Construction and Building Materials 265 (2020) 120193

Contents lists available at ScienceDirect

Construction and Building Materials

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

Mechanical properties of a balsa wood veneer structural sandwich core

material
Chao Wu a,b, Niloufar Vahedi b, Anastasios P. Vassilopoulos b, Thomas Keller b,⇑
a
School of Transportation Science and Engineering, Beihang University, 37 Xueyuan Road, Beijing 100191, China
b
Composite Construction Laboratory (CCLab), École Polytechnique Fédérale de Lausanne (EPFL), Station 16, Bâtiment BP, CH-1015 Lausanne, Switzerland

h i g h l i g h t s

Orthotropy on the material scale and system scale influenced the mechanical behavior.

The configuration of the veneer layers strongly affected the specimen responses.

The adhesive between the veneer layers did not affect the mechanical behavior.

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

Article history: Balsa wood is an appropriate core material for structural sandwich applications due to its high strength-
Received 6 August 2018 and stiffness-to-weight ratios. However, the mechanical properties vary considerably owing to the inher-
Received in revised form 18 June 2020 ent scattering of natural wood materials. One approach to reduce this scatter and tailor the mechanical
Accepted 8 July 2020
properties according to specific application needs is to recompose the natural material into a veneered
Available online 5 August 2020
material consisting of veneer layers of different grain orientations, which are adhesively bonded together.
The mechanical properties of such a veneered balsa wood, composed of alternating 0°/90° grain orienta-
Keywords:
tions, were investigated at ambient temperature according to corresponding standards. The properties
Balsa wood
Tensile properties
were significantly influenced by the orthotropy on the material scale within one veneer layer and on
Compressive properties the system scale within the assembled veneer layers. Standardized experimental set-ups and specimen
Shear properties geometries may produce artifacts such as buckling or strain hardening which deviate from the material
Sandwich panel behavior in real structures. The thin adhesive between the veneer layers did not negatively affect the
Core material mechanical behavior since failure occurred within the veneer layers and not in the interfaces.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction sheets) to resist indentation from heavy truck wheel loads. The
balsa wood core provides continuous support for the face sheets
Balsa is one of the lightest wood species used in structural in this case, thus avoiding stress concentration problems in the
applications, and due to its high strength- and stiffness-to- intersection of the face sheets and webs of multicellular decks
weight ratios is a preferred material for sandwich panel cores [1– [10,11].
3]. Such panels are found in major load-bearing applications, e.g. Orthotropic balsa wood used as core material in sandwich pan-
hulls of ships and aircraft [4–5] or in the rail, wind energy, aero- els is subjected to various stress states and a knowledge of the
space, and defense industries [6]. More recently, balsa wood has mechanical properties in all loading directions is therefore of crit-
been used as a core material for bridge decks [7–9] in bridge con- ical importance for appropriate design. Kotlarewski et al. [3] inves-
struction. One example is the glass fiber-reinforced polymer tigated the mechanical properties of balsa wood sourced from
(GFRP)-balsa sandwich deck installed on a road bridge across the Papua New Guinea. Static bending, hardness, compression (parallel
Avançon river in Bex, Switzerland [8]. The sandwich structure is and perpendicular to the grain) and shear (parallel to grain) exper-
composed of GFRP face sheets and a balsa wood core with end- iments were conducted. Specimens were cut from sapwood, pre-
grain orientation (with the wood fibers transverse to the face conditioned to 12% moisture content, and investigated according
to ASTM D143-09 [12]. This study indicated that the mechanical
⇑ Corresponding author. properties of balsa wood were closely related to its density. How-
E-mail addresses: [email protected] (C. Wu), [email protected] ever, high density did not guarantee a superior strength value.
(T. Keller). Specimens with a density range of 140–150 kg/m3 exhibited better

https://doi.org/10.1016/j.conbuildmat.2020.120193
0950-0618/Ó 2020 Elsevier Ltd. All rights reserved.
2 C. Wu et al. / Construction and Building Materials 265 (2020) 120193

mechanical properties compared to samples with lower and higher

densities. It was also noted that balsa wood from Papua New Gui-
nea exhibited slightly lower strength and stiffness than balsa
sourced from South America in terms of modulus of elasticity
and shear strength. Chowdhury et al. [13] also conducted experi-
ments concerning the characterization of mechanical properties
of balsa wood originating from Bangladesh. This research revealed
that the mechanical properties of balsa wood were mostly related Fig. 1. Balsa wood product BALTEKÒ VBC with veneer layers alternatingly oriented
in 0°/90° grain directions.
to the air- dry density of balsa. Osei-Antwi et al. [2] characterized
the shear properties of end-grain balsa panels from Ecuador. The
effects of the shear plane, density and adhesive joints between
lumber blocks on the shear properties of the balsa panels were Specimens were cut from the 1200  150  100-mm BALTEKÒ
evaluated. Similarly, the authors found that shear stiffness and VBC veneer blocks. According to the manufacturer, the average
strength increased with the increasing density of the balsa. Some density of BALTEK VBC is 180–240 kg/m3.
specimens exhibited significant ductility due to plastic deforma- The specimens were stored in a condition room with a constant
tions in the tracheids. temperature of 20 °C and a relative humidity of 65% for at least two
The above-mentioned studies were all conducted using balsa months to ensure that the moisture content of the specimens was
blocks or lumbers. Similar to other wood types [14], the mechan- equilibrated. The moisture content was then measured according
ical properties of balsa wood exhibit significant variations in the to ASTM 4442–07 [18] using cubic specimens of
radial direction of the tree’s cross sections and thus considerable 30  30  30 mm3 consisting of five veneer layers. The weights
scattering, which causes difficulties in the design of sandwich of the cubes were measured immediately after they were taken
panels incorporating a balsa core. In order to reduce this scatter, out of the condition room. Subsequently, the cubes were dried in
a new balsa product was developed for the deck of the above- an oven at a temperature of 105 °C for at least 48 h and then
mentioned Avançon bridge [8]. This balsa core consists of thin weighed. A total of ten cubes were measured; the average mea-
veneer layers bonded together with an adhesive. All veneer layers sured moisture content was 9.1% with a standard deviation of
are oriented in the same end-grain direction. The substitution of 0.27%. The average density before and after drying was 186.4 ± 11.
the larger lumber blocks by thin veneer layers allowed the reduc- 4 kg/m3 and 170.8 ± 10.9 kg/m3, respectively.
tion of density variation thanks to appropriate material selection,
thus reducing the scatter of the mechanical properties. Shir
2.2. Tensile experiments
Mohammadi et al. [15] compared the fracture toughness of this
laminated veneer lumber (LVL) balsa to the toughness of conven-
Three configurations of tensile specimens, T1, T2 and T3, were
tional balsa block material. The research demonstrated that LVL
prepared as shown in Fig. 2. In configuration T1 the load was
balsa material with a good lamination adhesive has improved
applied in the transverse-to-the-layers’ direction while in configu-
toughness and fiber bridging effects compared to bulk balsa.
rations T2 and T3 the load was parallel to the layers, i.e. in T2 par-
However, no systematic investigations and characterization of
allel to three 90°/0°/90° and in T3 parallel to three 0°/90°/0° layers.
the mechanical properties of such balsa wood veneers exist to
The detailed dimensions, shown in Fig. 3(a), were determined
date.
based on ASTM D3500-14 [19]. A special fixture was manufactured
This paper presents a comprehensive experimental study of
to apply the tensile load to the soft specimens, see Fig. 3(b). The
the tensile, compressive and shear properties of a veneered balsa
tensile experimental set-up is shown in Fig. 3(c).
wood. The material consisted of a series of veneer layers with
All experiments were conducted on an MTS Landmark 25 kN
alternating 0°/90° grain orientations. A total of 68 experiments
Servohydraulic Material Test System provided by MTS Systems
were conducted to consider all possible loading directions that
GmbH (Berlin, Germany) with a capacity of 25 kN. The built-in load
could be developed in the core of a sandwich panel. To achieve
this, three tensile, three compressive and four shear specimen
configurations were investigated. The failure modes, stress–strain
curves and modulus and strength values were reported. Finally,
the design-based 5% characteristic values, according to Eurocode
recommendations, were obtained for each configuration and com-
pared with corresponding values derived from statistical
distributions.

2. Experimental investigation

2.1. Materials and moisture content

The balsa wood used in the present study was BALTEKÒ VBC
provided by 3A Composites Core Materials (Sins, Switzerland)
[16]. This product consists of balsa veneer layers, which were pro-
duced by a rotary peeling process of balsa trunks using a roller
pressing bar. Detailed information about the rotary peeling process
can be found in [17]. Each veneer layer had a nominal thickness of
6 mm. They were alternatingly stacked in 0°/90° grain orientations,
as shown in Fig. 1, bonded together with a one-component cold-
curing and foaming polyurethane (PU) adhesive, Jowapur 687.22,
compressed, and cured at room temperature for at least 24 h. Fig. 2. (a) Front view and (b) side view of tensile specimens T1, T2, T3.
 C. Wu et al. / Construction and Building Materials 265 (2020) 120193 3

Fig. 3. (a) Dimensions of tensile specimens, and (b) fixture for applying tensile load (in mm), (c) experimental set-up of tensile experiments.

cell was calibrated to 20% of its full capacity. A video extensometer configurations C2 and C3; the dimensions were determined
system comprising a 10-bit Sony XCLU1000 CCD camera connected according to ASTM D3501-05 [20]. A special fixture was designed
to a Fujinon HF35SA-1 lens (with a focal length of 35 mm and an to load the specimens in compression via the tensile load of the
aperture of f1.4–22) was used to measure the axial strains. Black machine, see Fig. 5(a). Stoppers were used at both ends of the spec-
target dots of 1.5-mm diameter were applied on the specimen sur- imen to block any horizontal movement. The experimental set-up
face with a gage length of 24 mm (T1) and 30 mm (T2, T3), see of the compressive experiments is shown in Fig. 5(b).
Fig. 2(b). The video extensometer system captured the relative The same machine and video extensometer system were used
movements of the three pairs of target dots, which were measured as for the tensile experiments. Pairs of target dots with a gage
with an accuracy of ± 0.005 mm. The load obtained from the MTS length of 38 mm (50 mm) were applied to C1 (C2, C3) specimens,
load cell and the movement of the targets were recorded at a fre- see Fig. 4(a). Measurement frequency and displacement rate were
quency of 10 Hz. The experiments were performed at a displace- the same as for the tensile experiments. Eighteen compressive
ment rate of 1 mm/min. A total of 30 tensile experiments were experiments were conducted, i.e. six experiments for each of the
conducted, with 10 specimens in each direction. three configurations.

2.3. Compressive experiments

2.4. Shear experiments
Similar to the tensile specimens, three configurations of com-
pressive specimens, C1, C2 and C3, were cut from the veneer The possible shear stress directions and corresponding four
blocks, as shown in Fig. 4. Five veneer layers were considered in configurations of shear specimens are shown in Figs. 6 and 7. The
dimensions of the shear specimens were defined according to
ASTM D5379-12 [21]. S1 specimens consisted of two veneer layers
(0°/90°) with a corresponding thickness of 12 mm. The same thick-
ness was adopted for the remaining configurations.
A special shear-loading fixture was manufactured to transfer
the tensile load of the machine to a shear load on the specimen,
see Fig. 8(a). The experimental set-up of the shear experiments is
shown in Fig. 8(b).
The loading and video extensometer systems, displacement rate
and measurement frequency were the same as in the other exper-
iments. Four target dots were applied on the shear specimens as
shown in Fig. 7(a). In S1, S3 and S4 they were located at the four
corners of a 12  12 mm2 square area around the center of the
specimen. For S2, the size of the square area was 6  6 mm2, to
ensure that the target dots were on the same veneer layer. The
shear strain,c, was obtained from the deformed shape A0 B0 C 0 D0 ,
see Fig. 9, as follows [2]:
c ¼ a þ b, where a ¼ AA
0 0
AC
and b ¼ DD
CD
Considering that angles a and b are small, it is safe to assume
that cosa = a and cosb = b. The shear stress was calculated from
the load divided by the cross-sectional area between the notches.
A total of twenty shear experiments were conducted, i.e. five
Fig. 4. (a) Configurations of compressive specimens C1, C2 and C3, (b) dimensions. experiments for each of the four configurations.
4 C. Wu et al. / Construction and Building Materials 265 (2020) 120193

Fig. 5. (a) Fixture for applying compressive load, (b) experimental set-up of compressive experiments.

Fig. 6. Schematic of potential shear specimen configurations and shear stress directions.

3. Experimental results and discussions 3.1.2. Tensile stress–strain curves, modulus and strength
Selected representative tensile stress–strain curves of T1, T2
3.1. Tensile properties and T3 specimens are shown in Fig. 11. The strains are average val-
ues of the three measurements in the transverse direction. All
3.1.1. Failure modes specimens exhibited a linear elastic behavior under tensile loading.
The failure modes of the tensile specimens are shown in Fig. 10. The elastic modulus was derived from the slope of the curve over a
T1 specimens failed in one of the veneer layers in the transverse strain range of 0.001–0.003 according to ASTM D3039 [22] and the
grain direction. In the T2 and T3 specimens, tensile failures peak stresses were considered as the tensile strength. The corre-
occurred in the 90° veneer layer (with grain direction transverse sponding results are summarized in Table 1.
to load), while shear failures were observed in the 0° veneer layers Configuration T1 exhibited a much lower average modulus
(with grain direction parallel to load). and strength than T2 and T3 since the properties were dominated
 C. Wu et al. / Construction and Building Materials 265 (2020) 120193 5

3.2. Compressive properties

3.2.1. Failure modes

The failure modes of C1, C2 and C3 specimens are shown in
Fig. 12. Specimens C1 failed during buckling, exhibiting the typical
buckling mode of a column fixed on both sides. Correspondingly,
second-order bending moment peaks developed at (approxi-
mately) mid-height and the fixation points, which led to crack ini-
tiation on the tensioned side transverse to the grain at these three
locations. The cracks initiated and subsequently propagated in the
veneer layer and not in the adhesive interface between two veneer
layers.
C2 specimens, exhibiting the 90° layers on the outer sides, also
failed during buckling and associated seconder-order bending. A
series of shear cracks appeared in the mid 90° veneer layer due
Fig. 7. (a) Configurations of shear specimens S1, S2, S3 and S4, (b) dimensions.
to the shear stresses that developed during the bucking process.
This damage reduced the composite action between the adjacent
layers which, compared to the C1 configuration, changed the buck-
ling mode to that of a column with hinged supports, i.e. a rotation
at the supports occurred. Final failure of all C2 specimens occurred
by fracture of the outer veneer layer on the tension side and crush-
ing on the compression side at approximately mid-height.
C3 specimens were more compact due to the greater number of
0° layers that they comprised and they failed by a combined crush-
ing and kinking of all layers with associated delamination. It
should be noted that, owing to the nature of this balsa veneer
material, it is almost impossible to achieve a desirable crushing
failure mode for C1 and C2 configurations when standard dimen-
sions are used. Larger or shorter, i.e. more compact, specimens
could be used to determine whether pure crushing failure can
actually be achieved.

3.2.2. Compressive stress–strain curves, modulus and strength

Selected representative compressive stress–strain curves are
shown in Fig. 13. The strains are average values of the three or five
Fig. 8. Fixture and experimental set-up for shear loading. measurements in the transverse direction. C1 specimens initially
exhibited a linear response until they started to buckle, see
Fig. 13(a). Due to the horizontally layered configuration and thus
low axial stiffness, the specimens were able to sustain large lateral
deformations. The axial stress continued to increase nonlinearly
with increasing strain until crack initiation occurred as described
above. Subsequently the axial stress, i.e. the load, decreased during
crack propagation until failure as described above. The compres-
sive stress of C2 specimens also linearly increased with increasing
strain up to a peak point, see Fig. 13(b). The specimens then started
to buckle with gradually increasing lateral deformation. Due to the
flexural stresses developed during the buckling process, shear
cracks appeared in the middle 90° veneer layer and the axial stress
(load) gradually decreased until final failure occurred as described
above. C3 specimens also exhibited a linear elastic stress–strain
response up to a peak point, see Fig. 13(c), after which the stress
Fig. 9. Determination of shear strain using video extensometer [2].
suddenly dropped due to the crushing failure.
The results of all compressive experiments are listed in Table 2.
The modulus was determined by estimating the slope of the
by the transverse-to-grain direction. Configuration T3 with two 0° stress–strain curve’s linear elastic region in the range between 5
layers showed higher values than T2 with only one 0° layer. and 35% of the ultimate load according to ASTM D3501 [20]. The
However, the average modulus and strength of T3 were only stress at the peak point of the stress–strain curve was defined as
1.28 and 1.21 higher than those of T2, although the 0° the strength. For C1 and C2 specimens buckling stresses were also
cross-sectional area of T3 was 2 that of T2. The deviation of this defined and assumed as being the stress at the end of the linear
factor of 2 may be explained regarding the strength by the elastic stage. For C2 specimens this value was almost the same
sequential failure of the two 0° layers and regarding the modulus as that of the peak stress.
by the fact that the 90° layers already failed at very low strains, Similar to the tension results, configuration C1, loaded only in
i.e. below the strain interval at which the modulus was calcu- the transverse-to-the-grain direction, exhibited much lower
lated. In both cases the responses were influenced by resulting strength and stiffness than the C1 and C2 configurations. The
eccentricities. buckling stress was almost half of the compressive strength. C3
6 C. Wu et al. / Construction and Building Materials 265 (2020) 120193

Fig. 10. Typical failure modes of T1, T2 and T3 specimens.

Table 1
Results of tensile experiments.

Specimens Modulus (MPa) Tensile strength (MPa)

T1-1 50 0.25
T1-2 82 0.51
T1-3 52 0.18
T1-4 88 0.32
T1-5 57 0.20
T1-6 57 0.27
T1-7 53 0.28
T1-8 98 0.47
T1-9 75 0.36
T1-10 41 0.30
Mean ± standard dev. 65 ± 19 0.31 ± 0.12
Coefficient of variation (%) 29 38
T2-1 1129 7.78
T2-2 1631 13.84
T2-3 896 7.72
T2-4 2650 16.72
T2-5 1359 11.28
T2-6 965 7.37
T2-7 1959 15.23
T2-8 2501 10.87
Fig. 11. Tensile stress–strain responses of selected T1, T2 and T3 specimens.
T2-9 1795 13.95
T2-10 1071 6.62
Mean ± standard dev. 1596 ± 627 11.14 ± 3.66
specimens, with three 0° layers, exhibited higher strength and Coefficient of variation (%) 39 32
stiffness than C2 specimens with only two 0° layers. The C3 aver- T3-1 1362 10.07
age modulus and strength values were 1.25 and 1.50 higher T3-2 1934 14.45
than the C2 values, the former deviating from the theoretical value T3-3 1557 7.84
of 1.5 depending on the number of 0° layers. This difference can be T3-4 1622 8.47
T3-5 2378 21.42
attributed to the standard deviations of up to 20% of the average T3-6 3178 18.09
values. T3-7 2019 10.09
T3-8 2020 14.43
T3-9 1772 15.26
3.3. Shear properties T3-10 2553 15.63
Mean ± standard dev. 2039 ± 540 13.58 ± 4.40
3.3.1. Failure modes Coefficient of variation (%) 26 32
The failure modes of S1, S2, S3 and S4 specimens are shown in
Fig. 14. In S1 specimens, horizontal cracks initiated in the 0° veneer
layer from the notches of the specimens. Subsequently, under imens. Therefore, the same failure mode occurred, i.e. tensile fail-
increasing load, the continuous grains of the 0° layer between ure at the location of maximum curvature of the s-shape
the notches deformed to an s-shape and acted as tendons. The deformed layer where tensile stresses from the tendon action
specimen finally failed at the locations of maximum curvature of and local bending were superposed. In S3 specimens, failure initi-
these s-bands. The 90° veneer layer (on the back of the specimens ated with vertical cracks in the two 90° layers on both the left and
in Fig. 14) failed much earlier via a vertical shear crack connecting right sides of the notches, which resulted in the separation of a
the top and bottom notches. square element from the specimen. As the load was increased,
In S2 specimens, the 90° veneer layer failed first in shear and the square element started to rotate until it fractured in the 0°
the load was then sustained by the fibers in the 0° veneer layer layer between the notches. S4 specimens failed in the 90° layer
by the same s-shaped tendon mechanism as exhibited in S1 spec- between the notches.
 C. Wu et al. / Construction and Building Materials 265 (2020) 120193 7

element occurred. S4 specimens exhibited a progressive fracture

of the 90° layer as sheer stress decreased.
In S2 specimens, the shear stress also increased almost linearly
until the 90° layer started fracturing, see Fig. 15 (b). After the 90°
layer completely failed, the shear load was sustained only by the
0° veneer layer, i.e. the vertical component of its tension force,
which induced strain hardening until final failure in the 0° layer
occurred. In this case, in contrast to the S3 behavior, strain harden-
ing was primarily caused by the material behavior, as it may also
be activated in a real sandwich structure and not just represent
an artifact of the experimental set-up. Furthermore, in contrast
to the S1, S3 and S4 configurations, S2 specimens exhibited a sig-
nificant ductile behavior, i.e. the ability to sustain inelastic defor-
mation without loss of resistance.
The experimental results of all shear experiments are listed in
Table 3. The maximum shear stress of the stress–strain curve was
defined as the shear strength. The shear modulus was determined
from the slope of the stress–strain curves over a strain range of
0.004 with the lower strain point being in the range of 0.0015–
0.0025 according to ASTM D5379 [21]. For S2 specimens, the
stress at 0.05 shear strain was also extracted, which could repre-
sent the strength causing only limited activation of the tendon
Fig. 12. Typical failure modes of C1, C2 and C3 specimens.
mechanism. The average coefficients of variation (COV) for shear
strength and shear modulus were 12% and 22%, respectively (av-
erage values of all four configurations). Comparing these values to
3.3.2. Shear stress–strain curves, modulus and shear strengths the results published in [2] for commercial end-grain balsa pan-
The typical shear stress–strain curves of all configurations are els, i.e. COV values of 33% for shear strength and 28% for shear
shown in Fig. 15. For S1, S3 and S4 specimens, the shear stress ini- modulus, demonstrated the more uniform shear properties of
tially increased almost linearly with increasing shear strain up to a the LVL material.
peak point, see Fig. 15 (a). The nonlinearity was caused by the S1 and S2 specimens exhibited a significantly higher shear
damage accumulation during the progressive failure process. After strength and modulus than S3 and S4 specimens due to the acti-
the peak point, the shear stresses of S1, S3 and S4 specimens vation of the tendon mechanism. The modulus of S1, which was
dropped as the strain increased. S1 specimens lost their strength influenced by the axial and bending stiffness of the activated 0°
gradually because of the gradual fracture of the ‘‘tendon’’ layer. tendon, was higher than that of S2 since the cross section of
In S3 specimens, a first drop in shear stress was caused by the crack the 0° layer (12  6 mm2 in each case) was placed vertically
formation in the two 90° layers on both the left and right sides of and not horizontally as in the case of S2. For the same reason,
the notches. A second drop occurred due to the complete separa- the strength of S1 was however lower than that of S2 since the
tion of the square element from the specimen, as described above. activated bending moment and thus additional bending-tensile
After this second drop, the specimens exhibited a hardening stresses were higher in S1 than in S2. The S3 configuration exhib-
behavior that however was not related to the material, i.e. the sep- ited better properties than S4 since the shear force between the
arated square element was stuck between the grips and rotated notches acted parallel and not transverse to the grain as in the
and thus increased the resistance until complete fracture of the latter case.

Fig. 13. Compressive stress–strain curves of (a) C1 and (b) C2 and C3.
8 C. Wu et al. / Construction and Building Materials 265 (2020) 120193

Table 2
Results of compressive experiments.

ID Specimens Modulus (MPa) Strength (MPa) Buckling stress (MPa)

1 C1-1 81 0.69 0.38
2 C1-2 71 0.72 0.36
3 C1-3 86 0.71 0.39
4 C1-4 80 0.69 0.39
5 C1-5 63 0.54 0.37
6 C1-6 82 0.69 0.41
Mean ± standard dev. 77 ± 8 0.67 ± 0.07 0.38 ± 0.02
Coefficient of variation (%) 10 10 0.05
1 C2-1 2229 7.61 7.16
2 C2-2 2794 8.71 8.40
3 C2-3 2894 6.28 5.93
4 C2-4 2311 8.26 7.83
5 C2-5 2240 7.61 7.36
6 C2-6 2038 6.81 6.57
Mean ± standard dev. 2418 ± 344 7.55 ± 344 7.21 ± 0.88
Coefficient of variation (%) 14 12 12
1 C3-1 2592 9.42 N/A
2 C3-2 3064 10.62 N/A
3 C3-3 3648 13.62 N/A
4 C3-4 2176 8.21 N/A
5 C3-5 2905 12.34 N/A
6 C3-6 3711 13.08 N/A
Mean ± standard dev. 3016 ± 597 11.22 ± 2.15 N/A
Coefficient of variation (%) 20 19 N/A

Fig. 14. Failure modes of S1, S2, S3 and S4 specimens.

Fig. 15. Shear stress–strain curves of (a) S1, S3 and S4 and (b) S2 specimens.
 C. Wu et al. / Construction and Building Materials 265 (2020) 120193 9

Table 3
Results of shear experiments.

ID Specimens Modulus (MPa) Strength (MPa) Strength at 0.05 strain (MPa)

1 S1-1 53 1.37 NA
2 S1-2 85 1.79 NA
3 S1-3 69 1.16 NA
4 S1-4 94 1.74 NA
5 S1-5 58 1.20 NA
Mean ± standard dev. 72 ± 17 1.45 ± 0.30 NA
Coefficient of variation (%) 24 21 NA
1 S2-1 43 1.88 0.88
2 S2-2 49 1.98 1.07
3 S2-3 59 2.44 1.31
4 S2-4 46 2.00 0.84
5 S2-5 40 1.77 1.07
Mean ± standard dev. 48 ± 7 1.92 ± 0.10 1.03 ± 0.19
Coefficient of variation (%) 14 5 18
1 S3-1 26 0.64 NA
2 S3-2 13 0.51 NA
3 S3-3 26 0.57 NA
4 S3-4 25 0.63 NA
5 S3-5 22 0.52 NA
Mean ± standard dev. 23 ± 5 0.57 ± 0.06 NA
Coefficient of variation (%) 22 10 NA
1 S4-1 10 0.31 NA
2 S4-2 20 0.36 NA
3 S4-3 12 0.25 NA
4 S4-4 17 0.37 NA
5 S4-5 21 0.36 NA
Mean ± standard dev. 16 ± 5 0.33 ± 0.05 NA
Coefficient of variation (%) 31 15 NA

4. Characteristic values of strength and modulus

4.1. Statistical distributions

Many design codes are based on characteristic values of the

material properties, defined as the 5% fractile values [23]. These
values are dependent on the type of statistical distribution that
the data follows. Quantile-quantile (Q-Q) plots are used in the fol-
lowing to find the best distributions for the experimental data
presented above. In this goodness-of-fit method, the quantiles
of the data are set against the quantiles of a specific distribution
[24]. If the two sets originate from the same distribution, the data
points should fall approximately along the reference line [24]. An
example of a Q-Q plot is given in Fig. 16, which shows the quan-

Fig. 17. R-squared values derived from Q-Q plots of all strength data.

tiles of the strength data of the T1 specimens plotted against

those of the normal distribution. The corresponding R-squared
value representing the goodness of the normal distribution is
then calculated, resulting in a value of 0.92 in this case, see
Fig. 16. The R-squared values derivedfrom Q-Q plots of the
strength data of all the specimen configurations versus normal,
lognormal, Weibull and gamma distributions are illustrated in
Fig. 17.
The Weibull distribution exhibited the highest R-squared values
among all distributions for all specimen configurations. For the
modulus data, the results of the Q-Q plots were the same and Wei-
Fig. 16. Q-Q plot of strength data of T1 specimens vs normal distribution. bull was also shown to be the best fitting distribution.
10 C. Wu et al. / Construction and Building Materials 265 (2020) 120193

Table 4
Values of kn for 5% characteristic value [23].

n 1 2 3 4 5 6 8 10 20 30 1

kðnÞ 2.31 2.01 1.89 1.83 1.80 1.77 1.74 1.72 1.68 1.67 1.64

Table 5
5% characteristic values.

Configuration Eurocode Normal Weibull

Modulus Strength Modulus Strength Modulus Strength
T1 33 0.13 34 0.14 34 0.14
T2 517 4.84 564 5.11 648 5.45
T3 1110 6.02 1151 6.35 1083 6.61
C1* 62 0.35 63 0.35 65 0.35
C2* 1808 5.65 1851 5.76 1779 5.68
C3 1959 7.41 2034 7.68 2046 7.84
S1 40 0.92 43 0.97 44 0.97
S2+ 34 0.69 35 0.73 33 0.70
S3 13 0.47 14 0.47 16 0.47
S4 7 0.24 8 0.25 9 0.26
*
strength calculated based on buckling stress.
+
strength calculated based on strength at 0.05 strain.

4.3. Comparison and discussion

The 5% characteristic values of modulus and strength for all

configurations were calculated based on the Weibull and normal
distributions and compared with corresponding values obtained
from Eurocode EN1990, see Table 5.
The differences in the results are shown in Fig. 18 for the case of
the characteristic modulus of the C1 configuration. The best-fit
Weibull distribution gave higher values compared to the normal
distribution due to its narrower shape. The differences between
mean value and characteristic value of normal and Weibull distri-
butions were 1.5% and 2.4%, respectively. The Eurocode values
based on the normal distribution were slightly lower (1.6%) than
the corresponding values from the normal distribution since the
latter are based onkn = 1.64, i.e. on an infinite number of samples,
see Table 4.
Fig. 18. Comparison between 5% characteristic modulus of C1 corresponding to
Weibull, normal and Euorocode EN1990 [23]. 5. Conclusions

This paper presents a comprehensive experimental study to

4.2. Eurocode EN1990 characterize the tensile, compressive and shear properties of a
veneered balsa wood product, BALTEKÒ VBC, as a core material
Civil engineering structures in Europe, such as bridges or build- for sandwich applications. The failure modes, stress–strain curves
ings, are normally designed based on the Eurocode. According to and modulus and strength values were reported. The following
Eurocode EN1990 [23], the 5% characteristic values are based on conclusions were drawn:
a normal distribution and calculated as follows:

The mechanical properties were significantly influenced by the
orthotropy on two scales, i.e. on the material scale within one
X kðnÞ ¼ mx ð1  kn V x Þ ð1Þ veneer layer and on the system scale within the assembled
veneer layers. The material and system orientation with regard
to the loading direction must therefore be taken into account in
withX kðnÞ = characteristic value, mx = mean value of n sample results, the design verification.
kn = characteristic fractile factor and V x = coefficient of variation of
It is not easy to obtain the mechanical properties of these
X n . The values of kn depend on the number of samples and are listed veneer materials since standard set-ups and specimen geome-
in Table 4. tries may produce artifacts such as buckling or strain hardening.
The resulting 5% characteristic tensile, compressive and shear The material and system responses must be carefully checked
strength and modulus of all configurations calculated using Eq as to whether they correspond to the behavior in the real struc-
(1) are presented in Table 5. These values take into account the ture to be designed based on these values.
scatter of the experimental data and statistical uncertainty associ-
Although the veneer material is intended to reduce property
ated with the number of experiments [23]. It can be seen that the scatter, the latter was still significant. Due to the scatter, the
scattering nature of the wood resulted in characteristic values con- 5% characteristic values according to Eurocode EN1990 are sig-
siderably lower than the mean values shown in Tables 1–3. nificantly reduced.
 C. Wu et al. / Construction and Building Materials 265 (2020) 120193 11

The thin adhesive between the veneer layers did not seem to [2] M. Osei-Antwi, J. De Castro, A.P. Vassilopoulos, T. Keller, Shear mechanical
characterization of Balsa wood as core material of composite sandwich panels,
have negatively affected the mechanical behavior; failure
Constr. Build. Mater. 1 (41) (2013) 231–238.
occurred within the veneer layers and not in the interfaces. This [3] N.J. Kotlarewski, B. Belleville, B.K. Gusamo, B. Ozarska, Mechanical properties
may however change at elevated temperatures when the tem- of Papua New Guinea Balsa wood, Eur. J. Wood Wood Prod. 74 (1) (2016) 83–
perature approaches the glass transition temperature of the 89.
[4] C. Atas, C. Sevim, On the impact response of sandwich composites with cores of
adhesive. Balsa wood and PVC foam, Compos. Struct. 93 (1) (2010) 40–48.
[5] Y.W. Kwon, M.A. Violette, R.D. McCrillis, J.M. Didoszak, Transient dynamic
The results of this work demonstrate the advantage of veneered response and failure of sandwich composite structures under impact loading
with fluid structure interaction, Appl. Compos. Mater. 19 (6) (2012) 921–940.
balsa wood material compared to bulk/block-shaped balsa wood [6] S. Midgley, M. Blyth, N. Howcroft, D. Midgley, A. Brown, Balsa: biology,
when used as core material in sandwich structural elements. Nor- production and economics in Papua New Guinea, ACIAR Technical Reports
mally the stress state in sandwich cores is multidirectional. Series. 73 (2010).
[7] M. Osei-Antwi, J. De Castro, A.P. Vassilopoulos, T. Keller, Structural limits of
Veneered balsa wood can much better adapt to such complex FRP-Balsa sandwich decks in bridge construction, Compos. B Eng. 1 (63) (2014)
stress conditions and the scatter of the properties is reduced. It 77–84.
should however be taken into account that the specimen sizes [8] T. Keller, J. Rothe, J. De Castro, M. Osei-Antwi, GFRP-Balsa sandwich bridge
deck: concept, design, and experimental validation, J. Compos. Constr. 18 (2)
were small; the obtained properties, the strengths values in partic- (2013) 04013043.
ular, were influenced by size effects and are thus not representa- [9] J. Dai, H.T. Hahn, Flexural behavior of sandwich beams fabricated by vacuum-
tive for higher volumes, i.e. cannot be directly applied in the assisted resin transfer molding, Compos. Struct. 61 (3) (2003) 247–253.
[10] R.M. Reising, B.M. Shahrooz, V.J. Hunt, A.R. Neumann, A.J. Helmicki, M. Hastak,
design of balsa cores of large-scale sandwich elements.
Close look at construction issues and performance of four fiber-reinforced
polymer composite bridge decks, J. Compos. Constr. 8 (1) (2004) 33–42.
[11] L.N. Triandafilou, J.S. O’Connor, Field issues associated with the use of fiber-
CRediT authorship contribution statement reinforced polymer composite bridge decks and superstructures in harsh
environments, Struct. Eng. Int. 20 (4) (2010) 409–413.
Chao Wu: Conceptualization, Investigation, Formal analysis, [12] ASTM D143–09, Standard Test Methods for Small Clear Specimens of Timber
ASTM International, West Conshohocken, PA, 2009.
Writing - original draft. Niloufar Vahedi: Formal analysis, Writing
[13] M.Q. Chowdhury, S.K. Sarker, J.C. Deb, S.S. Sonet, Timber species grouping in
- original draft. Anastasios P. Vassilopoulos: Validation, Writing - Bangladesh: linking wood properties, Wood Sci. Technol. 47 (4) (2013) 797–
review & editing. Thomas Keller: Validation, Writing - review & 813.
[14] A.F. Moshtaghin, S. Franke, T. Keller, A.P. Vassilopoulos, Experimental
editing.
characterization of longitudinal mechanical properties of clear timber:
random spatial variability and size effects, Constr. Build. Mater. 1 (120)
Declaration of Competing Interest (2016) 432–441.
[15] M. Shir Mohammadi, J.A. Nairn, Balsa sandwich composite fracture study:
comparison of laminated to solid balsa core materials and debonding from
The authors declare that they have no known competing finan- thick balsa core materials, J. Compos. Part B: Eng. 1 (122) (2017) 165–172.
cial interests or personal relationships that could have appeared [16] https://www.3accorematerials.com/uploads/documents/Flyer-Baltek-VBC-
Prepress.pdf
to influence the work reported in this paper. [17] H. Bailleres, L. Denaud, J.C. Butaud, R.L. McGavin, Experimental investigation
on rotary peeling parameters of high density coconut wood, BioResources. 10
(3) (2015) 4978–4996.
Acknowledgments
[18] ASTM D4442–07, Standard Test Methods for Direct Moisture Content
Measurement of Wood and Wood-Base Materials ASTM International, West
The authors gratefully acknowledge the financial support pro- Conshohocken, PA, 2007.
vided by the Swiss Government Excellence Fellowship [19] ASTM D3500–14, Standard Test Methods for Structural Panels in Tension
ASTM International, West Conshohocken, PA, 2014.
(20150901), Swiss National Science Foundation (Grant No. [20] ASTM D3501-05a(2011), Standard Test Methods for Wood-Based Structural
200020_172520), and National Science Foundation of China Panels in Compression, ASTM International, West Conshohocken, PA, 2011.
(51978025, 51911530208). The materials used in this study were [21] ASTM D5379 / D5379M-12, Standard Test Method for Shear Properties of
Composite Materials by the V-Notched Beam Method, ASTM International,
kindly provided by 3A Composites Core Materials, Sins, West Conshohocken, PA, 2012.
Switzerland. [22] ASTM D3039 / D3039M, Standard Test Method for Tensile Properties of
Polymer Matrix Composite Materials, ASTM International, West
Conshohocken, PA, 2008.
References [23] EN 1990, Eurocode - Basis of structural design, The European Union Per
Regulation 305/2011, Directive 98/34/EC, Directive 2004/18/EC, 2002.
[1] G. Tsoumis, Science and technology of wood: structure, properties, utilization, [24] M.B. Wilk, R. Gnanadesikan, Probability plotting methods for the analysis of
Van Nostrand Reinhold, New York, 1991. data, Biometrika 55 (1) (1968) 1–7.