@ 2015 Taylor & Francis Group London ISBN: 978-1-138-02848-7 Frontiers in Offshore Geotechnics IiI Meyer (Ed.) Predicting monopile behaviour for the Gode Wind offshore wind farm F.C. Schroeder & A.S. Merritt Geotechuical Consurlring Group London UK K.W. Serensen A. Muir Wood & C.L. Thilsted DONG Energy: Gentofe Denmark D.M. Potts Imperial College London London UK ABSTRACT: Monopiles used as foundations for offshore wind turbines often have very large diameters and the API p-y method although this was developed for slender piles with relatively small diameters. Theoretical low length to diameter (L/D) ratios. Their response to lateral and moment loading is often modelled based on studies and field monitoring have shown that this method may not accurately describe monopile behaviour in particular the lateral load-displacement response for serviceability conditions. This paper presents advanced 3D Finite Element (FE) analyses undertaken during design of the Gode Wind offshore wind farm at a sand dominated site in the German North Sea. Two turbine locations with different soil profiles and varying monopile L/D ratios poa d id m san uoop pue udsp-peo a aedo o pseue m analyses predict a stiffer response of the monopile from initial loading until the design ULS condition which is consistent with field experience. 1INTRODUCTION API p-y curves can accurately capture the response of to suction caissons than traditional slender piles. large diameter monopiles with geometries much closer industry are monopiles; stiff piles with large diame- Most current foundations used in the offshore wind Secondly critical design issaes are not properly ters driven 2035 m into the seabed. Recently installed turbines that the foundations are subjected to strong taken into account. It is characteristic for offshore wind Gode Wind offshore wind projet will install larger monopiles have diameters of 5.56.5 m. In 2015 the cyclic loading originating from the wind and wave still 7.5 m diameter monopiles in dense sands off the coast of Germany. loads. This occurs not only during extreme conditions The current state-of-practice for design of monopiles but also during serviceability conditions. Therefore in sand is to use p-y curves according to API (2011). the primary design drivers for offshore wind tur- These p-y curves have gained broad recognition due stiffness rather than ultimate capacity: The API p-y bine foundations are often those of deformation and However when applied to offshore wind turbine foun- to the low failure rate of piles over several decades. curves are designed primarily for evaluation of the dations the design methodology is problematic for two ultimate lateral capacity: Important design issues such main reasons: small-strain stiffness are poorly accounted for. as accumulated rotation effects of cyclic loading and Firstly the p-y formulation is being used well Several numerical studies have been undertaken for sand was originally adopted by Murchison and monopiles see e.g. Augustesen et al. (2009) Lesny & beyond its verified range. The API p-y formulation toinvestigate thebehaviourof large diameter 1) e 1a u p (o) d s o sq o (6) ily obtained from two full-scale load tests reported by ever conclusions tend to diverge. In contrast results p po s sg s u psd see pe (6) xo slender and flexible piles with diameter D = 0.61 m that monopiles installed in sand bchave significantly et al. (1974). These tests were conducted using long (2009) and Kallehave et al. (2012) uniquely indicate and length L = 21 m. In contrast the next generation stiffer that predicted by API p-y curves. Furthermore the L/D ratios will bein the range of 34were designed using modified API p-y curves. 3D monopiles have diameters more than 10 times that. The monopiles for Gode Wind offshore wind farm as opposed to the original pile tests where the L/D ratioFE analyses were then used to benchmark and vali- was 34. Itistherefore notreasonable to assume that thedate the geotechnical design in the serviceability and 735
o ssd ed s ss analyses and the results which were benchmarked to the modified API p-y curves for sand The overall con- clusion is that modified API p-y curves for sands still se sd uao e od-pm full-scale measurements described above. as small-strain stiffness in agreement with the reported 2 GODE WIND OFFSHORE WIND FARM The Gode Wind offshore wind project is located in the German part ofthe North Sea 38 km north of the island bine generators each with a rated power output of 6 of Norderney. The wind farm consists of 97 wind tur- Figure 1. Monopile for Gode Wind offshore wind farm MW. (diameler = 7.5 m at seabed). The ground conditions at the project area are rea- sonably homogenous and typical for the German tion/urbine structure. The thickness of the remaining on the natural frequency of the integrated founda- North Sea; the seabed consist mainly of dense to extremely dense Pleistocene sands (CPT q values of cans were governed by a bination of ultimate 50110 MPa) with minor clay and silt layers spread and fatigue limit states as well as robustness for across the site. A thin layer of post-glacial Holocene installation. sediments prising loose to very loose silty sands To validate the monopile designs based on the and medium dense to dense sand layers is present at formed for two turbine locations with different ground modified API p-y curves 3D FE analyses were per- the seabed overlying the Pleistocene strata. The water depths vary from 27.933.6 m and large wave loads are conditions and varying L/D ratios. The analyses incor- present at this North Sea site; the height of the 50-year porated advanced soil models to accurately capture key extreme design wave is 18.5 m. features of the soil behaviour and determine the lat- All foumdations are driven monopiles (see Figure 1) eral load-displacement and moment-rotation response with grouted transition pieces. Diameters are 6.5 m at of the monopiles. turbine interface and 7.5 m at seabed. Each monopile was individually designed and optimised for the posi- tion specific loads and soil conditions. The monopile 3 DETAILS OF THE 3D FE ANALYSES weights are in the range of 700935 tonnes and can wall thickness’ lie in the range of 70100 mm. The 3.1C General penetration depths range from 2432 m resulting in L/D ratios of 3.24.2. The analysis of monopiles to lateral and moment load- The monopiles for Gode Wind were designed using however possible to use one plane of symmetry at the ing requires the performance of 3D analyses. It is modified API p-y curves. The major modifications included a) an un/reloading stiffness correction to load application thus modelling only one half of the centre of the pile in line with the direction of lateral ensure better prediction of as-built natural frequen- cies. The applied formulation was documented on the problem. basis of as-built measurements from existing wind 3D FE analyses were performed for turbine loca- farms b) addition of a toe-spring at the pile tip to tions Q03 and H03. These locations were chosen as the account for low L/D ratios c) lifetime accumulated monopiles at locations Q03 and H03 are at the extreme deformations were calculated according to LeBlanc ends of the L/D ratio spectrum across the site. For both locations the pile diameters below seabed were et al. (2010) while taking account of omni-directional 7.5 m with a wall thickness of generally 75 mm. For site specific soil conditions and d) ultimate limit loading and using model parameters derived for the location Q03 the penetration length considered in the q pabau se Z se paond aam ss 3.2 whereas for location H03 the penetration analysed analysis was 24.2 m resulting in a L/D ratio of around DIN 1054:2010-12 while taking account for poten- tial effects of cyelic degradations and pore pressure was 35.1 m giving a L/D of approximately 4.7. Further build-up during extreme storm events. FE analyses were carried out allowed a reduction of design iterations at this location for which no further The monopile design for Gode Wind was gov- erned by a bination of factors. The SLS criteria the pile length resulting in a maximum L/D across the for accumulated lifetime deformations governed the site of 4.2. The loads applied to the monopile at each location were govermed by the GEO-2 design proof The pile pile length of most piles whereas the remaining piles rent and wind loading. This resulted in a bination were carefully derived considering extreme wave cur- diameter andin some casesthe can thickness was gov erned by the un/reloading stiffness having a large effect of horizontal and moment loads acting at the level of the seabed which is considered herein as the ULS load 9
G (MPa) 30 100 150200250300350400 interpreted CPT data Q03 FE model Q03 • interpreted CPT data H03 FE model H03 15 Figure 2. 3D FE mesh for analysis of Location Q03. 20 condition. In the FE model the structure above seabed and moment load bination to be applied in a dis was modelled up to a height that allowed the horizontal 25 H03 these were heights of 39.4 m and 41.4 m above placement controlled manner. For locations Q03 and seabed respectively. 30 Figure 2 shows the FE mesh used for the analysis below seabed are given by a depth of 40 m and a radius of location Q03. The overall dimensions of the mesh 35 of 75 m. The analyses were performed using the FE code ICFEP. 20 node hexahedral clements were used in 40 bination with 16 node interface elements (Day & Potts 1994) and 8 node shell elements (Schroeder et al. 45 2007). Reduced integration was used for all elements and a modified Newton-Rhapson scheme with an error controlled sub-stepping algorithm was employed as 50 the non-linear solver (Potts & Zdravkovic 1999). Figure 3. Variation of Go with depth CPT data and FE model approximation (Location Q03). 3.2 Geotechnical conditions and soil models power of *n'. This can be reasonably modelled assum- The geotechnical conditions at each of the Gode Wind strains a value of = 1.0 is more appropriate (c.g ing = 0.5; however when approaching yield at larger amssd mea auod qte sos (sid) ss oogen turbine locations were investigated by Cone Pene- Porovic & Jardine 1994). To estimate the Go profiles measurements performed to 4060 m below scabed. (g-) profiles where used with the following correlation with depth below seabed the CPT cone resistance Some soil sampling and limited laboratory testing of the sands was also performed including particle size (Jardine et al. 2005): distributions and index tests. The basic geotechnical design profiles and param- G - 0.0203 0.001251.216×10*×n 9. eters for the locations considered were derived prin- cipally from the CPT records and correlations with where relative density and internal angle of shearing resis- lazy in the range of 3542*. tance. Internal angles of shearing resistance generally response of the sands it is necessary to account for In order to accurately model the stiffness-strain with P.
being the absolute atmospheric pressure the pre-yield variation of soil stiffness with strain level (100kPa) and beingthe free field vertical effective particularly at small strains and the dependency on the stress level. Consequently the sands were modelled stress. The variation of Gs with depth below seabed deter- as non-linear elastic perfectly plastic materials using is shown in Figure 3. Also shown are lines represent- mined from the CPT profile at both turbine locations a Mohr-Coulomb yield surface with non-associated plasticity conservatively assuming an angle of dilation ing costant values of (G/p) and the distributions equal to zero for all sand layers. of Gg with depth assumed in the FE analyses. It can The stiffness of sands at very small strain levels be seen that the variation in stiffness between the two (G) depends on the mean effective stress (pr) to the locations is relatively small. 737
rk-6470 are 00k=10 byb -12-16.2n leh -352-278n G3 FE 209 PY. *(0) l dain(4) a) Figure 4. Stiffness-strain curves for different depths and strata (Location Q03). - 463 py 903 FE 909 #4 The pre-yield constitutive model used in the analy- Modulus (E/pr) with deviatoric strain ina manner sim- ses describes the variation of the normalised Youngs ilar tothat proposed by Jardine etal. (1986) Itassumes a constant Poisson’s ratio of 0.25. Because the stiff- thanpdifferentstiffess-strainreltionships re ness in the model is normalised directly by p (rather used for different depths and soil layers resulting in 02% Disgte 0.% etat redio [5.0] ORN 164 the stepped profile of stiffness with depth shown in b) Figure 3. The stiffness-strain curves adopted for different depths and soil layers at location Q03 are shown in Figwe5. P-Y. Load-displacement curves; parison of FE and Figure 4. The shape of these curves was based on the idealised curve from an undrained triaxial test on North Sea sand from Dunkirk with an overconsoli- displacements of around 0.2D .e. 1.5m paring dation ratio of 2.0 (Kuwano 1999). The small strain the results obtained from the FE analyses with those shown in Figure 3 and all curves converge to a sim- stiffness values for each curve are based on the profile obtained from the modified p-y analysis. It can be seen that the predicted lateral load at these displacements ilar normalized stiffness for axial strains greater than forbothanalysis methodsis well inexcess ofthe design about 0.5%. ULS condition. rounding soil was modelled using speciallyformulated The interface between the monopile and the sur- For both locations the modified p-y analysis shows only small increases of lateral load at displacements interface elements (Day & Potts 1994) with a Mohr- approaching 0.2D while the load obtained from the Coulomb yield surface and non-associated plasticity. FE analyses are still increasing considerably. Further- Based on the average grain size (dso) obtained from more it should be noted that the assumption of zero the available grain size distribution curves and pub- spuuts o a s sosee 3d aq u moe lished correlations (API 2011 and Jardine et al. 2005) underestimate the ultimate capacity of monopiles in an angle of interface friction of 29° was considered to be appropriate for all strata. dense sands. As the measured behaviour of dense sands in the laboratory shows significant dilation it is reasonable to expect a substantial increase in the resis- Nonetheless for displacements ofaround 0.2D the FE tance offered by the soil particularly at the pile toe. 4 KEY RESULTS OF THE ANALYSES %0 aexdde e speo e a sasijeue 4.1 Load-displacement response - s mog paueqo speot oq ueq aou?q %o1 pue ‘0H pus go suogeso ao ssfeue &-d Bupuoxdsou The 3D FE analyses at locations Q03 and H03 were used to benchmark and validate the geotechnical respectively. Figure Sb) gives a more detailed view of the initial design carried out using the modified API p-y method- part of the load-displacement curves enabling a - ology in the serviceability and ultimate limit states. This was achieved by paring load-displacement parison of the FE and p-y analyses results in terms of serviceability conditions. It is clear that the FE anal.- and moment-rotation curves in terms of the ultimate eu uo ssuodsa sags Aeuesqns e ipad ssf capacity as well as the initial stiffness. sip jeoge sasjeue f-d popom a rs Bupeof Figure 5a) shows the normalised load-displacement curves at seabed level for both locations up to lateral placements obtained from the FE analyses at 10% of the ULS load are in the order of 50% of those 738
Horizontal displacement [%D] Horizontal displecement [%D] -0.4%-0.2%0.0%0.2%0.4%0.6%0.8%1.0% -10% 5% 0% 5% 0.0 99 10% %61 20% 0.1 0:1- 0:2 92 0.3 03 0:4 0:4 0.5- 8 05 0.6- 0.6- H03 0.5ULS 0:7 H03:2.0°ULS 0:7 - H03: 0.75°ULS H03: 4.5°ULS =H03 1.0ULS = H03: 6.0°ULS 08 003:0.5ULS Q03: 1.5°ULS 003: 0.75'ULS Q03: 2.5°ULS 003: 1.0ULS Q03:3.0′ULS ing failure at 0.2D seabed displacement. Figure 7. Horizontal pile displacememt profiles approach- initial stages of looding. Figure 6. Horizontal pile displacememt profiles during obtained from the modified p-y analyses indicating a suhstantially stiffer response of the monopiles under shows changing displacement profiles with increasing serviceability limit state loads. This agrees well with load levels. For small loads the monopiles are bending the findings from full-scale measurements reported kick). However as the load level increases the piles with only little backwards movement at the toe (toe is interesting to note that the initial part of the load. by Hald et al. (2009) and Kallehave et al. (2012). It increasingly rotate with increasing toe kick. For loca- displacement curve (up to aroumd 25% of the ULS tion Q03 toe kick mences for loads in excess of load) is very similar for both locations and does not HH03 toe kick only mences for loads in excess of approximately halfthe ULS load whereas for Location appear to be significantly influenced by the differ- ence in L/D ratio or soil profile. Given the relatively the ULS load. small areas of yield in the initial stages of loading the Foralload levels the influence ofthe L/D ratio can assumption in terms of the angle of dilation is not sig- the longer pile. However i is interesting to note tht be clearly seen with the shorter pile bending less than nificant in terms of the predicted pile response under operational load conditions. as failure is approached both monopiles rotate about a Very similar conclusions both in terms of the ulti- point at approximately 70% of their length. mate resistance and the serviceability conditions can The developing failure mechanism on the plane of be drawn for the moment-rotation behaviour. At the symmetry is illustrated in Figure 8 by the incremen- ULS loads the analyses predict otations at the seabe around 3 times the ULS loud. The overall lateral dis- tal displacement vectors for Location Q03 at a load of of approximately 0.3° and 0.25° for locations Q03 and H03 respectively. of 0.2D. Figure 8 indicates the following characteris placements at the seabed for this stage are in the order tic features of the likely ultimate failure mechanism in 4.2 Displacement profiles and failare mechanisms this plane: Normalised horizontal displacement profiles with nor- 1. Soil behind the pile “dropping" into the space malised depth below seabed for both locations are opened up by the forward movement of the pile sna pe g pe g sam u us 2. Soil in the upper part in front of the pile is pushed to significant depths; It should be noted that similar seabed displacements peo uagp an og poueqo ae suoexo oq o sidewzys and upwards towards the seabed; levels especially for larger loads. The figure clearly 3. At depth a rotational mechanism is developing. 739
