SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES

Felipe A. Villalobos Oxford University felipe.villalobos@eng.ox.ac.uk Guy T. Houlsby Oxford University guy.houlsby@eng.ox.ac.uk Byron W. Byrne Oxford University byron.byrne@eng.ox.ac.uk

SUMMARY
Suction caisson foundations are studied for offshore wind turbine applications. The research programme includes laboratory testing, larger scale field testing and theoretical modelling. This paper concentrates on the experimental results obtained in combined loading tests on monopod caissons. Results obtained from monotonic and cyclic tests on caissons installed either by pushing or by suction are presented and interpreted.

1. INTRODUCTION
The need for increased production of clean and sustainable energy in the near future has resulted in a search for alternatives to fossil fuels as sources of energy, such as nuclear power or a variety of renewable sources such as hydroelectric, solar or wind power. Wind energy is one of the most promising options for electricity generation, with optimistic forecasts for the near future (Byrne and Houlsby, 2003). This energy source is at present the fastest growing energy technology in the world. The wind speed is typically higher and steadier offshore than onshore, so offshore wind turbines can produce more power. However, capital costs (including installation and cable costs) are around 30-50% higher than onshore. The decision to go offshore can be justified by the extra revenue of 20% to 40% when compared to onshore wind turbines (Milborrow, 2003). Offshore electricity costs are dropping and, depending on the technological developments, could reduce to a third of present levels. As a result, offshore wind power is becoming more competitive when compared to other power sources. The UK government is implementing a renewable energy policy to reduce CO2 emissions. Currently, offshore wind farms are being built along the UK coasts, with the target to supply 10% of UK electrical energy requirements by 2010. At the time of writing, 32 offshore wind turbines were in operation and a further 60 in construction. It has been estimated that about 3000 turbines might be necessary to achieve the 10% target. In an offshore wind farm project, the cost of the foundations has been estimated as about 35% of the total installation cost (Byrne and Houlsby, 2003). Two types of foundations have so far been used for offshore wind turbines: gravity bases and piled foundations. The first option is

not suitable for large towers, since the size and weight required becomes excessive. The installation cost of piles is high. In addition, the time needed to drive a pile could be a considerable restriction, especially when weather conditions become harsh. Figure 1a depicts the order of magnitude of the size of an offshore wind turbine, in shallow water with a depth between 5m and 20m. For this sort of installation suction caissons might be a feasible solution to the foundation problem. This type of foundation has been used in the oil and gas industry in the construction of platforms and other offshore facilities (Sparrevik, 2002). However, the loading from a wind turbine structure differs from that for oil and gas structures for the wind turbine the moment loads are much larger in comparison with the vertical loads than for typical oil and gas applications (Byrne and Houlsby, 2003). Furthermore the total weight of the turbine structure is much lower, and many installations are required within a wind farm. The arrangement options for the wind turbine foundations could be a monopod, tripod or quadruped (see Figure 1b). For a tripod or quadruped the structural design approach must take into account the fact that the most unfavourable conditions involve the possibility of transient tensile loads in the upwind leg (for a discussion of this problem see Kelly et al., 2004). For monopods the most unfavourable loading condition results primarily in a large overturning moment. In both cases the design problem is not so much the ultimate load capacity of the foundation (which may be very large), but the accumulated deformations that might occur under cyclic loading. The research involved in this project includes laboratory testing, large scale field trials and numerical modelling. This paper will describe briefly these three components, and will concentrate on the experimental results obtained in combined loading tests of monopod caissons in sand.

100m

6MN 4MN
h = 30m

90m

(a)

(b)

Figure 1 - (a) Dimensions and magnitude of loads for a 3.5MW turbine structure founded on a monopod suction caisson; (c) Different configurations for offshore wind turbines foundations: multiple caissons and monopod caisson (adapted from Byrne and Houlsby, 2003)

2. THE RESEARCH PROGRAMME

2.1 Laboratory testing The laboratory tests were designed to provide the necessary data to develop the theoretical models for offshore foundations. The experimental results are interpreted within the framework of force resultant models. In this approach, a complex soil structure interaction problem is reduced to the analysis of resultant loads applied at a chosen reference point, at which the transfer of loads from the superstructure to the foundation is considered as occurring. The foundation behaviour can then be incorporated with the response of the superstructure in a numerical analysis. The force resultant models are expressed using plasticity theory, and the main aim of the tests was to define yield conditions and the evolution of plastic displacements. A three degree-of-freedom (3DOF) loading rig, designed by Martin (1994), was used to carry out the tests, (Figure 2). This rig can simultaneously apply vertical, rotational and horizontal displacements (w, 2R, u) to a footing by means of computer controlled stepper motors (Byrne, 2000). Therefore, loads typical of the offshore environment, consisting of gravity, wind, waves and currents can be reproduced with the rig by applying vertical, moment, and horizontal loads (V, M/2R, H).
293mm 293mm 202mm 202mm

(a)

(b)

(c)

Figure 2 (a) The three degree of freedom loading rig; (b) two of the caissons tested; (c) reference point and loads and displacements during loading

2.2 Large scale field trials Despite the versatility and lower cost laboratory tests as compared to field tests, there are scaling issues that need to be addressed in the study of foundations, such as the much higher level of stresses encountered in the field. It is necessary to know how the results obtained in the laboratory will scale for application to real foundations. For that reason, large field trials have been conducted using two caissons (see Table 1, last two columns) installed by suction in clay and sand soils (Kelly, 2002). For the clay tests a reaction frame was set up in an excavated rectangular pit 20m by 10m, 2m deep at the Bothkennar test site. The loads were applied to the caissons using hydraulic jacks for compression-tension and moment tests, see Figure 3a and b. A structural eccentric mass vibrator (SEMV) was used to apply large numbers of cyclic moment loads of very small amplitude, see Figure 3b. The field test results will be used to validate the numerical model, which is initially calibrated against the laboratory results. The field tests are not further discussed in this paper.

(a)

(b)

Figure 3 - Field trial frame set up in clay showing: (a) the hydraulic jacks over the 1.5m diameter caisson; (b) the hydraulic and SEMV used to test the caisson of 3m diameter

2.3 Theoretical modelling Force resultant models using work hardening plasticity theory have proved to be well suited to the analysis of the monotonic behaviour of spudcan and flat circular footings under combined loads (Martin, 1994; Houlsby and Cassidy, 2002). However, the response under cyclic loading is not so well modelled by this approach. Houlsby and Puzrin (2000) suggest that models using multiple yield surfaces may be suitable for modelling cyclic loading, and that these can be derived within a relatively compact mathematical framework by adopting the hyperplastic formulation which is based on thermodynamics. In the conventional plasticity it is necessary to define four components: the shape of the yield surface, a hardening law, flow rule and elastic behaviour inside the yield surface. Whilst hyperplasticity theory requires the definition of just two scalar functions, these in turn can be established from knowledge of the behaviour in conventional plasticity terms.

3.

MOMENT TESTS AND THEIR INTERPRETATION

3.1 Monotonic loading Moment loading tests were carried out to investigate the response of a monopod caisson under low vertical loads. Two aspect ratios of caisson were tested, L/2R = 0.5 and 1.0, as shown in Figure 2b. A range of aspect ratios is relevant as at this stage it is not yet clear will lead to an optimal design, although lower aspect ratios are more likely to be appropriate in sand and higher aspect ratios used in clay. The soil used in the experiments on dry sand was loose white Leighton Buzzard sand (average relative density, Rd = 30%). Experiments on saturated sand used Baskarp Cyclone sand saturated with 100 centistoke silicon oil. The details of the caissons tested are given in Table 1, and the soil properties in Table 2.

& = 0.5mm/s, until the Tests started by pushing the caisson into the ground, at a rate of w
Table 1 Geometry of the model caissons Diameter, 2R (mm) Length of skirt, L (mm) Thickness of the skirt wall, t (mm) Aspect ratio, L/2R Thickness ratio, 2R/t 293 150 3.4 0.5 86 Laboratory 202 200 3.4 1 59 Field 200 100 1.0 0.5 200 3000 1250 10 0.41 300 1500 1000 10 0.67 150

Table 2 White Leighton Buzzard sand and Baskarp Cyclone sand properties (after Schnaid, 1990 and Byrne, 2000) D10, D30, D50, D60: mm Coefficients of uniformity, Cu and curvature Cc Specific gravity, Gs 3 Minimum dry density, min: kN/m 3 Maximum dry density, max: kN/m Critical state friction angle, cs Leighton Buzzard 0.63, 0.70, 0.80, 0.85 1.36, 0.92 2.65 14.65 17.58 34.3 Baskarp Cyclone 0.178, 0.377, 0.577, 0.688 3.87, 1.16 2.69 12.72 16.85 32.5

underside of its lid made complete contact with the soil plug. At that point the maximum vertical load obtained, V0 will determine the size of the yield surface. Next the vertical load was & = 0.01mm/s. Once the target value was reached, it reduced to a chosen value at a rate of w & = 0.01mm/s with a constant was kept constant whilst the caisson was rotated at a rate of 2R ratio between the moment and horizontal load, M/2RH. Tests were conducted for a range of vertical loads from V = -50N (tension) to V = 100N, and at M/2RH values between -2 to 2. The ratio M/2RH can also be interpreted as the ratio between the height h where the horizontal load is applied, to the caisson diameter 2R, i.e. h/2R. The horizontal force is the resultant of the wind, waves and current forces. The low vertical load was held constant to reproduce the self weight of a light structure (wind turbine), whilst rotation is applied to reproduce the environmental loads. Figure 4a shows the load path applied. Yield points were obtained from the curves of: M/2R v. 2R and H v. u, as the intersection of the two straight lines, see Figure 4b. On the other hand, incremental plastic displacement vectors were calculated from the slopes curves of: u v. 2R and w v. 2R. The mathematical formulation adopted for the yield surface is given by an expression that represents an ellipsoid. Such a surface y can be expressed by:

H M H M y = hV + 2Rm V 2a hV f ( V ,Vt ,Vo ) = 0 o o o o o o 2RmoVo

(1)

(a)

(b)

Figure 4 (a) Load paths for monotonic loading tests and yield surface derivation for low vertical loads; (b) Curves of loads (M/2R, H) versus displacement (2R, u) and vertical displacement versus displacement (2R, u)

M/2R (N), 2R

120

V = -50 N (data) V = 0 N (data) V = 50 N (data)

80

40

H (N), u
0 -180 -140 -100 -60 -20 20 60 100 140 180

-40

-80

-120

Figure 5 Yield points fitted with ellipses curves in the M/2R H plane and experimental and normal flow vectors for V = -50N, 0N and 50N. Aspect ratio caisson L/2R = 0.5

Where a is the eccentricity of the yield surface, Vo is the maximum pure vertical load, Vt is the maximum pure pull-out load and ho and mo are the horizontal and moment dimension of the yield surface. The elliptical curves illustrated in Figure 5 were fitted to the experimentally determined yield points, using the least square error method. A separate curve was fitted to each set of tests at a particular vertical load. This fit shows clearly how expression (1) agrees very well with experimental results. The flow rule can be derived from the yield surface equation (1) using the following form:
p w &M y / V & p M = y / M &p y / H uH

(2)

p & , u & Vp , Where ( w M & H ) correspond to the increments of the plastic displacements and is a

positive scalar multiplier that accounts for the magnitude of these velocity vectors. Figure 5 shows the experimental flow vectors that represent the direction of the plastic displacements p &p , u &H ) in the M/2R v. H plane. The direction is similar to the vectors normal to the yield ( 2 R M surface, demonstrating that an associated flow rule is valid in this plane. Associated flow in the M/2R H plane has bee observed previously (Martin, 1994; Gottardi et al., 1999). However, when experimental flow vectors are plotted in the M/2R V plane they do not tend to follow the direction of the vectors normal to the yield surface, as can be observed in Figure 6. Further investigation is required to establish the correct form of non-associated flow rule. 3.2 Cyclic loading The environmental loads are cyclic, therefore an investigation of foundation behaviour under cyclic loading was conducted under similar conditions to those already explained for the monotonic combined loading tests. The same loading rig, caissons and soil were used (see Figure 2 and Tables 1 and 2). Tests were conducted holding a constant vertical load whilst a

100

M/2RH = 1 (data) M/2RH = -1 (data) M/2RH = 0.25 (data)

M/2R ( N), 2R
-20 80

60

40

20

0 -80 -60 -40 0 20 40 60 80 100

V (N), w

Figure 6 Yield points in the M/2R v. V plane and experimental and normal flow vectors for M/2RH = 1, 0.25 and 1. Aspect ratio of caisson L/2R = 0.5

cyclic rotational displacement of increasing amplitude was applied for ten cycles. Tests were performed for a range of vertical loads from V = -50N (tension) to V = 400N, and at M/2RH values between -2 to 2. Figure 7 shows ten rotational cycles applied to a caisson of diameter 2R = 293mm at a rate of & = 0.02mm/s. The response is hysteretic and it is possible to observe stiffness degradation 2R during each cycle. Figure 8 shows proof, however, that the shape of the cycles conform to the second Masing rule, which states that the shape of unloading and reloading curves is the same as that of the initial curve, but doubled in both dimensions. The first Masing rule is also confirmed by Figure 8. This states that the tangent to the slope of the reloading curves is identical to the tangent to the slope of the initial curve. The confirmation that the Masing rules apply offers the possibility of a relatively simple interpretation of the data, since Masing rules correspond to pure kinematic hardening.
60

40

Moment Load, M/2R (N)

20

-20

-40

-60 -2 -1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6 2

Rotational Displacement, 2R (mm)

Figure 7- Typical cyclic rotational test. V = 50N, M/2RH = 1 and L/2R = 0.5

120 100 Moment Load, M/2R (N) 80 60 40 20 0 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6

Rotational Displacem ent, 2R (m m )

Figure 8 - Second Masing rule. The initial loading is doubled, reversals and re-loadings are relocated.

In tests with V 0N the moment resistance approaches an asymptotic value. However, the remainder of the tests show an increase in their moment resistance after each cycle. There moment response increases as V is increased. Furthermore, there was an uplift of the caisson in tests with V < 100N. The caisson rotated almost without vertical displacement at V = 100N. Settlement occurred for very high vertical loads, V 200N. 3.3 Moment capacity tests of a suction installed caisson Model caissons were installed in loose dry sand by pushing. In reality caissons are installed by suction. A study of the effect of suction installation on moment capacity was therefore performed. Two series of combined loading tests, one using each of the installation methods, were carried out on a model scale suction caisson (4th column in Table 1). Both series of tests were in dense oil-saturated Baskarp Cyclone sand. The properties of this soil are in the last column of Table 2. Using the 3DOF rig (Figure 2) the caisson was first penetrated to 20mm using pushing to form a seal with the soil, and then a constant vertical load was held whilst suction was applied to install the caisson into the ground. Once installed, moment loading tests were conducted using the following sequence: (a) The footing was vertically displaced until a preset vertical load was reached. (b) The vertical load was held constant for a period of time, to allow excess pore pressure (measured by a pore pressure transducer under the middle of the caisson lid) caused by the loading in (a) to dissipate. (c) Rotational and horizontal movements were applied so that a load path in (V, M/2R, H) space was followed. A rotational displacement of 2R = 0.5mm was reached under a & = 0.0005mm/s. This corresponded to drained conditions. rate of 2R The relative density was estimated by driving a small cone penetrometer into the sample at the tested site (Mangal, 1999). The values obtained were Rd = 74% and 64% for the samples used for suction and pushing installation respectively. Figure 9 compares the two moment tests. No significant differences in moment capacity were observed between the methods of installation of the caisson.

5 Moment Load, M/2R (N)

3 2 V = 20N (by Suction) V = 20N (by Pushing) 1

0 0 0.125 0.25 0.375 0.5

Rotational Displacem ent, 2R (m m )

Figure 9 Comparison of moment capacity for a caissons installed by different methods, M/2RH = 0.5 and L/2R = 0.5

CONCLUSIONS A description of the research currently in progress to investigate suction caissons as an alternative foundation for offshore wind turbines has been presented. Laboratory testing has been carried out to provide the necessary data to construct and validate theoretical models. Field trial results are being used to assess the scale effect in the models. A hyperplasticity theory has been used to model monotonic and cyclic caissons response using multiple yield surfaces. This paper has focused mainly on laboratory testing, from which a yield surface and flow rule was determined, for two model caissons of different aspect ratios under low vertical load. The following conclusions are drawn: Monotonic and cyclic moment loading tests proved that higher moment resistance was obtained when the vertical load is increased. Furthermore, uplift of the suction caisson was observed when the vertical load was below a certain critical value. In cyclic tests a reduction of stiffness during each cycle was observed. Furthermore, all the tests obeyed the Masing rules. This makes the numerical modelling more straightforward since the entire response can be reproduced using the first loading part of the cyclic curve. Finally, analyses of the effect of the installation method on the moment capacity are in progress. Provisional results indicate that differences in capacity between the two methods are not significant.

REFERENCES
Byrne, B.W. (2000) Investigations of suction caissons in dense sand, DPhil thesis, University of Oxford Byrne, B.W. and Houlsby, G.T. (2003) Foundation for offshore wind turbines, Phil. Trans. of the Royal Society of London, Series A 361, 2909-2300 Gottardi, G., Houlsby, G.T. and Butterfield, R. (1999) The plastic response of circular footings on sand under general planar loading, Gotechnique 49, No. 4, 453-470 Houlsby, G.T. and Puzrin, A.M. (2000) A Thermomechanical Framework for Constitutive Models for Rate-Independent Dissipative Materials", International Journal of Plasticity 16, No. 9, 1017-1047 Houlsby, G.T. and Cassidy, M.J. (2002) A plasticity model for the behaviour of footings on sand under combined loading, Gotechnique 52, No. 2, 117-129 Kelly, R.B. (2002) Proposal for Large Scale Field Trials of Suction Caissons, Report FOT002, Department of Engineering Science, University of Oxford

Kelly, R.B., Byrne, B.W., Houlsby, G.T. and Martin, C.M. (2004) Tensile loading of Model Caisson Foundations for Structures on Sand, Proc. Int. Offshore and Polar Eng. Conference, Toulon Mangal, J.K. (1999) Partially drained loading of shallow foundations on sand, DPhil thesis, University of Oxford Martin, C.M. (1994) Physical and Numerical Modelling of Offshore Foundations under Combined Loads, DPhil thesis, University of Oxford Milborrow, D. (2003) Offshore wind rises to the challenge, Windpower Monthly, 19, No.4, 51-56 Schnaid, F. (1990) A study of the cone pressuremeter test in sand, DPhil thesis, University of Oxford Sparrevik, P. (2002) Suction Pile Technology and Installation in deep waters, Offshore Technology Conference, Houston, paper 14241 Villalobos, F.A. (2004) An experimental study of cyclically loaded monopod suction caisson foundations for offshore wind turbines, BGA Eighth YGE Symposium, Birmingham