Missouri University of Science and Technology

Scholars' Mine
International Conference on Case Histories in (2013) - Seventh International Conference on Case
Geotechnical Engineering Histories in Geotechnical Engineering

Apr 29th - May 10th

Probabilistic Three-Dimensional Model of an
Offshore Monopile Foundation: Reliability Based
Approach
Aalborg University, Denmark

Lars Vabbersgaard Andersen
Aalborg University, Denmark

Lars Bo Ibsen
Aalborg University, Denmark

J. Clausen
Aalborg University, Denmark

J. D. Sørensen
Aalborg University, Denmark

Part of the Geotechnical Engineering Commons

Recommended Citation
M. J. Vahdatirad, Lars Vabbersgaard Andersen, Lars Bo Ibsen, J. Clausen, and J. D. Sørensen, "Probabilistic Three-Dimensional Model
of an Offshore Monopile Foundation: Reliability Based Approach" (April 29, 2013). International Conference on Case Histories in
Geotechnical Engineering. Paper 7.
http://scholarsmine.mst.edu/icchge/7icchge/session08/7

This Article - Conference proceedings is brought to you for free and open access by Scholars' Mine. It has been accepted for inclusion in International
Conference on Case Histories in Geotechnical Engineering by an authorized administrator of Scholars' Mine. This work is protected by U. S. Copyright
scholarsmine@mst.edu.

partial or total safety factors in are usually not accounted for.g. These uncertainties a stochastic design approach. a reliability-based design process based on stochastic analysis of the soil parameters is proposed to obtain an efficient design with known reliability and smaller costs for tests and construction. Reliability indices of the monopile are obtained through an advanced reliability method and a probabilistic procedure is proposed regarding the 3D design of monopile foundations.g. It can also be noted that applying techniques and/or modeling procedures. Denmark 9000 Aalborg. This is the strategy which is typically utilized in the deterministic Several studies were conducted for developing stochastic design methodologies in the current design codes. e. A three-dimensional (3D) finite-element model is established and a stochastic model for the undrained shear strength of the soil is proposed using random field theory. deterministic design. Clausen J. the 5% quantile value of the undrained shear strength of the soil.09a 1 . of Civil Engineering.g. but the reliability of the foundation is unknown. expensive geotechnical in-situ and laboratory tests are placed on the soil surface was predicted analytically and conducted to estimate soil properties. Denmark 9000 Aalborg.PROBABILISTIC THREE-DIMENSIONAL MODEL OF AN OFFSHORE MONOPILE FOUNDATION: RELIABILITY BASED APPROACH M. Denmark J. of Civil Engineering. The Mohr–Coulomb constitutive model is used to model the soil behavior. a wind farm). In this models of foundations. Ibsen Dept. Denmark 9000 Aalborg. INTRODUCTION Designing offshore wind turbine foundations concerns several investigation can be cheaper than individual testing for each uncertainties due to material properties. over-consolidated clay is considered as an example. Denmark ABSTRACT When wind turbines are to be installed offshore. Aalborg Dept. Sørensen Dept.B. 8. Hence... 2003). or total safety factor on the resistance and/or on the loads. statistical and model uncertainties are not taken into account in code-based.V. by a reliability-based (2007) also studied the effect of soil spatial variability on the design procedure. Fenton and Griffiths unknown in this procedure. Furthermore. Andersen L. current methodology based design may be expensive.. Typically. or they are neglected by the deterministic design can be calibrated or modified and introducing either partial safety factors on material properties used in future designs. Aalborg Dept. but only deterministic verified via Monte Carlo simulation (MCS) when considering values (e. of Civil Engineering. the reliability of the structure remains of the soil (Fenton and Griffiths. of Civil Engineering. Aalborg Dept. In this study a monopile foundation in undrained. In a Paper No. expensive geotechnical in-situ tests are carried out at the location of each turbine and only a quantile value (typically the 5% quantile) of the measured strength parameters is used as design parameter. Instead.D. this can be cost (2011) proposed a reliability-based design procedure for effective using stochastic parameters of uncertain properties estimating the first natural frequency of an offshore wind which are already estimated through an optimized field turbine founded on a monopile. Furthermore.J. Aalborg University University 9000 Aalborg. measurement. a 5 percent quantile value) of them are used for spatially random fields for the cohesion and the friction angle design. of Civil Engineering. They applied a random field investigation for the whole region (e. measurement wind turbine in a wind farm. accounted for in a rational way. Vahdatirad L. Aalborg University University University 9000 Aalborg. Instead. Andersen et al. a design is obtained where uncertainties are settlement and ultimate load statistics of a pile. This model for the undrained shear strength of clayey soil. The bearing capacity of a footing regard.

(m) (m) Embedded Free properties 3. In this regard. Using this subroutine. An as proposed by Kellezi and Hansen (2003).06 35.5 m above the monopile cap with stiffness and applied an Asymptotic Sampling (AS) method to a horizontal direction. They used a semi-analytical model in monopile elements with a submerged unit weight are applied combination with Monte Carlo Simulation (MCS) for in a consolidation step. foundation in undrained. Ideally. This random field is used as Continuum 8-node solid elements (C3D8) were used for soil soil cohesion in the Mohr-Coulomb constitutive model. a free length above the soil turbines in the North Sea.81 m/s2 are used for study. This is implemented by a user- defined material subroutine (UMAT) which has been written in Fortran. This is ensured by the value of 52  10 N . a linear correlation model the bending along the pile. Three failure modes are the real situation. 2006). They at the pile cap. the linear relationship is applied as an approximation. Finally. the gravity loads of wind turbine. material random properties can be defined as solution-dependent state variables (SDV) in each integration point of the soil element. Furthermore. which is not corresponding to offshore standard (DNV. The undrained shear strength ( Cu ) of clayey soil is modeled by a LogNormal random field. Geometrical and material properties of the monopile “Model for limit state and design equations”). 2007). a cross-correlation should be monopile. A reliability analysis is performed for these failure states by means of the Outer radius Thickness Length (m) Geometrical AS method. a 3D finite-element model for a monopile 6 achieved. It is noted that a water density of 1000 estimating the distribution of the footing stiffness.05×1011 7872 0. Vahdatirad et al. as well as Abdel. (2012) estimated the stochastic calculating the submerged unit weights. The tangential behavior with a friction coefficient applied between Cu and E0 (Fenton and Griffiths. over-consolidated clay is developed and utilized as computational model. As shown in this table.00 0. A tie constraint was E0  k s Cu where k s  200 is the coefficient for over- used between the monopile and the soil elements inside the consolidated clay. A master-slave concept was between Cu and soil initial elastic modulus E0 is assumed as used for interaction between the monopile and the surrounding soil (Abdel-Rahman and Achmus. The amount of this load must be estimate rare events of the monopile stiffness.67 was applied for modeling the frictional behavior between the monopile and the surrounding soil. Furthermore. Afterwards. increasing trend over the depth is considered for the mean Rahman and Achmus (2006). It is assumed that the wind force is dominating considered a nonlinear p-y curve for the modeling of the soil and applied at a height of 61. The rotation at the pile cap is layer is considered for the monopile. an advanced reliability method was proposed to estimate rare events of the A stepwise execution is conducted for the finite element first natural frequency of an offshore monopile foundation. 8. An elastic–perfectly plastic Mohr-Coulomb constitutive model is used for the soil behavior.29 has the advantage that parametric analysis can be performed and used in the reliability assessment.50 COMPUTATIONAL MODEL Elastic Density Material modulus Poisson’s ratio 2 (kg/m3) A 3D finite element model has been constructed in the Abaqus properties (N/m ) numerical package by scripting in Python. considered: a serviceability limit state. analysis. 2003).09a 2 . considered large enough such that the lateral deformations plastify the soil completely and a full failure mechanism is In the present study. This prevents the soil to considered as a representative failure mode according to the go over the pile during failure. (2011) studied the application of a stochastic generation of the initial stress state using soil elements having dynamic stiffness model for a surface footing for an offshore a submerged unit weight only.00 4. (2012). but of 0. In another kg/m3 and gravitational acceleration of 9. the linear pressure-overclosure relationship with a contact 10 2 stiffness of 10 N/m was introduced in order to model the normal behavior at the interaction. Incompatible-mode 8-node solid elements (C3D8I) were used for the monopile in order to value of Cu (see Table 2). the geostatic step is first performed for Vahdatirad et al. Paper No.similar study by Andersen et al. a combination stiffness of a laterally loaded offshore monopile modeled by a of shear and bending moment is considered as external loads one-dimensional Finite Element Method (FEM) model. an ultimate limit state and a fully established failure in soil material (see the section Table 1. The geometrical and material properties of the monopile are close to the real site Table 1 shows the geometrical and material properties of the conditions for monopile foundations for large offshore wind monopile. Scripting in Python 2.

Figure 1 presents a realization of the random field for Cu . This method was originally proposed by Matheron (1973) and can be used for generation of realizations of a random field in a three-dimensional space by using a sequence of one-dimensional processes along lines crossing the domain. Mapping of the three-dimensional random field for A Matlab script has been developed for generating the random Cu in the applied finite-element model. y and z are spatial distances in the x . E0  200Cu E0 (N/m2) Paper No.499 boundaries. (degree) Dilation angle 8.5×105+2000z depth in meter) Mean value.40 Lognormally distributed undrained shear strength. Cu (N/m2) failure state for the same realization as in Fig.50 coefficient pressure 1. some parts close to the monopile and inside the failure region are not plastified. Density (kg/m3) COV 2200 Figure 2 illustrates plastic strains around the monopile at the 0.00 Elastic modulus. (degree) Friction angle 8. 8. 2. Further. 2006):  x y z    exp    (    ) (1)  x y z    where x . y and z directions. 1. Horizontal.00 0.09a 3 . The variables are saved as SDV and mapped on each integration point of a soil element by the UMAT Table 2. As shown in Fig. Deterministic and stochastic properties of the soil subroutine during the analysis. respectively. 1. An exponential 3D correlation function (  ) is used as proposed in (JCSS. field by TBM. which are representing the stronger area having higher values of Cu .01 δx Correlation length (m) Horizontal.  x   y is the correlation length in the horizontal directions and  z is the correlation length in the depth direction (see Table 2). For generating the random field. This example shows that a fully developed failure mechanism is obtained due to the large lateral deformations. The boundaries of the 1. Fig. the turning bands method (TBM) is utilized. δz Lateral earth 2. whereas the Stochastic properties properties white regions represent the weaker parts. The black regions in this figure Deterministic show the stronger parts with higher value of Cu . μ Poisson’s ratio computational domain are placed far enough away from the (z is layer pile inasmuch as there are no plastic strains near the (N/m2) 0. The deterministic and stochastic soil properties are shown in Table 2.01 δy Depth.00 0.

Fig. L with considered coefficient of variation (see Table 3). see table 4. namely the load P and the resistance or load  Lognormal . 0.00 0.00 0. Stochastic models for physical.  1.10 Mean This function is defined such that positive values of g Xstr Lognormal 1. W  (3) factor for the load effect. see below.15 98% Xdyn Lognormal 1. 2010): where Cuc is the characteristic value of undrained shear P  X dyn X exp X aero X str L (4) strength.05 Mean g Y P (2) Xexp Lognormal 1. Y is assumed to be assessed by the following model: be obtained from the FEM response by applying characteristic values of material parameters and  f is the partial safety Y   R  X. 8. and L is uncertainty related to the extreme load-effect due to wind loads.03 Mean correspond to safe states and negative values correspond to failure states. Partial safety factors for design equations land space topography.  m 1.  accounts for the model uncertainty.. (4) are shown in Table 3. see Table 3.  f 1.  R 1. X aero accounts for uncertainty in (partly based on IEC 61400-1. see Table 4. given as: L Weibull . 2005) assessment of lift and drag coefficients. see Table 2.50 5% bearing capacity Y. Partial safety factor for load resistance. X exp models the uncertainty related to the modeling of the exposure such as the terrain roughness and the Table 4. its characteristic value Lc is determined such that the following 2. 0. The load P and the resistance Y are supposed to where Yd is the design value of the load resistance which can be functions of relevant uncertainties. Three possibilities are considered to obtain the design value of the load bearing where X is the vector of random variables modeling soil capacity: strength parameters (here the undrained shear strength of clayey soil).  Cuc  Yd   c R  . Table 3. W is a set of deterministic parameters such as 1. simple load model is assumed to consist of  m  several uncertainties (see. The proposed statistical parameters for the uncertainties in Eq. including uncertainty in damping ratios and conversion factor.50 5% basic variables.00 0. accounting for bias in the model R () . W (6) A representative. e. The uncertainties in this study are assumed Partial safety factors for material properties. 2. Model two: Yd is determined from the characteristic value design equation is fulfilled: of load bearing capacity applying a partial safety factor for resistance: Yd   f Lc  0 (5) Yc Yd   (7) R Paper No.g. Sørensen & Toft.  c is the characteristic where X dyn accounts for uncertainty related to modeling of the value of the model uncertainty  in table 3.00 0. and  is a dynamic response.00* (IEC 61400-1.35 load. R () of soil strength parameter applying partial safety factors for material properties: represents the model for the load resistance which in this paper is represented by the FEM model described above. Finally. 0. Model one: Yd is determined using the characteristic value monopile properties or deterministic soil properties.09a 4 .  m is the partial safety factor for the material parameter. natural frequencies. model and statistical uncertainties MODEL FOR LIMIT STATE AND DESIGN EQUATIONS Variable Distribution Mean COV Quantile A generic form of a limit state function g is defined by two R Lognormal . Plastic strains at fully developed failure mechanism.3 to be representative for normal operation of wind turbines Conversion factor. 2005).20 Mean Xaero Gumbel 1. In this study.3 To obtain the distribution of the annual maximum load effect * Corresponding to no conversion (hidden) in the models. X str is uncertainty Variable Value related to the computation of the load-effects-given external Partial safety factor for load effect.

09a 5 . estimating low first passage probabilities of high-dimensional nonlinear systems (Sichani et. Then.0. For structural components. The implemented procedure and more details can be These levels are expressed as: found in (Bucher. 2009). the probability of failure can be expressed as: RELIABILITY ANALYSIS Pf  (  (1)) (13) Considering the limit state function given in Eq. leading to 2400 realizations in total. 2007). W  The basic idea of AS is to generate more simulations in the Yd   (8) target region (the failure domain) by increasing the excitation R power (Bucher. asymptotic sampling (AS) is applied to estimate the P . W   X dyn X exp X aero X str L is estimated. Therefore. where Yc is the characteristic value of the resistance Y (2009). If crude Monte Carlo simulation each realization of the limit state function. Then. and the probability of failure for each level of failure  (1)  A  B (12) is estimated.6. cf. 3. (3) is obtained by means of the FEM obtain a coefficient of variation of 0.25 degrees. In f factor.  (1) represents the un-scaled reliability index at (9) failure (Bucher. the reliability index at failure can the soil completely (total collapse of the soil is achieved). Hence. (2) or Eq. 0. the standard deviations of the random variables are increased artificially by the factor Based on the above models. Ultimate limit state (ULS) where the rotation of the f f 2 monopile cap is limited to 3 degrees. Model three: Yd is determined from the characteristic first natural frequency of an offshore wind turbine founded on a monopile. ( f ) B  A (11) 2. Serviceability limit state (SLS) where the rotation of the monopile cap is limited to 0. Having realizations of the resistance Y and the load this study. Hence. For this reason. a maximum annual probability of failure of the curve fitting by Eq.4 and 0. Andersen et al. fitting. (9). Asymptotic sampling was utilized as an efficient method for and  R is the partial safety factor for the resistance. Vahdatirad et al. a realization of the limit state value g can be determined probability of failure and the corresponding reliability index. (3) and quantile values of R () and  . a representative limit state of 1 / f to scale the results into the failure region. the function g can be written: scaled reliability index  ( f ) corresponding to scaled results g   R  X. 0. 0. the fitting equation proposed by Bucher (2009) is used: 1.3 for the probability and the f factor used for increasing the standard deviation of estimate.8. proposed for high-dimensional reliability analysis by Bucher Paper No. AS is an advanced Monte Carlo simulation method originally from Eq. Several values of the f factor were considered. three limit states for failure are considered based on the rotation of pile cap (DNV.  c R  Cuc . Pf  P( g  0) (10) 0. Simulation of this amount of realizations implies high computational cost inasmuch as one realization takes random variables (here Cu ). 400 realizations are made at each 3 4 order 10 to 10 is required for critical wind turbine value of the f factor. (2012) value of the random variable applying a partial safety proposed an improved AS method to estimate the stochastic factor for the resistance: stiffness of a monopile foundation by the FEM. (2011a) developed this method for high- dimensional dynamics problems such as wind turbines.al. (11). Andersen et al. Sichani et al. Fully developed failure limit state (FLS) where the lateral where A and B are coefficients which are determined through deformations of the pile are sufficiently large to plastify a regression analysis. application of advanced reliability the load P in Eq. Table 4. 10 to 10 realizations are needed to the resistance Y in Eq.3. 2011b). A corresponding realization of around 15 minutes. be estimated as: These failure criterions are considered through the reliability analysis. Herein. (9). 2009. (4) is obtained be simulation using the same methods is required such that fewer realizations are needed. 8. (2012) applied this method to estimation of rare events of the 3. Choosing the f factor is related to the desired probability level and having enough points for the Typically.5. the annual probability of failure can be written: where  is the standardized Gaussian distribution function. one realization of 4 5 (CMCS) is applied. obtained from Eq. including: 1. 2012). this relationship enables an estimation of  (1) by extrapolation techniques and curve In the present study. 2009). In another study..

5 ƒ factor in Asymptotic sampling method f=0. figure.Figure 3 shows ascending sorted values of the limit state function versus number of realizations. and the smallest one is obtained by the SLS. The corresponding expected) for all models is the largest..4 -6 f=0. (1) 6 Failure limit state.6 0 0. Corresponding reliability index to ƒ.1 0.5 0. FLS 4 10 Asymptotic fitted curve at FLS Limit state value. Paper No.6 0. SLS Fig. (13). Ascending sorted limit state values (g) versus number 6 Asymptotic fitted curve at SLS of realizations.. g=0 0 4 f=1 2 -2 f=0.3 0. f 2   f   1   1  2 Scaled reliability  1    ( f2 )  index. the un-scaled reliability index  (1) and the probability increases by decreasing the f factor.3 0. the number of samples in the failure domain ( g  0 ) Next.6 0.3 Fig. . g (N) 8 2 Scaled reliability index. a system of equations is constructed as: 4  1    ( f1 )  3 1.  (ƒi) Asymptotic fitted curve at ULS 5 Applying the reliability indices obtained from Eq..2 0. ULS Reliability index. 1 fi  ( fi )   (1  Pf ) (15) 16 6 x 10 Corresponding reliability index to ƒ.otherwise illustrate the reliability indices of the monopile using the three m design equations and three levels of failure defined by limit states SLS. the coefficients A and B are determined.9 1 -4 f=0.4 0. Number of realizations 7 Corresponding reliability index to ƒ.2 0.7 0. ULS Reliability index. (12) and Eq. 2. 4.8 0 f=0. (1) 1.8 0. . ULS and FLS. the reliability index  (1) using the FLS definition (as realizations for an f factor of fi . i  1.  (1) as well as reliability indices corresponding to different is the considered f factor and g m is the mth sample of f factors are illustrated in these figures. Figures 4 to 6 Pf   Im . (15) into Corresponding reliability index to ƒ. 8. (11). f 2    A   f  1  2   B   2  (16) 0 .4 0. fi .7 0.9 1 ƒ factor in Asymptotic sampling method  1    ( fn )  1. As shown in this Solving Eq.   .. 1 N  I m  1 if g f i  0 m Three models of the design equation presented in the previous fi section were used in the reliability analysis. FLS Asymptotic fitted curve at FLS Eq. (16). As shown in these fi figures. The related probability of failure Pf are estimated through Eq. of failure to each f can be estimated as: respectively. 3. This is in agreement with the design reliability index  ( fi ) is determined as: concepts inasmuch as the probability of failure in the FLS must be less than those using the ULS or SLS.  0 0.5 0.8 0.  (ƒi) 12 Asymptotic fitted curve at ULS Corresponding reliability index to ƒ. 5. 2     fn   fn  Fig. Reliability indices by AS method for three levels of 0 50 100 150 200 250 300 350 400 failure—design equation based on model 1.1 0. n .09a 6 .  (14) N m1 I  0. Reliability indices by AS method for three levels of failure—design equation based on model 2. SLS 14 Asymptotic fitted curve at SLS 6 Corresponding reliability index to ƒ. The AS fitted curves for finding where N  400 is the number of realizations.

pp.2×10-3 5. Reliability index (  ) 3.V. “Asymptotic sampling for high- ULS dimensional reliability analysis”.V. A stochastic 3D finite element model was developed for undrained.6 4. B.. Symp.8×10-4 1. ƒ factor in Asymptotic sampling method ACKNOWLEDGEMENTS Fig. J. Three design 6 Scaled reliability equations were proposed for reliability analysis at three levels index.9×10-5 failure ( Pf ) Reliability DNV-OS-J101 [2007]. ASCE. 24. Australia.3 0. Sørensen [2011]. Reliability indices by AS method for three levels of failure—design equation based on model 3. This illustrates the REFERENCES importance of choosing the model to obtain design values in a deterministic approach.5×10-4 5.T.3×10-6 Fenton. “Numerical factors in the design equation. Engrg. 54-65. 575-581.J. Civil- modes 1 2 3 Comp press. and D. Probability of No. over-consolidated clay. No. These variables Reliability index.1×10-5 failure ( Pf ) Approach”.D. ULS and FLS. No.4 index (  ) Structures”.e. Tsompanakis. No. ULS the random variables in the 3D random field.9 1 resulting in more optimized designs. “Wind Turbines Part 1: Design Paper No. 40(1). Thirteenth Intern. (1) of failure for SLS. 4 Furthermore. Andersen. 2. Appl.2×10-4 7. Failure Design models Engrg Compig. M.. “Bearing failure ( Pf ) Capacity Prediction of Spatially Random c   Soils”. Achmus [2006]. index (  ) SLS Sørensen [2012].. Proc. Reliability indices and probability of failures for the “Reliability-Based Assessment of the Natural Frequency of an three failure modes Offshore Wind Turbine Founded on a Monopile”. CONCLUSIONS Fenton. foundation. Perth.H. FLS were mapped on each integration points of the soil elements Asymptotic fitted curve at FLS by a user defined subroutine in Fortran. Table 5. in A reliability analysis was performed for an offshore monopile Geotech.0 3. pp. M.8 0. Reliability 3. Geotech. 504-510. Denver. “Natural Frequencies of Wind Turbines on Probability of Monopile Foundations in Clayey Soils: a Probabilistic 1. and D. how to apply the partial safety Abdel-Rahman. Vahdatirad and J. Gordon A.V. Geotech. thus 0 0 0.9 3. Scotland.6 0.. The authors kindly acknowledge financial support from the Danish Council for Strategic Research within the program Values of the reliability indices and probabilities of failure for “Reliability-based analysis applied for reduction of cost of the different design models are shown in Table 5. “Design of Offshore Wind Turbine 3. CO. and Environl. 170. Gordon A. 1-11. Stirlingshire. Griffiths [2007]. Comput. 6. C.7 0. Equivalently. The asymptotic sampling method was used for performing the reliability analysis. reliability is obtained. 8. Modeling of the Combined Axial and Lateral Loading of based procedure can be used to calibrate/modify the partial Vertical Piles”. Structrl.3 3. The 10 Asymptotic fitted curve at SLS turning-bands method was utilized to generate realizations of Corresponding reliability index to ƒ.6 4. The reliability-based procedure can 2 be used to calibrate/modify the partial safety factor for the soil properties such that a given target reliability is obtained. 43. the design equation for model 3 results in the most conservative results. [2009]. this table.09a 7 .D. Topping and Y. Intern. L. FLS Probability of 5. Probabilistic Eng. design model 3 results in smaller probabilities of failure compared to the other models. pp.V. Conf..2×10-4 1.5 0.5 3. paper 83.1 Bucher. Sichani and J.8×10-5 1. As shown in energy for offshore wind turbines”. the reliability.J. undrained shear strength of the soil was considered as 12 uncertain having a lognormal distribution based on the Corresponding reliability index to ƒ. Proc. and M. “Reliability- Based Deep Foundation Design”.8 Andersen. Can. Probab.4 0..1 0..  (ƒ i) Asymptotic fitted curve at ULS 8 Corresponding reliability index to ƒ. M.V. Vahdatirad. Proc. SLS concept of random field theory and spatial variation. pp. on Frontiers in Offshore safety factor for the soil properties such that a given target Geotech. on Civil. K.2 0. L. i. Mech. The IEC 61400-1 [2005]. Griffiths [2003].

J. Probabilistic Eng. Hansen [2003]. Conf. Inst. Stirlingshire. “Efficient estimation of first passage probability of high- dimensional nonlinear systems”. and Design Sichani. L. Bucher [2011b]. Engrg Compig.. Proc. pp. Saf. No (3).7: Soil 26. Sørensen. 468. “Section 3. B. M.R. on Testig. Method for Deep Found. Toft [2011]. pp. “The Dynamic Stiffness of Surface Footings Matheron.09a 8 . and Environl. Thirteenth Intern. Analysis of an Offshore Mono-pile Windmill Foundation”. [1973]. “Applications of Asymptotic Sampling on High Dimensional Vahdatirad. L. and P. Bayat. Nielsen and C. pp. Stochastic Finite Element Stiffness of a Laterally Loaded Monopile”.T.V. Ibsen Structural Dynamic Problems”. Revised Version. “Probabilistic Kellezi. Paper No..D. Their Applications”. on Civil. No. Properties”. 439. Civil- Comp press. Denmark. “Static and Dynamic design of wind turbines”.T. M.K. 33 (4-5). Andersen and L. G. Probab. Clausen and J.V. Sørensen [2011]. “An Improved Asymptotic Sampling Approach for 305–316. Scotland. Andersen.Requirements”. Kanasawa. .K.. 8. Mech. pp. Vahdatirad.. Ninth Intern. Lyngby. Tsompanakis. Structrl. J. Struct. Topping and Y. 539–549. S. J. Sichani. 5.. paper 79. “The Intrinsic Random Functions and for Offshore Wind Turbines: Reliability Based Assessment”.. M.V. Adv in Appl. GEO . paper 82. Nielsen and C.B.. L. S.H. and Henrik S.D. [2012]. 241–257. No. Energies.Danish Geotech. No.B.J. JCSS probabilistic model code [2006]. Bucher [2011a]. M. 3rd Edition. Conf. Proc..R. Japan. M.