3DFINITEELEMENTANALYSISOFMONOPILESANDITSAPPLICATIONIN OFFSHOREWINDFARMDESIGN Angeliki Grammatikopoulou Geotechnical Consulting Group London UK 442075818348 a.grammatikopoulou@gcg.co.uk Felix C. Schroeder Geotechnical Consulting Group London UK 442075818348 f.c.schroeder@gcg.co.uk Giuseppe Pedone Imperial College London UK 447580527841 g.pedone13@imperial.ac.uk formerly Geotechnical Consulting Group London UK Amandine Brosse Geotechnical Consulting Group London UK 442074818348 a.brosse@gog.co.uk Torben Sarensen @rsted Denmark 4599552622 tosor@orsted.dk David M.G. Taborda Imperial College London UK 442075946033 d.taborda@imperial.ac.uk ABSTRACT The PISA Joint Industry Project has led to significant advances in monopile design with 3D Finite Element (FE) analysis playing a key role in the development of the method as well as design process for monopile foundations for an offshore wind farm in the North Sea. The paper is divided into three parts. The first part demonstrates the ability of the adopted constitutive models to describe the soil response for the soils encountered on the basis of high quality ground investigation campaigns for the site. The second part presents the calibration of the constitutive model used in the FE analyses to simulate the sands at different monopile locations across the wind farm in terms of load displacement curves and structural forces within the monopiles. Keywords: monopile foundations finite element analysis constitutive modelling INTRODUCTION Monopile foundation design for offshore Wind Turbine Generators (WTG) requires consideration of many aspects of behaviour.The recently pleted PISA Joint Industry Project has led to significant advances in monopile design (Byrne et al. 2019 and Burd et al. 2019) with 3D Finite element (FE) analysis playing a fundamental role in the development of the PISA method was based on sophisticated 3D FE analyses using the Imperial College finite element code ICFEP (Potts & Zdravkovic 1999). This included Class A predictions of field tests on monopiles subjected to lateral loading at two sites;i.e.Cowden a glacial clay till site (Zdravkovic et al. 2019) and Dunkirk a marine sand site (Taborda et al. 2019). s reaction curves which can then be utilised in simpler 1D Winkler-type beam-spring models models are essential and in order to calibrate the model parameters high quality ground investigation information bining in-situ and laboratory testing results is required. using the FE code ICFEP which was used in the development of the PISA method. The
paper is divided into three parts; the first part demonstrates the ability of the constitutive models adopted in the 3D FE analyses to describe the soil response for clay and sand dominated materials encountered in the offshore wind farm site.In the FE analyses presented in this paper a different constitutive model to the one used for the development of the PISA method was adopted for the modelling of the sand deposits. Therefore a results of the 3D FE analyses of monopiles at different locations across the windfarm.These include load displacement curves and structural forces developed within the monopiles. SOILCONDITIONS CONSTITUTIVEMODELSANDASSOCIATEDPARAMETERS In the offshore wind farm site considered in this paper the soil deposits encountered consist of i) stiff clays of low and high plasticity and ii) sands (in most locations encountered in a dense state). Clay deposits The low plasticity clay deposits were modlelled in the FE analyses with the same constitutive model adopted in the PlSA work for the stiff glacial clay till at Cowden (Zdravkovic et al. 2019). This is an enhanced version of the Modified Cam Clay model featuring a non-linear Hvorslev surface on the dry side a generalised shape for the yield and plastic potential on the IC.G3S model by Taborda et al. 2016. the basis of laboratory tests across the whole site (which included oedometer tests triaxial CAUc and CAUe tests bender element tests resonant columns tests). Figure 1a shows the stiffness variation with strain obtained from undrained triaxial tests with local instrumentation which allowed the measurement of stiffness from very small strains for one of these glacial clay tills. Also shown on this figure are the results of bender element (BE) and resonant (for an isotropic material E=3G). It should be noted that the bender element tests as well as the available tests are not sufficient to fully define this aspect of behavior of the material and hence in line with the PISA work isotropic stiffness was assumed. In-situ P-S logging tests were also performed in some boreholes. Results of these tests were scattered and s reliable and hence the stiffness strain curve adopted in the FE analyses was based on the laboratory tests (Figure 1a). Figure 1b also shows the results of triaxial and BE tests on this test on this glacial tilltogether with the results of a single element finite element simulation of this test; good parisons between predictions and measurements can be seen. 12000 TriatlCAU 12000 BE-36h WL * 10000 e-RC 10000 BE36vh ft idealtati Ft idslhation 8000 8000 3Gah-neerly site 6000 6000 TriasilCAue -nearby sht 0009 4000 2000 (a) 2000 (b) 0.0001 0 0.001 0.01 0.1 0.0001 0.001 0.01 es [%] [%]s3 0.1 Fig. 1. Stiffness strain response of stiff glacial clay till
0 0.5 0.2 D Axial strain (%) -0.5 s"gs ries (-) 5.9 700 Experimental Gata 8000 Eperimental data =FE sm/aioe 5000 =FE simleton 4000 ARg 000 1000 Axial strain (%) 0.001 0.01 Axial strain (%) 10 Fig. 2. Comparison of experimental results and simulated response for CAUc test on stiff glacial tillclay However in another clay deposit index properties tests indicated significant variability of the material encountered. Figure 3 shows profiles of index properties with depth and the the clay can be seen even within the same borehole (i.e. BH06). As expected the response of this clay in undrained triaxial tests was also found to be variable. In some cases the when sheared in a manner similar to the one shown in Figure 2.In other cases the clay showed a stress-strain response typical of a stiff plastic clay i.e. when sheared it showed a varied between around 3% and 18% in the tests which showed strain softening. However it should be noted that undrained triaxial tests reported by Grammatikopoulou et al. (2017) on a similar clay deposit of high plasticity depicted a much higher reduction of strength from peak to post-peak i.e. in the order of 30%. As Grammatikopoulou et al. (2017) noted from CPT tests it is not always clear whether the low plasticity clay. As such it is important that laboratory tests are undertaken and the laboratory testing a conservative approach would need to be adopted. In the FE analyses presented here this clay was modelled using a Mohr-Coulomb type Tresca model (similar to Grammatikopoulou et al. 2017) but bined with a non-linear model which uses the same hyperbolic expression for the tangent shear modulus as the one used in the variant of the MCC model discussed above. Figure 4 presents the results of a single element finite element simulation. As discussed in Grammatikopoulou et al. (2017) the Tresca model provides a conservative representation of the stress-strain response although
it cannot reproduce the measured stress path and pore pressure response. It should be noted that the original PISA study considered the modelling of the glacial till deposit of low plasticity at Cowden as discussed above. The modelling of stiff plastic clays has subsequently been addressed by the PISA2 JIP which used a more sophisticated model than Tresca. Moisture Content Plastic Lime Liquid Limit [%] 10- 0 40 60 80 100 20- 70- 60- BH01 50- ees 30 A BH02 H High 0 BH03 BH04 0c BH01 BH05 BH02 E BH06 20- 10 BH06 10 20 40 50 Liquid Limat [%] 70 80 90 02011001 50 (b) 60 (a) 2.5 1.5 =FE sit 0.5 Axial strain (%) 10. 15 5′0 1.5 (-) 2.5 1200 000 800 600 400 200 1.5 0.000 0:081 Axial strain (%) Axial strain (%) 0.01 0.1 Fig. 4. Comparison of experimental and simulated response for CAUc test on stiff plastic clay
Sanddeposits In the PiSA work the sands were modelled with an advanced bounding surface plasticity model which can account for the effects of both stress level and void ratio as well as a number of other features of sand behavior as described in Taborda et al. (2019). In the FE softening variation of the Mohr-Coulomb model in which the angles of shearing resistance dilation and cohesion intercept (if necessary) are allowed to vary with strains as described by Potts and Zdravkovic (1999) and discussed in Grammatikopoulou et al. (2017). It should be noted that the peak strength and dilation of a sand will depend on the stress state and density of the material. Hence when using this model in the FE analyses the input parameters will have to vary to account for different in situ densities; unlike the bounding surface model used in the PISA work which is capable of reproducing the sand response under a wide range of states using a single set of parameters. This strain-softening Mohr-Coulomb elasto-plastic model was bined in the FE analyses with the non-linear elastic model IC.G3S (Taborda et al. 2016). The latter allows the simulation of a variety of features of sand response including non-linearity from early stages important for sands. It is generally accepted that the stiffness at very small strains (Go) depends on the mean effective stress (p) to the power of n where n = 0.5 is reasonably representative for sands. The formulation of the IC.G3S model allows the modelling of the dependence of G on p as well as on the sand density. The parameters of the IC.G3S model which control the stifness G at small strains were chosen for each sand deposit in bination with the in situ relative density D so as to reproduce the Rix & Stokoe (1992) Go interpretation of the CPT profiles for all monopile locations considered. Figure 5 shows an assumed for that location and Figure 5b the corresponding Go modelled by the IC.G3S model for the chosen model parameters (in this figure Go is normalized by the maximum value in the idealised profile). Figure 6 shows a simulation of a drained triaxial pression test on a sample from this dense sand (tested at p'=140kPa Dr~100%) using the strain softening Mohr Coulomb model in bination with the non-linear IC.G3S model. Relatve dersty [N| 0.5 Normalised Ge 20 40 60 100 120 15 Sand1 Sand1 Dr saturated (lamiokowski et al 2008) Dr (Baidi et al 1986) Clay1 Cay1 10 Sand2 Send 2 aay？
Chy 15 Sand2 30 35 Fig. 5. CPT interpretation of a) Dr and b) Go and idealised response for sand deposits (a) (q)
