Professional Documents
Culture Documents
© 2011 Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin · geotechnik 34 (2011), Heft 2 85
H. Sturm · Geotechnical performance of a novel gravity base type shallow foundation for offshore wind turbines
where Hext is the maximum horizontal load and δ the fric- Embedment depth, scour protection and suction affect al-
tion coefficient between plate and soil. The verification of so the bearing capacity. The resistance factors increase
safety against overturning can be used to estimate the and the actions decrease, viz.
minimum outer foundation diameter d (see Fig. 1) by as-
2Mext Vstat
suming a rotation around the geometrical centre of the R α = 0 ≥ + – Vsuc,α = 0 , for α = 0
plate(s) on compression side, viz. d–a 4
(2c)
2 2 Mext Vstat
⎧ 2Mext R α = 45 ≥ + – Vsuc,α = 45 , for α = 45
d–a 2
⎪ + a, for α = 0
⎪ Vstat
d≥⎨ (1b) The suction forces Vhsuc and Vvsuc, however, can hardly be
⎪ 2 2 Mext + a, for α = 45
⎪⎩ Vstat determined analytically. Also it has to be clarified if one
can rely in the design on the ability to mobilise these
forces during the whole life time of the foundation. This
where a denotes the edge length of a plate and Mext the can only be proved by means of Finite Element analyses,
maximum moment acting on the foundation measured at employing effective stress based constitutive models in ful-
seabed level. Both parameters, Vstat and d, serve as input ly coupled stress equilibrium and pore water flow simula-
for the verification of safety against bearing failure, viz. tions.
In order to assess analytically the behaviour of the
2Mext Vstat
R α = 0 ≥ + , for α = 0 foundation during cyclic loading, the procedures pro-
d–a 4 posed by Andersen [8] can be used if the soil during one
(1c)
2 2 Mext Vstat load cycle is almost undrained.
R α = 45 ≥ + , for α = 45
d–a 2
4 Numerical Model
where Ra = 0° and Ra = 45° are resistance factors determined 4.1 Finite Element models
according to Brinch Hansen [5] [6] or Meyerhof [7].
The resistance factors depend on the loading direction Three different Finite Element models were used in the
with: numerical study; two 3D models accounting for the differ-
ent loading directions α = 0° and a = 45° under utilisation
Hext of the corresponding symmetry conditions (Figs. 2a and
R α = 0 = f(H = , A = a 2, …)
4 2b), and a simplified 2D model based on a loading paral-
H lel to an arm, i.e. α = 0°. A detailed view of the 2D model is
R α = 45 = f(H = ext , A = 2a 2, …)
2 shown in Fig. 2c.
The 3D models were only used in static and cyclic
This analytical approach is an iterative procedure, which analyses under fully drained conditions2. They were used
may require several repetitions in order to find an opti- to study the influence of loading direction and to verify
mised geometry and weight of the foundation for a the 2D model; the verification is not presented here. The
specific site. In general, the foundation diameter should 2D model was also used in static and cyclic analyses un-
be minimised to reduce material and transportation der both fully drained and coupled stress equilibrium and
costs. pore water flow simulations. This model was employed in
Equations 1a to 1c become more complex if in addi- most presented simulations. Since only two of the four
tion the effect of skirts, scour protection and suction in- plates are considered in the 2D model, the weight and
side the skirt compartments are taken into account. The horizontal loads are halved so that the stresses in the soil-
sliding resistance increases due to passive earth pressure structure contact interface are the same in 2D and 3D
Ep and shear induced suction Vhsuc. Also the friction angle model. The moment loading remained unchanged, since
δ can be replaced with the (peak) friction angle ϕ of the the section modulus Iy,3D of the 3D models, where y de-
corresponding soil layer(s), since length and configuration notes the direction perpendicular to the considered sym-
of the skirts are generally designed to enforce a shear metry axis, viz.
plane reaching from tip to tip, preventing in particular
curved failure planes reaching up to the base plate or into 1 4 1
Iαy,=30d = Iαy,=345
d = I y, 3d = a + (d – a)2 a 2 (3)
weaker soil layers. Hence, Equation 1a becomes 3 2
⎧ 2Mext
⎪ + a, for α = 0
V
⎪ stat + 2 V v
suc, α = 0
2 Although a few coupled 3D analyses have been performed,
d≥⎨ (2b) they are, however, not considered in this study.
⎪ 2 2 Mext
+ a, for α = 45 3 The difference amounts ≈ 1.7 % for the dimensions shown
⎪V + 2 V v
⎩ stat suc, α = 45 in Fig. 1.
Table 1. Hypoplastic and intergranular strain parameters of the two different sands used in the presented simulations.
Tabelle 1. Stoffparameter der verwendeten Sande
In all FE models contact surfaces between plates, respec- general, the parameters of the Baskarp sand were used in
tively, skirts and soil were considered, allowing for open- this study, except for the simulations presented in the sec-
ing, closing and tangential sliding. A friction coefficient tion Failure and Stabilisation Behaviour in which the pa-
equal to the critical state soil friction angle was assumed, rameters of a Model sand were employed. Both sands con-
i.e. δ = ϕc, accounting for soil disturbance due to installa- sist mainly of quartz minerals. They have an uniformity
tion. The modelled soil block has a horizontal length of coefficient of U = 1.9 and an average grain size diameter of
150 m by 75 m and a depth of 90 m. This equates to 21 × d50 = 0.14 mm. The Baskarp sand is sub-angular, the Mod-
10.5 × 12 times the edge length a of the plates. All simula- el sand almost perfectly rounded. Both represent typical
tions were performed with the commercial Finite Element offshore sands as can be found in the North and Baltic
program Abaqus/Standard, using C3D8 (CPE4) elements Sea.
in drained simulations and C3D8P (CPE4P) in coupled Since typical sand deposits at the prospective off-
simulations for the 3D (2D) models. In general, the default shore wind parks are expected to be rather dense, an ini-
algorithms and settings of allowable tolerances were em- tial relative density of Dr = 80 % was assumed in all simu-
ployed. The only exception was the use of large deforma- lations.
tion theory, which means an updated Lagrangian formula- The foundation was modelled linear elastically. Box
tion. Numerical convergence studies with respect to ele- girder and skirts were divided into different segments with
ment type and size, increment length and other modelling equal cross sectional area, geometry and material parame-
parameters were performed in order to prove the suitabili- ters in order to represent a realistic bending behaviour.
ty of the models and the independence of the results on The cyclic loading was computed implicitly, i.e. the com-
the discretisation and modelling. As a matter of course, the plete loop of a cycle was followed numerically. This ap-
results of the 2D and corresponding 3D model are not proach requires a significant number of load increments
identical. But the differences are small and in general less and equilibrium iterations. In order to minimise the effect
than 10 % at maximum loading [9]. of the inevitably accumulated rounding error, only the
first 20 to 25 cycles were evaluated in this study.
4.2 Constitutive model and material parameters
5 Loading conditions
The mechanical behaviour of the soil was modelled with a
modified version of the hypoplastic constitutive model Only quasi-static loading was considered in this study, be-
proposed by von Wolffersdorff [10] incorporating the in- cause typical loads on OWTs have frequencies of less than
tergranular strain extension for improving the small strain 2 Hz. Thus any inertia effects of the soil can be neglected.
and cyclic behaviour, introduced by Niemunis and Herle The characteristic, i.e. unfactored, extreme load
[11], and an improvement of the stiffness response at very combination at the considered site amounts to Hext =
high densities proposed by Niemunis [12]. The parame- 17.83 MN acting on the structure at a height of hext = 20 m
ters of two different soils used in this paper are listed in above the seabed. The required characteristic weight of
Table 1. A procedure proposed to determine the hypoplas- the reference foundation without skirts was determined to
tic parameters is presented in Herle and Gudehus [13]. In be 41 MN, based on Equations 1a to 1c and the dimen-
H/ H ext (%)
35 m
50
Hext 20 m
α = 0°
0 α = 45°
0m 0 10 20 30
horizontal displacement (cm)
150
sions shown in Fig. 1. A half load period of the extreme
load combination of 20 s was assumed.
H/ H ext (%)
The design load case (DLC) found to be relevant for 100
the assessment of the serviceability of the foundation con-
sists of an asymmetric loading with an average value of
50
cyc = 0.09 · Hext and a cyclic amplitude of Hcyc = 0.14 ·
Hav ampl
isotropic permeability of k = 5.5 · 10–5m/s was assumed for (a) Horizontal displacement.
the soil. Hydrostatic pore pressure at rest was pre-defined
100 drained
at the outer boundaries and the free surface of the soil
block. coupled
H/ H ext (%)
50
To compare the influence of the drainage conditions on
the permanent displacement and rotation of the founda-
tion after unloading, a dissipation of the excess pore pres-
sure until a steady state is reached was allowed in the cou-
pled simulations. This was achieved after ≈ 500 s. 0
Fig. 5a shows the horizontal displacement, Fig. 5c
the vertical displacement and Fig. 5b the rotation of the 0 1 2 3 4
reference model. The soil response is softer under partial- horizontal displacement (cm)
ly drained conditions during the loading phase, because of
decreasing effective stresses in the soil on the compression (b) Rotation.
·10−2
side due to pore pressure generation.
In the subsequent consolidation phase, however the drained
8
larger horizontal displacement and rotation under par- coupled
tially drained conditions are compensated again. The
foundation slides and rotates back, approaching the same 6
state as the foundation under fully drained conditions. rotationϕ ( ° )
The curve shown in Fig. 5b can be interpreted as a creep- 4
type mechanism. For energetic reasons, the foundation
has to settle during this creep phase, as shown in Fig. 5c,
because the height of the centre of gravity cannot 2
increase again. That means that heave of the foundation
is physically impossible for the considered loading [14] 0
[15].
100 101 102
time t (sec)
6.2.2 Cyclic loading
Fig. 6 shows the average excess pore pressure develop- (c) Vertical displacement.
ment in the soil nodes adjacent to the soil-structure inter- 100 drained
face during cyclic loading for the first 20 cycles. Compres-
coupled
sion side denotes here the plate with increased stresses
due to the average (static) load Hav cyc.
Both amplitude and average value of the excess pore
H/ H ext (%)
(a) Rotation.
100 ·10−2
compression
uplift drained
excess pore pressure u (kPa)
coupled
50 2
rotation ϕ ( ° )
1
0
0
0 5 10 15 20
number of cycles 0 5 10 15 20
Fig. 6. Excess pore pressure development under the plates number of cycles
on compression and uplift side of the reference foundation
without skirts predicted with the 2D model (b) Settlement.
Bild 6. Porenwasserüberdruck unter den Platten auf der 1
Kompressions- und Entlastungsseite bei zyklischer Belas-
tung
·10−2 · 10−2
2 2
at ext. load
0 0
−2 −2 at rest
at rest
−4 −4
at ext. load
−6 soft −6 soft
stiff stiff
−8 −8
outside middle inside inside middle outside
plate on compression side plate on uplift side
Fig. 8. Footprints of the plates at compression (left) and uplift side (right) at rest and at extreme load for different structural
stinesses of the reference foundation without skirts predicted with the 2D model. Stiff here denotes a 10 times stiffer, soft a
10 times softer model than the reference model. Loading is applied from the right side
Bild 8. Setzungsmulden der Platten auf Kompressions- und Entlastungsseite im Ruhezustand und bei maximaler Belastung
in Abhängigkeit der Struktursteifigkeit
hi = skirt length
in the design should be that the weight G′ of the bal-
lasted foundation is sufficient for complete skirt penetra- di = inside diameter
tion. z
0 0
vertical pressure σ′ (kPa)
−100
−1,000 7.0 m 7.0 m
3.5 m 3.5 m
1.5 m 1.5 m
−1,500
−200
0.0 m 0.0 m
Fig. 10. Vertical effective stress distribution at the interface between plate and soil on compression and uplift side at peak of
the extreme loading predicted with the 2S model; loading is applied from the right side; after [20]
Bild 10. Vertikale effektive Spannungen in den Boden-Bauwerkkontaktzonen auf Kompressions- und Entlastungsseite bei
maximaler Belastung in Abhängigkeit der Schürzenlänge
skirts.
1
8 Failure and stabilisation behaviour
7.0 m Almost all standards and guidelines for the design of OWTs
0 3.5 m demand an analysis of the long-term behaviour of the
foundation. This is not possible with currently available
0 5 10 15 20 soil models. But recent results of 1g model tests
number of cycles and numerical simulations indicate that only a short
Fig. 11. Rotation of the foundation for different skirt lengths time frame with alternating static and cyclic loading is
during cyclic loading predicted with the 2D model sufficient to consider in an adequate serviceability analysis.
Bild 11. Rotation des Fundaments in Abhängigkeit der
Schürzenlänge bei zyklischer Belastung 8.1 ”Stabilisation” effect
0 0
7.0 m 7.0 m
40 −40
60 −60
80 −80
100 −100
outside middle inside inside middle outside
plate on compression side plate on uplift side
Fig. 12. Excess pore pressure and suction under the plate on compression and uplift sides at peak of the extreme loading for
two different skirt lengths predicted with the 2D model; loading applied from the right side; after [20]
Bild 12. Porenwasserüber- bzw. Unterdruck auf Kompressions- und Entlastungsseite bei maximaler Belastung in Abhängig-
keit der Schürzenlänge
accumulated rotations during cyclic loading, leading fi- tion without skirts (0 m). The corresponding settlement,
nally to a failure or collapse of the whole structure when shown in Fig. 14b, is normalised in a similar way.
it tilts over. In contrast to that, stabilisation denotes a de-
crease of the accumulated rotation, i.e. a back-rotation of ⎛ sav – sav ⎞ ⎛ sav ⎞
the structure towards the previous vertical position. Sta- sˆav = 100 · ⎜ 25 av 0 ⎟ · ⎜ av 0 av ⎟ [%] (8)
⎝ s0 ⎠ i ⎝ s25 – s0 ⎠ ref
bilisation in terms of physics also implies a decrease of
the total energy, which is accomplished here by a de-
crease of the potential energy. Thus, back-rotation is nec- In addition to the skirt length, the effect of structural stiff-
essarily accompanied by accumulated settlement of the ness studied shown in Figs. 14a and 14b. The term stiff de-
structure. notes a ten times larger, soft a ten times smaller Young’s
The rate of back-rotation is a function of the number modulus E than the one used in the reference model.
of applied load cycles, the cyclic amplitude and the aver- While the amount of back-rotation is larger for foun-
age cyclic load. In addition, foundation geometry, soil type dations with skirts, the actual length has only a minor in-
and soil conditions affect the stabilisation behaviour [22] fluence. The effect of structural stiffness is qualitatively
[23]. the same but the amount of normalised back-rotation is
larger for the softer and smaller for the stiffer model. This
8.2 Effect of skirt length and structural stiffness is again due to shear-induced dilatancy being smaller for
stiff and skirted foundations.
Fig. 13 shows the rotation of the cross foundation with dif-
ferent skirt lengths subjected first to the same extreme
load analysed in Section 6.2.1, followed by a cyclic load
package having a somewhat smaller amplitude and aver- ·10−2
age value than the load used in Section 6.2.2. Based on 10
e.g. Miner’s rule [26], classical theories assume a continu- 0m
ous increase of the accumulated rotation being in particu- 8 1m
lar independent of the order of the applied loads. Fig. 13, 2m
however, shows a back-rotation of the foundation during
rotation ϕ ( °)
6 4m
cyclic loading; i.e. a stabilisation. Within only 25 cycles,
the foundation with 4 m long skirts rotates back by
4
≈ 0.005° or ≈ 18 %, relative to the state after unloading
from the extreme load.
In order to quantify the back-rotation, a normalisa- 2
tion was introduced in Fig. 14a.
0
⎛ ϕ av – ϕ av ⎞ ⎛ ϕ av ⎞ 0 5 10 15 20 25
ϕˆ av = 100 · ⎜ 25 av 0 ⎟ · ⎜ av 0 av ⎟ [%] (7) number of cycles
⎝ ϕ0 ⎠ i ⎝ ϕ 25 – ϕ 0 ⎠ ref
Fig. 13. Rotation of the foundation with different skirt
lengths under fully drained conditions predicted with the 2D
where ϕ0av denotes the rotation after unloading from the model
extreme loading and ϕav25 the rotation of the foundation af- Bild 13. Rotation des Fundamentes in Abhängigkeit der
ter 25 cycles. The index ref refers to the reference founda- Schürzenlänge
soft
rotation ϕ̄ (%)
100
100
150
el tests in order to support the optimisation potential of [17] Andersen, K., Jostad, H. & Dyvik, R.: Penetration resistance
the proposed foundation solution compared to ordinary of offshore skirted foundations and anchors in dense sand.
GBSs without or very short skirts. Journal of Geotechnical and Geoenvironmental Engineering
134 (2008), No. 1, pp. 106–116.
[18] Houlsby, G. & Byrne, B.: Design procedures for installation
References
of suction caissons in clay Design procedures for installation of
[1] Houlsby, G., Ibsen, L. & Byrne, B.: Suction caissons for wind suction caissons in clay and other materials. Proceedings of the
turbines. Gourvenec, Cassidy (eds.): ISOFOG – International Institution of Civil Engineers – Geotechnical Engineering 158
Symposium on Frontiers in Offshore Geotechnics, 2005, pp. (2005), pp. 75–82.
75–93. [19] Houlsby, G. & Byrne, B.: Design procedures for installation
[2] Houlsby, G., Kelly, R. Huxtable, J. & Byrne, B.: Field trials of of suction caissons in sand, Proceedings of the Institution of
suction caissons in sand for offshore wind turbine foundations. Civil Engineers – Geotechnical Engineering 158 (2005), pp.
Géotechnique 56 (2006), No. 1, pp. 3–10. 135–144.
[3] Ibsen, L. & Brincker, R.: Design of a new foundation for off- [20] Dinier, M.: Entwurf und Bemessung einer Gründung mittels
shore wind turbines. Proceedings of the 22nd International Bucket Foundations für eine Offshore Windenergieanlage.
Modal Analysis Conference (IMAC), Detroit, 2004. Master’s thesis, Institute of Soil and Rock Mechanics in Karls-
[4] Grabe, J., Dührkop, J. & Mahutka, K.P.: Monopilegründun- ruhe, 2008.
gen von Offshore Windenergieanlagen – Zur Bildung von Po- [21] NGI: Skirted offshore foundations and anchors in sand:
renwasserdrücken aus zyklischer Belastung. Bauingenieur Phase 1, Activity 1.7: Back-calculation of model tests with
2004, Heft 9, S. 418-423. BIFURC. Tech. Rep. 524098-8, NGI – Norwegian Geotechni-
[5] Brinch Hansen, J.: A general formula for bearing capacity. cal Institute, 1999.
Geoteknisk Instituts Bulletin 5 (1961), pp. 38–46. [22] Sturm, H.:Numerical investigation of the stabilisation be-
[6] Brinch Hansen, J.: A revised and extended formula for bear- haviour of shallow foundations under alternate loading. Acta
ing capacity. Geoteknisk Instituts Bulletin 28 (1970), pp. 5–11. Geotechnica 4 (2009), No. 4, pp. 283–292.
[7] Meyerhof, G.: Some recent foundation research and its appli- [23] Sturm, H.: Stabilisation behaviour of cyclically loaded shal-
cation to design. The Structural Engineer 31 (1953) No. 6, pp. low foundations for offshore wind turbines. Ph.D. thesis, 2009.
151–167. [24] Sturm, H., Solf, O. & Kudella, P.: Self-healing effects of
[8] Andersen, K.: Bearing capacity under cyclic loading -off- shallow foundations for offshore wind turbine structures. 11th
shore, along the coast, and on land. Géotechnique 46 (2009), Baltic Sea Geotechnical Conference, 2008.
No. 5, pp. 513–535. [25] Solf, O., Kudella, P. & Triantafyllidis, T.: Investigation of
[9] Sturm, H.: Shallow Foundations of Offshore Wind Turbine the self-healing effect of monopile foundations. ICPMG -Inter-
Structures in the North and Baltic Sea. 18th European Young national Conference on Physical Modelling in Geotechnics,
Geotechnical Engineers Conference. Ancona, 2007. ETH Zurich (ETHZ), 2010.
[10] v. Wolffersdorff, P.A.: A hypoplastic relation for granular [26] Miner, M.: Cumulative damage in fatigue. Transactions of
material with predefined limit state surface. Mechanics of Co- the American Society of Mechanical Engineers 103 (67)
hesive-Frictional Materials 1 (1996), No. 3, pp. 251–275. (1945), pp. 159–164.
[11] Niemunis, A. & Herle, I.: Hypoplastic model for cohesion-
less soils with elastic strain range. Mechanics of Cohesive-Fric- Acknowledgement
tional Materials 2 (1997), pp. 279–299.
[12] Niemunis, A.: Extension to the hypoplastic model for soils. The research in this paper was supported, in chronological order,
Ph.D. thesis. Ruhr-Universität Bochum, 2003. by the construction company Ed. Züblin AG, the Federal Minis-
[13] Herle, I. & Gudehus, G.: Determination of parameters of a try for the Environment, Nature Conservation and Nuclear
hypoplastic constitutive model from properties of grain assem- Safety (BMU) under the promotional reference No. 0327618, the
blies. Mechanics of Cohesive-Frictional Materials 4 (1999), pp. Karlsruhe House of Young Scientists (KHYS) and the Norwegian
461–486. Geotechnical Institute (NGI).
[14] Gudehus, G.: Psammodynamics: attractors and energetic.
9th HSTM International Conference on Mechanics, 2010. Author
[15] Gudehus, G.: Wind parks in the German Bight – a chal- Hendrik Sturm, Norwegian Geotechnical Institute (NGI), P.O. Box. 3930,
lenge for geotechnics. geotechnik 34 (2011), No. 1, pp. 3–10. Ullevål Stadion, N-0806 Oslo, Norway, Hendrik.Sturm@ngi.no
[16] Wichtmann, T.: Explicit accumulation model for non-cohe-
sive soils under cyclic loading, Ph.D. thesis. Ruhr-Universität Submitted for review: 29 November 2010
Bochum, 2005. Accepted for publication: 5 January 2011