Future XL monopile

foundation design for a 10

MW wind turbine in deep
water

K.W. Hermans
J.M. Peeringa

December 2016
ECN-E--16-069
Acknowledgement

The author is grateful for the budget that was made available to carry out this study as
part of the knowledge finance of ECN.

‘Although the information contained in this report is derived from reliable sources and reasonable care
has been taken in the compiling of this report, ECN cannot be held responsible by the user for any
errors, inaccuracies and/or omissions contained therein, regardless of the cause, nor can ECN be held
responsible for any damages that may result therefrom. Any use that is made of the information
contained in this report and decisions made by the user on the basis of this information are for the
account and risk of the user. In no event shall ECN, its managers, directors and/or employees have any
liability for indirect, non-material or consequential damages, including loss of profit or revenue and loss
of contracts or orders.’
 Contents
Summary 4

1 Introduction 5

2 Preliminary design 7
2.1 Proposal design 7
2.2 Foundation stiffness 10
2.3 Natural frequency check 10
2.4 Extreme load check (buckling) 11
2.5 Fatigue check 18

3 Limitations of current models 25

3.1 Fatigue model 32
3.2 Structural model 25
3.3 Soil structure interaction 25
3.4 Hydrodynamic models: diffraction considered McCamy Fuchs 28

4 Conclusion 34

3
 Summary
The monopile foundation is the dominating support structure type for offshore wind
farms in construction today. It is anticipated that also in future wind farms the monopile
will remain a preferred choice due to the manufacturing speed and installation
experience. However, larger, heavier turbines are being developed and wind farms are
being installed in deeper waters, testing the technical and economic limits of monopile
foundations against other foundation types.

This report contains the results of an ECN internal knowledge project on future large
monopile foundations. One of the main objectives was to create a preliminary design
for a 10 MW support structure in 50 m water depth. This is a degree larger than the
current designs in the industry and is aimed to put our current models to the test. The
basic design features a foundation pile diameter of 9.3 m and a total steel mass above
mudline of 1850 tons. The design has not been optimized but was subjected to an
ultimate load, buckling and fatigue check. A global buckling check for a buckling failure
mode twice the length of the support structure was not passed and the applicability and
implications of this failure mode must be further investigated. The basic design passed
all other ultimate load checks and fatigue check at the mudline. Largest diameter in the
support structure is 9.3 m which is within manufacturing capabilities.

The second objective was to identify the limitations of current design methods and
simulation tools. The soil-structure interaction can be modelled by lateral springs (PY
curves) with the current tools available to ECN. However, for large diameter
foundations other physical load bearing mechanisms are at play and the PY curves do
not suffice for accurate soil-structure modelling. For a 9.3 m monopile, modelling the
hydrodynamic loading by diffraction does not have a very large influence on the
magnitude of the loads, hence Morison’s equation may still be used without being too
conservative. Finally, there are assumptions made in evaluating the fatigue damage by
using an SN curve and Miner’s rule.

4
 1
Introduction
Ever since the first offshore wind farms were built, the monopile support structure has
been a solid foundation choice. From a manufacturing standpoint the monopile is the
easiest to fabricate and in optimizing for example welding processes tighter tolerances
could be achieved and production volume could be increased. As the turbine size is ever
increasing and offshore wind farms are placed in deeper waters, monopile diameters
have been growing to accomplish the right stiffness. The limits of applicability of
monopile support structures have been shifting. Not long ago, water depth more than
25 meters or 5+ MW class turbines were considered the tipping point towards jackets
or other alternative support structures. Nowadays, ‘XL monopiles’ of over 1000 tons are
produced with diameters up to 8 meters for water depths up to 30-35 meters (Figure
1). This leads to the questions: ‘What are really the limits of monopile support
structures’? And can we design a monopile for a future scenario of a 10 MW turbine at
a site of 50m water depth?

In this report a simplified design methodology is set up and applied for a basic design
for a 10 MW turbine at 50 m water depth. Ultimate loads, buckling and fatigue checks
are performed, but no optimization or detailed design is carried out. Finally, an
evaluation is made of the current simulation tools at use in ECN. Are they still valid for
large diameter monopiles?

5
 Figure 1: The world’s heaviest monopile for Veja Mate offshore wind farm [1]

6
 2
Preliminary design

2.1 Proposal design

The proposed preliminary design is drafted by taking inspiration of current designs in

the waters. A conical section is incorporated in the monopile to reduce wave loads in
the splash zone and account for the increase in bending moment towards the mudline.
The taper angle, height at which the conical section should start and diameter at
mudline are taken as design inputs. In Table 1 and Table 2 the geometric design
parameters are given and the resulting preliminary design is sketched in Figure 3. The
overall design methodology is presented in the scheme of Figure 2.

Figure 2: Design methodology

Preliminary design 7
 Table 1: Geometry variables used for 10 MW design rev 00

Description value

Outer diameter at mudline 9.6 m

Height start taper MP above mudline 10 m
Length taper MP / taper angle MP 23.56 m / 2.5°
Top MP above mudline (bottom TP) 53.22 m
Platform height (top TP) 76.16 m
Height start taper tower above platform 10 m
Length taper tower / taper angle tower 47.96 m / 1.98°

Table 2: Diameter over thickness ratios used for 10 MW design rev 00

Description value

D/t Monopile 100

D/t Transition Piece 80
D/t tower 140

Table 3: Mass for preliminary 10 MW design rev 00

Description value

Mass of support structure steel (monopile above 1849 tons

mudline, tower, transition piece)
Mass of internal water 2774 tons

8
Figure 3: Preliminary design rev 00

9
 2.2 Foundation stiffness

In this first design iteration we limit ourselves to the design of the pile above the
mudline. However in order to model the structural dynamics right, the foundation
stiffness (and ideally also damping) must be taken into account. The soil layer
properties of OC3 have been used, as reported by Passon [2]. The foundation model in
Phatas includes a coupled lateral and rotation spring matrix. Recently, the Phatas soil-
structure interaction model has been improved by adding a PY module. This was not
used for creating the basic design, but the potential of the additional functionality is
elaborated upon in section 3.2.

2.3 Natural frequency check

The 10 MW turbine has an operational range of 6-9.6 rpm. Hence, the 1P frequency
band is from 0.10-0.16 Hz and the 3P frequency band 0.30 – 0.48 Hz. The fundamental
frequency of the entire support structure should be minimum 10% above 1P max (i.e.
pass if 𝑓1 > 0.176 𝐻𝑧).

In this section, an analytical expression is searched for that can deliver an accurate
instantaneous value of the first natural frequency such that the designer can adjust the
diameter on the fly.

As a first approximation of the first natural frequency a lumped mass on a stick model
can be used as presented in the PhD thesis of van der Tempel [3]. However, this does
not take into account the variation of flexural rigidity and mass over the height. And
since the diameter of the proposed design tapers from 9.6 to 4.2 m the point-on-a-stick
model is considered not accurate enough.

Alternatively, the analytical expression for a Rayleigh stepped tower may be used to
approximate the natural frequency. This method determines the unit mass and inertia
per discrete section and introduces an expected mode shape. In Appendix A the
expressions and principles for this method are worked out further.

10
Figure 4: ‘Point on a stick’ model [3] Figure 5: Rayleigh stepped tower [4]

The Rayleigh stepped tower method results in a first natural frequency of 0.245 Hz,
including the added mass of the submerged sections and foundation stiffness (see Table
3).

In Phatas, the natural frequency of the support structure is a result of the Craig-
Brampton modelling and is reported in towmod output files. As detailed loadcase
calculations will be performed with Phatas, the towmod natural frequency is compared
to the Rayleigh stepped tower in Table 4.

The frequency of 0.199 Hz is larger than 0.176 Hz and thereby passes the natural
frequency check. It must be noted that the Rayleigh stepped tower model shows a
distinctively higher eigenfrequency. This is cause of concern and deserves further
attention. For now, the natural frequency for both methods passes the natural
frequency check.

Table 4: Natural frequency estimation

Phatas towmod Rayleigh stepped tower

Rev00 0.199 Hz 0.245 Hz

2.4 Extreme load check (yield and buckling)

The local loads are expressed in three components: 𝑁 for the axial (compressive) force,
𝑉 for the shear force (positive downwind), and 𝑀 for the (effective) fore-aft bending
moment. For the design of the monopile, ultimate loads are evaluated by taking into
account the maximum rotor thrust force, self-weight of the support structure and rotor-
nacelle assembly, and hydrodynamic loading of the 50 year extreme wave. This
combination is conservative as it combines the maximum turbine load (when in
operation) with the maximum hydrodynamic load from a 50 year extreme wave. In case

11
 this sea state occurs, probably the turbine is idling. The combination is therefore not
prescribed by the IEC standard [5].
Only the extreme thrust force and self-weight of the rotor nacelle assembly are applied
at the tower top.

Thrust force
The maximum thrust force as reported in [6] is 4605 kN including a load safety factor 𝛾𝐹
of 1.35. A comparison is shown in Table 5. The DTU thrust force is considered to be
unrealistic because the ultimate loads have been extrapolated to a 50 year ultimate
value. Instead, the DTU 10 MW RWT has been modelled by ECN in aero-elastic load
simulation tool Phatas in the Focus6 suite. A set of 828 loadcases has been calculated
for an onshore and offshore model in 35 m water depth as created in other projects, for
which the maximum (absolute) thrust force is shown in Table 5. The value of the 35 m
monopile will be used further in the report and is hence indicated in bold writing in
Table 5.

Table 5: Maximum (absolute) thrust force

max|𝐹𝑡ℎ𝑟𝑢𝑠𝑡 | [𝑘𝑁] Includes 𝛾𝐹 ?

DTU onshore, at V=11 m/s 1508 No

DTU onshore, max (DLC 1.3) 4605 Yes
ECN trendwatcher, onshore 2442 No
(DLC 1.3, 12 m/s)
ECN trendwatcher, offshore, 2419 No
monopile 35 water depth
(DLC 1.3, 12 m/s)

Applying the thrust force results in an internal shear force 𝑉 that is constant over the
height of the support structure and an increasing bending moment 𝑀 to the mudline.

Self-weight
The rotor-nacelle assembly of the 10 MW RWT has a mass of 674 tons [6]. The self-
weight of the support structure and the weight rotor-nacelle assembly are multiplied by
a constant load factor of 1.25 to determine the local axial force 𝑁 as prescribed by DNV
for a static load .

Hydrodynamic loading
The hydrodynamic loading has been estimated by means of a simple calculation as
applied in the Upwind cost model [7]. The 50 year extreme wave height for the given
site is 17.67 m. The corresponding significant wave height is determined by Equation 1.
The period is found by Equation 2.

𝐻𝑟𝑒𝑑 = 1.1 ∗ 𝐻50𝑦𝑟 /1.86 Equation 1

𝐻𝑟𝑒𝑑
𝑇𝑟𝑒𝑑 = 11.1 √ Equation 2
𝑔

12
 1
The Morison equation is used for a first estimate of the extreme loads . It contains a
drag term and an inertia term. Below, drag coefficient 𝐶𝐷 = 1, inertia coefficient
𝐶𝑀 = 2 and water density 𝜌𝑤 = 1025 𝑘𝑔/𝑚3 .

𝐹ℎ𝑦𝑑 (𝑧𝑖 , 𝑡) = 𝐹𝑑 (𝑧𝑖 , 𝑡) + 𝐹𝑖 (𝑧𝑖 , 𝑡) Equation 3

1
𝐹𝐷 (𝑧𝑖 , 𝑡) = (𝑧𝑖,𝑡𝑜𝑝 − 𝑧𝑖,𝑏𝑜𝑡𝑡𝑜𝑚 ) ∙ 𝜌𝑤 ∙ 𝐶𝐷 ∙ 𝐷𝑖 ∙ 𝑢𝑛 (𝑧𝑖 , 𝑡) ∙ |𝑢𝑛 (𝑧𝑖 , 𝑡)| Equation 4
2
𝜋
𝐹𝑖 (𝑧𝑖 , 𝑡) = (𝑧𝑖,𝑡𝑜𝑝 − 𝑧𝑖,𝑏𝑜𝑡𝑡𝑜𝑚 ) ∙ 𝜌𝑤 ∙ 𝐶𝑀 ∙ 𝐷𝑖2 ∙ 𝑢̇ 𝑛 (𝑧𝑖 , 𝑡) Equation 5
4

Velocity 𝑢𝑛 (𝑧𝑖 , 𝑡) and acceleration 𝑢̇ 𝑛 (𝑧𝑖 , 𝑡) follow the airy wave theory, where the
motion of the water particles at section 𝑖 for wave 𝑛 can be described by the following
set of equations.

cosh(𝑘𝑛 𝑧𝑖 )
𝑢𝑛 (𝑧𝑖 , 𝑡) = 𝜁𝑛 ∙ 𝜔𝑛 ∙ ∙ sin(𝜔𝑛 𝑡) + 𝑢𝑐 (𝑧𝑖 ) Equation 6
sinh(𝑘𝑛 𝑑)
cosh(𝑘𝑛 𝑧𝑖 )
𝑢̇ 𝑛 (𝑧𝑖 , 𝑡) = 𝜁𝑛 ∙ 𝜔𝑛2 ∙ ∙ cos(𝜔𝑛 𝑡) Equation 7
sinh(𝑘𝑛 𝑑)

Where 𝜁𝑛 is equal to 𝐻𝑟𝑒𝑑 /2, 𝜔𝑛 is the wave frequency in rad/s, 𝑑 is the water depth,
𝑢𝑐 (𝑧) is the current velocity, and 𝑘𝑛 is the angular wave number which is iteratively
determined by Equation 8.
𝜔𝑛2
𝑘𝑛 = Equation 8
𝑔 tanh 𝑘𝑛 𝑑

𝑢𝑛 (𝑧𝑖 , 𝑡), 𝑢̇ 𝑛 (𝑧𝑖 , 𝑡), 𝐹𝑑 (𝑧𝑖 , 𝑡) and 𝐹𝑖 (𝑧𝑖 , 𝑡) are determined for each section for which the
average height 𝑧𝑖 is lower than LAT. The hydrodynamic force 𝐹ℎ𝑦𝑑 (𝑧𝑖 , 𝑡) is found for a
water depth of 29.6 m with a uniform current profile 𝑢𝑐 of 1.2 m/s and 50 yr extreme
wave conditions (𝐻𝑟𝑒𝑑 = 10.45 𝑚, 𝑇𝑟𝑒𝑑 = 11.46 s). The maximum 𝐹ℎ𝑦𝑑 (𝑧𝑖 ) in one wave
period is added to internal shear force 𝑉 after multiplication with variable load factor 𝛾𝐹
of 1.35. The wave loading also influences the distribution of the internal bending
moment 𝑀.

xxxxxxxxxxxxssssssssxxxxxxxxxxxxxx

1 Note that for the fatigue simulations, kinetics and hydrodynamic loads of the waves, including current, are
calculated by random ocean wave simulation tool ROWS [5]

13
 Figure 6: Internal loads distribution for extreme thrust force and 50 year extreme wave

2.4.1 Yield check

The yield check considers whether the local stress is smaller than the yield strength 𝑆𝑦 .
If the condition in Equation 9 the yield check is passed.

𝑁 𝑀 𝑆𝑦
+ ≤ Equation 9
𝐴 𝑊𝑒 𝛾𝑀

Where 𝑊𝑒 is the (elastic) section modulus that can be evaluated by Equation 10.

𝐼 𝐼 𝜋 𝐷4 − (𝐷 − 2𝑡)4 𝜋 3 2𝑡 4
𝑊𝑒 = = = ( )= 𝐷 [1 − (1 − ) ] Equation 10
𝑦 𝐷 ⁄2 32 𝐷 32 𝐷

Yield strength 𝑆𝑦 is set at 355 MPa for monopile, transition piece and tower (as
reported in [6]). A material safety factor 𝛾𝑚 = 1.1 is used in Equation 9.

As can be seen in Figure 7, the stress throughout the structure is lower than the yield
stress divided by the material safety factor. Hence, no yield failure is expected and the
yield check is passed.

14
Figure 7: Local stress (blue line)

15
 2.4.2 Global buckling check

The global buckling check is performed to determine column buckling of the entire
support structure (monopile, transition piece and tower). The equations used for this
check are presented in Appendix B. The parameter 𝛽 determines the buckling mode,
displayed in Figure 8. It is anticipated that 𝛽 = 2 is the failure mode for the entire
support structure. However, as can be seen by the right plot in Figure 9, the global
buckling check is not passed for 𝛽 = 2 (global buckling utilization ratio is above 1,
equaling a buckling failure).

Figure 8: Buckling failure modes

Figure 9: Global buckling check for two buckling failure modes

The validity of this buckling mode for the entire support structure is subject to further
evaluation considering that there are several flanges (and a stiff transition piece)
present in the structure, which can resist a global buckling mode as shown in Figure 8.

16
2.4.3 Local buckling check

In ‘Theory of elastic stability’ of Timoshenko & Gere [8], the critical local buckling stress
is given with Equation 11.

𝐸 𝑡𝑤 𝑁𝑐𝑟𝑖𝑡
𝜎𝑐𝑟 = = Equation 11
0.5𝐷√3(1 − 𝜈 2 ) 𝐴

According to DNV OS C201 [9], local buckling of tubular members with external pressure
need not to be considered for
𝐷 𝐸
≤ 0.5 √
𝑡 𝑆𝑦
𝐷
Or in other words, local buckling must be checked for ≥ 12.
𝑡
As for all sections the 𝐷/𝑡 ratio it well above 12, local buckling must be checked for
each cylinder section. In Figure 10 the critical stress and critical normal force are given.
As can be seen by the yellow striped line, the critical stress and normal force are not
reached by far.

Figure 10: Local buckling critical stress and normal force

17
 2.5 Fatigue check

2.5.1 Load case description (power production)

Design load case (DLC) 1.2 is taken for the calculation of fatigue damage, according to
IEC 61400-3 [5]. This load case describes a number of seeds per wind speed bin over the
operational range. Six unique seeds are used to generate the turbulent wind field for
each wind speed bin, running from 3 to 25 m/s. The loadcases are prepared using the
Focus program LCprep3ed1, following IEC 61400-3 edition 1 [5].

The occurrences are also calculated by LCprep and follow a Weibull distribution, defined
by a Weibull shape parameter and annual average wind speed for a twenty year
duration. The distribution used in fatigue processing is graphically depicted in Figure 11.

The environmental conditions are based on the Upwind deep water site, which is based
on measurements at the K13 platform and reported upon in Upwind design basis
document [10]. The reference hub height of the Upwind deep water design, annual
average wind velocity at this height and roughness length is shown in Table 6. Because
the hub height of the proposed monopile design is higher (due to the larger blade
diameter), the annual average wind velocity is extrapolated using the logarithmic shear
formulation in Equation 12.

ln(𝑧⁄𝑧0 )
𝑉(𝑧) = 𝑉(𝑧𝑟𝑒𝑓 ) ∙ Equation 12
ln(𝑧𝑟𝑒𝑓 ⁄𝑧0 )

Table 6: Annual average wind velocity at reference height Upwind and hub height 10 MW monopile

Upwind K13 deep water site

𝑧𝑟𝑒𝑓 Upwind 90.55 m (above SWL)

V (𝑧𝑟𝑒𝑓 ) 10.05 m/s

z0 0.002 m
10 MW monopile
z 121.521 M
V(z) 10.33 m/s

18
Figure 11: Distribution of 10 minute loadcases K13 deep [10]

2.5.2 Fatigue processing

After simulating six ten minute time series with different wind and wave seed per (1
m/s) wind speed bin, the resulting forces and moments in the structure are post-
processed into a fatigue equivalent value per loadcase. The rainflow algorithm is used
to count individual cycles with distinct range ∆𝑆𝑖 . Since the PHATAS simulation tool does
not calculate stresses, the force and bending moment time series are substituted for 𝑆.
The fatigue equivalent force or moment can be found by applying Equation 13. Here
𝑁𝑒𝑞 is the equivalent number of cycles and 𝑚 is the Wohler coefficient. For a ten
minute time series, when choosing 𝑁𝑒𝑞 = 600, the resulting ∆𝑆𝑒𝑞 is a 1Hz-equivalent
fatigue range. This means that a fixed amplitude 1Hz harmonic with range ∆𝑆𝑒𝑞 causes
the same level of fatigue damage as the mixed amplitude, multi-harmonic series S(t).

𝑛 1/𝑚
∆𝑆𝑖𝑚
∆𝑆𝑒𝑞 = (∑ ) Equation 13
𝑁𝑒𝑞
𝑖=1

19
 2.5.3 Fatigue load DLC 1.2

As was concluded from section 2.5.2 above, fatigue equivalent loads can be post-
processed by FocusReport. A set of 138 loadcases was simulated for Design Load Case
1.2, according to IEC 61400-3 [5]. Fatigue equivalent fore-aft bending moment at
mudline for these loadcases is presented in Figure 12.

Figure 12: Fatigue equivalent bending moment (using single Wohler component 𝑚=4)

An approximation can be made to the fatigue equivalent normal stress at the mudline
𝑀y,deql ∙𝐷𝑜
by evaluating the fore-aft bending moment into stress (𝜎𝑁 𝑑𝑒𝑞𝑙 ≈ ). However,
2𝐼
also the normal force contributes to the magnitude of stress in the material. Equation
14 shows all contributions to normal stress. The maximum normal stress is found at the
outer radius, for which the arm of the fore-aft bending moment is largest. As the fore-
aft bending moment 𝑀𝑦 is commonly larger than the sideways bending moment during
power production, Equation 15 holds as an expression for maximum normal stress in a
circular cross-section (for 𝑥 = 𝐷𝑜 /2, 𝑦 = 0).

𝑴𝒙 𝒚 𝑴 𝒚 𝒙 𝑵
𝝈𝑵 = + + Equation 14
𝑰 𝑰 𝑨
𝑴𝒚 𝑫𝒐 𝑵
𝝈𝑵,𝒎𝒂𝒙 = + Equation 15
𝟐𝑰 𝑨

20
The individual contribution of fore-aft bending moment and normal force to the normal
stress at the mudline is shown in a single time series of power production for 25 m/s
mean wind velocity.

Figure 13: Normal stress caused by bending moment (blue) and normal force / self-weight (red) at the
mudline for power production at 25 m/s mean wind velocity.

It can be seen from Figure 13 that there is little variation in the normal stress caused by
normal force, mostly consisting of self-weight of the upper structure. Hence, the
contribution of normal force to fatigue is limited.

The stress time series of Equation 15 (including normal force) are rainflow counted and
a damage equivalent value is created for each individual loadcase, taking into account
the multi-slope SN curve, see Figure 14.

21
 Figure 14: Distribution of fatigue equivalent normal stress 𝜎𝑁,𝑑𝑒𝑞𝑙 at mudline, for each loadcase (upper
plot) and averaged for each wind velocity (lower plot).2

2.5.4 Fatigue damage

The fatigue damage per loadcase 𝑗 is calculated by Equation 20.

𝑛𝑖
𝐷𝑗 = ∑ Equation 16
𝑁𝑖
𝑗

For which 𝑛𝑖 is the number of stress cycles and 𝑁𝑖 is the number of cycles to failure for
a given stress cycle range ∆𝜎𝑖 . 𝑁𝑖 can be calculated by Equation 21, where SCF
represents the stress concentration factor.

log 𝑁𝑖 = log 𝑎 − 𝑚 ∙ log (SCF ∙ ∆𝜎𝑖 ) Equation 17

The values for log 𝑎 and 𝑚 are dependent on the SN curve. According to Appendix A of
DNVGL-RP_C203 Fatigue Design of Offshore Steel Structures [12], longitudinal weld
seems fall under SN category B2, or C1-F3 for circumferential butt welds.

xxxxxxxxxxxxssssssssxxxxxxxxxxxxxx

2 A multi-slope SN curve was used in creating this graph.

22
Table 7: S-N curves in seawater with cathodic protection (DNVGL RP C203 [12])

𝑁 ≤ 106 cycles 𝑁 > 106 cycles

S-N curve 𝑚1 log 𝑎̅1 𝑚2 log 𝑎̅2 SCF

B2 4.0 14.685 5.0 93.59 N/A

C1 3.0 12.049 5.0 65.50 N/A
F3 3.0 14.576 5.0 14.576 1.61

Curve B2 is used as the representative SN relation. One has to take into account
possible hot spot stresses at geometric details. An example of this is the cable entry
hole, where the stress concentration factor (SCF) can be higher than 3.

The effect of plate thickness of welded joints on the fatigue damage is accounted for by
𝑘
𝑡
multiplying the stress range by ( ) . However, for class B2, 𝑘 = 0 and hence the
𝑡𝑟𝑒𝑓

thickness effect may be omitted.

For each individual loadcase, a fatigue damage 𝐷𝑖 is calculated using Equation 16 for all
identified stress ranges ∆𝜎𝑗 in the 10 minute time series. The lifetime fatigue damage is
determined by summing the individual damage times the occurrence of that loadcase,
which is shown per wind speed bin in Figure 11. The Weibull distribution shows that the
loadcases in DLC 1.2 (power production from 3-25 m/s wind speed) amount to 978780
ten minute time series, which equals 18.6 years of operation in the 20 year lifetime.

𝐷 = ∑occ𝑖 ∙ 𝐷𝑖 Equation 18

The results for lifetime fatigue damage, for a variation of stress concentration factors,
are shown in Table 8 and Figure 15.

Table 8: Fatigue damage at mudline for different stress concentration factors

SCF=1.0 SCF=1.5 SCF=2.0 SCF=2.5 SCF=3 SCF=3.5

Fatigue 0.0079 0.0601 0.25335 0.7721 1.9065 4.0218

damage D [-]

23
 Figure 15: Graph depicting fatigue damage at mudline for different stress concentration factors

The contribution of each wind speed to the lifetime fatigue damage is found by
summing the weighted damage per wind speed bin (Equation 19). Figure 16 displays the
contribution of each wind speed bin to the fatigue damage.
∑6𝑖=1 occ𝑖 ∙ 𝐷𝑖
∗ 100% Equation 19
𝐷𝑡𝑜𝑡𝑎𝑙

Figure 16: Contribution of each wind speed bin to total lifetime fatigue damage

24
 3
Limitations of current
models
In this chapter, the applicability of the current modelling approach for very large
monopiles is discussed. In upscaling the monopile concept, the validity of certain model
assumptions may be exceeded. The modelling capabilities of the structural model, soil-
structure interaction, hydrodynamic model and fatigue model are considered in this
chapter.

3.1 Structural model

In Phatas a Craig-Bampton (CB) model is created to account for the tower, transition
piece and monopile. This is useful in creating a beam-like model for a multi body
structure such as a jacket. In order to capture the dynamics of the support structure
enough CB modes must be taken into account. In the Phatas manual at least 5 internal
bending modes [14] are suggested for offshore turbines and in the simulations, 6
bending modes were applied. For more information on the Craig-Bampton method,
consult [15].

3.2 Soil structure interaction

Soil structure interaction is modelled in Phatas by a four degree of freedom spring

matrix [14]. Recently, advances have been made to incorporate multiple springs
throughout the soil making modelling of (non-linear) PY curves possible [16]. However,
PY curves have been developed for slender piles in the oil & gas industry bending in an
elastic fashion. A general rule of L/D ratio smaller than 7 indicates the bending mode is
more rigid [17]. In the PISA project, numerical modelling and field testing was

25
 performed and new design methods have been developed to account for large
diameter foundation piles [18], [19] . The findings include the increased importance of
distributed moment, base shear and base moment for shorter wider foundation piles
compared to the long slender piles where lateral P-Y springs may suffice.

3.2.1 Soil stiffness models

From the investigations into the various soil-structure stiffness models, the options are
listed below in increasing level of complexity:

- Rigid
- Apparent fixity length
- 2 DOF spring matrix (PHATAS)
- Linearized P-Y curves
- Non-linear P-Y curves (PHATAS + py module)
- Modified soil structure springs
- Strain wedge method
- FEM linear elastic solids
- FEM non-linear, anisotropic
- FEM multi-phase
- FEM + CFD of water

It can be seen that the additional PY module in PHATAS only increases the modelling
capabilities. Because of the custom (non-linear) spring curves that can be entered,
updated PY relations from the PISA project can be implemented. It is however advised
to carry out a validation study, and include other degrees of freedom (vertical and
rotation springs).

Influence of soil model on eigenfrequency

In the master thesis of Joey Velarde, the effect of the soil model to the natural
frequency has been investigated. It turns out that the API method results in a 4.0%
higher eigenfrequency than a finite element code for a 10 MW monopile foundation in
50 m water depth [20].

API method is overestimating soil stiffness, especially at greater soil depth, causing a
smaller lateral deflection at the seabed and pile toe [20]. The bending moment
distribution is however similar, see Figure 17.

26
Figure 17: Lateral deflection and bending moment distribution for pile of 50 m water depth [20]

Influence of variation of soil parameters on eigenfrequency and fatigue lifetime

Sebastien Schafhirt of NTNU has carried out a soil parameter variation study for the
generic NREL 5 MW turbine supported by a monopile structure as analysed in Phase II
of the OC3 Project [21]. The effect on eigenfrequency, fatigue load and fatigue lifetime
was quantified for a variation of soil parameters 𝑘 (initial modulus of subgrade
reaction), 𝜙 ′ (internal friction angle), and 𝜃 (accumulated rotation).
The fatigue lifetime was shown to vary between 18.21 – 20.80 years. A maximum
reduction of 6 % for the natural eigenfrequency was found by the author.

3.2.2 Foundation damping

The effective soil damping is a value to which much uncertainty is connected. The
effective soil damping influences the energy in the vibrations of the structure and as
such has an influence on the simulation of cyclic loads and the resulting fatigue life.
Carswell et al [22] have mapped the estimated critical damping values from literature
which ranges from 0.17 % to 3 % of critical damping. ECN has conducted a study to this
phenomenon in 2014 [23]. Soil damping is not modelled in Phatas however.

27
 3.3 Hydrodynamic models: diffraction

considered McCamy Fuchs

The monopile is one of the possible support structures for large sized modern wind
turbines in 30-50m water depth. For these conditions the size i.e. the diameter of the
monopile increases accordingly. The question is whether a XL monopile should be
considered a large structure compared to the wave and wave diffraction should be
accounted for.

An offshore structure is considered large in case the ratio of wave length L and the
cylinder diameter D is smaller than 5 (L/D < 5). For large structures the Morison
equation is not valid anymore for wave load calculations and diffraction should be
accounted for. For a vertical cylinder McCamy and Fuchs (1954) [24] derived an
analytical solution for the wave diffraction. The effect of diffraction can be expressed in
the inertia coefficient Cm, Figure 18, as applied in the Morison equation.

4𝐴(𝑘𝑎)
𝐶𝑚 = Equation 20
𝜋(𝑘𝑎)2

With
2 2
𝐴(𝑘𝑎) = √𝐽1′ + 𝑌1′ Equation 21

Where a = radius of cylinder, k = wave number and 𝐽1′ and 𝑌1′ are derivatives of the
Bessel functions of the first and second kind and order 1 respectively.

28
Figure 18: Inertia coefficient Cm for given water depth d, wave peak period Tp and diameter D.

For the inertia wave load Fi and the overturning moment Mi a response amplitude
operator (RAO) can be derived [25]. The equations are:

𝜋
𝐹𝑖 = 𝜌𝑔 𝐷2 tanh 𝑘𝑑 Equation 22
4
𝜋 1 1
𝑀𝑖 = 𝜌𝑔 𝐷 2 𝑑 [tanh 𝑘𝑑 + ( − 1)] Equation 23
4 𝑘𝑑 cosh 𝑘𝑑

Figure 19 shows the RAO of the inertia force Fi for both a constant Cm and a Cm with
the McCamy Fuchs diffraction correction. For wave frequencies above 1 rad/s the
effect of diffraction on the inertia force is clearly visible.

29
 Figure 19: Inertia load Fi based on constant Cm = 2 and MacCamy Fuchs correction.

The way an offshore structure responds depends on both the wave input and the RAO
of the inertia force. For a selected wave spectrum the response is calculated as follows.

𝑆𝐹𝑖 = |𝐹𝑖|2 S𝜁 Equation 24

Case study for three different cylinder diameters

As an example the approach is applied for three cylinder diameters: 7.5m, 10.m and
15m. The Pierson Moskowitz spectrum are applied for a seastate with significant
waveheight Hs = 1.1m and a peak period Tp = 5.9s.

Based on seatate wave period Tp = 5.900s and diameter D = 7.500m. The following
parameters are estimated:
- Wave number k = 0.116.
- L/D = 7.247.
- Diffracton parameter ka = 0.434.
- Inertia coefficient Cm = 2.042.

Based on the L/D number there is no need to include diffraction for this peak period.
This is also confirmed by the spectral load response Figure 20.

30
Figure 20: Response spectra of inertia force Fi for a cylinder diameter D = 7.5m.

Based on sea state wave period Tp = 5.900s and diameter D = 10.000m. The following
parameters are estimated:
- Wave number k = 0.116.
- L/D = 5.435.
- Diffraction parameter ka = 0.578.
- Inertia coefficient Cm = 1.940.

The L/D coefficient is now very close to five, where diffraction becomes more
important. Figure 21 shows that for the difference at higher wave frequencies becomes
visible.

Figure 21: Response spectra of inertia force Fi for a cylinder diameter D = 10.0m.

31
 Finally based on sea state wave period Tp = 5.900s and diameter D = 15.000m.
- Wave number k = 0.116.
- L/D = 3.623.
- Diffraction parameter ka = 0.867.
- Inertia coefficient Cm = 1.561.
The L/D ratio is below five. It is clear from Figure 22 that wave diffraction should be
accounted for.

Figure 22: Response spectra of inertia force Fi for a cylinder diameter D = 15.0m.

Conclusions
For a given sea state the peak period Tp can be used in the L/D ratio to identify whether
diffraction should accounted for. The response spectrum shows clearly which wave
frequencies are sensible to diffraction effects.

3.4 Fatigue model

In the fatigue damage evaluation of section 2.5, the SN curve has been used in
conjunction with Palmgren-Miner’s rule. This a commonly used approach described in
DNVGL recommended practises. However, there are large uncertainties connected for
this model. Both the SN parameters as the Palmgren-Miner’s rule include uncertainties.
In [13], the standard deviation of 0.2 for log 𝑎̅1 and 0.25 for log 𝑎̅2 are assumed. The
Palmgren-Miner rule is assumed log normal distributed with a median 1.0 and
Coefficient of Variation (CoV) of 0.3 [12]. This means that there is an inherent
uncertainty on the value of fatigue damage when using this simple approach. However,
the standard prescribes the use of these SN curves, which are in the end derived from
experiments, together with a safety factor. Hence conservatism is introduced that could
be reduced by investigating more physical effects of fatigue build-up and degradation.

32
For example, the assumption of linear fatigue damage over time does not account for
hardening/softening effects or accelerated crack growth. More complex fatigue models
have been developed to describe this effect. TNO is currently conducting a TKI Project
‘FeLoSeFI’ on the inclusion of retardation and load sequence effects and at institutes
such as WMC and Forwind fatigue tests of joints in jackets are carried out.

33
 4
Conclusion &
Recommendations
In this report, the design procedure for a monopile support structure is presented and a
basic design is created for the Innwind 10 MW offshore turbine in 50 m water depth.
This combination of a large turbine in deep water has not been applied by the industry
at this moment and may be considered outside the boundaries of applicability of a
monopile support structure. The preliminary design suggested as a first revision is
within the current manufacturing capabilities [26] with a bottom diameter of 9.3 m and
a combined foundation, transition piece and tower mass of 1840 tons (excluding the
embedded monopile section). Installing such an XL monopile will form a challenge.
However, initial contacts with the industry have indicated that if there is a demand for
larger crane capacity, hammer size or lifting reach, the installation contractors will
develop new vessels that can handle such sizes.

The basic design was subjected to a quasi-static ultimate load check, buckling check and
a fatigue check (at the mudline only). The static ultimate load was comprised of the
maximum rotor force from a previous dataset simulation (DLC 1.3) together with a 50
year extreme sea state. This is a conservative combination. It is recommended to
implement a more extensive ultimate load set that includes conditions as prescribed by
the design standards (where the turbine is idling for a 50 year sea state).
The nominal stress throughout the structure does not cause yielding for the ultimate
load, hence the yield check is passed. For stress concentrations however, the yield
stress may be surpassed. The global buckling check for column buckling of the entire
support structure is not passed for the most applicable buckling mode (𝛽 = 2).
However, the total buckling length for this failure mode of twice the entire support
structure length is very large and may not be relevant. For 𝛽 = 1, the global buckling
check is passed. Further investigation is needed to assess the applicability of this global
buckling mode.

Local buckling is checked for the entire support structure and the local stress and axial
force is found to be much lower than the critical stress and axial force. Fatigue damage
is calculated by performing a simulation of design load case 1.2, according to IEC 61400-

34
3 [5], for the wind and wave climate of Doggers bank. The time series were processed
by applying a rainflow algorithm, SN curve and Miner’s rule. The Wohler parameters for
a circumferential weld in water with cathodic protection (class B2) are taken. The
fatigue check is passed up to a stress concentration factor of 2.5. Limitations in the
current models include soil structure interaction, and simplifications of the structural
model and hydrodynamic load calculations. The increased functionality of Phatas by
using lateral springs over the length of the embedded foundation pile allow for
modelling of individual P-Y springs. However, it has been shown that other physical
effects should not be overlooked for large diameter foundation piles. It is hence
recommended to include flexibility in more degrees of freedom and perform validation
studies. The structural model and the use of Morison’s equation for wave loading are
considered to be sensible for large monopile foundations. Calculating lifetime fatigue by
applying a Wohler curve combined with the Palmgren-Miner’s rule is a conventional
way of determining fatigue life and follows the current standards and norms. However,
when accounting for crack growth, hardening and other effects will result in a more
realistic fatigue life estimation. Finally, the Rayleigh stepped tower algorithm suggested
in this report did not match the eigenfrequency from Phatas and therefore needs
further investigation before integrating it in a foundation design tool.

35
 Bibliography
[1] Offshorewind.biz, “EEW SPC rolls out world’s heaviest monopile,” Mar. 2016.
[2] P. Passon, Derivation and Description of the Soil-Pile-Interaction Models, IEA -
Annex XXIII Subtask 2, OC3 Project, 2006.
[3] J. van der Tempel, “Design of support structures for offshore wind turbines,” TU
Delft, 2006.
[4] L. Harland and J. Vugts, “Analytic expression for the first natural period of a
stepped tower,” Oct. 1996.
[5] Wind Turbines - Part 3 Design Requirements for offshore wind turbines, IEC, 2007.
[6] C. Bak, F. Zahle, R. Bitsche, T. Kim, A. Yde, L.C. Henriksen, A. Natarajan, and M.
Hansen, DTU 10-MW Reference Wind Turbine, 2013.
[7] W.E. de Vries, UpWind - Final report WP 4.2 - Support Structure concepts for
Deep Water Sites, Delft University of Technology, 2011.
[8] S.P. Timoshenko and J.M. Gere, Theory of elastic stability, 1961.
[9] Buckling strength of plated structures, DNV Recommended Practise C201, 2010.
[10] T. Fischer, W. deVries, and B. Schmidt, Upwind Design Basis - WP4: Offshore
foundations and support structures, 2010.
[11] H.B. Hendriks and B.H. Bulder, Fatigue equivalent load cycle method - a general
method to compare the fatigue loading of different load spectrums, ECN-C–95-
074, 1995.
[12] RP-C203: Fatigue design of offshore steel structures, DNVGL Recommended
Practise C203, .
[13] S. Márquez-Domínguez and J.D. Sorensen, “Fatigue reliability and calibration of
fatigue design factors for offshore wind turbines,” Energies, vol. 5, 2012, pp.
1816–1834.
[14] C. Lindenburg, “PHATAS Release ‘Jan-2014a’ User’s manual.”
[15] J.T. Young, Primer on the craig-bampton method, 2000.
[16] R. Brood, Soil modelling in phatas with P-Y reference guide, WMC, 2016.
[17] W.G. Versteijlen, J.M. de Oliveira Barbosa, van K.N. Dalen, and A.V. Metrikine,
“Method for extracting an equivalent winkler model of the 3D dynamic soil-
structure interaction or large-diameter offshore monopile foundations,”
Proceedings of XLIII International Summer School Conference APM, 2015.
[18] B. Byrne, R. McAdam, H. Burd, G. Houlsby, C. Martin, L. Zdravkovi, D. Taborda, D.
Potts, R. Jardine, and M. Sideri, “New design methods for large diameter piles
under lateral loading for offshore wind applications,” 3rd International
Symposium on Frontiers in Offshore Geotechnics (ISFOG 2015), Oslo, Norway,
June, 2015, pp. 10–12.
[19] B. Byrne, R. McAdam, H. Burd, G. Houlsby, C. Martin, K. Gavin, P. Doherty, D.
Igoe, L. Zdravkovi, D. Taborda, and others, “Field testest of large diameter piles

36
 under lateral loading for offshore wind applications,” Proceedings of the 16th
European Conference on Soil Mechanics and Geotechnical Engineering,
Edinburgh, UK, 2015.
[20] J. Velarde, “Design of monopile foundations to support the DTU 10 MW offshore
wind turbine,” DTU / NTNU / TU Delft, 2016.
[21] S. Schafhirt, A. Page, G.R. Eiksund, and M. Muskulus, “Influence of soil
parameters on the fatigue lifelife of offshore wind turbines with monopile
support structure,” 13th Deep Sea Offshore Wind R&D Conference, EERA
DeepWind’2016, 2016.
[22] W. Carswell, J. Johansson, F. Lovholt, S.R. Arwade, C. Madshus, D.J. DeGroot, and
A.T. Myers, “Foundation damping and the dynamics of offshore wind turbine
monopiles,” Renewable Energy, vol. 80, 2015, pp. 724–736.
[23] K.W. Hermans, Evaluation of modal parameters and damping from the structural
response of an offshore wind substructure. ECN-WIND-2014-242, 2014.
[24] R.C. MacCamy and R.A. Fuchs, Wave forces on piles: a diffraction theory, Beach
erosion board corps of engineers, 1954.
[25] T. Sarpkaya and I. Isaacson, Mechanics of wave forces on offshore structures, van
nostrand reinhold company inc., 1981.
[26] SIF group, “Monopile manufacturing State-of-the-art, https://sif-
group.com/en/quality/state-of-the-art,” 2016.

37
 Appendices

38
Appendix A. Rayleigh
stepped tower
The natural frequency of a tower with varying diameter can be approximated by the
analytic expression following the Rayleigh method. This has been worked out in a note
by Léon Harland and Jan Vugts in the OPTI-OWECS project [4] and was applied later in
the Upwind cost model.

The method is based upon the principle of maximum potential and kinetic energy in the
vibration modes and assumes a mode shape Ψ(𝑥) beforehand.
The mode shape used in the approximation can be described by Equation 25 and Figure
23.

𝝅𝒙
𝚿(𝒙) = (𝟏 − 𝐜𝐨𝐬 ( )) Equation 25
𝟐𝑳

Figure 23: Assumed mode shape Rayleigh stepped tower

The final expression for the first natural period in the note by Harland and Vugts [4] is
given in Equation 26.

𝟒𝝅𝟐 (𝑴𝒕𝒐𝒑 + 𝒎𝒆𝒒 𝑳)𝑳𝟑 𝟒𝟖

𝑻𝟐𝒕 = ∙ [ 𝟒 + 𝑪𝒇𝒐𝒖𝒏𝒅 ] Equation 26
𝟑𝑬𝑰𝒆𝒒 𝝅

The equivalent moment of inertia 𝐼𝑒𝑞 , mass per unit length 𝑚𝑒𝑞 and foundation
flexibility 𝐶𝑓𝑜𝑢𝑛𝑑 is found by the next expressions:
𝝅𝒙
∑𝒏𝒋=𝟏 𝑰𝒋 𝒍𝒋 𝐜𝐨𝐬 𝟐 ( 𝒋 )
𝑰𝒆𝒒 = 𝟐𝑳 Equation 27
𝑳
𝟐
𝝅𝒙
∑𝒏𝒋=𝟏 𝒎𝒋 𝒍𝒋 (𝟏 − 𝐜𝐨𝐬 ( 𝒋 ))
𝟐𝑳 Equation 28
𝒎𝒆𝒒 =
𝑳
𝟑𝑬𝑰𝒆𝒒
𝑪𝒇𝒐𝒖𝒏𝒅 = Equation 29
𝑲𝒆𝒒 𝑳
𝟐
𝑲𝒓𝒐𝒕 𝑲𝒍𝒂𝒕 𝑳
𝑲𝒆𝒒 = Equation 30
𝑲𝒓𝒐𝒕 + 𝑲𝒍𝒂𝒕 𝑳𝟐

39
 Additionally, when the individual section 𝑗 is submerged the mass of the internal water
is added to the mass 𝑚𝑗 . For revision 00 of the support structure design, the mass and
the calculated Rayleigh stepped tower frequency is shown in Table 9.

Table 9: Parameters for determining Rayleigh stepped tower frequency - design rev00

value

Mass of tower steel 1849.01 tons

Mass of internal water 2773.97 tons
Mass of tower (water + steel) 4622.97 tons
First (Rayleigh) frequency 0.2454 Hz

40
Appendix B. Global
buckling check
The following parameters are evaluated in the Upwind cost model [2] to perform a
global buckling check.

Elastic buckling force 𝑁𝑒

𝜋 2 𝐸𝐼
𝑁𝑒 = Equation 31
1.1 𝑠𝑘2

Buckling length 𝑠𝑘 = 𝛽 ∗ 𝐿.

For this check, 𝛽 = 2.0 and 𝐿 = hub height + water depth + scour depth.
The scour depth for this design is disregarded. In other words, scour protection is
assumed, preventing any scour developments.

Plastic compression resistance

𝐴 ∙ 𝑆𝑦
𝑁𝑝 = Equation 32
𝛾𝑀

(𝐴: cross sectional surface area. 𝑆𝑦 : yield strength, 𝛾𝑀 : material safety factor)

Plastic moment resistance

𝑊𝑝 ∙ 𝑆𝑦
𝑀𝑝 = Equation 33
𝛾𝑀

𝑊𝑝 is the plastic resistance capacity or plastic section modulus and is calculated with
Equation 34.
1 2𝑡 3
𝑊𝑝 = 𝐷3 [1 − (1 − ) ] Equation 34
6 𝐷

Reduced slenderness ratio 𝜆̅

𝑁𝑝 ∙ 𝛾𝑀
𝜆̅ = √ Equation 35
𝑁𝑒

Determine 𝜙
𝜙 = 0.5 ∙ [1 + 𝛼 ∙ (𝜆̅ − 0.2) + 𝜆2̅ ] Equation 36
(𝛼: 0.21 for hollow circular sections)

41
 Reduction factor 𝜅
1
for 𝜆̅ > 0.2
𝜅 = {𝜙 + √𝜙 2 − 𝜆2̅ Equation 37
1 for 𝜆̅ ≤ 0.2

∆𝑛 = 0.25 ∙ 𝜅 ∙ 𝜆2̅ ≤ 0.1 Equation 38

Unity check buckling due to axial loading

𝑁 𝛽𝑀 ∙ 𝑀
+ + ∆𝑛 ≤ 1 Equation 39
𝜅 ∙ 𝑁𝑝 𝑀𝑝
Note that Equation 39 has similar features to the yield check in in Equation 9. Due to
the introduction of factors 𝜅, 𝛽𝑀 and ∆𝑛 in the equation, the buckling criterion is stricter
than the yield check. In other words; if the buckling check is passed, the yield stress will
not be surpassed, so in this case the yield check can be omitted. However, a yield check
remains valuable for stress concentrations around geometric details such as the cable
entry hole or transition piece flange weld.

42
weg 3 P.O. Box 1
en 1755 LG Petten
nds The Netherlands

4949
8338

43