INTER / 49 - 7 - 3

Predicting the load‐deformation of bolted

timber connections up to failure

Y. Zhang, PhD candidate, University of Auckland

G. M. Raftery, Lecturer, University of Auckland
P. Quenneville, Professor of timber design, University of Auckland

Keywords: timber, connection, bolt, resistance, displacement, failure, seismic

1 Introduction
The design of connections is an important portion of the design exercise for timber
buildings. For buildings that are subjected to occupancy, wind and or snow loadings,
the predictions of the governing resistance can be sufficient. However, prediction of
the governing resistance for connections that are subjected to seismic loading is
insufficient. For timber connections subjected to seismic loading, it is extremely
important to be able to ensure a level of ductility. Thus, one needs to predict the
ultimate yielding resistance and the over‐capacity of the other failure modes (Park
and Priestly, 1992). Finally, it is also important to be able to predict the ultimate
displacement so as to have an indication of the ductility, given by Equation (1), that
the connection can provide. Thus, engineers need to be able to design timber bolted
connections with a known outcome for the ultimate resistance and yield and ultimate
displacements, as idealised in Figure 1, to quantify the ductility.
∆ ∆
(1)
∆

From past research on bolted

connections, it is well known that the
ultimate yielding resistance can be
well predicted. Research has allowed
timber designers to predict the
resistance of dowel‐type fasteners if
Figure 1. Load‐deformation curve for a ductile their governing mode of failure is
bolted connection showing ultimate resistance, yielding (Johansen, 1949). Yielding
yield and ultimate displacements. failure is the result of plastic
deformation of the wood fibres alone or in combination with the yielding of the bolt.
Described as the European Yield Model (EYM), yielding connections allow plastic

1
INTER / 49 - 7 - 3

deformation to occur through crushing of the wood fibres in bearing (Modes I and II)
and the combination of some crushing of the wood fibres and formation of plastic
hinges in the steel connector (Modes III and IV), as shown in Figure 2.
The EYM equations predict
the yielding capacity of
connections, based on
these four modes. For non‐
seismic design, the
connection resistance is
taken as the minimum
capacity of the four EYM
Figure 2. Yielding failure for bolted connections loaded parallel‐
modes, if none of the
to‐grain (double shear cases).
possible brittle failure
mode occurs. However, it is recognised that if brittle failures are avoided through
design, the ultimate yielding resistance will be given by the resistance of mode I or II,
as shown in Figure 3. It is thus imperative to understand connection resistances that
are associated with brittle failures in order to avoid them and target the ultimate
yielding capacity of the EYM failure modes.

Pu Pu

Py

(a) (b)

Figure 3. Idealised load‐displacement curves for EYM (a) modes I and II and (b) modes III and IV.
Subsequently, secondary failure occurs in one of the brittle failure modes identified
in Figure 4 (Jorissen, 1998, Quenneville and Mohammad, 2000). This secondary
failure typically occurs suddenly and without warning. Work on brittle failures of
bolted connections in recent decades has permitted some form of prediction of the
various brittle failure resistances, such as splitting, row shear, group tear‐out and net
tension (Quenneville and Morris, 2009).
It is understood that the resistance of brittle failure modes can govern the design of
timber connections and these brittle failures, if not controlled, can occur at a
resistance lower than Py, or in between Py and Pu in Figure 3b or below Pu in Figure
3a. Thus, to take full advantage of the maximum ductility that a timber connection
can offer, it is thus desirable to ensure that brittle failure mode resistances are well
above the resistance of modes I and II of the EYM.

2
 INTER / 49 - 7 - 3

If this is the case, the only

remaining unknown in the
determination of a connection’s
ductility is to quantify the
ultimate displacement.
When it comes to a bolted
connection displacement, little
or no research has allowed
anyone to be able to predict the
ultimate displacement.
Numerous researchers have
worked on ductility predictions
Figure 4. Types of timber bolted connection brittle failure and many have provided
modes. different procedures for
determining the yield point
from a known bolted connection load‐displacement curve (Munoz et al., 2008,
Karacabeyli and Cecotti, 1996, Yasamura and Kawai, 1998). However, the prediction
of the ultimate displacement is unknown. Until this is possible, determination of the
ductility available in a bolted connection will remain only approximate or available for
connection configurations that have been tested and replicated in practice. This is a
serious impediment for timber designers that need to design timber connections for
buildings in seismically‐active locations.
Thus, in order to palliate this situation, a research program was undertaken to predict
the ultimate displacement of timber bolted connections. It has been observed that
for highly ductile connections (with a significant amount of displacement), the final
failure mode is brittle. It is thus postulated that the ultimate displacement of a timber
bolted connection is related to the over‐capacity of the non‐governing brittle modes
of failure.

2 Methodology
Given that for the design of connections resisting seismic loads, one would ensure
that a yielding behaviour would be governing, it would thus mean that the resistance
of the other failure modes (splitting, row shear, group tear‐out, net tension) would
need to be well over the ultimate yielding failure mode resistance (modes I or II, not
modes III or IV).
Given that the resistance of the brittle failure modes is directly related to the
configuration of the connection, i.e. the amount of wood mobilised in resisting the
connection force, the brittle resistance over‐capacity will thus be affected and it was
postulated that the ultimate displacement would also be affected.

3
INTER / 49 - 7 - 3

Table 1: Bolted timber connection configurations tested. All Radiata Pine.

series material size bolt size nr x n a3t a1
(mm) (mm) (mm) (mm)
1 Glulam 90 x 180 12 2x2 36 36
2 Glulam 90 x 180 12 2x2 72 72
3 Glulam 90 x 180 12 2x2 108 108
4 Glulam 135 x 225 20 2x2 80 80
5 Glulam 135 x 225 20 2x2 160 160
6 Glulam 135 x 225 20 2x2 240 240
7 LVL 45 x 100 12 1 96 ‐‐‐
8 LVL 45 x 100 16 1 128 ‐‐‐
9 LVL 45 x 100 20 1 160 ‐‐‐
10 LVL 45 x 200 20 1 160 ‐‐‐

Note: a2 is 5d for all glulam configurations

All connection specimens consisted of steel‐wood‐steel connections. In the case of

multi‐bolt configurations, the end distance and bolt spacing were set equal to
remove any influence of the smaller one of the two variables. Connections were
tested in tension only under a monotonic load up to failure. A uniform displacement‐
control rate was maintained until failure
occurred. The tests were performed in
accordance with the standard testing
procedures outlined in ISO 10984‐2:2008.
Two linear variable differential transformers
(LVDTs) were symmetrically placed at the
end of the specimen to measure the average
deformation of the timber relative to the
steel plates. The test setup for the single bolt
in LVL is shown in Figure 5.
Using the load and displacement outputs,
the stiffness, resistance and displacement
values for each of the connections were
identified at multiple points along the curve.
The average values and coefficients of
variation (COV) were then calculated for
each connection type. This allowed the load‐
Figure 5. Test set‐up for the single bolt tests
deformation behaviour to be modelled.
in LVL.

4
 INTER / 49 - 7 - 3

3 Results and discussion

3.1 Load‐displacement response analysis
For the specimens of the first 6
series, the load‐displacement
response, when mode II governed
the resistance, consisted of three
distinctive regions, as shown in
Figure 6. The behaviour of the
connection starts with an elastic
phase (1), then moves on to
inelastic (2) and subsequently to
the plastic one (3). Between each
of these regions was a transition
zone, where interaction between
the two adjacent behaviours was
observed.
Referring to Figure 6, the initial
response is elastic, where Figure 6. Actual and idealised load‐displacement
recoverable displacement occurs response.
as the connection adjusts to the
applied load. As the load increases above a certain level, the connection begins to
yield in an inelastic response. This is identified as the point at which the load‐
displacement curve changes gradient, shown at location (Δ1, P1). P1 is associated with
the resistance given from the EYM mode III or IV. Experimental observations
demonstrate that at this point the bolts begin to bend. Plastic hinges start to form at
specific points, depending on the governing mode III (2 plastic hinges) or IV (4 plastic
hinges). As the load
increases, the rope effect
starts to take place and
multiple plastic hinges
form along the length of
the bolt, multiplying until
the entire length is
deformed, as shown in
Figure 7. Bolt deformed to ultimate EYM mode II after rope effect, Figure 7.
showing multiple plastic hinges.

5
INTER / 49 - 7 - 3

At this point, shown as

location (Δ2, P2), no additional
bending of the bolt can take
place and consequently the
timber is forced to fail in
embedment. A secondary
brittle failure then occurs (Δu,
Pu), resulting in a sudden
decrease in the load capacity
of the connection.
The configurations of each of
the connections tested
influenced the size of each of
Figure 8. Idealised average load‐displacement responses for the behaviour regions, and
series 1 to 6. the point at which failure
occurred. Connections of
series 1, 2 and 4 failed prior to the plastic region being reached, and hence the
ultimate capacity (Pu) was less than P2. Series 3, 5 and 6 typically failed within the
plastic region, which is the desired load‐displacement response for seismic
connections.
The co‐ordinate values and adjoining gradients recorded for the connections were
averaged for each of the six configurations, as shown in Figure 8 (series numbers
indicated in brackets). It was found that the configurations with the same bolt size
and cross‐sectional area had similar load‐displacement responses.
Classical design philosophy suggests that the capacity of a connection should be
defined as the minimum
resistance of any given failure 40 (9) (10)
mode. The minimum of the EYM 35 (8)
mode III and mode IV is 30
(7)
Load (kN)

determined to be the point at 25

which yielding failure begins to 20

occur. However, testing showed 15

10
that the capacity of the
5
connection continued to increase
0
after the bolts began to yield (Δ1, 0 2 4 6 8 10 12
P1) and a yield plateau only Displacement (mm)
formed when Mode II failure
began to occur (Δ2, P2). This Figure 9: Idealised average load‐displacement responses
indicates that the EYM Mode II for series 7 to 10.
accurately represent the ultimate

6
 INTER / 49 - 7 - 3

capacity of those connections for which a

brittle failure mode does not governs the
resistance.
With regards to the load‐displacement
relationships for the specimens of series 7 to
10, as shown in Figure 9, the behaviour was
slightly different as the cross‐section of the
specimen and the diameter of the bolts were
such that their slenderness ratio is very low
and this forced an EYM mode II yielding failure
to occur. The ultimate displacement was Elongated
bolt holes
attained when the specimen failed in splitting.
3.2 Reduced brittle resistance model
Assessment of a failed ductile bolted
connection from series 1 to 6, shown in Figure Compressed
fibres
10, allows one to make the following
observations. This specimen exhibited a
significant amount of deformation before
ultimately rupturing in row shear. As the
specimen starts to fail in yielding, timber
adjacent to the bolt hole is compressed. This
was accompanied by yielding of the bolt as Figure 10. Internal view of connection
well. As the connection load increased, the after secondary failure showing extent of
compressed zone increased and the damaged fibres.
connection deformation increased as well.
During this additional displacement, the bolt was subjected to the rope effect. The
wood fibres adjacent to the deformed bolt shape are bent, and there is a visible line
where the fibres have become dislodged from the surrounding timber matrix.
This compressed length is damaged and unable to provide resistance shear
resistance, resulting in a lower row shear capacity for the connection than that was
initially determined for the undamaged specimen. This means that existing brittle
failure models which rely on the shear resistance represent an upper bound of their
resistance. Thus, as the connection is deforming in a yielding mode, the brittle failure
resistances for row shear and group tear‐out are decreasing.
The failure hierarchy for series 8 to 10 is almost the same. Series 9 is taken as an
example. After the yielding capacity of the connection is reached, a crack is observed
next to the bolt hole. This crack is accompanied by crushing of the wood (Figure
11(a)). As the load increased, more wood was crushed but this crack didn’t develop
further. Instead, new cracks were observed a certain distance away from the
centreline of the specimen (Figure 11(b)).

7
INTER / 49 - 7 - 3

(a) (b) (c) (d)

Figure 11. Hierarchy of the splitting failure for specimens of series 7 to 10.

These cracks propagated downwards until another crack occurred at a location closer
to the centreline. This became the most significant crack and it gradually spread
towards the specimen’s end and in most cases, it spread towards the centreline at
the same time (Figure 11(c)). Shortly after this crack reached the specimen end, it
opened up from the end and this splitting caused the catastrophic failure and the
ultimate displacement was attained (Figure 11(d)). The failure hierarchy for series 7 is
more or less the same except that a crack initiated from the specimen end well
before the final failure. It propagated upwards and along a line close to the
centreline. This crack caused the catastrophic failure.
The nature of the failure indicates that failure was probably caused by both tensile
stresses perpendicular to grain and shear stresses parallel to grain. The load‐
displacement curves of the series 7 to 10 only consist of two distinctive regions
instead of three regions. The bolts from these tests maintained their original shape
throughout the test and no plastic deformation was evident. These observations
prove that only EYM mode II was involved. More tests are underway to assess the
effect of the end distance on the ultimate displacement when splitting is the
secondary failure mode.
Analysis of the secondary failure modes allows one to hypothesise on the causes of
the ultimate displacement (Δu). The elongation of the bolt holes reduces the volume
of timber providing shear or tension perpendicular‐to‐grain resistance to the
connection. This means that splitting, row shear and group tear‐out brittle failure
resistances decrease with any increase in permanent displacement (inelastic and
plastic deformation), as the capacity equations for these failure modes include the
end distance (a3t) and bolt spacing (a1) variables. This differs from the net tension
8
 INTER / 49 - 7 - 3

capacity, which remains constant, not being dependent on either of these variables.
This hypothesis is best illustrated in Figure 12.

1200

1000

800
Load (kN)

splitting
600
net tension
400 y ??
group tear‐out
200
row shear
0
0 10 20 30 u 40
Displacement (mm)
Figure 12. Assumed load‐displacement response for a yielding failure with decreasing brittle failure
resistances.

Figure 10 also reveals that the compressed length is greater than the permanent
displacement. It is not known at this time what is the actual relationship between the
connection deformation and the extent of the shear damage zone. In the resistances
relationship shown in Figure 12, the decrease in row shear and group tear‐out
resistances is assumed linear. Further research is underway to establish what is the
rate of change of the brittle failure mode resistances.

3.3 Prediction of the ultimate displacement

It is hypothesised that the ultimate displacement of a bolted connection is
corresponding to the connection displacement at which the EYM mode II resistance
intersects the reducing brittle resistance of splitting, row shear or group tear‐out
failures.
Using the reduced brittle resistance model in conjunction with the existing EYM, Jacks
(2015) and Novis (2015) thus proposed to determine the ultimate displacement (Δu)
using Equation (2). This equation is only applicable to those connections in which row
shear or group tear‐out is the governing brittle failure type, and the brittle resistance
(Pb) is greater than the ductile resistance (PEYM(II)).

9
INTER / 49 - 7 - 3

ac (Pb  PEYM (II ) )

u   1
FPb (2)
The variables ac and Δ1 can be defined according to Equation (2) and Equation (3)
respectively. F is an undetermined factor at this point which represents the rate of
change of the row shear and group tear‐out resistances
ac = min(a1, a3t) (3)

1 = P1 / K1 (4)

Where a1 and a3t are the bolt spacing and end distance prior to loading, and P1 and K1
are identified in Figure 6.
Equation (2) was developed using configurations three and five only, as they were the
only connections to fail in the plastic region and be governed by one of the applicable
brittle failure types (both failed by row shear). Additional research is required to
validate this model.
3.4 Model limitations
Equation (2) is proposed based on the assumption that the plateau of the load‐
displacement response is horizontal (that the plastic stiffness is zero). The net tension
model is also represented by a horizontal relationship which means that it is not
theoretically possible for a connection to fail in net tension once it has begun to fail
plastically. Nevertheless, a net tension failure was observed during the testing of
configuration 6, where friction played a significant role in the resistance of the
connection. This increase, when applied in the model, reduces the amount of
predicted deformation that will occur before the brittle failure resistance is
exceeded. Failure would thus occur on the load‐displacement curve at an earlier
point than expected, meaning that the model is non‐conservative if one assumes
zero stiffness in the plastic region.
It is also assumed that the rate of change of the resistances for the brittle failures is
constant for both the inelastic and plastic regions. Not enough experimental
observations are available to verify these assumptions. More observations of the full
connection load‐deformation curve (including the ultimate displacement) to quantify
the value of F. are being acquired through testing.
3.5 Ductility
As discussed above, seismic design requires knowledge of the ductility of the
connection. Therefore, in addition to the ultimate displacement, the displacement at
the yield point is also required. The location of the yield point is subjective and as the
bolt yielding begins to occur at location (Δ1, P1), the entire connection yielding does

10
 INTER / 49 - 7 - 3

not occur till the attainment of (Δ2, P2). Current models for identifying the yield point
do not acknowledge the presence of the inelastic region and hence may not be
accurate or neglect the plastic region altogether. Perhaps a review of the yield point
location is required in light of this research. However, it is suggested that an
appropriate location may be at the intersection of the extended elastic and plastic
lines. Verification of the yield point would then allow the ductility to be calculated; as
the ratio of the ultimate displacement (calculated using the proposed reduced brittle
resistance model) and the displacement at the yield point.

4 Conclusions
This paper presents an analysis of the load‐displacement response of bolted timber
connections up to failure. A hypothesis for the causation of the ultimate deformation
is proposed. From the study, it can be concluded that:
 For ductile seismic design, the EYM mode II best predicts the ultimate capacity.
This implies that the use of fasteners with low slenderness is recommended.
 The ultimate displacement occurs at the point where the load‐displacement
response meets the decreasing resistance of splitting, row shear or group tear‐out
failure modes.
More testing is underway to evaluate the effect of the end distance on the ultimate
displacement for secondary splitting failure.

5 Recommendations
It is recommended that more observations of the full connection load‐deformation
curve (including the ultimate displacement) be made to quantify the value of F.

6 Acknowledgements
The authors wish to acknowledge the initial research contribution of Sarah Novis and
Jessica Jacks, former undergraduate students at the University of Auckland, and the
experimental assistance of Mark Byrami and Ross Reichardt.

7 References
ISO 10984‐2 (2008): Timber structures – Dowel‐type fasteners. Determination of
embedding strength and foundation values.
Jacks, J. (2015): Predicting the load‐deformation behaviour of bolted timber
connections up to failure. Final Year Report submitted in partial fulfilment of the
undergraduate degree. Department of Civil and Environmental Engineering, The
University of Auckland.

11
INTER / 49 - 7 - 3

Johansen, K.W. (1949): Theory of Timber Connectors. Publications of the

International Association of Bridge and Structural Engineering. No. 9: 249‐262. Bern,
General Secretariat.
Jorissen, A. (1998): Double Shear Timber Connections With Dowel Type Fasteners,
PhD thesis, Delft University Press, The Netherlands, 264 p.
Karacabeili, E., Cecotti, A. (1996): Quasi‐static reversed cyclic testing of nailed joints.
Proceedings of the International Council for Building and Research, Working
Commission W18 – Timber Structures. Paper 29‐7‐7, Karlsruhe, Germany.
Munoz, w., Mohammad, M., Salenikovich, A. and Quenneville, P. (2008):
Determination of yield point and ductility of timber assemblies: in search for a
harmonised approach. Proceedings of the WCTE conference, Miyasaki, Japan.
Novis, S. (2015): Predicting the load‐deformation behaviour of bolted timber
connections prior to failure. Final Year Report submitted in partial fulfilment of the
undergraduate degree. Department of Civil and Environmental Engineering, The
University of Auckland.
Paulay, T., & Priestley, M. J. N. (1992): Seismic design of reinforced concrete and
masonry buildings. New York, Wiley.
Quenneville, J.H.P., and Mohammad, M. (2000): On the failure modes and strength of
steel‐wood‐steel bolted timber connections loaded parallel‐to‐grain. Can Journal Civ
Engr, 27, pp. 761‐773.
Quenneville, P., & Morris, H. (2009): Proposal for a mechanics‐based bolted
connection design approach for AS1720.1. Australian Journal of Structural
Engineering, 9(3), p. 195‐206.
Yasumura, M. and N. Kawai. (1998): Estimating seismic performance of wood‐framed
structures. Proceedings of 1998 I.W.E.C. Switzerland, Vol. 2, pp. 564‐571.

12