Paper_Risks
Paper_Risks
net/publication/352103618
CITATIONS READS
13 1,517
2 authors:
All content following this page was uploaded by Miguel Yurrita on 09 June 2021.
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A R T I C L E I N F O A B S T R A C T
Keywords: Timber connections can fail in a ductile or in a brittle way. A structural design that guarantees a ductile
Timber connection behaviour in case of failure is desirable, especially when facing extreme situations such as earthquakes. This
Brittle failure work discusses how the European Yield Model (based on a ductile failure mechanism), included in many stan
Ductile failure
dards, combined with a reduction of the effective number of fasteners may provide too conservative results,
Dowel-type fasteners
Parallel-to-grain
which may inadvertently lead to risky situations in which a connection assumed to fail under a ductile mech
Effective number of fasteners anism would actually fail in a brittle manner. Within this paper, a proposal to improve the discrimination ability
Eurocode 5 to correctly predict the failure mode is proposed.
1. Introduction The models dealing with brittle failure are more recent. Early at
tempts date to the 1980’s [8]. Since then, several authors have proposed
Timber construction is steadily growing, as an answer to an different models for brittle failure modes. Their inclusion in standards
increasing demand for sustainability in the building sector. An adequate such as Eurocode 5 [6] (as an informative annex), CSA Standard O86-09
design of the structure is therefore needed to ensure the safety and to [9] or the draft of the New Zealand standard [7] is quite recent and still
optimise the material resources. Connections play an utmost role in on going. An overview and comparison of the most representative
timber engineering. Several studies [1,2] pointed out that 25% of recent models was provided by Cabrero and Yurrita [10].
collapses from timber structures were related to failure of connections. In the case of Eurocode 5 [6], the EYM is combined with the
Timber connections may fail in a ductile or a brittle way, as quali reduction parameter nef [11], which multiplies the capacity of a single
tatively shown in Fig. 1. Since timber breaks in a brittle manner under fastener (obtained by the EYM) by the effective number of fasteners nef ,
bending and tension, the ductility of timber structures is usually pro usually lower than the actual one. This factor, together with the re
vided by connections. Therefore, high ductility values in connections are quirements of minimum spacings between fasteners, intends to prevent
desirable, especially in seismic regions [3]. On one hand, the sequential several brittle failure modes such as splitting (Fig. 2b) or row shear
deformation of a ductile failure (Fig. 1a) allows to identify a possible (Fig. 2c). Block shear (Fig. 2d) was not considered in the prenormative
failure in time to prevent it and contributes to a proper robustness of the version of the Eurocode 5 [12]. It was, however, included in the final
structure [4]. On the other hand, a brittle failure (Fig. 1b) implies a version as an informative (not compulsory) Annex A.
sudden collapse of the structure that may lead to human or material Several test campaigns, considering both connections with large
damage. Despite their relative importance, brittle failure modes are still diameter fasteners i.e. dowels and bolts [13] and small diameter fas
quite unknown, as demonstrated by a survey performed by Working teners i.e., nails, screws, and rivets [14], demonstrated that brittle fail
Group 3 of the Cost Action FP1402 [5], in which 30% of the respondents ure modes can be observed even when the prescribed minimum spacings
(mainly practitioners) were not aware of it. are met. In the case of large diameter fasteners (those that usually fully
The ductile design of connections has been traditionally based on the penetrate the timber member, such as dowels and bolts), Yurrita and
European Yield Model EYM, which assumes the plastic embedment Cabrero [15] used the experimental results from an comprehensive
failure of the wood and the yielding of the fastener (Fig. 2a). This model database to develop a new model dealing with all possible brittle failure
is included in many structural standards, as it is the case of Eurocode 5 modes of connections loaded parallel-to-grain, depicted in Fig. 2, which
[6] or the New Zealand Standard draft [7]. improved the prediction accuracy of the existing models [6,7,16,17].
* Corresponding author.
E-mail addresses: [email protected] (M. Yurrita), [email protected] (J.M. Cabrero).
https://doi.org/10.1016/j.engstruct.2021.112496
Received 15 February 2021; Received in revised form 23 April 2021; Accepted 2 May 2021
Available online 3 June 2021
0141-0296/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Moreover, as the nef takes account of brittle failure modes within a fastener slenderness (a stocky fastener requires a higher applied load to
ductile based model, it does not properly inform the designer on the yield and develop a ductile mechanism), or the joint configuration.
actual expected failure mode of the connection. The use of brittle failure Fig. 3 depicts typical configurations of timber-to-steel connections
models allows to separate the calculation of ductile and brittle failure (similar cases exist for timber-to-timber connections by replacing the
modes, so the designer is able to determine the expected failure mode. steel plate by another timber element). The number of shear planes ns
However, for that purpose, the prediction accuracy of both ductile and (one -Fig. 3a-, a-, two -Fig. 3b and Fig. 3c- or multiple shear planes
brittle models must be accurate enough to avoid false predictions. -Fig. 3d-), or the position of the timber element, as an outer element
The conservative trend resulting from the combination of the EYM (Fig. 3a, Fig. 3b or the outer elements from Fig. 3d) or an inner element
with the reduction factor nef , may imply unwanted situations in which a (Fig. 3c and the inner element from Fig. 3d) also affect to the resulting
brittle failure mode may govern the connection behaviour, although a yielding behaviour of the fastener.
ductile failure mode was predicted (as the ductile capacity was lower).
This paper discusses the possible existence of such situations and 2.1.1. Ductile failure mode of connections
quantifies them. Ductile failure mode is due to the combination of the embedment of
The paper is structured as follows: Section 2 includes a brief the wood and the yielding of the fastener.
description of the analysed type of structural connections, their brittle Different yielding modes may be produced depending on the mate
failure modes, and existing models. Section 3 analyse a database of rial and geometrical properties of the connections. The embedment
ductile tests and demonstrates the conservative trend of Eurocode 5 [6]. strength depends on both the timber density ρ and the fastener diameter
Section 4 presents an exhaustive parametric analysis that compares d, while the yielding of the fastener is related to its yielding moment My ,
Eurocode 5 [6] ductile model with the brittle model from Yurrita and defined by the steel yield strength fy and the fastener diameter d. Those
Cabrero [15] and identifies those cases where a brittle failure mode is factors, in combination with the thickness of the timber elements and
achieved even if a ductile one is expected. Finally, in Section 5, an in steel plates and the joint configuration, determine the number of plastic
crease of the load-carrying capacity predicted by Eurocode 5 [6] is hinges per shear plane in the connection, as depicted in Fig. 4.
suggested to minimise the risks of reaching an unexpected brittle failure Several authors [18,19] have developed analogue formulae for
mode. multiple shear planes connections, which are not explicitly included
within Eurocode 5 [6].
2. State of the art The EYM load-carrying capacity of fasteners submitted to large de
formations (plastic hinges) may be increased by additionally considering
2.1. Failure modes of timber connections the rope effect, which is produced by the axial capacity of fasteners
submitted to lateral loads. This extra contribution is limited to 25% and
As stated before, timber connections with dowel-type fasteners 0% of the calculated load-carrying capacity for bolts and dowels,
loaded parallel-to-grain may reach a brittle or ductile failure mode. The respectively.
resulting failure mode depends on the material and geometrical prop
erties of the connection. 2.1.2. Brittle failure mode of connections
A ductile failure mode implies that the yielding of the fastener occurs Brittle failure mode is due to a crack of the wood usually taking place
before cracking of the wood, while in a brittle failure mode, the failure of before the steel fasteners reach their plastic range, leading to a sudden
the wood happens before the plastic response of the fasteners is reached. collapse of the connection. The most common brittle failure modes of
Typical brittle failure modes are shown in Fig. 2. Of course, the material connections with large diameter fasteners loaded parallel-to-grain
properties of the structural elements play an active role on the governing (Fig. 2) are:
failure mode. But apart from them, geometrical parameters of the
connection may also become relevant, as the spacings between fasteners • Splitting (Fig. 2b): it is formed by a longitudinal crack along the row
(large spacings increase the chances of a ductile failure mode), the of fasteners due to tensile stresses perpendicular-to-grain. This
2
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 2. Possible failure modes of a timber connection with dowel-type fasteners: embedment (right hand-side, a) is the only ductile failure mode, the rest are brittle.
failure mode affects locally in a single row of fasteners and, there the described failure modes (both brittle and ductile) and are all studied
fore, it may not be the cause of the final global failure of a connection together in a same calculation process. They also consider the possible
with multiple rows. interaction between stresses related to different failure modes.
• Row shear (Fig. 2c): shear stresses generate two parallel cracks along The model proposed by Yurrita and Cabrero [15] also considers all
each row of fasteners of the connection. failure modes separately, except splitting (which was additionally
• Block shear (Fig. 2d): it is generated by the tear out of the loaded studied by Yurrita and Cabrero [23]). The effective thickness, defined by
timber area, defined by the connection perimeter. Two lateral shear Yurrita and Cabrero [13], was used both for the models for large
cracks along the exterior rows of fasteners and a head tensile crack diameter fasteners [15,23,24] and small diameter fasteners [14,25]. An
along the first column of fasteners are combined in this case. additional modification for brittle failure mode in connections with
• Net tension (Fig. 2e): it is defined by a crack generated on the net multiple shear planes was proposed by Yurrita et al. [24].
cross-sectional area at the beginning of the connection. In the present work, the model of Yurrita and Cabrero [15] has been
taken as the reference brittle model, since it was demonstrated to pro
There are different proposals dealing with brittle failure modes. vide the most accurate results. For a further insight about the model and
Eurocode 5 [6] only considers block shear in its informative Annex A its accuracy, the reader is referred to Yurrita and Cabrero [15].
directly. As already stated, row shear and splitting are implicitly The model considers the failure modes depicted in Fig. 5 by defining
considered by combining the EYM with the reduction factor nef and by the load carrying capacities of the involved failure planes (lateral shear
respecting some spacing limitations. planes L for row shear -Fig. 5a-; head tensile plane H for net tension
The parameter nef reduces the number of effective fasteners per row -Fig. 5c and both planes for block shear -Fig. 5b).
of fasteners. This factor is derived from the work by Jorissen [11]. It is
used both to reduce the design capacity of the connection to prevent 3. Analysis of the prediction accuracy of Eurocode 5 for ductile
from some brittle failure modes such as row shear and splitting and, at failure mode
the same time, to consider the uneven distribution of the load between
fasteners. This uneven distribution is redistributed when the fasteners 3.1. Database
yield, since they cannot be loaded above their plastic capacity [20]. For
large diameter fasteners such as dowels and bolts, nef is defined as: A database of experimental tests has been used to evaluate the pre
⎧ diction accuracy of the model included in Eurocode 5 [6]. As shown in
n
⎪
⎨ r Table 1, the tests from a total of 7 experimental studies (Ehlbeck and
nef = min √̅̅̅̅̅̅̅̅ (1) Werner [26], Ehlbeck and Werner [27], Jorissen [11], Blaß and Schmid
⎪ n0.9 4 a1
⎩ r
13d [28], Sandhaas [29], Hüner [30] and Misconel et al. [31]) were
considered.
where nr is the number of rows of fasteners, a1 is the spacing between The compiled database focuses on the ductile tests (although some
rows and d is the fastener diameter. researchers reported also some splitting cases), gathering a total of 221
The New Zealand standard draft [7], which can be considered as an configurations (1518 single tests). It comprises both timber-to-steel
evolution of the model included in the Canadian standard CSA Standard (wood-steel-wood, wsw) and timber-to-timber (wood-wood-wood,
O86-09 [9], is based on the model developed by Quenneville and www). Dowels were used as fasteners in 69.2% of the configurations,
Mohammad [21] and Mohammad and Quenneville [22]. It considers all whereas bolts in the rest. Solid wood is the main timber product
the described failure modes (except splitting, as it is assumed that it (65.2%), followed by glulam (25.8%) and LVL (9.0%). Regarding the
cannot lead to the entire failure of a connection with two or more rows type of wood, both softwood and hardwood are similarly represented
of fasteners) in separate calculations. (42.5% and 57.5%, respectively). Finally, almost half of the reported
The proposal from Hanhijärvi and Kevarinmäki [16,17] considers all tests (47.5%) included two or more rows of fasteners, whereas the rest of
3
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 4. Yielding modes of the fastener considered by the EYM for timber-to-steel connections. Similar cases are considered for timber-to-timber connections.
the configurations included only one row of fasteners. Table 1 provides fasteners nef and the rope effect (considered as 0% for dowels and 25%
further information about each of the studied works. for bolts).
The small number of replicates for each configuration in the existing
campaigns does not allow to obtain an accurate characteristic load-
3.2. Comparison between the test results and the predictions obtained carrying capacity and, hence, the validation is performed at the mean
from Eurocode 5 level. The required mean properties are obtained from the characteristic
values following the procedure explained by Jockwer et al. [32] and
As a first step, the test results have been compared with the load- Cabrero et al. [33] which is based on the probabilistic model for timber
carrying capacity predicted by the approach described in Eurocode 5 proposed by the Joint Committee on Structural Safety [34], previously
[6], that is, the combination of the EYM with the number of effective
4
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 5. Loading planes (lateral shear L, and head tensile H) related to each failure mode.
Table 1
Summary of the tests used for the validation of ductile failure.
Author Number of Joint scheme Fastener Type Timber product Timber class No columns
Config. Tests wsw www Bolt Dowel Glulam LVL Solid Softwood Hardwood 1 ⩾2
Total Number 221 1518 70 151 68 153 17 20 144 94 127 116 105
% - - 31.7% 68.3% 30.8% 69.2% 25.8% 9.0% 65.2% 42.5% 57.5% 52.5% 47.5%
used in other works [10,13–15,23–25,35]. and mean values close to 1 (around 1.05) are obtained. The joint
The boxplot graphic in Fig. 6 separately analyses the prediction ac configuration, timber product, type and size of fasteners used, or the
curacy obtained by Eurocode 5 [6] when compared with the tests of each number of rows have been considered as possible reasons. However, no
of the test campaigns (due to the small amount of tests from Hübner clear explanation for this dissimilar trend has been found. In the box
[30], it has been considered together with those from Misconel et al. plots in Fig. 6, two different series considering all the test together, and
[31]). The ratio between the predicted load capacity Fp and the tested without the ones from Jorissen [11] are given.
load capacity Ft is used as the comparison parameter, and the ideal ratio Fig. 7a studies the tests altogether, plotting the load-carrying ca
Fp /Ft = 1 is given by a vertical dashed line. pacity from tests in the abscissa axis, and the predicted values in the
The results show a consistent conservative trend on the prediction of ordinates axis. A dashed line depicts the ideal slope m = 1. The obtained
the load-carrying capacity. Average and median values from the analysis slope m = 0.564 of the fitting line confirms the conservative trend of
of the tests performed by Ehlbeck and Werner [26], Ehlbeck and Werner Eurocode 5 [6]. However, the high coefficient of correlation R2 = 0.917,
[27], Sandhaas [29], Hübner [30] and Misconel et al. [31] are between demonstrates that, despite this conservative trend, the model is quite
0.53 and 0.67. Slightly less conservative values (around 0.80) are ob consistent. Fig. 7b is a zoom of the former graphic, taking into account
tained in the case of Blaß and Schmid [28]. the range of tests with a load-carrying capacity within a range between
The only case with a different trend is Jorissen [11], where average 0 and 300 kN. The tests from Jorissen [11] are plotted differently, so
Fig. 6. Boxplot assessing the accuracy obtained by the Eurocode 5 [6] when compared with the test results from the studied authors, considering the accuracy of the
ratio between the predicted failure load Fp and the tested failure load Ft .
5
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 7. Comparison between the load capacity values obtained from the tests Ft and the corresponding theoretical values Fp predicted by the ductile approach from
the Eurocode 5 [6], and when applying the correction factor of 1.55.
6
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
some cases of brittle failure mode, like splitting or row shear. Other The studied parameters are: timber member thickness t (Fig. 9),
brittle failure modes such as block shear are not included. As a conse fastener diameter d (Fig. 10), spacing between columns of fasteners a1
quence, this model not only obtains too conservative results, it does not (Fig. 11), spacing between rows of fasteners a2 (Fig. 12) and distance to
include all brittle failure modes; and it additionally hinders the possi the loaded edge a3 (Fig. 13). The distance a4 has not been considered as
bility to determine whether a connection will fail in a brittle or in a its influence will affect mainly to net tension failure, which is calculated
ductile way. exactly in the same way in Eurocode 5 [6] and in the brittle model from
A parametric study has been conducted to study the effect of the Yurrita and Cabrero [15].
observed conservative trend when determining the failure mode. The For each case, results from the four possible combinations between
approach from Eurocode 5 [6] with its corrected version after applying the two fastener steel grades and timber products are provided. The
the factor of 1.55 derived from the analysis in Section 3, in combination combination of timber C24 with the steel grade 6.8 (Fig. 9a, Fig. 10a,
with the brittle failure model proposed by Yurrita and Cabrero [15] are Fig. 11a, Fig. 12a, Fig. 13a) is a good example of a low profile connec
also analysed. tion. Just in the opposite side, a high strength connection is obtained by
combining beech LVL80S and steel grade 12.9 (Fig. 9d, Fig. 10d,
4.1. Analysis of one connection Fig. 11d, Fig. 12d, Fig. 13d). In between, a case of softwood with high
steel grade -C24 + steel 12.9- (Fig. 9b, Fig. 10b, Fig. 11b, Fig. 12b,
4.1.1. Geometry and materials of the connection Fig. 13 and a hardwood with a mild steel grade -beech LVL80S + steel
Before presenting the overall results of the conducted analysis, an 6.8- (Fig. 9d, Fig. 10d, Fig. 11d, Fig. 12d, Fig. 13d) are also considered.
example of one of the studied connections is herein presented in Fig. 8. Since the former analysis of the tests considered the mean level of the
The considered case is a steel-wood-steel sws connection with a timber material properties, the same values are applied in the parametric
thickness t = 100 mm, fasteners of diameter d = 12 mm and steel plates analysis.
of tp = 12 mm (considered as thick plates according to Eurocode 5 [6]). In every graph, the variation of the studied parameter is shown in the
The connection includes a total of 12 fasteners distributed in a 3x5 mesh abscissa axis. In the case of distances between fasteners (a1 , a2 and a3 )
(number of rows nr = 3 and number of columns nc = 5). The basic the ratio between the parameter and the fastener diameter d is plotted
spacing distances (a1 , a2 , a3 and a4 are taken always as the minimum instead of the absolute distance. The predicted load Fp is plotted in the
prescribed in Eurocode 5 [6] to guarantee a ductile failure mode ordinate axis. The values from three models are plotted: the brittle
(smaller spacings are assumed to increase the risk of a brittle response). model from Yurrita and Cabrero [15], the ductile approach from Euro
Two different steel grades (6.8 and 12.9) for the fasteners and two code 5 [6], and the corrected version of the former one by applying the
timber products (C24 [36] and beech LVL80S [37]) were used. coefficient of 1.55 obtained in Section 3. Vertical dashed lines are used
to limit the intersection points between the brittle behaviour according
4.1.2. Parametric analysis of the connection example to Yurrita and Cabrero [15] and Eurocode 5 [6] with the correction
The parametric analysis studies the impact of varying several pa factor. Additionally a vertical continuous line is plotted as a reference of
rameters on connections as the one depicted in Fig. 8 which has been the spacing limits a1 = 5d, a2 = 3d and a3 = 7d provided by Eurocode 5
used as the basis geometry. [6].
7
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 9. Parametric analysis of the influence of the timber thickness t on the predicted load capacity Fp considering the brittle model from Yurrita and Cabrero [15],
the model from Eurocode 5 [6] and its corrected version according to Section 3.
The gray-filled areas of the graphs correspond to those where the Fig. 9. A range within 50 and 250 mm has been considered in the
brittle model [15] reaches a lower load-carrying capacity than the cor analysis, as the most representative of the thicknesses applied in timber
rected load capacity of Eurocode 5 [6], assumed as an accurate model engineering.
for ductile failure mode. The values from the models from Eurocode 5 In general, all cases with low timber thickness are expected to reach
[6] include the number of effective fasteners and, therefore, do not brittle failure mode. The limit thickness values depend mostly on the
discriminate between ductile and brittle failure modes. The configura steel grade of the fastener. When the lower steel grade 4.8 is considered,
tions in which brittle failure mode may govern, without the designer the limit stays in between 104 mm for C24 timber (Fig. 9 and 83 mm for
being aware, are represented by these shaded parts of the graphs. LVL80S (Fig. 9c). When the steel grade 12.9 is used, the thickness limit is
It may be observed how the current model in Eurocode 5 [6], increased up to 186 mm for C24 (Fig. 9 and 118 mm for beech LVL80S
including the nef parameter, consistently obtains lower load-carrying (Fig. 9d) In those timber thickness in which ductile failure mode is
capacities than the brittle model. As a result, the previously discussed guaranteed, the yielding mode of the fastener corresponds to two hinges.
conservative trend of Eurocode 5 [6] assumes a ductile prediction even Analysis of the influence of the fastener diameter d.
when brittle failure mode may occur. The analysis of the impact of the fastener diameter d is given in
Analysis of the influence of the thickness of the timber member t. Fig. 10. The range of the studied diameters matches with the existing
The influence of the thickness of the timber member is shown in diameters of dowels (between 6 and 30 mm).
8
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 10. Parametric analysis of the influence of the fastener diameter d on the predicted load capacity Fp considering the brittle model from Yurrita and Cabrero [15],
the model from Eurocode 5 [6] and its corrected version according to Section 3.
As shown, unexpected brittle failure modes (gray-shaded areas, those d = 5 to a1 /d = 10 is considered. The limit a1 /d = 5 established by
where the current Eurocode 5 [6] model conservatively predicts a Eurocode 5 [6] is given by a vertical black line.
ductile capacity where brittle failure mode may govern) may happen in A similar trend as above may be observed: almost no problems
the whole analysed range, as brittle failure modes may happen for both respecting the limitation of a1 /d = 5 are observed in the material
small and large diameters. Again, the cases with the high steel grade combinations with steel 6.8 (Fig. 11c and Fig. 11a, although in this latter
12.9 are riskier, with almost the whole part of the graphs shaded in gray case a distance of a1 /d = 6.2 would be required to obtain ductile failure
(Fig. 10b and Fig. 10d). The opposite trend is found when the steel grade mode). In contrast, a much higher spacing distance a1 is required to
6.8 is combined with LVL80S (Fig. 10c). achieve a ductile behaviour for the cases with steel grade 12.9 (Fig. 11d
Analysis of the influence of the spacing between columns of fasteners and Fig. 11b).
a1 . Analysis of the influence of the spacing between rows of fasteners a2 .
The impact of the variation in the spacing between columns of fas The study of the spacing between rows of fasteners a2 is given in
teners is given in Fig. 11. The variation plotted in the graphs is from a1 / Fig. 12. The same range of values as in the case of a1 is used, although
d = 1 to a1 /d = 10, although for the main analysis only the range a1 / this time, the limit established by Eurocode 5 [6] for dowels is a1 /d = 3.
9
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 11. Parametric analysis of the influence of the spacing between columns of fasteners a1 on the predicted load capacity Fp considering the brittle model from
Yurrita and Cabrero [15], the model from Eurocode 5 [6] and its corrected version according to Section 3.
In this case, very few cases are expected to reach brittle failure mode. mode are found when 12.9 steel is chosen as the timber product
Almost no cases under the limitation established by Eurocode 5 [6] are (Fig. 13b and Fig. 13d. In the case of C24 with 6.8 steel (Fig. 13a), a
found when using the steel grade 6.8 (Fig. 12c and Fig. 12a). In the case distance slightly higher than a3 /d = 9 is needed to achieve ductile
of the high grade steel 12.9, the limitation of Eurocode 5 [6] seems not behaviour. Finally, another time the combination of LVL80S with steel
enough. A distance a1 /d = 3.6 is required when combined with LVL80S 6.8 (Fig. 13c) appears as the safest combination, with no brittle cases.
((Fig. 12d), which increases to a1 /d = 4.15 when timber C24 is used
(Fig. 12b).
Analysis of the influence of distance to the loaded end edge a3 . 4.2. General parametric analysis
The last analysed parameter is the distance from the last column of
fasteners to the end-loaded edge of the connection a3 (see Fig. 13). The Once the methodology of the parametric study has been explained
studied range is in between a3 /d = 1 and a3 /d = 15, considering also with the previous example, the results from the general parametric
the limit given in Eurocode 5 [6] of a3 /d = 7. analysis, which considers a total of 1, 008 connections, are presented.
Another time, problems with the correct discrimination of the failure The analysis includes 21 different geometries of fastener distributions,
from a basic connection of 2 × 2 (nr × nc ) to 8 × 8 connection. Each
10
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 12. Parametric analysis of the influence of the spacing between rows of fasteners a2 on the predicted load capacity Fp considering the brittle model from Yurrita
and Cabrero [15], the model from Eurocode 5 [6] and its corrected version according to Section 3.
pattern has been studied for the cases of steel-wood-steel and wood- [15], a brittle failure mode is expected.
steel-wood connections, both assessed with dowels and with bolts Each graph presents four cases: steel-wood-steel connections with
(considering different rope effect for each fastener type, as established dowels (Fig. 14a, Fig. 15a and Fig. 16a); steel-wood-steel connections
by Eurocode 5 [6]). Finally, three timber thickness (100, 150 and 200 with bolts (Fig. 14b, Fig. 15b and Fig. 16b); wood-steel-wood connec
mm) were considered. tions with dowels (Fig. 14c, Fig. 15c and Fig. 16c) and wood-steel-wood
For each connection, the same process described above has been connections with bolts (Fig. 14d, Fig. 15d and Fig. 16d). In each case,
conducted, studying the four possible combinations of timber product four values are given, corresponding to the four material combinations
and steel grade of fasteners, and the variation of the already explained (timber C14 with steel grades 4.8 and 10.9 and LVL80S with the same
parameters. steel grades).
The following graphs (Fig. 14, Fig. 15 and Fig. 16 for timber thick Several conclusions may be obtained from the parametric analysis:
ness of 100, 150 and 200 mm) present the % of the studied cases in
which, after comparing the approach from Eurocode 5 [6] (with the • Joint configuration. The comparison between all studied cases of
correction factor of 1.55) and the brittle model from Yurrita and Cabrero steel-wood-steel sws and wood-steel-wood wsw connections shows
11
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 13. Parametric analysis of the influence of the distance to the end loaded edge of the timber element a3 on the predicted load capacity Fp considering the brittle
model from Yurrita and Cabrero [15], the model from Eurocode 5 [6] and its corrected version according to Section 3.
that in all cases, the % of wrong predictions are higher in sws con hardwood is combined with a low steel grade, as plastic hinges in the
nections (17.5%) than in the wsw case, where the predictions are fasteners are more likely. Therefore, in the combination of LVL80S
very accurate (1.9%). and 6.8 steel grade only 2.6% of brittle cases are expected.
• Timber thickness. The possibility of obtaining a wrong prediction of • Influence of the number of fasteners. Those connections with a
the failure mode decreases when the timber thickness (or the fastener reduced number of fasteners have a lower probability of reaching a
slenderness) is increased: 14.7% for t = 100 mm; 7.5% for t = 150 brittle failure mode (i.e., the configuration with a 2x2 mesh presents
mm and 4.3% for t = 200 mm. The rate of ductile cases relates to the 0% of brittle cases). The rate increases when the number of fasteners
yielding mode of the fastener. is higher: connections with 8x8 fasteners have, conversely, an
• Used materials. The most risky case is clearly the combination of C24 average rate of 38.4% of being wrongly predicted.
timber with 12.9 steel (17.3% of wrong cases). The combination of • Influence of the type of fastener. The rope effect considered by
softwood with a high steel grade makes it easier to produce a wood Eurocode 5 [6] for bolts is limited to 25% of the EYM capacity, and it
crack before the fastener yields. Just the opposite happens when a is dismissed for dowels. As shown in Fig. 17, the resulting variation is
12
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 14. Parametric study considering the % of brittle cases non predicted by the Eurocode 5 [6], considering a timber thickness t = 100 mm. Last column provides
the mean value for all the analysed cases.
insignificant: 9.7% of brittle cases for dowels and 7.9% for bolts. The 5. Proposal
slightly better performance of connections with bolts relates to the
increase in the ductile load-carrying capacity due to the consider As shown in Section 4, the conservative trend in the load capacity
ation of the rope effect. It has been found that mode H (which in prediction of Eurocode 5 [6] may lead to wrong predictions of the failure
cludes rope effect, Fig. 4h) was the most common in wsw mode. The used correction factor of 1.55 has been only developed for
configurations, while in the case of sws configurations, the resulting illustrative purposes in the parametric analysis, but being dependent on
yielding modes were more related to the slenderness of the fastener the compiled database, it should not be regarded as a general proposal.
t/d. An alternative way to reduce the observed conservative trend of
13
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 15. Parametric study considering the % of brittle cases non predicted by the Eurocode 5 [6], considering a timber thickness t = 150 mm. Last column provides
the mean value for all the analysed cases.
Eurocode 5 [6] could be not including the effective number of fasteners it may be considered that when the ductile load-carrying capacity is
nef . This reduction parameter was introduced to reduce the predicted reached, the initially uneven load distribution among fasteners has been
load-carrying capacity of the connection to include some of the brittle redistributed [20].
failure modes, but it also leads to an inaccurate prediction of the load- A boxplot, similar to the one shown in Fig. 6, but in which now
carrying capacity and, at the same time, it does not allow a correct Eurocode 5 [6] without the nef parameter is evaluated, is presented in
prediction of the failure mode of a connection. Fig. 18. It can be seen how the prediction accuracy obtained for the
The EYM being a plastic model, based on the yielding of the fastener, considered test campaigns (excluding the one from Jorissen [11]) is
14
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 16. Parametric study considering the % of brittle cases non predicted by the Eurocode 5 [6], considering a timber thickness t = 200 mm. Last column provides
the mean value for all the analysed cases.
improved, with average and median values closer to the ideal ratio correction factor of 1.55 (required to obtain a perfect slope m = 1 in
Fp /Ft = 1 (in between 0.74 from the tests of Ehlbeck and Werner [27] Fig. 7c) is reduced to 1.13.
and 1.04 of Blaß and Schmid [28]). The study of ductile and brittle failure modes separately can lead to a
As shown in Fig. 19, the overall prediction accuracy of Eurocode 5 more accurate prediction of the load-carrying capacity and, at the same
[6] without including the reduction factor nef improves. The slope time, the actual failure mode of the connection could be assessed. To
value increases to m = 0.838, while the correlation coefficient R2 evaluate the possible improvement, a similar parametric analysis as the
remains almost constant to the values previously given in Fig. 7a one performed in Section 4 has been conducted. Now, the model in
(from 0.917 to 0.909). As a consequence, if used, the former cludes the brittle failure approach from Yurrita and Cabrero [15] and
15
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 17. Parametric study considering the differences on the % of brittle cases non predicted by the Eurocode 5 [6] when using dowels or bolts. Last column provides
the mean value for all the analysed cases.
Fig. 18. Boxplot assessing the accuracy obtained by the Eurocode 5 [6] without the nef parameter when compared with the test results from the studied authirs,
considering the accuracy of the predicted ratio between the predicted failure load Fp and the tested failure load Ft .
the ductile approach of the EYM without the nef parameter, with the original value of 17.5% (Fig. 20a) is reduced to 6.6% (Fig. 20b). In
correction factor of 1.13. Fig. 20 compares the new obtained prediction addition, the increasing tendency related to connections with higher
accuracy of the new parametric analysis with the former one. Steel- number of fasteners is minimised, as the nef penalises specially those
wood-steel sws and wood-steel-wood wsw connections are evaluated cases.
separately. The three considered timber thickness (100, 150 and 200 Regarding wood-steel-wood connections, the average behaviour
mm) are plotted for each of the 21 different fastener distributions. remains almost similar. Eurocode 5 [6] (Fig. 20c) obtains 1.9%, while
In the case of steel-wood-steel connections, a clear reduction of the proposal (Fig. 20d) slightly increases the value to 2.2%. However,
the cases reaching brittle failure mode can be noticed. In average, the the same tendency of increasing the risk of reaching a brittle failure
16
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 19. Comparison between the load capacity values obtained from the tests Ft and the corresponding theoretical values Fp predicted by the ductile approach from
the Eurocode 5 [6] without the nef parameter before and after applying a correction factor of 1.13.
mode on connections with more fasteners is noticed. conservative trend, leads usually to a ductile failure mode prediction.
As a summary, Table 2 compares the results of the analysis, 70.7% of the cases are expected to reach a ductile failure mode, while
showing the % of brittle cases expected when applying both models only 39% of the configurations failed under this failure mode. A total of
(EYM combined with the nef and the proposal that dismiss this 33.3% false ductile cases are predicted.
reduction factor). The table includes all cases considered in the In contrast, the proposal reduces the number of ductile predictions to
already conducted parametric analysis. Moreover, the difference of 38.1%, with only 4.5% of false ductile predictions. Furthermore a good
% between both models and the ratio of improvement are provided. prediction of brittle cases (56.4% of true brittle and only 5.5% of false
The table divides the analysis regarding several classifications (type brittle) is achieved.
of connection, timber thickness, material combination, and type of In total, the original model from Eurocode 5 [6] (with nef ) gets 65.0%
fastener). It can be seen how the reached improvement is higher in of positive matches. The proposal clearly improves this trend with a total
the cases where brittle failure mode is more likely to happen such as of 90.0% of correct predictions.
with lower thicknesses or the combination of softwood with high
steel grade. In total, a reduction of 53.6% of the brittle cases is 6. Conclusions
achieved.
The ductile model included on Eurocode 5 [6] has been demon
strated to be consistent but, at the same time, conservative. Based on the
5.1. Comparison of the discrimination ability
results of analysis of the database of experimental tests, an illustrative
correction factor of 1.55 would lead, in average, to a more accurate
Finally, an analysis comparing the discrimination ability to
model.
determine the failure mode (ductile or brittle) between the model
A parametric study comparing both the ductile model of Eurocode 5
from Eurocode 5 [6] and the proposal has been conducted. Since the
[6] and the corrected one with the brittle model proposed by Yurrita and
combination of the EYM and the nef from Eurocode 5 [6] does not
Cabrero [15] has been performed. This study demonstrates that in many
allow to strictly determine the failure mode, this analysis has been
cases a brittle failure mode may be achieved when ductile failure mode
performed under the hypothesis that this model predicts a ductile
is expected. This can lead to risky situations, specially in seismic areas or
failure mode.
structures subjected to other accidental loads. Therefore, when looking
For the analysis, the brittle database ([21,22,24,31,38–49]) used by
at the resulting failure mode, the conservative trend of Eurocode 5 [6],
Yurrita and Cabrero [15] to validate the brittle model has been used
instead of being on the side of safety, may be unsafe, due to the wrong
together with the ductile database ([11,26–31]) described in this work
estimation of the failure mode.
(Table 1). This database includes a total of 420 configurations, 61.0%
To improve the model from Eurocode 5 [6], a proposal in which
brittle and 39.0% ductile. The load capacity of each test configuration
ductile (the model from Eurocode 5 [6] but without the nef param
predicted by Eurocode 5 [6] original model and the proposal without the
eter), and brittle (according to Yurrita and Cabrero [15]) failure
nef parameter were compared with the brittle predictions from Yurrita
modes are considered separately, is presented. The analysis demon
and Cabrero [15]. The expected failure mode (ductile or brittle), cor
strates how this combination of dedicated models allows to achieve
responds to the lower capacity, which it is compared with the experi
more accurate predictions of the ductile load-carrying capacity and it
mentally observed failure mode.
also improves the capacity to correctly identify the expected failure
Fig. 21 depicts the result of this analysis. The dark filled parts
mode of the studied connection.
correspond to the correct predictions (true ductile and true brittle),
The resulting improvement is confirmed by means of an extensive
while the clear filled parts show the % of wrong predictions (false ductile
database of brittle and ductile tests, which has been used to evaluate
and false brittle).
the discrimination ability between the failure modes of the original
It is clear how the original model from Eurocode 5 [6], due to its
17
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 20. Parametric study considering the % of brittle cases non predicted by the Eurocode 5 [6] (including nef ) and the proposal (without nef ). Last column provides
the mean value for all the analysed cases.
Table 2
Summary of the % of wrong predicted cases. Comparison between the Eurocode 5 [6] (with nef ) and the proposal (without nef ). All values are given in %.
Configuration Thickness t [mm] Material combination Fastener Total
sws wsw 100 150 200 C24 & 6.8 C24 & 12.9 LVL & 6.8 LVL & 12.9 Dowel Bolt Total
Eurocode 5 [6] 17.5 1.9 14.7 7.5 4.3 6.8 17.3 2.6 8.7 9.7 7.9 9.7
Proposal 6.6 2.2 7.8 5.1 3.9 4.2 7.3 1.9 4.6 4.5 4.5 4.5
Difference 10.9 − 0.3 6.9 2.4 0.4 2.6 10 0.7 4.1 5.2 3.4 5.2
Improvement 62.3 − 15.8 46.9 32 9.3 38.2 57.8 26.9 47.1 53.6 43 53.6
18
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
Fig. 21. Discrimination ability. Comparison between Eurocode 5 [6] (assuming that the EYM predicts a ductile failure), and the proposal.
model and the proposal. This analysis determines that Eurocode 5 [6] [15] Yurrita M, Cabrero JM. New design model for brittle failure in the parallel-to-grain
direction of timber connections with large diameter fasteners. Eng Struct 2020;
reaches a total of 65.0% of correct predictions, wile the proposal
217:110557. https://doi.org/10.1016/j.conbuildmat.2019.04.100.
increases the percentage of positive matches to 90.0%. [16] Hanhijärvi A, Kevarinmäki A. Design method against timber failure mechanisms of
dowelled steel-to-timber connections. In: CIB-W18 Timber Structures, Bled,
Declaration of Competing Interest Slovenia, Paper 40-7-3, 2007.
[17] Hanhijärvi A, Kevarinmäki A. VTT PUBLICATIONS 677: Timber Failure
Mechanisms in High-Capacity Dowelled Connections of Timber to Steel, VTT,
The authors declare that they have no known competing financial Espoo, ISBN 9789513870904, 2008a.
interests or personal relationships that could have appeared to influence [18] Pedersen MU. Dowel Type Timber Connections, Ph.D. thesis, Danmarks Tekniske
Universitet, 2002.
the work reported in this paper. [19] Sawata K, Sasaki T, Kanetaka S. Estimation of shear strength of dowel-type timber
connections with multiple slotted-in steel plates by European yield theory. J Wood
Acknowledgements Sci 2006;52(6):496–502. https://doi.org/10.1007/s10086-006-0800-9. ISSN
14350211.
[20] Blaß HJ, Sandhaas C. Timber Engineering - Principles for Design, Chapter 3, KIT
The research has been performed thanks to the COST Action FP1402. Scientific Publishing, Karlsruhe, ISBN 9783731506737, 2017, doi:10.5445/KSP/
The first author is supported by a PhD fellowship from the Programa de 1000069616.
[21] Quenneville P, Mohammad M. On the failure modes and strength of steel-wood-
Becas FPU del Ministerio de Educación y Ciencia (Spain) under the Grant steel bolted timber connections loaded parallel to grain. Can J Civ Eng 2000;27(4):
No. FPU15/03413. He would also like to thank the Asociación de 761–73.
Amigos of the University of Navarra for their help with a fellowship in [22] Mohammad M, Quenneville P. Bolted wood-steel and wood-steel-wood
connections: verification of a new design approach. Can J Civil Eng 2001;28(2):
early stages of this research.
254–63. https://doi.org/10.1139/l00-105. ISSN 0315-1468.
[23] Yurrita M, Cabrero JM. New design model for splitting in timber connections with
References one row of fasteners loaded in the parallel-to-grain direction. Eng Struct 2020;223:
111155. https://doi.org/10.1016/j.engstruct.2020.111155.
[1] Frühwald E, Serrano E, Toratti T, Emilsson A, Thelandersson S. Design of safe [24] Yurrita M, Cabrero JM, Quenneville P. Brittle failure in the parallel-to-grain
timber structures – How can we learn from structural failures in concrete, steel and direction of multiple shear softwood timber connections with slotted-in steel plates
timber? Design of safe timber structures – Report TVBK-3053, Tech. Rep.. Lund and dowel-type fasteners. Constr Build Mater 2019;216:296–313. https://doi.org/
University; 2007. 10.1016/j.conbuildmat.2019.04.100.
[2] Frühwald E. Analysis of structural failures in timber structures: Typical causes for [25] Yurrita M, Cabrero JM. Experimental analysis of plug shear failure in timber
failure and failure modes. Eng Struct 2011;33(11):2978–82. ISSN 01410296. connections with small diameter fasteners loaded parallel-to-grain. Eng Struct
[3] Jorissen A, Fragiacomo M. General notes on ductility in timber structures. Eng 2021;238:111766. https://doi.org/10.1016/j.engstruct.2020.111766.
Struct 2011;33(11):2987–97. https://doi.org/10.1016/j.engstruct.2011.07.024. [26] Ehlbeck J, Werner H. Tragfähigkeit von Laubholzverbindungen mit stabförmigen
ISSN 0141-0296. Verbindungsmitteln. Tech. Rep., Research report, Versuchsanstalt für Stahl, Holz
[4] Branco JM, Neves LA. Robustness of timber structures in seismic areas. Eng Struct und Steine, Karlsruhe University, 1989.
2011;33(11):3099–105. https://doi.org/10.1016/j.engstruct.2011.02.026. ISSN [27] Ehlbeck J, Werner H. Softwood and Hardwood Embedding Strength for Dowel-type
01410296. Fasteners, in: CIBW-18, Paper 25-7-2, 1992.
[5] Stepinac M, Cabrero JM, Ranasinghe K, Kleiber M. Proposal for reorganization of [28] Blaß HJ, Schmid M. Spaltgefahr von Nadelhölzern, Ph.D. thesis, TH Karlsruhe,
the connections chapter of Eurocode 5. Eng Struct 2018;170(May):135–45. https:// Karlsruhe, Germany, 2002.
doi.org/10.1016/j.engstruct.2018.05.058. ISSN 18737323. [29] Sandhaas C. Mechanical Behaviour of Timber Joints With Slotted-in Steel Plates,
[6] Eurocode 5. CEN:EN 1995-1-1:2004 - Eurocode 5: Design of timber structures - Part Ph.D. thesis, TU Delft, 2012.
1–1: General - Common rules and rules for buildings, Comité Européen de [30] Hübner U. Mechanische Kenngrößen von Buchen, Eschen und Robinienholz für
Normalisation (CTN), 2004. lastabtragende Bauteile, Ph.D. thesis, Technische Universität Graz, 2013.
[7] Quenneville P, Zarnani P. Proposal for the Connection Chapter of the New Zealand [31] Misconel M, Ballerini M, de Kuilen, V, J W G. Steel-to-timber joints of beech-lvl
Design of Timber Structures, Unpublished, 2017. with very high strength steel dowels. In: Proceedings of the World Conference on
[8] Smith I, Steck G. Influence of number of rows of fasteners or connectors upon the Timber Engineering, WCTE, Vienna, 269-276, ISBN 978-3-903024-35-9, 2016.
ultimate capacity of axially loaded timber joints. In: CIB W-18 Timber Structures, [32] Jockwer R, Fink G, Köhler J. Assessment of the failure behaviour and reliability of
Beit Oren, Israel, Paper 18-7-3, 1985. timber connections with multiple dowel-type fasteners. Eng Struct 2018;172:
[9] CSA Standard O86-09. Engineering design in wood. Canadian Standards 76–84. https://doi.org/10.1016/j.engstruct.2018.05.081.
Association, 2009. [33] Cabrero JM, Honfi D, Jockwer R, Yurrita M. A probabilistic study of brittle failure
[10] Cabrero J, Yurrita M. Performance assessment of existing models to predict brittle in dowel-type timber connections with steel plates loaded parallel to the grain.
failure modes of steel-to-timber connections loaded parallel-to-grain with dowel- Wood Mater Sci Eng 2019;14(5):298–311. https://doi.org/10.1080/
type fasteners. Eng Struct 2018;171:895–910. https://doi.org/10.1016/j. 17480272.2019.1645206.
engstruct.2018.03.037. [34] Joint Committee on Structural Safety (Ed.), Probabilistic Model Code, JCSS, 2006.
[11] Jorissen AJ. Double shear timber connections with dowel type fasteners, Ph.D. [35] Yurrita M, Cabrero JM. New criteria for the determination of the parallel-to-grain
thesis, TU Delft, 1998. embedment strength of wood. Constr Build Mater 2018;173:238–50. https://doi.
[12] Eurocode 5. ENV 1995-1-1:1993. Eurocode 5, Design of Timber Structures, Part org/10.1016/j.conbuildmat.2018.03.127.
1–1: General Rules and Rules for Buildings., Comité Européen de Normalisation [36] EN338. Structural timber - Strength classes, CTN 56, 2016.
(CTN), 1993. [37] E. 13183-1. Timber structures - Structural laminated veneer lumber -
[13] Yurrita M, Cabrero JM. Effective thickness of dowel-type fasteners at the elastic Requirements, Comité Européen de Normalisation, 2004.
range for steel-to-timber connections under brittle failure mode in the parallel-to- [38] Massé D, Salinas J, Turnbull J. Lateral strength and stiffness of single and multiple
grain direction: A new method based on a beam on elastic foundation. Eng Struct bolts in glue laminated timber loaded parallel to grain. Tech. Rep., Engineering and
2020;209:109959. https://doi.org/10.1016/j.engstruct.2019.109959. Statistical Research Centre, Research Branch, Agriculture, Ottawa, Canada, 1998.
[14] Yurrita M, Cabrero JM, Moreno-Zapata E. Brittle failure in the parallel-to-grain [39] Mishler A. Bedeutung der Duktilität für das Tragverhalten von Stahl-Holz-
direction of timber connections with small diameter dowel-type fasteners: A new Bolzenverbindungen, Ph.D. thesis, ETH Zürich, 1998.
design model for plug shear. Eng Struct 2021;241:112450. https://doi.org/ [40] Anderson GT. Experimental investigation of group action factor for bolted wood
10.1016/j.engstruct.2021.112450. connections. Master’s thesis. Virginia Polytechnic Institute and State University;
2001.
19
M. Yurrita and J.M. Cabrero Engineering Structures 242 (2021) 112496
[41] Dodson MA. The effects of row spacing and bolt spacing in 6-bolt and 4-bolt wood- [46] VTT. Monileikkeisten tappivaarnaliitosten kuormituskokeet (Loading Tests of
to-steel connections. Master’s thesis. Washington State University; 2003. Multiple Shear Dowelled Connections). Research report no RTE 1583/03. Ordered
[42] Reid MS. Predictions of two-row multi-bolted connections resistance subjected to by Wood Focus Oy, Tech. Rep., Technical Research Center of Finland, 2003.
parallel-to-grain loading. Master’s thesis. Ontario: Royal Military College of [47] Hanhijärvi A, Kevarinmäki A. Timber failure mechanisms in high-capacity
Canada; 2004. dowelled connections of timber to steel. Experimental results and design. VTT
[43] Sjödin J, Johansson C-J. Influence of initial moisture induced stresses in multiple Publications 677, Tech. Rep., Technical Research Center of Finland, 2008b.
steel-to-timber dowel joints. Holz als Roh- und Werkstoff 2007;65(1):71–7. [48] Legras BD. Effet de la teneur en humidité du bois sur la performance des
https://doi.org/10.1007/s00107-006-0136-6. ISSN 0018-3768. assemblages bois boulonnés de pin gris. Master’s thesis. Québec: Université Laval;
[44] SP. Dragprovning av limträbalkar. Research report number SP 91B1. Ordered by 2009.
Moelven Limträ A/S, Tech. Rep., Swedish National Testing and Research Intitute, [49] Ottenhaus L-M, Li M, Smith T, Quenneville P. Mode cross over and ductility of
1992. dowelled LVL and CLT connections under monotonic and cyclic loading. J Struct
[45] Kevarinmäki A. Ristipultilla vahvistetut puurakenteiden liitokset (Cross-bolt Eng 2018;144(7). https://doi.org/10.1061/(ASCE)ST.1943-541X.0002074.
Reinforced Joints of Timber Structures). HUT/LSEBP Publication 64. Tech. Rep.
Helsinki: Helsinki University of Technology; 1997.
20