Engineering Fracture Mechanics 77 (2010) 510–520

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

Crack equivalent based method applied to wood fracture

characterization using the single edge notched-three point bending test
M.F.S.F. de Moura a,*, N. Dourado b, J. Morais b
a
Faculdade de Engenharia da Universidade do Porto, Departamento de Engenharia Mecânica e Gestão Industrial, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
b
CITAB/UTAD, Departamento de Engenharias, Quinta de Prados, 5000-911 Vila Real, Portugal

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

Article history: In this work the single edge notched-three point bending test applied to wood fracture
Received 4 May 2009 characterization in mode I is analysed. A new data reduction scheme founded on the equiv-
Received in revised form 24 October 2009 alent linear elastic fracture mechanics with remarkable advantages relatively to classical
Accepted 29 October 2009
methods is proposed. The method is based on the beam theory accounting for a triangular
Available online 1 November 2009
stress relief region and on the crack equivalent concept. It does not require crack monitor-
ing during its propagation and provides a complete R-curve which is fundamental for a
Keywords:
clear identification of the fracture energy. The model was validated numerically by means
Wood
Fracture mechanics
of a bilinear cohesive damage model which allows the simulation of damage initiation and
Crack growth growth. The size of the stress relief region in the neighbouring crack zone was estimated
Civil engineering structures and a consistent value was found relatively to the one achieved by other authors.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Wood structural applications are increasing nowadays due to economical and ecological reasons. In general, the design of
wood structures is based on classical strength of materials approaches [1] which provide conservative predictions. More re-
cently, fracture mechanics approaches are gaining more prominence, namely the methods based on energetic analysis [2–5].
In this context the definition of accurate methods for measuring wood fracture properties, namely under pure mode I load-
ing, acquires special relevancy.
At the macroscopic scale wood is generally assumed to be a continuum, homogeneous and orthotropic material with
three well defined directions: longitudinal (L) along the fibres, radial (R) and tangential (T), which are normal and tangent,
respectively, to the annual growth rings (Fig. 1a). Six principal systems of crack propagation can be defined: TL, RL, LR, TR, RT
and LT (Fig. 1b). The first letter gives the normal direction to the crack plane and the second one indicates the direction of
crack propagation. The TL and RL crack propagation systems constitute the most frequent ones because of the particular de-
sign of timber structures. On the other hand, the crack propagation in the LR or LT systems, i.e., across wood fibres, is rare.
The Double Cantilever Beam test (DCB) is particularly adequate for mode I fracture characterization due to its simplicity and
the possibility to use the beam theory to measure the fracture energy [5,6]. However, the DCB is only appropriate for the RL
and TL crack propagation systems since in these cases crack propagates along wood fibres, i.e., in well defined planes. For the
remaining fracture systems (i.e., TR and RT) the DCB is not appropriate owing to a pronounced stick–slip effect. As a conse-
quence, the crack immediately deviates from the mid-plane thus originating a mixed-mode loading instead of a pure mode I.
Additionally, taking into account the curvature of the growth rings it is impractical to obtain DCB specimens for the fracture
systems TR and RT. In fact, wood is commonly regarded as a cylindrically orthotropic material at locations sufficiently far

* Corresponding author.
E-mail address: [email protected] (M.F.S.F. de Moura).

0013-7944/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2009.10.008
 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520 511

Nomenclature

a real crack length

a0 initial crack length
ae equivalent crack length
b specimen width
C specimen compliance
Em (m = L, R, T) Young’s moduli
ETf corrected flexural modulus
fb strength at the onset of bridging
ft local strength
GI strain energy release rate
GIb fracture energy due to fibre-bridging
GIc fracture toughness
GIc(inp) inputted value of fracture energy release rate
GIl fracture energy due to micro-cracking
Gm,n (m, n = L, R, T) shear moduli
H specimen height
h(x) specimen height of the effective section
I second moment of area of the entire section
Isr second moment of area of the effective section
L half length of the beam
L1 length of the lateral arms
L2 characteristic length defining the position of the SRR
lFPZ extent of the FPZ
Mf bending moment
P external applied central load
U bending strain energy
wb crack opening at the onset of bridging
b constant defining the SRR
d applied displacement
mm,n (m, n = L, R, T) Poisson’s ratios

Acronyms
CT compact tension
DCB double cantilever beam
FEA finite element analysis
FPZ fracture process zone
SEN-TPB single edge notched-three point bending
SRR stress relief region
WS wedge-splitting

from the pith (cylindrical origin) where the effect of curvature of the growth rings is not so pronounced. Additionally, for a
given specimen geometry, the tested volume should be free from knots, which is very difficult to achieve in the DCB owing to
its length. This is particularly important when a size effect study is pretended. In these cases alternative test methods must
be used as is the case of the compact tension test [7], wedge-splitting (WS) test [2] and the single edge notched-three point
bending (SEN-TPB) test [4,8,9] (Fig. 2). The compact tension test (CT) has been applied by Boström [7] in order to estimate
theoretically the initiation and growth of the fracture process zone (FPZ) in wood, which was considered fundamental to
determine the critical stress intensity factor. The author concluded that the load–displacement curve is sensitive to the mod-
ulus of elasticity perpendicular to the grain, the tensile strength, the fracture energy and the specimen size. The wedge-split-
ting (WS) test proposed by Stanzl-Tschegg et al. [2] consists of two load transmission pieces and a slender wedge with an
angle of 5–10°. The energy losses are drastically diminished by means of roll bodies which transmit the loads from the wedge
to the specimen. Therefore, the vertical load applied by the compression machine originates high horizontal splitting forces,
which leads to stable crack initiation and growth. The SEN-TPB is easier to execute than the WS test since it does not require
special devices, and it is more appropriate to undergo tests involving different specimen sizes, which are frequently per-
formed in wood fracture analyses (size effect studies). Gustafsson [8] reported that the SEN-TPB requires relatively small
dimensions in order to get consistent fracture parameters. This is important to obtain a test volume free from knots, as it
is recognized that knots increase considerably the fracture energy. Daudeville [9] used the SEN-TPB specimen to measure
the fracture energy in tension perpendicular to grain in spruce (denoted as RL and TL in the literature). The author used this
geometry to perform size effect studies. More recently, Dourado et al. [4] employed the SEN-TPB to estimate the cohesive
fracture properties in two wood species used in timber construction: Maritime pine (Pinus pinaster Ait.) and Norway spruce
512 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520

TL RL

LR TR

L
RT LT

(a) (b)
Fig. 1. Wood anatomic axis: (a) longitudinal (L), radial (R) and tangential (T) directions; (b) fracture systems in wood.

P, δ

a0

2L
Fig. 2. Schematic representation of the SEN-TPB test.

(Picea abies L.). The fracture process zone (FPZ) interaction with the specimen boundaries has revealed to be a crucial issue on
the definition of the appropriate specimen size. The authors verified that the extent of the self-similar crack growth depends
on the wood species used in the experiments, with spruce exhibiting a considerable larger FPZ extent than pine. As the inter-
action of the FPZ with the specimen boundaries leads to the overestimation of the energy release rate, then the appropriate
choice of the ligament length in the SEN-TPB becomes a key issue.
Notwithstanding the cited advantages of the SEN-TPB, this test presents several drawbacks that hinder a clear identifica-
tion of the fracture energy using the classical data reduction schemes. The objective of this work is to develop a suitable data
reduction scheme to be applied on the SEN-TPB test for wood fracture characterization under pure mode I loading. The meth-
od is based on the beam theory and crack equivalent concept thus overcoming the difficulty of crack measurement during its
propagation. A triangular stress relief zone in the vicinity of the crack is taken into account. A cohesive zone model was ap-
plied to validate the proposed method. The obtained results demonstrate that with the proposed procedure the inputted
fracture energy in the cohesive zone model is well captured which validates the proposed data reduction scheme. Addition-
ally, the proposed data reduction scheme was used to estimate the minimum specimen size which avoids the confinement of
the FPZ thus providing adequate measurements of the fracture energy.

2. Numerical analysis

The specimen is constituted by a central region with bonded arms (Fig. 3). The central part is oriented to fit with the cho-
sen wood fracture system (i.e., TL) and is rigidly bonded to the remaining elements (lateral arms). These elements are aligned
with the wood fibre direction (i.e., L) to assure high bending stiffness. The use of both arms in the SEN-TPB is crucial, since as
 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520 513

the loading-span gets higher it turns impossible to induce fracture in the TL system in a continuum wood specimen. It should
be noted, that the lateral arms can be reused several times.
In the present analysis, the central part of the specimen allows the fracture characterization of the TL crack propagation
system. Although the objective of this test is to be applied in planes where the DCB does not perform well (TR and RT), this
system was chosen to validate the proposed data reduction scheme since it was already characterised in a previous work [4].
Geometrically similar specimens of different sizes (homothetic) (Table 1) were analysed with the objective to identify appro-
priate dimensions to evaluate fracture energy. This issue is fundamental on the context of a size effect study in wood [10].
Hence, a finite element analysis (FEA) using plane strain elements and a cohesive damage model [11] was performed. Due to
symmetry conditions only half specimen was simulated. In regards to the FE-model (Fig. 4) used in the numerical simula-
tions, the mesh has been implemented to provide a ligament length uniformly divided in every 0.5 mm for the totality of
the performed analyses. In Fig. 4 the mesh corresponding to the series D2 with 837 solid elements (8-node rectangular
and 6-node triangular) and 64 cohesive elements (6-node) is presented. Norway spruce (Picea abies L.) material properties
(Table 2) were used in the simulations.

3H H 3H

T
H
L L L
a0 = H /2
R R
T R T
b
Initial crack not
notch

Fig. 3. Parts setup before bonding showing the wood anatomic axis.

Table 1
Dimensions of the set of specimens used in the numerical analyses.

Series H (mm) L1 2.5 H (mm) L (mm) a0 (mm) b (mm)

D0 17.5 43.75 52.5 8.75 5
D1 35 87.5 105 17.5 10
D2 70 175 210 35 20
D3 140 350 420 70 40
D4 280 700 840 140 80
D5 560 1400 1680 280 160

y
5 1
2 H 2H

H
H
2
x

Fig. 4. Mesh used in the numerical simulations.

Table 2
Elastic properties of Norway spruce (Picea abeas L.) [12, Guitard].

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
514 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520

ft

G I μ (Component: "micro-cracking")

Stress

fb G I b (Component: "fibre-bridging")

w
wo wb wc
Crack opening

Fig. 5. Bilinear Petersson’s cohesive damage model [11].

Table 3
Cohesive parameters of Norway spruce according to [4].

ft (MPa) fb (MPa) wb (mm) GIc (N/mm)

1.660 0.300 0.090 0.145

2000

1600
1/C
Load, P (N)

1200

800

400

0
0 1 2 3 4 5 6 7 8
Displacement, δ (mm)

Fig. 6. P–d curve obtained for series D5.

A bilinear damage model (Fig. 5) was applied to simulate damage initiation and growth, using the set of cohesive param-
eters shown in Table 3. This model is particularly appropriate for wood since two different fracture phenomena are easily
identified during fracture in this material [2,10]: micro-cracking simulated by the first branch of the softening law and fi-
bre-bridging by the second one (Fig. 5). More details of the cohesive damage law can be found in references [4,5]. The loading
is applied by displacement control with very small increments to avoid unstable propagation. During crack growth the load–
displacement curve (Fig. 6) is the only data to be recorded since it will be fundamental to get the R-curve. Finally, the per-
formance of the method will be assessed by comparing the GIc measured from the plateau of the R-curve with the inputted
value (GIc(inp)) in the cohesive damage model.

3. Data reduction scheme

3.1. Model presentation

The classical data reduction schemes usually applied to measure the fracture energy are not adequate for the SEN-TPB
specimen. The compliance calibration methods require crack monitoring during its growth which is not possible to perform
 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520 515

in this test. In fact, crack propagates rapidly and the presence of fibre-bridging hinders a clear identification of its tip. In addi-
tion, the extent of the FPZ developed ahead of the crack tip can be affected in some specimen sizes by the normal compres-
sive stresses induced by bending above the specimen neutral axis [4]. This spurious effect alters the conditions of self-similar
crack growth thus influencing the measured energy. In order to overcome these drawbacks and to identify adequate spec-
imen sizes, a new data reduction scheme based on a modified beam theory and crack equivalent concept is proposed.
One of the aspects that cause difficulties on the direct application of the beam theory concepts is related to the presence
of the crack which alters the bending stresses profile in its neighbourhood. Kienzler and Herrmann [13] suggested that in this
specimen geometry an approximately triangular stress relief region (SRR) can be identified, as sketched in Fig. 7. Inside the
stress relief region the normal stresses induced by bending drop approximately to zero and this must be taken into account
in the analysis. Following the notation used to define the specimen geometry in Fig. 7, the SRR develops for L2 6 x 6 L, which
means that for x < L2 the stress profiles are well approached by beam theory. Thus, the strain energy due to bending can be
written as
"Z Z Z #
L2
M2f L1
M 2f L
M 2f
U¼2 dx þ dx þ dx ð1Þ
0 2EL I L2 2EL Isr L1 2ET Isr

where L1 is the useful length of the lateral arms, Mf is the bending moment (Mf = Px/2) and EL and ET the Young’s modulus on
the longitudinal and transverse direction, respectively (Fig. 3). The parameters I and Isr represent the second moment of area of
the entire section (height H) and of the effective section in the SRR (height h(x)), respectively (Fig. 7). They can be written as
3
bH bðhðxÞÞ3
I¼ ; Isr ¼ ð2Þ
12 12
being b the specimen width (Fig. 3) and
a
hðxÞ ¼ H þ ðL2  xÞ ð3Þ
L  L2
where a is the crack length (Fig. 7). Substituting Eqs. (2) and (3) in Eq. (1), and applying the Castigliano theorem,
@U
d¼ ð4Þ
@P
the applied displacement d (Fig. 2) is obtained. The resulting equation of the specimen compliance (C = d/P) can be written as
( "  #  2 #
2L32 6ðL  L2 Þ 1 L21 L22 L  L2 L2 L1 L  L2 H
C¼ 3
þ  þ  þ ln
EL bH ba EL 2ðhðL1 ÞÞ2 2H2 a H hðL1 Þ a hðL1 Þ
"     #)
2
1 L2 L21 L  L2 L1 L L  L2 hðL1 Þ
þ 2
 2
þ  þ ln ð5Þ
ET 2ðH  aÞ 2ðhðL1 ÞÞ a hðL1 Þ H  a a Ha

where h(L1) is obtained from Eq. (3) considering x = L1. The previous equation establishes the C = f(a) relationship taking into
account a triangular SRR (Fig. 7). However, there are additional issues that can affect the accuracy of Eq. (5) before the dam-
age onset. Shear effects were not considered and they can be non negligible namely for higher specimens. On the other hand,
due to stress concentrations in the neighbourhood of the crack tip the typical stress profile induced by bending is disturbed.
Finally, it is known that wood presents important scatter in its elastic properties [14], which means that ET (Fig. 3) can vary
markedly from specimen to specimen. In order to overcome these inaccuracies, a corrected flexural modulus (ETf) should be
defined and used instead of the nominal one. This ETf can be easily obtained for each specimen from Eq. (5) considering the
initial values of compliance C0 and crack length a0, which must be accurately measured. For the modulus EL no correction
was made since the specimen lateral arms used in the experiments are reused (normally the same arms are used for all tests

L
L1

h (x) H
a
x

L2 βa SRR

Fig. 7. Notation used for the geometry definition.

516 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520

performed for a given specimen size), being previously characterised. Moreover, Morel et al. [3] demonstrated that the con-
tribution of EL to the specimen compliance is negligible for the SEN-TPB test (Fig. 3).
As already referred the crack length during its growth cannot be accurately measured in these tests. However, an equiv-
alent crack (ae) can be defined as a function of the current compliance via Eq. (5). It should be noted that this equivalent
linear elastic fracture mechanics concept is restricted to quasi-brittle materials in the sense that the unloading specimen
compliance accounts for all the nonlinear phenomena linked to the FPZ development and/or crack propagation. Since no ana-
lytical solution is possible to obtain, Eq. (5) has been solved numerically using the bisection method in order to obtain the ae
for each point of the P–d curve. The strain energy release rate is obtained by means of the Irwin-Kies equation

P2 dC
GI ¼ ð6Þ
2b da
which, considering ae instead of a in Eq. (5), gives
 " 2 !  #
3P 2 ðL  L2 Þ
1 L2 L21 ðH þ 3kae Þ ðL  L2 Þ L1 ð2H þ 3kae Þ 2L2 ðL  L2 Þ2 3 H k
GI ¼ 2
 þ   ln þ
b EL 2H2 a2e 2a2e ðH þ kae Þ3 a3e ðH þ kae Þ2 H a3e ae H þ kae H þ kae
" !
1 L2 ð3ae  HÞ L2 ðH þ 3kae Þ ðL  L2 Þ L1 ð2H þ 3kae Þ Lð3ae  2HÞ
þ 3
þ 1  þ
ETf 2a2e ðH  ae Þ 2a2e ðH þ kae Þ3 a3e ðH þ kae Þ2 ðH  ae Þ2
  #)
ðL  L2 Þ2 Hð1 þ kÞ 3 H þ kae
þ  ln ð7Þ
a3e ðH  ae ÞðH þ kae Þ ae H  ae

being k = (L2  L1)/(L  L2). Following this procedure an R-curve is obtained without measuring the crack length during its
propagation. The sole parameter that must be identified is the length L2 which defines the SRR (Fig. 7). This will be discussed
in Section 4.

0.030
-15 6 -12 5 -9 4
C = 2.692×10 a 0 - 5.924×10 a 0 + 5.411×10 a 0 -
0.025 -6 3 -4 2 -1
2.623×10 a 0 + 7.118×10 a 0 - 1.024×10 a 0 + 6.108
Compliance, C (mm/N)

0.020

0.015

0.010

0.005

0.000
260 300 340 380 420 460 500 540
a 0 (mm)

Fig. 8. Compliance calibration curve.

400
Equivalent crack length , a e (mm)

300

200 a e ELEFM
a_ELFM

a e NDRS
a_CBBM

100

0
0.001 0.0012 0.0014 0.0016 0.0018 0.002
Compliance, C (mm/N)

Fig. 9. Comparison between equivalent cracks provided by the two methods: the new data reduction scheme (ae NDRS) and the equivalent LEFM (ae ELEFM)
proposed in [3].
 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520 517

3.2. Model validation

In order to verify the performance of the compliance function presented in Eq. (5) it was decided to compare the equiv-
alent crack length provided by this function with the one obtained through an alternative equivalent linear elastic fracture
mechanics (LEFM) approach. This second method was already applied successfully in a previous work [3] and is based on the
establishment of a compliance calibration curve as a function of different initial crack length values (a0) for a given specimen
size. Thus, the initial compliance obtained by FEA for each selected a0 is used to define a polynomial function (C = f(a0))
which is subsequently applied to the load–displacement data of the fracture test to estimate the equivalent crack length. This
method was applied to specimen size D5 (Table 1) and the C = f(a0) curve obtained considering a step of 20 mm for a0 in the
range 280 6 a0 6 500 mm (Fig. 8). The fitted polynomial function C = f(a0) was then used to estimate the ae from the current
unloading compliance C = d/P (Fig. 6). Fig. 9 shows the good agreement obtained for the ae provided by the two data reduc-
tion schemes in the selected range.

1.2 1.0 1.2 1.0

β = 1.25 β = 1.0
1.0 0.8 1.0 0.8
G I / G Ic(inp)

G I / G Ic(inp)
l FPZ/(H /2)

l FPZ/(H /2)
0.8 0.8 G I/G Ic(inp)
0.6 0.6
0.6 l FPZ/(H /2) 0.6 l FPZ/(H /2)
0.4 0.4
0.4 G I/G Ic(inp) 0.4

0.2 0.2 0.2

0.2

0.0 0.0 0.0 0.0

8.75 10.75 12.75 14.75 16.75 17.50 20.00 22.50 25.00 27.50 30.00
a e (mm) a e (mm)
(a) (b)
1.2 1.0 1.2 1.0
β = 1.07 β = 1.07
1.0 1.0 0.8
0.8
G I / G Ic(inp)
G I / G Ic(inp)

l FPZ/(H /2)
l FPZ/(H /2)

0.8 G I/G Ic(inp) 0.8

0.6 0.6
0.6 0.6
0.4 G I/G Ic(inp) 0.4
0.4 l FPZ/(H /2) 0.4
0.2 l FPZ/(H /2) 0.2
0.2 0.2

0.0 0.0 0.0 0.0

35 40 45 50 55 70 75 80 85 90 95

a e (mm) a e (mm)
(c) (d)
1.2 1.0 1.2 1.0
β = 1.07 β = 1.07
1.0 0.8
1.0 0.8
G I / G Ic(inp)
G I / G Ic(inp)

l FPZ/(H /2)
l FPZ/(H /2)

0.8 0.8
0.6 0.6
0.6 0.6
l FPZ/(H /2) 0.4 0.4
G I/G Ic(inp) G I/G Ic(inp)
0.4 0.4 l FPZ/(H /2)

0.2 0.2 0.2

0.2

0.0 0.0 0.0 0.0

140 147 154 161 168 175 280 290 300 310 320 330
a e (mm) a e (mm)
(e) (f)
Fig. 10. R-curves for the six modelled specimen sizes presented in Table 1: (a) D0, (b) D1, (c) D2, (d) D3, (e) D4, (f) D5.
518 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520

4. Results

Six homothetic specimen sizes were analysed (Table 1). The goal was to verify which were the dimensions allowing an
adequate replication of the inputted GIc (i.e., GIc(inp)), in the plateau region of the R-curve. On the other hand, the estimation
of the length L2 was also necessary. In the following discussion the horizontal cathetus of the triangular SRR was assumed to
be a linear function of the crack length, i.e., L  L2 = ba, being b a constant to be defined. In fact, when the crack grows the
resistant ligament extent (H–a) diminishes (Fig. 7), thus decreasing L2 and increasing the SRR. Kienzler and Herrmann [13]
argued that b is of the order of unity for this type of test thus pointing to an isosceles triangle defining the SRR.
Figs. 10a–f exhibit the R-curves obtained for the specimen sizes D0 to D5 (Table 1), by means of the data reduction scheme
detailed in Section 3. The strain energy release rate is normalized by the inputted value (GIc(inp)). For each case, the value of
b originating a plateau approximately equal to unity is presented. It is verified that for H P 70 mm a constant value is ob-
tained (b = 1.07). This value is consistent with the results of Kienzler and Herrmann [13]. However, for H < 70 mm (D0 and
D1) different values for b are obtained, and the plateau practically does not exist (H = 17.5 mm) or is quite short (H =
35 mm). These observations clearly indicate that these dimensions are not adequate for fracture characterization of this
material with this test. As discussed in [4], depending on the specimen size and the extent of the FPZ of the tested material,
spurious effects can arise. As a matter of fact, the normal compressive stresses induced by bending above the specimen neu-
tral axis hinder a self-similar crack growth from a certain point. In smaller specimens this effect occurs prematurely hinder-
ing self-similar crack propagation from the beginning of the fracture process thus influencing decisively the obtained R-
curves. An alternative way to define a critical specimen size is the extent of the FPZ that should be approximately constant
during self-similar crack growth. Hence, from Figs. 10a–f the normalized extent of FPZ versus the equivalent crack length
demonstrates that a clear plateau is visible for the higher dimensions, i.e., for H P 140 mm. This is illustrated in Fig. 11 which
represents the normalised fracture energy as a function of the normalised equivalent crack length a = ae/H. As it can be seen
the value of a corresponding to the plateau onset increases with the decrease of the specimen size (D5 to D0). Consequently,

1.2

1.0
D5 D4 D3 D2 D1 D0

0.8
G I / GIc (inp)

0.6

0.4

0.2

0.0
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85

α = a e /H

Fig. 11. Superposition of R-curves (Table 1).

2400 1.4
E T = 401 MPa
2000 1.2
E T = 334 MPa β = 1.07
Load, P (N)

1.0
G I/G Ic (inp)

1600 E T = 267 MPa

0.8 E T = 267 MPa
1200 E T = 334 MPa
0.6
800 E T = 401 MPa
0.4
400 0.2
0 0.0
0.0 1.0 2.0 3.0 4.0 280 286 292 298 304 310 316
Displacement, δ (mm) a e (mm)
(a) (b)
Fig. 12. Effect of the elastic property ET on the parameter b. (a) Load–displacement curves and (b) corresponding R-curves.
 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520 519

2.5 1.4
ft 1.2
2.0 β = 1.07
Stress f (MPa)
1.0

G I/G Ic(inp)
1.5 +20% Reference
0.8
Reference
1.0 0.6
-20% Reference
0.4
0.5 fb
wc 0.2
wb
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 280 286 292 298 304 310 316
Crack opening, w (mm) a e (mm)

(a) (b)
Fig. 13. Effect of the cohesive parameters on the R-curve. (a) Stress-softening diagram and (b) corresponding R-curves.

when the plateau is attained the remaining resistant section becomes tiny for smaller specimen sizes, thus leading to spu-
rious effect of the compressive stresses induced by bending above the neutral axis, which hinders the self-similar propaga-
tion. It should be noted, that the conclusions in regards to the adequate dimensions for the present case might not be applied
if a different material is analysed. In fact, since the fracture process zone (FPZ) extent may change drastically with the com-
bination of cohesive parameters (namely ft and GIc), it is expectable that an higher FPZ requires an higher specimen size
(H > 70 mm), from which the self-similar crack growth is verified. It should be noted that the FPZ confinement is responsible
for size effect phenomenon of quasi-brittle materials, i.e., the effect of the structure size on its nominal strength and on its
resistance to crack growth at the peak load [15,16]. These achievements confirm the results obtained by Stanzl-Tschegg et al.
[2] who observed that no size effect is present in spruce wood from certain specimen size using the WS test (ligament length
and specimen width higher than 70 mm and 30 mm, respectively). These authors atributed this effect to the microprocess
zone that does not form fully in the smaller specimens, thus affecting the measured fracture energy. In order to verify the
eventual influence of the material properties on the parameter b, a sensitive analysis was performed. Hence, the elastic mod-
ulus ET of the specimen central part (Fig. 3) and the cohesive parameters (Fig. 5) were varied ±20% (typical scattering in wood
properties) from the reference values used in this study (Tables 2 and 3). As it can be observed in Figs. 12 no significant dif-
ference on the plateau values of the R-curves was detected using the same parameter b. In regards to the cohesive properties
variation (Fig. 13) no influence was detected as in fact the three R-curves are coincident. This means that b does not depend
on the material properties in the considered range.

5. Conclusions

A new data reduction scheme for the single edge notched-three point bending test applied to mode I wood fracture char-
acterization was presented. The method is based on the beam theory and crack equivalent concept and takes into account
the presence of a stress relief region of triangular shape. The method does not require the crack length monitoring during its
growth which is very difficult to perform in this test and in this material. A bilinear cohesive damage model was used to
simulate damage initiation and progression considering six different specimen sizes. The goal was to verify the capacity
of the data reduction scheme to replicate the inputted fracture energy, thus validating the procedure. The proposed method
only requires the load–displacement data to obtain an R-curve. It was verified that the inputted value of fracture energy is
well captured in the plateau region of the R-curves except for the smaller specimens where a plateau region practically does
not exist. For the remaining ones a constant value for the parameter defining the size of the stress relief region was found and
it was confirmed to be consistent with previous results of other authors. It was also verified that this parameter is not sen-
sible to material properties.
Finally, it should be noted that the presented methodology is not exclusive for wood and that it can be easily applied to
mode I fracture characterization of other different quasi-materials using this test.

Acknowledgement

The authors would like to thank the Portuguese Foundation for Science and Technology for supporting the work here pre-
sented, through the research project PDTC/EME-PME/64839/2006.

References

[1] European Committee for Standardization. EN 1995-1-1. Design of timber structures. Part general rules and rules for buildings, Brussels; 2004.
[2] Stanzl-Tschegg SE, Tan DM, Tschegg EK. New splitting method for wood fracture characterization. Wood Sci Technol 1995;29:31–50.
[3] Morel S, Dourado N, Valentin G, Morais J. Wood: a quasibrittle material – R-curve behavior and peak load evaluation. Int J Fracture 2005;131:385–400.
520 M.F.S.F. de Moura et al. / Engineering Fracture Mechanics 77 (2010) 510–520

[4] Dourado N, Morel S, de Moura MFSF, Valentin G, Morais J. Comparison of fracture properties of two wood species through cohesive crack simulations.
Composites Part A 2008;39:415–27.
[5] de Moura MFSF, Morais J, Dourado N. A new data reduction scheme for mode I wood fracture characterization using the double cantilever beam test.
Engng Fract Mech 2008;75:3852–65.
[6] Yoshihara H, Kawamura T. Mode I fracture toughness estimation of wood by DCB test. Composites Part A 2006;37:2105–13.
[7] Boström L. The stress–displacement relation of wood perpendicular to the grain. Wood Sci Technol 1994;28:319–27.
[8] Gustafsson PJ. A study of strength of notched beams. Paper 21-10-1. In: Proceedings of CIB-W18A meeting, Parksville, Canada; 1988.
[9] Daudeville L. Fracture in spruce: experiment and numerical analysis by linear and non linear fracture mechanics. Holz Roh Werkst 1999;57:425–32.
[10] Dourado N. R-Curve behaviour and size effect of a quasibrittle material: wood. PhD thesis. University of Trás-os-Montes Alto Douro and University of
Bordeaux I (Under Co-tutorship), Vila Real, Portugal; 2008.
[11] Petersson PE. 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; 1981.
[12] Guitard D. Mécanique du matériau bois et composites. Cepadues-editions. ISBN 2.85428.152.7:108-23; 1987.
[13] Kienzler R, Herrmann G. An elementary theory of defective beams. Acta Mech 1986;62:37–46.
[14] de Moura MFSF, Silva MAL, Morais J, de Morais AB, Lousada JJL. Data reduction scheme for measuring GIIc of wood in end-notched flexure (ENF) tests.
Holzforschung 2009;63:99–106.
[15] Bazant ZP. Scaling theory for quasibrittle structural failure. Proc Natl Acad Sci 2007;101:13400–7.
[16] Morel S. R-curve and size effect in quasibrittle fractures: case of notched structures. Int J Solids Struct 2007;44:4272–90.