Available online at www.sciencedirect.

com

Procedia Engineering 10 (2011) 13091314

ICM11

Numerical modelling and strength prediction of welded

double lap joints made of timber
Till Valle a*, Yves Weinandb, Benjamin Hahnb, Bernhard Stammb
a
College of Engineering and Architecture, University of Applied Science Fribourg, CH 1705 Fribourg / Switzerland
b
IBOIS, Laboratory for Timber Construction, EPF Lausanne, CH-1015 Lausanne / Switzerland

Abstract

Welding of wood is a relatively new method of joining load bearing elements, which holds the potential to compete
with structural adhesion, since it allows for an almost instantaneous bond.
First experimental and numerical investigations showed that failure of wood-welded connections is linked to a
combination of shear and transverse-tensile stresses in the welded interface. One of the main conclusions drawn from
previous investigations is the inability of simple stress-based approaches to allow for an estimate of strength, which is
due to the huge stress peaks generated in such joints.
To achieve a reliable dimensioning of wood-welded joints, it is paramount to determine the stress-strain state inside
the connection, taking into account the mechanical complexity of the involved components, to determine the
mechanical resistance towards the acting stresses and to combine both in the framework of a strength prediction
routine.
This paper addresses the aforementioned issues by summarizing experimental and numerical results on wood-welded
joints, suggesting an experimental method to characterize the strength of the welded interface, and combining the
previous results in the form of a strength prediction method. Lastly, experimental and numerical results are compared,
which allows for a validation of the suggested approach.

2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.

Selection and peer-review under responsibility of ICM11
Keywords: Wood-to-wood connections, friction welding, welded double-lap joints, numerical modelling of welded joints

* Till Valle, Tel.: +41-79-287-2153; fax: +41-21-693-2394.

E-mail address: Till.Valle@fibreworks.org.

1877-7058 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of ICM11 doi:10.1016/j.proeng.2011.04.218
1310 Till Valle et al. / Procedia Engineering 10 (2011) 13091314

1. Introduction

1.1. Load-bearing timber joints

Today, engineers have at their disposal two principle tools to connect load bearing timber structures;
the first fraction of them are mechanical connectors (i.e. dowels, screws, bolts) and the second and
increasingly considered possibility are adhesives like glues. It is less known that a recently developed
adhesive-free bonding method commonly labelled welding of wood also holds potential for load bearing
connections of timber elements. As a fast and direct joining technique welding of wood offers potential
advantages over adhesively bonding, with regard to the relatively long time needed for glues to cure. As
research passed over from small scale tests to the welding of larger timber boards, the question to which
extent welded joints can be considered for load bearing joints in structural timber engineering, and how
such joints can be dimensioned, has to be approached.

1.2. Capacity prediction of timber joints

Before addressing the problem of capacity prediction of welded timber joints the following four points
have to be taken into account:
 The stress-strain state in lap cemented joints has to be determined (herein, the term cemented is used
as a generic term for adhesively bonded and welded joints [1]);
 Quantifying the failure criterion by identifying the failure mechanism of welded timber joints.
 The first two points in combination with the brittle nature of wood make it indispensable to understand
the influence of size effects and its relation to material strength;
 Developing an actionable capacity prediction routine implementing points (i)(iii).

Volkersen [2] and Goland and Reissner [1] determined stresses inside cemented joints in a purely
analytical way. Significant contributions were developed by Hart-Smith [3], followed by more recent
publications by Tsai et al. [4] and Zou et al. [5]; the short list is not claiming to be exhaustive. An
analytical approach of the problem of stress contribution is almost impossible without mechanical
idealizations, i.e. isotropic and linear-elastic material properties. In order to achieve an accurate estimate
of stresses, Finite Element Analysis (FEA) can be used to overcome the limitations due to analytical
simplifications, as it has been shown by Da Silva et al. [6,7].
A number of failure criteria applicable to timber have been developed, and various in depth reviews
were published, e.g. Kasal and Leichti [8]. A commonly applied criterion was proposed by Norris [9], the
latter being based on von Mises-Hencky distortion energy hypothesis; it basically corresponds to a Tsai-
Hill formulation. Regarding failure criteria for interfaces, no such generalized theories exist, and
investigations have to be carried out considering each individual case. Interface failure, frequently
associated with energetic failure criteria, rather than being stress-based, can also be tracked back to
probabilistic concepts [10]. Adding to the complexity of the matter, the associated experimental setups
turn out to be quite complex.
Failure of timber can be characterized as being extremely brittle, especially when it is submitted to
tensile and shear stresses simultaneously. In order to be considered conceptually, such a failure mode can
be triggered by a single weak element which is randomly distributed in the bulk material. Thus, with
increasing the size of the cemented joint, the probability that such a defect is encountered increases too. In
effect, this is the way how Weibull defined the size effect, offering a straightforward quantifiable
relationship between the size of material samples and their respective failure strength [11]. Since failure
 Till Valle et al. / Procedia Engineering 10 (2011) 13091314 1311

of cemented joints is usually associated to a combination of transverse tensile and shear stresses, size
effects have to be considered when handling capacity prediction methods.
Such a prediction method has to integrate all above mentioned considerations. Previous investigations
on FRP adherends [12] and timber adherends [13] showed that the use of a stress-based metric in order to
predict the capacity of cemented joints composed of brittle adherends is not adequate. However, a
probabilistic dimensioning method has been successfully tested against experimental data in [14] and
[15]. This method yields accurate predictions and offers an explanation for the increased material
resistance towards steep stress gradients. It has the additional benefit of relying solely on objective
geometrical and mechanical parameters, excluding any empirical input.

2. Capacity prediction

2.1. Experimental results on welded joints and the welded interface

All experimental data related to the capacity predictions made herein are presented in detail in the
companion paper of the present one; please refer to [16]. To summarize: Experimental and numerical
investigations were carried out on a series of wood welded double-lap joints (Fig. 1-left), in which the
overlap length was varied from 40 mm to 160 mm in steps of 40 mm, considering three different
combinations of thicknesses. The resulting capacities, listed in Table 1, besides exhibiting large
variability, were relatively low, if compared to the load capacity of the members. Subsequently, the
mechanical resistance of the welded connection was experimentally investigated using an adapted off-
axis test, see Fig. 1-right. The latter relatively simple test-setup allowed characterizing the failure
criterion. The experimental evidence confirmed the large variability of the strength of welded bonds.

2.2. Modelling of the welded interface

For the off-axis samples, the tensile load, P, results in stresses in the principal material axis (1,2) by
transformation as follows:

V X V 0 cos 2D V Y V 0 sin 2D W XY V 0 sinD cosD (1)

Where 0 = P/A, A being the cross-VHFWLRQDO DUHD RI WKH VSHFLPHQ DQG  WKH RII-axis orientation
related to the weld line. For each sample, the acting stress, 0, has to be transformed into the
FRUUHVSRQGLQJXY, and XY using (Eq. 1), which act simultaneously.
It can be shown tKDW WKH LQIOXHQFH RI WKH ORQJLWXGLQDO X, is almost negligible. By means of a
subsequent statistical analysis, it was found that the off-axis data is best modelled by Eq. 2:
2 2
2
Vy W xy (2)
I w  1
fW , y fW , xy
Where the parameters fW,Y = 0.61 MPa and fW,XY = 1.07 MPa representing the strength towards tensile
out-of-plane stresses, resp. the shear strength.

2.3. Determination of the Weibull parameters

As welded joints fail under a combination of stresses, it is necessary to extend the concept of the
Weibull distribution towards stresses acting simultaneously. Since, following Eq. 2, W2 = 1 defines
1312 Till Valle et al. / Procedia Engineering 10 (2011) 13091314

failure, it was decided to (i) use the expression W as a stress operator, and subsequently (ii) express all
experimental results from the off-axis samples, disregarding their respective off-D[LVDQJOHDFFRUGLQJO\
The corresponding staWLVWLFDO SDUDPHWHUV LH N DQG 0, were estimated using the least squares/rank
regression method, see [11-15] on how to proceed, and yielded in k = 2.7UHVS0 = 1.05.

2.4. Algorithm of joint capacity prediction

Since the off-axis samples exhibited systematic brittle failure, it was decided to achieve the statistical
modelling of the latter by a Weibull distribution, see Eq. 3, which has proved to be the most accurate
approximation for such failure type.

V k
Ps exp  dV (3)
V V 0

In Eq. 3 Ps is the probability of survival corresponding to the stress acting over a volume V, 0 is the
characteristic stress or scale parameter and k is the shape parameter that gives a measure of the strength
variability, with low values of k corresponding to a high variability. One consequence of Eq. 2 is that for
two sizes S1 and S2 submitted to constant stresses 1 and 2 at failure, assuming equal probabilities of
survival, the relationship given in Eq. 4 is obtained:
1
k
V1 S2
(4)
V2 S1
The quantity size, labelled herein Si, is a length, an area or a volume. If the whole joint is idealized as
being constituted by n elements, its survival depends on simultaneous non-failure of all elements.
Consequently, if each constituent element i, with a volume Ai is subjected to F,i, the probability of
survival of the joint is given by:

n A V m n A V m
PS exp  i F ,i exp  i F ,i (5)
A0 V F ,0 i 1 A V
i 1
0 F ,0

Where Ai are the areas of the considered surface elements constituting the overlap, A0 is the areas of
the friction welded areas that failed in the off-axis samples, W,i is given by (Eq. 2), 0 is the characteristic
stress, and k is the shape parameter according to Eq. 3$VVWDWHGSUHYLRXVO\WKHVWUHVVHVXYDQGXY,
needed to formulate W, were gathered using FEA. Thus, after having determined all stresses, element by
element, all W,i were computed using a spreadsheet, and eventually the corresponding probability of
failure, PS, associated to each element. The global failure is thus defined as the load level, FFEA, for which
(Eq. 5) delivers a global probability of survival, PS, equal to 50%.

2.5. Results of joint capacity prediction

The above described probabilistic method was used to determine the strength of all investigated joint
configurations. Besides the prediction of the mean values, which correspond to the above mentioned
50%-quantile, the 5%-quantile values were determined and listed in Table 1. The joint strengths as
 Till Valle et al. / Procedia Engineering 10 (2011) 13091314 1313

predicted using the probabilistic method increase almost linearly with the overlap length up to L=160
mm, beyond which the strength increase flattens asymptotically for the joints.
Table 1 also compares the predicted capacities to the experimentally determined values; good
agreement was obtained in the general trend, however, due to inconsistent test data, isolated predictions
are of lower accuracy.
It appears that, on average, capacities are slightly underestimated by 17%. Because of the high
scattering in experimental data, a subsequent comparison, at the lower 5%-quantile values, was also
performed, and subsequently listed in Tab. 1: both indicate similarly good agreements (deviations of resp.
19% on average). The good agreement at the lower 5% quantile values sets the basis for characteristic
design values in engineering applications.

Table 1: Experimental joint capacities vs. Numerical predictions, on both the averages and the lower 5%-quantile (all values in kN)

Averages 20/40/20 20/20/20 40/40/40 20/40/20 20/20/20 40/40/40 20/40/20 20/20/20 40/40/40
40 11.29 17.99 8.93 17.7 17.1 14.3 56% -5% 61%
80 17.43 17.3 -1%
120 21.34 31.47 18.53 23.9 31.2 17.8 12% -1% -4%
160 22.14 26.5 20%
Lower 5% 20/40/20 20/20/20 40/40/40 20/40/20 20/20/20 40/40/40 20/40/20 20/20/20 40/40/40
40 8.44 13.35 6.72 14.1 13.7 11.1 67% 3% 65%
80 13.01 8.1 -38%
120 15.89 23.37 14.10 17.9 17.7 10.6 13%
160 16.43 17.4 6%

Figure 1: Specimens after being tested: welded double-lap joints (left) and off-axis tests (right)

3. Conclusions

Welding of wood was investigated in the light of application as a structural joint technique.
Experimental and numerical investigations were carried out on a series of friction welded double-lap
joints, in which the overlap length was varied. Subsequently, the mechanical resistance of the welded
connection was experimentally investigated using an adapted off-axis test. The latter relatively simple
1314 Till Valle et al. / Procedia Engineering 10 (2011) 13091314

test-setup allowed characterizing the failure criterion, which turned out to be best described using Weibull
statistics, in terms of statistical variability. The numerical results show that the load transfer in wood
welded joints is conceptually comparable to adhesively bonded joints, i.e. steep stress profiles with the
associated stress peaks. Lastly, a probabilistic joint strength method allowed to accurately predicting the
experimental mean values (on average by 17%), besides accurately quantifying the inherent variability
(on average by around 19%), thus objectively validating the method.

Acknowledgements

The present research work is funded by the Swiss National Foundation and is part of the project
SNF-Synergia Project Nr. CRSI22_127467/1.

References

[1] Goland M, Reissner E. The stresses in cemented joints, J. Appl. Mech. 1944: A1727
[2] Volkersen O. Nietkraftverteilung in zugbeanspruchten konstanten Laschenquerschnitten. Luftfahrtforschung 1938;15:41-47
[3] Hart-Smith LJ. Analysis and design of advanced composite bonded joints. NASA CR-2218 1974
[4] Tsai MY, Oplinger DW, Morton J. Improved theoretical solutions for adhesive lap joints. International Journal of Solids and
Structures 1998; 35(12):1163-1185
[5] Zou GP, Shahin K, Taheri F. An analytical solution for the analysis of symmetric composite adhesively bonded joints.
Composite Structures 2004; 65(3-4):499-510
[6] Da Silva LFM, Das Neves PJC, Adams RD, Spelt JK. Analytical models of adhesively bonded joints--Part I: Literature
survey. International Journal of Adhesion and Adhesives 2009; 29(3):319-330
[7] Da Silva LFM, Das Neves PJC, Adams RD, Wang A, Spelt JK. Analytical models of adhesively bonded joints--Part II:
Comparative study. International Journal of Adhesion and Adhesives 2009; 29(3):331-341
[8] Kasal B, Leichti RJ. State of the art in multiaxial phenomenological failure criteria for wood members. Progress in
Structural Engineering and Materials 2005; 7(1):3-13
[9] Norris CB. Strength of orthotropic materials subjected to combined stresses. Report No.1816, Forest Products Laboratory,
Madison, USA. 1962
[10] Lamon J. A micromechanics-based approach to the mechanical behavior of brittle-matrix composites, Composites Science
and Technology 2001; 61(15):2259-2272
[11] Weibull W, A statistical theory of strength of materials. Proceedings of the Royal Swedish Institute, Research No.151,
Stockholm, Sweden. 1939
[12] Valle T, Correia JR, Keller T. Probabilistic strength prediction for double lap joints composed of pultruded GFRP
profilesPart I: experimental and numerical investigations. Composites Science and Technology 2006; 66(13):19151930
[13] Tannert T, Valle T, Hehl S. Probabilistic design of adhesively bonded timber joints, Part I: experimental and numerical
investigations. Submitted to Wood Science and Technology 2010
[14] Valle T, Correia JR, Keller T. Probabilistic strength prediction for double lap joints composed of pultruded GFRP
profilesPart II: Strength prediction. Composites Science and Technology 2006; 66(13):19151930
[15] Tannert T, Valle T, Hehl S. Probabilistic design of adhesively bonded timber joints, Part II: Strength prediction. Submitted
to Wood Science and Technology 2010
[16] Hahn B, Weinand Y, Stamm B, Valle T. Experimental investigations on welded double lap joints composed of timber.
Submitted to International Conference of mechanical behaviour of materials ICM11 2011