Field Testing of Suction Caissons at Bothkennar and Luce Bay

Transcription

1 Field Testing of Suction Caissons at Bothkennar and Luce Bay by G.T. Houlsby, R.B. Kelly, J. Huxtable and B.W. Byrne Report No. OUEL 2276/5 University of Oxford Department of Engineering Science Parks Road, Oxford, OX1 3PJ, U.K. Tel /2833 Fax

2 Field Testing of Suction Caissons at Bothkennar and Luce Bay G.T. Houlsby, R.B. Kelly, J. Huxtable and B.W. Byrne This report consists of three papers that have resulted from a joint industry project investigating the application of suction caissons to offshore wind turbines. The first two papers report field scale testing at two different locations in Scotland. The final paper presents some scaling relationships for use in comparing laboratory testing to the field testing. a) Field trials of suction caissons in clay for offshore wind turbine foundations. Houlsby, G.T., Kelly, R.B., Huxtable, J. and Byrne, B.W. Abstract: A programme of testing of caisson foundations in clay at the Bothkennar test site is described. The tests are relevant to the design of foundations for offshore wind turbines, either in the form of monopod or tetrapod foundations. Records are presented for installation of the caissons, cyclic moment loading under both dynamic and quasi-static conditions, cyclic inclined vertical loading and for pullout of the caisson. Variation of stiffness of the foundation is observed, with high initial stiffness followed by hysteretic behaviour at moderate loads and degradation of response at high loads. Some implications for the design of wind turbine foundations are briefly discussed. b) Field trials of suction caissons in sand for offshore wind turbine foundations. Houlsby, G.T., Kelly, R.B., Huxtable, J. and Byrne, B.W. Abstract: A programme of testing on suction caisson foundations in an artificially prepared sand test bed near Luce Bay, in Scotland, is described. The tests are relevant to the design of either monopod or tetrapod foundations for offshore wind turbines. Records are presented for suction installation of the caissons, cyclic moment loading under both quasi-static and dynamic conditions to simulate the behaviour of a monopod foundation, and cyclic vertical loading and pullout of caissons to simulate one footing in a quadruped foundation. Variations of stiffness with loading level of the foundation are observed, with high initial stiffness followed by hysteretic behaviour at moderate loads and degradation of response at high loads. Some implications for the design of wind turbine foundations are briefly discussed. c) A comparison of field and laboratory tests of caisson foundations in sand and clay. Kelly, R.B., Houlsby, G.T. and Byrne, B.W. Abstract: Laboratory tests applying vertical and moment loads to suction caissons founded in sand and clay have been conducted to simulate an equivalent series of field tests. The caissons used in the laboratory were.15m,.2m and.3m in diameter, while those for the field tests were 1.5m and 3.m diameter. The loads applied to the caissons in the laboratory tests were scaled from those in the field tests, and the models were loaded in a near identical manner to the field trials. The test results are presented in non-dimensional form for comparison. The non-dimensional laboratory moment test data were similar to the field data in most cases. The non-dimensional data from vertically loaded caisson tests in the laboratory and in the field show some differences, and possible reasons for these are discussed.

3

4 1 Field trials of suction caissons in clay for offshore wind turbine foundations G.T. Houlsby 1, R.B. Kelly 1, J. Huxtable 2 and B.W. Byrne 1 Keywords: bearing capacity, clay, dynamic test, foundations, stiffness ABSTRACT A programme of testing of caisson foundations in clay at the Bothkennar test site is described. The tests are relevant to the design of foundations for offshore wind turbines, either in the form of monopod or tetrapod foundations. Records are presented for installation of the caissons, cyclic moment loading under both dynamic and quasi-static conditions, cyclic inclined vertical loading and for pullout of the caisson. Variation of stiffness of the foundation is observed, with high initial stiffness followed by hysteretic behaviour at moderate loads and degradation of response at high loads. Some implications for the design of wind turbine foundations are briefly discussed. INTRODUCTION The offshore wind energy industry is a very rapidly expanding sector of vital economic importance in the UK, and foundation costs are an important part of the costs of offshore wind turbine installations (Byrne and Houlsby, 23). Most current foundations for offshore wind turbines are large monopiles, although some have been founded on gravity bases. However, with the current expansion of the offshore wind energy industry, alternative foundation types are being considered. One possibility is the use of suction caisson foundations (Houlsby and Byrne (2), Byrne and Houlsby (23)). Suction caissons are now widely used as anchors for floating structures, and have also been used offshore as foundations for a small number of fixed platforms (Bye et al., 1995). They are large cylindrical structures, open at the base (see Figure 1). During installation they cut a small distance into the seabed under their own weight, but are then installed to their full depth (with the caisson lid flush with the seabed) by pumping out the water that is trapped within the caisson. They can be installed in either clays or sands. The principal advantage for the offshore wind application is that the caissons can be installed rapidly, using relatively inexpensive equipment. Methods for designing caisson foundations for offshore wind applications are in their infancy, and in response to the need for design methods, a programme of research has been sponsored by the DTI, EPSRC and a consortium of companies (see Acknowledgements) (Byrne et al. 22). In this 1 Department of Engineering Science, Oxford University 2 Fugro Structural Monitoring Ltd.

5 2 study, design methods are being developed principally on the basis of small-scale model testing (Byrne and Houlsby (22, 24), Byrne et al. (23), Kelly et al. (23, 24)), but an important part of the research is a programme of intermediate scale field trials to check on the scalability of the results. In this paper tests on 1.5m and 3.m diameter caissons at the Bothkennar test site are reported. Typical sizes of prototype caissons are discussed below. Other studies of design methods for suction caissons for foundation (as opposed to anchor) applications have concentrated on analytical procedures (e.g. Bransby and Randolph, 1998) or finite element analysis (e.g. Gourvenec and Randolph, 23). A key feature of offshore wind turbine structures is that (for their size) they are relatively light (with a mass of the order of 6t for a 3.5MW turbine structure), yet they are subjected to large horizontal forces and overturning moments from wind and waves (Byrne and Houlsby, 23). The horizontal load may, for instance, be of the order of 65% of the vertical load. Thus the challenge to the foundation engineer is to carry large (and repetitive) horizontal loads and overturning moments, but relatively little vertical load. Two main structural configurations using caissons are being considered: either a monopod consisting of a single large caisson (typically 2m to 25m in diameter for a modern large turbine structure), or a tetrapod in which the load is transferred through a truss structure to four smaller caissons, see Figure 2 (preliminary calculations indicate that the obvious alternative of a tripod is a less favourable configuration). Each of the smaller caissons might be say 6m to 8m in diameter. For the monopod the most important load on the foundation is the overturning moment. In the case of the tetrapod the moment loading is principally carried by push-pull action by opposing footings, and it is the variation of vertical load (and in particular the possibility of tension on a footing) that is most important. In both cases the design objective is to select an appropriate diameter D and depth h of the caisson, and in the tetrapod case the spacing s must also be determined (see Figure 2). The testing programme described below includes tests directed towards the design of both the monopod and tetrapod. Data were obtained from the installation phases for each caisson. Loading of the caissons was by means of a combination of dead weights, hydraulic jacks and inertial loading from a Structural Eccentric Mass Vibrator (SEMV). The test programme was designed by Oxford University, and site operations were managed by Fugro Structural Monitoring Ltd. The tests were carried out in December 23 and January 24.

6 3 EQUIPMENT AND TESTING PROCEDURES The caissons were fabricated from mild steel, with the principal dimensions given in Table 1. The lids of the caissons were stiffened by I-sections. Ports were provided for attaching the pump and for venting the caissons. An A-frame structure was attached by pins to the 3.m caisson to transfer loads from the SEMV or hydraulic jack. Load cells were fitted to measure the axial load in all four legs of the frame. Reaction loads for testing were provided by a steel frame, and the layout of the entire testing assembly is shown in Figure 3. The reaction frame was supported on two square foundations (2m x 2m x 1m deep), which were also installed by the suction method. The frame was installed in a pit of depth approximately 1.5m at the Bothkennar test site, see Figure 4. Details of the soil properties at Bothkennar are reported in the collection of papers in Géotechnique, Volume 42, Number 2 (June 1992). The best estimate is that the base of the pit corresponds to a depth of 1.75m in Nash et al. (1992), so that the undrained shear strength (measured by the undrained triaxial test), taken from the figures in Nash et al. (1992) is s u = z where s u is in kpa and z is the depth in metres below the base of the excavation. Salient values are therefore s =11.4kPa at the soil surface, 13.4kPa at the base of the 1.5m caisson and 14.4kPa at the base of the 3.m caisson. The bulk density of the clay at relevant depths is estimated as 168kg/m 3. Throughout the testing period the base of the test pit was covered by about.25m of water. The vertical bearing capacities of the small and large caissons were estimated (using the method of Houlsby and Martin, 23) as 164kN and 746kN respectively. In addition to the clevis pin load cells at the base of the A-frame on the 3.m caisson, loads applied by all hydraulic jacks used in the testing were measured by further load cells. Displacements of the caissons were measured by draw-wire transducers attached to a scaffold frame based at least 1.5m away from the outside of the caisson. The draw-wires were attached to the caisson via upstand frames to keep the transducers clear of the water. Six transducers were used to resolve all six degrees-of-freedom of movement. The co-ordinates of the transducers were first determined using surveying techniques. During SEMV tests, displacements were also monitored by means of accelerometers. One pore water pressure transducer was fixed centrally to the underside of the lid of each caisson, and two more on the inside of the caisson just above the base of the skirt (at opposite ends of a diameter). The general layout of the instrumentation is shown in Figure 5. All transducer data were logged at 5Hz for most tests, with averages over 5 readings used for further analysis, and at 4Hz for the SEMV tests (without data averaging). u

7 4 The caissons were installed according to the following procedure. (a) The caisson was lowered to the soil surface and allowed to penetrate under its own weight with the interior vented to atmosphere. (b) Water was pumped into the caisson until it was full. (c) The vent was closed and water pumped out of the caisson to install it to full depth. During phase (c) dead weights were added to the caisson to correct (as far as possible) any errors in levelling of the caisson. All subsequent testing (except where noted) was carried out with the caisson vents sealed. The tests on the 3.m caisson were relevant to a monopod. Small amplitude cyclic horizontal loads were applied at the top of the A-frame (4.23m above the lid of the caisson) by means of the SEMV, operating at 1Hz, at which frequency the applied load was ± 5.kN. These loads are intended to be primarily representative of wave loads experienced by a prototype structure, but scale to waves of different magnitudes (and return periods) for different sizes of prototype. No attempt has been made to scale the wave frequency, as undrained conditions are assumed in both test and prototype, and dynamic effects may be accounted for as discussed below. A fixed bias to the horizontal loading (representing a wind and/or current loading) was achieved by suspending a 4kg block from a pulley system attached to the top of the A-frame (see Figure 3(b)). The vertical load on the caisson throughout these tests was augmented by a 24kg concrete block. During the first series of SEMV tests an interesting observation was that, whilst the caisson moved only imperceptibly, vibration at 1Hz transmitted through the ground set up a resonance in the scaffold reference frame for displacement measurement (thus effectively rendering the displacement measurements useless). This could not be eliminated satisfactorily by stiffening the frame, and in subsequent tests displacements were measured by accelerometers attached directly to the caisson. A second set of tests on the 3.m caisson involved large amplitude (but low frequency) cyclic horizontal loading from a hydraulic jack placed approximately horizontally between the top of the A-frame and the main reaction frame, Figure 3(a). The amplitude of the loads was steadily increased until large (>2mm) movements of the loading point occurred. These tests were principally intended to assess the performance of the caisson under extreme conditions. The tests on the 1.5m caisson were relevant to the tetrapod design. It was first loaded to a fixed vertical load by means of a hydraulic jack. Cyclic inclined loading was then applied using a second jack (inclined at 2:1 to the horizontal). Packets of 1 cycles of increasing load amplitude were applied. The intention was that during these cycles the load in the vertical jack would be held constant, but the stiffness of the hydraulic system rendered it difficult to control this load, which therefore showed significant fluctuations. At the end of the testing the inclined jack was

8 5 disconnected, and the caisson pulled out rapidly by the vertical jack to assess the ultimate tensile capacity. After the tests were completed the caissons were removed from the ground simply by reversing the installation process: i.e. by reconnecting the pump and pumping water back into the caisson. The 1.5m and 3.m caissons were each installed and tested at two locations in the same test pit at Bothkennar: a brief summary of the tests completed is given in Table 2. TEST RESULTS Installation Figure 6 shows the records of measured suction against penetration depth for two installations of the 3.m caisson and one of the 1.5m caisson. Interruptions to the applied suction have been removed from the records for clarity. It can be seen that the variation of suction with depth is reasonably repeatable for the 3.m caisson. Also shown on the figure are the computed profiles of suction, using the procedure described by Houlsby and Byrne (24) (modified slightly to account for the fact that the caisson is not entirely submerged). The calculations used the strength profile quoted in the previous section, together with an adhesion factor α =. 5 both inside and outside the caisson and N = 9 on the rim of the caisson. c It can be seen that in general the computed suction pressure agrees well with the observations as the full depth of the caisson is approached, but underestimates the suction at shallower depths. The most probable explanation is that the fitted strength profile from Nash et al. (1992) is appropriate principally for depths greater than about 3.5m from original ground surface (i.e. z >1.75m approximately). At shallower depths the evidence from vane tests (Nash et al., 1992) is that the strength increases significantly, probably due to past desiccation. The strength of the shallow soil is probably underestimated in the calculation, leading in turn to an underestimation of the suction. Vibration tests on 3.m caisson Figure 7 shows the record of the applied moment (deduced from the load cells in the legs of the A-frame) against time for an SEMV test. The test starts at an offset moment of approximately 16.6kNm (from the block and pulley system shown in Figure 3(b)). As the eccentric mass on the SEMV starts to rotate, it exerts an inertial force at the top of the A-frame which varies with the square of the angular velocity. The amplitude of loading therefore builds up steadily with the

9 6 frequency, with a minor fluctuation at about 7Hz ( t 11s in Figure 7) due to a resonance of the suspended mass providing the offset moment. Figure 8 shows the resulting moment-rotation response. The initial small amplitude cycles (at low frequency) plot as the densely packed curves in the centre of the diagram, showing a high stiffness and relatively little hysteresis. The line through the data shows a rotational stiffness of 225MNm/radian. As the amplitude of moment increases the moment-rotation loop opens up gradually, until the steady state of 1Hz cycling is reached, at which stage an approximately elliptical loop (the outer loop in the diagram) is continually retraced. The open loop arises from damping which has three possible causes (a) viscous material damping, (b) plastic dissipation of energy in the soil and (c) radiation damping. The data can be interpreted by first taking the Fast Fourier Transform of both the moment and rotation to convert to the frequency domain, and then taking the ratio between the two FFT s to obtain the complex, frequency-dependent impedance. The real part of the impedance represents the stiffness and inertial effects, and the imaginary part the damping. The main information on the effects of frequency on the response is contained within the transients at the beginning and end of the test, and Figure 9 shows the real and imaginary parts of the transfer function computed for two 2s periods covering these transients. Also shown at 1Hz are the real and imaginary components computed directly from the steady state response. The data may be compared with theories for the behaviour of a circular foundation on an elastic material. Wolf (1994) describes two lumped-parameter models for this case. Wolf presents models both for surface and embedded footings. Whilst recognising that the caisson is in fact embedded into the soil, we use here a preliminary analysis based on factors for surface footings. The analysis of embedded footings requires consideration of cross-coupling between moment and horizontal loading terms, and these coupling effects differ for the stiffness and damping. Furthermore the role of the inertial terms in the rocking mode is not fully resolved. The principal effect of ignoring the footing embedment is that, in the following, the stiffness coefficients may be underestimated (by a factor which may be in the region of 1.5, but depending on the assumptions about interactions on the side of the caisson, and the variation of stiffness with depth), and the shear modulus correspondingly overestimated. Whilst this affects the absolute values of the moduli discussed below, it does not affect their relative values.

10 7 Wolf s first model is a 3-parameter model represented conceptually by Figure 1(a). For a rigid circular footing at the surface of an elastic soil subjected to moment loads, so that F M (the 3 8GR applied moment) and u θ (the corresponding rotation), the rotational stiffness is k =, 3 ( 1 ν ) where G is the shear modulus and ν is Poisson s ratio. The damping and mass coefficients are 2 R R calculated by c = γ k and m = µ k vs vs 2, where γ and µ are dimensionless coefficients and v s = G ρ is the shear wave velocity ( ρ is the clay density). Note that although the soil is assumed to be purely elastic, there is loss of energy through radiation damping, which accounts for transmission of energy away from the foundation to an infinitely distant boundary. Adapting the methods of Wolf (1994) and Das (1993) suggests the values γ =. 242 and µ =. 24. The real part of the impedance is comparison with the data, computed for 2 k ω m and the imaginary part is ω c, and these are shown on Figure 9 for G = 12.5MPa. Examining first the real part, the theory provides a reasonable fit to the data at low frequency, but overestimates the stiffness at higher frequencies: this is because the higher frequencies involved higher amplitude cycling, for which the secant shear stiffness of the soil would be expected to reduce. Wolf (1994) suggested the alternative 5-parameter model shown conceptually in Figure 1(b) as a more accurate representation of the foundation behaviour. The stiffness K is as for the 3- parameter model, and damping and mass coefficients are defined in a similar way to those in the 3- parameter model, but with suggested values for ν =. 5 of γ =, µ =. 267, γ 1 =. 345, µ 1 =.29. (Note that for the moment-rotation case the general 5-parameter model thus reduces effectively to a 4-parameter model). In this case the real and imaginary parts of the impedance can 2 be calculated as a µ k 1 1 a µ and 1+ a µ 1 γ1 ( ) a µ ak 1 + γ where a = ωr v s is γ1 1+ a µ 1 γ1 the dimensionless angular velocity. Figure 9 shows that the 5-parameter model gives a very similar variation of the real part of the impedance to the 3-parameter model. However, within the range of frequencies tested it gives a lower imaginary part (which represents the damping). Comparing with the data, it is clear that, in order to fit the real part of the impedance the stiffness of the foundation would have to be reduced as the frequency increased (and amplitude of loading increased). Of course this reduction in stiffness is accompanied by an increase in the material damping, which is not taken into account by this model. One therefore expects the

11 8 damping (imaginary part of the impedance) to be underestimated at higher frequencies by the simple elastic model. The 5-parameter model therefore provides a better representation of the real behaviour, since this underestimates the damping. It is worth noting that whilst the rotation of the caisson during steady state cycling is very closely fitted by a 1Hz sinusoid, the moment contains significant higher frequencies, with the 2Hz component being about 15% of the amplitude of the 1Hz fundamental. Such a response is indicative of non-linearities (such as plastic dissipation) not accounted for in the simple 3- or 5-parameter models. Jacking tests on 3.m caisson Following the SEMV tests, the 3.m caisson was subjected to further cycles of moment, but under quasi-static conditions, by loading with a hydraulic jack, Figure 3(a). Figure 11(a) shows an example of the resulting moment-rotation curve for the first few, small amplitude, cycles of test Jack_3._2, showing significant hysteresis even at this stage. Figure 11(b) shows the continuation of the same test to larger amplitude, showing that not only does hysteresis increase with amplitude, but also that there is a degradation of the stiffness over several cycles of loading. The degraded response does, however, appear to be gradually stabilising. Figure 12 shows very large amplitude cycles from test Jack_3._1. These cycles are largely of curiosity value, since at such large displacements a full-size foundation would have failed for all practical purposes. It is worth noting, however, the characteristic shape of the cycles in which, after an initially stiff unloading, a very flexible response is observed, followed by a slight stiffening. This behaviour is typical of a gapping response in which the stiffening occurs as a gap (created by the previous half cycle) is closed. Indeed gaps several tens of millimetres wide and up to 1.2m deep were measured down the side of the caisson during these cycles. The SEMV and jacking tests may be compared as follows. At each frequency the real part of the impedance can be used to deduce a secant shear stiffness of the soil. By making use of the fact that the SEMV applies a load proportional to the square of the frequency, an amplitude of loading can also be attributed to each frequency. Dividing the amplitude of loading by the impedance allows an amplitude of rotation to be determined. Hence the secant stiffness can be expressed as a function of amplitude of rotation. The results are shown on Figure 13, with the set of points for the initial ramp up showing slightly higher stiffness than the ramp down, possibly because some degradation of stiffness is attributable to the cycling. Also shown on the figure is a single point deduced from the steady state conditions: the confidence attached to this value is much higher than for any of the

12 9 other data, because it is based on many more readings. It is, however, entirely consistent with the transient data. Also shown on Figure 13 are the secant stiffness values from jacking test Jack_3._2 (conducted immediately after SEMV_2_2), and Jack_3._1 (conducted during an earlier installation of the caisson). These are consistent with the SEMV data in that they show a continuing reduction of the stiffness with increasing amplitude of cycling, indeed the shape of the ( θ) G log curve is very similar to the familiar pattern for variation of stiffness with strain on a ( γ) G log plot. The rotation of the caisson is of course approximately proportional to the shear strain amplitude in the soil. Test Jack_3._1 shows a somewhat higher stiffness than Jack_3._2, probably because better control of level was achieved during installation of this caisson, so that a better contact between soil and the lid of the caisson was probably achieved. (In practice it is likely that any void between the soil and the lid would be grouted to ensure best performance of the caisson). Finally Figure 13 shows a simple fit to the variation of the secant shear stiffness based on the hyperbolic momentrotation relation =, where K is the initial value of the rotational stiffness, M M max M K θ M + AM max M max is the maximum moment, and K K5 = 2 + A where K 5 is the secant rotational stiffness at half the maximum moment. The curve is constructed for to G = 14MPa ), M max = 135kNm and A = 8. K = 252MNm/radian (corresponding As mentioned above, a more accurate interpretation of the moment loading tests could be made by accounting for the embedment of the foundation in the calculation of stiffness factors (see e.g. Doherty and Deeks (23)), but this would simply reduce the absolute values of the estimated stiffness, and not significantly change the relative values or the overall interpretation of the pattern of response. Jacking tests on 1.5m caisson The 1.5m caisson was first loaded to approximately 12kN by the vertical jack, and then subjected to cyclic loading from the inclined jack. Because of practical difficulties with simultaneous control of pressures in two hydraulic jacks, the load path was not, however a simple line in V-H space, but involved a rather complex path as shown in Figure 14 (for clarity in this figure a considerable amount of data has been removed, and the paths plotted only for a few cycles near the beginning, middle and end of the test). Importantly, however, the path involved the main element of loading in the field, in that horizontal loading is accompanied by changes in vertical

13 1 loading too: the path was selected to represent a realistic ratio between these changes. Note, however, that in the prototype tetrapod the caisson would be restrained against rotation, whereas in the tests the caisson was free to rotate. Figure 15(a) shows the horizontal load against displacement for the first few cycles of test Jack_1.5_2. In spite of the complications caused by scatter in data, the hysteresis loops are clear. Figure 15(b) shows the continuation of the test to a series of packets of ten cycles at increasing amplitude of load. For the tests up to H = ±3kN there is very little degradation of response, but the tests at H = ±4kN show a clear degradation with cycling. Using the bearing capacity factors of Houlsby and Martin (23) the computed bearing capacity for the foundation is 163kN, so this degradation may well be due to the bearing capacity of the foundation being reached during the compressive cycles, see Figure 16. Figure 16 shows the vertical movements throughout test Jack_1.5_2. During the first few cycles there is very little vertical movement. During the intermediate cycles some vertical downward movement accumulates during the cycling, but rapidly stabilises, and finally the largest cycles cause ongoing significant vertical movements. It appears therefore that for both horizontal load and moment cycling there is a pattern of stiff response with little hysteresis at very small cyclic loads only. As loads increase the stiffness reduces and hysteresis increases, but the loops are fairly stable. Eventually a load level is reached at which a rapid deterioration of performance with number of cycles is observed. Pullout tests on 1.5m caisson At the end of the jacking tests on the 1.5m caisson, the caisson was pulled out rapidly by means of the vertical jack. The results for test Pull_1 are shown in Figure 17. The tensile load decreases rapidly to about -15kN, and after some minor fluctuations pullout occurs at a relatively constant load. Figure 18 shows the variation with time of pore pressure measured under the lid of the caisson and at the tip of the caisson. Also shown is the total vertical load, converted to the dimensions of pressure by dividing by the area of the caisson. It can be seen that the pore pressure under the lid is approximately -8kPa (relative to atmospheric pressure) throughout pullout, indicating that cavitation has probably occurred, with the formation of a void between the caisson lid and the soil. The tip pore pressure is approximately -1kPa, indicating cavitation beneath the tip. The relatively small difference between the total load (converted to pressure) and the pressure beneath the lid represents the friction of the sides of the caisson. The pressure difference of about

14 11 12kPa converts to shear stress of about 2.25kPa on the inside and outside of the caisson, indicating an α value (shear stress divided by undrained shear strength) of only.2. Note also on Figure 17 that once the tensile load exceeds about 15kN (corresponding approximately to the estimated friction) the stiffness drops significantly below that encountered during compressive loading. (The initial compressive loading curve is shown on Figure 17 for comparison, reversed and shifted to the same origin as the pullout). Experience from model testing in sands (Kelly et al. 24) suggests that if the friction value is exceeded during cycling, that rapid degradation of the foundation would occur. After about 115mm of pullout (at time 16:44) the vent to the caisson was opened, at which stage the pressure within the caisson rose to near atmospheric and extraction occurred at much lower loads. IMPLICATIONS FOR FULL-SCALE FOUNDATIONS The principal purpose of the tests described here is for calibration of force resultant theoretical models based on work-hardening plasticity to describe the response of caisson foundations (Houlsby, 23). However, some simple scaling can be applied to the results of the tests to make some preliminary estimates of the sizes of caissons that would be needed for full scale wind turbine installations. A 3.5MN wind turbine in typical offshore conditions would result in an overturning moment, in extreme conditions, of approximately 12MNm (Byrne and Houlsby, 23). If at this stage it is determined that (say) an acceptable one-way rotation of the foundation is.1 radian, then the results from Figure 11 indicate that for a 2-way rotation of.2 radian a typical mobilised value for a soft clay as at Bothkennar would be about 175. Assuming a soil of strength G su s = 6kPa, it can be estimated that a caisson of diameter 26m would be required to provide a sufficiently stiff response. The cyclic nature of the applied loading is principally due to the waves, which may have a period of about 1s. For this case the dimensionless frequency a would be about.34, indicating that the dynamic effects would be small, and a quasi-static analysis of the foundation would be justified. If alternatively an approach based on strength were adopted, then it may be estimated that the 3.m caisson was able to sustain cyclic moments of about 7kNm without significant degradation of response. Since the moment capacity scales linearly with the shear strength, and with the cube of the foundation size, it is concluded that a foundation of 22m diameter would be required in 6kPa u

15 12 clay with similar properties to that at Bothkennar. Either a strength or a stiffness criterion therefore results in a foundation of comparable magnitude, but serviceability considerations (i.e. deformations) lead to a requirement for a larger foundation. If a tetrapod were to be designed then first the caisson spacing must be determined. For an overturning moment of 12MNm and a weight of the structure of say 6MN, then a spacing of 4m is needed if tension is to be avoided completely. The maximum loading on an individual caisson would be 3MN, which could be carried in a clay of strength 6kPa with a factor of safety of about 1.5 by a caisson of diameter 4.m. The estimated shear load of 4MN could also be carried by foundations of this size. It is difficult, however, to assess the influence such a caisson would have on the stiffness of the structure without more detailed knowledge of the structure itself. CONCLUSIONS A series of field trials of caisson foundations in soft clay are described. The tests are relevant to both monopod and tetrapod designs for foundations for offshore wind turbines. Installation of the caissons was achieved by suction. High frequency, low amplitude cyclic moment tests on a 3.m caisson showed that the response was affected by stiffness, inertial and damping effects. Low frequency cyclic moment tests on the 3.m caisson indicate a stiff response at low amplitude, with a gradual reduction of stiffness and increase of hysteresis at large amplitude. There was evidence of gapping at the side of the caisson under very large amplitude cycles. Cyclic inclined loading tests on a 1.5m diameter caisson also show a reduction of stiffness and increase of hysteresis as load amplitude increases, with a significant reduction in stiffness after the compression to tension boundary is crossed and frictional capacity exceeded. Pullout of the 1.5m caisson indicated that ultimate tensile resistance is governed by cavitation beneath the foundation. The tests contribute to the development of design procedures for offshore wind turbines founded on caissons. ACKNOWLEDGEMENTS This research was sponsored by the DTI and a consortium of companies (Fugro Ltd, SLP Engineering Ltd, Garrad Hassan, General Electric Wind Ltd, Aerolaminates Ltd and Shell Renewables Ltd). The authors are very grateful to the Royal Society for the Protection of Birds (and in particular to Mike Trubridge) for making the site available for this testing. The authors thank Dr A. Blakeborough for use of the SEMV designed by him, and for advice on interpretation of the SEMV tests.

16 13 REFERENCES Bransby, M.F. and Randolph, M.F, (1998) Combined loading of skirted foundations, Géotechnique, Vol. 48, No. 5, Bye, A., Erbrich, C., Rognlien, B. and Tjelta, T.I. (1995) Geotechnical design of bucket foundations, Proc. Offshore Technology Conference, OTC 7793 Byrne, B.W. and Houlsby, G.T. (22) Experimental Investigations of the Response of Suction Caissons to Transient Vertical Loading, Proc. ASCE, Journal of Geotechnical Engineering, Vol. 128, No. 11, Nov., pp Byrne, B.W. and Houlsby, G.T. (23) Foundations for Offshore Wind Turbines, Phil. Trans. of the Royal Society of London, Series A, Vol. 361, December, pp Byrne, B.W., Houlsby, G.T., Martin, C.M. and Fish, P. (22) "Suction Caisson Foundations for Offshore Wind Turbines", Wind Engineering, Vol. 26, No. 3, pp Byrne, B.W., Villalobos, F. Houlsby, G.T. and Martin, C.M. (23) Laboratory Testing of Shallow Skirted Foundations in Sand, Proc. Int. Conf. on Foundations, Dundee, 2-5 September, Thomas Telford, pp Byrne, B.W. and Houlsby, G.T. (24) Experimental Investigations of the Response of Suction Caissons to Transient Combined Loading, Proc. ASCE, Jour. of Geotechnical and Geoenvironmental Engineering, Vol. 13, No. 3, pp Das, B.M. (1993) Principles of Soil Dynamics, Brooks/Cole, California Doherty, J.P. and Deeks, A.J. (23) Elastic response of circular footings embedded in a nonhomogeneous half-space, Géotechnique, Vol. 53, No. 8, October, pp Gourvenec, S. and Randolph, M.F. (23) Bearing capacity of a skirted foundation under V,H,M loading, Proc. 22 nd Int. Conf. On Offshore mechanics and Arctic Engineering, 8-13 June, Cancun, Mexico, OMAE Houlsby, G.T. (23) "Modelling of Shallow Foundations for Offshore Structures", Invited Theme Lecture, Proc. Int. Conf. on Foundations, Dundee, 2-5 September, Thomas Telford, pp Houlsby, G.T. and Byrne, B.W. (2) Suction Caisson Foundations for Offshore Wind Turbines and Anemometer Masts, Wind Engineering, Vol. 24, No. 4, pp Houlsby, G.T. and Byrne (24) Calculation procedures for installation of suction caissons in clay, submitted to Geotechnical Engineering Houlsby, G.T and Martin, C.M. (23) "Undrained Bearing Capacity Factors for Conical Footings on Clay", Géotechnique, Vol. 53, No. 5, June, pp Kelly, R.B., Byrne, B.W., Houlsby, G.T. and Martin, C.M. (23) "Pressure Chamber Testing of Model Caisson Foundations in Sand", Proc. Int. Conf. Foundations, Dundee, 2-5 September, Thomas Telford, pp Kelly, R.B., Byrne, B.W., Houlsby, G.T. and Martin, C.M. (24) "Tensile Loading of Model Caisson Foundations for Structures on Sand", Proc. ISOPE, Toulon, in press Nash, D.F.T., Powell, J.J.M. and Lloyd, I.M. (1992) Initial Investigations of the Soft Clay Test Site at Bothkennar, Géotechnique, Vol. 42, No. 2, pp Wolf, J.P. (1994) Foundation Vibration Analysis Using Simple Physical Models, Prentice Hall, New Jersey

17 14 Diameter Skirt length L D ratio Wall thickness Approximate mass (including appurtenances) 1.5m 1.m.67 8mm 67kg 3.m 1.5m.5 8mm 2kg Table 1: Details of caisson dimensions Caisson Installation Test type Code Notes 1.5m 1 Installation Inst_1.5_1 Jacking test Jack_1.5_1 Pull out Pull_1 2 Installation Inst_1.5_2 No suction data Jacking test Jack_1.5_2 Pull out Pull_2 Inclined jack attached, but not pressurised 3.m 1 Installation Inst_3._1 SEMV tests SEMV_1_1 SEMV_1_2 SEMV_1_3 No accelerometer data No accelerometer data No accelerometer data Jacking test Jack_3._1 2 Installation Inst_3._4 SEMV tests SEMV_2_1 SEMV_2_2 5Hz logging Jacking test Jack_3._2 Table 2: Outline of caisson tests carried out at Bothkennar Flow Pressure differential W Flow Figure 1: Installation of a suction caisson

18 15 Turbine support structure Water surface Seabed h h Caisson D D s Caissons Figure 2: Possible configurations for suction caisson foundations for wind turbines H L V V A W A H H B B L L L L L L R 15 C R 3 C C (a) Figure 3: Outline of field testing equipment, dimensions in mm (water in excavation and displacement reference frames not shown). (a) arrangement for jacking tests on 1.5m and 3.m caissons, (b) alternative arrangement during SEMV tests. Labels indicate (A) A-frame, (B) concrete block, (C) caissons, (H) hydraulic jacks, (L) load cells, (R) foundations of reaction frame, (V) SEMV, (W) weight providing offset load for SEMV tests (b)

19 16 Figure 4: Test rig showing the 1.5m caisson installed and 3.m caisson in place for installation Draw-wire displacement transducers Vertical Accelerometer 3-axis Accelerometer Pressure sensors Figure 5: Outline of instrumentation on caisson (draw wire transducer reference frame not shown)

20 17 Suction (kpa) Displacement (mm) m Inst_1.5_1 1.5m calculated 3.m Inst_3._1 3.m Inst_3._2 3.m calculated Figure 6: Records of suction during penetration 4 3 Moment M (knm) Time (s) Figure 7: Moment v. time for initial phase of test SEMV_2_2

21 Moment M (knm) Rotation θ (radians) Figure 8: Moment-rotation response of caisson in test SEMV_2_ Impedance (knm/radian) 15 1 Real, 3-parameter Real, 5-parameter Real SEMV, ramp up Real SEMV, ramp down Real, 1Hz steady state Imaginary, 3-parameter Imaginary, 5-parameter Imaginary SEMV, ramp up Imaginary SEMV, ramp down Imaginary, 1Hz steady state Frequency (Hz) Figure 9: Complex θ M transfer function for test SEMV_2_2 compared with theoretical expressions for 3- and 5-parameter models

22 19 F u F u m m c 1 u 1 k c k c m 1 (a) (b) Figure 1: Conceptual models for stiffness, damping and mass of a foundation (a) 3-parameter model, (b) 5-parameter model Moment M (knm) Moment M (knm) Rotation θ (radians) -1 Rotation θ (radians) (a) Figure 11: Moment-rotation curve for loading of 3.m caisson, test Jack_3._2. (a) detail of small amplitude cycles, (b) medium amplitude cycles (b)

23 Moment M (knm) Rotation θ (radians) Figure 12: Moment-rotation curve for loading of 3.m caisson, test Jack_3._1: large amplitude cycles SEMV_2_2, ramp up SEMV_2_2, ramp down SEMV_2_2, 1Hz steady state Jack_3._2 Jack_3._1 Hyperbolic fit 1 G (MPa) θ (radians) Figure 13: Computed secant shear modulus from test SEMV_2_2 and jacking tests on 3.m caisson

24 Horizontal Load (kn) Vertical Load (kn) Figure 14: Load path in cyclic inclined loading test Jack_1.5_2 on 1.5m caisson Horizontal Load (kn) Horizontal Load (kn) Horizontal Displacement (mm) -5 Horizontal Displacement (mm) (a) Figure 15: Horizontal movement during inclined loading test Jack_1.5_2 on 1.5m caisson. (a) detail of small amplitude cycles, (b) large amplitude cycles (b)

25 Vertical Load (kn) Vertical Displacement (mm) Figure 16: Vertical movement during inclined cyclic loading test Jack_1.5_2 on 1.5m caisson Vertical displacement (mm) Pullout Loading curve (shifted and reversed) Vertical load (kn) Figure 17: Load v. displacement during pullout test Pull_1

26 :25 16:3 16:35 16:4 16:45 16:5-2 Pressure (kpa) Lid pressure Skirt pressure Total load/area -12 Figure 18: Record of pull out of 1.5m caisson, test Pull_1

27

28 1 Field trials of suction caissons in sand for offshore wind turbine foundations G.T. Houlsby 1, R.B. Kelly 1, J. Huxtable 2 and B.W. Byrne 1 Keywords: bearing capacity, sand, dynamic test, foundations, stiffness ABSTRACT A programme of testing on suction caisson foundations in an artificially prepared sand test bed near Luce Bay, in Scotland, is described. The tests are relevant to the design of either monopod or quadruped foundations for offshore wind turbines. Records are presented for suction installation of the caissons, cyclic moment loading under both quasi-static and dynamic conditions to simulate the behaviour of a monopod foundation, and cyclic vertical loading and pullout of caissons to simulate one footing in a quadruped foundation. Variations of stiffness with loading level of the foundation are observed, with high initial stiffness followed by hysteretic behaviour at moderate loads and degradation of response at high loads. Some implications for the design of wind turbine foundations are briefly discussed. INTRODUCTION The offshore wind energy industry is a very rapidly expanding sector of vital economic importance in the UK, and the foundation represents an important part of the costs of offshore wind turbine installations (Byrne and Houlsby, 23). This paper describes a series of tests directed towards the understanding of the behaviour of suction caisson foundations in sand, as possible foundations for offshore wind turbines. An earlier paper, by Houlsby et al. (24), describes in detail the background motivation to this testing, and presents equivalent data for tests in clay. The testing programme described here follows a similar pattern to the tests in clay, and includes tests directed towards the design of both monopod and quadruped foundations for offshore wind turbines. The most important load case for a monopod foundation is the applied overturning moment, whilst for a quadruped foundation thee vertical loading on an individual caissons is most important. Data were obtained from the installation phases for each caisson. Loading of the caissons was by means of a combination of dead weights, hydraulic jacks and inertial loading from a Structural Eccentric Mass Vibrator (SEMV). The test programme was designed by Oxford University, and site operations were managed by Fugro Structural Monitoring Ltd. The tests were carried out in February and March Department of Engineering Science, Oxford University 2 Fugro Structural Monitoring Ltd.

29 2 THE LUCE BAY TEST SITE The tests were carried out in a specially prepared sand bed in an aggregate extraction quarry near Luce Bay, Dumfries and Galloway, Scotland. The bed was prepared by placing selected fill in layers approximately 25mm thick and compacting them by multiple passes of a wheeled loader. The sand bed was approximately 4m x 15m x 3.5m deep. A grading curve for the sand is shown in Figure 1. Although the sand itself is almost single-sized at about.3mm to.4mm, it can be seen that there is a significant (15%) gravel content. The sand bed was also observed to contain a proportion of small rounded stones up to about 8mm in size, even though every attempt was made to remove this coarser material in the preparation of the bed. The fines content is negligible. The fill was placed in an unsaturated condition above the water table in the summer. The entire sand bed was allowed to flood slowly over a period of several months, and by the time of the testing had been submerged for four months. During testing there was about 15mm depth of water over the sand surface. The bed of sand was characterised principally by means of in situ testing. Three CPT tests with pore pressure measurement, two cone pressuremeter tests and three seismic cone tests were conducted. The cone resistance records of three CPT and two cone pressuremeter tests are shown superimposed in Figure 2, showing (within some variability) a strong increase of cone resistance with depth, principally due to the increasing stress level. Using standard correlations, the estimated relative density of the sand bed was 8-85%. In the lower part of the sand layer there appears to be some looser material, and beneath the sand are much softer deposits that were not investigated in detail. Figure 3 shows profiles of shear modulus with depth inferred from the cone pressuremeter and seismic cone tests. Down to a depth of 2m there is generally good agreement between the seismic and cone pressuremeter data, with the exception of that from seismic cone test 1. At greater depths, the cone pressuremeter data suggests a higher shear modulus than the seismic cone data. The shear stiffness can be characterised by the relationship G p a = 25 p pa, where a p is atmospheric pressure and p is the mean effective stress estimated on the basis that the saturated unit weight was 1.3kN/m 3 and estimating K as.5. This relationship has been fitted to the upper bound of the test data, as disturbance during testing is expected to lead to underestimates of the small strain shear modulus. We note, however, that the shear moduli observed at this site are quite high for sand at shallow depths, and this may be due to the compaction procedures used in preparation of the bed.

30 3 EQUIPMENT AND TESTING PROCEDURES The caissons, loading frame and instrumentation methods used for the tests in sand were essentially identical to those employed in the previous clay tests at Bothkennar (Houlsby et al., 24). A caisson of diameter of 3.m and a with skirt 1.5m deep was used for moment loading tests. A second caisson of diameter of 1.5m and with a 1.m skirt was used for vertical loading tests. An outline diagram of the test set-up is given in Figure 4, and a photograph of the rig installed at the Luce Bay testing site shown in Figure 5. Prior to suction-assisted installation the caissons were allowed to penetrate under their own weight with the vent to the caisson open to air. In the case of the 3m caisson a 24kg mass was added to the caisson to cause a slight further penetration. The caisson vent was then closed and suction applied by pumping out the trapped air from inside the caisson. The pump used was capable of pumping both air and water. A flow meter was installed between the caisson and pump. The testing procedures in sand were very similar to those adopted for the tests in clay (Houlsby et al, 24). The principal exception related to the SEMV tests using a Structural Eccentric Mass Vibrator to apply relatively high-frequency cyclic loading. Experience from the earlier clay tests indicated the value of having a spread of data obtained at different frequencies of loading. At Luce Bay a series of tests were therefore carried out, in which the frequency was increased in steps, with approximately 15s of cycling at each frequency. A more reliable definition of the variation of the complex impedance with frequency was possible using this method. The 1.5m and 3.m caissons were each installed and tested at two locations in the test bed at Luce Bay: a brief summary of the tests completed is given in Table 1. TEST RESULTS Installation Figure 6 shows the records of measured suction against penetration depth for one installation of the 3.m caisson and one of the 1.5m caisson. It can be seen that the variation of the required suction with depth is similar for both caissons. Also shown on the figure are the computed profiles of suction, using the procedure described by Houlsby and Byrne (24) (modified slightly to account for the fact that the caisson is not entirely submerged). The parameters used in the calculations were a soil friction angle of 45, Ktanδ = 1, an effective unit weight of 1.3kN/m 3, the ratio of permeability inside the caisson to outside the caisson was 3 and stress distribution factors m and n were taken as 1. The vertical load, including self weight, applied to the caissons was

31 4 7kN and 6kN for the 1.5m and 3.m diameter caissons respectively. Using these values, the final pressure for installation of the 3m diameter caisson is predicted closely, while the final pressure required to install the 1.5m diameter caisson is under-predicted. During the actual installations there were a number of stoppages in pumping for a variety of reasons. It was observed that after each stoppage movement only occurred once the suction returned to a value similar to that before the stoppage. Figure 7 shows the excess pore pressure measured by sensors placed inside the caissons, 5mm above the tip of the skirts, plotted against suction measured immediately beneath the lid of the caisson. Although there is some scatter, largely related to the stoppages, there is a strong correlation between the two pressures, as would be expected. Approximately 65% of the suction pressure is measured near the tip of the skirts, and this fraction appears to remain approximately constant throughout the installations. Figure 8 shows a comparison between the volume of air and water removed from the caisson (by integration of the flow measurement) and the volume computed from the cross sectional area of the caisson multiplied by the depth of penetration. There is a close relationship between the two. It can be concluded that during the installation (a) little heave occurred within the caisson and (b) the volume of water seeping through the sand was small. Either of the above phenomena would have resulted in significantly larger volumes from the flow measurements. Vibration tests on 3.m caisson Figure 9 shows the record of the applied moment (deduced from the load cells in the legs of the A-frame) against time for an SEMV test in which the excitation is ramped to a constant frequency of 8Hz and held for about 18s. As the eccentric mass on the SEMV starts to rotate, it exerts an inertial force at the top of the A-frame which varies with the square of the angular velocity. The amplitude of loading therefore builds up parabolically with the frequency. Figure 1 shows results demonstrating the moment-rotation response, and was compiled by extracting data from a number of tests, each similar to the one shown in Figure 9, but at different frequencies. From each test the steady state response has been extracted. The small amplitude cycles (at frequencies less than 6Hz) are not shown, as the movements induced in the caisson were too small to be accurately recorded by the accelerometers. As the amplitude of moment increases the moment-rotation loop opens up gradually, until the steady state of 1Hz cycling is reached, at which stage an approximately elliptical loop is continually retraced. The open loop arises from

32 5 damping which has three possible causes (a) viscous material damping, (b) plastic dissipation of energy in the soil and (c) radiation damping. The minor axes of these ellipses are much smaller than those recorded at Bothkennar, indicating that much less damping occurs in dense sand than in soft clay. The data can be interpreted by first taking the Fast Fourier Transform of both the moment and rotation signals to convert to the frequency domain, and then taking the ratio between the two FFT s to obtain the complex, frequency-dependent impedance (a generalised form of the rotational stiffness). The real part of the impedance represents stiffness and inertial effects, and the imaginary part the damping. At Bothkennar, the main information on the effects of frequency on the response was obtained from the transients at the beginning and end of the test, while at Luce Bay the information was gathered from the steady state of individual tests, stepped at 1Hz intervals from 1Hz to 1Hz. Figure 11 shows the real and imaginary parts of the transfer function computed from 5Hz-1Hz. Also shown as discrete points are the average responses computed at each steady state. The averaged points can be regarded as much more reliable, as they are derived from much more data, but it can be seen that in fact the transfer function deduced from the transient response agrees well with the averaged points. The data may be compared with theories for the behaviour of a circular foundation on an elastic material. As discussed by Houlsby et al. (24), Wolf (1994) describes two lumpedparameter models for this case, a 3-parameter model and a 5-parameter model. The fitted response from a 3-parameter model is shown on Figure 11 for comparison with the test data, computed for G = 85MPa and ν =.2. Examining first the real part, the stiffness is chosen to provide a reasonable fit to the data at frequencies in the range 5Hz to 6Hz, but overestimates the stiffness at higher frequencies. This is of course consistent with the common observation that the stiffness of soils reduces with strain amplitude, so that a constant stiffness model cannot fit behaviour across a wide range of strain amplitudes. No fit from a 5-parameter model is shown in Figure 11, as the 5- parameter model gives a very similar variation of the real part of the impedance to the 3-parameter model. However, within the range of frequencies tested the 5-parameter model gives a much lower imaginary part of the impedance (representing the damping). The 3-parameter model appears to be more satisfactory in this respect. Jacking tests on 3.m caisson Following the SEMV tests, the 3.m caisson was subjected to further cycles of moment, but under quasi-static conditions, by loading with a hydraulic jack, Figure 4(a). Figure 12 shows the

33 6 moment-rotation curve for cycles from test LB_Jack_3._1. The moment-rotation response is initially stiff with little hysteresis. At larger amplitudes of rotation the secant stiffness decreases and hysteresis increases as the amplitude increases. The cycles at very large amplitude have a characteristic shape in which, after an initially stiff unloading, a very flexible response is observed, followed by slight stiffening. This behaviour is typical of a gapping response in which the stiffening occurs as the gap created during the previous half cycle is closed. Gaps were observed down the side of the caisson during these cycles. The moment-rotation curve for jacking test LB_Jack_3._2 is shown in Figure 13. In this test packets of 1 cycles were applied to the caisson with amplitudes of ±42kN/m, ±85kN/m, ±169kN/m and ±254kN/m. As for the data in Figure 12, the displacement amplitude increases with load amplitude and hysteresis occurs during the larger amplitude cycles. The unload-reload parts of the curve in Figure 13 are stiffer than those in Figure 12. There is a slight shakedown apparent in the data at low amplitudes of rotation, with a slight stiffening occurring over several cycles of the same amplitude. The moment-rotation response during each packet of cycles appears to approach a steady state by the end of the packet. Byrne and Houlsby (24) have made similar observations about hysteretic behaviour during cycling from small-scale moment loading tests on caissons in the laboratory. The SEMV and jacking tests may be compared by deducing the equivalent secant shear modulus from the moment-rotation response of the footing using standard formulae for response of a surface footing to moment loads. For the SEMV tests the moment range is taken as the difference between the maximum and minimum values in any given cycle, and the corresponding rotations are computed. The secant stiffness can then be expressed as a function of amplitude of rotation as shown in Figure 14. Also shown on Figure 14 are the secant stiffness values from jacking test LB_Jack_3._1. These are consistent with the SEMV data in that they show a continuing reduction of the stiffness with increasing amplitude of cycling: indeed the shape of the ( θ) G log curve is very similar to the familiar pattern for variation of stiffness with strain on a ( γ) G log plot. The rotation of the caisson is of course approximately proportional to the shear strain amplitude in the soil. Finally Figure 14 shows a simple fit to the variation of the secant shear stiffness based on the hyperbolic moment-rotation relation M K θ = M M max max M + AM, where K is the initial value of the rotational stiffness, M max is the maximum moment, and K K5 = 2 + A. The curve is

34 7 constructed for A = 4. K = 1125MNm/radian (corresponding to G = 1MPa ), M max = 45kNm and More accurate interpretation of the moment loading tests could be made by accounting for the embedment in the calculation of stiffness factors (see e.g. Doherty and Deeks (23)), but this would simply reduce the absolute values of the estimated soil stiffness, and not significantly change the relative values in the overall interpretation of the pattern of response. Jacking tests on 1.5m diameter caissons Two jacking tests were conducted on the 1.5m caisson. Test LB_Jack_1.5_1 involved an application of combined vertical and horizontal loads to the caisson via vertical and inclined jacks. A vertical jack applied the mean vertical load, while the inclined jack provided a cyclic loading component. This test was not successful, as it was not possible to control either the mean or cyclic loads accurately, because the soil/caisson/jack system was sufficiently stiff that the jacks were not able to operate independently of each other. Cyclic vertical loading was applied to the 1.5m diameter caisson in Test LB_Jack_1.5_2 about a mean load of 6kN (not including the self-weight of the caisson). Packets of 1 cycles were applied to the caisson with load amplitudes starting at ±1kN and increasing in steps of ±1kN to ±1kN, and these are shown in Figure 15. The data show that the secant stiffness decreases as the amplitude of the load increases. Furthermore the displacements increase markedly during load packets where the caisson was cycled into tension. It is of interest to note that the net displacement after each cycle in the ±1kN load packet was downward, even though large tensile displacements had occurred during each cycle. The implication of this is that, as long as the mean vertical load is compressive, a caisson foundation cycled into tension will ratchet into the sand rather than out of the sand. However, the data also suggest that caissons should not be loaded in tension for serviceability reasons, as the stiffness of the caisson reduces to a level where foundation movements would render a wind turbine inoperable. The data show that the caisson has a significant cyclic tensile capacity (in excess of -4kN) but extremely large displacements are required to mobilise this capacity. These data support conclusions made from small-scale laboratory tests (Byrne and Houlsby 22, Kelly et al., 23, 24), and this comparison was an important objective for the large-scale testing programme.

35 8 Pullout tests on 1.5m caisson At the end of the jacking tests on the 1.5m caisson, the caisson was pulled out rapidly by means of the vertical jack. The results for test LB_Pull_2 are shown in Figure 16. The tensile load decreases rapidly to about -12kN, when a vent plug in the lid of the caisson was opened to prevent damage to the load cell. The tensile load had not reached a maximum by this stage. The proportion of the tensile load generated by suction pressure inside the caisson is also shown in Figure 16. The difference between the total load carried by the caisson and the suction load represents the friction acting on the skirts of the caisson. The frictional load generated on the skirts of the caisson, as it was pulled out of the sand, is shown in Figure 17, along with an estimate of the friction load computed using the method described by Kelly et al., (23). The calculation was conducted with Ktanδ =.75. This calculation is based on similar principles to that used to predict the suction during installation shown in Figure 6. The parameters used in the calculations are identical, with the exception that Ktanδ = 1 was used for installation. The differences between the values of Ktanδ used are considered to be relatively small, given the approximations involved in modelling caisson installation/extraction in an idealised soil. IMPLICATIONS FOR FULL-SCALE FOUNDATIONS The principal purpose of the tests described here is for calibration of force resultant theoretical models based on work-hardening plasticity to describe the response of caisson foundations (Houlsby, 23). However, some simple scaling can be applied to the results of the tests to make some preliminary estimates of the sizes of caissons that would be needed for full-scale wind turbine installations. A 3.5MN wind turbine in typical offshore conditions would result in an overturning moment, in extreme conditions, of approximately 12MNm (Byrne and Houlsby, 23). If at this stage it is determined that (say) an acceptable two-way rotation of the foundation is.2 radian, then the results from Figure 14 indicate that for this rotation the shear modulus would be about 15MPa. If it is assumed that the shear modulus scales with the square root of the mean stress, then it can be estimated that a caisson of diameter 2m would be required to provide a sufficiently stiff response in sand with a relative density of about 8%. The cyclic nature of the applied loading is principally due to the waves, which may have a period of about 1s. For this case the dimensionless frequency a = ωr/c s, (where ω is the frequency, R is the caisson radius and c s is the shear wave velocity =

36 9 2m/s), would be about.3, indicating that the dynamic effects would be minimal (Wolf, 1994), and a quasi-static analysis of the foundation would be fully justified. If alternatively an approach based on strength were adopted, then it may be estimated that the 3.m caisson was able to sustain cyclic moments of about.12mnm without significant degradation of response. Since the moment capacity scales linearly with the effective unit weight, and with the fourth power of the foundation size, it is concluded that a foundation of 17m diameter would be required in sand with a unit weight of 1.3kN/m 3 with similar properties to that at Luce Bay. Byrne and Houlsby (23) proposed Equation 1 as another method for estimating the diameter of a monopod foundation at low vertical loads: M 2R = 1 1 f3 2RH f + f2 + M ( V W) (1) where, M is the overturning moment, R is the radius of the caisson, H is the horizontal load, V is the vertical load and W is the effective weight of the sand inside the caisson. The factors f 1, f 2 and f 3 were obtained from a limited number of small-scale laboratory tests and are equal to 3.26, 1.7 and.71 respectively. If the weight of the structure was 6MN and the horizontal load was 4MN then Equation 1 indicates that the diameter of the monopod foundation would be about 19m if the ratio of the length of its skirt to its diameter were.5. Either the strength or the stiffness criterion therefore results in a foundation of comparable size, but the latter, which is related serviceability considerations (i.e. deformations) leads to a requirement for a slightly larger foundation. If a quadruped were to be designed then first the caisson spacing must be determined. For an overturning moment of 12MNm and a weight of the structure of 6MN, then a spacing of 4m is needed if tension is to be avoided completely. The confirmation on the basis of large scale tests that tension should be avoided is considered an important, though negative, outcome from this research. The maximum loading on an individual caisson would be 3MN, which could be carried in sand like that at Luce Bay with a factor of safety of about 1.5 by a caisson of diameter 3.5m. The estimated shear load of 4MN can be carried by a caisson with a diameter of 4m and a skirt length of 2.67m. In this case the shear loading appears to govern the size of the caisson required, but it is considered likely that this would change when deformations are taken into account. Unfortunately, however, it is difficult to assess the influence the caisson size would have on the overall stiffness of the structure, without more detailed knowledge of the structure itself.

37 1 A comparison of the estimated caisson diameters for a wind turbine foundation in dense sand with those estimated for soft clay by Houlsby et al., (24) shows that the size of a mono-caisson foundation in dense sand would be about 6% of that in soft clay, whereas the sizes of caissons in a quadruped foundation would be rather similar. CONCLUSIONS A series of field trials of caisson foundations in sand are described. The tests are relevant to the design of both monopod and quadruped foundations for offshore wind turbines. Installation of the caissons was achieved by suction. High frequency, low amplitude cyclic moment tests on a 3.m caisson showed that the response was affected by stiffness, inertial and damping effects. Low frequency cyclic moment tests on the 3.m caisson indicate a stiff response at low amplitude, with a gradual reduction of stiffness and increase of hysteresis at large amplitude. There was evidence of gapping at the side of the caisson under very large amplitude cycles. Cyclic vertical loading tests on a 1.5m diameter caisson also show a reduction of stiffness and increase of hysteresis as load amplitude increases, with a significant reduction in stiffness after the compression to tension boundary is crossed and frictional capacity exceeded. Pullout of the 1.5m caisson indicated that a sizable ultimate tensile resistance can be generated but is accompanied by extremely large displacements. The tests contribute to the development of design procedures for offshore wind turbines founded on caissons. ACKNOWLEDGEMENTS This research was sponsored by the DTI and a consortium of companies (Fugro Ltd, SLP Engineering Ltd, Garrad Hassan, General Electric Wind Ltd, Aerolaminates Ltd and Shell Renewables Ltd). The authors are very grateful to Mr Adam Macintosh of Luce Bay Plant Hire for making the site available for this testing. The authors thank Dr A. Blakeborough for use of the SEMV designed by him, and for advice on interpretation of the SEMV tests. Dr B.W. Byrne acknowledges the support provided by Magdalen College, Oxford. REFERENCES Byrne, B.W. and Houlsby, G.T. (22) Experimental investigations of the response of suction caissons to transient vertical loading, Proc. ASCE, Journal of Geotechnical Engineering, Vol. 128, No. 11, Nov., pp Byrne, B.W. and Houlsby, G.T. (23) Foundations for offshore wind turbines, Phil. Trans. of the Royal Society of London, Series A, Vol. 361, December, pp

38 11 Byrne, B.W. and Houlsby, G.T. (24) Experimental investigations of the response of suction caissons to transient combined loading, Proc. ASCE, Journal of Geotechnical and Geoenvironmental Engineering 13, No 3, Mar., pp Doherty, J.P. and Deeks, A.J. (23) Elastic response of circular footings embedded in a nonhomogeneous half-space, Géotechnique, Vol. 53, No. 8, October, pp Houlsby, G.T. (23) "Modelling of Shallow Foundations for Offshore Structures", Invited Theme Lecture, Proc. Int. Conf. on Foundations, Dundee, 2-5 September, Thomas Telford, pp Houlsby, G.T. and Byrne (24) Design procedures for installation of suction caissons in sand, Proc. ICE, Geotechnical Engineering, in press Houlsby, G.T., Kelly, R.B., Huxtable, J. and Byrne, B.W. (24) Field trials of suction caissons in clay for offshore wind turbine foundations, submitted to Géotechnique, Kelly, R.B., Byrne, B.W., Houlsby, G.T. and Martin, C.M. (23) "Pressure Chamber Testing of Model Caisson Foundations in Sand", Proc. Int. Conf. Foundations, Dundee, 2-5 September, Thomas Telford, pp Kelly, R.B., Byrne, B.W., Houlsby, G.T. and Martin, C.M. (24) "Tensile Loading of Model Caisson Foundations for Structures on Sand", Proc. ISOPE, Toulon, Vol. 2, pp Wolf, J.P. (1994) Foundation Vibration Analysis Using Simple Physical Models, Prentice Hall, New Jersey Caisson Installation Test type Code Notes 1.5m 1 Installation LB_Inst_1.5_1 Jacking test LB_Jack_1.5_1 Combined vertical and horizontal loading: unsuccessful because of control problems 3.m Pull out Installation LB_Pull_1 LB_Inst_1.5_2 Jacking test LB_Jack_1.5_2 Vertical loading only Pull out Installation LB_Pull_2 LB_Inst_3._1 SEMV tests LB_SEMV_1 With offset weight (asymmetric cycling) Jacking test LB_Jack_3._1 Increased amplitude cycling Installation LB_Inst_3._2 SEMV tests LB_SEMV_2 Without offset weight (symmetric cycling) Jacking test LB_Jack_3._2 Multiple cycles at each amplitude Table 1: Outline of caisson tests carried out at Luce Bay

39 Percent Passing Particle Size (mm) Figure 1: Grading curve for test bed at Luce Bay Cone Resistance (MPa) Depth (m) Figure 2: CPT data at Luce Bay

40 13 Shear modulus G (MPa) Depth z (m) Fitted curve Cone pressuremeter Seismic cone 1 Seismic cone 2 Seismic cone Figure 3: Estimated profile of shear modulus with depth at Luce Bay H L V V A W A H H B B L L L L L L R 15 C R 3 C C (a) Figure 4: Outline of field testing equipment, dimensions in mm (water level and displacement reference frames not shown). (a) arrangement for jacking tests on 1.5m and 3.m caissons, (b) alternative arrangement during SEMV tests. Labels indicate (A) A-frame, (B) concrete block, (C) caissons, (H) hydraulic jacks, (L) load cells, (R) foundations of reaction frame, (V) SEMV, (W) weight providing offset load for SEMV tests (b)

41 14 Figure 5: Test rig showing the 1.5m caisson installed in foreground and 3.m caisson in background after a jacking test had been completed Suction (kpa) Displacement (mm) LB_Inst_1.5_2 LB_Inst_3._2 1.5m prediction 3.m prediction 14 Figure 6: Records of suction during penetration