You are on page 1of 11

Construction and Building Materials 25 (2011) 3078–3088

Contents lists available at ScienceDirect

Construction and Building Materials


journal homepage: www.elsevier.com/locate/conbuildmat

Investigation on bonding between timber and ultra-high performance


concrete (UHPC)
Martin Schäfers b,1, Werner Seim a,⇑
a
Department of Structural Engineering, University of Kassel, D-34125 Kassel, Germany
b
Forschungsvereinigung Kalk-Sand eV, Entenfangweg 15, D-30419 Hannover, Germany

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

Article history: Bonded composite constructions from timber and ultra-high performance concrete (UHPC) are investi-
Received 5 July 2010 gated as a highly innovative structural element. In a first step, mechanical models for the description
Received in revised form 14 December 2010 of the bond between timber and UHPC are documented and analyzed. The theory of the compliant bond,
Accepted 24 December 2010
which was already introduced by Volkersen (1953) [28], and the linear elastic fracture mechanics (LEFM)
Available online 12 February 2011
are possible approaches for this purpose. Extensive experimental investigations on the bond behaviour
are accomplished. From this it is found that the failure of the bond occurs in timber close to the bond-line
Keywords:
in most cases. The kind of the surface treatment of UHPC does not affect the bond behaviour. In general,
Bond
Composite
the ultimate loads can be estimated by Volkersens theory, but in the case of short bond lengths, tension
Adhesive stresses in timber, orthotropic to the bond-line affect bending strength. With numerical studies, this
UHPC effect could be examined and it was found that the mentioned tension stresses do not affect the strength
Timber for bond lengths of 300 mm and more. On the basis of the bond investigations, recommendations for the
Fracture mechanics design of bending members are given, and validated by two bending tests.
Volkersen Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction ising, because they exhibit a high surface tensile strength. A further
advantage that results from the steel fibre reinforcement is that the
Combining timber and concrete to composite structures is an concrete slabs can be very thin because there is no need for a con-
efficient way to create high performance bending members, hence ventional steel-bar reinforcement. Thereby, the weight of the com-
each material is exposed to stresses that correspond to its charac- posite construction can be reduced further.
teristics (timber – tension, concrete – compression). This leads to In the following studies, the basic information on the behaviour
lightweight constructions with a high global bending stiffness of bonded joints of UHPC and timber will be developed by theoret-
and load capacity. One weak point of common timber–concrete ical and experimental investigations. In a first step, mechanical
composites is the connection between the two materials. Mostly models for the description of the bond behaviour are documented.
point shaped joints like screws or nails are applied, which leads After that, experimental investigations on the bond, which con-
to a flexible bond. Compared to rigid bond, this causes a reduction sider several aspects like the bond length, the surface treatment
in bending stiffness and the concrete cannot be exposed to its full of concrete, or the type of timber are carried out. The results of
strength. the experimental investigations will be validated with mechanical
To overcome this disadvantage, the use of adhesives can assure models as mentioned above, as well as with numerical investiga-
a stiff compound between the two materials. At present, there are tions. Finally, conclusions for the design of bending members will
only few investigations that consider this approach. Pincus [20] be derived on the background of these investigations and validated
and [21], Brunner et al. [5], Haiman et al. [13] and Negrao et al. by two full-scale bending tests.
[18] used adhesives for bonding of timber beams to standard con-
crete slabs. Within all these cases, the transfer of shear forces
through the bond was limited by the low surface tensile strength 2. Modelling of the bond
of the standard concrete. In this context, the use of steel fibre rein-
forced high-, resp. ultra-high-performance concretes seems prom- Beside the properties of the applied materials the failure load of
a bond-line depends on its length (e.g. [3] or [26]). Conventional
⇑ Corresponding author. Tel.: +49 561 804 2625; fax: +49 561 804 7647. stress-based design criteria are not able to consider this phenom-
E-mail addresses: martin.schaefers@kalksandstein.de (M. Schäfers), wseim@
enon. In the following, mechanical models which consider this
uni-kassel.de (W. Seim). ‘‘size-effect’’ will be reviewed. Preliminary investigations [24,23]
1
Tel.: +49 511 279 5442; fax: +49 511 279 5419. as well as first results from bond tests depicted that the fracture

0950-0618/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2010.12.060
M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088 3079

of bonded UHPC–timber-joints takes place in timber close to the In the following section it will be considered in which cases the
bond-line. Due to high shear strength, this does not apply to beech LEFM can be used for the prediction of the ultimate load of a bond
specimens. Nevertheless, in the following the focus lies on and in which cases other methods are more reasonable. In Sec-
mechanical models which are adequate to describe bonding failure tion 4, the failure loads calculated according to Eq. (5) are com-
as shear failure of the timber. In this context, it must be stated that pared to the results of the experimental investigations from
the theories which are discussed below only consider stresses Section 3.
which act in the direction of the bond-line, and neglect bending ef-
fects and stresses normal to the bond-line. A detailed observation 2.2. Generalized Volkersens theory according to Gustafsson
on the influence of these different stress combinations will be
accomplished within the numerical investigations in Section 5. On the basis of Volkersens theory, which was already published
in 1953 [28], Gustafsson derives a generalized method for the eval-
2.1. Prediction of the bond strength by LEFM uation of failure loads of bonded composites. Gustafsson estab-
lishes a normalized s  d-relation by dividing s through sf and d
The application of linear elastic fracture mechanics (LEFM) is through Gf/s. Furthermore Gustafsson inserts a ratio of the axial
one way to evaluate the strength of timber members if size-effects stiffnesses:
have to be considered. On the basis of LEFM, the failure load for the
t 1  E1
deployed test specimen (Fig. 1) is derived by considering the a¼ 6 1:0: ð6Þ
t 2  E2
change of the compliance C if a crack is propagating over the length
da Generally the behaviour of a bond is influenced by the elastic
modulus and the thickness of the bond partners (E1, E2, t1, t2), by
da da
dC ¼  : ð1Þ the size of the bond region (b and l), by the shear strength sf, the
ðEAÞw ðEAÞw þ ðEAÞc fracture energy Gf and the special shape of the constitutive relation
of the bond g (see Fig. 3). Gustafsson summarizes these nine
1 2
dW ¼  F  dC: ð2Þ parameters and proposes a normalized failure load in dependency
2 of the the brittleness ratio k, the ratio of the axial stiffnesses a and
According to the definition of the fracture energy Gf, the applied the shape of the constitutive stress–strain relation g.
load F reaches a maximum and the crack propagation becomes 2 3
unstable if: 6 2 7
Fu 6 l  s2 7
dW ¼ GF  dA: ð3Þ ¼ f6 ; a; g 7: ð7Þ
sf  b  l 4 t 1  E1  Gf 5
|fflfflfflffl{zfflfflfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl}
With Eqs. (2) and (3) and for dA = b  da Eq. (1) can be converted to: norm: failure load brittleness ratio k

GF  b  da ¼ ð4Þ The brittleness ratio k characterizes the bond behaviour. By


  means of limit considerations, Gustafsson shows that k can be used
1 2 da da
F   ; to characterize the type of failure and that different methods can
2 ðEAÞw ðEAÞw þ ðEAÞc
be applied to predict the failure load in each case:
and the fracture load can be written as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1. For small k, the failure load Fu only depends on the local shear
2  GF  b strength sf but is independent of the fracture energy Gf and
Fu ¼ 1
: ð5Þ
ðEAÞ
 ðEAÞ 1þðEAÞ the shape of the constitutive relation g.
w w c

2. If k is high, Fu only depends on the fracture energy Gf and is


It can be pointed out that the ultimate load, which is calculated
independent from the local shear strength sf and the shape of
on the basis of LEFM (Eq. (5)) is only dependent on the value of the
the constitutive relation g.
fracture energy Gf. The shape of the constitutive stress–strain rela-
3. In the transition zone between high and small values of k, the
tion and the size of the local shear strength of the bond have no
local shear strength sf and the fracture energy Gf, as well as
influence on the ultimate load according to LEFM.
the shape of the constitutive relation g, have an effect on the
size of the failure load Fu (see Fig. 2).

In the first case, the joint behaves ductilely and the failure load
appears at a completely plasticized joint-line. In the second case,
the joint behaves brittely and the failure load is identical to the
solution according to LEFM (Eq. (5)). Between these two limiting
cases exists a transition zone, in which the failure load can be cal-
culated on the basis of Volkersens theory.
By means of numerical calculations on the basis of this theory
and adopting four different shapes of the s–d-curve (see Fig. 3),
Gustafsson defines magnitudes for the brittleness ratio k that as-
sign the limits between the three zones in Fig. 2.

F u  sf  b  l if k 6 0:1ð1 þ aÞ: ð8Þ

F u ¼ sf  b  l  f ðk; a; gÞ generally: ð9Þ

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fu  2ð1 þ aÞb  t1  E1  Gf : ð10Þ

Fig. 1. Specimen with rigid compound. if k P 10ð1 þ aÞ


3080 M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088

constitutive relation IV LEFM

constitutive
relations
l ti I - III
malized joint strength
norm

ductile transition zone


failure
brittle
failure

b ittl
brittleness ti λ
ratio

Fig. 2. Normalized failure load over brittleness ratio k, from [22].

Fig. 4. Glued-in rods under: (a) pull–pull; (b) pull–compression; (c) pull–distrib-
uted loading, from [12].

ð1 þ kÞ  sinhðxÞ
F u;a ¼ sf  dst  lg  ð11Þ
x  ðk þ coshðxÞÞ
tanhðxÞ
F u;b ¼ sf  dst  lg  ð12Þ
x
ð1 þ kÞ
F u;c ¼ sf  dst  lg  : ð13Þ
x= tanhðxÞ þ k
sffiffiffiffiffiffiffi
lgeo
I: Linear-elastic II: Bi-linear (ξ = 1,0) with x ¼
lm
p  dst  l2g  ð1 þ kÞ
lgeo ¼
2  Ast
Est  Ast
lm ¼
Ew  Aw
Est  Gf
k¼ :
sf
2

The Eqs. (11)–(13) result from the solution of the differential


equation of the compliant bond, whereas Eq. (11) is the solution
for the pull–pull configuration (Fig. 4a), Eq. (12) for the pull–com-
III: Bi-linear (ξ = 0,02) IV: Pile-shaped pression configuration (Fig. 4b) and Eq. (13) for an inclined orien-
tation of the glued-in rod (Fig. 4c). The experimental test setup in
Fig. 3. Four types of the constitutive relation for the bond, from [22].
Section 3 follows the pull–pull configuration in principle (Fig. 4a) –
with compression – instead of tension loads. Based on the assump-
tion that the constitutive relation is independent to the direction of
As Fig. 2 shows, the influence of the shape of the constitutive
the strain, Eq. (11) is applied for the prediction of the ultimate load
relation (Fig. 3) has only a small influence on the magnitude of
of the bond tests in Section 4.
the calculated failure load (cf. also [1]). The influence of the shape
of the constitutive relation is more significant, if the relative dis-
placement of the bond partners shall be considered. For the design 3. Bonding tests
of bending members, the calculation of the relative displacement
at the scale of the bond test specimens is not essential. It only 3.1. Preliminary considerations and parameters
has to be determined if the bond can be regarded as a stiff connec-
tion or not. This will be evaluated experimentally later on. Hence, One aim of the following investigations on the bond between
further calculations concerning the bond tests are carried out with UHPC and timber is to develop basic information for the design
a linear elastic constitutive relation. This has the advantage that a of full-scale bending members. In this context the predominant
closed solution of the differential equation exists and that the fail- failure criteria in the bond region and the bond capacity have to
ure loads can be estimated analytically. be identified. Further questions that have to be considered are
Based on Volkersens theory and adopting a linear elastic consti- the influence of the type of timber materials and the influence of
tutive relation, Gustafsson et al. propose design equations for the the surface treatment of concrete. Another central question is the
ultimate load of joints under three different loading conditions influence of the bond length on the behaviour of the joint. Conclud-
(cf. Fig. 4): ing a standard test setup for experimental investigations on
M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088 3081

Table 1 Table 3
Test design and varied parameters for the bond tests. Timber materials: E-modulus, compression and shear strength.

Test Timber Surface Concrete Bondlength (mm) Ec,0 (N/mm2) q (kg/m3) fc,0 (N/mm2) fv (N/mm2)
V01-V03 Sp G UHPC 150 Spruce 9,940 397.0 37.8 5.7
V04-V06 Be G UHPC 150 Beech 13,560 622.4 56.6 11.3
V07-V09 Ke G UHPC 150 Kerto 13,460 523.7 47.2 5.4
V10-V12 Sp Fl UHPC 150
V13-V15 Be Fl UHPC 150
V16-V18 Ke Fl UHPC 150
V19-V21 Sp G HPC 150 shown in Fig. 5. The load is applied at the top of the specimen by
V22-V24 Ke G HPC 150 a hydraulic jack over a load cell through a steel plate. At the top
V25-V27 Sp G UHPC 100
of the specimens, the width of the cross-section was reduced be-
V28-V30 Ke G UHPC 100
V31-V33 Sp G UHPC 200 cause only the timber over the length of the bonding-line should
V34-V36 Ke G UHPC 200 be activated. On the right side in Fig. 5 the load is transferred
V37-V39 Sp G UHPC 250
V40-V42 Ke G UHPC 250
V43-V45 Sp G UHPC 300
V46-V48 Ke G UHPC 300 2F
V49-V51 Sp G UHPC 400
V52-V54 Ke G UHPC 400
V55-V57 Sp G UHPC 500
V58-V60 Ke G UHPC 500

The bold characters highlight the parameters which were varied within the
experimental observations. load cell
20 20
bonded timber–concrete joints shall be proposed. Table 1 gives an
overview on bonding tests and varied parameters. For the basic
type, glued-laminated timber (GL24h) is bonded to sandblasted
UHPC slabs with the epoxy resin ‘‘Sikadur 330’’. The bond length

30
steel plate
for the basic type was 150 mm. Besides the kind of timber (spruce,
unidirectional laminated veneer lumber and beech), the concrete
design (UHPC and HPC), the surface treatment of the concrete
(sandblasted and ground) and the bond length (100, 150, 200,
250, 300, 400 and 500 mm) are varied. Preliminary investigations

170
depicted that the thickness of the bond-line and the surface treat- gluelam, GL24h
ment of timber have no influence on the bond strength (see [23]).
3 screws
Hence, these parameters were not varied for the bonding tests. The ,00
6 x 50 R85
thickness of the bond was chosen to 1.5 mm and the timber surface
was planed in all cases. For each parameter combination a set of

370
three specimens was tested, which leads to overall 60 bond tests.
The concrete design for the UHPC and HPC as well as the com- A A
W1 + W2
pressive strengths of these concretes are listed in Table 2. Both
steel-bearing

concretes were steelfibre-reinforced. The UHPC exhibited a com- HPC/UHPC


pression strength of 152.4 N/mm2 and the HPC of 82.2 N/mm2.
150
joint
For the timber materials, material properties were evaluated
experimentally according to DIN EN 408 [9]. The average values
of these properties are given in Table 3. In the case of spruce, 12
specimens were tested and in the case of beech and Kerto, three
20

resp. five.
W3 + W4

3.2. Test setup F F

The test setup for the experimental investigations on bonding


was developed on the basis of preliminary investigations and is cross-section A - A

Table 2 W1 + W3
Concrete mixture and strength.

per m3 Concrete UHPC (kg) HPC (kg)


100

Water 166.0 170.0


CEM I 52.5 R HS-NA 832.0 418.5
Basalt sand 0/2 559.4 1671.0
Sand 0/2 518.8 0.0
Microsilika MS983 135.0 60.0 W2 + W4
Silica fume W12 207.0 291.5
Steel fibres 196.3 196.3
Superplasticizer 29.4 20.0
Compression strength 20 200 20
in N/mm2 according to
DIN EN 12390-3 152.4 82.2
Fig. 5. Experimental setup for the bonding tests.
3082 M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088

through the bond-line into the concrete slab. To minimise the by compression failure of the concrete. The tension failure oc-
needed measurement instrumentation, the load is transferred into curred only at short bond lengths and might be caused due to ten-
a steel bearing on the left side which ensures a bonding failure at sion stresses that were induced into the specimens due to the test
only one side of the test specimen (Fig. 5). The relative displace- geometry. Particular friction on the concrete affects the size of
ments between concrete and timber were measured at four points these tension stresses. This question will be studied in more detail
with displacement transducers (nominal measuring length of in Section 5 by numerical investigations. For further examinations
±1 mm). The displacement transducers are placed on the front of the test results, only the specimens that showed a bond
and back side of the specimen on the top (W1 + W2) and at the bot- failure – resp. shear failure are regarded.
tom (W3 + W4) of the bond-line. All spruce and LVL specimens that broke down due to a shear
For four specimens (V03, V31, V43 and V57) strain gradients failure showed fracture surfaces almost exclusively in timber. Nei-
were measured with one-dimensional strain gauges. These strain ther the type of surface treatment of the concrete nor the type of
gauges (6 mm gauge length) were applied on the sandblasted sur- concrete had any influence on that phenomenon. The spruce spec-
face of the concrete slab directly at the bonding area (Fig. 6). The imens that failed due to bonding failure exhibited two different
wires of the strain gauges were transferred through drilled holes shapes of fracture surface in general. In most cases, a fracture sur-
with a diameter of 3 mm, which were placed 8 mm aside the cen- face along the growth rings appeared (see Fig. 7). A smaller part,
ter of the strain gauges. To minimise the impact of drilling the however, exhibited a fracture surface very close to the bonding
holes on the behavior of the concrete, the holes were drilled with- area but not following the direction of the growth rings (see
out hammer and with low driving speed. The spacing between the Fig. 8). Overall, the thickness of timber which was still attached
single strain gauges was chosen according to the expected strain to the concrete slab was situated between 2 mm and 25 mm in
gradients in load direction. The load was applied under displace- case of the spruce specimens. The fracture surfaces of the LVL spec-
ment control by a hydraulic jack with a displacement rate of imens were quite irregular, as shown in Fig. 9. The thickness of the
0.01 mm/s. remaining timber was between 6 mm and 21 mm in this case. The
failure of the beech specimens was characterized by a fracture in
3.3. Test results concrete, starting from the end face of the concrete plate and
changing into the timber after about a third of the bond length.
Within the documentation of the results, the loads are assigned
according to the definition in Fig. 5, which means that the load that
is introduced into the timber specimen is 2F and the loads that act
on the bonding area resp. on the steel bearing are F each.

3.3.1. Basic observations


During the bond tests, clicking noises could be heard for all
specimens from spruce and LVL, starting up from loads of about
F = 30–75 kN, depending on the length of the bond-line.
Failure occurred abruptly within all experiments. Most of the
overall 60 specimens failed – as expected – by a bond- respective
shear failure. Altogether, 10 specimens collapsed by other failure
modes, 9 by tension failure orthogonal to the bond-line and one

500
50

9. 8. 7. 6. 5. 4. 3. 2. 1.
50

Fig. 7. Fracture surface along the growth rings, specimen V32 - spruce.
10 25 50 75 90 90 75 50 25 20

300
50

7. 6. 5. 4. 3. 2. 1. LEGENDE

DMS 1-LY41-6/120
50

10 40 45 55 55 45 40 10 20

200 150

7. 6. 5. 4. 3. 2. 1. 5. 4. 3. 2. 1.
100
100

10 2 5 30 35 35 30 25 10 20 10 25 40 40 25 10 20

Fig. 6. Arrangement of the strain gauges on the UHPC slab. Fig. 8. Fracture surface close to the bond-line, specimen V38 - spruce.
M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088 3083

Table 4
Mean values of the ultimate loads and share of fracture area.

Test n(⁄) F u (kN) Share of fracture area in %


Timber Concrete Bonding area
V01–V03 2 100.0 100 0 0
V04–V06 2 155.0 35–75 15–35 10–30
V07–V09 3 91.0 100 0 0
V10–V12 1 98.0 100 0 0
V13–V15 2 152.8 40–60 15–25 15–45
V16–V18 3 87.6 100 0 0
V19–V21 1 116.0 100 0 0
V22–V24 3 93.0 100 0 0
V25–V27 2 54.1 99–100 0 0–1
V28–V30 3 59.4 100 0 0
V31–V33 2 117.5 100 0 0
V34–V36 3 114.4 100 0 0
V37–V39 3 135.6 80–100 0–15 0–5
V40–V42 3 135.2 100 0 0
V43–V45 3 150.8 80–100 0–18 0–2
V46–V48 2 148.3 98–100 0–2 0
Fig. 9. Fracture surface close to the bond-line, specimen V53 - LVL.
V49–V51 3 182.3 85–100 0 0–15
V52–V54 3 180.1 98–100 0 0–2
V55–V57 2 166.0 100 0 0
250 V58–V60 3 185.5 98 0 2
(⁄)
Number of evaluable tests with a bond failure.
200
failure loads Fu [kN]

200
150

160
100
load F [kN]

spruce 120
50 lvl
beech
V011,2 - = 150 mm
80 v

V013,4 - = 150 mm
0 v

V431,2 - v = 300 mm
0 200 400 600 V433,4 - v = 300 mm
40
joint length v [mm] V591,2 - v = 500 mm
V593,4 - v = 500 mm
Fig. 10. Failure loads in dependency of the bond length.
0
0 0.2 0.4 0.6 0.8
3.3.2. Ultimate Loads of the bond tests. average relative-displacement δrel [mm]
Fig. 10 shows the ultimate loads over bond length for all spec-
imens. In Table 4 the ultimate loads and the share of fracture sur- Fig. 11. Relative displacements between timber and concrete for the specimens
face for timber, concrete and adhesive are documented. Regarding V01, V43 and V59.

the type of timber, the beech specimens exhibited significantly


higher ultimate loads compared to the other specimens. The LVL
displacements at the bottom of the bond-line (W3 + W4). With
specimens did not show higher ultimate loads than the spruce
an increasing bond length, the relative displacements at the bot-
specimens but the deviation was – apart from the longest bond
tom became greater compared to the relative displacements at
length – considerably smaller. An influence of the type of surface
the top of the bond-line (see Fig. 11). Also, nonlinear effects in-
treatment and of the concrete mix cannot be detected. With
crease with an increasing bond length. The results of the strain
increasing bond length, the failure loads of the spruce and the
measurement will be documented and discussed within the
LVL specimens both increase in a similar range. If the mean values
numerical calculations on the bond behaviour in Section 5.
are regarded, it can be depicted that the ultimate loads do not in-
crease any more from a maximum bond length of about 400 mm.

4. Evaluation of the experimental results


3.3.3. Relative displacements between concrete and timber
The load over relative displacement behaviour of the specimens For the evaluation of the bonding strength according to Volker-
depends on the kind of timber and on the bond length. For the sens theory, the geometrical dimensions of the bond partners, their
beech specimens, the behaviour is linear until the failure and over- Young’s modules, the fracture energy Gf, the local shear strength sf
all very stiff. In case of the spruce resp. LVL specimens, slight non- and a definition on the shape of the stress–strain relation g are re-
linearities can be observed for the short bonding lengths. This quired. In the following investigations, only a linear shape of the
observation conforms to the clicking noises from the spruce and constitutive relation (Fig. 3, type I) of the bond shall be adopted.
the LVL specimens starting at a low load level. Aside from these parameters, the capacity of the joint is also
At the short bonding lengths, the displacements that were dependent on the thickness of the active part of timber tw, as Pört-
observed at the top of the bond-line (W1 + W2) were close to the ner documented comprehensively [22].
3084 M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088

The geometrical dimensions of the bond partners and their 250


ultimate load
Young’s modules are well known. The remaining three parameters according to LEFM
(Gf, sf and tw) cannot be estimated directly. Based on that reason, a
parametric study will be carried out in the following, to identify 200
these three parameters.

ultimate load F [kN]


In the case of spruce and LVL, almost 100 % of the failure was
150
located in the timber. Hence, the values of Gf and sf for the para-
metric study were chosen on the basis of experimental data for
softwood from literature. For the local shear strength of spruce, 100 Spruce
values between sf = 3.6 N/mm2 and sf = 15.6 N/mm2 can be found lvl
(cf. [27], [29] and [16]). On the background of this variety of data, τf = 14 N/mm2

a range of values of 10 N/mm2 ±4 N/mm2 is defined for the follow- 50 τf = 10 N/mm2

ing study. τf = 6 N/mm2

For the fracture energy of spruce, Aicher et al. [2] identified val-
0
ues between Gf = 0.56 N/mm and Gf = 1.00 N/mm. Hence, for the
0 200 400 600
present work a range of values of 1.0 N/mm ± 0.4 N/mm is deter-
mined. The upper value is chosen slightly higher than documented joint length v [mm]

in literature, because the bond between timber and UHPC is ex-


Fig. 13. Ultimate loads over joint length: Variation of shear strength sf.
pected to exhibit a higher fracture energy than timber only
specimens.
The active timber thickness will be varied between tw = 60 mm
and tw = 100 mm. That corresponds to the values which were iden- 250
tified by Pörtner [22] for glued-in CFK rods. An overview of the val- ultimate load
ues for the varied parameters is given in Table 5. In the following according to LEFM
200
diagrams (Fig. 12–14 ), the results of the parametric study are doc-
ultimate load F [kN]

umented and compared to the experimental data. If the three dia-


grams are looked at, it is obvious, that the analytical solution 150
overestimates the experimental failure loads for short bond
lengths in almost all cases. This can be explained due to tension
stresses which are introduced into the specimens because of the 100 Spruce
shape of the test setup orthogonal to the bond-line. This phenom- lvl
tw = 100 mm
enon stands in good agreement to the tension failures that were
50 tw = 80 mm
only observed for short bond lengths and it will be studied in more tw = 60 mm
detail within the numerical investigations in the following section.
For the fracture energy Gf, the median value of 1.0 N/mm2 delivers 0
the best correlation to the experimental data. For the lowest value
0 200 400 600
joint length v [mm]

Fig. 14. Ultimate loads over joint length: Variation of active timber thickness tw.
Table 5
Values for the parametric study.
(Gf = 0.6 N/mm2), the analytical match the experimental failure
Parameter Range of values
loads very well for medial bond lengths. But the failure loads are
Gf (N/mm) 1.0 ± 0.4 overestimated for short bond lengths and underestimated for long
sf (N/mm2) 10 ± 4
bond lengths . Consequentially, a fracture energy of Gf = 1.0 N/mm2
tw (mm) 80 ± 20
is defined for further calculations.
In case of the variation of the local shear strength, the compar-
ison of the theoretical and the experimental results suggests that a
250
shear strength of 6 N/mm2 is adequate. On the other hand, preli-
ultimate load minary shear tests delivered a shear strength of the bond of about
according to LEFM
11 N/mm2 ([23]), and the average local shear strengths of spruce
200 which were found in literature are broadly higher. Hence it can
be assumed that there is some considerable influence of the al-
ultimate load F [kN]

ready mentioned tension stresses. Based on these considerations,


150 the median value of the local shear strength of sf = 10 N/mm2 will
be taken for further considerations. For the active timber thickness,
an amplitude of tw = 80 mm is chosen, because the smaller and
100 Spruce
lvl higher considered values lead to an under- resp. an overestimation
Gf = 0,6 Nmm/mm2 of the failure loads.
50 Gf = 1,0 Nmm/mm2
Gf = 1,4 Nmm/mm2
5. Numerical modelling
0
0 200 400 600 The finite element method (FEM) was adopted to get more
joint lengh detailed information on the stress–strain relation of the bond. In
v [mm]
particular the influence of tension stresses perpendicular to the
Fig. 12. Ultimate loads over joint length: Variation of Fracture Energy Gf. grain – which are dependent on the magnitude of friction that
M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088 3085

appears at the front side of the concrete slab between concrete and
steel bearing – should be investigated. Hence the magnitude of the
friction coefficient at that contact area will be varied, and its im-
pact on the tension stresses will be studied.

5.1. Development of the numerical model

For the numerical investigations, the commercial software ABA-


QUS 6.5-6 [14] was used. The geometry of the test specimens was
discreticized by isoparametrical 3D-elements with eight nodes and
reduced integration (C3D8R). The load was applied – as in the
tests – into the specimen through a steel plate. This steel plate
was also discretized with the above mentioned 3D-elements. The
load was applied as a compression stress which was equivalent
to the particular situation into a circular area of 15 mm diameter.
The contact between the concrete plate and the bearing was
simulated by different contact conditions. Orthogonal to the con-
tact zone, a load transfer was realized by a rigid coupling. Tangen-
tial to the contact area, the coupling was simulated by an
Fig. 15. Mesh geometry for the numerical investigations, bond length 150 mm.
interaction property with different magnitudes of friction coeffi-
cients. Firstly a friction coefficient of l = 0.5 was chosen. Further-
more, the limit cases of l = 0 and l = 1 were considered as well,
2.4
to evaluate the influence of friction. The steel bearing at the bottom
of the concrete slab was modeled by a rigid plane which was cou-
FEM 100 %
pled to a reference point. The boundary conditions were then ap- DMS 100 %
plied on to this reference point. This approach has the advantage FEM 50 %
DMS 50 %
that the resulting force can be taken directly at the reference point. 1.6 FEM 10 %
Concerning the constitutive relations, the adhesive as well as DMS 10 %
Strain ε [‰]

the UHPC and the steel were regarded as isotropic materials. For
the numerical calculations, only the specimens with glued lami-
nated timber specimens will be considered. An orthotropic consti-
0.8
tutive relation was chosen, because the growth rings of the glued
laminated timber are orientated almost into the direction of the
longer face of the lamellas. All materials were considered as linear
elastic. Material properties are given in Table 6. For the modelling
of the bond tests, symetry was considered and only one quarter of 0
the specimen was discreticized with the adequate boundary condi- 0 50 100 150 200
tions. As the basic model, the specimen geometry for a bond length Joint Length [mm]
v
of 150 mm was examined first. The mesh geometry of the numer-
ical model is shown in Fig. 15. The bond-line was discreticized with Fig. 16. Normal strain e11 in the UHPC-Slab adjacency to the bond-line, experi-
three elements over the joint thickness of 1.5 mm. The global axes mental and numerical results.

for all strains and stresses are also given in Fig. 15. Overall the
model contains 42000 elements and 157000 degrees of freedom.
documented. This comparison of calculated and measured strains
The models for increasing bond lengths were generated analo-
showed a good agreement as Fig. 16 depicts.
gously and the element size was chosen as for the bond length of
Furthermore, the relative displacement at the position of the
150 mm.
displacement transducers W3 and W4 was calculated with the
numerical model and compared to the experimental results. It
was found that the gradient of the load–relative displacement
5.2. Model validation
curve of the test result was almost identical to the numerical
result.
To validate the numerical model, strains, reaction forces and
displacements were calculated and compared to the experimental
results. 5.3. Investigation on the stress interaction of shear and tension
At first, the plausibility of the reaction force at the reference
point was controlled. From the four tests V03, V31, V43 and V57, Tension stresses orthogonal to the bond-line seem to affect the
strains along the sandblasted surface of the concrete plate are bonding strength for different bond lengths (cf. Sections 3 and 4).

Table 6
Material properties for the numerical investigations.

E11 (N/mm2) E22 (N/mm2) E33 (N/mm2) m12 (–) m13 (–) m23 (–) G12 (N/mm2) G13 (N/mm2) G23 (N/mm2)
Timber 11,000 800 450 0.5 0.31 0.24 740 620 42
UHPC 50,000 – – 0.2 – – – – –
Adhesive 38,000 – – 0.4 – – – – –
Steel 210,000 – – 0.3 – – – – –
3086 M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088

The numerical model was used to investigate this effect in more 10


detail. For this effort, the friction coefficient in the contact area be-
tween the loading area of the concrete slab and the steel bearing V300
was varied, and the shear stresses and the tensile stresses in timber 8

Shear Stress S12 [N/mm2]


V150 V200
close to the bond-line were calculated. V400
V250 V100
As upper and lower boundaries, friction coefficients of l = 0 and 6
l = 1 were regarded. Between the concrete and the steel bearing, a V500
thin LVL plate was situated during the bond tests and the sliding
was observed between the LVL plate and steel. In literature, a fric- 4
tion coefficient of about l = 0.5 can be found for the combination
Failure Criterion
‘‘steel-timber’’ (cf. [11], [17] and [6]). Hence, this value was re- according to Norris
2
garded as a third variant. The Figs. 17 and 18 show the results of Results of the experi-
the investigation for the bond lengths of 100 mm and 400 mm. mental Investigations
The two diagrams depict that the maximum shear stresses (S12) 0
occur at the side of the load application at about 10 mm distance
0 0.6 1.2 1.8
from the begin of the bonding area, for all cases. The maximum
Tensile Stress S22 [N/mm2]
shear stresses increase if the friction coefficient increases.
The stresses perpendicular to the grain are compression at the Fig. 19. Interaction of maximum shear and tension stresses - results from Norris
beginning of the bond-line in all cases and increase with the Criterion versus calculation.
bond-line, until they reach a maximum at about 20–40 mm from
the begin of the joint. The peak values of the tension stresses de-
crease as expected with an increasing friction coefficient. In the The tensile strengths of spruce, which are documented in liter-
case of a rigid coupling (l = 1), no tensile stresses occur. ature, vary between f90,t = 1.1 N/mm2 and f90,t = 2.5 N/mm2 (cf.
[15], [4] and [10]). The evaluation of the numerical results of all
investigated bond lengths shows that the tension stresses for a fric-
8 tion coefficient of l = 0.5 are situated in the lower range of these
S21 values. For the longer bond lengths (lv = 300–500 mm) however,
they are far below these limits.
4 Based on the previous calculations, the interaction of the max-
imal shear stresses with the maximal tensile stresses for all bond
S12 and S22 [N/mm2]

S22 lengths is plotted in Fig. 19. The positions of the maximum shear
0
stresses and the maximum tensile stresses were not exactly iden-
tical within the numerical result. But taking into account the non-
-4 linear effects of the bond behaviour (cf. Fig. 11), it seems justified
μ=0 to compare the maximum values of shear and tensile stresses.
μ = 0,5 Additional to the numerical results, the failure criterion of Nor-
-8 μ=∞ ris [19] is plotted in Fig. 19 for a shear strength of fv = 8 N/mm2 and
a tension strength of f90,t = 1.4 N/mm2. For these strengths, a good
agreement between the numerical results and the Norris Criterion
-12
can be stated, except for the longest bond length.
0 20 40 60 80 100 As a preliminary result it can be remarked that the bond tests
Joint Length v [mm] can be described by Volkersens theory very well, and that the bond
length from which no increase of the failure loads occurs is about
Fig. 17. Development of shear (S21) and tension stresses (S22) along the bond-line,
400 mm. The influence of the tension stresses orthogonal to the
bond length = 100 mm.
bond-line can be neglected from about 300 mm on. Hence the ap-
plied test setup for the bond tests is proposed as a standard test
method to estimate the strength of bonded timber concrete com-
8 posites, with a bond length of 400 mm.

S21
6. Conclusions for the design of bending members
4
S12 and S22 [N/mm2]

Based on the previous investigations, the following conclusions


S22
0 for the basic design of bending members from timber und UHPC
can be drawn. Due to the rigid bond that is obtained by glued con-
nections of timber and UHPC, it is recommended to adopt the c-
-4
μ=0 method according to DIN 1052 [8] for the design of the hybrid
bending members, with a value of c = 1 for rigid bonding.
μ = 0,5
As the bond tests showed, the bond failure of the test specimens
-8 μ=∞ occurred almost exclusively in timber (except for the beech speci-
mens). Hence the bond failure can be regarded as a shear failure of
-12 the timber surface and the design of the bond-line results in a
0 100 200 300 400 shear design of timber in the cases of spruce and LVL.
Considering the shear design, the results of the present investi-
Joint Length v [mm]
gations suggest that the shear strength of glued laminated timber
Fig. 18. Development of shear (S21) and tension stresses (S22) along the bond-line, from DIN 1052 [8] (fv = 2.5 N/mm2) can be applied. The results of
bond length = 400 mm. the bond tests show that this value is far on the safe side. But
M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088 3087

because of the limited number of available tests, until now there is the ultimate loads, the failure modes and the global stiffnesses of
no basis to allow higher shear stresses. the hybrid members, will be documented. More detailed results
One further parameter, which is needed if the c-method is ap- are to be found in [25].
plied, is the active width of the concrete slab beff. Numerical inves- The two specimens were loaded displacement controlled with a
tigations depicted that – for common distances in between single displacement rate of 0.1 mm/s up to the global failure. Both spec-
timber beams – the whole width of the concrete slab is involved imens showed linear elastic behaviour until the ultimate load. In
homogeniously in load carrying. Hence the effective width beff is Table 7, the main results of the two bending tests are documented.
equivalent to the distance between two beams. This numerical re- The estimated stiffnesses, failure criterion and ultimate loads were
sult will be proven within the following bending tests. calculated on the basis of the c-method, for a concrete E-module of
50,000 N/mm2. The E-modules of the glued laminated beams were
identified experimentally according to DIN EN 408 [9].
7. Full-scale testing The results in Table 7 show an overall acceptable agreement be-
tween experimental findings and the applied theory, if the com-
The basic design recommendations will be verified by two full- mon variation of the material parameters of timber and concrete
scale bending tests. The geometry of the bending members was se- is considered. The experimental global stiffness of the two bending
lected on the background of preliminary investigations regarding members is very close to values as calculated.
the stress distribution in the bending members [24] and is depicted Concerning the failure mode, both bending members failed – as
in Fig. 20. The cross-section consists of two glue lam beams (Gl24h) expected – due to a tension failure of timber in the outermost la-
with a width of 100 mm and a height of 200 mm both and a con- mella of the glue-lam beam, between the to load application
crete plate (20 mm thickness, UHPC from Table 2) that is bonded points, starting from a knot. The achieved ultimate loads were
to the glue-lam beams. The adhesive which was applied was the broadly higher than the expected loads, which is self evident be-
same epoxy resin which was already used for bonding tests in sec- cause the characteristic tension strength of 24 N/mm2 from DIN
tion 3 (Sikadur 330). Based on the findings from Section 3, the 1052 [8] was employed for the calculation. To be able to interpret
UHPC-surface was sandblasted and the timber surface was planed. the magnitude of the failure loads, a mean failure load was calcu-
The length of the members is 4.0 m, which leads to a span of lated, considering a typical mean tension strength of Gl24h of
3.84 m. 40 N/mm2 (e.g. [7]). For this mean shear strength, the calculated
The test setup was a four-point bending test according to DIN ultimate load is about Fu,m,cal = 170 kN for both members. If the test
EN 408 [9] (see Fig. 21). Well instrumentation was installed to con- results are regarded, it can be stated that one ultimate load was
sider absolute and relative deformations and strains at different higher and one smaller than this mean ultimate load.
locations. In the following, only the main results of the tests, like If the complete results of the two bending tests are regarded, it
can be stated that the assumptions that were made for the design
of hybrid bending members from timber and UHPC seem to be
legitimated.

8. Summary and prospects

Within the present work, extensive theoretical and experimen-


tal investigations on the bond between timber and UHPC were car-
ried out. On the basic of the results of these investigations,
assumptions for the design of bending members were made which
were finally verified by two bending tests.
At first, theoretical models were documented that are able to
Fig. 20. Cross section for the bending tests.
describe the behaviour of the bond between timber and concrete.
In this context, on the one hand the theory of Volkersen seems
applicable, and on the other hand, LEFM-formulations are able to
predict the joint strength above a certain bond length.
In a second step, experimental investigations on the bond
behaviour were carried out. The tests depicted that a failure of
the bond in timber close to the bond-line is relevant in almost all
cases, and that nonlinear effects in the load–relative displacement
curve occur with increasing bond length. The results of the bond
tests were in good agreement with the theoretical considerations.
But in the case of short bond lengths, tension stresses affected the
bond behaviour and the experimental ultimate loads were lower
than the estimated ones.
With the help of numerical investigations, this effect was stud-
Fig. 21. Test setup for the bending tests. ied and it could be allocated that tension stresses orthogonal to the

Table 7
Results of the two bending tests in comparison theoretical estimated values.

Test (EI)ef,cal (N/mm2) (EI)ef,exp (N/mm2) (EI)ef,cal/(EI)ef,exp (–) Failure mode Fu,k,cal (kN) Fu,exp (kN)
Theory = experiment
BP1 5.138  1012 4.696  1012 1.09 Tension in timber 113.4 130.0
BP2 5.487  1012 5.001  1012 1.10 Tension in timber 112.6 188.0
3088 M. Schäfers, W. Seim / Construction and Building Materials 25 (2011) 3078–3088

bond-line are secondary for bond lengths beyond 300 mm. The elasto-plastischen Werkstoffmodells. Holz Roh Werkst Heft 2007;65:
S.301–31.
missing parameters for the bond theory were established and a
[11] Guan N, Thunell B, Lyth K. On the friction between steel and common swedish
characteristic bond length of 400 mm was evaluated. wood species. Holz Roh Werkst, Heft 1983;41:S.55–60.
On the basis of the previous investigations it was emphasized [12] Gustafsson PJ. Serrano E, Aicher S, Johansson C-J. A strength design equation
that the c-method can be applied for the design of bending mem- for glued-in rods. In: International symposium on joints in timber structures;
Band 22 der Reihe Rilem Proceedings, Stuttgart; 2001.
bers, considering the bond as a stiff connection. The design of the [13] Haiman M, Rak M, Krolo J, Herceg J, Calogovic V. EPS concrete composite
bond can be carried out through a shear design of timber close to structures – lab testing and FEA modelling. In: Proceedings of the international
the bond applying the corresponding shear strengths that are given conference on experimental mechanics, Conf 12, Milan, Italy; 2004.
[14] Hibbit, Karlsson, Sorensen. ABAQUS/CAE, Version 6.5. 1080 Main Street,
in the design codes or the approvals of the materials. Finally, these Pawtucket, RI 02860-4847, USA.
assumptions could be verified by two bending tests. [15] Logemann M, Schelling W. Die Bruchzähigkeit von Fichte und ihre
A field of interest for further investigations is the long term wesentlichen Einflussparameter – Untersuchungen im Mode-2. Holz Roh
Werkst 1992;50:S.117–21.
behaviour of the bond under changing thermal and humidity con- [16] Morel S, Valentin G. Size effect in crack shear strength of wood. J Phys IV –
ditions, with a focus on internal stresses due to the swelling and Colloq C6, Suppl J Phys III 1996;6:S.385–94.
contracting behaviour of timber and concrete. Also the vibration [17] Möhler K, Herröder W. Obere und untere Reibbeiwerte von sägerauhem
Fichtenholz, Holz als Roh- und Werkstoff. Heft 1979;37:S.27–32.
characteristics and the behaviour of bending members under [18] Negrao JH, Oliveira FM, Oliveira CL. Investigation on timber–concrete glued
asymmetric loads should be investigated in more detail in the composites. In: 9th World conference on timber engineering, Portland, OR,
future. USA; 2006.
[19] Norris CB. Strength of orthotropic materials subjected to combined stresses.
Report 1816, Forest Products Laboratory, Madison, WI, USA; 1962.
[20] Pincus G. Behaviour of wood–concrete composite beams. J Struct Div, Proc Am
References Soc Civ Eng 1970:S.2009–19.
[21] Pincus G. Epoxy bonded wood–concrete and corrugated steel–concrete
[1] Aicher S. Structural adhesive joints including glued-in bolts. In: Thelandersson composite. In: International conference on materials of construction for
S, Larsen HJ, editors. Timber engineering. West Sussex, England: John Whiley & developing countries, Bangkok, Thailand;1978. p. S.1003–15
Sons; 2003. p. S.333–63. [22] Poertner C. Untersuchungen zum Verbund zwischen eingeklebten stiftförmigen
[2] Aicher S, Boström L, Gierl M, Kretschmann D, Valentin G. Determination of faserverstärkten Kunststoffen und Holz. Dissertation, Schriftenreihe
fracture energy of wood in mode II. Technical report – RILEM TC 133. Swedish Bauwerkserhaltung und Holzbau, Heft 2, Universität Kassel; 2006.
National Testing and Research Institute; 1997. [23] Schäfers M, Seim W. Development of adhesive bonded timber–UHPC
[3] Aicher S, Reinhardt HW. Einfluss der Bauteilgröße in der linearen und composites – experimental and theoretical investigations. In: 10th World
nichtlinearen (Holz-)Bruchmechanik. Holz Roh Werkst 1993;51:S.215–20. conference on timber engineering, June 2–5, 2008, Miyazaki, Japan; 2008.
[4] Blass HJ, Schmidt M. Querzugfestigkeit von Vollholz und Brettschichtholz. Holz [24] Schäfers M, Seim W. Development of adhesive-bound UHPC–timber
Roh Werkst 2001;58:S. 456–46. composites. In: Proceedings of the second international symposium on ultra
[5] Brunner M, Schnüriger M. Holz-Beton-Elemente mit Klebeverbund. high performance concrete; March 05–07, 2008. p. S.259–66
Forschungsbericht Nr. 2637-SB-01, Hochschule für Architektur, Bau und Holz HSB, [25] Schäfers M. Entwicklung von hybriden Bauteilen aus Holz und hochfesten
Burgdorf, Biel; 2005. bzw. ultra-hochfesten Betonen – Experimentelle und theoretische
[6] Böge A. Technische Mechanik; Statik, Dynamik, Fluidmechanik, Untersuchungen. Dissertation, Universität Kassel; 2010.
Festigkeitslehre. Friedrich Vieweg und Sohn Verlag; 2006. [26] Smith I, Landis E, Gong M. Fracture and fatigue in wood. John Wiley & Sons,
[7] Denzler JK. Modellierung des Größeneffektes bei biegebeanspruchtem Ltd.; 2003.
Fichtenschnittholz. Dissertation, TU-München; 2007. [27] Ukyo S, Masuda M. A new method for measuring the true shear strength of
[8] DIN 1052: Entwurf, Berechnung und Bemessung von Holzbauwerken – wood. In: 9th World conference on timber engineering, Portland, OR, USA;
Allgemeine Bemessungsregeln und Bemesungsregeln für den Hochbau 2004- 2006.
08. [28] Volkersen O. Die Schubkraftverteilung in Leim-, Niet- und Bolzenverbindungen
[9] DIN EN 408: Holzbauwerke – Bauholz für tragende Zwecke und Brettschichtholz – Teil 1. EnergieTechnik 1953;3:S.68–71.
– Bestimmung einiger physikalischer und mechanischer Eigenschaften 2004-08. [29] Wernersson H, Gustafsson PJ. The complete stress–slip curve of wood-
[10] Fleischmann M, Krenn H, Eberhardsteiner J, Schickhofer G. Numerische adhesives in pure shear. In: Verchery G, Cardon AH (Herausgeber).
Berechnung von Holzkonstruktionen unter Verwendung eines orthotropen Mechanical behaviour of adhesive joints; 1987. p. S.139–50.

You might also like