Professional Documents
Culture Documents
Publication date:
2009
Document Version
Publisher's PDF, also known as Version of record
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.
ISSN 1901-726X
DCE Technical Report No. 57 Department of Civil Engineering
Aalborg University
Department of Civil Engineering
Water & Soil
by
K.T. Brødbæk
M. Møller
S.P.H. Sørensen
A.H. Augustesen
January 2009
© Aalborg University
Scientific Publications at the Department of Civil Engineering
Technical Reports are published for timely dissemination of research results and scientific work
carried out at the Department of Civil Engineering (DCE) at Aalborg University. This medium
allows publication of more detailed explanations and results than typically allowed in scientific
journals.
Technical Memoranda are produced to enable the preliminary dissemination of scientific work by
the personnel of the DCE where such release is deemed to be appropriate. Documents of this kind
may be incomplete or temporary versions of papers—or part of continuing work. This should be
kept in mind when references are given to publications of this kind.
Contract Reports are produced to report scientific work carried out under contract. Publications of
this kind contain confidential matter and are reserved for the sponsors and the DCE. Therefore,
Contract Reports are generally not available for public circulation.
Lecture Notes contain material produced by the lecturers at the DCE for educational purposes. This
may be scientific notes, lecture books, example problems or manuals for laboratory work, or
computer programs developed at the DCE.
Theses are monograms or collections of papers published to report the scientific work carried out at
the DCE to obtain a degree as either PhD or Doctor of Technology. The thesis is publicly available
after the defence of the degree.
Latest News is published to enable rapid communication of information about scientific work
carried out at the DCE. This includes the status of research projects, developments in the
laboratories, information about collaborative work and recent research results.
Published 2009 by
Aalborg University
Department of Civil Engineering
Sohngaardsholmsvej 57,
DK-9000 Aalborg, Denmark
ISSN 1901-726X
DCE Technical Report No. 57
Review of py relationships in cohesionless soil
K. T. Brødbæk1 ; M. Møller2 ; S. P. H. Sørensen3 ; and A. H. Augustesen4
Aalborg University, January 2009
Abstract
Monopiles are an often used foundation concept for oshore wind turbine converters.
These piles are highly subjected to lateral loads and thereby bending moments due to
wind and wave forces. To ensure enough stiness of the foundation and an acceptable
pile-head deection, monopiles with diameters at 4 to 6 m are typically necessary. In
current practice these piles are normally designed by use of the py curve method
although the method is developed and veried for small diameter, slender piles with a
diameter up to approximately 2 m. In the present paper a review of existing py curve
formulations for piles in sand under static loading is presented. Based on numerical
and experimental studies presented in the literature, advances and limitations of
current py curve formulations are outlined.
1
2 Review of py relationships in cohesionless soil
the lateral deection of the pile, y . Note Full-scale tests though substantiate a non-
that there in present paper is distinguished linear relationship between soil resistance
between soil resistance, p, and ultimate and pile deection. The subgrade reac-
soil resistance, pu . The soil resistance is tion method must therefore be considered
given as the reaction force per unit length too simple and highly inaccurate. In addi-
acting on the pile until reaching the ulti- tion the subgrade reaction method is not
mate soil resistance. The ultimate soil re- able to predict the ultimate lateral resis-
sistance is given as the point of maximum tance. The py curve method assumes
soil resistance. a non-linear dependency between soil re-
sistance and pile deection and is there-
Several formulations of py curves exist fore able to produce a more accurate so-
depending on the type of soil. These for- lution. Furthermore the ultimate lateral
mulations are originally formulated to be resistance can be estimated by using the
employed in the oshore gas and oil sec- py curve method.
tor. However they are also used for o-
shore wind turbine foundations, although In both the py curve method and the
piles with signicantly larger diameter and subgrade reaction method the winkler ap-
signicantly smaller slenderness ratio are proach, cf. section 2, is employed to cal-
employed for this type of foundation. culate the lateral deection of the pile and
internal forces in the pile. When employ-
In this paper the formulation and imple-
ing the Winkler approach the pile is con-
mentation of py curves proposed by Reese
sidered as a beam on an elastic foundation.
et al. (1974) and API (1993) for piles in
The beam is supported by a number of un-
sands due to static loading will be anal-
coupled springs with spring stiness given
ysed. However, alternative methods for
by py curves. When using the Winkler
designing laterally loaded piles have been
approach the soil continuity is not taken
proposed in the literature. According to
into account as the springs are considered
Fan and Long (2005) these alternative ap-
uncoupled.
proaches can generally be classied as fol-
lows: The elasticity method, e.g. Banerjee and
Davis (1978), Poulos (1971), and Pou-
The limit state method. los and Davis (1980), includes the soil
continuity. However, the response is as-
The subgrade reaction method. sumed to be elastic. As soil is more likely
to behave elasto-plastically, this elasticity
The elasticity method. method is not to be preferred unless only
small strains are considered. Hence, the
The nite element method.
method is only valid for small strains and
thereby not valid for calculating the ulti-
Simplest of all the methods are the limit mate lateral resistance.
state methods, e.g. Broms (1964), consid-
ering only the ultimate soil resistance. Another way to deal with the soil con-
tinuity and the non-linear behaviour is
The simplest method for predicting the to apply a three-dimensional nite ele-
soil resistance due to a given applied hor- ment model, e.g. Abdel-Rahman and
izontal deection is the subgrade reaction Achmus (2005). When applying a three-
method, e.g. Reese and Matlock (1956) dimensional nite element model both de-
and Matlock and Reese (1960). In this formations and the ultimate lateral re-
case the soil resistance is assumed lin- sistance can be calculated. Due to the
early dependent on the pile deection. complexity of a three-dimensional model,
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 3
As a consequence of the oil and gas indus- Murchison and O'Neill (1984) com-
try's expansion in oshore platforms in the pare the py curve formulation pro-
1950s, models for design of laterally loaded posed by Reese et al. (1974)
piles were required. The key problem is with three simplied expressions
the soil-structure interaction as the sti- (also based on the Mustang Island
ness parameters of the pile, Ep , and the tests) by testing the formulations
soil, Es , may be well known but at the against a database of relatively well-
soil-pile interface the combined parameter documented lateral pile load tests. A
Epy is governing and unknown. In order to hyperbolic form is found to provide
investigate this soil-pile interaction a num- better results compared to the origi-
ber of full-scale tests on fully instrumented nal expressions formulated by Reese
piles have been conducted and various ex- et al. (1974).
pressions depending on the soil conditions
have been derived to predict the soil pres-
sure on a pile subjected to lateral loading. Research has been concentrated on deriv-
ing empirical (e.g. Reese et al. 1974)
Historically, the derivation of the py and analytical (e.g. Ashour et al. 1998)
curve method for piles in sand is as fol- py curve formulations for dierent types
lows: of soil giving the soil resistance, p, as a
function of pile displacement, y , at a given
Analysing the response of beams on point along the pile. The soil pressure at
an elastic foundation. The soil is a given depth, xt , before and during a sta-
characterised by a series of linear- tic excitation is sketched in g. 1b. The
elastic uncoupled springs, introduced passive pressure on the front of the pile is
by Winkler (1867). increased as the pile is deected a distance
Hetenyi (1946) presents a solution to yt while the active pressure at the back is
the beam on elastic foundation prob- decreased.
lem.
An example of a typical py curve is shown
McClelland and Focht (1958) as well in g 2a. The curve has an upper horizon-
as Reese and Matlock (1956) suggest tal limit denoted by the ultimate soil resis-
the basic principles in the py curve tance, pu . The horizontal line implies that
method. the soil behaves plastically meaning that
4 Review of py relationships in cohesionless soil
Following relations are used: In (5) the shear strain, γ , in the beam is
neglected. This assumption is only valid
M = Ep Ip κ (2)
for relatively slender beams. For short and
dV rigid beams the Timoshenko beam theory,
= −p (3)
dx that takes the shear strain into account,
p(y) = −Epy y (4) is preferable. The following relations are
used:
Ep and Ip are the Young's modulus of elas-
ticity of the pile and the second moment V = Gp Av γ (6)
of inertia of the pile, respectively. κ is the dy
curvature strain of the beam element. γ= −ω (7)
dx
dω
κ= (8)
dx
3 Formulations of py
curves for sand
Pile 1 Pile 2
2.7
Load
arrangement
0.9
0.3
0.0
Elevation
Instrumented
[m]
section
Diaphragm
8.8
9.5
Reaction
piles
Embedment length = 21 m
5 2
21.0 EpIp = 1.7144 x 10 kNm
Figure 6: Set-up for Mustang Island tests, after Cox et al. (1974).
Prior to pile installation, two soil borings determined as no failure occurred in the
were made, each in a range of 3.0 m (10 pile. After the static load test on pile 1 two
ft) from a test pile. The soil samples cyclic load tests were conducted. 52 days
showed a slight dierence between the two after installation a pull-out test was con-
areas where the piles were installed, as ducted on pile 2. A maximum of 1780 kN
one boring contained ne sand in the top (400000 lb) was applied causing the pile
12 m and the other contained silty ne to move 25 mm (1 inch). After another
sand. The strength parameters were de- week pile 2 was subjected to three cases
rived from standard penetration tests ac- of cyclic loading and nally a static load
cording to Peck et al. (1953). The stan- test. The static load case on pile 2 was per-
dard penetration tests showed large vari- formed immediately after the third cyclic
ations in the number of blows per ft. Es- load case which might aect the results.
pecially in the top 40 ft of both borings Reese et al. (1974) do not clarify whether
the number of blows per ft varied from 10 this eect is considered in the analyses.
to 80. From 40 to 50 ft beneath the mud-
line clay was encountered. Beneath the
clay layer the strength increased from 40
3.2 Method A
to 110 blows per ft. From 60 ft beneath
the mudline to the total depth the num-
ber of blows per ft decreased from 110 to Method A is the original method based on
15. the Mustang Island tests, cf. Reese et al.
(1974). The py curve formulation con-
The piles were in total subjected to seven sists of three curves: an initial straight
horizontal load cases consisting of two sta- line, p1 , a parabola, p2 , and a straight line,
tic and ve cyclic. Pile 1 was at rst sub- p3 , all assembled to one continuous piece-
jected to a static load test 16 days after in- wise dierentiable curve, cf. g. 7. The
stallation. The load was applied in steps last straight line from (ym ,pm ) to (yu ,pu )
until a maximum load of 267 kN (60000 is bounded by an upper limit characterised
lb) was reached. The maximum load was by the ultimate resistance, pu .
8 Review of py relationships in cohesionless soil
Direction
of
movement
Fp
Figure 7: py curve for static loading using
method A, after Reese et al. (1974).
D
ically admissible failure modes. At shal- Figure 9: Failure mode for deep depths, after
low depths a wedge will form in front of Reese et al. (1974).
the pile assuming that the Mohr-Coulomb
failure theory is valid. Reese et al. (1974)
calculated according to (12) and (13):
uses the wedge shown in g. 8 to analyt-
ically calculate the passive ultimate resis- K0 x tan ϕtr sin β
tance at shallow depths, pcs . By using this pcs = γ 0 x (12)
tan(β − ϕtr ) cos α
failure mode a smooth pile is assumed, and tan β
therefore no tangential forces occur at the + γ0x (D − x tan β tan α)
tan(β − ϕtr )
pile surface. The active force is also com-
+ γ 0 x(K0 x tan ϕtr (tan ϕtr sin β − tan α)
puted from Rankine's failure mode, using
the minimum coecient of active earth − Ka D)
pressure. pcd = Ka Dγ 0 x(tan8 β − 1) (13)
+ K0 Dγ 0 x tan ϕtr tan4 β
At deep depths the sand will, in contrast pcs is valid for shallow depths and pcd for
to shallow depths, ow around the pile and deep depths, γ 0 is the eective unit weight,
a statical failure mode as sketched in g. 9 and ϕtr is the internal angle of friction
is used to calculate the ultimate resistance. based on triaxial tests. The factors α and
The transition depth between these failure β measured in degrees can be estimated by
modes occurs, at the depth where the ul- the following relations:
timate resistances calculated based on the ϕtr
two failure modes are identical. α= (14)
2
ϕtr
β =45◦ + (15)
2
The ultimate resistance per unit length of Hence, the angle β is estimated according
the pile can for the two failure modes be to Rankine's theory which is valid if the
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 9
x [m]
and horizontal earth pressure at rest, re- −15
spectively:
−20
ϕtr
Ka = tan2 (45 − ) (16) −25
2
K0 = 0.4 (17) −30
0 1 2 3 4 5 6
pc [kN/m] 4
x 10
The value of K0 depends on several fac- (a) D = 1.0 m
tors, e.g. the friction angle, but (17) does
0
not reect that. φtr=30°
−20 °
φtr=40
The theoretical ultimate resistance, pc , as
function of depth is shown in g. 10. −40
−60
with diameter and angle of internal fric-
tion. Hence, for piles with a low slender- −80
3
50000
k [kPa/m]
4
x 40000
x
D D >5, A=0.88
5
30000
Below the
water table
6
20000
of the py curves, in which the analyti- is a considerable dierence in soil resis-
cal expressions for the ultimate resistance, tance predicted by the two methods when
(12) and (13), are approximated using the considering the section between the points
dimensionless parameters C1 , C2 and C3 : (yk ,pk ) and (yu ,pu ) as shown in g. 14.
µ ¶ The soil parameters from tab. 2 has been
pus = (C1 x + C2 D)γ 0 x used to construct the py curves shown in
pu = min
pud = C3 Dγ 0 x g. 14.
(28)
6 120 2500
C3
p [kN/m]
2000
5 C1
100 x=5m
1500
4 80 Method A
1000
C 1 and C 2 [-]
Method B
C 3 [-]
k
3 60 500 m
C2
u
0
2 40 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
y [m]
Figure 13: Variation of the parameters C1 , C2 Table 2: Soil parameters used for plotting the
and C3 as function of angle of internal friction, py curves in g. 14.
after API (1993).
γ0 φtr D k
[kN/m3 ] [◦ ] [m] [kN/m3 ]
A hyperbolic formula is used to describe
10 30 4.2 8000
the relationship between soil resistance
and pile deection instead of a piecewise
formulation as proposed by method A:
3.4 Comparison of methods
kx
p(y) = Apu tanh( y) (29)
Apu A comparison of both static and cyclic py
The coecient A could either be deter- curves has been made by Murchison and
mined from g. 11a or by: O'Neill (1984) based on a database of 14
µ ¶ full-scale tests on 10 dierent sites. The
H pile diameters varied from 51 mm (2 in.)
A = 3.0 − 0.8 ≥ 0.9 (30)
D to 1.22 m (48 in.). Both timber, concrete
and steel piles were considered. The soil
Since: friction angle ranged from 23◦ to 42◦ . The
kx test piles' slenderness ratio's were not pro-
dp Apu vided.
|y=0 = Apu |y=0 = kx (31)
dy cosh2 ( kxy ) Apu
Murchison and O'Neill (1984) compared
the py curve's initial slope is then simi- the dierent py curve formulations with
lar using the two methods, cf. (20). Also the full-scale tests using the Winkler ap-
the upper bound of soil resistance will ap- proach. The predicted head deection,
proximately be the same. However, there maximum moment, Mmax , and the depth
12 Review of py relationships in cohesionless soil
of maximum moment were compared ac- Murchison and O'Neill (1984) state that
cording to the error, E : the sizes of the errors in tab. 3 can-not be
explained by parameter uncertainty. The
|predicted value - measured value| amount of data included in the database
E=
measured value was very small due to the unavailability of
(32)
appropriately documented full-scale tests
In the analysis it was desired to assess
and Murchison and O'Neill (1984) there-
the formulations ability to predict the be-
fore concluded that a further study of the
haviour of steel pipe monopiles. Multi-
soil-pile interaction was needed.
plication factors were therefore employed.
The error, E , was multiplied by a factor of
two for pipe piles, 1.5 for non-pipe driven
piles and a factor of one for drilled piers. 4 Limitations of py curves
When predicted values were lower than the
measured values the error was multiplied
The py curve formulations for piles in co-
by a factor of two. By using these factors
hesionless soils are, as described, devel-
unconservative results are penalised and
oped for piles with diameters much less
pipe piles are valued higher in the com-
than 4 to 6 m which is often necessary
parison. In tab. 3 the average value of E
for nowadays monopiles. Today, there
for static py curves are shown for the two
is no approved method for dealing with
methods. As shown, method B results in
these large diameter oshore piles, which
a lower average value of E for all the cri-
is probably why the design regulations
teria considered in the comparison. The
are still adopting the original py curves,
standard deviation of E was not provided
cf. Reese et al. (1974), API(1993), and
in the comparison.
DNV(1992).
Table 3: Average values of the error, E . The The py curve formulations are, as de-
methods are compared for pile-head deection, scribed, derived based on the Mustang Is-
maximum moment and depth to maximum mo- land tests which included only two piles
ment. and a total of seven load cases. Further-
Pile-head Mmax Depth to more, the tests were conducted for only
deection Mmax one pile diameter, one type of sand, only
Method A 2.08 0.75 0.58 circular piles etc. Taken into account the
Method B 1.44 0.44 0.40 number of factors that might aect the
behaviour of a laterally loaded pile and
Murchison and O'Neill (1984) analysed the very limited number of full-scale tests
the sensitivity to parameter variation for performed to validate the method, the in-
method B. The initial modulus of sub- uence of a broad spectre of parameters
grade reaction, k , the internal friction an- on the py curves are still to be claried.
gle, ϕ, and the eective unit weight, γ 0 , Especially when considering oshore wind
were varied. They found that a 10 % in- turbine foundations a validation of sti
crease in either ϕ or γ 0 resulted in an in- piles with a slenderness ratio of L/D < 10
crease in pile-head deection of up to 15 is needed as the Mustang Island test piles
and 10 %, respectively. For an increase had a slenderness ratio of L/D = 34.4. It
of 25 % in k an increase of up to 10 is desirable to investigate this as it might
% of the pile-head deection was found. have a signicant eect on the initial sti-
The sensitivity analysis also shows that ness which is not accounted for in the py
k has the greatest inuence on pile-head curve method. Briaud et al. (1984) postu-
deection at small deections and that ϕ late that the soil response depends on the
has a great inuence at large deections. exibility of the pile. Criteria for sti ver-
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 13
sus exible behaviour of piles have been slope and hereby the initial stiness of the
proposed by various authors, for example soil-pile system.
Dobry et al. (1982), Budhu and Davies
(1987), and Poulus and Hull (1989). The When using the py curve method the pile
dierence in deformation behaviour of a bending stiness is employed when solving
sti and a exible pile is shown in g. the beam on an elastic foundation prob-
15. A pile behaves rigidly according to lem. However, no importance is attached
the following criterion, cf. Poulus and Hull to the pile bending stiness in the formu-
(1989): lation of the py curves, hence Epy is inde-
pendent of the pile properties. The valid-
µ ¶0.25
Ep Ip ity of this assumption can be questioned as
L < 1.48 (33) Epy is a parameter describing the soil-pile
Es
interaction.
Es is Young's modulus of elasticity of the
soil. The criterion for a exible pile be- When decoupling the non-linear springs
haviour is: associated with the Winkler approach an-
µ ¶ other error is introduced since the soil in
Ep Ip 0.25 reality acts as a continuum.
L > 4.44 (34)
Es
In the following a number of assumptions
According to (33) a monopile with an and not claried parameters related to the
outer diameter of 4 m, an embedded length py curve method are treated separately.
of 20 m and a wall thickness of 0.05 m be- The treated assumptions and parameters
haves rigidly if Es < 7.6 MPa. In con- are:
trast, the pile exhibits a exible behaviour
if Es > 617 MPa. Even dense sands have
Es < 100 MPa, so the recently installed Shearing force between soil layers.
monopiles behave, more like a rigid pile
than a exible. The ultimate soil resistance.
Figure 15: Rigid versus exible pile behaviour. 4.1 Shearing force between soil
layers
For modern wind turbine foundations only
small pile head rotations are acceptable. Employing the Winkler approach the soil
Furthermore, the strict demands to the response is divided into layers each rep-
total stiness of the system due to reso- resented by a non-linear spring. As the
nance in the serviceability mode increase springs are uncoupled the layers are con-
the signicance of the py curve's initial sidered to be independent of the lateral
14 Review of py relationships in cohesionless soil
pile deection above and below that spe- soil resistance. The method for estimating
cic layer, i.e. the soil layers are consid- pu is therefore evaluated in the following.
ered as smooth layers able to move rel-
atively to each other without loss of en-
ergy to friction. Pasternak (1954) modi- Failure modes
ed the Winkler approach by taking the
shear stress between soil layers into ac-
When designing large diameter monopiles
count. The horizontal load per length of
in sand, the transition depth will most of-
the pile is given by:
ten occur beneath the pile toe, cf. g.
p dy 10b. There are however several uncertain-
p(y) = −Epy y − Gs (35) ties concerning the ultimate resistance at
dx
shallow depths.
where Gs is the soil shear modulus. The
subgrade reaction modulus Epy given in The prescribed method for calculating the
tab. 1 may indirectly contain the soil shear ultimate resistance at shallow depths as-
stiness as the py curve formulation is sumes that the pile is smooth, i.e. no
p
tted to full-scale tests. Epy is a modulus skin friction appears and a Rankine fail-
of subgrade reaction without contribution ure mode will form. However, in real-
from the soil shear stiness. ity a pile is neither perfectly rough nor
perfectly smooth, and the assumed failure
Belkhir (1999) examines the signicance
mechanism is therefore not exactly cor-
of shear between soil layers by compar-
rect. According to Harremoës et al. (1984)
ing the CAPELA design code, which takes
a Rankine failure takes place for a per-
the shear between soil layers into account,
fectly smooth wall and a Prandtl failure
with the French PILATE design code,
for a perfectly rough wall, cf. g. 16a and
which deals with smooth boundaries. The
g. 16b, respectively. Due to the fact that
two design codes are compared with the
the pile is neither smooth nor rough a com-
results of 59 centrifuge tests conducted
bination of a Rankine and Prandtl failure
on long and slender piles. Analyses show
will occur. Furthermore, the failure modes
concordance between the two design codes
are derived for a two-dimensional case.
when shear between soil layers is not taken
into account. Furthermore, the analyses In (12) the angle α, which determines the
shows a reduction from 14 % to 5 % in the horizontal spread of the wedge, appears.
dierence between the maximum moments Through experiments Reese et al. (1974)
determined from the centrifuge tests and postulate that α depends on both the void
the numerical simulations when taking the ratio, friction angle, and the type of load-
shear between the soil layers into account. ing. However, the inuence of void ratio
However, it is not clear from the paper and type of loading is neglected in the ex-
whether or not the shear between soil lay- pression of α.
ers is dependent on pile diameter, slender-
ness ratio etc. Furthermore it is not clar- Nowadays monopiles are non-slender piles
ied whether the authors distinguish be- with high bending stiness. The piles will
p
tween Epy and Epy . therefore deect as almost rigid piles and
rather large deformations will occur be-
neath the point of zero deection. How-
4.2 The ultimate soil resistance ever, when calculating the ultimate resis-
tance according to method A and B the
The py curve formulations according to point of zero deection is disregarded. For
Method A and Method B, cf. (12) and non-slender piles a failure mode as shown
(29), are both dependent on the ultimate in g. 17 could form. This failure mode
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 15
0 0
API Broms
method method
1 1
Depth [m]
Depth [m]
2 2
3 3
0.0 1.0 2.0 0.0 1.0 2.0
0
Hansens
method
1
Depth [m]
Legend:
3
0.0 1.0 2.0
pult/pult( fe m) [-]
Figure 18: Comparison of the ultimate resistance estimated by Broms' method, Hansen's method and
API's method with a nite element model, after Fan and Long (2005). pult /pult(f em) denes the ratio of
the ultimate resistance calculated by the analytical methods and the ultimate resistance calculated by
the nite element model.
side friction as illustrated in g. 19 is ne- associated with the tests, have diameters
glected. To take this into account Briaud of 9, 12, and 16 mm and a slenderness ra-
and Smith (1984) has proposed a model tio larger than 20. The magnitude of the
where the ultimate resistance is calculated acceleration in the centrifuge is not given.
as the sum of the net ultimate frontal re- The ultimate resistance is compared for
sistance and the net ultimate side friction. four depths and the error between mea-
In the model both the net ultimate frontal sured and computed values is found to be
resistance and the net ultimate side fric- less than 10 %. The methods proposed by
tion are taken to vary linearly with pile Broms (1964) and Reese et al. (1974) are
diameter. Zhang et al. (2005) refer to also compared with the model tests. Con-
a comparison made by Barton and Finn clusions similar to Fan and Long (2005)
(1983) of the model made by Briaud and are reached.
Smith (1984) and lateral load tests per-
formed in a centrifuge. The circular piles, Side friction is due to the model proposed
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 17
cal load were applied. The conclusion of The SW model parameters are related to a
the analyses were that the lateral capac- three-dimensional passive wedge develop-
ity of piles in sand increases under vertical ing in front of the pile subjected to lateral
load. The increase in lateral capacity de- loading. The wedge has a form similar to
pended on how the vertical load was ap- the wedge associated with method A, as
plied as the highest increase was in the shown in g. 8. However the angles α and
case of VPL. For the dense sand with a β are given by:
lateral deection of 5 % of the side length
α = ϕm (36)
the increase in lateral capacity was, in the
ϕm
case of SAVL, of up to 6.8 %. The same β = 45◦ + (37)
2
situation in the case of VPL resulted in
an increase of up to 39.3 %. Furthermore, where ϕm is the angle of mobilised internal
the analyses showed a large dierence in friction.
the increase of lateral capacity between the The purpose of the method is to relate
two types of sand as the dense sand re- the stresses and strains of the soil in the
sulted in the highest increase. Due to ver- wedge to the subgrade reaction modulus,
tical loads higher vertical soil stresses and Epy . The SW model described by Ashour
thereby higher horizontal stresses occur, et al. (1998) assumes a linear deection
which also mobilise larger friction forces pattern of the pile over the passive wedge
along the length of the pile. Therefore, the depth, h, as shown in g. 20. The dimen-
lateral capacity increases under the inu- sion of the passive wedge depends on two
ence of vertical loading. types of stability, local and global, respec-
Although the analyses made by tively. To obtain local stability the SW
Karthigeyan et al. (2006) indicated model should satisfy equilibrium and com-
a considerable increase in the lateral patibility between pile deection, strains
capacity at a relative high deection, the in the soil and soil resistance. This is ob-
improvement at small displacements are tained by an iterative procedure where an
not as signicant. Therefore it might initial horizontal strain in the wedge is as-
not be of importance for wind turbine sumed.
foundations.
head deection can be calculated from the coecient of horizontal earth pressure,
the one-dimensional Winkler approach. K , needs to be 1, which is not the case for
Global stability is obtained when concor- most sands. Eects of cyclic loading are
dance between the pile head deection cal- not implemented in the SW model which
culated by the Winkler approach and the is a large disadvantage seen in the light
SW-model is achieved. The passive wedge of the strict demands for the foundation
depth is varied until global stability is ob- design.
tained.
Ashour et al. (2002) criticise the py curve
The pile bending stiness inuence the de- method as it is based and veried through
ection calculated by the one-dimensional a small number of tests. However the
Winkler approach and hereby also the SW model, has according to Lesny et al.
wedge depth. Hence, the pile bending sti- (2007) been veried only for conventional
ness inuence the py curves calculated by pile diameters.
the SW-model.
Ashour et al. (2000) investigate by means
The equations associated with the SW mo- of the SW model, the inuence of pile sti-
del are based on the results of isotropic ness on the lateral response for conditions
drained triaxial tests. Hereby an isotropic similar to the Mustang Island tests. A py
soil behaviour is assumed at the site. The curve at a depth of 1.83 m is shown in
SW model takes the real stresses into ac- g. 21 for dierent values of Ep Ip . The
count by dealing with a stress level, de- py curve proposed by Reese et al. (1974)
ned as: is also presented in the gure. It is seen
∆σh that there is a good concordance between
SL = (38)
∆σhf the experimental test and the SW model
for similar pile properties. Furthermore,
where ∆σh and ∆σhf are the mobilised
the soil resistance increases with increas-
horizontal stress change and the horizontal
ing values of Ep Ip .
stress change at failure, respectively. The
spread of the wedge is dened by the mo- 500
Reese et al . (1974a)
Hence the dimensions of the wedge de- 400
300
EpIp
Although the SW model is based on the
three-dimensional soil-pile interaction and 200
0.1 EpIp
it is dependent on both soil and pile prop-
erties there still are some points of crit- 100
0.01 Ep Ip
icism or doubt about the model. The
model does not take the active soil pres- 0
0 40 80 120
nite element model. The results are the determination of k , which might seem
shown in g. 22. The investigation showed surprisingly. Dierent studies on the con-
that the pile stiness has neither signi- sequences of neglecting the pile parame-
cant inuence on the ultimate bearing ca- ters have been conducted over time with
pacity nor the initial stiness of the soil- contradictory conclusions. Ashford and
pile system. The eect of pile stiness Juirnarongrit (2003) point out the follow-
shown in g. 21 and 22 has not been ver- ing three conclusions.
ied trough experimental work.
The rst signicant investigation was the
600 analysis of stress bulbs conducted by
Depth = 1 m
Terzaghi (1955). Terzaghi concluded that
by increasing the pile diameter the stress
p [kN/m]
400
Depth = 2 m
portional to the inverse of the diameter
giving that the modulus of subgrade reac-
p [ kN/m]
400
tion, Epy is independent on the diameter,
cf. denitions in tab. 1.
200
400
for circular and squared piles.
200
Thirdly, Pender (1993) refers to two re-
ports conducted by Carter (1984) and Ling
0.00 0.02 0. 04 0.06
(1988). Using a simple hyperbolic soil mo-
y [m ]
del they concluded that Epy is linear pro-
1.689x10
6
kNm
2
portional to the pile diameter.
5 2
1.689x10 kNm
1.689x10
4
kNm
2
The conclusions made by Terzaghi (1955),
Rigid pile Vesic (1961), and Pender (1993) con-
cerns the subgrade reaction modulus, Epy .
Figure 22: Eect of pile bending stiness, after
Their conclusions might also be applicable
Fan and Long (2005).
for the initial modulus of subgrade reac-
∗ .
tion, k , and the initial stiness, Epy
4.5 Diameter eect on initial Based on the investigations presented by
stiness of py curves Terzaghi (1955), Vesic (1961), and Pen-
der (1993), it must be concluded that no
The initial modulus of subgrade reac- clear correlation between the initial mod-
tion, k , is according to API (1993), DNV ulus of subgrade reaction and the pile dia-
(1992), and Reese et al. (1974) only de- meter has been realised. Ashford and
pendent on the relative density of the soil Juirnarongrit (2005) contributed to the
as shown in g. 12. Hence, the methods discussion with their extensive study of
A and B do not include Ep Ip and D in the problem which was divided into three
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 21
Depth = 1 m
proach gave more similar deections to the
numerical modelling than by using the tra-
30
ditional linear approach in the py curve
D)
10
mination of pile length. The py curves
still underestimates the pile head deec-
0. 00 0.02 0.04 0. 06
y/D
tions even though the parabolic approach
50
Depth = 2 m is used.
a
æ x ö
D)
30
E pyparabolic = E pyref × çç
* *
÷÷
è xref ø
v
p/(
10 x ref
0. 00 0. 02
y/D
0. 04 0.06
E pylinear = k × x
*
Depth
50
Depth = 3 m
D)
30
v
p/(
10
D = 1.2 m D = 0. 61 m
E pyref
*
Initial stiffness E py
*
D = 0.3 m
Table 4: Chronological list of investigations concerning the diameter eect on the initial stiness of the py curve
formulations.
Author Method Conclusion
Terzaghi (1955) Analytical Independent
Vesic (1961) Analytical Independent
Carter (1984) Analytical expression calibrated
against full-scale tests Linearly dependent
Ling (1988) Validation of the method proposed
by Carter (1984) Linearly dependent
Ashford and Juirnarongrit (2005) Numerical and large-scale tests Insignicant inuence
Fan and Long (2005) Numerical Insignicant inuence
Lesny and Wiemann (2006) Numerical Initial stiness is
non-linear for long
and large diameter piles
curves for a 0.65 m diameter pile in order 4.6 Choice of horizontal earth
to predict the response of a 0.15 m pile. pressure coecient
The calculations showed a good approx-
imation of the moment distribution, but
the deections however were considerably When calculating the ultimate resistance
underestimated compared to the measured by method A the coecient of horizon-
values associated with the 0.15 m test pile. tal earth pressure at rest, K0 , equals 0.4
even though it is well-known that the rela-
tive density/the internal friction angle in-
uence the value of K0 . In addition, pile
Based on published lateral pile load tests driving may increase the coecient of hor-
Stevens and Audibert (1979) found that izontal earth pressure K .
deections computed by the method pro-
posed by Matlock (1970) and API (1987) The inuence of the coecient of horizon-
were overestimated. The overestimation tal earth pressure, K , are evaluated by
increases with increasing diameter leading Fan and Long (2005) for three values of
to the conclusion that an increase in mod- K and an increase in ultimate resistance
ulus of subgrade reaction, Epy , occurs with were found for increasing values of K . The
an increase in diameter. increase in ultimate resistance is due to
the fact, that the ultimate resistance is
primarily provided by shear resistance in
the sand, which depends on the horizontal
By testing laterally loaded piles with di- stress.
ameters of 0.27 m, 1.22 m, and 1.83 m in
an overconsolidated clay, O'Neill and Dun- Reese et al. (1974) and thereby API
navant (1984) and Dunnavant and O'Neill (1993) and DNV (1992) consider the initial
(1985) found that there were a non-linear modulus of subgrade reaction k to be in-
relation between deection and diameter. dependent of K . Fan and Long (2005) in-
They found that the deection at 50 % vestigated this assumption. An increase in
of the ultimate soil resistance generally K results in an increase in conning pres-
decreased with an increase in diameter. sure implying a higher stiness. Hence, k
Hence, Epy increases with increasing pile is highly aected by a change in K and k
diameter. increases with increasing values of K .
24 Review of py relationships in cohesionless soil
4.7 Shearing force at the pile toe The py curve method was originally de-
veloped to be used in the oshore oil
Recently installed monopiles have diame- and gas sector and has been veried with
ters around 4 to 6 m and a pile slender- pile diameters up to approximately 2 m.
ness ratio around 5. Therefore, the bend- Nowadays monopiles with diameters of 4
ing stiness, Ep Ip , is quite large compared to 6 m and a slenderness ratio around 5
to the pile length. The bending deforma- are not unusual.
tions of the pile will therefore be small and
In the present review a number of the as-
the pile will almost move as a rigid object
sumptions and not claried parameters as-
as shown in g. 25.
sociated with the py curve method have
been described. The analyses considered
in the review state various conclusions,
some rather contradictory. However most
of the analyses are based on numerical
models and concentrates on piles with a
slenderness ratio much higher than 5. In
order to calibrate the py method to nowa-
days monopiles further numerical and ex-
perimental work is needed. Important
ndings of this paper are summarised as
follows:
Figure 25: Deection curve for non-slender pile. When employing the Winkler ap-
proach the soil response is assumed
As shown in g. 25 there is a deection at
independent of the deections above
the pile toe. This deection causes shear-
and below any given point. The eect
ing stresses at the pile toe to occur, which
of involving the shear stress between
increase the total lateral resistance. Ac-
soil layers seems to be rather small,
cording to Reese and Van Impe (2001) a
and from the analysis it is not clear
number of tests have been made in order
whether the results are dependent on
to determine the shearing force at the pile
pile properties.
toe, but currently no results from these
tests have been published and no meth- The failure modes assumed when
ods for calculating the shearing force as a dealing with the ultimate soil resis-
function of the deection have been pro- tance at shallow depth seems rather
posed. unrealistic. In the employed meth-
ods the surface of the pile is assumed
smooth. Furthermore, the method
5 Conclusion does not take the pile deection into
account, which seems critical for rigid
piles.
Monopiles are the most used foundation
concept for oshore wind energy convert- Soil dilatancy aects the soil re-
ers and they are usually designed by use sponse, but it is neglected in the py
of the py curve method. The py curve curve formulations.
method is a versatile and practical design
method. Furthermore, the method has a Determining the ultimate soil resis-
long history of approximately 50 years of tance by the method proposed by
experience. Hansen's (1961), seems to give more
K. T. Brødbæk, M. Møller and S. P. H. Sørensen 25
reasonable results than the method A pile which behaves rigidly will have
associated with the design codes. a deection at the pile toe. Tak-
Moreover, side friction is neglected in ing this deection into consideration
the design codes. might give an increase in net soil re-
sistance.
In current practice piles are anal-
ysed separately for vertical and hor-
izontal behaviour. Taking into ac- 6 References
count the eect of vertical load seems
to increase the lateral soil resistance.
However the eect is minor at small Abdel-Rahman K., and Achmus M., 2005.
lateral deections. Finite element modelling of horizontally
loaded Monopile Foundations for Oshore
Analyses of py curves sensitivity to Wind Energy Converters in Germany. In-
pile bending stiness, Ep Ip , gives ternational Symposium on Frontiers in
rather contradictory conclusions. Ac- Oshore Geotechnics, Perth, Australia,
cording to the Strain Wedge model Sept. 2005. Balkerna.
the formulations of py curves are API, 1993. Recommended practice
highly aected by pile bending sti- for planning, designing, and construct-
ness. This is in contradiction to the ing xed oshore platforms - Working
existing py curves and a numerical stress design, API RP2A-WSD, American
analysis performed by Fan and Long Petrolium Institute, Washington, D.C.,
(2005). 21. edition.
Ashour M., Norris G., and Pilling P., Au- Journal of the Southeast Asian Geotechni-
gust 2002. Strain wedge model capa- cal Society, 34(2), pp. 123-133.
bility of analyzing behavior of laterally
loaded isolated piles, drilled shafts and pile Cox W. R., Reese L. C., and Grubbs B. R.,
groups. Journal of Bridge Engineering, 1974. Field Testing of Laterally Loaded
7(4), pp. 245-254. Piles in Sand. Proceedings of the Sixth
Annual Oshore Technology Conference,
Banerjee P. K., and Davis T. G., 1978. Houston, Texas, paper no. OTC 2079.
The behavior of axially and laterally
Desai C. S., Sharma K. G., Wathugala G.
loaded single piles embedded in non-
W., and Rigby D. B., 1991. Implementa-
homogeneous soils. Geotechnique, 28(3),
tion of hierarchical single surface δ0 and δ1
pp. 309-326.
models in nite element procedure. Inter-
national Journal for Numerical & Analyt-
Barton Y. O., and Finn W. D. L., 1983.
ical Methods in Geomechanics, 15(9), pp.
Lateral pile response and py curves from
649-680.
centrifuge tests. Proceedings of the 15th
Annual Oshore Technology Conference , DNV, 1992. Foundations - Classication
Houston, Texas, paper no. OTC 4502, pp. Notes No 30.4, Det Norske Veritas, Det
503-508. Norske Veritas Classication A/S.
Belkhir S., Mezazigh S., and Levacher, D., Dobry R., Vincente E., O'Rourke M., and
December 1999. Non-Linear Behavior of Roesset J., 1982. Stiness and Damping of
Lateral-Loaded Pile Taking into Account Single Piles. Journal of geotechnical engi-
the Shear Stress at the Sand. Geotechni- neering, 108(3), pp. 439-458.
cal Testing Journal, GTJODJ, 22(4), pp.
308-316. Dunnavant T. W., and O'Neill M. W.,
1985. Performance analysis and inter-
Briaud J. L., Smith T. D., and Meyer, B. pretation of a lateral load test of a 72-
J., 1984. Using pressuremeter curve to inch-diameter bored pile in overconsoli-
design laterally loaded piles. In Proceed- dated clay. Report UHCE 85-4, Dept. of
ings of the 15th Annual Oshore Technol- Civil Engineering., University of Houston,
ogy Conference, Houston, Texas, paper no. Texas, p. 57.
OTC 4501.
Fan C. C., and Long J. H., 2005. As-
Broms B. B., 1964. Lateral resistance of sessment of existing methods for predict-
piles in cohesionless soils. Journal of Soil ing soil response of laterally loaded piles
Mechanics and Foundation Div., 90(3), in sand. Computers and Geotechnics 32,
pp. 123-156. pp. 274-289.
Ling L. F., 1988. Back Analysis of Lat- Pender M. J., 1993. Aseismic Pile Foun-
eral Load Test on Piles. Rep. No. 460, dation Design Analysis. Bull. NZ Nat.
Civil Engineering Dept., Univ. of Auck- Soc. Earthquake Engineering., 26(1), pp.
land, New Zealand. 49-160.
Matlock H., and Reese L. C., 1960. Gener- Poulos H. G., 1971. Behavior of laterally
alized Solutions for Laterally Loaded Piles. loaded piles: I - Single piles. Journal of the
Journal of the Soil Mechanics and Foun- Soil Mechanics and Foundations Division,
dations Division, 86(5), pp. 63-91. 97(5), pp. 711-731.
Matlock H., 1970. Correlations for De- Poulos H. G., and Davis E. H., 1980. Pile
sign of Laterally Loaded Piles in Soft foundation analysis and design, John Wi-
Clay. Proceedings, Second Annual O- ley & Sons.
shore Technology Conference, Houston,
Texas, paper no. OTC 1204, pp. 577-594. Poulus H., and Hull T., 1989. The Role
of Analytical Geomechanics in Foundation
McClelland B., and Focht J. A. Jr., 1958. Engineering. in Foundation Eng.: Current
Soil Modulus for Laterally Loaded Piles. principles and Practices, 2, pp. 1578-1606.
Transactions. ASCE 123: pp. 1049-1086.
Reese L. C., and Matlock H., 1956.
Murchison J. M., and O'Neill M. W, 1984. Non-dimensional Solutions for Laterally
Evaluation of py relationships in cohe- Loaded Piles with Soil Modulus Assumed
sionless soils. Analysis and Design of Pile Proportional to Depth. Proceedings of the
28 Review of py relationships in cohesionless soil