile load tests used for the calibration of the p-y approaches were done
with flexible piles. Monopiles have a very large bending stiffness and
behave thus more like rigid piles, which means that a significant
deflection of the pile toe (toe kick) usually occurs under extreme
loading.
The method was originally derived by Reese et al. (1974) based on the
results of a number of pile load tests (Cox et al. 1974). The piles in
these tests had a diameter of 0.61m. Murchison & ONeill (1984)
modified the approach by introducing the tanh-function. The parameter
A is a calibration factor introduced to match the theoretical approach
for the maximum soil reaction pu with the experimental results.
Wiemann et al. (2004) showed that the p-y approach for sand
overestimates the soil stiffness for large diameter piles in larger depths.
This has a minor effect for flexible piles, since here the deflections in
larger depths are very small or zero. However, for very stiff piles this
overestimation matters. Abdel-Rahman & Achmus (2005) and also
Achmus et al. (2008) carried out numerical simulations and showed
that the p-y method for monopiles in sand significantly underpredicts
pile head deflections. Recently, Soerensen et al. (2010) proposed an
approach with under-linear and diameter-dependent increase of the
initial stiffness of the p-y curves, based on numerical simulations.
For cohesive soils, the following p-y approach is given in DNV (2007):
The static ultimate lateral resistance shall be calculated as
(3 c + ' z ) D + J cu z
pu = u
9 cu D
(3)
144
modified p-y approach can then be used for the design of the other
monopiles in the wind farm.
Of course the material law used for the soil in the simulations must
be capable to predict the pile behavior sufficiently accurate. To
check that, a comparison of the results for a small-diameter pile
with the results from p-y method calculations can be carried out.
Fig. 3 left shows the deflection lines for monopiles with different
diameters in homogeneous dense sand, derived once by the p-y method
and once by a numerical simulation with an elastoplastic material law
for the sand. To ensure a similar pile behaviour, different pile lengths
were examined, and typical service loads for the different pile
geometries were applied.
Both methods have already been used in wind farm projects with
monopile foundations. The latter procedure is elucidated in Fig. 4. First
the numerical model is validated by comparison of results for a pile
with an ordinary (small) diameter. Then the large-diameter pile is
analyzed and a modified p-y approach is seeked which predicts similar
deflections as obtained in the numerical simulation.
For piles supporting offshore wind towers, up to 109 load cycles have to
be considered over the structures lifetime of usually 20 or 25 years.
Although the load amplitudes vary and the largest part of the cycles
have small amplitudes, the question has to be answered how the actual
number of cyclic loads can be considered in the design. An important
aspect in that regard is how to determine the minimum pile length to
ensure sufficient stability, i.e. limited deflections, under cyclic load
action.
145
Accumulation of Deflections
It is known from different experimental investigations that the
deflections of a horizontally loaded pile increase under one-way cyclic
loading. As an example, model test results of Hettler (1981) for flexible
piles in homogeneous sand are shown in Fig. 5.
Lin & Liao (1999) proposed to determine the parameter t in Eq. 8 with
the following equation:
t = 0,032
Here yN and y1 are the horizontal pile head deflections after N load
cycles and after 1 load cycle (static loading), respectively. fN(N) is a
function which describes the increase of deflections. As long as the
cyclic load amplitude is well below the ultimate pile capacity, sedation
behavior can be expected, which means that the deflection rate
decreases with increasing number of load cycles. The most common
functions for piles under cyclic loading that are found in literature are
of the exponential type such as Eq. 7 (e.g. Little & Briaud 1988) and of
logarithmic type such as Eq. 8 (e.g. Hettler 1981):
(7)
f N = 1 + t ln N
(8)
(10)
(6)
fN = N m
l0
E sN cp , N =1
a
E s1
cp , N
a
(9)
146
(11)
Fig. 6. Degradation of secant modulus under cyclic loading in the pilesoil model (schematic)
(12)
Here N is the number of load cycles and X is the cyclic stress ratio
defined by Huurman (1996) for cohesionless material as follows:
X =
1,cyc
1, sf
(13)
where 1,sf is the major principal stress at static failure state and 1,cyc is
the major principal stress for the cyclic stress state under consideration.
The cyclic stress ratio is thus dependent on the confining pressure and
on the cyclic stress level.
From cyclic triaxial test results documented in the literature, typical
regression parameters b1 and b2 were found for dense sand to be b1 =
0.20, b2 = 5.76 and for medium dense sand b1 = 0.16, b2 = 0.38 (Kuo
2008).
A problem to be dealt with is that the Equations 11 and 12 are valid for
triaxial test conditions with isotropic initial stress conditions and a
constant confining pressure 3 during cyclic loading. In the pile-soil
system, the initial stress conditions (before application of the horizontal
load) are anisotropic and the minor principal stress in the elements as
well as the direction of the principal stress axes in general change with
the application of the load. To overcome this problem, a characteristic
cyclic stress ratio Xc is defined here as
XC =
X (1) X ( 0)
1 X ( 0)
(14)
Here the index (1) means the cyclic stress ratio at loading phase and the
index (0) means at unloading phase (cf Fig. 7). At the initial (and
unloading) phase, only the vertical load V due to the tower weight is
considered, and the lateral load H is applied subsequently in the loading
phase. The characteristic cyclic stress ratio is derived from the
difference between the stress ratios in the loading and the unloading
phase. Due to the denominator in Eq. 14 this value varies from 0 to 1.
The accumulation of plastic strain and the degradation of stiffness of
the soil element can be obtained from Eq. 12 by replacing X by Xc.
Lin & Liao (1999) proposed the use of a strain superposition method.
Here the static pile head deflections under the different loading classes
are used to assess the effect on the accumulation of deflections. The
resultant head deflection under n different loading classes Bi (defined
by bending moment and horizontal load at pile head) is obtained by the
following equation:
In the last step of the simulation (model C in Fig. 7), the deformation
response of the system is analyzed using the degradation stiffnesses
obtained from evaluation of models A and B.
147
y res = y1,1 1 + t ln N1 + N k*
k =2
with
N k*
=e
(15)
1 y1, k
(1 + t ln N k )1
t y1,1
Here y1,i is the (static) pile deflection under the load Bi and Ni is the
respective number of load cycles. The approach assumes the validity of
Eq. 8 regarding the cyclic increase of deflections.
CSR =
( 1 3 ) cyc
( 1 3 ) f ,stat
(16)
The critical cyclic stress ratio CSRcrit is reached when the excess pore
pressure progressively increases with increasing time or load cycle
number, respectively. In the design of the pile it has then to be ensured
that the actual cyclic stress ratio remains well below the critical value.
148
The problem with the assessment of such tests is that the boundary
conditions in the tests do not coincide with the actual boundary
conditions in situ. At the pile, in particular the drainage conditions are
different to the test conditions. At the pile high excess pore pressures
occur only locally, and drainage can occur upwards, downwards or
horizontally around the pile, which was shown in numerical simulations
from Grabe et al. (2004). Moreover, the magnitude of excess pore
pressures in cyclic tests is frequency dependent.
It should be noted that the calculation methods give the pile head
deflection and rotation after N load cycles under load. Thus, the elastic
portion has to be subtracted to obtain the (plastic) permanent deflection.
As an approximation, the elastic pile deflection line can be determined
by calculating the pile behavior using linear springs having the initial
stiffnesses of the p-y curves.
CONCLUSIONS
For the design of horizontally loaded piles for offshore wind energy
foundations, explicit consideration of capacity and stiffness reductions
due to cyclic loading is required from German authorities. Also, the
permanent deflections to be expected over the structures lifetime has to
be predicted.
The paper in hand presents the knowledge regarding these questions
and makes evident that research is needed to clarify several aspects of
pile behavior under cyclic loading.
Methods and possibilities to account for cyclic load effects in the
design are presented. In general the p-y method can be used in the
design, For ULS design, additionally the results of undrained cyclic
triaxial tests can be used to verify the results. With regard to
accumulated deflections and rotations, a numerical simulation method
is presented. This method makes use of the results of drained cyclic
triaxial tests and helps to assess the stability of cyclic horizontally
loaded piles.
REFERENCES
149
150