Engineering Structures 56 (2013) 1103–1116

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Strain-based calculation model for centrically and eccentrically loaded

timber columns
Matthias Theiler a,⇑, Andrea Frangi a, René Steiger b
a
ETH Zurich, Institute of Structural Engineering (IBK), Wolfgang-Pauli-Strasse 15, 8093 Zurich, Switzerland
b
Empa, Materials Science and Technology, Structural Engineering Research Laboratory, Ueberlandstrasse 129, 8600 Dübendorf, Switzerland

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

Article history: The behaviour of timber members subjected to axial compression or combined axial compression and
Received 19 December 2012 bending is characterised by the non-linear increase of the deformation due to the increasing eccentricity
Revised 14 June 2013 of the axial load and due to the non-linear material behaviour. The paper presents a strain-based model
Accepted 25 June 2013
taking into account these effects.
Available online 26 July 2013
Design approaches given in timber structures design codes are compared and differences in the results
obtained with the different approaches are identified. Furthermore, a strain-based model to analyse the
Keywords:
load-bearing capacity of centrically and eccentrically loaded timber columns is described and its power is
Timber structures
Columns
assessed. It is shown that in particular the non-linear behaviour of timber when subjected to compression
Compression parallel to the grain considerably influences the load-bearing capacity.
Stability The model is validated on the basis of experimental investigations on solid Norway spruce beams
Global buckling loaded in combined axial compression and bending. A good agreement was found between the estimated
P-delta effect values using the strain-based model and the experimentally derived values.
2nd Order structural analysis A comparison of the model with the design approaches given in the codes shows that the load-bearing
capacity can be overestimated under certain conditions. Finally, it is illustrated how the design
approaches can be modified in order to reach a more consistent design.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
Axial compression or combined axial compression and bending
are encountered in many types of timber members such as col-
umns, frame structures or compression members of truss girders.
The behaviour of these structural members is primarily character-
ised by the non-linear increase of the deformation due to the
increasing eccentricity of the axial load (P-delta effect). In addition
to this geometric non-linear behaviour, the non-linear material
behaviour of timber members subjected to compression parallel
to the grain has to be accounted for.
The influence of the P-delta effect on the load-bearing capacity
of timber members subjected to axial compression was investi-
gated first by Tetmajer [1] in 1896. Tetmajer’s studies set up the
basis for the design of timber members subjected to axial compres-
sion for a long time. Tests performed by Larsen and Pedersen [2]
confirmed the results obtained by Tetmajer. The experimental
investigation showed the great influence of the varying material
properties on the load-bearing capacity. In order to account for
⇑ Corresponding author. Tel.: +41 44 633 61 69; fax: +41 44 633 10 93.
E-mail addresses: theiler@ibk.baug.ethz.ch (M. Theiler), frangi@ibk.baug.ethz.ch
(A. Frangi), rene.steiger@empa.ch (R. Steiger).
these variations and hence to estimate the resistance of glued lam-
inated timber members subjected to compression more accurately,
Blaß [3,4] performed Monte Carlo simulations. The buckling curves
given in different design codes [5–7] were derived from these
investigations. For timber members subjected to combined axial
compression and bending, Buchanan [8,9] developed a numerical
model capable of investigating the influence of the non-linear
material behaviour on the moment – axial force interaction. In
addition Buchanan investigated the influence of the size of the
member.
Current design codes (e.g. Eurocode 5 [5] or the Swiss national
code for the design of timber structures SIA 265 [6]) provide two
different approaches for the design of centrically and eccentrically
loaded timber columns:
– a simplified calculation model based on the Effective Length
Method (ELM),
– 2nd order analysis of the structure.
In ELM, the buckling problem of a structural system is re-
duced to that of an equivalent simply supported (pinned) col-
umn. The 2nd order analysis of the structure is a method
which takes into account the non-linearity by studying the equi-

0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.06.032
1104 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
librium of the deformed structural system. In general, non-line-
arity caused by the increasing eccentricity of the external load
as well as non-linearity caused by the non-linear material
behaviour of timber subjected to compression should be consid-
ered. However, the 2nd order analysis is often understood as a
theory based on linear elastic material behaviour and the effects
caused by the non-linearity of the material are neglected. Even
the design codes [5–7] only provide rules for this 2nd order lin-
ear elastic analysis of the structure. In this paper, a clear distinc-
tion between the 2nd order linear elastic analysis and the
generalised 2nd order analysis is made.
The two approaches (ELM and 2nd order linear elastic analysis
of the structure) given in the codes are not consistent and can lead
to different results. This situation led to controversial discussions
in the scientific community [10–13]. The discussion in particular
showed that there are inconsistencies concerning the consider-
ation of the effect of moisture content (MC) and duration of load
(DOL) as well as inconsistencies concerning the implementation
of the 2nd order linear elastic analysis in the design codes. While
recent research on the load-bearing behaviour of timber members
subjected to axial compression or combined axial compression and
bending was mainly focused on MC and DOL [10–12,14,15], this
publication deals with the influence of the non-linear material
behaviour and with the implementation of 2nd order linear elastic
analysis in the design codes such that there are only minor differ-
ences between 2nd order linear elastic analysis and the Effective
Length Method. Hence, the results presented here are only valid
for short-term response under load at constant interior climate.
In fact, MC and DOL and in particular the creep behaviour and
the climate take a major impact on the load-bearing behaviour of
timber columns and should also be considered for the design of
timber members subjected to compression or combined compres-
sion and bending [16–18].
2. Design of timber members in compression parallel to the
grain
The current design codes such as Eurocode 5 EN 1995-1-1 [5],
the Swiss design code SIA 265 [6] or the withdrawn German code
DIN 1052 [7] provide two different approaches for the design of
timber member subjected to either compression parallel to the
grain or combined axial compression and bending. In general, the
Effective Length Method is used for simple design situations while
the 2nd order linear elastic analysis of the structure provides some
advantages for more complex design situations.
2.1. Effective Length Method (ELM)
The simplified calculation model is based on the Effective
Length Method (ELM). The buckling problem of a structural system
is reduced to that of an equivalent simply supported (pinned) col-
umn [19].
For the design, the internal forces and moments are calculated
based on a simple 1st order analysis and the non-linear P-delta ef-
fect is taken into account by means of a buckling factor kc. This fac-
tor describes the ratio between the axial stress at buckling failure
of a member subjected to axial compression and its compressive
strength parallel to the grain. kc depends on the effective length
of the structural system which can be expressed by the slenderness
ratio k (Eq. (1)).
‘cr
k¼
i Timber members subjected to combined axial compression and
rffiffiffi
I sion zone before failure occurs. This non-linearity leads to a curved
i¼
A shape of the moment – axial force interaction diagram depending
where k is the slenderness ratio; ‘cr is the effective length; i is the
radius of gyration; I is the 2nd moment of inertia and A is the area
of the cross-section.
The buckling factor kc as used in [5–7] is based on extensive
investigations performed by Blaß [3]. In order to determine the
characteristic (i.e. 5th percentile) load-bearing capacity of timber
columns a Monte Carlo simulation technique was used. The
numerical model and the parameter study considered the P-delta
effect, the variability of the strength and the stiffness properties
within the timber members, the geometric imperfection of the tim-
ber members and the non-linear material behaviour of timber
when subjected to compression parallel to the grain and bending.
For the ultimate limit state analysis, the design codes [5–7] rec-
ommend to use a linear interaction model for combined axial com-
pression and bending as given in Eq. (3). In this interaction model,
the buckling factor kc is used to reduce the compressive strength
parallel to the grain of the timber member in order to account
for probable buckling.
rc;0;d rm;I;d
þ 61 ð3Þ
kc  fc;0;d fm;d
rc,0,d is the design value of the acting compressive stress parallel to
the grain; kc is the buckling factor; fc,0,d is the design compressive
strength parallel to the grain; rm,I,d is the design value of the acting
bending stress derived by means of a 1st order structural analysis
and fm,d is the design bending strength.
2.2. 2nd Order linear elastic analysis of the structure
As an alternative to the calculation model based on the Effective
Length Method, timber members subjected to axial compression or
combined axial compression and bending can be designed by per-
forming a 2nd order linear elastic analysis of the structure. The 2nd
order linear elastic analysis is a method which takes into account
the geometric non-linearity by studying the equilibrium of the de-
formed structural system. An initial deformation is introduced into
the calculation in order to account for the geometric imperfection
of the structural member as e.g. deviation from a perfectly straight
shape.
For a simply supported, axially loaded column the 2nd order
linear elastic analysis can easily be performed, assuming sinusoidal
distributed initial deformations. The initial deformation in combi-
nation with the axial load leads to an initial bending moment MI.
The P-delta effect causes a magnified moment MII. MII can be calcu-
lated by multiplying the initial bending moment MI by a magnifi-
cation factor l [20]:
MII ¼ MI  l ð4Þ
1
l¼ ð5Þ
1 NN
Euler
NEuler is the Euler buckling load:
p2  EI
NEuler ¼ ð6Þ
‘2cr
MII is the magnified bending moment (2nd order linear elastic the-
ory, deformed structural system); MI is the initial bending moment
(1st order theory, undeformed structural system); l is the magnifi-
cation factor; N is the normal force acting on the column; NEuler is
the Euler buckling load; E is the modulus of elasticity (MOE); I is
the 2nd moment of inertia and ‘cr is the effective length.
ð1Þ
bending tend to develop non-linear deformations of the compres-
ð2Þ
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
on the ratio between the tensile strength ft,m,0 and the compressive
strength fc,m,0 parallel to the grain [9]. For the ultimate limit state
analysis of timber columns the design codes [5–7] consider this
non-linear interaction behaviour by squaring the compression part
in the interaction model (Eq. (7)). However, the non-linear material
behaviour also influences the deformations of the structural sys-
tem, and as a consequence, also the magnified moment MII is influ-
enced by the non-linear material behaviour. However, these effects
are neglected when performing a 2nd order linear elastic analysis
of the structure.
 2
rc;0;d rm;II;d
þ 61
fc;0;d fm;d
rc,0,d is the design value of the acting compressive stress parallel to
the grain; fc,0,d is the design compressive strength parallel to the
grain; rm,II,d is the design value of the acting bending stress derived
from a 2nd order structural analysis and fm,d is the design bending
strength.
2.3. Comparison of ELM and 2nd order linear elastic analysis of the
structure
The two calculation models described above differ in terms of
the way the effects caused by the non-linear material behaviour
of timber subjected to axial compression as well as the geometric
and structural imperfections are considered. While these effects
are considered implicitly in the buckling factor when using the
Effective Length Method, the non-linear material behaviour is ne-
glected when performing a 2nd order linear elastic analysis of
the structure. The effect of the non-linear material behaviour is
covered by the selection of the initial deformations and by the de-
sign value of the stiffness.
However, the rules for the 2nd order linear elastic analysis dif-
fer between the different design codes. In particular, there are dif-
ferences in the design value of the stiffness on which the
calculation is based. According to Eurocode 5 EN 1995-1-1, the de-
sign value for the stiffness is defined as the mean value of the mod-
ulus of elasticity (MOE) divided by the partial safety factor cM. The
Swiss code SIA 265 recommends reducing the mean value of the
MOE by dividing it by cM/gM. The factor gM permits the conversion
of the value of strength obtained from short-term standard tests to
the situation of structural elements in building practice. According
to the former German code DIN 1052, the 5th percentile of the
MOE reduced by the partial safety factor cM has to be used for sta-
bility verifications of single structural members. In Table 1 a sum-
mary and a comparison of these values are given. Exemplary, the
design values are listed for solid timber of strength class C24
according to EN 338 [21] and for glued laminated timber of
strength class GL24 h according to EN 1194 [22] for standard con-
ditions (MC = 12%).
Furthermore, the implementation of the two calculation methods
in the different codes differs with respect to the way the impact of
the moisture content (MC) and the duration of load (DOL) is ac-
Table 1
Design values of the stiffness of solid timber of strength class C24 and glued laminated members of strength class GL24 h to be assumed in a 2nd order linear elastic analysis
according to different design codes.
Eurocode 5 EN 1995-1-1 [5]
Design value of the stiffness E0;mean
cM
Solid timber C24 b 8460 N/mm2
c
Glued laminated timber GL24 h 9280 N/mm2
a
Values for single structural members. Additional values are provided for the analysis of structural systems.
b
According to EN 338 [21].
c
According to EN 1194 [22].
1105
counted for. Both, the strength and the stiffness of timber are af-
fected by MC and DOL. However, the influence of MC and DOL are
different for strength and stiffness [23]. The codes only provide sim-
plified approaches for the consideration of these effects.
According to Eurocode 5 EN 1995-1-1 and DIN 1052, the influ-
ence of MC and DOL are accounted for by using a modification fac-
tor kmod. The Swiss code SIA 265 considers the influence of MC and
DOL by using modification factors gw and gt respectively. While
according to Eurocode 5 EN 1995-1-1 and DIN 1052 the modifica-
tion factor kmod has to be applied only on strength properties SIA
265 provides modification factors gw for the strength properties
as well as for the stiffness properties. Eurocode 5 provides a mod-
ð7Þ
ification factor kdef to account for the influence of MC and DOL on
the stiffness properties. However, according to Eurocode 5 the fac-
tor kdef shall not be applied when performing a 2nd order linear
elastic analysis.
The implementation of the modification factors in the design
codes is not consistent for the two calculation methods [10–12].
This situation leads to additional differences between the different
calculation methods.
Fig. 1 shows an exemplary comparison between the two calcu-
lation methods for the different codes. The comparison was made
for a simply supported column with a cross-section of
200 mm  200 mm made of solid timber of strength class C24
(EN 338 [21]). The design value of the load-bearing capacity of
the column was calculated according to Eurocode 5 EN 1995-1-1,
SIA 265 and DIN 1052 for both design approaches (ELM and 2nd
order linear elastic analysis of the structure). In this comparison,
the modification factors were chosen as follows: kmod = 0.8,
gw = 1.0 and gt = 1.0. These values reflect a situation of a timber
member in service class 1 (the average moisture content in the
timber member does not exceed 12%) with medium term action
(duration of load in the order of one week to six months). Applying
Eurocode 5 EN 1995-1-1 and DIN 1052 provides similar results for
stocky columns. The load-bearing capacity calculated according to
SIA 265 is slightly lower since the influence of the moisture con-
tent and the duration of load is considered in a different way. For
columns with slenderness ratios k higher than 60 large deviations
can be observed, in particular when regarding the results obtained
with the 2nd order linear elastic analysis. The deviations can partly
be explained by the differences in the consideration of the mois-
ture content and the duration of load. In addition, the design value
of the stiffness substantially influences the results of the 2nd order
linear elastic analysis. The accuracy of the different design ap-
proaches and how a better fit between them can be reached will
be discussed in detail later in this paper (see also Fig. 12).
3. Strain-based model
In order to try to better assess the global buckling behaviour of
timber columns, a strain-based model was developed. Such models
so far have not been commonly applied for the design of timber
structures, because the failure mechanism in timber elements is
a
SIA 265 [6] DIN 1052 [7]
E0;mean E0;05
cM =gM cM
6470 N/mm2 5640 N/mm2
7330 N/mm2 7435 N/mm2
1106 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
Fig. 1. Design value of the load-bearing capacity obtained with different design methods according to Eurocode 5 EN 1995-1-1 [5], SIA 265 [6] and DIN 1052 [7].
influenced by a distinct non-linear stress–strain relationship. This
leads to a complex calculation procedure. In literature, some appli-
cations for timber members are reported [3,4,8,24].
The core of the model is the equilibrium between the acting
external and the resulting internal force and moment. In addition,
the model relates the curvature caused by the internal force and
moment to the deformation due to the external force and moment.
One major simplification is made: Only uniaxial stress condi-
tions are considered. Actually, timber members loaded in com-
bined axial compression and bending are subjected to multi-axial
stresses. The consideration of these effects would lead to a more
complex calculation procedure. When multi-axial conditions are
considered, a multi-axial failure criterion has to be applied in con-
sequence. Many different failure criteria for timber (e.g. Tsai-Wu,
Norris or van der Put criterion) are reported in literature [25].
However, these failure criteria require interaction coefficients for
the multi-axial stress condition which are difficult to evaluate for
timber members in structural size. Hence, the failure criteria often
lack in precision or are only valid under certain conditions (e.g.
clear wood). Furthermore, the failure of timber members subjected
to axial compression and bending is dominated by the longitudinal
stresses. Hence, an approach using uniaxial stress conditions is
more appropriate. In addition, a strain-based model with uniaxial
stress condition is more suitable to demonstrate the influence of
the non-linear material behaviour.
The strain-based model consists of two parts. In a first step
(Fig. 2, left) the internal force and moment are regarded. For a gi-
ven cross-section and a given strain distribution, the resulting
Fig. 2. Calculation procedure: internal force and moment (Ni, Mi) are calculated for a given strain distribution parallel to the grain (left) and are then compared with external
force and moment (Ne, Me) acting on a timber column (right).
internal force Ni and moment Mi can easily be calculated, assuming
that plane sections remain plane. The calculation starts by select-
ing values for the strain parallel to the grain e0 at location of the
mass centre of the cross-section and for the curvature vy. These
two parameters define the strain distribution within the whole
cross-section. Based on the strain distribution the stress distribu-
tion is calculated using a material model represented by a
stress–strain relationship. Basically, any stress–strain relationship
can be used in the model. However, the selection of an appropriate
material model is of essential impact on the result of the calcula-
tion. Once the stress distribution is specified, the internal force
and moment can be determined by integrating the stresses over
the cross-section.
In the second step the deformation of the column is accounted
for (Fig. 2, right). The deformation of the column depends on the
curvature and in particular on the distribution of the curvature
along the column. The curvature can be calculated as the 2nd der-
ivation of the deformation curve. Hence, the maximal deformation
due to the P-delta effects can be written as:
ZZ
2 ‘2cr
eII ¼ vðxÞ  dx ¼ vy  ð8Þ
c
where eII is the maximal deformation due to the P-delta effects; v(x)
is the curvature along the column; vy is the curvature at the col-
umn’s mid-height; ‘cr is the effective length and c is an integration
constant.
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116 1107

The integration constant c depends on the shape of the defor- 4.2. Timber subjected to bending
mations. For a simply supported column with sinusoidal distrib-
uted deformations the constant c is equal to p2. When designing timber members loaded in bending, usually a
Finally, the external bending moment can be calculated linear elastic material model is assumed. In fact, the actual bending
depending on the external force Ne and the curvature vy: behaviour is more complex and depends on the behaviour in com-
! pression and in tension parallel to the grain. The assumption of a
‘2 linear elastic material behaviour is only appropriate for stresses
M e ¼ Ne  ðe0 þ eI þ eII Þ ¼ Ne  e0 þ eI þ vy  cr ð9Þ smaller than the proportional limit. Beyond the proportional limit
c
the stiffness of the compression zone of the member is reduced and
non-linear deformations in the compression zone can develop. This
where Me is the acting external bending moment; Ne is the acting
leads to a shift of the neutral axis towards the tension side and a
external force (compression); e0 is the initial deformation of the col-
non-linear increase of the deformation due to the reduction of
umn; eI is the eccentricity of the external force; eII is the maximal
the bending stiffness [9,26]. Finally, the bending failure is governed
deformation due to the P-delta effects; vy is the curvature at the
by the tensile strength of the member. The degree of the non-line-
column’s mid-height; ‘cr is the effective length and c is the integra-
arity of the stress distribution is directly related to the ratio be-
tion constant.
tween the tensile strength ft,m,0 and the compressive strength
Both the internal and the external bending moment are func-
fc,m,0 parallel to the grain. Clear wood and high strength timber
tions of the curvature of the cross-section at the column’s mid-
members exhibit considerably higher tensile strength than com-
height.
pressive strength parallel to the grain. Therefore, such members
In other words, the strain-based model can be seen as a gener-
fail after the appearance of non-linear deformations in the com-
alised 2nd order structural analysis with the following special
pression zone. For normal strength solid timber or glued laminated
characteristics:
timber (e.g. C24 or GL24 respectively), the tensile strength parallel
to the grain is not higher than the compressive strength. As a con-
– When performing a 2nd order linear elastic analysis of the
sequence, no or only minor non-linear deformations can be ob-
structure only the geometric non-linearity is considered and
served when such structural members fail in bending.
the non-linearity due to the change in member stiffness for
increasing load is neglected. With the strain-based model both
effects are accounted for. 4.3. Available material models
– There are differences in the calculation procedure. When per-
forming a 2nd order structural analysis the acting (external) While the behaviour of timber subjected to tension parallel to
force and moment are firstly calculated (assuming a deformed the grain can be adequately modelled by means of the linear elastic
structural system). Afterwards, the acting force and moment Hooke’s law, the more complex non-linear material behaviour of
are compared to the resistance of the cross-section subjected timber subjected to compression has to be taken into account by
to combined axial force and bending moment. Within the an appropriate mathematical formulation. Many models being
strain-based model the external force and moment are directly capable to account for the non-linear material behaviour of timber
linked to the internal force and moment by using the curvature. in compression parallel to the grain have been investigated and are
As a consequence, there is no need to calculate the resistance of reported in literature [26–29]. All these models are based on the
the cross-section. Bernoulli hypothesis (plane sections remain plane) and therefore,
the bending behaviour can directly be derived from the behaviour
in compression and tension parallel to the grain. Fig. 3 presents the
4. Stress–strain relationship
most common models.
The simplest model is the bilinear elastic–plastic stress–strain
Generally, the material properties of timber are markedly influ-
relationship. This model was proposed first by Neely [27] in
enced by inhomogeneity (knots, deviated grain, etc.), moisture
1898. In many situations this relationship is a reasonable approx-
content, temperature and duration of load. Furthermore the mate-
rial properties of timber strongly depend on type and orientation of
stresses. For centrically and eccentrically loaded timber columns,
the material behaviour in case of compression parallel to the grain
as well as in bending are of particular interest.

4.1. Timber subjected to axial compression or tension

The material behaviour of timber when subjected to compres-

sion or tension parallel to the grain is different. When timber is
loaded in compression it shows linear stress–strain relationship
up to a proportional limit smaller than the compressive strength.
Beyond this limit the stiffness is reduced, leading to a ductile fail-
ure with a softening effect after reaching the compressive strength.
In tension parallel to the grain, the stress–strain relationship of
timber is linear elastic up to failure. At ultimate stress, timber fails
in brittle manner, usually starting from knots or locally deviated
grain. As a consequence, the tensile strength parallel to the grain
of timber members is considerably influenced by structural defects
and in addition it distinctly depends on the size of the member.
The size effect is less significant for timber members subjected to
compression parallel to the grain [8,23]. Fig. 3. Different material models for timber.
1108 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
imation of the actual material behaviour. Zakić [26] assumed that
the behaviour of wood subjected to compression parallel to the
grain could be represented by a parabolic function. Bazan [28] pre-
sented a model allowing for a linear softening after reaching the
compressive strength parallel to the grain. The models by Zakić
and Bazan were based on experimental investigations on clear
specimens of limited size subjected to bending.
Glos [29] proposed a more comprehensive model, being valid
not only for clear wood but also for structural timber. Based on
extensive experimental studies on solid timber boards, a polyno-
mial formulation for the stress–strain relationship was obtained.
The advantage of this model is that it represents the actual shape
of the material behaviour more accurately than other models.
The model takes into account both the non-linear material behav-
iour before reaching the compressive strength and the subsequent
softening. Since the model was actually developed for timber
members subjected to axial compression and bending [30], the
stress–strain relationship proposed by Glos was used for the fol-
lowing comparative calculations.
4.4. Comparison of the material models presented in 4.3
The impact of the chosen material model on the load-bearing
capacity of a column was evaluated by means of the strain-based
model. The load-bearing capacity was estimated for a simply sup-
ported column made of solid timber of strength class C24 accord-
ing to EN 338 [21] by applying the different material models
described in Section 4.3. As strength and stiffness properties the
characteristic values (5th percentile) for solid timber of strength
class C24 according to EN 338 [21] were used for all material mod-
els. Therefore, the differences in results are caused only by the dif-
Fig. 4. Influence of the material model on the load-bearing capacity of a simply supported C24 solid timber column with a cross-section of 200 mm  200 mm calculated with
the strain-based model.
Fig. 5. Qualitative representation of the stress–strain relationship for the material model proposed by Glos [29].
ferent consideration of the non-linear material behaviour of timber
when subjected to compression parallel to the grain.
Fig. 4 shows the comparison for simply supported columns with
a cross-section of 200 mm  200 mm. The slenderness ratio k of
the columns was varied between 0 and 150. The calculations were
performed by assuming an initial deformation e0 of 0.0025‘cr
which corresponds to the value proposed by Eurocode 5
EN 1995-1-1.
When comparing the different material models, substantial dif-
ferences of up to 20% can be observed. While for stocky timber col-
umns (k < 20) subjected to pure compression parallel to the grain
all material models provide identical results, the results for timber
columns of intermediate or high slenderness (k P 20) are mark-
edly influenced by the non-linear material behaviour.
4.5. Input data
Appropriate input data is essential to obtain correct results with
the strain-based model. In particular the parameters needed for
the material model are of high importance. When using the mate-
rial model proposed by Glos, six material properties are needed to
mathematically describe the full stress–strain relationship as
shown in Fig. 5. The modulus of elasticity in tension Et,0 and the
tensile strength ft,m,0 describe the linear elastic material behaviour
of timber subjected to tension parallel to the grain. The material
behaviour in compression parallel to the grain is defined by means
of the modulus of elasticity in compression Ec,0, the compressive
strength fc,m,0, the strain at the compressive strength ec,0 and the
residual strength fc,m,u,0 after reaching the compressive strength.
Fig. 6 shows the impact of the above mentioned material prop-
erties on the load-bearing capacity of a simply supported column.
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
Fig. 6. Influence of the material properties on the load-bearing capacity of a simply supported column calculated with the strain-based model.
The sensitivity of the material properties on the load-bearing
capacity was estimated by a comparative calculation for simply
supported columns of different slenderness ratios. One material
property was varied while the rest of the material properties were
kept constant. The figure shows the ratio between the load-bearing
capacity calculated with the initial value X0 of the respective mate-
rial property and a reduced or increased value 0:9 X 0 and 1:1 X 0 . As
initial values X0, characteristic strength and stiffness values for so-
lid timber of strength class C24 according to EN 338 were used.
It can be seen, that for stocky columns the compressive strength
parallel to the grain fc,m,0 is the governing material property. With
increasing slenderness ratio the influence of the compressive
strength decreases while the influence of the MOE (Et,0, Ec,0) in-
creases. In addition, the strain at compressive strength (ec,0) has a
substantial impact on the load-bearing capacity, especially for col-
umns of intermediate slenderness. On the other hand, the tensile
strength ft,m,0 as well as the residual compressive strength fc,m,u,0
have no or only minor influence on the load-bearing capacity. In
the following paragraphs, the different material properties are dis-
cussed in detail.
4.5.1. Stiffness
As already mentioned, the modulus of elasticity (MOE) is one of
the most important material properties. In general, the MOE in ten-
sion and in compression parallel to the grain differ. In various
investigations e.g. [31] differences in the range of several per cent
between tension MOE, compression MOE and bending MOE were
found. However, with regard to a simple calculation procedure to
later on be applied in design codes, the use of one single MOE for
tension and compression is reasonable. Mean values as well as
5th percentile values for the modulus of elasticity are defined in
EN 338 [21] for solid timber and EN 1194 [22] for glued laminated
timber depending on the strength class of the member. In both
standards EN 338 and EN 1194 the postulated MOE parallel to
the grain are valid for compression and tension as well as for bend-
1109
ing. Hence, differing the MOE with respect to type of loading would
be very difficult, since there are no specific values available in the
standards.
4.5.2. Tensile strength
For the strain-based model, not the tensile strength ft,0 of a tim-
ber member which is defined in in EN 338 [21] for solid timber and
EN 1194 [22] for glued laminated timber but rather the ultimate
tensile strength at extreme fibres position ft,m,0 is of interest. This
material property is strongly influenced by the quality of the tim-
ber (timber grade), by the loading situation and by the size of the
member. The strength at extreme fibres position can roughly be
estimated based on the bending strength fm,0 (Fig. 7). Non-linear
deformations in the compression zone lead to a shift of the neutral
axis towards the tension side and a redistribution of the stresses.
Hence, the tensile strength at extreme fibres is higher than the
bending strength, particularly for high strength timber.
However, the tensile strength ft,m,0 influences the load-bearing
capacity of timber members subjected to axial compression or ax-
ial compression and bending only to a very small extent (see
Fig. 6). In general, the failure is governed by the compressive
strength of the member or by the loss of the stability. Hence, a va-
lue for the tensile strength at extreme fibres position 10% higher
than the bending strength is a reasonable approach for all timber
grades.
4.5.3. Compressive strength
As it can be seen in Fig. 6, the compressive strength fc,m,0 plays
an important role, in particular for stocky timber members. Again,
not the compressive strength fc,0 of a timber member is of interest
but rather the strength fc,m,0 which is the compressive strength of
the plastic zone of a member loaded in combined axial compres-
sion and bending. In contrast to the tensile strength, timber loaded
in compression shows a ductile behaviour and as a consequence,
stress redistribution can occur. Additionally, size effects are of min-
1110 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
Fig. 7. Differences in the determination of the bending strength fm,0 and the actual
tensile strength at extreme fibres position ft,m,0.
or importance for timber subjected to compression. Buchanan [8]
found a strength reduction of a few per cent when doubling the
length of the timber member. This size effect factor is in good
agreement with results published in [32]. In this study, also the
width effect was investigated for visually graded timber. It was
found that length effect factors and width effect factors are similar.
As a consequence, the compressive strength of the plastic zone
may be slightly higher compared to the compressive strength of
a timber member. However, both strength values reach almost
the same level and for simplification, the use of fc,m,0 = fc,0 seems
appropriate. The compressive strength of timber was investigated
in different studies [29,33] and characteristic values for solid tim-
ber and glued laminated timber are given in product standards as
e.g. EN 338 and EN 1194.
4.5.4. Strain at ultimate compressive stress
For timber members subjected to axial compression, the stiff-
ness decreases before the compressive strength is reached. In the
material model, this effect is described with the parameter ec,0
which is the strain at ultimate compressive stress. The product
standards EN 338 and EN 1194 do not give any values for this
material property and the strain at ultimate compressive stress le-
vel ec,0 has been investigated in very few studies only [29,33,34]. In
addition, the evaluation of exact values is demanding since there is
no distinct peak in the stress–strain relationship at the ultimate
compressive stress level and the shape of the falling branch of
the curve strongly depends on the test setup and on the loading
rate [8]. One possible approach is to derive ec,0 from the elastic
strain by enlarging the latter with a factor be:
fc;m;0
ec;0 ¼ be  ð10Þ
Ec;0
where ec,0 is the strain at ultimate compressive stress; be is the mod-
ification factor; fc,m,0 is the compressive strength of the plastic zone
and Ec,0 is the modulus of elasticity in compression parallel to the
grain.
O’Halloran [34] performed tests on small clear wood specimens
and reported a value of be = 1.25. This agrees quite well with the re-
sults published by Glos [29]. Based on mean values taken from this
investigation on solid timber boards tested in standard conditions
(controlled environment at 20 °C and a relative humidity of 60%) a
value of be = 1.26 could be found. In [33] compression tests on
specimens made of glued laminated timber are reported, leading
to similar results. Even though these ratios are strictly speaking
only valid on the mean level, here a value of be = 1.25 was used
to derive the strain at ultimate stress from the elastic strain.
4.5.5. Residual compressive strength
As described above, the falling branch of the stress–strain rela-
tionship is markedly influenced by the test setup and the loading
rate. Hence, it is difficult to define exact values for the residual
strength fc,m,u,0 after reaching the ultimate stress and accordingly,
this value is not defined in design codes or product standards. As
shown in Fig. 6, the impact of the residual strength is small. It is
therefore accurate enough to use a simplified approach according
to Eq. (11) with a bf-value of 0.85 which corresponds to a mean va-
lue found in [29].
fc;m;u;0 ¼ bf  fc;m;0 ð11Þ
fc,m,u,0 is the residual compressive strength; bf is the modification
factor and fc,m,0 is the compressive strength of the plastic zone.
4.6. Shear deformations
In the strain-based model, the shear deformations have been
neglected. One possibility to account for the influence of the shear
deformations in the calculation procedure would be to reduce the
modulus of elasticity of the column. For a simply supported col-
umn with sinusoidal deformations, the reduced modulus of elastic-
ity can be written as:
1 1
Ered ¼ E  ¼E ð12Þ
1 þ eev 2
1 þ 2p EI
b ‘cr GAv
where Ered is the reduced modulus of elasticity accounting for shear
deformations; E is the modulus of elasticity (MOE); ev is the maxi-
mal shear deformations; eb is the maximal deformation due to
bending; I is the 2nd moment of inertia; ‘cr is the effective length;
G is the shear modulus and Av is the effective shear area.
For a rectangular cross-section the effective shear area Av can be
written as:
5
Av ¼ A ð13Þ
6
where A is the area of the cross-section.
Alternatively, when performing finite element calculations,
shear deformations may be easily taken into account by consider-
ing the shear stiffness in the stiffness matrix. The influence of shear
deformations on a simply supported column is shown in Fig. 8. The
calculation was performed using characteristic strength and stiff-
ness values for solid timber of strength class C24 according to EN
338 [21]. As it can be seen from Fig. 8, the influence of shear defor-
mations is relatively small. For slender columns, only a few per
cent of the total deformations are caused by the shear deforma-
tions. For stocky columns the total deformation are insignificant
and neither the shear deformations nor the bending deformation
have a major impact on the load-bearing capacity. As a conse-
quence, shear deformations do not considerably influence the
load-bearing behaviour of columns subjected to axial compression
or combined axial compression and bending. For the further calcu-
lations presented in this paper, the shear deformations were
neglected.
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
Fig. 8. Influence of the shear deformations on the load-bearing capacity of a simply supported C24 solid timber column with a cross-section of 200 mm  200 mm calculated
with the strain-based model.
5. Validation of the model with experimental data
5.1. Experimental data
The strain-based model was benchmarked to results of experi-
mental investigations [35,36]. Steiger examined the load-bearing
capacity of solid timber members subjected to combined axial
force N and bending moment M. Extensive experimental tests were
performed using a four point bending test setup (Fig. 9). Thereby,
solid timber members made of Norway spruce (picea abies) were
examined under combined axial load (tension or compression)
and bending moment. The timber members had a length of
2.76 m and a rectangular cross-section of 80 mm  160 mm. These
dimensions correspond to a slenderness ratio k of 60. In order to
study the influence of the timber strength and stiffness on the
interaction behaviour, the tests were carried out with members
of two different timber grades (C27 and C40 according to EN 338
[21]). For each grade, eleven test series were carried out, including
test series with pure axial force, a test series with pure bending
moment and test series with different ratios of axial force N and
lateral forces Flateral. The ultimate loads and the corresponding
deflections at mid-span were recorded. Furthermore, various mate-
rial properties of the test specimens (e.g. modulus of elasticity,
density, moisture content) were determined. In order to get control
of the variations in strength and stiffness and to assess the strength
class of the C27 and C40 sub-samples, the raw material had been
graded by means of ultrasound.
Fig. 9. Test setup for the investigation of solid timber members subjected to combined axial force and bending moment [35,36].
1111
5.2. Modelled data
In the present study, the experimentally derived MOE was used
to estimate the compressive and bending strength of the timber
specimens. Correlations between the material properties as pro-
posed in [37] were used. In [37] correlation measurements were
determined and regression equations were given for solid timber
from the same growing area and for specimens of the same dimen-
sion as tested in [35].
The expected value for the compressive strength parallel to the
grain was taken from [37]. The expected value for the bending
strength fm,0 according to this literature was derived from the test
data assuming a linear stress distribution within the cross-section.
Hence, the values for the tensile strength at extreme fibres position
ft,m,0 had to be estimated according to the model shown in Fig. 7.
The strain corresponding to the ultimate compressive stress ec,0
and the residual strength after reaching the ultimate compressive
stress fc,m,0,u were estimated according to Section 4.5. To simplify
the calculation procedure, constant values were assumed for these
two parameters (be = 1.25 and bf = 0.85). Furthermore, the initial
deflection of the specimens was neglected, since the P-delta effect
in the tests by Steiger [35,36] were not caused by initial deflection
but rather by lateral loading.
Based on these parameters, the load-bearing capacity of the test
specimens was predicted using the strain-based model. The calcu-
lation procedure described in Section 3 was slightly modified to ac-
count for the particular test setup: The lateral load applied to the
1112 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
specimens influenced the shape of the member’s deflection. As a
consequence, the integration constant c had to be adjusted itera-
tively for each load increment.
5.3. Comparison of modelled data and experimental data
A total of 148 test results were used to validate the strain-based
model. The investigation was performed for six different ratios of
axial compression and bending moment (including test series with
pure bending and pure axial compression).
In Fig. 10 a comparison between the predicted values and the
experimental data is shown for three different ratios of axial com-
pression and bending moment. The calculated and the experimen-
tally determined load-bearing capacities are plotted on the Y- and
X-axis, respectively. For points below the solid black line inclined
under 45°, the predicted values are lower than the actual values,
while points above the solid line represent test results, where
the strain-based model overestimates the load-bearing capacity
of the test specimens.
In Table 2 a more detailed comparison is made. For the different
test series the mean value l of the ratio between the predicted
load-bearing capacity Nmodel and the experimentally determined
load-bearing capacity Ntest is given. Furthermore, the coefficient
of correlation q is listed.
The test series with pure axial compression or combined axial
compression and bending exhibited good agreement between the
model and the experiments. The mean value of the ratio Nmodel/
Ntest for all test series lies in the range of 0.91–0.97. Also a high
coefficient of correlation between the modelled and the experi-
mentally determined data was found.
For the tests conducted in pure bending, it can be observed that
the model overestimates the load-bearing capacity considerably. In
addition, the coefficient of correlation for this test series is quite
Fig. 10. Comparison between the experimentally determined load-bearing capacity Ntest and the calculated load-bearing capacity Nmodel.
Table 2
Mean value l and coefficient of correlation q of the predicted and the experimentally determined load-bearing capacity.
Test series Ratio Number of tests
MI/N n Nmodel/Ntest
Pure bending 1 23
Compression and bending 0.307 m 20
0.204 m 20
0.144 m 20
0.100 m 21
Pure compression 0 44
Total 148
low. This has various reasons. First of all, the load-bearing capacity
of members subjected to pure bending strongly depends on the
tensile strength at ultimate fibres ft,m,0. As explained in Sec-
tion 4.5.2, this material property is very difficult to evaluate and
it is characterised by a high variability. In addition, only 23 exper-
iments in pure bending had been performed. The test series with
pure axial compression and combined axial compression and
bending contain a larger database with 44 and 81 tests,
respectively.
In general, the investigations showed a good agreement be-
tween the estimated and the experimentally derived values. The
strain-based model slightly underestimates the load-bearing
capacity of the timber members; the mean value of the ratio
Nmodel/Ntest is 0.93.
6. Comparison of the strain-based model to 2nd order linear
elastic analysis of the structure and ELM
The strain-based model was finally compared to the codified
design approaches (see Section 2). The load-bearing capacity was
calculated for a simply supported solid timber column of strength
class C24 with a constant cross-section of 200 mm  200 mm. The
comparison was done on the level of characteristic values. For that
reason, 5th percentile values for the strength properties were cho-
sen for the calculations. For the modulus of elasticity characteristic
(5th percentile) values were taken as well. This is in contradiction
to the design rules given in Eurocode 5 EN 1995-1-1 or SIA 265
where the design is based on mean values of the modulus of elas-
ticity. However, it ensures a consistent material model and compa-
rable results.
Furthermore, the calculations were performed considering an
initial deformation e0 of 0:0017 ‘cr . This value corresponds to a
95th percentile estimated from in situ measurements performed
Axial force Bending moment
Mmodel/Mtest
l q l q
1.22 0.61
0.97 0.94
0.92
0.93
0.91
0.94 0.78
0.93 1.22
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
by Ehlbeck and Blaß [38]. In this investigation the initial deforma-
tion of 142 real columns made of solid timber were measured. The
measured values were considerably smaller than the initial defor-
mation recommended by Eurocode 5 EN 1995-1-1 (0:0025 ‘cr ).
However, the results of [38] were the basis for the estimation of
the buckling factor kc used in ELM. Consequently, the calculation
with the 2nd order linear elastic analysis and the strain-based
model were performed with the same initial deformation to ensure
consistency.
6.1. Comparison of the strain-based model and ELM
As described above, the same 5th percentile values were chosen
for the material properties as well as for the geometric imperfec-
tions for both calculation procedures. However, a direct compari-
son of the strain-based model and the simplified calculation
model based on the Effective Length Method (ELM) is not possible.
The buckling factor kc used in ELM as codified in Eurocode 5 EN
1995-1-1 was estimated on the basis of Monte Carlo simulations.
Hence, some inherently stochastic characteristics of the material
are already included in the buckling factor kc. The load-bearing
capacity calculated with ELM corresponds to a 5th percentile value
of the load-bearing capacity. On the other hand, in the strain-based
model the 5th percentile was used for every single material prop-
erty to calculate the load-bearing capacity. As a consequence, the
calculated load-bearing capacity does not necessarily correspond
to a 5th percentile value of the load-bearing capacity. Conse-
quently, it is not surprising that the values calculated with the
Effective Length Method are slightly higher than the results ob-
tained with the strain-based model as it can be seen from Fig. 11.
6.2. Comparison of the strain-based model and the 2nd order linear
elastic analysis of the structure
The strain-based model and the 2nd order linear elastic analysis
can be directly compared. Both models are based on the same in-
put data. Hence, occurring differences were caused only by the dif-
ferent model assumptions. Actually, the comparison shows
considerable differences (Fig. 11). When comparing the two calcu-
lation models, a difference of more than 15% is found for columns
with slenderness ratios between 50 and 70.
In addition, Fig. 11 shows the differences between the Effective
Length Method and the 2nd order linear elastic analysis. A similar
comparison has already been shown in Fig. 1. The comparison
shown in Fig. 1 is based on design values provided by the codes
while in Fig. 11 the comparison is based on 5th percentile values.
As a consequence, the differences between the Effective Length
Fig. 11. Comparison between the strain-based model, ELM and the 2nd order linear elastic analysis of the structure.
1113
Method and the 2nd order linear elastic analysis observed in
Fig. 1 are caused by differences in the stiffness values, by the influ-
ence of moisture content and the duration of load, and by the non-
linear material behaviour. In Fig. 11 both calculations are based on
the same stiffness values and the influences of moisture content
and duration of load are neglected. Accordingly, the differences ob-
served in Fig. 11 are mainly caused by the fact that the 2nd order
linear elastic analysis neglects the influence of the non-linear
material behaviour.
6.3. How to reach a better fit?
The 2nd order linear elastic analysis of the structure is based on
the assumption of linear elastic material behaviour, which means
that the reduction of the stiffness caused by the non-linear mate-
rial behaviour is neglected and the load-bearing capacity is overes-
timated. In order to reach a better agreement a correction factor
has to be introduced in the design equation. Possibilities to modify
the design approach are (i) to enlarge the initial deformations or
(ii) to reduce the design stiffness.
In particular, the reduction of the stiffness seems to be a practi-
cable solution, since it describes the physical phenomena quite
well. In 1895 Engesser [39,40] firstly introduced the reduction of
the stiffness for steel columns and suggested to use a tangent mod-
ulus instead of the modulus of elasticity. Engesser’s theory was
first questioned by other scientists but Shanley showed in 1947
[41], that Engesser’s method was a valuable possibility to account
for non-linear deformations of the compression zone.
In most of the national and international design codes the con-
cept of performing a 2nd order linear elastic analysis of the struc-
ture using a reduced value for the stiffness is implemented.
However, as mentioned in Section 2.3, there are large differences
regarding the design value of the stiffness used in the calculation.
Furthermore, as it can be seen in Fig. 1, none of the results obtained
with the 2nd order linear elastic analysis agree well with the re-
sults derived with the Effective Length Method.
There are many parameters influencing the results, thus it is dif-
ficult to give concluding advices concerning the design rules. How-
ever, some suggestions for a more accurate design procedure can
be given as follows:
– For stocky columns (k < 20) the different design approaches
and the strain-based model provided similar results. The load-
bearing capacity is primarily governed by the compressive
strength parallel to the grain. This material property is well
known and the member fails in more or less ductile manner.
As a consequence, the design of stocky columns is accurate
enough.
1114 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
– For columns with a slenderness ratio higher than 20, the influ-
ence of the reduction in stiffness cannot be neglected (Fig. 11).
This influence is particularly distinctive for columns of slender-
ness ratios k between 40 and 90. The use of a tangent modulus
instead of the modulus of elasticity (MOE) is reasonable for
these columns. However, the tangent modulus is not constant
but depends on the loading and in particular on the axial com-
pression applied. Dubas [42] suggested to reduce the MOE
depending on the slenderness ratio for compression members
of truss girders with a slenderness ratio k smaller than 100. A
simplified approach would be to use a tangent modulus of
T 0;05  0:8 E0;05 as suggested in [39–41].
– For slender columns, the MOE rather than the compressive
strength governs their load-bearing capacity (Fig. 6). Hence, it
is suggested to perform the 2nd order linear elastic analysis
based on a 5th percentile value of the MOE rather than a mean
value. This concept was already proposed in the withdrawn
German code DIN 1052. According to the European standards
EN 338 and EN 1194 the 5th percentile of the MOE is defined
as E0:05 ¼ 0:67 E0;mean for coniferous solid timber and as
E0:05 ¼ 0:85 E0;mean for glued laminated timber.
However, the use of the 5th percentile of the MOE is only correct
for columns of infinite slenderness ratios. For columns with a finite
length, the load-bearing capacity is governed not only by the MOE
but by an entire set of parameters (such as the MOE, the compres-
sive strength, the strain at compressive strength and the initial
deformation). The load-bearing capacity calculated with the 5th
percentile of a set of parameters is in general slightly higher than
the load-bearing capacity calculated with the 5th percentile of
every single parameter. Since in general only 5th percentile values
for the single parameters are available, the use of 5th percentile
values seems to be reasonable for columns of high slenderness
(k P 100).
– When calculating design values for the load-bearing capacity,
the strength properties as well as the stiffness properties have
to be divided by the partial safety factor cM.
– The stiffness of timber is influenced by the moisture content
and the duration of load [23]. As a consequence, not only the
strength properties but also the moduli of elasticity have to
be reduced by a modification factor taking into account these
influences, as it was already done in the former Swiss design
code SIA 164 [43] and still is in the actual one (SIA 265).
Fig. 12 shows a comparison between the design approaches gi-
ven in Eurocode 5 EN 1995-1-1 [5], the strain-based model and the
Fig. 12. Design value of the load-bearing capacity as obtained with different design approaches.
modified approaches using a tangent modulus as described above.
The comparison is performed for the same example as shown in
Fig. 1. Consequently, the curves representing the design ap-
proaches given in Eurocode 5 EN 1995-1-1 were taken from this
figure. The curve representing the strain-based model is calculated
based on 5th percentile values. This characteristic load-bearing
capacity is divided by the factor cM/kmod to obtain design values.
For the modified design approaches, a 2nd order linear elastic anal-
ysis was performed using a tangent modulus instead of the modu-
lus of elasticity. On the one hand, the tangent modulus was defined
as T d ¼ ð0:8 E0:05 Þ=cM as suggested in [39–41]. On the other hand,
the calculation was performed using a tangent modulus which de-
pends on the slenderness ratio as suggested in [42]. For the estima-
tion of the tangent modulus, a parabolic function with Td = 0 for
k ¼ 0 and T d ¼ E0:05 =cM for k ¼ 100 was chosen for columns with
a slenderness ratio smaller than 100. For k > 100 the tangent mod-
ulus was defined as T d ¼ E0:05 =cM . To ensure a consistent design,
characteristic values (i.e. 95th percentile) were chosen for the ini-
tial deflection. The influence of moisture content and duration of
load on the stiffness properties is neglected for all curves shown.
Fig. 12 shows that the strain-based model as well as the two
modified design approaches correspond quite well to the Effective
Length Method. When using the 2nd order linear elastic analysis as
described in [5] the load-bearing capacity is overestimated. How-
ever, the figure only shows one single example. Depending on
the strength and stiffness properties (timber grade) as well as on
the influence of the moisture content and the duration of load,
the results can be different. In addition, also the selection of appro-
priate initial deflections influences the results of the calculation.
7. Conclusions
When designing timber members subjected to simultaneously
acting axial compression and bending moment, the increase of
the bending moment due to the eccentricity of the axial force
and due to the non-linear material behaviour of timber subjected
to compressive stress has to be taken into account. The current de-
sign codes provide two different approaches for the design of
respective members (simplified analysis based on the Effective
Length Method and 2nd order linear elastic analysis of the struc-
ture). However, the two design approaches are not consistent
and can lead to different results.
The load-bearing behaviour of centrically and eccentrically
loaded timber columns can be described by means of a strain-
based model. In principle, the strain-based model can be seen as
a 2nd order analysis with the following special characteristics:
 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
– When performing a 2nd order analysis of the structure it is
often based on the assumption of linear elastic material behav-
iour. In contrast, non-linear material behaviour can easily be
implemented in the strain-based model. The non-linearity in
the material behaviour has a distinctive influence on the load-
bearing behaviour of timber members subjected to axial com-
pression or combined compression and bending.
– There are differences in the calculation procedure. When per-
forming a 2nd order structural analysis the acting force and
moment are firstly calculated and then compared to the resis-
tance of the cross-section. Within the strain-based model the
external force and moment and the internal force and moment
are directly linked to each other by the curvature. As a conse-
quence, the resistance of the cross-section is directly imple-
mented in the calculation model. There is no need for the use
of a moment – axial force interaction formula.
By benchmarking calculated values to experimental results the
model has been proven capable in providing adequate estimations
of the load-bearing capacity of solid timber columns subjected to
axial compression or combined compression and bending. A de-
tailed investigation using the strain-based model showed that:
– The choice of the material model strongly influences the result
of the calculation. While the simplified bilinear elastic–plastic
stress–strain relationship may be appropriate for modelling
timber subjected to bending, a more advanced model is needed
when modelling timber subjected to combined axial compres-
sion and bending where failure is governed by loss of stability.
– The load-bearing capacity of stocky columns (k < 20) is gov-
erned by the compressive strength. For slender columns
(k P 100) the modulus of elasticity (MOE) is the dominant
material property. For columns of intermediate slenderness
ratio, the compressive strength, the MOE and the non-linear
material behaviour have an impact on the load-bearing
capacity.
– Shear deformations do not considerably influence the load-
bearing capacity of timber columns.
Based on these investigations, some conclusions concerning the
design approaches given in selected current design codes can be
made:
– When performing a 2nd order linear elastic analysis of the
structure, the non-linear material behaviour of timber cannot
be taken into account. Consequently, an adjustment of the
results obtained with this method is required. This can be done
by reducing the stiffness of the structural member. The use of a
tangent modulus for columns of intermediate slenderness ratio
(20 6 k < 100) would be an appropriate solution. It is suggested
to adopt this concept in a future version of the Eurocode 5 EN
1995-1-1 [5]. If so, however, also the values for the initial defor-
mations have to be adjusted in order to ensure a consistent
design approach.
– The design should be based on 5th percentile values divided by
the partial safety factor cM for the strength properties as well as
for the stiffness properties. This concept was already proposed
in the withdrawn German code DIN 1052 [7]. This is a simple
adjustment, leading to more consistent results, in particular
for situations where the results are not affected by the moisture
content or the duration of load.
– The results presented here are strictly speaking only valid for
short-term response under load at constant interior climate.
However, the influence of the moisture content (MC) and the
duration of load (DOL) on the modulus of elasticity should be
considered. In the current version of the Eurocode 5 EN 1995-
1115
1-1 [5] this so called long-term stability is not taken into
account. When developing the next generation of Eurocode 5
this topic should be considered.
Based on this conclusions it is recommended to use the Effec-
tive Length Method when designing timber members subjected
to axial compression or combined axial compression and bending
according to the current version of the Eurocode 5 EN 1995-1-1
[5] or the Swiss code SIA 265 [6]. Alternatively a 2nd order s linear
elastic analysis can be performed using the design values for the
stiffness properties as proposed by the withdrawn DIN 1052 [7].
Acknowledgement
The authors gratefully acknowledge the financial support by the
Swiss National Science Foundation (SNSF) provided through the re-
search Project No. 200021-126655.
References
[1] Tetmajer L. Die Gesetze der Knickungs- und der zusammengesetzten
Druckfestigkeit der technisch wichtigsten Baustoffe. Zurich, Switzerland:
Materialprüfungs-Anstalt am Schweiz. Polytechnikum Zurich; 1896.
[2] Larsen HJ, Pedersen SS. Tests with centrally loaded timber columns. In:
Proceedings of CIB-W18 meeting 4: Paper No. 4-2-1. Paris, France; 1975.
[3] Blaß HJ. Tragfähigkeit von Druckstäben aus Brettschichtholz unter
Berücksichtigung streuender Einflussgrössen [Dissertation]. Karlsruhe,
Germany: Universität Fridericiana Karlsruhe; 1987.
[4] Blaß HJ. Design of columns. In: Proceedings of the 1991 international timber
engineering conference, vol. 1. London: TRADA; 1991. p. 1.75–1.81.
[5] CEN. EN 1995-1-1. Eurocode 5: Design of timber structures. Part 1–1: General
– Common rules and rules for buildings. Brussels, Belgium: European
Committee for Standardization; 2004.
[6] SIA. SIA 265: Timber structures. Zurich, Switzerland: Swiss Society of
Engineers and Architects; 2012.
[7] DIN. DIN 1052: Entwurf, Berechnung und Bemessung von Holzbauwerken –
Allgemeine Bemessungsregeln und Bemessungsregeln für den Hochbau.
Berlin, Germany: Beuth; 2008.
[8] Buchanan AH. Strength model and design methods for bending and axial load
interaction in timber members [Dissertation]. Vancouver, Canada: University
of British Columbia; 1984.
[9] Buchanan AH, Johns KC, Madsen B. Column design methods for timber
engineering. Can J Civil Eng 1985;12(4):731–44.
[10] Kessel MH, Schönhoff T, Hörsting P. Zum Nachweis von druckbeanspruchten
Bauteilen nach DIN 1052:2004–08, Teil 1. Bauen mit Holz
2005;107(12):88–96.
[11] Kessel MH, Schönhoff T, Hörsting P. Zum Nachweis von druckbeanspruchten
Bauteilen nach DIN 1052:2004–08, Teil 2. Bauen mit Holz 2006;108(1):41–4.
[12] Möller G. Zur Traglastermittlung von Druckstäben im Holzbau. Bautechnik
2007;84(5):329–34.
[13] Köhler J, Frangi A, Steiger R. On the role of stiffness properties for ultimate
limit state design of slender columns. In: Proceedings of CIB-W18 meeting 41:
Paper No. 41-1-1. St. Andrews, Canada; 2008.
[14] Becker P, Rautenstrauch K. Time-dependent material behavior applied to
timber columns under combined loading. Part II: Creep-buckling. Holz als Roh-
und Werkstoff 2001;59(6):491–5.
[15] Hartnack R, Schober K-U, Rautenstrauch K. Computer simulations on the
reliability of timber columns regarding hygrothermal effects. In: Proceedings
of CIB-W18 meeting 35: Paper No. 35-2-1. Kyoto, Japan; 2002.
[16] Hartnack R. Langzeitverhalten von druckbeanspruchten Bauteilen aus Holz
[Dissertation]. Weimar, Germany: Bauhaus-Universität Weimar; 2004.
[17] Becker P. Modellierung des zeit- und feuchteabhängigen Materialverhaltens
zur Untersuchung des Langzeittragverhaltens von Druckstäben aus Holz
[Dissertation]. Weimar, Germany: Bauhaus-Universität Weimar; 2002.
[18] Hartnack R, Rautenstrauch K. Long-term load bearing of wooden columns
influenced by climate – view on code. In: Proceedings of CIB-W18 meeting 38:
Paper No. 38-2-1. Karlsruhe, Germany; 2005.
[19] Blaß HJ. Buckling length. In: Timber engineering, STEP 1. Almere, Netherlands:
Centrum Hout; 1995.
[20] Bažant ZP, Cedolin L. Stability of structures: elastic, inelastic, fracture, and
damage theories. New York, USA: Oxford University Press; 1991.
[21] CEN. EN 338: Structural timber – strength classes. Brussels, Belgium: European
Committee for Standardization; 2009.
[22] CEN. EN 1194: Timber structures – glued laminated timber – strength classes
and determination of characteristic values. Brussels, Belgium: European
Committee for Standardization; 1999.
[23] Madsen B. Structural behaviour of timber. North Vancouver, Canada: Timber
Engineering Ltd.; 1992.
[24] Hörsting OP. Zum Tragverhalten druck- und biegebeanspruchter Holzbauteile
[Dissertation]. Braunschweig, Germany: TU Braunschweig; 2008.
1116 M. Theiler et al. / Engineering Structures 56 (2013) 1103–1116
[25] Cabrero J, Blanco C, Gebremedhin K, Martin-Meizoso A. Assessment of
phenomenological failure criteria for wood. Eur J Wood Wood Prod
2012;70(6):871–82.
[26] Zakić BD. Inelastic bending of wood beams. J Struct Div 1973;99(10):
2079–95.
[27] Neely ST. Relation of compression-endwise to breaking load of beam. In:
Progress in timber physics: USDA Division of forestry, Circular No. 18; 1898. p.
13–7.
[28] Malhotra SK, Bazan IMM. Ultimate bending strength theory for timber beams.
Wood Sci 1980;13(1):50–63.
[29] Glos P. Zur Bestimmung des Festigkeitsverhaltens von Brettschichtholz bei
Druckbeanspruchung aus Werkstoff- und Einwirkungskenngrössen
[Dissertation]. Munich, Germany: TU München; 1978.
[30] Glos P. Zur Modellierung des Festigkeitsverhaltens von Bauholz bei Druck-,
Zug- und Biegebeanspruchung. Munich, Germany: Sonderforschungsbereich
96 & Laboratorium für den konstruktiven Ingenieurbau – TU München; 1981.
[31] Steiger R, Arnold M. Strength grading of Norway spruce structural timber:
revisiting property relationships used in EN 338 classification system. Wood
Sci Tecnol 2009;43(3–4):259–78.
[32] Barrett JD, Lam F, Lau W. Size effects in visually graded softwood structural
lumber. J Mater Civil Eng 1995;7:19–30.
[33] Frese M, Enders-Comberg M, Blaß HJ, Glos P. Compressive strength of spruce
glulam. Eur J Wood Wood Prod 2012;70(6):801–9.
[34] O’Halloran MR. Curvilinear stress-strain relationship for wood in compression
[Dissertation]. Fort Collins, USA: Colorado State University; 1973.
[35] Steiger R. Versuche an Fichten-Kanthölzern: Biegemoment-Normalkraft-
Interaktion. Basel, Switzerland: Birkhäuser; 1995.
[36] Steiger R, Fontana M. Bending moment and axial force interacting on solid
timber beams. Mater Struct 2005;38(5):507–13.
[37] Steiger R. Mechanische Eigenschaften von Schweizer Fichten-Bauholz
bei Biege-, Zug-, Druck- und kombinierter M/N-Beanspruchung
[Dissertation]. Zurich, Switzerland: ETH Zurich; 1996.
[38] Ehlbeck J, Blaß HJ. Imperfektionsannahmen für Holzdruckstäbe. Holz als Roh-
und Werkstoff 1987;45(6):231–5.
[39] Knickfragen Engesser F. Schweizerische Bauzeitung 1895;25(13):88–91.
[40] Engesser F. Ueber Knickfragen. Schweizerische Bauzeitung 1895;26(4):24–6.
[41] Shanley FR. Inelastic column theory. J Aeronaut Sci 1947;14(5):261–7.
[42] Dubas P. Stabilitätsprobleme. In: Einführung in die Norm SIA 164 (1981) –
Holzbau. Zurich, Switzerland: Lehrstuhl für Baustatik und Stahlbau, ETH
Zurich; 1981.
[43] SIA. SIA 164: Holzbau. Zurich, Switzerland: Swiss Society of Engineers and
Architects; 1981.