You are on page 1of 8

Proceedings of the ASME 2022 41st International

Conference on Ocean, Offshore and Arctic Engineering


OMAE2022
June 5-10, 2022, Hamburg, Germany

OMAE2022-79114

FATIGUE LIFE EVALUATION OF A TIDAL TURBINE BLADE: FROM SIMULATIONS USING


BEMT/FEM AND CFD/FEM COUPLINGS TO FULL-SCALE TEST

Stéphane Paboeuf Meryem Guisser Sébastien Loubeyre Peter Davies


Bureau Veritas Bureau Veritas Bureau Veritas Solutions Ifremer Centre Bretagne
4 rue Duguay-Trouin 4 rue Duguay-Trouin 4 rue Duguay-Trouin Marine Structures
44800 Saint-Herblain, 44800 Saint-Herblain, 44800 Saint-Herblain, Laboratory,
France France France 29280 Plouzané, France

Maël Arhant Nicolas Dumergue Erwann Nicolas


Ifremer Centre Bretagne Ifremer Centre Bretagne Sabella
Marine Structures Laboratory, Marine Structures Laboratory, 7, rue Félix Le Dantec,
29280 Plouzané, France 29280 Plouzané, France 29000 Quimper, France

ABSTRACT results are then compared with measurements performed on a


In the race towards a greener and safer tomorrow, full-scale blade at Ifremer facilities. Furthermore, a Blade
renewable energies are playing an uncontestably key role. The Element Momentum Theory (BEMT) tool and Computational
energetic transition to cleaner energy has been the go to strategy Fluid Dynamics (CFD) software are used to generate
for countries around the globe. Nature offers many power operational and extreme environmental loads applying on the
sources and tidal turbine is one of marine systems able to convert blade. BEMT/FEM one-way coupling and CFD/FEM two-way
tide energy in electricity. However, tidal turbine technology is coupling numerical simulations have been realised and results
still under development and research area remains an important between both methods have been compared. Finally, the fatigue
aspect before commercial deployment. life, in operational condition, has been evaluated with the recent
RealTide, a European research project on advanced fatigue methodology, based on a ply-by-ply approach, developed
monitoring, simulation and control of tidal devices in unsteady, by Bureau Veritas. Several design optimisations including glass
highly turbulent realistic tide environments, is the perfect fibres plies have also been studied in order to reduce costs of
example of the investment that goes into research and production.
development to create better technologies in order to help with Keywords: Tidal turbine, composite materials, Fatigue
the energetic transition. The goal of the project is to combine
advanced in flow measurement in highly energetic tidal currents 1. INTRODUCTION
with condition monitoring systems to design reliable tidal The main objective of tidal turbine developers is to design,
turbines. This implies the study of the turbine components, manufacture and install efficient and reliable machines at sea.
chiefly the blades, and the causes of their failure. During the 45 months of the research project RealTide1 [1],
The aim of the present paper is to hand over the work FMEA (Failure Modes and Effect Analysis) [2] and RAM
performed within RealTide project on a blade, developed by (Reliability, Availability and Maintainability) [3] analyses have
Sabella for the next tidal turbine generation, made in composite been carried out in order to identify critical components and the
materials. Firstly, a Finite Element Model (FEM) is carried out blade of turbines has been identified as one of them.
of the 5-meters length blade in carbon fibre. A modal and static The study, presented in this paper, focuses on the Sabella
analysis under various testing loads are run on the FEM and D12 turbine, the next tidal turbine generation developed by

1
RealTide project has received funding from the European Union’s
Horizon 2020 research and innovation programme under Grant Agreement No
727689.

1 © 2022 by ASME
Sabella. The turbine consists of a 12 meters diameter rotor with Within the RealTide project, a full scale blade prototype has
5 composite materials blades, see Figure 1. Each blade has a been produced, see Figure 3, and tested at Ifremer facilities [10].
length of 5 meters and is composed of two half shells, bonded to A special testing frame, see Figure 4, has been designed and
a longitudinal spar (stiffener), with a winglet at the tip. The blade assembled at the Ifremer facilities in Brest, France, to load the
is made of carbon fibre and epoxy resin using an infusion full scale blade under design loads and up to the failure [10]. The
process. root of the blade is bolted to a thick steel plate to provide a
realistic clamping condition.

FIGURE 3: Full scale blade prototype

FIGURE 1: Sabella D12 tidal turbine

The aim of the paper is:


- to compare two loading methods based on: 1) Boundary
Element Momentum Theory (BEMT) by using
StarBlades [4], a Bureau Veritas Tool; 2)
Computational Fluid Dynamic with StarCCM+, a
Siemens software.
- to perform a fatigue analysis by using the Bureau
Veritas methodology [5], [6], [7] and to evaluate the
influence of materials on the fatigue life.

2. FEM MODEL AND PROTOTYPE TESTING


A Finite Element Model (FEM) representing the full scale
Sabella D12 blade has been created in FEMAP, a Siemens
software, as shown in Figure 2. The mesh is made of 4-nodes
shell elements and laminate properties have been imported from
ComposeIT [8], a Bureau Veritas tool for composite application.
Mechanical properties are based on theoretical values indicated
in Bureau Veritas Rules NR546 [9] and have been subsequently
confirmed by tests performed by Ifremer [10]. The different
colours visible on the model, see Figure 2, correspond to FIGURE 4: Full scale blade and test bench, Ifremer facilities
laminate property evolution, the thickness of the skin being
thicker at the root than at the tip. 2.1 Modal Analysis
A modal analysis using the Finite Element Model described
above was carried out with NX Nastran and results are compared
with measurement of eigenvalues. Table 1 shows the
frequencies, in Hz, measured with accelerometers on the full
scale blade and those calculated with the FEM. The blade is
cantilevered, with a clamp at the root and free at the tip.
As measured frequencies are around 10% less than the
calculations and as the weight of the prototype and the weight of
FIGURE 2: Finite Element Model of Sabella D12 tidal turbine the model are similar, modal analysis results indicate that the
blade model is a little bit more rigid than the prototype.

2 © 2022 by ASME
TABLE 1: Blade frequencies, measured vs computed Bureau Veritas. The tool has been programmed in Python and
inputs are given through a Graphical User Interface, see Figure
Frequency (Hz) 6. Inputs necessary for the prediction of forces are:
Mode Type Mode Shape - Geometry of the blade,
measured computed - Lift and drag coefficients,
First - Flow environment:
bending 17.3 19.2 o sea depth,
mode o rotor position,
Second o sea density,
bending 56.7 65.1 o flow velocity,
mode o TSR (Tip Speed Ratio),
Transverse o power law (velocity versus water depth).
bending / 83.4 For the present study, the Sabella D12 turbine is considered
mode installed in Passage du Fromveur, Ushant, France. Two
First operating conditions have been selected, normal operation with
torsional 100.0 110.5 TSR1 and flow velocity U1, and extreme condition with TSR2
mode and flow velocity U2.
Third The fatigue analysis is performed in considering the turbine
bending in normal operation for a period of 25 years and the comparison
110.1 148.6
mode between BEMT and CFD coupling is carried out for the extreme
condition.
2.2 Static Analysis
Loads have been applied at three points along the blade to
reproduce the experimental set-up and in order to compare
simulation and testing results. Various load sequences have been
performed by Ifremer before testing to failure, three will be
examined here: 6 tons, 7.5 tons and 9 tons. The analysis of
difference between the experimental deflection and the Finite
Element model values for the last loading case (9 tons) is less
than 4%. The data has been plotted in the graph illustrated in
Figure 5.
Despite of modal analysis showing a stiffer model, the
measured and computed deflection are similar. The results of the
modal analysis and the static analysis are consistent and validate
the global behaviour of the FE model. FIGURE 6: Distributed BEMT forces

The blade model has been discretised into thirty sections in


StarBlades in order to obtain the variation of axial and tangential
forces along the blade. The FE model is also divided in the same
thirty sections and an element able to distribute force over the
whole section has been created for each of these sections (yellow
elements in the Figure 7) in order to apply forces generated by
StarBlades. Moreover, in order to fit with the wing theory, each
resultant force is applied on a point at 25% along the chord [11].

FIGURE 5: Blade displacement comparison, measured vs


computed

3. BEMT/FEM COUPLING
In this section, loads applied on the blade are estimated with
StarBlades, a tool based on Blade Element Momentum Theory
(BEMT) and developed within the RealTide project [4] by FIGURE 7: BEMT forces distributed along the FEM

3 © 2022 by ASME
4. CFD/FEM COUPLING Figure 9 illustrating the convergence curve shows that the
In addition to BEMT calculations, a CFD analysis has been convergence has been reached from the 4th iteration. The
performed using the extreme condition defined previously. The difference in displacement at the tip between the first (1) and last
coupling achieved between the FEM and CFD models is weak in iteration (7) is around 5.6%. On the other hand, the difference in
the sense that it does not enable dynamic update of the pressure maximum stresses in fibre direction (1,max), in perpendicular
field and resulting blade deformation. One CFD calculation will direction of fibre (2,max), and shear (12,max), does not exceed 7%,
result in a given pressure field to be used in FEMAP, leading to as shown in Table 2. The error, for this application case, between
a given deformation profile. However, the deformation will also a first calculation without any blade deformation consideration
have an influence on the pressure field and a new CFD and a completely deformed blade (around 200 mm at the tips), is
calculation has to be carried out with the new profile. This small in comparison to the CPU time invested for such numerical
creates the need for an iterative approach to reach a converged simulations.
deformation of the blade.
TABLE 2: Comparison of results, iteration 1 vs iteration 7

Iteration Displac. 𝝈𝟏,𝒎𝒂𝒙 𝝈𝟐,𝒎𝒂𝒙 𝝉𝟏𝟐,𝒎𝒂𝒙

1 vs 7 5.6% 4.98% 6.66% 0%

5. COMPARISON BEMT/FEM and CFD/FEM


COUPLINGS
Results obtained by BEMT/FEM coupling and CFD/FEM
coupling have been compared and the maximum displacement
obtained by BEMT is 10% higher than the 7th iteration in CFD.
Axial and tangential forces obtained with both methods have
been also compared, see Figure 10, and the maximum difference
FIGURE 8: CFD-FEM coupling, iteration process is 12% for axial forces and 5% for tangential forces. Such results
are achieved by considering the induced drag effect on tangential
The iterative process applied is presented in Figure 8. At the forces [12]. Indeed, as the winglet is not considered in the BEMT
start, the CFD calculation is launched for the conditions of model with StarBlades, a new calculation has been performed to
interest with the blade in its static position. The obtained pressure evaluate the impact of this parameter on the resultant forces. The
field is extracted and interpolated on the FEM mesh. This output tangential forces have been reduced by 23% and fit now very
is then imported into the FEMAP model which provides the well with CFD results.
resulting blade deformation. A deformation file is generated
which contains the displacement of the FEM mesh nodes. This
file is used in STAR-CCM+ to deform the CFD mesh and a new
CFD calculation is launched with the same inflow condition and
the new deformed blade shape. This process is repeated until
converging toward a stable deformation of the blade indicating
that an equilibrium between the fluid loads and structure loads
has been reached, see Figure 9.

FIGURE 9: CFD-FEM coupling, convergence curve FIGURE 10: Axial (top) and tangential (bottom) forces
obtained from BEMT and CFD computations

4 © 2022 by ASME
6. FATIGUE
6.1 Loads
Loads applied on a tidal turbine blade are due to the fluid
force acting on it while it rotates in the environment. The rotation
of the turbine creates a distribution of fluid force along its length,
and this causes a bending moment on the blades along the axial
flow direction. In operation, the force distribution fluctuates due
to:
1. variations in the inflow velocity (tides, waves,
turbulence),
2. variations in the flow environment (varying density,
viscosity, etc.),
3. variations in the rotational speed of the turbine.
FIGURE 12: Linear interpolation of velocity data
In accordance with [6] the present fatigue analysis considers
the major varying stresses for current velocity spectra calculated The linear approximation of the velocity corresponding to
for both: (1) the low frequency, due to the variation in the inflow the annual tidal coefficients for the design life (25 years) is given
velocity characterized by sinusoidal fluctuations of the tidal in the following spectra, see Figure 13.
stream velocity; (2) high frequency, due to the rotation and the
operational profile of the turbine.

6.1.1 Low frequency cycles


The low frequency cycles, are due to tidal ebb and flow
fluctuations with a period of 12 hours. The tidal coefficient
varies from a minimum of 20 up to a maximum of 120. The
variation of tidal coefficients within the year is obtained from
[13] and is represented as a histogram, in Figure 11, with respect
to their occurrence frequency.

FIGURE 11: Tidal coefficient spectrum

The low frequency velocity is generated by the tidal flow.


The velocity data for a limited number of tidal coefficients was FIGURE 13: Low Frequency velocity spectrum, Ebb Tide (top)
provided by Sabella. The remainder of the data are approximated and Flow Tide (bottom)
by linear interpolation. This method has been proven to give
decent results in terms of accuracy, especially during the design 6.1.2 High frequency cycles
stage of the turbine. The high frequency velocity is linked to the turbine’s
Figure 12 indicates that there is no ebb nor flow dominance rotation. The velocity is non-uniform with sea depth, resulting in
on the site; both are almost symmetrical. a variation of the force distribution during the rotation of the
blade between the highest and lowest points it passes through.
This fluctuation is taken into account by a power law
incorporated into the BEMT load calculations. The velocity
spectra are illustrated hereafter in Figure 14.

5 © 2022 by ASME
FIGURE 14: High Frequency velocity spectrum, Ebb Tide (top)
and Flow Tide (bottom) FIGURE 16: Low frequency bending moment spectra

6.1.3 Bending moment


The maximum stress has been calculated at 2.85m from the
tip with the static analysis. The bending moment at this section
is calculated at different current velocities with StarBlades and
plotted in Figure 15. As observed, the bending moment is
proportional to the velocity squared and can be expressed by the
following formula:
BM = f(U²) (1)
BM = Bending Moment in N.m
U = Flow velocity in m/s

FIGURE 15: Bending Moment vs Flow Velocity at 2.85 m from


the tip FIGURE 17: High frequency bending moment spectra

From equation (1) velocity spectra are converted into 6.2 Ply-by-ply analysis
bending moment spectra at the section at 2.85 m from the tip for The fatigue analysis is based on a ply-by ply approach and has
the stress calculations needed for the fatigue analysis, see Figure been performed in accordance with the methodology described
16 and Figure 17. in [6]. S-N curves for three load ratios have been drawn based on

6 © 2022 by ASME
the static and fatigue experimental results presented by Kawai
and Itoh [14] for unidirectional carbon/epoxy composites, see
Table 3. Curves have been extrapolated to the accurate load ratio
using the Goodman extrapolation as indicated in [6], see Figure
18. Damage is calculated as the ratio of the number of cycles in
each stress bin to the corresponding number of cycles of the S-N
curve at the corresponding stress level. Thus damages of each
bin can be summed using a Miner sum [15].

TABLE 3: Kawai and Itoh experimental results for


carbon/epoxy UDs

𝝈𝒖𝒍𝒕,𝒕 𝝈𝒖𝒍𝒕,𝒄 Load


Alpha
[MPa] [MPa] ratio R*
0.1 (T) 0.033
Fibres
1887 -806.5 -1 (T-C) 0.036
([0°])
10 (C) 0.017
0.1 (T) 0.041
Matrix
30.1 -154 -1 (T-C) 0.054
([90°])
10 (C) 0.005
*: T = Tensile and C = Compression

The load ratios mentioned in Table 3 have been chosen FIGURE 18: S-N curves
based on the low frequency and high frequency conditions. More
explicitly, the low frequency load ratio corresponds to the 6.3 Optimisation
alternating stress case, and is defined by the fluctuations between In order to reduce the cost of the blade, a new design has
the ebb and flow velocities, which oscillates between -0.94 and been proposed mixing carbon and glass fibres. Grogan et Al. [17]
-1.02. The chosen load ratio is thus R = -1. As for the high have already highlighted the interest of using glass and carbon
frequency load ratios, they are calculated based on the Root fibres for the optimisation of tidal turbine blade design. Here, the
Bending Moment (RBM). The obtained load ratios for the ebb number of carbon fibre plies has been reduced by approximately
and flow tides are 0.85 and 1.18 respectively. The tensile-tensile 60% for the skins and the spar is now completely made from
fatigue cycles corresponding to the ebb tide will thus be analysed glass fibre composite. Both static and fatigue analyses have been
with a load ratio R = 0.1, whereas the compressive-compressive performed on this new design and for the fatigue, S-N curves
fatigue cycles will be analysed with a load ratio R = 10. defined in NI603 [16] for glass fibres have been considered.
The static analysis shows that the weight of the blade and
From stress spectra generated for both fibres and matrix, and the maximal displacement increased by 10% and 20%
the fatigue life S-N curves, the linear accumulation of damage respectively. With the exception of 2 plies for which the safety
has been calculated for the two most stressed elements, ply A factor is 5% below the allowable criteria, all stresses are within
and ply B, identified at the upper and lower skin. The cumulative BV criteria.
damage is summarized in the Table 4. The fatigue analysis shows that the most critical area now is
in the spar instead of the skin. The total damage obtained in the
TABLE 4: Cumulative damage in carbon fibre blade skin led to a safety factor of 182 (to compare with 385 with
carbon fibre). On the contrary, the damage calculated in the spar
Cumulative damage during the design life is now significantly greater than 100, leading to an unacceptable
safety factor.
Ply Fibre Matrix
The replacement of carbon fibre by glass fibre seems to be
A 9.74E-05 0.00111 a solution for reducing the cost and respecting the design criteria
Upper skin
B 9.23E-05 0.00111 especially for the skin. However, despite acceptable stresses in
A 0.00011 0.0026 the spar with the static analysis, the fatigue analysis highlighted
Lower skin the lower resistance of glass fibre composite to cyclic loading.
B 0.0001 0.0024
7. CONCLUSION
The damage obtained is less than 1 for the blade made in The present study has been performed on a full scale
carbon fibre and in considering only the operational loads for a composite tidal turbine blade. The comparison of measured
period of 25 years. For such condition, the safety factor is about eigenvalues and deflection with numerical simulations
385.

7 © 2022 by ASME
confirmed the representativeness of the Finite Element Model. [3] Reliability, Availability and Maintainability Assessment
The FEM could therefore be used with confidence in the next report, RealTide Project, https://www.realtide.eu/
analyses. [4] Generalised Tide-to-Wire model, RealTide Project
Hydro-structure couplings have been performed using 2 Consortium
different methods. The first one is based on Blade Element [5] Tomy J.P., Mouton L., Paboeuf S., Comer A., Haldar
Momentum Theory (BEMT) for the prediction of forces applied A.K., Portela A., “Analytical Approach for Global Fatigue of
along the blade. The second, Computational Fluid Dynamics Composite-hull vessels”, MARTECH 2020, Lisbon, Portugal
(CFD) using the Navier-Stokes equation, is more accurate but [6] Paboeuf S., Mouton L., Tomy J.P., Arhant M.,
more time consuming, meshing and CPU, and computes Dumergue N., Davies P., “Application of ply-by-ply fatigue
interaction of the fluid with surfaces. Comparison between the analysis methodology in the design of a full-scale tidal turbine
two showed only a small difference between the results obtained blade”, EWTEC 2021, Plymouth, UK
with BEMT/FEM coupling and CFD/FEM coupling. Despite [7] Mouton L., Paboeuf S., Tomy J.P., Caous D., Valette J.,
some assumptions in BEMT, results seem sufficient at a “Implementing realistic environmental loads on the fatigue
preliminary design stage in order to validate solutions. This testing of a prototype tidal turbine blade”, EWTEC 2021,
approach will be also suitable to optimize blade section profiles Plymouth, UK
or for screening of environmental conditions for the [8] Bureau Veritas Marine & Offshore, “Bureau Veritas
identification of the most critical case. CFD calculations will Software,” 2021. [Online]
give more accurate results but could only be used for a detailed [9] Bureau Veritas, NR546 Hull in Composite Materials and
analysis at a final stage. Plywood, Material Approval, Design Principles, Construction
The BV methodology for fatigue analysis has been applied. and Survey, DT R02 ed., 2021.
Some remaining validation tests need to be carried out to confirm [10] Davies P., Dumergue N., Arhant M., Nicolas E.,
the assumptions but the method is promising due to the limited Paboeuf S., Mayorga P., “Material and structural testing to
number of tests and specimens required. The study also improve composite tidal turbine blade reliability”, EWTEC
highlights the importance of performing a fatigue analysis in 2021, Plymouth, UK
addition to static analysis. The static analysis is not sufficient for [11] H. Abbott, Ira and E. Von Doenhoff, Albert, “Theory of
validating the behavior of composite materials under cyclic Wing Sections”, 1949
loads. Moreover, the study demonstrated the difference between [12] NASA, Induced Drag Coefficient, [Online],
the fatigue resistance of carbon fibre and glass fibre composites. https://www.grc.nasa.gov/www/k-12/airplane/induced.html
[13] Tides4Fishing. [Online] 2021.
https://tides4fishing.com/tides/tidal-coefficient#coef
ACKNOWLEDGEMENTS [14] M. Kawai, N. Itoh, “A failure-mode based
This work were carried out within RealTide project. Authors anisomorphic constant life diagram for a unidirectional
acknowledge all partners of the project: Bureau Veritas, Sabella, carbon/epoxy laminate under off-axis fatigue loading at room
Ifremer, University of Edinburgh, EnerOcean, Ingeteam and 1- temperature”, 2014, Journal of Composite Materials, pp. 571-
Tech. 592.
RealTide project has received funding from the European [15] M. Miner, “Cumulative damage in fatigue”, Journal of
Union’s Horizon 2020 research and innovation programme Applied Mechanics, vol. 12, pp. 149-164, 1945.
under Grant Agreement No 727689. [16] Bureau Veritas, NI603 – Current and Tidal Turbine, DT
R01 E, May 2015
[17] D.M. Grogan, C.R. Kennedy, S.B. Leen, C.
REFERENCES O’Bradaigh, “Design of composite tidal turbine blades”,
[1] RealTide project website, https://www.realtide.eu/ Renewable Energy, vol. 57, pp 151-162, Sept. 2013
[2] Failure Modes Effect Analysis, RealTide Project,
https://www.realtide.eu/

8 © 2022 by ASME

You might also like