Accepted Manuscript

Fatigue failure of a composite wind turbine blade at its root end

Hak Gu Lee, Min Gyu Kang, Ji-sang Park

PII: S0263-8223(15)00687-X
DOI: http://dx.doi.org/10.1016/j.compstruct.2015.08.010
Reference: COST 6708

To appear in: Composite Structures

Please cite this article as: Lee, H.G., Kang, M.G., Park, J-s., Fatigue failure of a composite wind turbine blade at its
root end, Composite Structures (2015), doi: http://dx.doi.org/10.1016/j.compstruct.2015.08.010

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Fatigue failure of a composite wind turbine blade at its root end
Hak Gu Lee*, Min Gyu Kang, Ji-sang Park

Wind turbine Technology Research Center, Korea Institute of Materials Science,

797 Changwondaero, Changwon, Gyeongnam, 641-831, Republic of Korea

Abstract

As blade failures at wind farms have increased, the structural safety of composite wind

turbine blades is ever more important. The recent implementation of considerably larger

blades has made the problem even more crucial. One of the critical failure modes is the

blade root failure, which can result in the blade being pulled out from its wind turbine

during operation. In this study, we experienced delamination failure at the blade root

during fatigue testing of a 3MW full-scale wind turbine blade according to international

standard IEC 61400-23: Full-scale structural testing of rotor blades. Comparing the

measured data with the FE analysis results, we simulated the situations the blade had

experienced, and then found what caused the delamination failure as well as the

problem of the conventional design approach. The bumping motions of the blade shell

caused by geometric complexities between the maximum chord and the root alter

significantly the load distribution at the end of the blade root. Therefore, to enhance the

structural safety of a large composite wind turbine blade, a more detailed FE analysis on

the blade root in the design stage is needed.

Keywords: Fatigue; Failure; Delamination; Wind turbine blade; Bumping motion

*
Corresponding author.
1/24
Tel.: +82-55-280-3261; fax: +82-55-280-3498.
E-mail address: [email protected] (H.G. Lee).
1. Introduction

With the recent trend toward large slender wind turbine blades, questions are being

raised regarding their reliability. In order to evaluate the static strengths and fatigue

lives of these larger blades, static and fatigue tests of full-scale prototype blades should

be conducted according to international standards or equivalent guidelines [1-4]. Testing

methodologies developed so far are well described in the two review papers of Malhotra

et al. [5] and Yang et al. [6].

Previous studies pertaining to static strength of a full-scale wind turbine blade are as

follows. Jensen et al. [7] tested a 34 m composite wind turbine blade until its structural

collapse. Debonding of the outer skin was the initial failure mechanism, followed by

delamination buckling which led to the blade’s collapse. Jensen et al. believed the main

root cause was the Braizer effect of the shell structure due to bending. Overgaard et al.

[8-9] carried out a static flapwise bending test of a 25 m wind turbine blade to collapse.

The Brazier effect had a large influence on the local out-of-plane deflection, but its

influence on the longitudinal strain level in the primary load-carrying laminate was

insignificant. Overgaard et al. assert that the structural stability of the generic wind

turbine blade has been governed by buckling and the delamination phenomena. Yang et

al. [10] conducted a static test of a 40 m wind turbine blade under flapwise loading to

collapse. Yang et al. concluded the Brazier effect was not the dominant failure

mechanism, but debonding between the pressure-side and the suction-side aerodynamic

shells was the initial failure mechanism followed by its instable propagation which

leads to collapse.

Previous studies pertaining to fatigue of wind turbine blades are divided into two

categories: a material fatigue behavior and a structural fatigue behavior. The material

2/24
fatigue behavior has extensively studied with uni-axial, in-plane loading of balanced

and symmetrical, relatively thin laminates [11-12], but they are only remotely

representative for blade structures [13]. Tests with a full-scale wind turbine blade to

study the structural fatigue behavior are so expensive that few studies have been

conducted to date. Leeuwen et al. [14] had carried out fatigue tests of 37 wind turbine

blades 3.4 m in length as well as 35 coupons to compare fatigue strength from full-scale

blade tests with coupon-based predictions. Flapwise failures occurred at the tensile side,

but edgewise failures were the result of crack initiation starting in the bonding at the

trailing edge followed by further crack propagation in the laminate. Blade fatigue data

compared with coupon data fitted reasonably with flapwise tests, but they did not

compare well for edgewise tests. Marín et al. [15-16] inspected fatigue damage of a

300kW wind turbine blade, and then performed a FE analysis to reveal the root cause of

the fatigue damage. The crack initiated at the abrupt geometric-transient region between

the root zone and the aerodynamic zone had been propagated into the laminate. It

should be noted that the aforementioned studies dealt with small wind turbine blades

that have relatively higher fatigue margins than the large slender wind turbine blades

presently in development.

Another approach to enhance the reliability of a wind turbine blade is an FE fatigue

simulation for the blade. Kong et al. [17-18] designed a 750 kW wind turbine blade,

factoring in its fatigue life of 20 years based on the well-known S-N linear damage

equation, the load spectrum, and Spera’s empirical equations. Shokrieh et al. [19]

performed a case study with a 23 m wind turbine blade. Using its FE shell model with a

stochastic approach on fatigue loads, the fatigue life was bounded between 18.66 years

and 24 years as lower and upper limits. Toft et al. [20] estimated the reliability of a wind

3/24
turbine blade for a single failure mode, considering statistical uncertainties.

Despite such contributions by many researchers, about 30 blade failures are occurring

per year throughout the world, and the number of blade failure are increasing over time

[21]. There have been blade failures not observed in previous studies which have

occurred in the field at Eclipse wind farm and Ocotillo wind farm in 2013. In these

instances, the wind turbine blades were pulled out from the wind turbines due to

delamination at the root. The authors of this paper have also experienced a similar

phenomenon during fatigue testing of a 3MW full-scale wind turbine blade. To find a

root cause of the phenomenon, loading conditions calculated by its FE shell model were

applied to the more detailed FE solid model of the root subcomponent. Comparing the

analysis results with measured strain data of the T-bolt, we adjusted loading conditions

of the FE solid model to simulate deformations and stress distributions of the blade root.

Based on the simulation results, this study has found one of the plausible root causes

able to incur delamination at the root of a wind turbine blade.

2. Test blade and its failure during fatigue testing

The test blade is a 3MW glass/epoxy composite blade as shown in Fig. 1. Blade length

and weight are 56 m and 14.5 ton, respectively. It has been developed as a result of a

R&D project funded by Korean government. After mounting the test blade on a stand

fixture like a horizontal cantilever beam, the fatigue test setups including an

aerodynamic fairing, two additional masses, a flapwise exciter, and an edgewise exciter

were attached on the test blade. Then we carried out a dual-axis resonance fatigue test of

the blade using two different resonance frequencies according to international standard

IEC 61400-23 [1]. The tip motion of the test blade during dual-axis resonance fatigue

4/24
testing is shown in Fig. 2. After flapwise 510,000 cycles under the equivalent amplitude

of 5,352 kNm and the mean of 5,970 kNm and edgewise 780,000 cycles under the

equivalent amplitude of 4,454 kNm and the mean of 0 kNm at the end of the blade root,

fatigue failure was found as shown in Fig. 3.

Strain values of the T-bolt, located as shown in Fig. 4, were measured during the fatigue

testing. The shape and specification of the T-bolt are represented in Fig. 5 and Table 1.

The graphs in Fig. 6 are the measured strain values of the two different strain gages

attached at the same cross section of the T-bolt. In each graph the amplitude of tensile

strains larger than that of compressive strains means separation of the T-bolt joint has

occurred. Furthermore, the tensile or compressive amplitude of strain gage 2 being

larger than that of strain gage 1 means bending of the T-bolt. Separation and bending of

the T-bolt joint are unexpected phenomena that must be avoided in the design stage.

Before going into the detailed root cause analysis, the conventional approach for a blade

root design will be explained in the next section to understand why the approach failed

to predict the separation and bending observed.

3. Conventional blade root design

Fig. 7 represents a schematic diagram of a blade root part including a pitch bearing

assembled by T-bolts and nuts. The blade root is a very thick composite laminate able to

enclose T-bolts and cross nuts, its thickness being about 100 mm. Thus, blade designers

have believed that, compared with blade shell sandwich structures whose thickness are

less than about 30 mm, a blade root is so stiff that the stress distribution in it is similar

to that of a hollow circular cylinder structure when subjected to bending. Based on this

presumption, the distributions of local moments and axial stresses or forces have been

5/24
calculated as shown in Fig. 7. This conventional approach for the test blade gives 14

kNm for the maximum local moment and 267 kN for the maximum axial force at each

T-bolt joint. The axial force is much smaller than the pretension 340 kN for the T-bolts

in Table 1. Thus the separation and bending of the T-bolt cannot be observed during the

fatigue testing based on this calculation. The presumption regarding the blade root had

worked well before wind turbine blades became larger and more slender. However, the

current trend is requiring a more detailed analysis on the blade root part.

4. FE simulation for root cause analysis

A flowchart of the root cause analysis carried out in this study is represented in Fig. 8.

Using the test setup and the loading conditions aforementioned in section 2, a static

analysis with the FE shell model for the test blade was conducted to calculate the

plausible amplitudes of an axial force and a local moment at each T-bolt. The calculated

values do not represent the real situation during the fatigue testing because the clamped

boundary conditions used in the FE shell model does not match with the separation of

the T-bolt joint observed. The axial force and the local moment were applied to the

subcomponent FE solid model for the test blade root, and then we modified them,

comparing the calculated T-bolt strains with the measured strains during the fatigue

testing. After several modifications, stress distributions at the end of the blade root were

able to explain the observed delamination that occurred.

The information on the FE shell model for the test blade is in Fig. 9 and Table 2. The

shell model reflects the shape of the test blade and laminating sequences of composites.

We used a commercial FE solver, ABAQUS 6.13, and its 4 node shell element, S4R. The

number of the elements is 57,969. The boundary condition was the clamped condition at

6/24
the end of the blade root, and the loading conditions were the flapwise and the edgewise

test bending moment distributions along the positive y- and x-directions, where the

positive y means a chord direction toward the trailing edge in the pitch zero section and

the positive x means the cross product of the positive y with the pitch axis.

Properties of unidirectional glass NCF/epoxy composites used in this study are in Table

4. Four properties were measured from coupon tests: E1 of 40.14 GPa, E2 of 12.30 GPa,

v12 of 0.26, and G12 of 3.40 GPa, where the subscripts 1, 2, 3 mean the fiber direction,

the transverse direction, and the thickness direction, respectively. A transversely

isotropic material needs 5 independent material properties, so we assumed v23 of glass

NCF/epoxy composites as 0.38 compared with S-2 glass/epoxy composites in reference

22, the ratio of 1.48 between two Poisson’s ratios v12 and v23. Then G23 can be

calculated from equation (1), resulting in 4.44 GPa, and the other properties E3, v13, and

G13 are the same as E2, v12, and G12, respectively.



 = (1)



The information on the subcomponent FE solid model for the test blade root is in Fig.

10 and Table 3. ABAQUS 6.13 was also used as the FE solver for the subcomponent

model, which was constructed with the 8 node solid element, C3D8I. The number of the

element was 53,629. The boundary conditions used in the model were the axially

symmetric and symmetric condition along the hoop direction, the fixed condition along

the blade length (spanwise) direction on the positions of two bearing ball arrays, and the

contact conditions with the friction coefficient of 0.3 on the several contact surfaces of

the T-bolt and the nuts. The loading conditions were the axial force and the local

moment calculated by the FE shell model, which were applied to the cross section of the

blade root part along the spanwise direction and the hoop direction, respectively.

7/24
Equivalent orthotropic properties of the blade root laminate whose stacking sequence is

[45/0/-45]n were generated for the convenience of FE modeling with solid elements.

Classical laminated plate theory (CLPT) cannot calculate whole equivalent orthotropic

properties because even interlaminar stresses at the interface of two laminae are

discontinuous [23]. CLPT gives us only in-plane laminate properties such as Ex, Ey, vxy,

and Gxy and through-thickness Poisson’s ratios such as vxz and vyz, which are calculated

from equations (2), (3), and (4) [22].

  
      
 = (2)



  
      
 = (3)



̅ !" + 
 = ∑'( ̅ !" + $
̅ !"$ % & ( i = 1, 2, 6 ) (4)

̅ , !"* , % , and 2H are the i, j component of A matrix, the i, j component

where )* , *

of the transformed compliance matrix in the k-th lamina, the i, j component of the

transformed reduced stiffness matrix in the k-th lamina, the thickness of the k-th lamina,

and the total thickness of the laminate, respectively. The six properties were calculated

using Table 4 and the stacking sequence [45/0/-45]s, in which the thicknesses of 45/-45

lamina and 0 lamina were 0.15 mm and 0.60 mm, respectively. The large staking

number of the unsymmetric laminate [45/0/-45]n makes its properties converge into

those of the symmetric laminate [45/0/-45]s, so we used the symmetric stacking

sequence instead of the unsymmetric one. The remaining three properties, Ez, Gxz, and

Gyz were calculated from a FE cube model whose stacking sequence is [45/0/-45]10s. By

applying normal forces or shear forces to the cube surfaces, we obtained a pertinent

deformation value at each case. From the deformation value and loading conditions, the

cube stiffness was calculated at each case. Table 5 shows the calculated equivalent

8/24
orthotropic properties of the blade root laminate.

5. Analysis results

The analysis result of the FE shell model shows that the maximum values of the axial

force and the local moment per each T-bolt are greater than those of the conventional

model. As shown in Fig. 11(a), the blade root end near 110 degrees from the leading

edge receives severe tension, 387 kN, even larger than pretension of the T-bolt, 340 kN,

resulting in the separation of the T-bolt joint. The local moment and the axial force in

the FE shell model greater than those in the conventional model in Fig. 11 come from

the inward bumping motion in Fig. 12, and the lower values come from the outward

bumping motion. The strangest location for these differences is the trailing edge. The

location is expected to move inward because the flapwise and the edgewise bending

moment are applied along the positive y- and the positive x-direction. However, it

moved outward, as shown in Fig. 12, resulting in a small axial force and a small local

moment. This opposite moving direction at the trailing edge may be caused by the

geometric complexities from the maximum chord of the blade to the root. In the same

cross section of the blade, the alleviation of the axial force and the local moment in

some location incurs an increase of the axial force and the local moment in another

location. Thus the bumping motions of the blade are thought to be the main reason of

the unfavorable load distribution at the end of the blade root.

The subcomponent FE solid model simulated well the T-bolt joint when applying 100%

of the axial force and 73% of the local moment at the location of 90 degrees. The

measured and calculated strain values at the top and bottom positions of the T-bolt are

shown in Fig. 13; the measured strain value at the top position was extrapolated based

9/24
on the two measured data of different strain gages. The simulation shows us partial

separation between the blade root and the pitch bearing as shown in Fig. 14.

Furthermore, the interlaminar shear stress, σ23, and the peel stress, σ33, are very severe

when the partial separation occurs. The positons of the two severe stresses are well

matched with the observed delamination positions as shown in Fig. 15. Therefore, we

conclude that the delamination at the end of the blade root due to partial separation

followed by crack propagation into the root laminate, as shown in Fig. 16, would bring

about the failures in which blades are pulled out from their wind turbines.

6. Conclusion

In this study we have experienced delamination failure at the end of the blade root

during its full-scale fatigue testing. To find what caused the failure, FE analyses were

carried out using a subcomponent FE solid model as well as a full-scale FE shell model.

The analysis results reveal that for a slender and large wind turbine blade the real load

distribution at the root is very different from that calculated by the conventional

approach, which assumes the blade root has enough stiffness to be modeled as a

bending of a hollow circular cylinder. The bumping motions of the blade shell alter load

distribution at the end of the blade root, resulting in the alleviation of load in some

locations and the increase of load in other locations. The severe increase of load incurs

partial separation of the T-bolt joints followed by delamination at the end of the root,

which may lead to pulling of the blade out from its wind turbine during operation.

Therefore, detailed analyses on the blade root should be carried out to enhance its

structural safety especially for a slender and large wind turbine blade.

10/24
ACKNOWLEDGEMENTS

This work was supported by the New & Renewable Energy of Korea Institute of Energy

Technology Evaluation and Planning (KETEP) grant funded Korea government

Ministry of Trade, Industry and Energy (No.2012T100201707).

REFERENCES

[1] International Electrotechnical Commission. International Standard IEC 61400-23

Wind turbines – Part23: Full-scale structural testing of rotor blades, 2014.

[2] Det Norske Vertias. Standard DNV-DS-J102 Design and manufacture of wind

turbine blades, offshore and onshore wind turbines, 2010.

[3] Germanischer Lloyd. Rules and Guidelines IV Industrial Services 1 Guideline for

the Certification of Wind Turbines, 2010.

[4] Germanischer Lloyd. Rules and Guidelines IV Industrial Services 2 Guideline for

the Certification of Offshore Wind Turbines, 2012.

[5] Malhotra P, Hyers RW, Manwel JF, McGowan JG. A reveiw and design sutdy of

blade testing systems for utility-scale wind turbines. Renewable and Sustainable

Energy Reveiws 2012;16:284-292.

[6] Yang B, Sun D. Testing, inspecting and monitoring technologies for wind turbine

blades: A survey. Renewable and Sustainable Energy Reviews 2013;22:515-526.

[7] Jensen FM, Falzon BG, Ankersen J, Stang H. Structural testing and numerical

simulation of 34 m composite wind turbine blade, Composite Structures

2006;76:52-61.

[8] Overgaard LCT, Lund E, Thomsen OT. Structural collapse of a wind turbine blade.

11/24
 Part A: Static test and equivalent single layered models. Composites: Part A

2010;41:257-270.

[9] Overgaard LCT, Lund E. Structural collapse of a wind turbine blade. Part B:

Progressive interlaminar failure models. Composites: Part A 2010;41:271-283.

[10] Yang J, Peng C, Xiao J, Zeng J, Xing S, Jin J, Deng H. Structural investigation of

composite wind turbine blade considering structural collapse in full-scale static

tests. Composite Structures 2013;97:15-29.

[11] Kensche CW. Fatigue of composites for wind turbines. International Journal of

Fatigue 2006;28:1363-1374.

[12] Mandell JF, Samborsky DD, Agastra P, Sears AT, Wilson TJ. Analysis of

SNL/MSU/DOE fatigue database trends for wind turbine blade materials. Sandia

National Laboratory report: SAND2010-7052, 2010.

[13] Nijssen RPL. Fatigue life prediction and strength degradation of wind turbine

rotor blade composites. Sandia National Laboratory report: SAND2006-7810P,

2006.

[14] van Leeuwen H, van Delft D, Heijdra J, Braam H, Jørgensen ER, Lekou D, Vionis

P. Comparing fatigue strength from full scale blade tests with coupon-based

predictions. Transactions of the ASME 2002;124:404-411.

[15] Marín JC, Barroso A, París F, Cañas J. Study of damage and repair of blades a 300

kW wind turbine. Energy 2008;33:1068-1083.

[16] Marín JC, Barroso A, París F, Cañas J. Study of fatigue damage in wind turbine

blades. Engineering Failure Analysis 2009;16:656-668.

[17] Kong C, Bang J, Sugiyama Y. Structural investigation of composite wind turbine

blade considering various load cases and fatigue life. Energy 2005;30:2101-2114.

12/24
[18] Kong C, Kim T, Han D, Sugiyama Y. Investigation of fatigue life for a medium

scale composite wind turbine blade. International Journal of Fatigue

2006;28:1382-1388.

[19] Shokrieh MM, Rafiee R. Simulation of fatigue failure in a full composite wind

turbine blade. Composite Structures 2006;74:332-342.

[20] Toft HS, Sørensen JD. Reliability-based design of wind turbine blades. Structural

Safety 2011;33:333-342.

[21] http://www.caithnesswindfarms.co.uk/AccidentStatistics.htm , 2015.

[22] Herakovich CT. Mechanics of fibrous composites. New York: John Wiley & Sons,

Inc., 1998. p. 14, 112-183.

[23] Reddy JN. Mechanics of laminated composite plates theory and analysis. New

York: CRC Press, Inc., 1997. p. 651-657.

13/24
 Fig. 1 3MW test blade 56 m in length.

Fig. 2 Tip motion during dual-axis resonance fatigue testing.

14/24
Fig. 3 Failure at the end of the blade root.

15/24
Fig. 4 Location of the T-bolt where bolt strains were measured.

Fig. 5 Schematic diagram of the M36 T-bolt used in this study.

16/24
 (a) (b)
Fig. 6 Strain values measured by two different strain gages attached at the same cross
section of the T-bolt: (a) strain gage 1 and (b) strain gage 2.

Fig. 7 Schematic diagram of a blade root part and a pitch bearing.

17/24
Fig. 8 Flowchart of the root cause analysis taking the fatigue testing conditions into
account.

18/24
Fig. 9 FE shell model used in this study.

Fig. 10 FE solid model used in this study.

19/24
 (a) (b)
Fig. 11 Comparison between the results of the FE shell model and those of the
conventional approach: (a) axial force distribution and (b) local moment distribution.

Fig. 12 Bumping motion of the blade shell during the fatigue testing.

20/24
Fig. 13 Comparison between the measured and the calculated strain ranges.

Fig. 14 Partial separation between the blade root and the pitch bearing.

21/24
 Fig. 15 Stress distributions that incur delamination at the end of the blade root.

Fig. 16 Schematic diagram of the blade root failure caused by delamination followed by
crack propagation.

22/24
Table 1 Specifications of the M36 T-bolt used in this study
Min. diameter Pretension Prestress Min. yield strength
Grade
[mm] [kN] [MPa] [MPa]
10.9 28 340 552 940

Table 2 FE shell model for the test blade

FE solver Element type No. of Boundary condition
the elements
4 node
ABAQUS 6.13 shell element 57,969 Clamped condition at the root
(S4R)

Table 3 Subcomponent FE solid model for the test blade root

FE solver Element type No. of Boundary conditions
the elements
Axially symmetric condition
8 node
Symmetric condition
ABAQUS 6.13 solid element 53,629
Fixed condition
(C3D8I)
Contact condition (µ = 0.3)

23/24
Table 4 Properties of the glass NCF/epoxy unidirectional lamina
E1 E2 E3 G12 G13 G23
v12 v13 v23
[GPa] [GPa] [GPa] [GPa] [GPa] [GPa]
40.14 12.30 12.30 0.26 0.26 0.38 3.40 3.40 4.44

Table 5 Properties of the [45/0/-45]n laminate

Ex Ey Ez Gxy Gxz Gyz
vxy vxz vyz
[GPa] [GPa] [GPa] [GPa] [GPa] [GPa]
30.69 13.44 12.70 0.42 0.21 0.34 6.26 3.55 4.21

24/24