Mechanical Behavior and Modeling of Dowelled

Steel-to-Timber Moment-Resisting Connections

Bo-Han Xu 1; Abdelhamid Bouchaïr 2; and Patrick Racher 3

Abstract: This paper investigates the mechanical behavior of dowelled steel-to-timber moment-resisting connections. Two configurations of
joints were tested, and the experimental results were used to validate a three-dimensional (3D) nonlinear finite element method (FEM) model.
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

The model uses the Hill criterion to manage the plastic yielding of timber. The brittle character of timber in shear and tension perpendicular to
grain is managed through the damage evolution based on the Hoffman failure criterion. The numerical model represents with good accuracy
the behavior observed during experiments. The analytical model based on the theory of plastic limit analysis was evaluated through the
application to the tested connections. DOI: 10.1061/(ASCE)ST.1943-541X.0001119. © 2014 American Society of Civil Engineers.
Author keywords: Moment-resisting connections; Steel-to-timber joints; Finite element; Load distribution; Analytical approach; Wood
structures.

Introduction grain (Sawata and Yasumura 2003; Gattesco and Toffolo 2004).
Because of large interest in applications of moment-resisting con-
Because of the environmental benefits, timber is widely used in the nections, several studies were reported on timber frame moment
construction sector. Through the application of structural wood joints (Lam et al. 2008). However, the number of experimental
based products such as glued-laminated timber (glulam), laminated studies on dowelled steel-to-timber moment-resisting connections
veneer lumber (LVL), parallel strand lumber (PSL), and laminated available in the literature is rather limited.
strand lumber (LSL), the spans and load-carrying capabilities of To study the effect of the connection geometry, the experimental
structural members have increased to a level whereby they are eco-
way would be very time-consuming. For the effective and economi-
nomically competitive with other materials traditionally associated
cal reason, it is generally combined with the numerical approach. In
with large modern structures. Greater attention must be paid to con-
the last decade, some numerical research based on two-dimensional
nections, because of the complexity of the internal forces transfer
(2D) FEM models were performed (Chui and Li 2005; Bouchaïr
between the components (Bainbridge and Mettem 1998). Moment-
resisting connections are commonly encountered in timber struc- et al. 2007). However, it has been recognized that these 2D FEM
tures for transferring the internal forces between the structural models only give reasonable predictions for very specific situations
members (beams and columns). Dowelled joints are commonly such as very thin or very thick timber members. Typically, dowel-
used as structural timber moment-resisting connections. To prevent type connections are three-dimensional (3D) problems because of
splitting and to enhance the ductility, the timber-to-timber joints are the nonuniform stress distribution across the thickness of the mem-
usually reinforced in the vicinity of the fasteners, using plywood bers combining the nonlinear behavior of dowel in bending and the
or densified veneer wood plates glued onto the interfaces of the timber embedding.
jointed members (Guan and Rodd 2001; Bouchaïr et al. 2007). In this study, a 3D FEM model including the multidimensional
Complying with the actual architectural trends and high strength failure criterion is proposed to predict the moment-carrying capac-
requirements, timber-to-timber joints are mainly replaced by steel- ity of dowelled steel-to-timber joint in bending. The model is de-
to-timber joints. veloped considering the elastoplastic behavior of steel and timber
At present, the majority of the performed studies have focused with the damage evolution of timber. The nonlinear material model
on the steel-to-timber joints in tension parallel or perpendicular to has been successfully applied to simulate the behavior of timber
connections with glued-in rods in bending (Xu et al. 2012).
1
Associate Professor, State Key Laboratory of Coastal and Offshore Besides, the contact between the connection components is mod-
Engineering, Dalian Univ. of Technology, No. 2 Linggong Rd., Ganjingzi, eled using contact element models with friction.
Dalian 116024, People’s Republic of China (corresponding author). The experimental results are reported to observe the real behav-
E-mail: [email protected] ior of the connection and to validate the numerical model. In
2
Professor, Clermont Université, Université Blaise Pascal, Institut
Pascal, BP 10448, F-63000 Clermont-Ferrand, France; and CNRS, UMR
addition, the moment-carrying capacity of the connections is ana-
6602, Institut Pascal, F-63171 Aubière, France. E-mail: abdelhamid lytically calculated on the basis of single fastener resistance that is
[email protected] defined by the well-known yield theory (Johansen 1949; Larsen
3
Associate Professor, Clermont Université, Université Blaise Pascal, 1973; STEP 1 1995; Moss 1997). The 3D numerical model is used
Institut Pascal, BP 10448, F-63000 Clermont-Ferrand, France; and CNRS, to investigate the interaction effect among the dowels and to check
UMR 6602, Institut Pascal, F-63171 Aubière, France. E-mail: patrick the validity of the analytical approach that can be applied with vari-
[email protected] ous standard codes defining the fastener resistance. The validated
Note. This manuscript was submitted on July 17, 2013; approved on
May 23, 2014; published online on August 7, 2014. Discussion period open
numerical model is also used to perform a comprehensive analysis
until January 7, 2015; separate discussions must be submitted for individual of the main parameters characterizing the mechanical behavior of
papers. This paper is part of the Journal of Structural Engineering, the connections, such as the stress distribution and the position
© ASCE, ISSN 0733-9445/04014165(11)/$25.00. of the failure zones.

© ASCE 04014165-1 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 Table 1. Measured Steel Properties (MPa) rotation of timber member measured by digital cameras is less, the
Steel components Property Value relative rotation (θ) between the connected members can be approx-
imately calculated by the measured displacements at two positions
Steel plate fy 309
(u1 and u2 ) on the steel plate (Fig. 1). In this case, tanðθÞ ¼
fu 434
Steel dowel fy 366
ðu2 − u1 Þ=208.
fu 543 The experiments were conducted in accordance with the pro-
cedure defined in the European Norm EN 26891 (CEN 1993),
achieved with a constant displacement rate of 4 mm= min. Because
Experimental Program of unavoidable small construction tolerances, a loading-unloading
cycle at the initial stage was performed to obtain the initial stiffness.
Two configurations of steel-to-timber joints were tested in bending Then, the joints were loaded until 40% of the maximum estimated
at Blaise Pascal University (Clermont-Ferrand, France; Xu et al. load (Fest ), and were unloaded until 0.1Fest in order to determine the
2008). Each specimen consists of two lateral glulam members elastic stiffness. Another loading-unloading cycle was carried out to
(2,000 × 486 × 75 mm) connected to a main steel plate (1,200 × present the evolution of stiffness at the nonlinear behavior stage.
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

260 × 8 mm) using six 16 mm dowels. The glulam members were For configuration A, the joints showed a ductile failure with
manufactured with Scandinavian spruce and MUF adhesives by plastic deformation of the dowels in bending and the timber in em-
commercial manufacturer James. The measured timber density bedding. For configuration B, after a significant deformation of
ρmean is equal to 438 kg=m3 , and the moisture content hmean is the dowels in bending, the splitting of timber member appeared
equal to 10%. The mechanical characteristics of steel plate and in the joint area. The crack propagated along the wood grain in
dowel, such as the yield strength (fy ) and the ultimate strength the two side members near the dowels, which resulted in a decrease
(fu ), were determined from tensile tests (Table 1). of the applied load representing the beginning of the joint failure.
The configurations and geometry of the specimens with the Figs. 4 and 5 show the timber embedding on the outside of joint and
positions of the main forces and displacement measurements are on the side close to steel plate, and also the deformed dowels for
shown in Fig. 1. Although the circular, trapezoidal, and rectangular configurations A and B, respectively. Moreover, no significant
patterns are usually used, the rectangular pattern was chosen in this deformation of the dowel holes was observed in the steel plate
study because it promotes the combination of high shear and ten- for the two configurations of tested joints.
sion perpendicular to grain stresses (STEP 1 1995). Two configu- Fig. 6 shows the experimental moment-rotation curves of
rations of connections (A and B) were tested to obtain the failure two configurations A and B. Although only two specimens were
mode induced by dominant shear or tension perpendicular to grain tested for each configuration, the experimental curves of the same
configuration present a similar trend. The configuration A shows
stresses. The edge distances were extended to take into account the
the more ductile behavior, where there is no significant increase
fire resistance conditions for timber joints exposed to fire according
of load after 3° of rotation. The configuration B also shows a global
to the design rules of EN 1995-1-2 (CEN 2004b).
ductile behaviour; however, a crack occurs on the row of dowels.
The load was applied by a displacement control process on the
The tests were stopped when the load reaches a plateau with large
steel plate, and three supports were set on the glulam members.
displacement and rotation.
Fig. 2(a) shows the experimental set-up where several smooth tim-
For the structural analysis and design approach, the value of the
ber planks were placed between glulam members and steel brackets
yield capacity of the connection is defined as the intersection be-
for lateral stabilization of the specimen. The joint detail is illus-
tween the elastic stiffness and the tangent stiffness at the final part
trated in Fig. 2(b). The forces were measured by load cells at
of the moment-rotation curve (Fig. 6). The Table 2 summarizes the
the loading point and at the supports [Fig. 2(c)]. The LVDT dis-
yield and ultimate moment-carrying capacities of the connections.
placement transducers called cd and digital cameras were used
The joint A exhibits a lower capacity than the joint B.
to measure the displacements at different locations in the joint area
[Fig. 2(c)], where cd1 measured the relative displacement between
main steel plate and lateral glulam member, and cross-shaped star Analytical Approach
represents the location of data collected by digital cameras. To il-
lustrate the global behavior of the joint, the main load-displacement The analytical model used to predict the moment-carrying capacity
curves (F-u) of configuration A are shown in Fig. 3. Because the of the joint in bending is usually based on an elastic approach.

R1

115 128 128 115 115 256 115

1 2 3 a c LVDT F
a c 1 2 u1 u2 u3
1860

100
200 A 3 4 B θ
100 325 208
b d 5 6
4 5 6 b d 988

R1
R2

Fig. 1. Geometry of the steel-to-timber joints in bending and positions of the main measured forces and displacements

© ASCE 04014165-2 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 Load cell

cd1 cd2 cd3

cd4

Load cell
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

82 208

745
(a)

Load cell
(c)

(b)

Fig. 2. Experimental set-up: (a) the whole joint bending set-up; (b) detail of the joint zone after test; (c) positions of the load cells, displacement
transducers, and data collected by digital cameras

In this case, the members are assumed to be rigid because they are X
n

stiffer and stronger than the joint. Therefore, the joint rotation re- M¼ FM;i ri ð1Þ
sults from the rotational displacement ω of the fasteners (Fig. 7). i¼1

Defining the rotation center as the geometric center of the joint, the where FM;i = load on the fastener i, and ri = distance
equilibrium condition is given by between the fastener i and the rotation center of the joint. As the
fasteners behave linearly, the following relationships apply

FM;i ds;i F
ds;i ¼ and ω¼ ¼ M;i ð2Þ
K α;i ri K α;i ri
30
where K α;i = slip modulus in the force direction (αi þ π=2).
From Eqs. (1) and (2), the load on the fastener i is expressed as
F (kN)

20
follows:
K α;i ri
FM;i ¼ M ð3Þ
Kr
10
A1(u1) with the rotational stiffness K r
A1(u2)
A1(u3) X
n
0 Kr ¼ K α;i r2i ð4Þ
0 30 60 90 i¼1
u (mm)
Under the applied moment, fasteners are loaded at a varying
Fig. 3. Example of load-displacement curves (specimen A1)
angle to grain direction. Because of the orthotropic behavior of

© ASCE 04014165-3 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 Fig. 4. Failure modes of specimen A after test (timber in embedding and dowels in bending)
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Failure modes of specimen B after test (crack line, timber in embedding, and dowels in bending)

Table 2. Experimental Results of the Tested Connections

( θ y, My) Series Fu (kN) θy (°) M y (kN · m) Mu (kN · m) Mu =M y
Joint A 33.6 1.06 31.4 33.2 1.06
30
Joint B 38.8 1.03 34.2 38.3 1.12
Ratio A/B 0.87 1.03 0.92 0.87 —
M (kN⋅m)

20
FM,i

10
A (1)
ds,i
A (2) ri
0 Y
0 1 2 3 4 5 ω
(a) θ (°)
αi
O
( θ y, M y) X

30
Fig. 7. Geometric definition and force on fasteners
M (kN⋅m)

timber, each fastener load depends on the slip modulus in the

20 direction of the load. However, the effect of angle to grain on
slip modulus is not considered in current European code (CEN
2004a). For joints made with dowel-type fasteners, the slip
10 modulus kser per shear plane per fastener under load should be
taken as
B (1)
B (2)
0 kser ¼ 2ρ1.5
m d=23 ð5Þ
0 1 2 3 4
(b) θ (° ) As the slip modulus is assumed to be constant in all load direc-
tions, the load on the dowel i because of bending moment is calcu-
Fig. 6. Moment-rotation curves
lated using the following formula

© ASCE 04014165-4 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 r resistance (M y ) is defined as the yield moment resistance, when
FM;i ¼ Pn i 2
M ð6Þ
i¼1 ri the first dowel reaches its ultimate resistance. When all dowels
reach their ultimate resistances, and still considering the geometric
In addition, the shear and normal forces (V, N) are assumed to center as the rotation center of the connection, the moment resis-
be uniformly distributed on the dowels (FV;i ¼ V=n, FN;i ¼ N=n). tance (M u ) is defined as the ultimate moment resistance. For the
The resultant load applied on each dowel is calculated by summing two configurations of joints, the yield moment resistance (M y )
the vectors FM;i , FV;i , and FN;i . The resistance of each dowel de- under pure bending is higher than that under bending moment with
pends on the wood grain direction, thus it is calculated considering shear force; however, the shear force slightly influences the ulti-
the direction of the resultant applied load. According to the yield mate moment resistance (M u ).
theory (CEN 2004a), the resistance of a steel-to-timber joint with a In order to assess the values obtained using the analytical
single dowel in double shear is given by the Eq. (7). This approach approach, the comparison of experimental and analytical results
can be used whatever the direction of the applied load regarding the is shown in Table 4. M y and M u correspond to the yield and ulti-
direction of wood grain. The resistance Fv;R for each shear plan is mate moment resistance.
the weakest among the values associated to the three failure modes
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

(Fig. 8). The parameters influencing the failure modes are the em-
bedding strength of timber, the yield limit of steel, and the dowel Finite Element Model
slenderness. In general, a large plasticity of the joint can be pro-
vided when relatively slender fasteners are used. In that case, the The 3D FEM model developed to simulate the behavior of the steel-
more ductile failure modes 2 and 3 are governing. Mode 1 concerns to-timber joint in bending is built using the MSC.MARC software.
the timber joint with rigid fasteners (low slenderness) where the The nonlinear response of the connection is because of the bending
timber embedding is dominant. The application to the tested joints of dowels, the embedding of wood, and the contact evolution
shows that failure mode 2 is dominant. It is because of the combi- between wood and steel. In this study, steel is modeled using an
nation of dowel yield failure in bending with the timber failure in elastoplastic model based on von Misès yield criterion. Wood is
embedding simulated by a nonlinear anisotropic model based on Hill yield
criterion combined with Hoffman failure criterion. The nonlinear
8
> f h;α td mode 1 evolution of contact between steel plate, fasteners, and timber is
>
< hqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i also taken into account.
4M y;R
Fv;R ¼ min f h;α td 2 þ fh;α dt2 − 1 mode 2 ð7Þ The nonlinear wood model is developed on the basis of a con-
>
>
: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi stitutive law, which deals with the anisotropic behavior of wood
2.3 M y;R fh;α d mode 3
and the ductile characteristic or brittle failure depending on the
stress direction. In fact, wood exhibits ductile compressive behav-
with: M y;R ¼ 0.3f u d2.6 ior and brittle behavior in shear and tension perpendicular to grain.
f h;0 In this material model, the anisotropic elastoplastic constitutive law
fh;α ¼ ; k90 ¼ 1.35 þ 0.015d without strain-hardening according to Hill yield criterion is used
k90 sin2 α þ cos2 α to represent the timber compressive behavior. Brittle behavior in
fh;0 ¼ 0.082ð1 − 0.01dÞρ tension perpendicular to grain and shear is modeled using the
progressive failure analysis approach, which is based on Hoffman
Because of limited test data, the mean values of material proper- failure criterion representing the damage evolution in timber by a
ties are used to calculate the embedding strength instead of char- reduction of the elastic modulus.
acteristic design values. Table 3 shows the moment resistance of
the tested joints according to the analytical approach, considering
Meshing and Boundary Conditions
the bending moment with or without shear force. The moment
Considering the symmetry, only a half of the joint geometry is mod-
eled. The meshing based on 8-noded hexahedral elements, is shown
2F 2F 2F in Fig. 9. The loads are introduced by increasing controlled dis-
placements applied on nodes of steel member with the conditions
of supports reproducing the real conditions of the experimental
set-up.

Material Properties
F F F F F F
Steel is considered as an isotropic (E ¼ 210,000 MPa, ν ¼ 0.3)
and elastoplastic material. The stress-strain curve used in the
Mode 1 Mode 2 Mode 3 numerical model is a nonlinear one obtained from tests with
Fig. 8. Plastic failure modes of steel-to-timber joints
Table 4. Comparison of the Analytical and Experimental Values of
Moment Resistance
Analytical
Table 3. Analytical Values of Moment Resistance (kN·m)
(kN · m) Test (kN · m) My Mu
M MþV
Moment Analytical/ Analytical/
resistance Joint A Joint B Joint A Joint B Series My Mu My Mu test test
My 25.0 27.6 23.3 25.4 Joint A 23.3 26.2 31.4 33.2 0.74 0.79
Mu 26.4 30.8 26.2 30.7 Joint B 25.4 30.7 34.2 38.3 0.74 0.80

© ASCE 04014165-5 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 where dλ = plastic multiplier. In this study, perfect plasticity is con-
sidered for timber without strain-hardening.
However, the Hill criterion does not take into account the differ-
ence of strength between tension and compression. In order to con-
sider this dissymmetry of resistance, Hoffman failure criterion built
into the MSC.MARC is used, which is an extension of the Hill cri-
terion. In association with the unequal ultimate stresses in tension
and compression, the progressive failure is simulated through a re-
duction of the elastic modulus in both parallel and perpendicular to
grain directions, to represent the damage evolution in timber. When
the criterion is reached at an integration point, the elastic modulus
in the direction parallel to grain E0 is set equal to the modulus in the
direction perpendicular to grain E90 , and the modulus E90 is re-
duced to 10% of its initial value. This hypothesis is a standard
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

method for orthotropic materials proposed by MSC.MARC soft-

ware (MSC 2007), which has been successfully applied to predict
the behavior of other joint configurations such as timber joints with
glued-in rods (Xu et al. 2012) and traditional timber mortise-tenon
joints (Xu et al. 2013). The Hoffman criterion is associated with the
Hill criterion to manage the plastic yielding combined with the
Fig. 9. Meshing of the tested specimen: (a) tested specimen (load and damage evolution. It can be expressed as follows:
supports); (b) dowel zone
C1 ðσy − σz Þ2 þ C2 ðσz − σx Þ2 þ C3 ðσx − σy Þ2 þ C4 σx þ C5 σy
þ C6 σz þ C7 τ 2yz þ C8 τ 2xz þ C9 τ 2xy ¼ 1
1 1 1
some simplifications (piece-wise curve). Timber is an ortho- C1 ¼ − ; C2 ¼ C3 ¼ ;
tropic material with identical properties in the radial (R) and f t;90 fc;90 2ft;0 f c;0 2ft;0 fc;0
tangential directions (T). This combined direction is referred to as 1 1 1 1
C4 ¼ − ; C5 ¼ C6 ¼ − ;
perpendicular to grain (⊥) while the longitudinal direction (L) is f t;0 fc;0 f t;90 f c;90
referred to as parallel to grain (//). The glulam used for steel-to- 1
timber joint corresponds to resistance class GL24 h according to C7 ¼ C8 ¼ C9 ¼ 2 ð10Þ
EN 1194 (CEN 1999), which is equivalent to the class 24 F-1.8 E fv
according to AITC 117 (AITC 117 2004). The mean values of In this study, the mean values of the compressive strength are
modulus of elasticity (CEN 1999) and Poisson’s ratios (Guitard determined by the test according to EN 408 (CEN 2004c): f c;0 ¼
1987) are given in Table 5. 39 MPa and f c;90 ¼ 3.24 MPa. Ten specimens were cut from the
Because of the anisotropic plastic behavior of timber, the Hill’s timber member in the tested timber joints. It should be noted that the
criterion built into the MSC.MARC is chosen. The Hill criterion is a variability of wood can be high. On the basis of statistical analysis
generalized version of the von Misès yield criterion to consider the from test results in the literature, the coefficient of variation for ten-
anisotropy of the materials. Its stress potential can be expressed as sile strength ranges from 10 to 40%, and the one for shear strength
follows: ranges from 10 to 20% (Xu 2009). In this model, the low coefficient
σ̄ ¼ ½a1 ðσy − σz Þ2 þ a2 ðσz − σx Þ2 þ a3 ðσx − σy Þ2 of variation equal to 10% is adopted, which corresponds to the low
pffiffiffi bound strengths in tension and shear. As a normal distribution is
þ 3a4 τ 2zx þ 3a5 τ 2yz þ 3a6 τ 2xy
1=2 = 2 adopted, the mean values of the timber strengths in tension and
2 1 1 2 shear can be derived from the characteristic values (CEN 1999) us-
a1 ¼ − ; a2 ¼ a3 ¼ ; a4 ¼ a5 ¼ a6 ¼ ing a multiplier coefficient equal to 1.2. In this case, mean values
f 2c;90 f 2c;0 f 2c;0 3f 2v
considered for the timber strengths in tension and shear are: f t;0 ¼
ð8Þ 19.8 MPa, f t;90 ¼ 0.48 MPa, and f v ¼ 3.24 MPa.
Moreover, the size effects on the brittle failure of timber are not
The hypothesis of associated plasticity is considered. Thus the taken into account in this model. Although the existence of size
Hill criterion is used as a plastic flow law. The relationship between effects is well known, no conclusive evidence has yet arisen con-
the plastic strain increment and the stress increment is given by cerning the accuracy of probabilistic strength theories to describe
the size in the strength of timber (Tannert et al. 2010). This paper
∂ σ̄
dεp ¼ dλ ð9Þ presents a simple material model using common timber property
∂σ parameters, which will allow the engineer to develop an FEM
model for the prediction of timber joint behavior.

Table 5. Timber Properties Used in the Model Contact Description

Property Value
The contact between the connection components is usually mod-
E0 (MPa) 11,600 eled using deformation contact elements. Contact occurs in the in-
E90 (MPa) 390 terface between the timber member and the steel plate, the dowel
G (MPa) 750 and the timber member, the dowel and the steel plate, and finally
ν TR ¼ ν LT 0.41 the stiff steel plates at the supports and the timber member. Contact
ν RL 0.02
is modeled using the direct constraint method in MSC.MARC.

© ASCE 04014165-6 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 The method requires the definition of the contact body that poten-
tially may come in contact with the other. Contact bodies can
simply be the physical bodies themselves (e.g., timber, dowels,
and steel plate). In this method, when a node of a body contacts 30
another body, a multipoint constrain is imposed. This constraint

M (kN⋅m)
allows the contacting node to slide on the contacted segment.
In this way, a contacting node is forced to be on the contacted 20
segment. During the iteration procedure, a node can slide from
one segment to another, changing the retained nodes associated
with the constraint. A node is considered sliding off a contacted
10 A (1)
segment if it passes the end of the segment over a distance more
than the contact tolerance. A (2)
Contact may be developed with friction based on the Coulomb FEM
criterion. The method allows no movement until the friction force is 0
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

0 1 2 3 4 5
reached. After that, the movement initiates and the friction forces
remain constant. The friction coefficient between dowels and tim- (a) θ (°)
ber is set equal to 0.3, which can be considered as the limit value
between the smooth and rough surface dowel (Sjodin et al. 2008).
In the case of no consideration of friction, the predicted elastic
stiffness is the same with one in the case of friction; however,
30
the predicted moment-carrying capacity in the nonlinear stage is
significantly less one in the case of friction. The friction coefficient

M (kN⋅m)
between dowels and the main steel plate is set equal to 0.001, be-
cause of the thin steel plate mobilizing a small contact zone with 20
dowels and negligible influence of the relative displacement be-
tween them. The comparison considering the different values of
the friction coefficient (e.g., 0.1 and 0.3) shows no significant 10 B (1)
influence on numerical results. The stiff steel plates used on the
B (2)
supports in the tests are modeled with a control of the contact
conditions between the stiff steel plates and timber. The friction FEM
coefficient between the stiff steel plates and timber is set equal 0
0 1 2 3 4
to 0.2 (Xu et al. 2009). In fact, contact in the normal direction (b) θ (°)
between timber and stiff steel plates on the supports, is more im-
portant than their friction (tangential direction). Several simulations Fig. 10. Experimental and numerical moment-rotation curves
with friction coefficients varying from 0.1 to 0.5 show no signifi-
cant difference of numerical curves.

Table 6. Comparison of Results (Test and FEM Model)

Validation and Application of the Numerical Model Rotational stiffness Yield moment
(kN · m=°) resistance (kN · m)
The 3D FEM model is validated by comparison of the numerical
and experimental moment-rotation curves. The validated model is Series Test Model Test Model
used to present the load distribution among the dowels and to check Joint A 43.6 40.0 31.4 29.5
the validity of the analytical approach. Moreover, the stress distri- Joint B 43.6 43.7 34.2 32.0
bution and the failure index according to Hoffman criterion are
observed to understand the evolution of the failure zones.
resistance of joint B characterized at the ultimate phase by a brittle
failure mode.
Global Moment-Rotation Behavior
The comparison between the numerical and experimental moment-
Load Distribution among the Dowels
rotation curves shows that the numerical model represents well the
experimental curve (Fig. 10). However, at the initial phase of load- The load distribution among the dowels obtained from the numeri-
ing, the stiffness of the numerical model is higher than the exper- cal model is calculated for the two configurations of joints tested,
imental one. This can be explained by the fact that the numerical with an applied force equal to 23.6 kN, which induces the moment
model does not take into account the 1 mm hole clearance in the corresponding to the lowest yield moment (23.3 kN · m) defined
steel plate and the initial defects of the joint induced by production by the analytical model. The resultant force on each dowel is
tolerances. The numerical initial stiffness and experimental elastic numerically calculated considering the resultant contact force be-
stiffness taken from the unloading-reloading phases are quite sim- tween the dowel and the hole of the steel plate. A nonuniform dis-
ilar. The numerical model is in good agreement with experimental tribution of the forces is observed among the dowels depending on
observation regarding the main parameters such as the stiffness and their positions in the joint (Table 7). The values of load for each
the yield resistance (M y ) of the joint (Table 6). Because of high dowel are similar between the numerical and analytical models.
variability of timber properties, the low bound strengths in tension The dowel numbering is shown in Fig. 1.
and shear that govern the brittle failure mode are adopted in the Fig. 11 shows the directions of load for each dowel that are
model. Thus, the model may underestimate the ultimate moment given by the numerical and analytical models with an applied load

© ASCE 04014165-7 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 Table 7. Load Distribution among the Dowels in the Joint Table 8. Load Distribution among Dowels for Joint in Pure Bending
Series Load 1 2 3 4 5 6 Series Load 1 2 3 4 5 6
Joint A FEM (kN) 27.8 18.5 33.2 28.7 19.1 33.1 Joint A FEM (kN) 31.8 18.4 32.9 32.8 19.0 33.1
Analytical (kN) 27.1 19.0 33.3 27.1 19.0 33.3 Analytical (kN) 32.4 19.9 32.4 32.4 19.9 32.4
Analytical/FEM 0.97 1.03 1.00 0.94 0.99 1.01 Analytical/FEM 1.02 1.08 0.98 0.99 1.05 0.98
Joint B FEM (kN) 21.9 28.8 21.9 28.6 22.5 29.2 Joint B FEM (kN) 26.7 25.8 26.6 25.0 26.9 25.6
Analytical (kN) 24.4 30.5 17.6 25.5 24.4 30.5 Analytical (kN) 29.4 29.4 23.1 23.1 29.4 29.4
Analytical/FEM 1.11 1.06 0.80 0.89 1.08 1.04 Analytical/FEM 1.10 1.14 0.87 0.92 1.09 1.15

150 2.0
τ
120
σ 90
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

1.5
90
Dowels vertical positions (mm)

Stress (MPa)
60 1.0

30
0.5
0
-150 -120 -90 -60 -30 0 30 60 90 120 150
-30 0.0
0 50 100 150 200
-60
-0.5
-90 Distance on line a-b (mm)
2.0
-120
FEM τ
Analytical
-150
1.5 σ 904
Dowels horizontal positions (mm)
Stress (MPa)

150
1.0
120
0.5
90
Dowels vertical positions (mm)

60 0.0
0 50 100 150 200
30
-0.5
0 Distance on line c-d (mm)
-150 -120 -90 -60 -30 0 30 60 90 120 150
-30 Fig. 12. Stress distribution along the dowel line for joint A
-60

-90 joint B. This result is in a favorable agreement with that from

the analytical model. It implies that in an elastic analysis, the hy-
-120
FEM pothesis of the center of rotation coinciding with the geometric
Analytical center of the connection is acceptable.
-150
Dowels horizontal positions (mm)

Fig. 11. Load vector for each dowel in the joints A and B Stress Analysis
During the tests of joint B, the crack parallel to grain appeared and
propagated in the timber member along the row of dowels. The
crack is mainly because of the shear stresses along the row of dow-
equal to 23.6 kN. The average angular difference between the els combined to tensile stresses in the direction perpendicular to
directions given by the numerical model and the analytical model grain induced by the transfer of the bending moment. Figs. 12–13
is about 12° for the joint A and 9° for the joint B. show the stress distribution on the main row of dowel defined in
In addition, the numerical model is also used to estimate the load Fig. 1. The maximum shear stress in the joint B is higher than that
distribution among the dowels considering the pure bending load- in the joint A; however, they are lower than the shear strength of
ing case. Table 8 shows the values of load for each dowel calculated timber. The maximum values of tensile stresses in the direction
using the numerical and the analytical models with an applied perpendicular to grain, for the two configurations of joints, are very
bending moment equal to 25 kN · m, which corresponds to the close and higher than the strength of timber. For the analysed con-
lowest yield limit defined by the analytical model. The average figurations of joints, the interaction between shear and tension
angular difference between the directions given by the numerical perpendicular to grain characterize the failure with a higher contri-
and analytical models is about 11° for the joint A and 9° for the bution of tensile stress perpendicular to grain.

© ASCE 04014165-8 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 3.0
τ
2.5 σ 90
2.0
Stress (MPa)

1.5

1.0

0.5

0.0
0 50 100 150 200
-0.5
Distance on line a-b (mm)
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

2.5
τ
2.0
σ 90
1.5
Stress (MPa)

1.0

0.5

0.0
0 50 100 150 200
-0.5

-1.0

-1.5
Distance on line c-d (mm)

Fig. 13. Stress distribution along the dowel line for joint B Fig. 15. Contour plot of the Hoffman failure index in the section of
(a) a-b; (b) c-d for joint B

The failure index of Hoffman criterion based on the stresses’

interaction provided by the numerical model is used to evaluate the appearance of crack. For joint B, the limit of the failure criterion
the position of the first potential failure zones. Figs. 14 and 15 is reached at the zone between two dowel holes. This is consistent
show the zones where the limit of the Hoffman failure criterion with the experimental observation of the crack in joint B.
is reached at 23.6 kN. It indicates the potential failure zones before
Conclusion

Experimental tests were carried out to observe the behavior of a

dowelled steel-to-timber joint in bending considering two configu-
rations of dowel distribution. A 3D numerical model based on FEM
has been developed to understand the mechanical behavior of the
joint. The model uses the plastic flow law based on the Hill cri-
terion associated to the Hoffman failure criterion representing the
evolution of damage in timber by a reduction of the elastic modu-
lus. Although the number of specimens is limited, the experimental
results can partly validate the numerical model. Comparison with
experimental results shows that the proposed model is promising. It
is capable of simulating the behavior of timber joints such as the
global moment-rotation response, the bending moment-carrying
capacity, and the failure mode. Besides, the numerical results show
that the analytical approach used in design represents correctly
the load distribution among the dowels. This implies that the geo-
metric center of the dowels group can be used as a mechanical ro-
tation center of the joint. The analytical approach, based on the
assumption of elastic distribution of internal forces in the connec-
tion, can be applied with various standard codes defining the fas-
tener resistance. Furthermore, the comparison shows that the yield
theory is a good tool to predict the resistance of dowelled joints in
bending. The numerical model with Hoffman criterion gives the
possibility to follow and investigate the evolution of the failure
zones in the dowelled timber connections with various materials
Fig. 14. Contour plot of the Hoffman failure index in the section of
and geometric configurations. This approach can be profitably
(a) a-b; (b) c-d for joint A
used to limit the expensive experiments. It allows performing a

© ASCE 04014165-9 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 sensitivity study of the input parameters in the FEM model with ρ = density;
application to various types of connections. Then, the model will ρmean = mean density;
be applied to timber joint configurations varying the dimension of σi , τ ij = stresses in timber;
the timber members, the number of fasteners, and the load ratio ν = Poisson’s ratio;
between bending moment and shear with normal forces. The ν TR , ν LT , ν RL = TR, LT, and RL Poisson’s ratio; and
aim is to develop a more general analytical approach to predict ω = rotational displacement of the fasteners.
the behavior of timber joints including the brittle failure.

Acknowledgments References
The authors gratefully acknowledge the support of the National American Institute of Timber Construction (AITC). (2004). “Standard
Natural Science Foundation of China (Grant No. 51108055), Crea- specifications for structural glued laminated timber of softwood spe-
tive Research Groups of the National Natural Science Foundation cies.” AITC 117, Englewood, CO.
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

of China (Grant No. 51121005), Research Projects of Institut Bainbridge, R. J., and Mettem, C. J. (1998). “A review of moment-resistant
structural timber connections.” Proc., Inst. Civ. Eng. Struct. Build.,
Pascal, Université Blaise Pascal, and the Chinese-French coopera-
128(4), 323–331.
tion program Cai Yuanpei 2012 – 27938UM. Bouchaïr, A., Racher, P., and Bocquet, J. F. (2007). “Analysis of dowelled
timber to timber moment-resisting joints.” Mater. Struct., 40(10),
1127–1141.
Notation Chui, Y. H., and Li, Y. (2005). “Modeling timber moment connection under
reversed cyclic loading.” J. Struct. Eng., 10.1061/(ASCE)0733-9445
The following symbols are used in this paper: (2005)131:11(1757), 1757–1763.
d = fastener diameter; European Committee for Standardization (CEN). (1993). “Timber
ds;i = fastener displacement; structures—Test methods—Determination of embedding strength and
dεp = plastic strain increment; foundation values for dowel type fasteners.” EN 26891, Brussels,
dλ = plastic multiplier; Belgium.
E = elastic modulus of steel; European Committee for Standardization (CEN). (1999). “Timber
E0 , E90 = parallel and perpendicular to grain elastic structures—Glued laminated timber—Strength classes and determina-
modulus; tion of characteristic values.” EN 1194, Brussels, Belgium.
European Committee for Standardization (CEN). (2004a). “Eurocode 5—
F = load;
Design of timber structures. Part 1-1: General—Common rules and
Fest = estimated load; rules for buildings.” EN 1995-1-1, Brussels, Belgium.
FM;i = load on the dowel i due to bending moment; European Committee for Standardization (CEN). (2004b). “Eurocode 5—
FN;i = load on the dowel i due to normal force; Design of timber structures. Part 1-2: General—Structural fire design.”
Fu = ultimate resistance of the joint; EN 1995-1-2, Brussels, Belgium.
FV;i = load on the dowel i due to shear force; European Committee for Standardization (CEN). (2004c). “Timber
FV;R = load-carrying capacity per shear plane per fastener; structures—Structural timber and glued laminated timber—Determination
f c;0 , f c;90 = compressive strength parallel and perpendicular to of some physical and mechanical properties.” EN 408, Brussels,
grain; Belgium.
f h;α = embedding strength at an angle α to the grain; Gattesco, N., and Toffolo, I. (2004). “Experimental study on multiple-bolt
steel-to-timber tension joints.” Mater. Struct., 37(2), 129–138.
fh;0 = embedding strength parallel to grain;
Guan, Z. W., and Rodd, P. D. (2001). “Hollow steel dowels—A new
ft;0 , ft;90 = tensile strength parallel and perpendicular to grain; application in semi-rigid timber connections.” Eng. Struct., 23(1),
f u = ultimate strength of steel; tensile strength of 110–119.
fastener; Guitard, D. (1987). Mechanics of wood and composite materials, Cepadues
fv = shear strength of timber; édition, Toulouse, France, 238 (in French).
fy = yield strength of steel; Johansen, K. W. (1949). Theory of timber connections, international asso-
G = shear modulus; ciation for bridge and structural engineering, Vol. 9, Bern, Switzerland,
hmean = mean moisture content; 249–262.
i = dowel numbering; Lam, F., Schulte-Wrede, M., Yao, C. C., and Gu, J. J. (2008). “Moment
K α;i = slip modulus in the force direction (αi þ π=2); resistance of bolted timber connections with perpendicular to grain rein-
forcements.” Proc., 10th World Conf. on Timber Engineering (WCTE
K r = rotational stiffness;
2008) (CD-ROM), Miyazaki, Japan.
kser = slip modulus; Larsen, H. (1973). “The yield load of bolted and nailed joints.” IUFRO-5
M = bending moment; Conf., Int. Union of Forest Research Organizations, London, U.K.,
M y = yield moment resistance of the joint; 646–654.
M y;R = yield moment of fastener; Moss, P. J. (1997). “Multiple-bolted joints in wood members—A literature
M u = ultimate moment resistance of the joint; review.” General Technical Rep. FPL-GTR-97, Forest Products
N = normal force; Laboratory, Madison.
n = amount of dowel; MSC.MARC. (2007). “Theory and user information.” User’s manual,
ri = distance between the dowel i and the rotation Vol. A, MSC Software Corporation, Newport Beach, CA.
center of the joint; Sawata, K., and Yasumura, M. (2003). “Estimation of yield and ultimate
strengths of bolted timber joints by nonlinear analysis and yield theory.”
t = timber member thickness;
J. Wood Sci., 49(5), 383–391.
u = displacement; Sjodin, J., Serrano, E., and Enquist, B. (2008). “An experimental and
V = shear force; numerical study of the effect of friction in single dowel joints.” Eur.
α = angle between a force and the direction of grain; J. Wood Wood Prod., 66(5), 363–372.
θ = rotation; STEP 1. (1995). Timber engineering, basis design, material properties,
θy = rotation corresponding to yield moment; structural components and joints, 1st Ed., Centrum Hout, Netherlands.

© ASCE 04014165-10 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165

 Tannert, T., Lam, F., and Vallée, T. (2010). “Strength prediction for rounded Xu, B. H., Bouchaïr, A., and Taazount, M. (2008). “Assemblages bois-
dovetail connections considering size effects.” J. Eng. Mech., 10.1061/ metal sollicites en flexion.” Université Blaise Pascal, Clermont-Ferrand,
(ASCE)0733-9399(2010)136:3(358), 358–366. France (in French).
Xu, B. H. (2009). “Numerical modelling of the behaviour of timber con- Xu, B. H., Bouchaïr, A., Taazount, M., and Racher, P. (2013). “Numerical
nections considering the damage evolution.” Ph.D. thesis, Université 3D finite element modelling and experimental tests of rounded dovetail
Blaise Pascal, Clermont-Ferrand, France (in French). connection.” Eur. J. Environ. Civ. Eng., 17(7), 564–578.
Xu, B. H., Bouchaïr, A., and Racher, P. (2012). “Analytical study and finite Xu, B. H., Bouchaïr, A., Taazount, M., and Vega, E. J. (2009). “Numerical
element modeling of timber connections with glued-in rods in bending.” and experimental analyses of multiple-dowel steel-to-timber joints in
Constr. Build. Mater., 34, 337–345. tension perpendicular to grain.” Eng. Struct., 31(10), 2357–2367.
Downloaded from ascelibrary.org by University of Waterloo on 02/14/25. Copyright ASCE. For personal use only; all rights reserved.

© ASCE 04014165-11 J. Struct. Eng.

J. Struct. Eng., 2015, 141(6): 04014165