Design, Fatigue Test and NDE of a

Sectional Wind Turbine Rotor Blade

F. H

AHN

,

1

C.W. K

ENSCHE

,

1

R.J.H. P

AYNTER

,

2

A.G. D

UTTON

,

2

*

C. K

ILDEGAARD AND

J. K

OSGAARD

3

1

DLR, Pfaffenwaldring, Stuttgart, Germany

2

CLRC, Rutherford Appleton Lab., Chilton Didcot,

Oxon., United Kingdom

3

LM Glasfiber A/S, Lunderskov, Denmark

ABSTRACT: Future sizes of wind turbine rotor blades will exceed 50 m. For
transportation, it is favourable to make them in two parts and connect them in a
suitable way at the operational site. With the co-operation of industry and research
institutes in three countries, a spar beam-connection principle was selected as a
possible solution.

The pre-design, including finite element analysis, for the structural details were

carried out at DLR, Stuttgart (Germany). LM Glasfiber A/S (Denmark) realised the
concept in a 13.4 m GRP blade with a wound tube as a transverse load and bending
moment transferring spar stump. The sectional blade was investigated for static and
fatigue integrity at DLR up to 5 million load cycles in a sinusoidal one-step test.
Strain gauges were applied at those locations shown by the finite element analysis to
be critical.

During the fatigue tests, the blade was observed by means of a thermoelastic stress

analysis camera (TSA) by CLRC (United Kingdom) with the aim to observe the
stress distributions. These measurements located stress concentrations not otherwise
predicted by the finite element analysis. At those locations more strain gauges were
applied and found to show relatively high stresses. The strain measurements can be
used to calibrate the signals shown by the TSA camera.

Thus, on the basis of the combined use of different design and measurement

methods, a promising way is shown to find stress ‘‘hot spots’’ in complex composite
components and to inform directly and immediately the manufacturers of those
articles about possible or necessary modifications.

Journal of

T

HERMOPLASTIC

C

OMPOSITE

M

ATERIALS,

Vol. 15—May 2002

267

0892-7057/02/03 0267–12 $10.00/0

DOI: 10.1106/089270502021455

ß 2002 Sage Publications

[12.3.2002–12:21pm]

[267–278]

[Page No. 267]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

*Author to whom correspondence should be addressed.

INTRODUCTION

I

N ORDER TO

investigate the structural integrity of a sectional wind turbine

rotor blade, a bending spar connection principle was chosen, as used for

example in wings of sailplanes and motor gliders. In order to test the
connection principle, the ‘‘tube connection’’ was embedded into an LM
13.4 m standard blade. This should give a realistic load introduction in the
connecting structure. The blade was tested statically and in fatigue (with
the same loads as calculated for an undivided blade).

DEVELOPMENT OF THE CONNECTION

Pre-design

The pre-design was done by DLR. The outline of the design is given in

Figure 1.

A GRP tube is fixed in the outer part of the blade by two ribs. This tube

comes out of the blade and fits into another GRP tube, which is embedded
in the inner part of the blade in the same manner. This method is used to
mount the wingtips to the rest of the wing in many sailplanes. The locking
device against the flying forces used in sailplanes (where no centrifugal
loading exists) is a snap bolt. For rotor blades, the locking device has to be
constructed more massively. Two shear bolts prevent the blade halves from
rotating in relation to each other.

Ribs for load distribution

and tube embedding

Snap bolt

Outer and Inner tubes

Blade Tip

View from leading edge

View from top

Figure 1. Pre-design for the tube connection.

268

F. H

AHN ET AL

.

[12.3.2002–12:22pm]

[267–278]

[Page No. 268]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

Finite Element Calculation

DLR did an FE calculation of the blade [1] with the embedded new

structure for the dividing mechanism. This FE calculation determined the
lay-up for the new connecting parts. The FE model was modified, until the
resulting strains were lower than the allowable strains given by LM
Glasfiber. The strains in longitudinal direction for flapwise loading are
shown in Figure 2. The values on the right side of the contour plot are in
percent of the allowable strain.

The results of the FE calculation were then used to determine the final

lay-up of the blade, as it was built. Thus we had the possibility to compare
the FE results with reality during the test.

Final Design

LM Glasfiber developed the final design from this pre-design. Instead of

transferring the loads by ribs from the tube into the shell, LM Glasfiber
decided to use a construction with which they have good experience with
their method of tip brake attachment.

In the outer part of the blade, the space between the two main shear webs

and the tube is filled with two blocks of foam coated with GRP. In the inner

Figure 2. Finite element results (strains in lengthwise direction). The outlines show: the line
of the main shear webs and the location of the connecting tube. The highest stress is
predicted just outboard of the connecting tube (label MX).

Design, Fatigue Test and NDE of a Rotor Blade

269

[12.3.2002–12:22pm]

[267–278]

[Page No. 269]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

part, two steel rings of different diameter are embedded between similar
blocks. The tube coming out of the outer blade is conical with two
cylindrical bushes at the inner and the outer end. Two shear bolts in the
bulkheads prevent the blade halves from rotating in relation to each other.

Figure 3 shows the two blade halves being connected. The two cylindrical

sections of the GRP tube fit into two steel rings in the inner section of the
blade.

MECHANICAL TESTING OF THE BLADE

The LM 13.4 m tube connected sectional blade was tested in three stages.

First it was tested statically in the flapwise direction towards the suction side
and in edgewise directions (both towards the leading edge and the trailing
edge). Then it was tested in fatigue in the flapwise direction. Finally, residual
strength tests were performed with the same loading conditions as for the
initial static tests.

Static Testing

The blade was tested to the design loads of an undivided LM 13.4 m blade

in flapwise and in edgewise direction. Forty strain gauges were applied in
order to control the tests and to detect possible failure in the structure. Thus
also the opportunity to compare the results with the behaviour of the
undivided blade was available.

The blade went through the static tests without any failure. The maximum

measured strains were relatively low (about 0.3%). Because of the higher
flapwise loading and the cylindrical shape of the bending spar, the flapwise

Figure 3. Connecting the blade halves.

270

F. H

AHN ET AL

.

[12.3.2002–12:22pm]

[267–278]

[Page No. 270]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

load case dominates. For this reason, the dynamic test was carried out in
this orientation.

Dynamic Testing

The blade was tested in fatigue to 5 million load cycles with an R-value of

0.27 at the separation 4.5 m from the root. An exciter drove the blade at

its resonant frequency, which was found to be about 2 Hz. The exciter was
mounted to the blade on a steel beam construction clamped to the blade, see
Figure 4. In order to get the correct moment distribution, extra mass was
added to the blade tip by extending the steel beam construction. With the
applied strain gauges, the moment distribution during the dynamic set-up
could be controlled.

The moment range during the fatigue test was measured continuously by

an accelerometer fixed to the blade. The deflection range d can easily be
calculated from the frequency f and the acceleration range a:

d ¼

a

!

2

¼

a

2f

ð

Þ

2

ð

1Þ

From the calibration of the dynamic test the moment range correspond-

ing to the deflection range is known. Figure 5 shows the moment range at
the separation versus the load cycles. Figure 6 shows the frequency
distribution of the moment range. The target moment range was 84.72 kNm.

The resonance frequency, as a means to define a possible stiffness change,

was monitored during the dynamic test. Since there was only a marginal
change during the whole test and the resonance frequency is very sensitive to
the stiffness of the blade, no decrease in stiffness could be stated.

Figure 4. Blade ‘‘in action’’.

Design, Fatigue Test and NDE of a Rotor Blade

271

[12.3.2002–12:22pm]

[267–278]

[Page No. 271]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

The only failure during the whole dynamic test was a screw (of bad

quality) within the locking device against the fly forces. After the screw was
replaced by one of better quality, this problem was solved.

Residual Strength Testing

The blade was loaded with the same conditions as the initial static tests. It

could not be broken, because the test rig at DLR was not able to bear the
loads of a destructive test. The tests before and after the dynamic test were
compared in order to check if the blade suffered some structural failure.

The residual strength tests were again in flapwise direction towards the

suction side and in both edgewise directions.

Figure 7 compares the plots of load versus deflection for the flapwise load

case. The two curves lie exactly upon each other, which also indicates that
there was no decrease in the stiffness of the blade.

The measured strains for the same tests were also compared. The strains

were normalised to the maximum load of the first static test in order to

0

500

1000

1500

2000

2500

75

80

85

90

95

Moment at 4,5 m from root [kNm]

Number of datapoint

s

Figure 6. Frequency distribution of moment range.

0

10

20

30

40

50

60

70

80

90

1

2

3

4

5

Number of load cycles [10

6

]

Moment at seperation [kNm]

Figure 5. Moment range at separation versus load cycles.

272

F. H

AHN ET AL

.

[12.3.2002–12:22pm]

[267–278]

[Page No. 272]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

make the values comparable. As with the load versus deflection curve the
strains did not show any change due to the test procedure. The blade
survived all tests, which were the same as for an undivided blade, without
failure.

THERMOELASTIC MEASUREMENTS

Personnel of CLRC carried out the thermoelastic measurements during

the dynamic tests. A summary of the measurement method is given followed
by the application to the blade tests.

Thermoelasticity

This phenomenon was predicted by Lord Kelvin (William Thomson) in

the mid-nineteenth century with his theory on the thermodynamics of solid
materials under stress. When an elastic material is stressed under adiabatic
conditions, the temperature changes; in most materials the temperature
increases when compressed and drops when stretched. Lord Kelvin’s
theories were subsequently shown experimentally with progressing accuracy
over the next hundred years.

The main relationships have been developed for isotropic materials and

can be combined to give, for the transient temperature change:

T ¼

T

C

p

,

ð

2Þ

where , and C

p

are the material’s coefficient of thermal expansion,

density and specific heat capacity respectively, T is the temperature of the

–10

0

10

20

30

40

50

–100

0

100

200

300

400

500

Displacement [mm]

Load [kN]

before dyn. test
after dyn. test

Figure 7. Comparison of load/deflection curves – the two curves are almost identical, thus
cannot be seen separately.

Design, Fatigue Test and NDE of a Rotor Blade

273

[12.3.2002–12:22pm]

[267–278]

[Page No. 273]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

material, and is the applied total direct stress transient. Shear stress does
not contribute to the effect. A typical value for steel is about 0.0009 K per
MPa applied at a temperature of 300 K, so the effect is small.

For orthotropic materials such as directional composites, an alternative

proposed relationship is:

T ¼

T

C

p

ð

L

11

þ

M

22

Þ

:

ð

3Þ

This is appropriate for observation at a surface, thus

33

is not included

because it is necessarily zero, the other directions (11 and 22) refer to the
principal directions of the lay-up. The parameters L, M are intended to
combine the differing expansion coefficients, moduli and Poisson’s ratios
and must be found from measurement.

This and other proposed relationships for application to composite

materials are under investigation during this and other work by the
investigators of CLRC.

Infra-Red Radiometry

Detection of the temperature differences is difficult, because they are small

and short-lived, due to conduction into the rest of the specimen and convection
into the surroundings. The method used here is to use fairly fast repeated
dynamic loading of the specimen (typically sinusoidal) and detect the
temperature amplitude from the infra-red radiation from the surface.

The history of IR radiometry for elasticity measurements starts with

M.H. Belgen working for NASA in 1967. In 1978, a joint patent was granted
to MoD/Sira Ltd in the United Kingdom which led to a commercial system
called ‘‘SPATE’’ (Stress Pattern Analysis by measurement of Thermal
Emission). To generate an image of a stress field, the SPATE system uses a
single detector and mirrors to scan the field of point by point measurement.

The system used in this work is a ‘‘DeltaTherm’’ which has an array of

many detectors (128 by 128) scanned electronically at about 430 frames per
second. Because the temperature changes are small (in the order of 0.1 K)
and the noise associated with infra-red detection is relatively high, the
measurement is made by signal processing (in a separate electronics system)
of the detector output, typically requiring about 5 s each. The main advantage
of the DeltaTherm is that it can process all the sensors in this time compared
with SPATE which would require about 23 h (128 128 5 s) to produce an
equivalent image. See [2] for a more detailed treatment.

The user interface to the system is via a PC which controls the

measurement system and provides data presentation and storage.

274

F. H

AHN ET AL

.

[12.3.2002–12:22pm]

[267–278]

[Page No. 274]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

Observations of the Blade

The camera (which is about the size of a 1970s video camera) and a small

mirror (coated for infra-red use) were placed on the floor below the blade
under test to view the lower surface. As the view is limited, larger coverage
can be obtained by setting images next to each other in post processing (with
a library of MATLAB subroutines prepared by CLRC).

Two such images of data sets are presented in Figure 8. These correspond

to points where the connecting tube is fixed at either end. The height of the
main feature is the zone of attachment in the inside of the blade skin. It is
characterised by a zone of low stress, presumably due to the extra material
(glues and local reinforcement) here, and a surrounding zone of higher stress.

The immediate use of the images is a fast general view of the stress field,

which shows substantial detail and can thus guide location of strain gauges.
Extra gauges were placed at ‘‘hot-spots’’ in the stress pattern.

Calibration of data

It is not possible to give an absolute calibration of the measured data

because there are so many factors in the relationship. Thus a compromise is

(a)

(b)

Figure 8. Thermoelastic data at the ends of the connecting tube. (a) Blade root end,
(b) Blade tip end. Brighter shows greater total stress. The horizontal line at zero is the
centreline between the shear webs. The sloping line at about 300 in (b) is the leading edge.
Other small features are strain gauges, wires and markers on the blade surface.

Design, Fatigue Test and NDE of a Rotor Blade

275

[12.3.2002–12:22pm]

[267–278]

[Page No. 275]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

to find a linear relationship (find L and M of Equation 3) with strain gauge
measurements. Thus the measurement at these points can be extended to the
whole field of view.

COMPARISON WITH STRESS MODELLING

Once the calibration has been found, comparison can be made with model

results by combining the predicted stresses in each direction with
the estimated coefficients (L and M) to produce an equivalent stress
intensity map.

This gave the general findings:

.

The thermoelastic measurements find the same large features as model
results.

.

The full-field thermoelastic measurement can find small details that may
not be predicted by structural modelling and can be used as data when
refining models.

.

The full-field thermoelastic measurement can find details impossible to
find with a few point measurements such as strain gauges.

Modelling is limited by detail: either by the coarseness of the mesh, or the

lack of exact information about a real component – assumptions and
simplifications are used to enable the modelling. Experiment often has the
opposite problem: everything is measured at once. A closer match between
the modelled and measured would have required a refined mesh and
inclusion of more details in the model.

CONCLUSIONS

If a rotor blade for a wind turbine has to be divided for any reason, the

connection principle with connecting tubes is one practicable possibility.
The 13.4 m sectional blade survived all performed tests without failure. The
only failure during a full test program was the breakage of a screw of poor
quality. Thus the connection principle is suitable for sectional rotor blades.

The testing method with driving the blade in resonance by an exciter

mounted to the blade was performed for the first time at DLR. This method
made the test possible, because the needed power of the hydraulic system is
enormous and would influence other hydraulic tests running in the same
laboratory.

The thermoelastic measurements performed by CLRC showed that there

are relatively small spots with relatively high loading, which were not
predicted by the FE calculations. Also for fatigue test monitoring purposes,
this method will be examined further.

276

F. H

AHN ET AL

.

[12.3.2002–12:22pm]

[267–278]

[Page No. 276]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword

Other connection methods are under investigation within the same

project. The results will be compared to indicate effective methods of joining
these large composite components.

ACKNOWLEDGEMENT

This research was supported by the EC Non-nuclear Energy Programme

under contract number JOR3-CT97-0167.

REFERENCES

1. Hahn, F. FEM calculation of the tube connection for the LM 13.4 m blade. DLR report

internal to the project.

2. Lesniak, J.R. and Boyce, B.R. (1995). A High-Speed Differential Thermographic Camera.

Spring SEM Conference.

Design, Fatigue Test and NDE of a Rotor Blade

277

[12.3.2002–12:22pm]

[267–278]

[Page No. 277]

FIRST PROOFS

i:/Sage/Jtc/Jtc15-3/JTCM-21455.3d

(JTCM)

Paper: JTCM-21455

Keyword