Engineering Structures 65 (2014) 52–61

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Joints and wood shear walls modelling I: Constitutive law, experimental

tests and FE model under quasi-static loading
J. Humbert a, C. Boudaud b, J. Baroth c,⇑, S. Hameury d, L. Daudeville c
a
Korea Forest Research Institute, Seoul, Republic of Korea
b
LUNAM Univ., Groupe Ecole Superieure du Bois, Nantes, France
c
UJF-Grenoble 1, Grenoble-INP, CNRS UMR 5521, 3SRLab, Grenoble F-38041, France
d
Centre Scientifique et Technique du Bâtiment, Marne-la-Vallée, France

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

Article history: This study is the first of two companions papers that present a finite element (FE) model of timber-frame
Received 18 June 2013 structures. It introduces a versatile hysteretic constitutive law developed for various joints with steel fas-
Revised 14 January 2014 teners commonly used in timber structures (nails, screws, staples, bracket-type 3D connectors, punched
Accepted 29 January 2014
plates). Relative to previous models available in the literature, the proposed model improves numerical
Available online 3 March 2014
robustness and represents a step forward by taking into account the damage of joints with metal fasten-
ers. More than 300 experimental tests are carried out on joints and used to calibrate the constitutive law
Keywords:
for nails and bracket-type 3D connectors. An average calibration method is presented to take into account
Timber-frame structures
Shear walls
the experimental variability. 14 experimental tests are performed on different configurations of shear
Finite element walls and are used to validate the proposed FE model. Both monotonic and reversed cyclic loadings
Cyclic loading are used in these quasi-static tests. The FE model predictions are in good agreement with the experimen-
Constitutive law tal results. The second paper will present dynamic experiments and numerical predictions of the tests, as
well as the development and validation of a computationally efficient simplified modelling of timber-
frame structures based on a simplified finite element model for shear walls.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction local behaviours. This study focuses on shear walls, as they contrib-
ute the most to the energy dissipation of structures.
This paper is motivated by two facts. First, timber-frame con- The work presented in these papers is based on a coupled
struction is an increasingly common building system in Europe, experimental/FE modelling approach. One should note that the
primarily for residential single or two-story houses. These struc- behaviour of shear walls can also be estimated through an analyt-
tures present many qualities, including good earthquake resistance ical approach [26,21], but such a method would not allow the anal-
due to the excellent strength-to-density ratio of timber and to the ysis of both the global and local behaviour of a timber-framed
ductility of joints with metal fasteners, providing limited inertia structure. Therefore, in this study, quasi-static experimental tests
forces and good energy dissipation, respectively. Second, the most on metal fasteners (nails, bracket-type 3D connectors and punched
recent European code for the design of earthquake-resistant build- plates) are performed to calibrate their hysteretic constitutive
ings (Eurocode 8 [15]) has been accompanied by a new seismic behaviour. Quasi-static and dynamic tests on shear walls are car-
hazard map in some countries. Generally, based on these revised ried out to validate the numerical model for shear walls. Because
maps, earthquake resistance calculations are now mandatory in a nonlinear dissipative phenomena in timber-frame structures are
lot more cases and the design ground accelerations are greater mainly concentrated in joints, simplified force–displacement mod-
than previously. Therefore, the seismic behaviour of timber-frame els for joints can be derived from refined analytical or FE models
structures must be studied, to better understand their global and [7,10,1,32,35] or by fitting the results of tests performed on joints
[30]. The proposed approach is based on a multi-scale concept, as
proposed previously by various authors [38,18,42,36]. Such an ap-
proach requires a behaviour law to represent the force–displace-
⇑ Corresponding author. Address: 3SR Laboratory, Domaine Universitaire, BP53,
ment evolution on each scale. Numerous constitutive laws have
38041 Grenoble Cedex 9, France.
been developed over the years, from the nonlinear laws for mono-
E-mail addresses: [email protected] (J. Humbert), clement.boudaud@
ecoledubois.fr (C. Boudaud), [email protected] (J. Baroth). tonic loads [20,22,27] to hysteretic models of various complexities

http://dx.doi.org/10.1016/j.engstruct.2014.01.047
0141-0296/Ó 2014 Elsevier Ltd. All rights reserved.
 J. Humbert et al. / Engineering Structures 65 (2014) 52–61 53

[8,39,28,40,6,11,17,43,33]. Henceforth, only the hysteretic laws can be identified. Although most of the models already meet
capturing a damage process are discussed. Richard et al. [38] pro- this condition, the BWBN model does not.
posed a strength reduction based on a cumulative factor calculated
in one direction in respect to the previously achieved strength in It is important to notice that the hysteretic behaviour of nailed
the opposite direction. Collins et al. [9] defined a quite similar wood joints governs the response of many wood systems when
damage process. Although most of these constitutive laws use subjected to lateral loadings; the force–displacement backbone
exponential functions for the pre-peak backbone curve and hyster- and hysteresis curves of shear walls and joints are then similar in
esis loops, the model of Ayoub [3] is defined with trilinear func- shape. Thus, a common feature to all the abovementioned force–
tions, in this model the damage process is described in detail and displacement models is that they can be used to describe the con-
can be divided into four degradation phenomena: strength reduc- stitutive behaviour of joints as well as the global shear wall re-
tion, unloading stiffness decrease, accelerated stiffness decrease sponse to lateral forces.
and cap degradation. The evolutionary parameter hysteretic model In this study, a new hysteretic constitutive behaviour law for
(EPHM) proposed by Pang et al. [37] is only defined by exponential joints and timber-frame structures is proposed, and its application
functions (pre and post-peak backbone, unloading and loading to the modelling of oriented strand board (OSB) and particleboard
hysteretic loops) and damage is not cumulative. The latest version sheathed shear walls is presented. More than 260 tests on nailed
of the Bouc-Wen-Baber-Noory (BWBN) model has been presented joints and 50 tests on connections made with bracket-type 3D con-
by Xu and Dolan [42]. The BWBN model is analytical and phenom- nectors are performed. The calibration of the law at the joint scale
enological. Its history-dependent stiffness and strength degrada- is detailed, and particular emphasis is given on how to take into ac-
tion provide accurate fitting of reversed-cyclic experimental tests count the variability of the experimental results. Tests performed
on nailed connections and shear walls, but the BWBN does not rely on 7 different configurations (combining different specimens and
on physical parameters such as displacements, forces and stiffness- vertical loadings) of shear walls are described. The development
es. A new model, developed by Humbert [23], can be considered an of the numerical model of shear walls is then explained. To assess
improvement of the Richard et al. [38] and Yasumura et al. [43] its capability to predict the behaviour of different configurations of
models and fulfils the following needs: shear walls, its predictions are compared to the experimental re-
sults of the 14 tests under quasi-static loading. Experimental tests
 Richard’s behaviour law shows that for some sets of parameters present a certain variability and the large sample size allows its
(e.g. for a metal punched plate), an exponential function does quantification. Then, when comparing the deterministic predic-
not provide a strict analytical continuity at one end of the tions of the model to the experimental results, the experimental
branch leading to numerical issues [23]. This issue is shared variability can be considered. Moreover, tests on different configu-
by all models using the exponential functions originally intro- rations are designed to estimate the model versatility.
duced by Foschi [20].
 The law should model asymmetric behaviour, such as that of 2. Force–displacement hysteretic constitutive law
punched metal plates for roof trusses and bracket-type 3D con-
nectors. To the best of our knowledge, all of the aforementioned The one-dimensional constitutive law is shown in Fig. 1.
behaviour laws would require new developments to meet this The following notations are used to describe the asymmetric
need. feature of the modelled systems and the notion of force sign. The
 For the reliability analysis of structures, it is convenient to þ direction corresponds to the first direction of loading in the
develop a robust model defined by physical parameters such case of reverse loading ( refers to the opposite direction). In
as displacements, forces, and stiffnesses, whose variabilities the absence of a superscript, the parameters refer to both sides

Fig. 1. Proposed force–displacement constitutive law [23].

54 J. Humbert et al. / Engineering Structures 65 (2014) 52–61

Table 1 Table 2
Model parameters governing the constitutive behaviour under monotonic loading. Model parameters governing the shape of hysteresis loops.

Parameter Unit Description Parameter Unit Controls

K0 N/m Initial elastic stiffness C1 – Unloading stiffness
dy m Yield limit C2 – Reloading stiffness
d1 m Displacement at peak force C3 – Tangent stiffness at F ¼ 0
F1 N Peak force C4 – Residual displacement
K1 N/m Pre-peak tangent stiffness
d2 m Intermediate displacement limit
F2 N Force at intermediate limit d2 residual displacement u0 ¼ upk  F pk =K 4 after a hypothetical linear
du m Ultimate displacement
elastic unloading with stiffness K 4 from (upk ; F pk ). Alternatively,
Fu N Force at ultimate displacement
for some other cases, dc is related to the residual displacement
after a linear elastic unloading with a stiffness j C 4 j K 4 propor-
tional to the secant stiffness. These control parameters C i¼1;...;4 (Ta-
of loading. A bullet superscript  is used to refer to the side op-
 ble 2) mainly depend on the phenomena involved, and therefore on
posed to the implied one. For example, a ¼ b implies that
þ  the configuration of the modelled system. They are constant for a
a ¼ b and aþ ¼ b . The branches of the force-displacement mod-
given configuration.
el are grouped into two distinct categories and numbered from (0)
Lastly, a third set of 3 parameters controls the damage process
to (5). A first group, formed by branches (0) to (3), describes the
of the model. The word damage refers here to the decrease in
behaviour under monotonic loading. The initial linear branch (0)
strength under cyclic loading. It is based on the hypothesis that
ranges from the zero displacement up to the yield displacement
the hysteresis loops are bound by the backbone curve, which mod-
dy . The corresponding elastic stiffness is K 0 . This branch is followed
els the force-displacement evolution of the joint under monotonic
by branch (1), which models the nonlinear phenomena in the joint
loading. During the first loading, the peak (upk ; F pk ) is located on the
up to the force peak at (d1 ; F 1 ). After the force peak, branches (2)
backbone curve. The damage process defines the evolution of the
and (3) model up to the ultimate displacement du at force F u asso-
ratio (1-D) between the ‘‘nondamaged load’’ F mono and the ‘‘dam-
ciated with the collapse of the joint. F u is generally chosen to be
aged load’’ F pk . The scalar damage indicator D ranges from 0 to 1,
null to ensure a correct continuity of forces and prevent numerical
where D ¼ 0 corresponds to a nondamaged mechanical system
issues. This first set of 9 parameters is summarised in Table 1.
and D ¼ 1 corresponds to a fully collapsed mechanical system. D
These parameters are similar to those proposed by Richard et al.
is increased by DD at each change of the force sign ((4) to (5) in
[38]. Branch (1) is defined using a rational quadratic Bézier curve,
Fig. 1). To ensure the damage stabilisation after a few cycles of con-
replacing the original exponential function introduced by Foschi
stant amplitude occurs as experimentally observed, the increment
[20] and providing a strict analytical continuity of forces.
DD is defined as DD ¼ gðD1  DÞ, where g defines the proportion of
A second group of branches describes the hysteresis loops typ-
damage at constant amplitude cycles and D1 is an upper limit of D
ically observed when the joint undergoes a reversed loading. To de-
for the displacement dmax . This limit is constant for a given maxi-
scribe these branches, the max subscript is defined, which
mum displacement dmax . D1 is defined by Eq. (1) as a function of
corresponds to the absolute maximum value reached over the past
dmax using a power law. This function is referred to as the ‘‘damage
loading history up to the current time step. Additionally, the pk
limit function (DLF)’’. A power term Br > 1 ensures that the damage
subscript corresponds to the value at the last loop peak. Starting
remains moderate before the force peak and becomes severe after
from a previously reached loop peak (upk ; F pk ), branch (4) models
the peak. This model is consistent with experimental observations.
the nonlinear elastic unloading down to a null force. A residual dis-
Table 3 summarises the three damage parameters.
placement dc –0 is commonly observed due to prior plastic defor-
Br
mations. The unloading stiffness K 4 is either (a) proportional to D1 ¼ Bc ðdmax =d1 Þ ð1Þ
the elastic stiffness K 0 of the joint or (b) proportional to the secant
stiffness F pk =upk when modelling a stiffness decrease with displace-
3. Scale 1: Joints with metal fasteners
ments of increasing amplitude. Examples of (a) and (b) include
cases in which the overall behaviour of the joint is dominated by
In this section, quasi-static experimental tests carried out on
the embedment of wood dominating and by the pulling-out of
joints with metal fasteners are presented. Then, the calibration of
the fastener from the timber members and/or the plastic yielding
the parameters of the constitutive law of joints is described.
of the fastener, respectively. Following this unloading, loading in
the opposite direction is modelled with branch (5). The stiffness 3.1. Experimental tests
at dc between branches (4) and (5) is denoted by K c and is used
as a tangent for both branches for the sake of continuity. Branch Experimental tests on joints with metal fasteners (scale #1) are
(5) eventually reaches the previous loop peak (upk ; F pk ) in the oppo- conducted to provide input data for the numerical model of the
site loading direction. Like the unloading stiffness K 4 , the reloading shear wall (scale #2). There are three different steel joints in a
stiffness K 5 is proportional to the elastic stiffness K 0 or the secant shear wall:
stiffness F pk =upk . A second set of 4 control parameters C i¼1;...;4 gov-
erns the shape of the hysteresis loops, allowing the modelling of  Panel-to-frame (P2F) joint: In this study, these joints are made
several mechanical behaviours, e.g., the adjustment of the thick- with nails. Fig. 2a shows a nail after a shear wall test. Only ring
ness of the pinching area. Parameters C 1 and C 2 control the unload- shank and square masonry nails are used, as it is common in
ing stiffness K 4 and reloading stiffness K 5 , respectively. Parameter
C 3 controls the tangent stiffness K c at location (dc ; 0). Values of C 3
Table 3
greater than 1 model a thick pinching area, while values less than 1 Parameters governing the damage indicator calculation.
model a thin pinching area. A physical limitation is also imposed so
Parameter Unit Description
that K c cannot exceed the initial elastic stiffness K 0 . Finally, param-
eter C 4 controls the value of the residual displacement dc after the Bc – Linear coefficient of the DLF
Br – Power term of the DLF
nonlinear elastic unloading. Based on observations from experi-
g % Damage proportion at constant amplitude cycles
mental tests, this displacement can be related to the fictive
 J. Humbert et al. / Engineering Structures 65 (2014) 52–61 55

Fig. 2. Photographs of the three joints with metal fasteners used in a shear wall.

France. Their dimensions ranges from 2:1 mm to 3:1 mm in considered in this study: 9 and 12 mm OSB with 2:1  45 mm nails
diameter and 45 mm to 90 mm in length. Electrogalvanization and and 16 mm particleboard with 2:5  50 mm nails, respectively
hot-dip galvanizing coatings, as well as stainless steel are used. named OSB9, OSB12 and P16. The wood member dimensions are
 Frame-to-frame nail (F2F nail): The joints between the top/sill 45  115  250 mm. Fig. 3a shows the principle of these tests,
plate and the studs are composed of 3 to 5 nails. Their usual which consists of a shear test, first under a monotonic loading
length is approximately 90 mm for a diameter of 3:1 mm. and then under reversed-cyclic loading (EN 12512 [13]). The grain
Fig. 2b presents such a connection after a pull-out test. orientation of the wood member and the orientation of the panel
 Frame-to-frame bracket (F2F bracket): At both ends of the shear are not variable parameters of the tests, based on the results pre-
wall, the F2F connections must be strengthened to prevent sented by [19] for parallel and perpendicular to grain tests on nail
uplift of the exterior studs. Thus, an additional bracket-type joints. The high number of configurations and the repeated tests
3D connector is used. Fig. 2c displays a F2F bracket joint at per configuration led to a total of 263 tests for P2F nail joints.
the end of a shear wall test. The 3D connector is connected by F2F nail joints were not tested and the results of tests achieved
an anchor or bolt to the foundation or the lower story. by Richard et al. [38] are used instead. These tests (Fig. 3b) consist
of a cyclic pull-out load on a joint. The same tests are performed on
Thirty-three configurations of P2F nail connections are tested, 3 F2F joints made of bracket-type 3D connectors only (Fig. 3c). In
of which are used in the 3 tested configurations of shear walls that case, wood member dimensions are 45  140  400 mm. For

Fig. 3. Experimental tests on metal fastened joints.

56 J. Humbert et al. / Engineering Structures 65 (2014) 52–61

each configuration, the tests were repeated twice for monotonic distinguished: the first level is a direct calibration, which consists
loading and 5 times for reversed cyclic loading, which led to a total of reproducing one particular test, and the second level is an aver-
of 56 quasi-static tests. age calibration, which consists of calibrating the parameters to
Although it is not the main concern of this paper, some results reproduce the average behaviour observed in several experiments.
of the experimental tests are provided herein. For each monotonic The hysteretic model is based on three types of parameters: back-
test, the yield displacement V y , the ultimate displacement V u (from bone curve (Table 1), pinching (Table 2), and damage (Table 3).
which the ductility is derived Ds ¼ V u =V y ), the maximal force F max Fig. 5a presents a direct calibration of parameters for a P2F nail
are calculated according to the Equivalent Energy Elastic–Plastic joint. This calibration is achieved by calculating the backbone
(EEEP) method [2]. curve parameters from a single test under monotonic loading
For the nail connections, the failure mode according to Johan- and calibrating the pinching and damage parameters by successive
sen’s theory [24] (only one or two plastic hinges, because no wood simulations. The average calibration is based on the observation
crushing modes were observed) and the ultimate failure mode that backbone curve parameters display some variability, while
(withdrawal of the nail shank or pulling through of the nail head) the pinching parameters do not. Using the direct calibration as a
are also provided. Three monotonic tests are performed for each starting point, the backbone curve and damage parameters are
configuration of nail joint and Table A.5 presents the average re- re-calibrated so that the simulation now reproduces the average
sults. The EEEP method provides consistency in the V y calculation envelope curve of all available cyclic tests. Fig. 5b presents the
method but, as it has been show by Munoz et al. [34] and con- average envelope curve and the calibrated model. This process pro-
firmed by Malo et al. [31], is hardly satisfying for some tests re- vides the joint models used in the subsequent shear wall
sults. Moreover, the values of V y greatly affects the static modelling.
ductility Ds and explain why it reaches unexpected values, such It is worth noticing that, in Fig. 5a, the force–displacement curve
as 43 for configuration N15. Under monotonic loading the domi- under reversed cyclic loading is asymmetric. The resistant forces
nant mode of failure is the pulling through of the nail head. With- are greater in compression (when the panel moves downward,
drawal of the nail shank is also observed, it is generally partial and see Fig. 3a). It is believed that the asymmetry of the timber assem-
becomes a failure mode for smaller nails dimensions. As expected, bly and slight misalignments of the test machine induce more fric-
it is observed that smaller diameters tends to fail with two plastic tion in compression. This phenomenon was also observed by
hinges, indeed this failure mode only happens for diameter equal Fonseca et al. [19] and Li et al. [29], who present experimental tests
or inferior to 2:3 mm. It is also confirmed that the two plastic on similar connections (panel to stud with only one nail). Li et al.
hinges failure mode leads to greater values of ductilities. The [29] explain this behaviour by the fact that wood and metal dam-
Fig. 4a shows a typical force–displacement evolution of a nail con- ages on one side also affect the strength on the other side. However,
nection. It is obvious that damages due to the cyclic loading re- although the tests described in this study were conducted by load-
duces the maximal force F max (by 20% in average) and the ing first in traction and then in compression, the forces were great-
ultimate displacement V y . er in compression, not in traction. It is then believed that the
For the bracket connections, the failure modes are complex as phenomenon described by Li et al. [29] does not have a significant
they include the nails behaviour under shearing and strict with- effect relative to the friction phenomenon we described. This point
drawal loading, the bracket folding or unfolding and wood crush- is further discussed later in this study for tests on shear walls. As
ing. For that reason, the results of the 3D bracket tests presented the loading conditions of the joint tested are not strictly similar
in Table A.6 are limited to the values of V y ; V u ; Ds and F max . The to the conditions for P2F connections in the full size shear wall, nei-
Fig. 4b displays a typical force–displacement evolution of a single ther in compression nor in traction, the calibration is achieved on
bracket specimen in the Y direction. In that case, the joint is stiffer the average behaviour between compression and traction.
in Y as it corresponds to the wood compression, while Yþ only Fig. 6a shows the numerical behaviour of a F2F nail joint in pull-
correspond to the unfolding of the bracket. In average, the maximal out and compression. The compression (contact between the two
force F max under cyclic loading is 15% smaller than under mono- timber elements) is linear and the stiffness is calculated according
tonic loading, and the ductility is reduced by 32%. to the material characteristics and the dimensions of the section in
contact. The pull-out behaviour is bilinear, and the parameters are
3.2. Calibration of the force–displacement model estimated from tests carried out by Richard et al. [38]. The shear
behaviour is linear and symmetric. The stiffness is calculated
The results of the tests achieved on P2F nails are used to cali- according to Eurocode 5 [14]. Note that for all joints (P2F and
brate the constitutive model. Two levels of calibration are F2F), no rotational stiffness is implemented. For F2F joints, it is

1.5 8
Cyclic
6 Monotonic
1 4
2 Z
Force (kN)

Force (kN)

0.5 X
0 Y-
−2 Y+
0
−4

−0.5 −6
Cyclic −8
Monotonic
−1 −10
−10 −5 0 5 10 15 20 25 30 −40 −30 −20 −10 0 10 20 30
Displacement (mm) Displacement (mm)

Fig. 4. Typical monotonic and cyclic force–displacement evolutions for nail and 3D brackets connections.
 J. Humbert et al. / Engineering Structures 65 (2014) 52–61 57

ties and set as bilinear. Shear behaviour is linear and symmetric.

The E5Ò pull-out behaviour parameters are calibrated using the
same method as P2F connections.

4. Scale 2: Shear wall

In this section, quasi-static experimental tests carried out on

shear walls are presented. Then, numerical modelling of timber
framed shear walls is detailed. Finally, the predictions of the FE
model of shear wall are confronted to the experimental results.

4.1. Experimental tests

The shear walls studied in this study correspond to structural

elements found in conventional timber-framed houses in Europe,
which are quite similar to typical North American shear walls, ex-
cept for the nail diameter of the P2F joints. The diameter of these
nails rarely exceeds 3 mm in France (common dimensions are be-
tween 2:1 mm and 2:5 mm) while it is generally greater than 3 mm
in America. Fig. 7 describes the shear wall technology and its
dimensions. The frame is made of C24 timber (strength class
according to the European standard EN 338 [16]). To provide
shearing resistance, wood-based panels are nailed to the frame.
These panels are usually OSB, particleboard or plywood. In this
study, OSB-3 panels and P5 particleboards are used (panels
classification according to EN 12369-1 [12]). The P2F joints can
be formed with nails, screws, or staples. The spacing between
two P2F joints along the perimeter of the panel is set to 150 mm
(sext ) and 300 mm (sint ) along intermediate studs. The connections
between horizontal frame elements (sill or top plates) and
vertical frame elements (studs) are formed with long nails
(3:1 mm  90 mm in this study). The anchorage of the shear walls
– on the foundation or the lower story – is achieved with bolts.
Current anchorages refer to regularly settled bolts, one in every
span between two studs. Their purpose is to transfer the shearing
load. Exterior anchorages refer to the addition of a 3D connector af-
Fig. 5. Calibration of the hysteretic model for a 2:1 mm  45 mm P2F nail in a 9 mm
fixed to the exterior stud and a bolt. Their purpose is to transfer the
OSB panel.
vertical uplifting loads. Two 3D bracket-type connectors are used,
both provided by Simpson Strong-TieÒ: E5Ò standard and
justified by the fact that their rotational stiffness is insignificant AH2950/2Ò reinforced connectors (referred to as AH). AH brackets
compared to the stiffness provided by the sheathing panels. are specifically designed to withstand high uplifting loads.
Fig. 6b shows the numerical behaviour of a F2F bracket joint The experimental tests on shear walls are detailed in Boudaud
with an E5Ò bracket-type 3D connector in pull-out and compres- et al. [4]. Fig. 7 displays the principle and protocol (EN 12512
sion. This combination behaves the same in compression as the [13]) of the tests. OSB9, OSB12, and P16 configurations of shear
F2F nail joints. Experimental tests were carried out on E5Ò connec- walls are tested. One push-over and two reversed-cyclic tests were
tors in shear and pull-out. In contrast, the behaviour of joints made achieved for each configuration tested. Maximal forces for cyclic
with AH connectors was estimated based on the connector proper- tests can be seen on Table 4.

1 10

0.5 5

0
0
Force (kN)

Force (kN)

−5
−0.5
−10
−1
−15
−1.5 −20

−2 −25
−2 0 2 4 6 8 10 −2 0 2 4 6 8 10 12
Displacement (mm) Displacement (mm)

Fig. 6. Behaviour law for Frame-to-frame joints.

58 J. Humbert et al. / Engineering Structures 65 (2014) 52–61

Fig. 7. Shear wall description and principle of reversed cyclic tests.

Table 4
Experimental and numerical comparison for reversed-cyclic loading.

Configuration Results
Panel F2F Angle P2F Vertical F max (kN) Ref.
(£  L) load (kN) Exp Num D ð%Þ
OSB9 E5 2:1  45 0 12.0 11.7 +2.5 SW1
12.2 +4.1 SW2
5.5 11.9 11.3 +5.3 SW3
12.7 +3.4 SW4
OSB12 E5 2:1  45 0 12.1 11.8 +2.2 SW5
6 13.2 13.0 +1.7 SW6
AH 2:5  50 3 12.4 13.7 10 SW7
14.0 +1.5 SW8
P16 E5 2:5  50 6 14.2a 14.8a 3.7a SW9
18.7 18.8 0.38 SW10
AH 0 23.2 20.9 +10 SW11
23.4 +11 SW12
3 22.0 20.8 +5.4 SW13
22.9 +8.9 SW14
a
For this shear wall sext ¼ 300 mm along the central stud (instead of 150 mm).

4.2. Numerical modelling

The finite element modelling of shear walls was conducted

using beam, plate, and two-node spring-like finite elements. The
constitutive behaviour presented previously was implemented
using the free software Code_Aster.1 Fig. 8 presents a schematic
view of the finite element model of a shear wall. Euler beam ele-
ments model the frame, and their elastic modulus and density are
the mean values corresponding to C24 timber according to EN 338
[16] provisions. Four-node plane stress elements model the panels,
and their material properties are isotropic and correspond to OSB-
3 or P5 according to EN 12369-1 [12] provisions. Each two-node
spring-like element models a metal fastener joint whose properties
are given by the previous calibration. The resulting mesh is com-
posed of 108 P2F, 8 F2F nail and 2 F2F angle elements. The F2F joints
have different behaviours in shear and pull-out/compression: there-
fore, each behaviour law is affected to the corresponding transla-
tional degree of freedom (DOF). For each P2F connection, a local
basis is oriented according to the relative displacement direction,
and the constitutive law is affected to both translational DOF in this
Fig. 8. Finite element modelling of a shear wall.
basis. Richard et al. [38] showed that this direction is globally con-
stant over the calculation. Judd [25] developed an oriented spring
strictly constant, its change over the calculation has been quantified
pair model based on this observation. As the drift direction is not
by calculating the ratio F y =F x (F x and F y are the forces associated
with the translational DOFs of the element in the local basis). This
1
All documentation available at www.code-aster.org. ratio is less than 0:3 for 80–90% of the P2F connections. As a result,
 J. Humbert et al. / Engineering Structures 65 (2014) 52–61 59

500
Experimental FE model

Dissipated Energy (J)

400

300

200

100

0
0 5 10 15 20 25 30
Cycle

15
Experimental FE model
Peak Force (kN)

10

0
0 5 10 15 20 25 30
Cycle

Fig. 9. Dissipated energy and peak forces: comparison between the experimental results and FE model predictions.

the effect of the overestimation of the resistance of the joints is configuration, which limits the effects of the variability when com-
insignificant at the scale of the wall. paring the experimental behaviour to the numerical predictions.
The sill plate is considered to be embedded because insignifi- Table 4 presents the results in terms of peak forces. It shows that
cant displacements are recorded in tests. The top plate and the the peak forces predicted by the numerical model are in good
highest node of each external stud undergo an imposed displace- agreement with the experimental results.
ment. To comply with the experimental set-up, as the actuator To assess the quality of the model predictions, the dissipated
pushes a metal plate attached to the wood top plate and the top energy is often used [5,9,37,41]. This approach consists of calculat-
of the external studs. Experimentally, a vertical load is applied to ing the areas inside the hysteresis loops, but similar values do not
the shear wall in approximately half of the tests (6 kN on each necessarily mean that both force–displacement evolutions are
stud). The global responses show very limited influence of the ver- equivalent. As a result, the area and peak force of each half-cycle
tical load. Nevertheless, the uplift of the external studs is affected should be compared simultaneously to assess the similarity be-
by the vertical load. With loading, no uplift is observed. Without tween experimental and numerical hysteresis loops. Fig. 9 presents
loading, the uplift of exterior studs is significant, and the 3D con- such a comparison for one of the tests (OSB12 C2), showing that
nectors are strained beyond their yield limit. Numerically, a verti- the detailed FE model is able to predict the peak force and the area
cal load can be applied to the model. In that case, the load is inside the hysteresis loops, and therefore their shapes, fairly accu-
uniformly distributed along the plate. rately. It also shows that the errors of the FE model predictions are
more significant for the last few cycles. Nevertheless, authors point
4.3. FE model predictions vs. experimental results out that the same observations can be made from the force–dis-
placement comparison presented in Fig. 10a, for the same test.
The model predictions are compared to 14 experimental results Fig. 10b compares the experimental and calculated force–displace-
obtained from quasi-static tests under reversed-cyclic loading. ment curves for the OSB9 shear wall. From the two examples of
Thus, the numerical predictions can be compared to several config- predictions presented in Fig. 10, and the results presented in Ta-
urations of shear walls (different nails, sheathing panels and verti- ble 4, it can be seen that the model predictions are in good agree-
cal loading conditions). Moreover, two tests are performed for each ment with the experimental behaviour. Indeed, the pinching and

15 15

10 10

5 5
Force (kN)

Force (kN)

0 0

−5 −5

−10 Experimental −10 Experimental

FE model FE model
−15 −15
−80 −60 −40 −20 0 20 40 60 80 −80 −60 −40 −20 0 20 40 60 80
Displacement (mm) Displacement (mm)

Fig. 10. Force–displacement curves of shear walls: comparisons of experimental results and FE model predictions.
60 J. Humbert et al. / Engineering Structures 65 (2014) 52–61

peak forces of the hysteresis loops are in accordance with the describe the hysteretic behaviour, notably by considering the dam-
experimental data. Fig. 9a and b also show that the last cycle in age effects (strength reduction), and its numerical robustness com-
the negative side is over predicted. Experimentally, this cycle is pared to existing models using exponential functions.
clearly asymmetric in terms of force, which is explained by the fact In the second part of this study, the results of more than 300
that the failure of the wall occurs during this cycle. Therefore, these tests performed on timber joints with metal fasteners show a sig-
over predictions are not considered to be an issue, because the nificant variability. An average calibration method is thus devel-
model is ‘‘only’’ intended to predict the behaviour up to the failure. oped for the identification of the model parameters. Using these
One should note that the error estimation between experimen- calibrated models of joints, a FE model of a shear wall is developed.
tal and numerical hysteretic curves equally concerns the The frame is modelled by Euler beams elements, the panels by
constitutive law calibration at scale 1 and the detailed FE model four-node plates elements, and every joint by a two-node spring-
predictions at scale 2 presented herein. In this study, errors quan- like element. The FE model predictions are compared to the results
tification has not been used, as the quality of the numerical predic- of 14 experimental quasi-static tests for validation. These compar-
tions (Table 4) were considered to be satisfactory. isons show that the FE model accurately predicts the experimental
Like nails, shear walls present a strength asymmetry. For nails, behaviour of different configurations of shear walls.
this behaviour is believed to be due to the asymmetry of the wood In part II of these two companions papers, dynamic experimen-
assembly. This cannot be the case for shear walls because they are, tal tests and numerical calculations are addressed, and a simplified
like the test machine, symmetric. Moreover, a comparison of the nail FE model of shear wall is presented. This simplified model is used
and shear wall results (Figs. 5a and 10, respectively) show that the to build a FE model of the structure.
behaviours are quite different. For nails, the asymmetry is significant
at all magnitudes, whereas it appears to increase with the cycle mag- Acknowledgements
nitude in the case of shear walls. This can be explained by the phe-
nomenon described by Li et al. [29] (page 12). In conclusion, it is The SISBAT research project is funded by the French Agency
believed that the two phenomena can induce a strength asymmetry. ANR (Agence Nationale de la Recherche ANR-08-RISKNAT2008).
For tests on asymmetric joints, the effect of friction is dominant, Authors would like to express their gratitude for this support and
while the effect of damage prevails for tests on symmetric elements. thank the members of this project for their contributions and dat-
abases, especially Mrs C. Faye (FCBA Technological Institute) and
5. Conclusion her colleagues, for the effort provided.

This study is dedicated to the development of a versatile hyster- Appendix A. Experimental data
etic constitutive behaviour law for timber joints made of metal fas-
teners. The main features of the model are its ability to accurately Tables A.5 and A.6.

Table A.5
Experimental results on nail joints (Fig. 3a) for monotonic loading.

No. Configuration Results

£L Shape Mat Panel Vy Vu Ds F max Mode Hinge

N1 2:1  38 RSN SS OSB9 0.8 21.7 28.2 962 W 2

N2 OSB12 1.3 21.9 22.1 934 W+P 2
N3 OSB15 0.6 15.0 26.6 734 W 2
N18 2:5  60 OSB9 2.1 17.5 8.4 1626 P 1
N19 OSB12 2.5 21.8 8.7 1470 P 1
N20 OSB15 1.6 21.8 14.0 1142 P 1
N25 3:1  85 OSB9 2.2 17.5 8.1 1913 P 1
N26 OSB12 2.6 22.9 9.0 1661 P 1
N27 OSB15 2.7 23.3 9.0 1293 P 1
N4 2:1  45 EG OSB9 1.2 15.5 15.3 855 P 1
N5 OSB12 2.3 21.1 9.9 1005 P 1
N16 2:5  50 P10 2.8 24.1 8.8 1577 W 1
N17 P16 1.1 23.2 20.9 1276 W 1
N6 2:1  45 OSB9 1.6 15.3 9.5 925 P 1
N7 OSB15 0.7 17.3 29.2 1001 W+P 2
N21 2:8  80 OSB9 2.1 15.7 7.9 1525 P 1
N22 OSB15 1.3 21.6 16.9 1962 W+P 1
N8 2:3  60 HdG OSB9 1.4 13.9 10.7 1153 P 1
N9 OSB15 3.4 24.2 7.4 1247 W+P 1
N30 3:1  90 OSB9 1.6 14.3 9.3 1244 P 1
N31 OSB15 1.8 18.1 10.3 1871 P 1
N14 2:3  60 SMN SS OSB9 1.1 26.0 32.4 1013 P 2
N15 OSB15 0.8 29.6 43.0 965 W 2
N28 3:1  85 OSB9 0.6 14.7 23.2 1797 P 1
N29 OSB15 1.0 30.1 34.8 1329 P 1
N12 2:3  60 HdG OSB9 1.4 13.9 10.3 1279 P 1
N13 OSB15 0.9 19.2 21.4 1629 P 2
N23 3:1  75 OSB9 1.2 15.9 20.8 2073 P 1
N24 OSB15 1.0 18.2 18.0 1864 P 1
N10 2:3  60 RSN P10 2.2 24.3 11.3 2208 P 1
N11 P16 2.7 32.9 12.5 1583 W 2
N32 3:1  90 P10 1.6 20.5 13.9 2015 P 1
N33 P16 2.0 23.3 12.4 2187 P 1

£  L: Nail diameter and length in mm. Mat: Nail material. V y and V u in mm. F max in N. Mode: Mode of failure. Hinge: Number of hinges. RSN: Ring Shank Nail. SMN: Square
Masonry Nail. SS: Stainless Steel. EG: Electrogalvanization. HdG: Hot-dip Galvanising. W: Nail withdrawal. P: Nail pulling through the panel.
 J. Humbert et al. / Engineering Structures 65 (2014) 52–61 61

Table A.6 [14] EN 1995-1-1. Design of timber structures. Part 1-1: General – common rules
Experimental results on bracket joints (Figs. 3c and 4b) for monotonic loading. and rules for buildings; 2005.
[15] EN 1998-1. Design of structures for earthquake resistance – general rules,
No. Configuration Results seismic actions and rules for buildings; 2005.
[16] EN 338. Structural timber – strength classes; 2003.
Type Dir. Support # Vy Vu Ds F max
[17] Folz B, Filiatrault A. Cyclic analysis of wood shear walls. J Struct Eng
B1 E5 X Wood 4 4.5 35.9 8.1 32.9 2001;127(4):433–41.
B2 Steel 4 3.1 31.7 10.1 19.4 [18] Folz B, Filiatrault A. Seismic analysis of woodframe structures. II: Model
B3 Y Wood 1 2.8 41.4 14.9 9.8 implementation and verification. J Struct Eng 2004;130(9):1361–70.
3.8 10.9 2.7 5.8 [19] Fonseca F, Rose S, Campbell S. Nail, wood screw, and staple fastener
B4 2 3.7 26.5 7.5 14.7 connections. Caltech wooframe project. Chapter: Testing setup; 2002.
[20] Foschi R. Analysis of wood diaphragms and trusses, Part 1: Diaphragms. Can J
B5 Steel 1 3.2 28.0 8.7 14.2
Civil Eng 1977;4(3):345–62.
1.3 10.6 9.0 4.5
[21] Girhammar U, Kallsner B. Elasto-plastic model for analysis of influence of
B6 2 2.1 28.5 13.9 15.1 imperfections on stiffness of fully anchored light-frame timber shear walls.
B7 Z Wood 2 1.8 9.0 5.0 7.3 Eng Struct 2009;31:2182–93.
B8 Steel 2 3.2 19.4 6.3 18.9 [22] Gupta A, Kuo G. Behavior of wood-framed shear walls. J Struct Eng
B9 XY Wood 1 12.3 50.4 4.1 11.6 1985;111(8):1722–33.
5.5 23.0 4.2 5.5 [23] Humbert J. Characterization of the behavior of timber structures with metal
B10 Steel 1 11.3 49.8 4.4 12.3 fasteners undergoing seismic loadings. Ph.D. thesis, Grenoble University;
5.9 23.3 3.9 9.2 2010.
B11 2 11.1 42.1 4.5 18.4 [24] Johansen K. Theory of timber connections. Int Assoc Build Struct Eng
B12 E14 X Wood 4 6.1 32.6 5.3 29.0 1949:249–62.
B13 Steel 2 5.1 42.0 8.3 17.2 [25] Judd J. Analytical modeling of wood-frame shear walls and diaphragms.
B14 Y Wood 1 4.6 26.0 5.7 10.8 Master’s thesis, Brigham Young University; 2005.
[26] Kallsner B, Girhammar U. Plastic models for analysis of fully anchoared light-
3.2 13.5 4.2 4.9
framed timber shear walls. Eng Struct 2009;31:2171–81.
B15 Steel 1 7.2 22.4 5.4 16.3
[27] Kasal B, Leichti R. Nonlinear finite-element model for light-frame stud walls. J
1.4 15.5 11.0 6.4
Struct Eng 1992;118(11):3122–35.
B16 Z Wood 2 1.3 5.7 4.3 6.5 [28] Kivell T, Moss P, Carr A. Hysteretic modelling of moment-resisting nailed
B17 Steel 1 3.6 23.6 6.6 18.8 timber joints. Bull. New Zealand Natl Soc Earthquake Eng 1981;14(4):233–43.
B18 XY Wood 1 11.3 38.6 3.4 9.2 [29] Li M, Foschi R, Lam F. Modeling hysteretic behavior of wood shear walls with a
5.6 19.1 3.4 4.9 protocol-independent nail connection algorithm. J Struct Eng
B19 Steel 1 20 50 2.5 16.5 2012;138(1):99–108.
6.8 24.3 3.6 9.9 [30] Loo W, Quenneville P, Chouw N. A numerical study of the seismic behaviour of
B20 2 21.5 38.7 1.8 26.6 timber shear walls with slip-friction connectors. Eng Struct 2012;34:233–43.
[31] Malo K, Siem J, Ellingsbo P. Quantifying ductility in timber structures. Eng
V y and V u in mm. F max in kN. Dir.: Direction. #: Number of brackets in the joint. Struct 2011;33:2998–3006.
[32] Meghlat E, Oudjene M, Ait-Aider H, Batoz J. A new approach to model nailed
and screwed timber joints using the finite element method. Construct Build
References Mater 2013;41:263–9.
[33] Meireles H, Bento R, Cattaro S, Lagomarsino S. A hysteretic model for frontal
[1] Andreasson S, Yasumura M, Daudeville L. Sensitivity study of the finite walls in pombalino buildings. Bull Earthquake Eng 2012.
element model for wood-framed shear walls. J Wood Sci 2002:171–8. [34] Munoz W, Mohammad M, Salenikovich A, Quenneville P. Determination of
[2] ASTM E 2126. Standard Test method for cyclic (reversed) load for shear yield point and ductility of timber assemblies: in search for a harmonised
resistance of vertical elements of the lateral force resisting systems for approach. Engineering Wood Products Association; 2008.
buildings. American Society for Testing and Materials (ASTM); 2007. [35] Oudjene M, Meghlat E, Ait-Aider H, Batoz J. Non-linear finite element
[3] Ayoub A. Seismic analysis of wood building structures. Eng Struct modelling of the structural behaviour of screwed timber-to-concrete
2007;29:213–23. composite connections. Compos Struct 2013:102:20–28.
[4] Boudaud C, Hameury S, Faye C, Daudeville L. European seismic design of shear [36] Pang W, Rosowsky D. Beam-spring model for timber diaphragm and shear
walls: experimental and numerical tests and observations. World Conf Timber walls. Struct Build 2010;163(SB4):227–44.
Eng Proc 2010. [37] Pang W, Rosowsky D, Pei S, Lindt JVD. Evolutionary parameter hysteretic
[5] Ceccotti A, Karacabeyli E. Validation of seismic design parameters for wood- model. J Struct Eng 2007;133(8):1118–29.
frame shear wall systems. Can J Civil Eng 2002;29:484–98. [38] Richard N, Daudeville L, Prion H, Lam F. Timber shear walls with large
[6] Ceccotti A, Vignoli A. A hysteretic behavioural model for semi-rigid joints. Eur openings: experimental and numerical prediction of the structural behaviour.
Earthquake Eng 1989;3:3–9. Can J Civil Eng 2002:29:713–724.
[7] Chui Y, Ni C, Jiang L. Finte-element model for nailed wood joints under [39] Saiidi M, Sozen M. Simple and complex models for nonlinear seismic response
reversed cyclic load. J Struct Eng 1998;124(1):96–103. of reinforced concrete structures. Structural research series No. 466, Civil
[8] Clough R. Effect of stiffness degradation on earthquake ductility requirements. Engineering Studies, University of Illinois et Urbana-Champaign; 1979.
Technical report No. SESM 66-16, University of California, Berkeley; 1966. [40] Stewart W. The seismic design of plywood-sheathed shear walls. Ph.D. thesis,
[9] Collins M, Kasal B, Paevere P, Foliente G. Three-dimensional model of light Univ. of Canterbury, Christchurch, New Zealand; 1987.
frame wood building. I: Model description. J Struct Eng 2005;131(4):676–83. [41] Xu J, Dolan J. Development of a wood-frame shear wall model in abaqus. J
[10] Daudeville L, Davenne L, Yasumura M. Prediction of the load carrying capacity Struct Eng 2009;135(8):977–84.
of bolted timber joints. Wood Sci Technol 1999:15–29. [42] Xu J, Dolan J. Development of nailed wood joint element in abaqus. J Struct Eng
[11] Dolan J. The dynamic response of timber shear walls. Ph.D. thesis, University of 2009;135(8):968–76.
British Columbia, Vancouver, BC, Canada; 1989. [43] Yasumura M, Kamada T, Imura Y, Uesugi M, Daudeville L. Pseudodynamic tests
[12] EN 12329-1. Wood-based panels – characteristic values for structural design – and earthquake response analysis of timber structures ii: two-level
Part 1: OSB, particleboards and fibreboards; 2002. conventional wooden structures with plywood sheathed shear walls. J Wood
[13] EN 12512. Timber structures – test methods – cyclic testing of joints made Sci 2006;52:69–74.
with mechanical fasteners; 2002.