International Journal of Solids and Structures 48 (2011) 234–242

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

A numerical study on the SEN-TPB test applied to mode I wood

fracture characterization
N. Dourado a, M.F.S.F. de Moura b,⇑, J. Morais a
a
CITAB/UTAD, Departamento de Engenharias, Quinta de Prados, 5000-911 Vila Real, Portugal
b
Faculdade de Engenharia da Universidade do Porto, Departamento de Engenharia Mecânica e Gestão Industrial, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: A numerical study involving the single edge notched-three point bending test was methodically per-
Received 13 February 2010 formed to evaluate its adequacy for wood fracture characterization under mode I loading. Stress analysis
Received in revised form 30 August 2010 along the ligament length using cohesive zone modeling allowed understating the influence of the com-
pressive stresses induced by bending on the stress profile for geometrically similar specimens with dif-
ferent sizes. The stress relief region induced by the presence of the crack was identified by a stress
Keywords: analysis procedure along the beam length. A data reduction scheme based on the beam theory and on
Wood
the crack equivalent concept was then developed considering an approach using a rectangular stress
Fracture characterization
Mode I
relief region. The method allows estimating the Resistance-curve without monitoring the crack length
Stress relief region during propagation. Important remarks regarding the aspects which contribute to explain the size effect
phenomenon in wood are highlighted in this study. Following the presented procedure the application of
the single edge notched-three point bending test for wood fracture characterization under mode I has
become considerably user-friendlier.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Wood has ever been used by the humanity justified by its broad
availability and remarkable mechanical properties at reasonably
low weight, such as its high strength, bending stiffness and tough-
ness. Today there is a renewed interest on structural applications
of wood due to environmental reasons and energy shortages.
Therefore, the development of adequate failure criteria used in
wood design acquires special relevancy. In this context, Fracture
Mechanics criteria are particularly adequate owing to their ability
to deal with the stress singularities that are widely present in woo-
den structural components. Furthermore, Fracture Mechanics cri-
teria based on energetic analysis (Stanzl-Tschegg et al., 1995;
Morel et al., 2005; de Moura et al., 2008; Dourado et al., 2008) is
particularly adequate when strong heterogeneity in the material
is present. In this context, it becomes relevant the evaluation of
fracture energy under pure mode I loading for the different fracture
planes, typical of wood. In fact, wood is usually regarded as an
orthotropic material with three well defined directions: longitudi-
nal (L) along the fibers, radial (R) and tangential (T) to the annual
growth rings. Therefore, six principal systems of crack propagation
can be defined in wood: TL, RL, LR, TR, RT and LT. According to this
⇑ Corresponding author.
E-mail address: mfmoura@fe.up.pt (M.F.S.F. de Moura).
notation, the first letter indicates the normal direction to the crack
plane, while the second addresses the direction of crack propaga-
tion. It is generally accepted that fracture usually occurs in the
TL and RL crack propagation systems. In these cases, the double
cantilever beam (DCB) test is the best choice owing to its simplicity
(Yoshihara and Kawamura, 2006; de Moura et al., 2008). However,
fracture can also occur in the TR and RT systems, for which the DCB
is not appropriate due to observed stick–slip effect and dimen-
sional restrictions. In these cases, the single-edge-notched beam
loaded in three-point-bending (SEN-TPB) (Nordtest Method,
1993) test (Fig. 1) is particularly adequate owing to its simplicity
and smaller required dimensions than the DCB to measure fracture
energy GIc. Gustafsson (1988) argued that the SEN-TPB requires
smaller dimensions to provide reliable fracture measurements,
which is vital to get a test volume free from defects, which can al-
ter markedly the measured fracture energy. Daudeville (1999) used
Damage Mechanics (DM) and Linear Elastic Fracture Mechanics
(LEFM) models to simulate fracture in wood using the SEN-TPB
test. The author analyzed the size effect and gave the range of
applicability of LEFM. He argued that fracture energy can be con-
sidered as a material parameter and it was verified that a non-lin-
ear approach is necessary for the simulation of fracture of small
specimens. Nevertheless, the above mentioned advantages are mit-
igated due to the difficulty associated to definition of an user
friendly data reduction scheme to evaluate GIc. Indeed, it is known
(Dourado et al., 2008; de Moura et al., 2010) that spurious effects

0020-7683/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2010.09.020
 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242
y the use of this test appealing. The proposed method is based on
P, δ
x 2. Stress analyses
H
L1 a0
2L
Fig. 1. Schematic representation of the SEN-TPB test.
induced by compressive stresses due to bending restrict the avail-
able region necessary for self-similar crack growth. These condi-
tions are fundamental to perform rigorous evaluations of GIc.
There are two main techniques commonly used: the compliance
calibration method (CCM) and the beam theory (BT) based meth-
ods. The CCM (Dourado et al., 2010) is usually based on the compli-
ance (C = d/P) calibration as a function of crack length (C = f(a))
considering several specimens with different initial crack lengths
a0. However, the compliance calibration involving different speci-
mens can lead to errors on GIc measurements due to material var-
iability for specimen to specimen, typical of natural materials as is
the case of wood. Alternatively, the crack can be monitored during
its growth to establish the C = f(a) relationship. However, this task
is very difficult to perform due to unstable propagation and ambig-
uous detection of the crack tip motivated by an extensive fracture
process zone developed ahead of the crack tip (micro-cracking and
fiber-bridging). On the other hand, the presence of the crack ren-
ders difficult the application of beam theory based methods, fre-
quently used in other specimen geometries (e.g., DCB de Moura
et al., 2008). Indeed, the crack induces stress concentrations and
originates a stress relief region leading to the modification of the
stress profile predicted by the BT.
The objective of this work is to analyse numerically several fea-
tures intrinsic to the SEN-TPB specimen when applied to mode I
fracture characterization of wood. Detailed stress analyses were
conducted with the aim to develop a suitable data reduction
scheme, overcoming the above referred drawbacks and rendering
b
3H
T
L
R b
R L
T
a0 = H /2
Fig. 2. Specimen parts before bonding.
235
the beam theory and on the crack equivalent concept, and accounts
for the stress relief region induced by the presence of a crack.
The SEN-TPB (Fig. 1) specimen test applied to wood is formed
by two arms oriented along the longitudinal (L) direction and a
central part bonded to the referred arms (Fig. 2). The central part
represented in the figure is oriented to carry out fracture experi-
ments in the TL system (the first letter indicates the normal direc-
tion of the crack plane, while the second specifies the direction of
crack propagation). Although the SEN-TPB test is particularly use-
ful in the TR and RT fracture systems due to inadequacy of the DCB,
the TL system was chosen to validate the procedure, since it was
characterized in a previous work (Dourado et al., 2008).
A stress analysis of the SEN-TPB specimen was performed in or-
der to understand several issues affecting the stress profiles ob-
tained by means of the beam theory. Effectively, the presence of
the crack influences drastically the stress distributions in a non-
negligible region around the crack. Therefore, two effects might
be discussed. The first one is the influence of compressive stresses
due to bending on the natural evolution of the stress profile along
the ligament length during crack propagation. The second one is
related to the presence of a stress relief region near the pre-crack
faces, which alters the stress profiles estimated by the beam
theory.
2.1. Stress profile along the ligament
Stress profiles along the ligament length were obtained by
means of cohesive elements located at the crack section for differ-
ent specimen sizes (Table 1). The elastic properties of spruce (Table
2) and the respective cohesive parameters (Table 3), which are fun-
damental to define the cohesive zone (CZ) ahead of the crack tip,
are used in the numerical simulations. The constitutive law used
in the set of cohesive elements is presented in Fig. 3, and is formed
by three branches. The first one (0 6 w 6 wo) simulates the linear
behavior before damage onset. The value of wo is defined by the ra-
tio between the local strength (ft) and the interface stiffness, which
H
3H
H
L
R
T
236 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242

Table 1 Table 3
Summary of the specimen dimensions used in the numerical analyses. Cohesive zone parameters of Norway spruce used in the bilinear model (Dourado
et al., 2008).
Series H (mm) L1 (mm) L (mm) a0 (mm) b (mm)
ft (MPa) fb (MPa) wb (mm) GIc (N/mm)
D1 140 350 420 70 40
D2 280 700 840 140 80 1.66 0.30 0.09 0.145
D3 560 1400 1680 280 160
D4 1120 2800 3360 560 320
D5 2240 5600 6720 1120 640
D6 4480 11,200 13,440 2240 1280 f

should be the largest value that avoids pronounced interpenetra-

tions and do not produce numerical problems (106 N/mm3) (de
ft
Morais et al., 2003). This bilinear softening law (Fig. 3) describes
both micro-cracking (first linear descending branch between wo G Iμ (Component: "micro-cracking")

Stress
and wb) and fiber-bridging (second linear descending branch be-
tween wb and wc) phenomena typical of crack propagation in wood
(Stanzl-Tschegg et al., 1995; Dourado et al., 2008). The ultimate
crack opening (wc) is obtained from the critical energy release rate fb G I b (Component: "fibre-bridging")
(GIc), which is equated to the area under the cohesive law

ft wb þ fb ðwc  wo Þ
GIc ¼ : ð1Þ w
2
wo wb wc
Cohesive elements allow simulating damage initiation and growth
Crack opening
along the crack path, and will be used to simulate crack propagation
in the SEN-TPB test. They were positioned all through an upright Fig. 3. Softening law (Petersson, 1981) used in the numerical computations.
central line sited ahead of the initial crack notch a0 = H/2 (Fig. 4)
dividing the specimen ligament length in every 0.5 mm. This mesh
size was kept constant for all specimen sizes (Table 1).
proving that the CZ is already being confined, due to compressive
The analysis was performed under displacement control.
stresses induced by bending. As a result, the respective stress pro-
Fig. 5(a) presents the evolution of normalized normal stresses f/ft
files (curves 5 and 6 in Fig. 5(a)), are different from the previous
(in this case, the normal stress f is oriented along the x-axis:
ones, and remarkably non-conformable with each other. It should
Fig. 1) as a function of the normalized ligament length y/(H/2) for
be noted that two different trends on the stress profiles are identi-
six stages during crack propagation considering the intermediate
fied (Fig. 5(a)). In fact, for profiles (1)–(3) the peak stress moves
dimension D3 (Table 1). Since the ligament length varies with the
away relatively to the crack extremity, demonstrating the develop-
crack growth, the plots were all moved to the graph origin, which
ment phase of the CZ. On the other hand, the opposite trend can be
thus corresponds to the crack extremity in all cases. Fig. 5(b) rep-
observed for profiles (4)–(6), which puts into evidence the phe-
resents the normalized cohesive zone length as a function of the
nomenon of CZ confinement. Between (3) and (4) there is practi-
normalized equivalent crack length ae (to be defined later, see Eq.
cally no difference, demonstrating that fracture is occurring
8), showing the corresponding values for the above referred six
under self-similar conditions.
stages. The cohesive zone length (Fig. 5(b)) is accessed (for each
The effect of the specimen size on the stress profiles in the cohe-
increment of displacement) via the coordinates of the integration
sive zone at the crack initiation was also analyzed (Fig. 6). As it can
points of the cohesive elements undergoing the softening process
be seen, these stress profiles at crack initiation are differently af-
(see softening branches in Fig. 3). Consequently, the lCZ is easily ob-
fected for the six specimen sizes (Fig. 6). Effectively, for smaller
tained in each increment by subtracting the coordinates (in the
dimensions the confinement of the stress profile motivated by
crack path) of the last integration point undergoing the softening
bending takes place on the CZ profiles from the beginning. Never-
process and the one corresponding to crack tip position. Accord-
theless, a convergence trend between those profiles is achieved
ingly, the stress profiles along the normalized ligament length
with the increase of the specimen size (D1 to D6), clearly demon-
are also obtained for the selected stages (Fig. 5(a)). The first two
strating the existence of a size effect phenomenon.
stages (points 1 and 2) correspond to the CZ development before
crack initiation occurs. Effectively, the stresses at the crack extrem-
ity in these cases are non-null (Fig. 5(a)), demonstrating that the 2.2. Stress relief region
pre-crack has not yet propagated, although damage is already pres-
ent. As can be observed in Fig. 5(b), the cohesive zone length (lCZ) is The stress relief region (SRR) located in the vicinity of the crack
practically constant between points (3) and (4), which means that plane plays an important role in the bending behavior. Therefore, it
self-similar crack propagation conditions are occurring (Morel turns evident that its configuration might be estimated numeri-
et al., 2010). Consequently, the corresponding stress profiles during cally, to be included in the formulation of the data reduction
these stages are practically coincident (Fig. 5(a)). For the remaining scheme to evaluate GIc. With this aim, a stress analysis along the
stages (i.e., points 5 and 6 in Fig. 5(b)), the lCZ clearly decreases, beam length was performed. Fig. 7 presents profiles of the normal

Table 2
Elastic properties of Norway spruce (Picea abeas L) (Guitard, 1987).

EL (MPa) ER (MPa) ET (MPa) mLT mLR mTR GTL (MPa) GRT (MPa) GRL (MPa)
9900 730 334 0.435 0.430 0.249 610 22 500
 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242 237

y
5 1
2 H 2H

Cohesive
elements
x
H
a0 = H /2

Fig. 4. Mesh used in the numerical analysis.

1.5 1.2
1 2 6 5
3 D1 D2 D3 D4 D5
1.0 4 1.0
D6
0.5 0.8
3
/ft

f/f t
0.0 1
0.6

-0.5 2 0.4
Crack extremity 6
5

4
0.2
-1.0
0.0
-1.5
0 10 20 30 40 50 60
0.0 0.2 0.4 0.6 0.8 1.0 y (mm)
y /(H /2)
Fig. 6. Normalized stress profiles (f/ft) in the cohesive zone at crack initiation for
(a) different specimen sizes.

0.4

1.2
0.3 2.6%H 9.9% H 10.6%H
1.0 15.7%H 19.3%H 24.4%H
l CZ /(H /2)

3 4
5
σxx/σxx max (BT)

0.2 6 0.8

2
0.1
1 0.2
0.0
0.0 0.2 0.4 0.6
a e / (H /2) - 1
(b)
Fig. 5. Plotting of (a) normalized stress profiles (f/ft) as a function of the
normalized crack ligament (y/(H/2)) and the (b) normalized cohesive zone length
(lCZ/(H/2)) as a function of the normalized equivalent crack length (ae/(H/2)-1), for
specimen size D3. Included points refer to following phases of crack propagation:
development of CZ (1, 2); maximum and constant CZ length (3, 4); decrease of CZ
length (5, 6).
stress (rxx) induced by bending for different positions along the
beam height (H) obtained by means of the FEM analysis, normal-
ized with the maximum normal stress obtained through beam the-
ory in the absence of initial crack ðrxx maxðBTÞÞ. As plotted in this
figure, the linear profile is drastically affected by the presence of
the crack and tends to zero in the central region where the crack
is located. It can also be observed that the profiles are differently
affected as a function of their position along the beam height.
With the purpose of clearly identify the SRR shape, plots of the
normal stresses for several cross sections along the beam axis were
also made. Fig. 8 shows the normalized stress profiles for four
0.6
0.4
0.0
0.8 1.0 0.5 0.6 0.7 0.8 0.9 1
x /L
Crack
Fig. 7. Normalized stress profiles ðrxx =rxx maxðBTÞÞ at different heights in the
specimen for size D3 along the normalized (x/L) specimen length.
representative sections. Hence, up to 60% of the specimen
mid-span (i.e., x = 0.6L) good agreement between finite element
(FE) results and the beam theory predictions is observed
(Fig. 8(a)). However, non-negligible differences can be identified
for x = 0.9L. In the vicinity of the crack region (i.e., x = 0.98L),
roughly 25% of the specimen height (H) is free from stresses
(Fig. 8(c)), which confirms the existence of the SRR. Clearly, in this
region a remarkable difference between the FE and the BT results is
evident. Finally, at the specimen mid-span (i.e., x = L), a stress con-
centration at the crack tip (y/H = 0.5) is observed (Fig. 8(d)).
In order to obtain the profile of the SRR, a refined analysis of the
stress profile considering several cross-sections (every 1% of L for
0.9L 6 x 6 L) has then been carried out. The normalized fraction
of the specimen height with null stresses in each cross-section
(hSRR/H) was estimated (Fig. 9) using the normalized position of
238 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242

1.0 1.0
FEA FEA
0.8 BT 0.8 BT

0.6 0.6
y/H

y/H
0.4 0.4

0.2 0.2

0.0 0.0
-1.0 -0.5 0.0 0.5 1.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
σ xx σ xx max
/ σ xx /σ xx max
(a) (b)
1.0 1.0
FEA FEA
0.8 BT 0.8 BT

0.6 0.6
y/H

y/H

0.4 0.4

0.2 0.2

0.0 0.0
-3 -2 -1 0 1 2 3 -4 -1 2 5 8 11 14 17
σ xx /σ xx max σ xx /σ xx max
(c) (d)
Fig. 8. Normalized stress profiles ðrxx =rxx maxÞ along the normalized specimen height (y/H) in four cross-sections (a) x = 0.6L; (b) 0.9L; (c) 0.98L and (d) L.

1.0 linear stress profile is achieved by FE analyses in the compressive

FEA stress region and partially on the tensile region, before the pertur-
BT bation induced by the crack takes place. Therefore, the size of the
stress relief region, hSRR/H, is estimated through
0.8 BT (SRR)
 
hSRR NAFEA NABT
¼2  : ð2Þ
H H H
0.6
This approach provides the adequate estimation of the normalized
y /H

SRR, since the application of the BT to the remaining useful section

(1  hSRR/H) leads to a stress distribution (labeled BT (SRR) in Fig. 9)
NA FEA/H 0.4 that agrees with the stress profile given by the FEA in the above re-
ferred region. This means that in the SRR, the beam with crack be-
NA BT /H haves as an equivalent one with a reduced normalized cross height
h SRR/H equal to 1  hSRR/H. The resulting SRR outline obtained from this
0.2
procedure is presented in Fig. 10. In a previous work (de Moura
et al., 2010), a triangular SRR profile (Fig. 10) was assumed through
a data reduction scheme used to estimate the GIc. However, the
0.0
compliance equation requires an iterative numerical procedure to
-3 -2 -1 0 1 2 3 estimate the equivalent crack length ae, which is not user-friendly.
σ xx /σ xx max To overcome this drawback, a rectangular SRR profile (Fig. 10) is as-
sumed in this work, since it leads to a much more simplified
Fig. 9. Relative position of the neutral axis for cross-section x = 0.98L, estimated
by FEA (NAFEA/H) and by the BT (NABT/H), and respective normalized SRR height
analysis.
(hSRR/H).
2.3. Data reduction scheme

the neutral axes of the stress profiles given by FEA (NAFEA/H) and The size of the adopted rectangular configuration for the SRR is
BT (NABT/H = 0.5). From Figs. 8(a)–(d) it can be seen that an almost assumed to be a function of the current crack length a. In fact, each
 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242 239

1.0 different regions may be defined in the specimen. Dimension L1 de-

fines the arm length (Fig. 11), which means that the elastic modu-
lus used in the bending formulation is EL(Fig. 2). For x < L2 (Fig. 11)
the stress profile is well managed by the beam theory (BT). How-
0.8 Lateral arm Specimen
ever, for L2 6 x 6 L, the rectangular SRR is considered and the size
of the working section is reduced accordingly. Consequently, the
equation of the strain energy due to bending is
0.6 Triangular "Z #
Rectangular Z Z
M2f M 2f M 2f
h SRR / H

L1 L2 L
SRR
SRR U¼2 dx þ dx þ dx ; ð3Þ
0 2EL I L1 2ET I L2 2ET Isr
0.4
being Mf the bending moment (Mf = Px/2) and EL and ET the Young’s
modulus on the longitudinal and transverse direction, respectively
(Fig. 2). The parameters I and Isr represent the second moment of
0.2 area of the entire section (height H) and of the effective section in
Real SRR
Estimated SRRprofile
profile
Crack the SRR (height of H  ka), respectively
3
0.0 bH bðH  kaÞ3
I¼ ; Isr ¼ ð4Þ
0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 12 12
x/L with b standing for the specimen width (Fig. 2). Thus, applying the
Castigliano theorem
Fig. 10. Plotting of real and simplified approaches of the SRR in the normalized SRR
height (hSRR/H) versus normalized specimen length (x/L) space. @U
d¼ ; ð5Þ
@P
to Eq. (3), leads to the equation of the specimen compliance (C = d/P)
" #
y L31 L32  L31 L3  L32
C¼2 3
þ 3
þ : ð6Þ
EL bH ET bH ET bðH  kaÞ3
L
There are several aspects that are not accounted for in this equation.
L1
Effectively, the shear effects were not considered, although they can
be non-negligible for large specimens. Moreover, there is a stress
x concentration at the crack-tip, which alters the BT stress profile.
H On the other hand, it is known that wood (being a natural material)
ka a presents a strong variability on its elastic modulus (de Moura et al.,
2008; Dourado et al., 2010). The aforementioned problems are ac-
L2 ka SRR counted for by means of the definition of a corrected value of the
flexural modulus (ETf), defined on the basis of the initial values of
Fig. 11. Notation used for geometry definition. the crack length (a0) and the corresponding measured specimen
compliance (C0) using Eq. (6)
! !1
time the crack grows by a small increment da, an additional quan- L32  L31 L3  L32 C0 L31
ETf ¼ þ  : ð7Þ
tity of strain energy is released from the newly-unloaded material bH
3
bðH  ka0 Þ3 2 EL bH3
near the crack. Considering the half-specimen (Fig. 11), the SRR be-
comes a square with edge length of ka, being k a non-dimensional In regards to the elastic modulus of the specimen arms (EL), no cor-
parameter to be discussed below. According to this figure three rection is necessary, since it was proved by Morel et al. (2005) that

50000
2400 D6

40000
D3
1800
Load (N)

30000
Load (N)

1200
20000
D5
D2
600
10000
D4
D1
0 0
0 2 4 6 8 0 2 4 6 8
Displacement (mm) Displacement (mm)

Fig. 12. Load–displacement curves obtained in FEM computations for all specimen sizes (see Table 1 for included labels).
240 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242

1.2 1.0 1.2 1.0

D1 D2
1.0 1.0
0.8 0.8
G I/GIc (inp)
G I / G Ic (inp)

0.8 0.8

G I / G Ic (inp)
G I/G Ic (inp) 0.6

l CZ /(H /2)
0.6

lCZ/(H /2)
k = 0.82
0.6 k = 0.84 0.6
0.4 0.4
0.4 l CZ/(H /2)
0.4 l CZ/(H /2)
0.2 0.2
0.2 0.2

0.0 0.0 0.0 0.0

0.00 0.05 0.10 0.15 0.20 0 0.05 0.1 0.15
ae / (H /2) - 1 a e /(H /2) - 1

(a) (b)
1.2 1.0
1.2 1.0
D3
1.0 D4
0.8 1.0
0.8
0.8 0.8
G I/G Ic (inp) G I/G Ic (inp)
0.6
G I / G Ic (inp)

0.6

l CZ /(H /2)
k = 0.80
G I / G Ic (inp)
l CZ /(H /2)

0.6 0.6 k = 0.80

0.4 0.4
0.4 0.4
l CZ/(H /2) 0.2 0.2
0.2 0.2 l CZ/(H /2)

0.0 0.0 0.0 0.0

0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08
ae /(H /2) - 1 a e /(H /2) - 1

(c) (d)
1.2 1.0 1.2 1.0
D5 D6
1.0 1.0
0.8 0.8

0.8 G I/G Ic (inp) 0.8 G I/G Ic (inp)

G I / G Ic (inp)

G I / G Ic (inp)

k = 0.80 0.6 0.6

lCZ /(H /2)

l CZ/(H /2)
k = 0.80
0.6 0.6
0.4 0.4
0.4 0.4

0.2 0.2
0.2 l CZ/(H /2) 0.2 l CZ/(H /2)

0.0 0.0 0.0 0.0

0 0.01 0.02 0.03 0.04 0 0.00625 0.0125 0.01875 0.025

ae /(H /2) - 1 ae /(H /2) - 1

(e) (f)
Fig. 13. Normalized R-curves (GI/GIc(inp)) and cohesive zone lengths (lCZ/(H/2)) for analyzed specimen sizes: (a) D1; (b) D2; (c) D3; (d) D4; (e) D5; (f) D6. The dots identify the
points corresponding to Pu.

8 2 !1 31=3 9
its contribution to the specimen compliance is negligible. Another
1< 4 C L31 L32  L31 L3  L32 5 =
problem typical of fracture tests in wood is related to the difficulty ae ¼ H   : ð8Þ
of crack length monitoring during its growth with the necessary k: 2 EL bH3 ETf bH3 ETf b ;
accuracy, owing to unstable propagation and development of pro-
nounced fracture process zones (FPZ) at the crack tip. Concurrently,
Consequently, fracture energy can now be obtained by means of the
the crack length evaluation is a fundamental parameter for tough-
Irwin–Kies equation
ness evaluation using the classical data reduction schemes. Hence,
to overcome this difficulty, an equivalent crack length approach
based on the specimen compliance is proposed in this work. P2 dC
GI ¼ ; ð9Þ
Accordingly, an estimate of the equivalent crack length ae can be 2b da
obtained, isolating ae from Eq. (6) and using ETf instead of ET, as
follows: applied to Eq. (6), which leads to
 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242 241

 
3P2 L3  L32 k 60
GI ¼ 2 4
: ð10Þ
b ETf ðH  kae Þ
50
D4 (1120) D5 (2240) D6 (4480)
Using this methodology, the Resistance-curve (R-curve) as a function 40 D3 (560)

(mm)
of the equivalent crack (i.e., GI = f(ae)) is easily obtained by means of
D2 (280)
Eqs. (7), (8) and (10). The R-curve propitiates an indubitable estima- 30
D1 (140)
tion of the fracture energy, providing that a plateau value is

CZ
attained.

l
20

3. Results
The specimen sizes used in this analysis are listed in Table 1. It
should be noted that the presented sizes were selected in order to
study numerically some aspects involving this specimen geometry
(SEN-TPB). In fact, bigger sizes of the selected range are obviously
out of scope of any experimental campaign.
The application of the above described data reduction scheme
(Eqs. (7), (8) and (10)) and load–displacement curves (Fig. 12) pro-
vide the computation of the R-curves (Figs. 13(a)–(f)). The value of
parameter k defining the dimension of the SRR that provides a
good agreement between the plateau of the R-curve and the input-
ted fracture energy (GIc (inp)) in the cohesive model, is presented
for each R-curve. It can be observed that the value of k converges
to 0.8 from specimen size D3 to D6. Additionally, from this size
(i.e., D3) the extent of the cohesive zone (lCZ) presents an indubita-
ble plateau concurrently to the plateau of the R-curve. This means
that the conditions of self-similar crack propagation are present for
a certain crack extent, thus providing correct measurements of the
fracture energy. It can be concluded that the specimen with the
lowest size (i.e., D1) is clearly being affected by spurious effects,
i.e., the compressive stresses due to bending affect the fracture
process from the beginning, owing to the specimen small size.
Effectively, the plateau of the R-curve is not followed by a corre-
sponding trend on the lCZ curve (Fig. 13(a)), which means that frac-
ture is being affected by spurious effects from the beginning. As a
consequence, the adequate value of k (i.e., 0.84) is higher than the
converged one (i.e., 0.8). Specimen of size D2 can be considered as
being in the limit, since a quite short plateau on the lCZ curve is
formed simultaneously with the plateau of the R-curve, and only
a slight difference (0.82 instead of 0.8) was detected in the value
of k. It has been verified that an acceptable error of 5.0% on the
toughness (GIc) is found when parameter k is set to 0.8 for this
specimen size. This error magnitude is negligible in wood since a
10
0
0 500 1000 1500 2000 2500
H /2 (mm)
Fig. 14. Damage zone extent (lCZ) at crack initiation for the analyzed specimen
sizes.
scatter of 20% is frequently found on its properties (de Moura
et al., 2009).
Another interesting aspect has to do with the position, in these
curves (Fig. 13(a)–(f)), of the point corresponding to the ultimate
load Pu. It can be observed that as the specimen increases in size
the plateau onset is reached closer and closer to the ultimate load
Pu. When the maximum load is reached before the stable condi-
tions of crack propagation take place, this means that the extent
of the damaged zone has already attained a value relatively to
the ligament length that induces load reduction, even that no crack
growth had occurred.
In addition to this analysis, it can be seen from Fig. 14 that the
size of the CZ tends to a constant value as the specimen increases
in size, reinforcing the idea that spurious effects vanish for these
cases. This highlights that higher specimen sizes propitiate better
conditions of self-similar crack growth. Moreover, this numerical
analysis demonstrates that considered higher sizes (i.e., D4–D6)
are unnecessary, since they propitiate similar results to size D3.
It should be emphasized that these achievements can be altered
if a different material is analyzed. In fact, the FPZ size depends on
fracture toughness and local strength, which means that higher
dimensions could be necessary to achieve self-similar crack growth
conditions if a material with, for example, much higher toughness
is tested. In the context of wood a scatter of ±20% on its mechanical
properties is normal. Consequently, the influence of this scatter on
the measured toughness was assessed maintaining the value of

1.4
2500
1.2
2000 k = 0.80
1.0
G I / G Ic (inp)
Load (N)

1500 0.8 E T = 267.2 MPa

E T = 334.0 MPa
0.6
1000 E T = 400.8 MPa E T = 400.8 MPa
E T = 334.0 MPa 0.4
500 E T = 267.2 MPa
0.2

0 0.0
0 1 2 3 4 0 0.015 0.03 0.045 0.06 0.075 0.09
Displacement (mm) ae /(H /2) - 1
(a) (b)
Fig. 15. Effect of ET on the measured toughness considering the same value of k. (a) Load–displacement curves and respective (b) R-curves.
242 N. Dourado et al. / International Journal of Solids and Structures 48 (2011) 234–242
2.0
1.5 1.2 G ref
Stress, f (MPa)
G ref
1.0
0.8 G ref
0.5
0.0
0.00 0.05 0.10 0.15
Crack opening, w (mm)
(a)
Fig. 16. Effect of the cohesive parameters on the R-curve. (a) Stress-softening diagram and (b) respective R-curves.
parameter k and considering the specimen size D3. The sensitivity
analysis was performed by varying ±20% the parameters ET and GIc,
since it was verified to be the most influencing on the specimen
mechanical behavior. Figs. 15 and 16 show a small influence on
the plateau values given by the R-curves, thus demonstrating that
the proposed methodology to determine toughness is adequate.
4. Conclusions
In this study the single edge notched beam loaded in three-
point bending test was thoroughly analyzed in order to verify its
adequacy to be used for wood fracture characterization under
mode I loading. This test is particularly suitable for wood crack
propagation systems where the double cantilever beam test is
not applicable. However, there are several aspects hindering a
straightforward determination of fracture toughness using this
specimen geometry which deserve a deep analysis.
Initially, a stress analysis along the ligament length, obtained
through cohesive modeling, showed that the stress profile is af-
fected by spurious effects induced by compressive stresses due to
bending. It was verified that these spurious effects increase as
the specimen decreases in size, owing to the reduction of the liga-
ment length.
On the other hand, a stress analysis along the beam length was
performed to identify the profile of the stress relief region in the
vicinity of the crack. A new data reduction scheme accounting
for the stress relief region was then developed assuming for sim-
plicity a rectangular shape for this zone. This approach, although
less rigorous than the triangular one used in a previous work, pro-
vides a much easier procedure to estimate the Resistance-curve.
The method based on the beam theory and on the crack equivalent
concept does not require crack monitoring during its growth and
accounts for scatter of the elastic properties characteristic of wood.
A cohesive zone modeling was performed using the bilinear
softening law to simulate damage initiation and growth, involving
six homothetic specimens. It was observed that the value of the
parameter defining the stress relief region that propitiates the pla-
teau equal to the inputted GIc tends to a constant for higher spec-
imen sizes. Additionally it was verified that the small specimen
sizes do not provide a constant cohesive zone length concurrently
to the plateau of the R-curve. Moreover, the analysis of the position
of the point corresponding to the ultimate load in the R-curves and
lCZ curves allowed reinforcing the idea that the lowest specimen is
clearly inadequate. This study allowed verifying that size effect
practically does not exist for sizes greater or equal to D3. Taken into
1.2
1.0
0.8
G I /G Ic (inp)
1.2 G ref
G ref
0.6
0.8 G ref
0.4
0.2
0.0
0.20 0.25 0 0.015 0.03 0.045 0.06 0.075 0.09
a e /(H/ 2) - 1
(b)
account the reasonable dimensions associated with the acceptable
error observed, the size D2 can be assumed to be the best compro-
mise for this material system. It should be noted that this conclu-
sion can be altered if a different material is used. In fact, the size of
the fracture process zone, which depends on the material, plays a
fundamental role on the referred constraints.
Acknowledgement
The authors thank the Portuguese Foundation for Science and
Technology for supporting the work here presented, through the
research project PDTC/EME-PME/64839/2006.
References
Daudeville, L., 1999. Fracture in spruce: experiment and numerical analysis by
linear and non linear fracture mechanics. Holz als Roh-und Werkstoff 57, 425–
432.
de Morais, A.B., de Moura, M.F.S.F., Gonçalves, J.P.M., Camanho, P.P., 2003. Analysis
of crack propagation in double cantilever beam tests of multidirectional
laminates. Mechanics of Materials 35, 641–652.
de Moura, M.F.S.F., Morais, J., Dourado, N., 2008. A new data reduction scheme for
mode I wood fracture characterization using the double cantilever beam test.
Engineering Fracture Mechanics 75, 3852–3865.
de Moura, M.F.S.F., Silva, M.A.L., Morais, J.J.L., de Morais, A.B., Lousada, J.J.L., 2009.
Data reduction scheme for measuring GIIc of wood in end-notched flexure (ENF)
tests. Holzforschung 63, 99–106.
de Moura, M.F.S.F., Dourado, N., Morais, J., 2010. Crack equivalent based method
applied to wood fracture characterization using the single edge notched-three
point bending test. Engineering Fracture Mechanics 77, 510–520.
Dourado, N., Morel, S., de Moura, M.F.S.F., Valentin, G., Morais, J., 2008. Comparison
of fracture properties of two wood species through cohesive crack simulations.
Composites Part A 39, 415–427.
Dourado, N., de Moura, M.F.S.F., Morais, J., 2010. Single-edge-notched beam loaded
in three-point-bending. Holzforschung 64, 119–126.
Guitard, D., 1987. Mécanique du matériau bois et composites. Cépadues-Editions.
ISBN: 2.85428.152.7, pp. 108–123.
Gustafsson, P.J., 1988. A study of strength of notched beams. In: Proceedings of CIB-
W18A Meeting, Parksville, Canada, Paper 21-10-1.
Morel, S., Dourado, N., Valentin, G., Morais, J., 2005. Wood: a quasibrittle material.
R-curve behavior and peak load evaluation. International Journal of Fracture
131, 385–400.
Morel, S., Lespine, C., Coureau, J.-L., Planas, J., Dourado, N., 2010. Bilinear softening
parameters and equivalent LEFM R-curve in quasibrittle failure. International
Journal of Solids and Structures 47, 837–850.
Nordtest Method, 1993. NT BUILD 422. Wood: Fracture Energy in Tension
Perpendicular to the Grain.
Petersson, P.E., 1981. Crack Growth and Development of Fracture Zone in Plain
Concrete and Similar Materials. Report No. TVBM-1006, Division of Building
Materials, Lund Institute of Technology, Lund, Sweden.
Stanzl-Tschegg, S.E., Tan, D.M., Tschegg, E.K., 1995. New splitting method for wood
fracture characterization. Wood Science of Technology 29, 31–50.
Yoshihara, H., Kawamura, T., 2006. Mode I fracture toughness estimation of wood
by DCB test. Composites: Part A 37, 2105–2113.