Downloaded from orbit.dtu.

dk on: Apr 08, 2021

Determination of the J integral for laminated double cantilever beam specimens: The
curvature approach

Rask, Morten; Sørensen, Bent F.

Published in:
Engineering Fracture Mechanics

Link to article, DOI:

10.1016/j.engfracmech.2012.06.017

Publication date:
2012

Link back to DTU Orbit

Citation (APA):
Rask, M., & Sørensen, B. F. (2012). Determination of the J integral for laminated double cantilever beam
specimens: The curvature approach. Engineering Fracture Mechanics, 96, 37-48.
https://doi.org/10.1016/j.engfracmech.2012.06.017

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
 You may not further distribute the material or use it for any profit-making activity or commercial gain
 You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
 Determination of the J integral for laminated double cantilever beam
specimens: The curvature approach

M. Raska,∗, B.F. Sørensena

a DTU Wind Energy, Risø Campus, Technical University of Denmark, 4000 Roskilde, Denmark

Abstract
A new approach is proposed for measuring the J integral (and thus the fracture resistance) of interface
cracks in multiply laminates. With this approach the J integral is found from beam curvatures and applied
moments. Knowledge of ply layup and stiffness is not required. In order to test the accuracy of the proposed
approach, double cantilever beam specimen loaded with uneven bending moments (DCB-UBM) specimens
were tested and analysed using the curvature approach and a method based on laminate beam theory. Beam
curvatures were determined using a configuration of strain gauges. Good agreement was obtained between
the two approaches.
Keywords: Fibre reinforced materials, Bridging, Delamination, Interface fracture, J integral

1. Introduction

The rotor blades of modern wind turbines are made primarily of composite laminates and sandwich
structures [1]. A type of failure often seen in laminates and sandwich structures is delamination, e.g. cracking
along interfaces between plies, along adhesive/laminate interfaces and along sandwich/core interfaces.
There is an increasing interest in the use of cohesive zone modelling in the simulation of delamination
of components. In cohesive zone modeling, the fundamental constitutive law that describes the mechanics
of the fracture process is a traction-separation law called a cohesive law. Several commercially available
finite element programs have capabilities for simulations using cohesive zone modelling. However, the
determination of cohesive law parameters remains an important experimental challenge.
Most approaches for experimental determination of cohesive laws involve numerical simulations of a test
specimen loaded to failure. The fracture plane is modelled by a cohesive zone and the cohesive law is
determined indirectly by iteratively guessing and comparison of measured and predicted overall response,
e.g. load-displacement relations [2, 3, 4]. A more direct approach - one that does not require modelling of
the test specimen with a cohesive zone - is to derive the cohesive law by differentiation of the J integral
by the end-opening of the fracture process zone [5, 6, 7]. This approach has recently been extended to
determination of mixed-mode cohesive laws [8, 9].
The J integral is particularly useful in the analysis of beam-type specimens that exhibit a large-scale
bridging zone [10, 7]. A remarkable J integral analysis is that of a Double Cantilever Beam (DCB) specimen
loaded with wedge forces [6, 11]. For this specimen configuration, the J integral result for each arm is
simply the product of the instantaneous values of the applied force per unit width and the beam rotation
at the loading point. The prior load-displacement history, the specimen thickness and the elastic (linear
and non-linear) properties of the beams are not required [6, 11]. This analysis is valid also for large-scale
bridging zones. This specimen configuration has been used for the deduction of the cohesive laws of adhesive
joints using the approach involving J and the end-opening, as described above [12, 13].

∗ Corresponding author
Email addresses: [email protected] (M. Rask), [email protected] (B.F. Sørensen)

Preprint submitted to Elsevier June 8, 2012

 The analysis of laminates and multilayers consisting of elastic materials (assuming small-scale yielding)
can also be analysed by evaluating the J integral around the external boundaries. For DCB specimens loaded
with axial forces and/or bending moments, the J integral can be expressed in terms of the instantaneous
values of the applied forces and moments. However, for laminates consisting of many layers (lamina), the
calculations are involving and may best be handled numerically, e.g. using an analysis based on laminate
theory [14, 15, 16]. A drawback of this approach is that the thickness and elastic properties of each layer
must be known; this is particularly challenging for laminates with significant amount of off-axis plies for
which the elastic properties may not be accurately known. An approach based on gluing thick and stiff skins
with known elastic properties to the specimens can be used for reducing the sensitivity of elastic properties
of thin layers [16].
The study of large-scale bridging and large-scale active cohesive zones involves special considerations. In
general, J integral solutions differ from potential energy release rate solutions based on linear-elastic fracture
mechanics analysis of stress intensity factors [10]. Exceptions are beam-type specimens loaded with axial
forces and moments [7]. As such, a DCB specimen loaded with Uneven Bending Moments (DCB-UBM)
was recently proposed for characterization of mixed-mode crack growth in composites exhibiting large-scale
bridging [17]. For specimens consisting of elastic beams, the J integral can be determined in closed analytical
form and the only experimental data required is the instantaneous value of the applied moments (along with
specimen dimensions and elastic properties); no measurement of beam rotation or crack length is required.
For this test configuration, the J integral result is independent of crack length, giving stable crack growth
under any mode I - mode II combination.

1.1. Problem statement

In this paper, we propose an approach for the determination of the J integral for the DCB-UBM test
specimen from the instantaneous values of the applied moments and the beam curvatures. The analysis
remains valid for large-scale bridging problems as long as the beams behave linear-elastically. No knowledge is
required of the thickness and elastic properties of the individual layers. Therefore, the method is particularly
suited for the analysis of large-scale bridging problems of laminates consisting of many layers having different
elastic properties. As such, the present study answers the question raised by Paris and Paris [11] whether
other specimen configurations can be found for which the J calculation are equally direct as the DCB loaded
with wedge forces.

Figure 1: The homogeneous Double Cantilever Beam loaded with Uneven Bending Moments (DCB-UBM) specimen under
different load combinations; a) a nominal mode I load, b) a mixed mode load and c) a nominal mode II load.

As mentioned, the DCB-UBM test configuration is loaded by a combination of bending moments, as

shown in Fig. 1. A nominal mode I loading is shown in Fig. 1a. With moments defined positive downwards
as shown in Fig. 1b, the nominal mode I load state occurs when M1 = −M2 . In Fig. 1b, a mixed mode
loading is shown. Then, the moments applied to the beam fulfills M1 6= |M2 |. In Fig. 1c, a nominal mode

2
II loading is seen (M1 = M2 ). By varying the relations between M1 and M2 any mode combinations can be
attained.
In this study we develop the curvature model, based on Bernoulli beam theory, and test the proposed
approach using adhesive joint specimens, i.e. laminate beams joined by an adhesive layer. The methodology
is valid also for specimens made of dissimilar beams. The cracking will occur along one of the adhesive/beam
interfaces, meaning that the DCB specimens were not fully symmetric. Therefore, the crack growth is always
mixed mode, and the terms nominal mode I and mixed mode is used in the remainder of this text.

2. Theory

2.1. The curvature model

The J integral is valid for small strains, rotations and displacements. Generally, the J integral is defined
as [18, 19]: Z  
∂ui
J= Φdx2 − σij nj dS (1)
Γ ∂x1
where Φ is the strain energy density, σij is the stress tensor, ui is the displacement vector, nj is the outwards
unit vector normal to the integration path Γ and S denotes the length of the path Γ. The J integral is
path independent [19], implying that J takes the same value for any integration path that goes around the
fracture process zone from the lower crack face to the upper crack face in the counter clockwise direction,
as indicated with the dotted line in Fig. 2. As the beams are loaded with pure moments, the shear stresses

Figure 2: Generic DCB-UBM specimen with a large-scale bridging zone. The counter-clockwise integration path is shown.

along the boundaries vanish away from the crack tip and bridging zone, i.e. σij → 0 for i 6= j and both σij
and the strain tensor εij are independent of x1 away from the fracture process zone, where the beams are
subjected to pure bending. Then, for linearly elastic materials, the strain energy density is
1
Φ= σ11 ε11 . (2)
2
We now examine the integration path Γ1 illustrated in Fig. 2. With nj = (−1, 0, 0) the second term in
Eq. (1) can be rewritten using
∂ui ∂u1
σij nj = σ11 n1 = −σ11 ε11 . (3)
∂x1 ∂x1
Since the x1 -components of displacement and strain relate by
∂u1
= ε11 (4)
∂x1
eqs. (2) and (3) are the J integral-contribution along the integration path Γ1 . With dS = −dx2 at x1 = −a,
the J integral contribution along Γ1 (x1 = −a and − H2 ≤ x2 ≤ 0) becomes
Z −H2
σ11 (x2 )ε11 (x2 )
J1 = − dx2 . (5)
0 2

3
Figure 3: Moment and curvature and defined positive downwards. The strain distribution across the beam, resulting from a
pure moment is indicated.

This expression is valid for plane stress as well as plane strain. Note that in a multi-layer, the stress σ11 is,
unlike the strain ε11 , discontinous from layer to layer, and in a laminate based analysis [16] the integration
paths would be subdivided into one path per layer.
In the following, it is convinient to use multiple coordinate systems - a global x1 , x2 -system with origo at
the crack tip, and a local x, y-system for each beam with origo at the neutral axis of each beam positioned
a distance δ from the bottom of the beam. Both systems are defined in Fig. 3, where a positive bending
moment and curvature is shown. The neutral axis is shown as a dotted line at y = 0 with y = x2 + H − δ.
For a symmetric beam, the neutral axis coincides with the geometrical midplane, δ = H/2.
Changing Eq. (5) from the x1 , x2 system to the x, y system we get
Z −δ
σ11 (y)ε11 (y)
J1 = − dy. (6)
H2 −δ 2

Under the assumptions of small rotations and displacements, the strain profile along the y-axis for a multi-
layered beam subjected to bending is linear, as indicated in Fig. 3.

ε11 (y) = κ0 y (7)

where κ0 denotes the curvature of the neutral axis. Still examining integration path Γ1 Eqs. (6) and (7) are
combined to give

κ01 −δ κ0 M2
Z
J1 = − σ11 (y)ydy = 1 ·
2 H2 −δ 2 B

with B denoting the width of the specimen and the moment is defined by
Z H2 −δ
M2 = B σ11 (y)ydy. (8)
−δ

Similar results can be obtained for the other beam ends. There are no contributions to the J integral from
integration paths Γ2 and Γ4 where dx2 = 0 and σij nj = 0. Thus, the total J integral can be found by
summing the three beam-end contributions, i.e. along Γ1 , Γ3 and Γ5 . The result is

κ01 M1 κ0 M2 κ0 M1 + M2
J= · + 2· − 3· (9)
2 B 2 B 2 B
where κ02 and κ03 are the curvatures at the neutral axes of beams 2 and 3 respectively. The third term is
negative in agreement with the direction of the integration path Γ3 .
The result Eq. (9) shows that we can determine the J integral by measuring the applied moment and the
curvature of the neutral axis for each beam. Note, that the elastic properties (or layup sequence, anisotropy
etc.) of the beams of the DCB specimen are not required. This implies that for a complex multilayered
DCB specimen, even with unknown materials and unknown layup sequence, the J integral can obtained by
measuring the curvature of the beams, the applied moments and the width of the beams.
4
 The proposed curvature approach was verified analytically against known results from the literature.
The verification examples were a bi-layer specimen [16], sandwich specimens with center or interface cracks
[9, 20] and five-ply laminate [16]. In all cases, complete analytical agreement was found between the curvature
model and the results from the literature.

2.2. Curvatures from strains

It can fairly easily be shown that the curvature of the neutral axis of a beam κ0 can be found from the
normal strain ε11 at the top and bottom of the beam. This is beneficial since, if we know the strains in the
top and bottom, the curvature can be calculated, without knowledge of the position of the neutral axis. It
follows from Eq. (7) that the strain in the top and bottom of the beam can be obtained as

εt11 = κ0 (H − δ) and εb11 = −κ0 δ. (10)

with H denoting the height of the beam in question, and superscripts t and b denoting respectively the top
and bottom of the beam in a global system of reference. δ is eliminated to arrive at

εt11 − εb11
κ0 = . (11)
H
For instance for the top beam, beam no. 1, in Fig. 2 the curvature of the beam can be expressed as

ε11 (y = H1 − δ) − ε11 (y = −δ)

κ01 = . (12)
H1
In the derivation of Eq. (11) it was assumed that the beams experience pure bending. This means
that no matter which method (strain gauges, optical methods, etc.) is chosen for determining the necessary
strains, it must be done in a region of the beam free of effects from the beam end and the fracture process
zone (crack tip and bridging zone) in correspondence with de Saint-Venant’s principle.
If the beam in question is symmetric such that the neutral axis lies in the midplane of the beam (δ = H/2),
the strains will be opposite and numerically equal at the top and bottom of the beam, i.e. εt11 = −εb11 . Then
it is only necessary to determine the strain at either the top or bottom surface of each beam. For instance,
if the strain is determined on the top of the beam, the curvature can be calculated as

2εt11
κ0sym = (13)
H
where the subscript sym indicates that Eq. (13) is only valid for symmetric beams.

3. Experimental evaluation of the curvature approach

3.1. Overall idea

In order to validate the curvature approach, a test programme was set up in which a number of specimens
were tested, and the data were analysed using both the laminate theory approach of Lundsgaard-Larsen et
al [16] and the curvature approach to extract the J integral values. This enables direct comparison of the
two approaches. The DCB test specimens were made from two glass fibre/polyester composite beams joined
by an adhesive layer. The beams had two different layups giving different stiffnesses. The layups consisted
of two different lamina denoted material A and material B, with material A being less stiff than material
B. However, the material/adhesive interface to be tested was the same in all samples. The samples were
subjected to two different load configurations. The load configurations were chosen to minimize the number
of required strain gauges. The first was a nominal mode I configuration with M1 = −M2 and M3 = 0.
With this configuration, the uncracked end will have zero loading and thus zero curvature. The second was
a mixed mode load configuration with M1 = 0 and M2 = M3 . Beam 1 will have zero loading and thus zero
curvature. This means that in both cases, only two of the three beam ends would have non-zero curvature,

5
 Parameter Variants Comments
Materials A and B Material B is stiffer than material A
Layups 1 and 2 Layup 1 is stiffer than layup 2
Load configurations Nominal mode I M1 = −M , M2 = M and M3 = 0
Mixed mode M1 = 0, M2 = M and M3 = M
Analysis methods Curvature approach Beam curvatures from strain gauges
Laminate theory approach Lamina thicknesses and stiffnesses must be measured

Table 1: Overview of the test programme.

Mode I Layup 1
Material A h1=0.8mm
M

Adhesive layer Material B h2=8.4mm

M
Material A h3=0.8mm

= strain gauge
Mixed mode Layup 2

Material A h1=3.8mm

Adhesive layer M Material B h2=2.4mm

M Material A h3=3.8mm

B=30mm

Figure 4: Left: DCB specimens made by joining two symmetric 3-ply beams (thickness 10 mm) with an adhesive layer (thickness
5 mm). The two load configurations are illustrated. The mode I load configuration corresponds to M1 = −M , M2 = M and
M3 = 0, while the mixed mode configuration corresponds to M1 = 0, M2 = M and M3 = M . The locations of the strain
gauges are illustrated. Right: Sketches of the two layups.

so that only two of the three beam ends would need to be fitted with strain gauges, see Fig. 4. An overview
of the test programme is shown in Table 1.
Laminates consisting of six layers joined by an adhesive were used, see Fig. 4. The layups were made
from combinations of two different glass fibre/polyester composite materials, see Section 3.2.1. For reasons
of confidentiality these are referred to simply as material A and material B. Material B is approximately
three times as stiff as material A, due to a large amount of 0◦ -fibres, whereas material A mainly contains
off-axis fibres.
Two different types of specimens with different layups were made. In layup 1, the central layer of material
B has a nominal thickness of 8.4 mm, while the central layer in layup 2 has a nominal thickness of 2.4 mm.
All beams had nominal thicknesses of 10 mm. This means that layup 1 contains a higher ratio of the stiffer
material B, whereby layup 1 is expected to be stiffer than layup 2. However, since fracture occurs along
similar material interfaces for the two specimen types, similar J integral curves are expected for the two
layups.
The test programme included twelve specimens, divided in four groups of three: L1MI, L1MM, L2MI
and L2MM, with L1 and L2 denoting layup 1 and layup 2 while MI and MM denotes nominal mode I and
mixed mode, respectively. The thickness of all seven layers (six composite plies and the adhesive layer) of
each specimen are needed for the analysis based on the laminate beam theory approach. Thus, the thickness
of each of these layers were measured at three positions along the specimen, and the average of each set of

6
three measurements were entered in a database.

3.2. Experimental work

3.2.1. Specimen manufacturing
The laminates for the test specimens were supplied by Danish manufacturer of wind turbine blades LM
Wind Power A/S. The DCB specimens were manufactured as follows. First, a laminate panel (600 by 600
mm, thickness approximately 10 mm) of each glass fibre/polyester layup were made by the techniques of
hand layup and matrix impregnation by vacuum infusion. After consolidation, 15 beams (length 500 mm,
width 30 mm) were cut from each panel. To make six DCB specimens of each layup, the beams were bonded
together in pairs with a thermoset adhesive (Oldopal 0555 VE from Büfa, hardened at 40◦ C for 16h). A slip
foil (thickness 12.7 µm) was placed at the one end of the beams to define a sharp crack initiation location.
Spacers (5 mm) were used to control the thickness of the adhesive layer. Steel grips were fixed to each beam
with four steel screws and an epoxy adhesive (Scotch-Weld DP 460 from 3M, hardened at 40◦ C for 2h).
Holes (diameter 1.5 mm) were drilled at the end of the slip foil (where crack growth will start) and steel
pins were inserted. These were to be used for mounting of extensometers. Finally, strain gauges (0◦ , gauge
length 10 mm) were fixed to the beams to be subjected to bending according to Fig. 4. In accordance with
de Saint-Venant’s principle, strain gauges were placed a distance of two times the beam height away from
the gripping fixtures on the specimens. Fig. 5 shows a detailed drawing of a finished specimen.

Figure 5: Drawing of specimen. Positions for both mode I and mixed mode tests are shown.

3.2.2. Measurement of elastic properties

Spare beams, not used for manufacturing DCB specimens, were subjected to tensile tests on a standard
servo-hydraulic test machine fitted with a 100 kN load cell (Instron UK1029). Three beams of each layup
were tested to strains of 0.3%. Strains were determined with a extensometer (Instron 1562, gauge length 50
mm). The tests were run at a displacement rate of 0.01 mm/s. Readings from the extensometer and load
cell were collected at a PC at 10Hz. The stiffness for each layup were found by making linear regression to
the stress-strain data in the strain interval 0.05-0.25%. From these values, the Young’s moduli of the two
materials A and B were determined. By performing tensile tests of both layups, the effective stiffness Ē of
each layup was determined. Then, the rule of mixtures could be employed to back-calculate the effective
stiffness values of the materials A and B.

7
 Note that knowledge of the Young’s moduli of the two materials is not required for the curvature approach
(Eq. 9), but is a required input in an analysis based on laminate theory, see section 3.3.2.

3.2.3. DCB test method

The DCB-UBM specimens were tested on a special test device, shown schematically in Fig. 6. The testing
fixture and procedure is described in detail in [17]. With two load cells and an arrangement of wires and

Figure 6: A schematic drawing of the testing device. Adapted from [9].

rollers, the test fixture allows for the test specimen to be tested under any combination of bending moments.
The wires apply the same force to two rollers located at each of the transverse arms mounted on each beam
of the DCB specimen, creating pure bending moments. As moment = force × arm, a configuration with
different moment arms on each side gives different moments on the two beams, even though the forces P are
of identical magnitude at both transverse arms. Adjusting the device from one load configuration to another
is simply a matter of altering the lengths of the moments arms by changing the position of the rollers on
the transverse arms. The moments are written as

M1 = P `1 , M2 = P `2 . (14)

By adjusting `1 and keeping `2 fixed any moment combination can be obtained. The wire force is increased
by moving the lower part of the test device (denoted lower beam in Fig. 6) downwards at a constant
displacement rate.
The test was conducted as follows. First, a DCB specimen was mounted at the test device. The
transverse arms and the wire was mounted. Next, an extensometer (Instron type 2620–601, range ±5 mm),
8
was mounted at the steel pins to record the end-opening. The strain gauges were connected to a strain gauge
amplifier (HBM MGCplus AB22A). The displacement rate of the lower beam was 10 mm/min. Pictures
were taken during the experiments, and data for the two load cell readings (N), the extensometer opening
(mm) and the four strain gauge readings (%) were logged at 25Hz.

3.3. Data analyses

The datasets for each specimen were analysed using both the curvature approach and the laminate beam
theory approach, whereby the outcome of the two approaches can be compared. In both cases, the adhesive
layer is ignored. This is a reasonable approximation, as the adhesive is much softer than the glass fibre
beams.

3.3.1. Curvature approach

From the loads and the length of the moment arms, `1 and `2 , the moments were calculated for each
data point using Eq. (14). The curvature at each data point were determined from the strain gauge data
using Eq. (11). From the moment and curvature values, the J integral was calculated for each beam and
the total J integral value was obtained as the sum of the contributions from each beam, according to Eq.
(9).

3.3.2. Laminate beam theory approach

For the laminate beam theory data analysis, a Matlab code acquired from [16] was modified so that the
J integral was calculated for each data point. The J integral was determined using the result from [16]
p̃
X Ep Mb2  2 3
Ab yp−1 − yp3 − 3Ab Bb yp−1
2
− yp2 + 3Bb2 (yp−1 − yp )



J= 2 (15)
p=1 6 (Db Ab − Bb2 )

where p = 1, 2, . . . , p̃ is the integration path number indicated in Fig. 2, b = 1, 2, 3 is the beam index, Ep
is the Young’s modulus of the ply related to integration path p and yp is the lower (in the global system)
interface of that ply. Ab , Bb and Db are the extension, coupling and bending terms for beam b. For the sake
of brevity, the definitions of these terms are not given here; they can be found in [16].

4. Results
4.1. Material stiffnesses
Table 2 lists the effective Young’s modulus Ē in the x1 -direction for the two materials used to make the
layups. The term effective is used to point out that these stiffness values are found experimentally as the
values when the materials are fixed in the described layups, i.e. not from tests of the individual layers alone.

Material Ē (GPa)
A 14.6 ± 0.09
B 49.1 ± 1.4

Table 2: Effective Young’s modulus with standard deviation for the two materials.

4.2. Overview of failure modes of DCB-UBM specimens

Crack growth initiated from the slip foil along the adhesive/laminate interface. It was, however, observed
that the four groups of specimens developed different damage mechanisms. As shown in Fig. 7a-b, the mode I
test specimens (groups L1MI and L2MI) cracked along a single crack path at the material/adhesive interface
with the formation of large-scale bridging by glass fibres. Layup 1 tested in mixed mode (L1MM), Fig. 7c,
cracked in two planes with one fracture plane at the adhesive/material A interface and a second plane at the
material A/material B interface. Layup 2 tested in mixed mode (L2MM), Fig. 7d, cracked in three planes
with one plane at the adhesive/material A interface and two planes in the material A ply.
9
 L1M1 x2 L1MM 2 tips
1 tip x1

L2M1 L2MM 3 tips

1 tip

10 mm

Figure 7: Photographs of four test specimens during crack growth. The pictures show a characteristic specimen from each of
the four test groups a) L1MI, b) L2MI, c) L1MM and d) L2MM.

4.3. Fracture resistance data

J integral results from the two approaches are plotted as a function of the end-opening δ ∗ for mode I
experiments and mixed mode experiments in Figs. 8 and 9 respectively. The data plots in Figs. 8 and 9

Figure 8: J integral as a function of end-opening δ ∗ for specimens tested in nominal mode I configuration. The solids lines show
results from the curvature model, while the dotted lines show results from the laminate theory model. The specimens L1MI-1
and L1MI-3 are not included since the crack kinked into the adhesive layer right after crack initiation for these specimens.

all have the same characteristic appearance which can be described in four phases (phases indicated in Fig.
8): In phase 1, the J integral value increases without causing crack growth. Then, in phase 2 crack growth
initiates. In phase 3, the J integral value increases rapidly with a continuous increase in δ ∗ . During this
phase, the crack tip advances and a fracture process zone of bridging fibres develops. At an end-opening
of roughly 1 mm (phase 4) the J integral increases more slowly and almost linearly with δ ∗ , approaching
a steady state value. During the tests, a significant amount of fibre bridging was seen, see Fig. 7. Fibre
bridging is known to increase the fracture resistance [21].

10
Figure 9: J integral as a function of end-opening δ ∗ for specimens tested in mixed mode configuration. For the curvature model
curves (solid lines) it has been indicated whether the crack front had two or three crack tips.

The crack growth initiation values J0 for all samples are listed in Table 3. Crack initiation was defined as
an end-opening of 20 µm. The average values with standard deviations are 314 ± 49J/m2 and 392 ± 60J/m2
for the mode I tests and the mixed mode tests, respectively.

Specimen J0 (J/m2 ) Specimen J0 (J/m2 )

L1MI-1 - L1MM-1 437
L1MI-2 358 L1MM-2 319
L1MI-3 - L1MM-3 320
L2MI-1 320 L2MM-1 425
L2MI-2 333 L2MM-2 460
L2MI-3 245 L2MM-3 390

Table 3: Values of fracture resistance for crack growth initiation along the laminate/adhesive interface. Values are taken from
the curvature approach analyses. Values from specimens L1MI-1 and -3 have been omitted as the crack in these specimens
initiated in the adhesive layer.

The results from the nominal mode I tests in Fig. 8 show that the two approaches gives J, δ ∗ -curves with
roughly the same shape and deviations of 0 − 10%.
Fig. 9 shows the results from mixed mode tests of the two approaches. Again, the J, δ ∗ -curves follow the
same pattern for the two approaches. L1MM-1 shows deviations of about 1%. For L1MM-2, the laminate
theory approach gives J values about 15% larger the the curvature approach, and for L1MM-3, the curvature
approach gives J values about 6% larger than the laminate theory approach. The layup 2 experiments shows
curves of similar appearances for the two approaches. However, the deviations are very large: The curvature
approach gives J integral values about 83-102% higher than the laminate theory approach.
The large deviations for the L2MM group was investigated by studying the beam curvatures: From
the pictures taken during the experiments, the actual curvatures of the beams were determined by fitting
circles to the photographs of the bent beams. With the laminate theory formulation, the beam curvatures
were calculated from the geometry, layup configuration, stiffness and thickness of each ply and applied
moments. Thereby, it was found that the predictions of beam curvature from the curvature approach and
11
the laminate theory approach were respectively 1-5% and 47-51% below the actual curvature determined
from the photographs.

5. Discussion

5.1. Comparison of the two approaches

From Fig. 8 we saw that the two analyses of the L1MI group gives comparable results with non-
systematic deviations between 0 and 10%. The analysis based on laminate beam theory depends on a
number of variables which can not be determined with great accuracy, such as the Young’s moduli and
thickness of individual layers. The curvature approach, on the other hand, depends on variables which can
be determined with high accuracy, such as strains, moments and laminate beam thickness. It is therefore
reasonable to assume that the curvature approach is the more accurate approach.
From Fig. 9 it is seen that the L1MM group has steady-state J values, Jss , about 1650 J/m2 , while
the laminate theory approach treatment of the L2MM group gives Jss values around 1500 J/m2 and the
curvature approach treatment gives Jss values around 3000 J/m2 . As mentioned above, it is expected
that cracks along identical interfaces under identical moment combinations give similar Jss values. The
results in Fig. 9 therefore seem to indicate that the curvature approach based analysis of the L2MM group
overestimates the Jss values. However, from Fig. 7c-d it was observed that the L1MM group has two crack
tips (and two fracture process zones), while the L2MM group has three crack tips. In effect, the cracks
in the L2MM specimens create roughly 50% more new fracture surface area per unit length crack growth
than cracks in the L1MM specimens. Therefore, as a first approximation, it is reasonable to expect that
the specimens in the L2MM group have steady-state J values that are approximately 50% higher than
the values for the specimens in the L1MM group. This hypothesis is largely consistent with the results in
Fig. 9. Furthermore, this hypothesis is in agreement with the finding that the laminate theory approach
underestimates the beam curvature, while the curvature approach is capable of determining the curvatures
with low deviations. Hence it is reasonable to conclude that the laminate theory approach underestimates
the J integrals for the L2MM group. It is not known why the laminate theory approach performs poorly
for the L2MM group, considering that the two approaches are in good agreement for all other test groups.
It is an interesting finding, that by changing the ply thicknesses in the layup, one can change the cracking
mechanism and thereby the steady-state fracture resistance of identical interfaces. The Jss value for all single
crack tip tests are around 750 J/m2 , while the two- and three crack tip tests have Jss values around 1650 and
3000 J/m2 , respectively. It is an interesting idea that there might be an approximately linear dependency
between the number of crack tips/fracture process zones and the steady state J values.

5.2. Crack growth initiation

From Table 3 it is seen that all specimens have comparable crack growth initiation energy values. The
values are of the same order of magnitude as results from the literature for similar materials/interfaces
systems [22]. When dividing the values into layup 1 and layup 2, the values are 359 ± 55J/m2 and 362 ±
78J/m2 respectively. Thus, no difference in crack initiation energy is seen between layup 1 and layup 2.
This is not surprising since the conditions at the crack tip are identical for the two layups. For mode I and
mixed mode, the crack growth initiation values are 314 ± 49J/m2 and 392 ± 60J/m2 respectively. Even
though the difference between these results are not significant, there is a tendency that the mixed mode
experiments have the highest crack initiation energies. It is reasonable to expect that there is a difference
in the initiation energies, as the stress state at the crack tip is different for the two load configurations. For
many interface systems, a higher amount of mode II results in a higher fracture energy [23].

5.3. Practical advantages and disadvantages of the curvature approach

For a homogenous DCB specimen consisting of a single material with well-known elastic properties, the
laminate beam theory approach is rather straight-forward. However, when the examined DCB specimen
consists of multilayer-beams, the laminate beam theory approach becomes more labour-intensive as the
stiffness and thickness of all layers are required.
12
 The curvature approach, on the other hand, is equally extensive for all types of DCB specimens. Per-
forming the experiments using the curvature approach with the strain gauge measurement involves fixing
strain gauges to the DCB beams and connecting (soldering) these to the strain gauge amplifier. This pro-
cedure takes some time. However, as layer thicknesses and stiffnesses are not required, it is not necessary
to spend time determining these. Furthermore, not depending on values of layer thickness and stiffness
implies that the curvature approach is likely to give more precise results than the laminate beam theory
approach. The necessary inputs (moments and top and bottom beam strains) for the curvature approach
can be determined with high accuracy, while the inputs for the laminate beam theory approach (lamina
thicknesses and stiffnesses) are more difficult to determine. From the laminate beam theory expression in
Eq. 15 it is seen that J ∝ E and J ∝ h3 . This implies than measurement errors in lamina thickness are
more critical than measurement errors in the lamina stiffness.
As mentioned in Section 2.2, when characterising a DCB specimen with symmetric beams it is only
necessary to use one strain gauge per beam. Eventhough the beams in the present test were consided to
be symmetric, two strain gauges were used per beam, in order to validate the above claim. Indeed, it was
found that the readings from all strain gauge pairs had agreements within 2.5%, with an average value of
1.3%. Thus, with symmetric beams, it would be most convinient to place one strain gauge at the outside
surface of each beam, thereby avoiding having to place strain gauges in the gap between the beams. The
curvature approach is thus not only limited to adhesive joint testing but can also be used for delamination of
laminates having symmetric beams. For a delamination test with unsymmetric beams it might be difficult
to place find space for all necessary strain gauges. Furthermore, the agreements within strain gauge pairs
justifies the decision to ignore the adhesive layer in the data analyses in Section 3.3.
An alternative to using strain gauges would be to use a digital image correlation system such as Aramis
from Trilion Quality Systems [24] for the curvature measurements. Thereby time could be saved in specimen
preparation, but the data analysis would probably be more extensive and time consuming.
As such, it is argued that the curvature approach is suitable for complex layups, where it has a number
of advantages over the laminate theory approach. Furthermore, the curvature approach can be used for
characterization of materials and assemblies with unknown parameters, such as test specimens cut from
older constructions.
Finally, although it was not the scope of the present work, the mode I cohesive laws can be determined
from the measured data on the J integral and the end-opening δ ∗ by fitting a suitable funtion to the J, δ ∗ -data
and differenting the function [25].

6. Conclusions
An approach for determining the J integral for multilayer Double Cantilever Beams with Uneven Bending
Moments (DCB-UBM) has been developed. This approach is based on the idea that the J integral can be
found from beam curvatures and moments alone. The approach is especially well-suited for multilayer
laminates, since no knowledge is needed on the thickness and Young’s modulus of individual plies in the
beams. This gives two advantages: The curvature approach is comfortable to use, as the equations and data
treatment routines are much less extensive than the ones needed when applying the traditional laminate
theory approach. The curvature approach is very accurate as the parameters on which it depends can be
determined with great accuracy.
In the present work, the beam curvatures were determined by the use of strain gauges. The test specimens
included two different layups subjected to two different load situation, such that four groups of experiments
were conducted. Three of these showed excellent agreement between the curvature approach and the laminate
theory approach, while it was found that the laminate theory approach underestimated the J integral
massively for the fourth group.

Acknowledgments
The authors wish to thank LM Wind Power A/S (in particular Rune Vestergaard for donating test
materials and Christian Lundsgaard-Larsen for kindly providing the Matlab code used for the laminate
13
approach). We also thank Christian H. Madsen for assistance with specimen preparation and Erik Vogeley for
assistance with specimen testing. The DCB experiments leading to the results in this paper was conducted
as part of a master project by Morten Rask. The research leading to these results has received funding
from the European Comunity’s Seventh Framework Programme (FP7/2007-2013) under grant agreement no
214467 (NATEX).

References
[1] Sørensen BF, Holmes JW, Brønsted P, Branner K. Blade Materials, testing methods and structural design, in Wind
Power Generation and Wind Turbine Design (ed. Tong W). Southampton: WIT Press; pp. 417-465. 2010.
[2] Kafkalidis MS, Thouless MD. The effect of geometry and material properties of the fracture of single lap-shear joints.
Int J Solids Struct 2002;39:4367-83.
[3] Yang QD, Thouless MD, Ward SM. Numerical simulations of adhesively-bonded beams failing with extensive plastic
deformation. J Mech Phys Solids 1999;47:1337-53.
[4] Yang QD, Thouless MD. Mixed-mode fracture analysis of plastically-deforming adhesive joints. Int J Fract 2001;110:175-
87.
[5] Li VC, Ward RJ. A novel testing technique for post-peak tensile behaviour of cementitious materials. Fracture toughness
and fracture energy - testing methods for concrete and rocks (eds. Mihashi H, Takahashi H and Wittmann FH).
Rotterdam: A. A. Balkema Publishers; pp. 183-95. 1989.
[6] Olsson P, Stigh U. On the determination of the constitutive properties of thin interphase layers - an exact inverse
solution. Int J Fract 1989;41:R71-6.
[7] Suo Z, Bao G, Fan B. Delamination R-curve phenomena due to damage. J Mech Phys Solids 1992;40:1-16.
[8] Sørensen BF, Kirkegaard P. Determination of mixed mode cohesive laws. Eng Fract Mech 2006;73:2642-61.
[9] Sørensen BF, Jacobsen TK. Delamination of fibre composites: determination of mixed mode cohesive laws. Compos Sci
Tech 2009;69:445-56.
[10] Bao G, Suo Z. Remarks on crack-bridging concepts. Appl Mech Rev 1992;45:355-61.
[11] Paris AJ, Paris PC. Instantaneous evaluation of J and C ∗ . Int J Fract 1988;8:R19-R21.
[12] Stigh U, Andersson T. An experimental method to determine the complete stress-elongation relationship for a structural
adhesive layer loaded in peel, in Fracture of Polymers, Composites and Adhesives (eds. Williams JG and Pavan A).
Amsterdam: ESIS Publication 27, Elsevier Science Ltd; pp. 297-306. 2000.
[13] Högberg JL, Sørensen BF, Stigh U. Constitutive behaviour of mixed mode loaded adhesive layer. Int J Solids Struct
2007;44:8335-54.
[14] O’Brian TK, Raju IS, Garber DP. Residual thermal and moisture influences on the strain energy release rate analysis
of edge delamination. J Compos Tech Res 1986;8:37-47.
[15] Ozdil F, Carlsson LA. Characterisation of mixed mode delamination growth in glass/epoxy composite cylinders. J
Compos Mater 2000;33:420-41.
[16] Lundsgaard-Larsen C, Sørensen BF, Berggreen C, Østergaard RC. A modified DCB sandwich specimen for measuring
mixed-mode cohesive laws. Eng Fract Mech 2008; 75: 2514-30.´
[17] Sørensen BF, Jørgensen K, Jacobsen TK, Østergaard RC. DCB-specimen loaded with uneven bending moments. Int J
Fract 2006;141:159-72.
[18] Rice JR. Mathematical analysis in the mechanics of fracture. Fracture, an advanced treatise, Vol. II (ed. H. Liebowitz).
New York: Academic Press; 191-311. 1968.
[19] Rice JR. A path independent integral and the approximate analysis of stain concentration by notches and cracks. J
Appl Mech 1968;35:379-86.
[20] Bao G, Ho S, Suo Z, Fan B. The role of material orthotropy in fracture specimens for composites. Int J Solids Struct
1992;29:1105-16.
[21] Albertsen H, Ivens J, Peters P, Wevers M, Verpost I. Interlaminar fracture toughness of CFRP influenced by fibre
surface treatment: Part 1. Experimental results. Compos Sci Technol 1995;54:133-145.
[22] Sørensen BF, Goutianos S, Jacobsen TK. Strength scaling of adhesive joints in polymer–matrix composites. Int J Solids
Struct 2009;46:741-61.
[23] Hutchinson JW, Suo Z. Mixed mode cracking in layered materials. Adv Appl Mech 1992;29:63–191.
[24] http://www.trilion.com/
[25] Sørensen BF, Jacobsen TK. Determination of cohesive laws by the J integral approach. Eng Fract Mech 2003;70:1841-58.

14