INTER

International Network on Timber Engineering Research

INTER ‐ International Network on Timber Engineering Research

2014 the International Network on Timber Engineering Research (INTER) was

founded.

Scope
Presentation, discussion and documentation of research results in timber
engineering and development of application rules for timber design codes or
standards related to timber engineering.
Approach
Annual meetings in different countries/places hosted by meeting participants
Presentation and discussion of papers
Peer review of the abstracts before the meeting and of the papers during the
meeting
Decision of the acceptance of the abstracts before the meeting by a well‐
defined review process
Decision of the acceptance of the papers for the proceedings during the
meeting
Publication of the papers and the discussion in proceedings
MEETING FORTY‐EIGHT 2015

Rules
All decisions including the appointment of the chairperson or the location of
annual meetings are made by the participants attending a meeting.
Membership
Persons contributing to or being interested in research related to timber
engineering. MEETING FORTY‐EIGHT
ŠIBENIK, CROATIA
INTER PROCEEDINGS

AUGUST 2015
 INTER
International Network on Timber Engineering Research

Proceedings
Meeting 48

24 - 27 August 2015

Šibenik, Croatia

Edited by Rainer Görlacher

Timber Scientific Publishing

KIT Holzbau und Baukonstruktionen
Karlsruhe, Germany
2015
Publisher:
Timber Scientific Publishing
KIT Holzbau und Baukonstruktionen
Reinhard-Baumeister-Platz 1
76131 Karlsruhe
Germany
2015

ISSN 2199-9740
Table of Contents

1 List of Participants 1

2 Minutes of the Meeting 5

3 INTER Papers, Šibenik, Croatia 2015 23

48 - 2 - 1 Proposal of a Eurocode-based Method for the Buckling 27

Design of Timber Log-walls - C Bedon, M Fragiacomo,
C Amadio
48 - 2 - 2 Design of Timber Members Subjected to Axial 45
Compression or Combined Axial Compression and
Bending Based on 2nd Order Theory - A Frangi, M
Theiler, R Steiger
48 - 6 - 1 Rolling Shear Properties of some European Timber 61
Species with Focus on Cross Laminated Timber (CLT):
Test Configuration and Parameter Study - T Ehrhart, R
Brandner, G Schickhofer, A Frangi
48 - 7 - 1 A Universal Approach for Withdrawal Properties of 79
Self-Tapping Screws in Solid Timber and Laminated
Timber Products - A Ringhofer, R Brandner, G
Schickhofer
48 - 7 - 2 Characteristic Withdrawal Capacity and Stiffness of 99
Threaded Rods - H Stamatopoulos, K A Malo
48 - 7 - 3 Load-carrying Capacity of Dowelled Connections - H J 115
Blass, F Colling
48 - 7 - 4 Evaluation of the Reliability of Design Approaches for 131
Connections Perpendicular to the Grain - R Jockwer, R
Steiger, A Frangi
48 - 7 - 5 Simplified Fatigue Design of Typical Timber-Concrete 147
Composite Road Bridges - K Kudla, U Kuhlmann
48 - 12 - 1 Concentrated Load Introduction in CLT Elements 163
Perpendicular to Plane - T Bogensperger, R A Jöbstl
48 - 12 - 2 Shear Properties of Cross Laminated Timber (CLT) 181
under in-plane Load: Test Configuration and
Experimental Study - R Brandner, P Dietsch, J
Dröscher, M Schulte-Wrede, H Kreuzinger, M Sieder, G
Schickhofer, S Winter
48 - 12 - 3 Advanced Modelling for Design Helping of 203
Heterogeneous CLT Panels in Bending - L Franzoni, A
Lebée, F Lyon, G Foret
48 - 12 - 4 Performance of Canadian Glulam Columns with New 217
Laminae E Requirements - F Lam, Jung Kwon Oh, BJ
Yeh, Jun-Jae Lee
48 - 12 - 5 Design of CLT Beams with Large Finger Joints at 233
Different Angles - M Flaig
48 - 15 - 1 Performance-Based Seismic Design of Light-frame 251
Structures – Proposed Values for Equivalent Damping -
J Hummel, W Seim
48 - 15 - 2 Simplified Wall Bracing Method Using Wood Structural 271
Panel Continuous Sheathing - T Skaggs, B J Yeh, E Keith
48 - 15 - 3 Structural Characterization of Multi-storey CLT 285
Buildings Braced with Cores and Additional Shear
Walls. - A Polastri, L Pozza, C Loss, I Smith
48 - 15 - 4 Dissipative Connections for Squat or Scarcely Jointed 301
CLT Buildings - Experimental Tests and Numerical
Validation - R Scotta, L Pozza, D Trutalli, L Marchi, A
Ceccotti
48 - 16 - 1 Analysis of Fire Resistance Tests on Timber Members 317
in Compression with Respect to the Reduced Cross-
Section Method - J Schmid, M Klippel, A Liew, A Just, A
Frangi
48 - 19 - 1 A Beam Theory Fracture Mechanics Approach for 335
Strength Analysis of Beams with a Hole - H Danielsson,
P J Gustafsson
48 - 19 - 2 A Strongest Link Model Applied to Fracture 351
Propagating Along Grain - P-J Gustafsson, R Jockwer, E
Serrano, R Steiger
48 - 102 - 1 A Proposal for a new Background Document of 369
Chapter 8 of Eurocode 8 - M Follesa, M Fragiacomo, D
Vassallo, M Piazza, R Tomasi, S Rossi, D Casagrande
48 - 102 - 2 Aspects of Code Based Design of Timber Structures - J 391
Köhler, G Fink
48 - 102 - 3 The Reality of Seismic Engineering in a Modern Timber 405
World - T Smith, D Moroder, F Sarti, S Pampanin, A H
Buchanan
48 - 102 - 4 An Execution Standard Initiative for Timber 423
Construction - T Toratti
4 INTER Notes, Šibenik, Croatia 2015 439

Start Time of Charring of Timber Members Protected with Gypsum 441

Plasterboards - A Just, K Kraudok
Protective Effect of Insulation Materials on Charring of Timber 445
Elements - M Tiso, A Just
Investigation of Thread Pattern for Self-tapping Screws as 449
Reinforcement for Embedment Strength - Cong Zhang, Wen-Shao
Chang, R Harris

5 Peer Review of Papers for the INTER Proceedings 453

6 Meeting and List of CIB-W18 and INTER Papers 455

1 List of participants
AUSTRIA
T Bogensperger holz.bau forschungs gmbh
M Retuznik proHolz Austria, Vienna
G Schickhofer Technical University of Graz

CANADA
M Kovacs McMaster University, Hamilton
F Lam University of British Columbia, Vancouver
M Popovski FPInnovations, Vancouver
A Salenikovich Université Laval, Québec
I Smith University of New Brunswick
T Tannert University of British Columbia, Vancouver

CHINA
Minjuan He Tongji University
Zheng Li Tongji University

CROATIA
M Jelec University of Osijek
V Rajcic University of Zagreb
M Stepinac University of Zagreb
J Barbalic University of Zagreb

ESTONIA
A Just Tallinn University of Technology
M Tiso Tallinn University of Technology

FINLAND
A Kevarinmäki VTT, Expert Services Ltd.
T Toratti Wood Working Industries Federation, Helsinki

FRANCE
L Franzoni Université Paris-Est
R Pommier University of Bordeaux
GERMANY
H J Blaß Karlsruhe Institute of Technology (KIT)
P Dietsch Technical University of Munich
M Flaig Karlsruhe Institute of Technology (KIT)
M Frese Karlsruhe Institute of Technology (KIT)
R Görlacher Karlsruhe Institute of Technology (KIT)
M Gräfe Technical University of Munich
J Hummel University Kassel
M Kleinhenz Technical University of Munich
K Kudla University of Stuttgart
U Kuhlmann University of Stuttgart
S Loebus Technical University of Munich
W Seim University Kassel
S Winter Technical University of Munich

ITALY
C Bedon University of Trieste
A Ceccotti University of Venice
M Follesa University of Sassari
L Marchi University of Padova
C Nuti RomaTre
L Pozza University of Padova
A Polastri CNR-IVALSA

JAPAN
M Yasumura Shizuoka University

NEW ZEALAND
D Moroder University of Canterbury
F Sarti University of Canterbury
F Scheibmair University of Auckland
P Zarnani University of Auckland

NORWAY
P AAnensen Norwegian Institute of Wood Technology (NTI)
A S Nygård Norwegian University of Life Sciences
G Glaso Norwegian Institute of Wood Technology (NTI)
K A Malo NTNU Trondheim
H Stamatopoulos NTNU Trondheim
SLOWENIA
A Blatnik Plata d.o.o
B Dujic Contemporary building design, Celje

SWEDEN
H Danielsson Lund University
P J Gustafsson Lund University
J Schmid SP Wood Technology
E Serrano Lund University

SWITZERLAND
T Ehrhart ETH Zürich
G Fink Swiss Federal Laboratories for Material Science (EMPA)
S Franke Bern University of Applied Sciences
R Jockwer Swiss Federal Laboratories for Material Science (EMPA)
P Kobel ETH Zürich
J Ogrizovic ETH Zürich

UNITED KINGDOM
R Harris University of Bath
K Ranasinghe TRADA, High Wycombe
Wen-Shao Chang University of Bath
Cong Zhang University of Bath

USA
T Skaggs American Plywood Association, Tacoma
B Yeh American Plywood Association, Tacoma
4
2 Minutes of the Meeting
by F Lam, Canada

CHAIRMAN'S INTRODUCTION

Prof. Hans Blass welcomed the delegates to the International Network on Timber
Engineering Research (INTER). More than 70 participants registered for the meeting.
This shows the continued interest in and the tradition of our meeting in light of
changing the name of our group to INTER in Bath last year. The chairman thanks our
host V Rajčić for the arrangement of the venue and looks forward to interesting
presentations and discussions during our meeting. INTER will continue our tradition
of yearly meetings to discuss research results with the aim of transferring the
information into practical application.
There are 24 papers accepted for this meeting. Papers brought directly to the
meeting would not be accepted for presentation, discussions, or publication. The
same rule applies to papers where none of the authors are present and papers which
are not defended by one of the authors. In the second case, we have made a one-
time exception as the flights of authors for Paper 48-2-2 were unexpectedly cancelled
and R Jockwer will present the work.
The presentations are limited to 20 minutes each, allowing time for meaningful
discussions after each presentation. The Chair asked the presenters to conclude their
respective presentations with a general proposal or statements concerning impact of
the research results on existing or future potential applications and development in
codes and standards.
There are 7 topics covered in this meeting: Timber columns (2), Stresses for solid
timber (1), Timber joints and fasteners (5), Laminated members (5), Structural
stability (4), Fire (1), Fracture Mechanics (2), and Structural design codes (4).
Numbers in parentheses are the number of papers presented in each topic.
The participants have the possibility of presenting notes towards the end of the
technical session. R Görlacher has brought a list of intended note presentations.
Participants intending to present notes that are not on the list should notify R
Görlacher accordingly.
An address list of the participants will be circulated for verification of accuracy.
The Vice Dean of International Cooperation from the University of Zagreb, Faculty of
Civil Engineering gave a welcome address to our group. V Rajčić introduced the host
University of Zagreb and the current state of using wood in buildings in Croatia. V
Rajčić also welcomed the delegates.

5
GENERAL TOPICS

S Winter gave a presentation on the status report of CEN/TC 250/SC5. New items in
Eurocode 5 include: CLT & reinforcements and Timber concrete composites. I Smith
commented that Canadian code also faces similar challenges. E Serrano received
clarifications about project team formation, timing and next phase.

TIMBER COLUMNS

48 - 2 - 1 Proposal of a Eurocode-based Method for the Buckling Design of Timber

Log-walls – C Bedon, M Fragiacomo, C Amadio
Presented by C Bedon
H Blass asked about the G value of 500 MPa in Slide 10 and suggested that shear in
buckling would be rolling shear in these walls. C Bedon responded that the chosen
value was based on calibration with test results. Preliminary study showed that using
a lower G value did not agree with data. Also they did not notice a lot of shear
deformation as the rotation was almost a rigid body motion.
G Schickhofer commented that it would be better to consider a minimum
requirement rather than such a complicated approach for a simple block wall house.
F Lam asked why the cases with design properties were higher than the cases with
mean properties. C Bedon agreed and responded that the process also involved a
normalization process.
A Ceccotti asked if the authors checked how far the designs are away from capacities
in terms of safety. C Bedon answered that cases where stiffness of walls with large
openings can be important and cases where top restraints were not available can
also be important.
U Kuhlmann commented that compressive strength versus critical loading is
confusing.

48 - 2 - 2 Design of Timber Members Subjected to Axial Compression or Combined

Axial Compression and Bending Based on 2nd Order Theory – A Frangi, M
Theiler, R Steiger
Presented by R Jockwer
S Winter asked why machine grading was not used to select the laminates. R Jockwer
responded that no machine grading was available at the time of the project and
visual grading is more common.

6
R Harris commented that the text of the paper seemed to suggest used of Emean for
single member design but the slides clarified that Emean is used for system approach.
P Dietsch commented that there did not seem to be a difference between Eurocode
5 and the approach proposed by the paper where γm is already used.
I Smith asked how often the failures occurred at mid height. R Jockwer responded
that the columns quite often failed at mid height. I Smith commented that failure
locations in timber can be more random.

STRESSES FOR SOLID TIMBER

48 - 6 - 1 Rolling Shear Properties of some European Timber Species with Focus on

Cross Laminated Timber (CLT): Test Configuration and Parameter Study –
T Ehrhart, R Brandner, G Schickhofer, A Frangi
Presented by T Ehrhart
S Franke commented that the test did not consider different laminate thicknesses but
only focused on the increase of board width. He stated that Beech CLT showed
influence of laminate thickness rather than the laminate thickness to width ratio. T
Ehrhart asked whether the ratio was kept constant in the Beech CLT tests so that
thickness effect could be isolated. S Franke said no. G Schickhofer added that in
future standardized procedures for LCT, 20 30 and 40 mm thickness will be available;
therefore, in this study 30 mm thickness was used.
H Blass asked about the low G value. G Schickhofer responded that higher G values
up to 100 MPa can be used.
P Zarnani commented that the rolling shear test results using the two plate
configuration are not pure shear as tension can play a role.
M Flaig commented that the work was well done and agreed with higher values of
~100 MPa for G values. The rolling shear strength of spruce of 1.4 MPa cannot be
confirmed with bending test results. G Schickhofer discussed 3 layer CLT has lower
strength values possibly due to rolling shear volume effect. Also rolling shear
strength should be between 1.2 and 1.5 MPa.
F Lam commented that in CLT bending tests for rolling shear strength, a high speed
camera was used to capture the actual failure mode confirming existing tension
perpendicular to grain stresses causing cracks that preceded rolling shear like failure.
The interaction is complicated and requires more attention. He also commented that
the current study for basic rolling strength is valid.

7
TIMBER JOINTS AND FASTENERS

48 - 7 - 1 A Universal Approach for Withdrawal Properties of Self-Tapping Screws in

Solid Timber and Laminated Timber Products – A Ringhofer, R Brandner, G
Schickhofer
Presented by G Schickhofer
H Blass commented and discussed about the usefulness of Kser (screw’s withdrawal
stiffness) as Kser is very much dependent on test configuration. G Schickhofer agreed
as the test configuration dependency of stiffness is different from that of strength. R
Jockwer commented that there are discussions in EC5 about test configuration. G
Schickhofer agreed that the availability of a standardized test configuration would be
good.
S Winter commented that he did not doubt the accuracy of the equations but they
should be further simplified for practising engineers. G Schickhofer disagreed as
there are so many products but the equation has three main variables for
consideration.
H Stamatopoulos commented that withdrawal stiffness had more factors affecting
the values.
P Zarnani asked about different screw manufacturers. G Schickhofer answered that
different screws should have little effect in terms of the values. T Tannert
commented that some believe in Canada that the design method for self-tapping
wood screws should be extended to all types of screws. He asked whether the model
fits to other data or other types of screws. G Schickhofer answered that this would be
a good idea.
A Salenikovich agreed that diameter mattered and not the product type and that the
difference between self-tapping and non self-tapping screws is small. H Blass stated
that this would depend on density of the wood as some species would need
predrilling.
K Malo supported that withdrawal stiffness would be important for vibration cases.
M Flaig commented that using a probabilistic approach about gap influence would be
valuable. F Lam commented that angle application of screws would lessen the
influence of gaps. G Schickhofer agreed.

48 - 7 - 2 Characteristic Withdrawal Capacity and Stiffness of Threaded Rods – H

Stamatopoulos, K A Malo
Presented by H Stamatopoulos

8
H Blass commented about the stiffness dependence on the embedment and asked
why FEM predictions were higher than experimental values except for the 90 degrees
case. H Stamatopoulos answered that the rod slipped with the interface more at 0
degrees than with 90 degrees. Therefore, with less relative slip at 90 degrees, there
was better agreement.
S Franke asked how was the load slip evaluated in relation to the strength. H
Stamatopoulos answered that small embedment length was used only. FEM
calculation for strength would be more difficult as crack formation could be issues
that needed to be considered; therefore, only used for stiffness prediction. S Franke
stated that he has a student working on strength prediction.
P Zarnani and H Stamatopoulos discussed local shear failure mode such as block
shear failure issues.
W Seim asked about the definition of fracture energy and from where these values
were obtained. H Stamatopoulos clarified that he did not use fracture energy and
just used a bilinear constitutive law. E Serrano commented that you needed fw and
two slopes for the bilinear constitutive law; therefore, you have defined the fracture
energy. H Stamatopoulos agreed.
I Smith and H Stamatopoulos discussed progressive collapse and the use of a long rod
to get steel yielding rather than withdrawal.
R Jockwer asked about pull-pull rather than push-pull test configuration. H
Stamatopoulos responded that there was work done and support conditions played
an important role in withdrawal stiffness.

48 - 7 - 3 Load-carrying Capacity of Dowelled Connections – H J Blass, F Colling

Presented by H Blass
K Malo asked whether the approach is valid for stainless steel. H Blass responded
yes. S Franke and H Blass discussed about the fitting process for screws are more
difficult.
A Salenikovich asked for comments for multiple fasteners in a row. H Blass
responded that EC5 equations were used.
S Franke commented that the attempt was to justify changes to EC5. H Blass
responded that the old allowable values were not based on tests of steel strength,
therefore over-strength situations were not correctly considered. Here the old code
is still non-conservative by ~ 10% but not 25% as previously thought.
V Rajčić and H Blass discussed the lack of conservatism of the old code when different
failure cases were considered.

9
R Jockwer commented the yield strength of the dowels were very important. H Blass
responded that high strength steel dowel compared to mild steel would still be more
beneficial although it would be dependent on cost and economics.
I Smith commented that this is a manifestation of system effect.
U Kuhlmann commented about target failure mode in relationship to the type steel
used.

48 - 7 - 4 Evaluation of the Reliability of Design Approaches for Connections

Perpendicular to the Grain – R Jockwer, R Steiger, A Frangi
Presented by R Jockwer
F Lam commented that 3 P Weibull distribution would provide a better fit for the
distribution tails. R Jockwer agreed.
I Smith suggested use of reinforcement rather than depending on design approaches
that rely on stable crack growth.
H Blass asked about reinforcement between the dowels. R Jockwer responded that
they thought about this but did not use it because of the arrangement of steel plates
and dowels were rather close. H Blass and R Jockwer further discussed the diameter
of the dowel in relationship to the size of the potential reinforcement screws.
P Zarnani and R Jockwer discussed issues related to recording the crack opening.
P Gustafsson asked if this is only valid for symmetric test set up. R Jockwer answered
that they had considered one sided loading and it seemed to be okay. However the
experimental base for unsymmetrical loading is quite weak.

48 - 7 - 5 Simplified Fatigue Design of Typical Timber-Concrete Composite Road

Bridges – K Kudla, U Kuhlmann
Presented by K Kudla
I Smith commented that we do not really know what the damage accumulation rule
should be. Creep rupture rather than cyclic load effect would be more important
with the concept of killer load. He asked about the experimental verification of the
work. K Kudla answered that experimental verification was not done. Miner’s rule
should be on the safe side.
W Seim received clarification that the S-N line results are conservative. U Kuhlmann
commented that this is a typical approach.
P Zarnani asked about the geometry of the concrete slab compared to the beam. K
Kudla answered that it would not make sense to change the geometry as they had
aimed for optimal conditions.

10
K Malo asked if temperature rise was experienced during testing. U Kuhlmann
responded yes, but that it should not be a big issue.

LAMINATED MEMBERS

48 - 12 - 1 Concentrated Load Introduction in CLT Elements Perpendicular to Plane –

T Bogensperger, R A Jöbstl
Presented by T Bogensperger
F Lam received confirmation that the equation involved fr,90mean and the characteristic
strength was incorrect and should be modified. F Lam asked about the softening
procedure used in FEM analysis. T Bogensperger provided some explanation and
agreed to add information in text of paper.
H Blass received clarification for the justification of punching factor k of 1.75 as
possible localized effect where rolling shear failure was assumed not to be brittle. F
Lam asked about the deflection in compression perpendicular to grain. T
Bogensperger stated that the deflection was small because reinforcement screws
were used. These screws did not add to the shear strength.

48 - 12 - 2 Shear Properties of Cross Laminated Timber (CLT) under in-plane Load:

Test Configuration and Experimental Study – R Brandner, P Dietsch, J
Dröscher, M Schulte-Wrede, H Kreuzinger, M Sieder, G Schickhofer, S
Winter
Presented by P Dietsch
F Lam received clarification that the effect of compressive stresses on rolling shear
strength was taken into considered using adjustment factors.
H Blass and P Dietsch discussed the roller bearing details and member under pure
shear such as the case of shear wall. They also discussed the case of CLT as bending
members on edge where standard bending test would be available but difficult to
achieve shear failure of the members. Here torsional shear and rolling shear would
coexist. A Ceccotti stated that in a 3x3 m wall in shear connection design would be
more critical.
I Smith asked how the specimen size was selected. P Dietsch responded that they
were chosen to avoid localized stresses therefore a ratio of 1:3 for width to height
was chosen. G Schickhofer added that the tests are good for wall and diaphragm
cases but beam solutions would need more work.
P Zarnani asked whether mechanistic modelling approach would be used. P Dietsch
answered that the model is already mechanics based.

11
48 - 12 - 3 Advanced Modelling for Design Helping of Heterogeneous CLT Panels in
Bending – L Franzoni, A Lebée, F Lyon, G Foret
Presented by L Franzoni
H Blass asked if the effect of board width was considered. L Franzoni said not yet and
3.5 is the ratio between boards. H Blass asked why were tension and compression
properties established using small clear tests. L Franzoni answered that they tried to
be consistent with modelling approach using clear properties and agreed that the
material in reality has defects.
K Malo asked why there were sharp changes shown in the curves in slide 9 where
with in-plane shear failure one would expected smooth rather than sharp changes. L
Franzoni said that this could be caused by numerical issues
G Fink received clarification that the model was based on small clear properties.
K Malo and L Franzoni further discussed sharp changes were due to change in failure
mode because the model could not determine mixed failure mode and therefore
sharp changes were seen.
G Schickhofer asked whether the end goal should be aimed towards design code. L
Franzoni said that this was not their goal at this moment.

48 ‐ 12 ‐ 4 Performance of Canadian Glulam Columns with New Laminae E

Requirements –F Lam, J K Oh, B J Yeh, J J Lee
Presented by F Lam
T Bogensperger received clarification of the definition of column length being the
total length of the column minus the length of the supporting shoes.
H Blass asked about the orientation of the column in relation to the buckling
direction. Since the columns buckled in the direction parallel to the glue-line of the
laminae, would the bending strength of the column be different if the columns
buckled in the other direction. F Lam responded the “bending” strength is governed
by the compression strength of the laminae when maximum load was reached. The
final failure of tension occurred at a lower load level; therefore, orientation should
not matter.
I Smith asked whether pin-pin conditions were tried. F Lam said that it was
considered but not adopted given that the literature reported friction issues from
pin-ended cases. H Blass commented that in reality pin-ended conditions do not
exist so the code would be conservative if one assumed pin-ended case. F Lam
agreed especially at high load levels.
A Salenikovich received clarification that the laminae were sampled from a mixed
pool with MOE ranged of 11 to 13.1 GPa.

12
48 ‐ 12 ‐ 5 Design of CLT Beams with Large Finger Joints at Different Angles –M Flaig
Presented by M Flaig
G Schickhofer asked about the 0.3 reduction factor as a constant and suggested it
would be better to have this as a function.
S Winter asked why there was a nominal reduction of 14% but found a 30% reduction
for the profile. M Flaig responded that these specimens were from industry and they
did not have control over the production accuracy. Here, optimized production of
CLT as beams was not yet available.
W Seim received clarification that Elocal as shear free MOE. He asked for examples of
where these members could replace glulam. M Flaig responded that CLT beam has
higher strengths in the perpendicular to grain direction therefore, these members
could be suitable for cases where higher stress in perpendicular to grain directions as
well as cases where cracks might be important.
L Franzoni asked given a straight beam why were those dimensions given. M Flaig
said that bending strength of straight beams would depend on the specimen size; so
standardized test span to depth ratio of 18 to 25 are typically used. In the current
tests finger joint governed so the specimen length did not matter.
G Fink commented that based a sample size of 5 to get a characteristic strength
might be too low. M Flaig agreed but stated that it would be too expensive to test
many beams and that they relied on prior knowledge about CLT as beam element
and used this information as guidance. G Fink commented they could consider
different models to estimate characteristic properties.
T Bogensperger was surprised by the large stiffness degradation and asked if the
strength depended on the glue. M Flaig stated that they did not know whether
different glue would affect the strength but did not expect so. He also did not expect
the large stiffness degradation at first but the test results clearly showed this.
T Bogensperger also asked about gluing large finger joint on site. M Flaig stated that
it could be done but would not be recommended.

STRUCTURAL STABILITY

48 ‐ 15 ‐ 1 Performance‐Based Seismic Design of Light‐frame Structures ‐ Proposed

Values for Equivalent Damping –J Hummel, W Seim
Presented by J Hummel
M Popovski received confirmation that light frame and not CLT was shown in slide 7;
also 2.8 mm diameter nails 65 mm in length and 75 mm spacing were used in the
test. M Popovski commented that µ=1.64 from the data seemed too low.

13
D Moroder and J Hummel discussed the use of elastic spectrum and equal
displacement approach. J Hummel also clarified issues related to rocking in the steel
plate and the provision of ball bearing guides.
M Yasumura commented that in Japan a similar approach is available via the BSL
using equivalent single mass. Here, long wall would depend on the joint at the end of
the panel. Also in multi-story cases, different rotations of wall exist in each story. He
asked how one would handle these cases. J Hummel answered that long walls were
not considered yet. This could be considered in a model. Also multi-story cases need
hold down to confine rotation such that shear would govern. M Yasumura
commented that one would need a load displacement relationship and one should
model the structure not just the wall. J Hummel agreed.
A Ceccotti stated that the CLT hysteresis loops would depend on different boundary
conditions and the vertical load and paper reference on the topic is available.
B Dujič asked if this method would be applicable to timber structures. J Hummel
responded that it has the potential to, but they would still need to work on inelastic
spectrum for timber structures.
S Winter commented on editorial issues where details about the test specimens
should be added. He also questioned the test with gypsum fibre boards and what
would be the difference. J Hummel has the gypsum results where higher but similar
load characteristics were observed. Also fatigue failures of staples and nails were
observed.
WS Chang and J Hummel discussed about the flag shape hysteresis loops for CLT and
the real building would behaviour differently because of centring effect. Also shear
walls have relatively high equivalent viscous damping and the issue of building
information from a wall to the whole structure.
F Lam received clarification the PGA used was 1g and it was chosen to drive the walls
into the inelastic range. He commented that the spectrum of time histories looked
strange.
I Smith discussed about missing wood based material with other material.
B Dujič commented about viscous damping issues.
D Moroder and J Hummel discussed the spectrum, time history shifting and scaling
procedures.

48 ‐ 15 ‐ 2 Simplified Wall Bracing Method Using Wood Structural Panel Continuous

Sheathing –T Skaggs, B J Yeh, E Keith
Presented by T Skaggs

14
I Smith received confirmation that there are plan limitations for diaphragm and wall
to match each other.
G Schickhofer asked if it would be possible to introduce CLT with this approach. T
Skaggs responded yes and that it could be possible although the market demand in N
America for CLT in residential construction is low.
A Salenikovich asked if calculation tools would be available. T Skaggs said yes but the
calculations are so simple that the tools would not really be needed.
W Seim and T Skaggs discussed how to deal with symmetrical and asymmetrical cases
where location of the bracing elements would have to be within certain distance of
the centre of the building plan.
F Lam and T Skaggs discussed consideration of CLT for tornado resistance structures
although tornado forces are not considered in N American codes. Here, CLT for safe
room tornado has been considered.

48 ‐ 15 ‐ 3 Structural Characterization of Multi‐storey CLT Buildings Braced with

Cores and Additional Shear Walls –A Polastri, L Pozza, C Loss, I Smith
Presented by A Polastri
D Moroder commented that the proposed interactive effort for design included FEM
model for tall building but no practicing engineers are doing this. A Polastri agreed
but stated that the FEM model is important and needed for tall buildings. D Moroder
asked about higher mode effect. A Polastri said that only regular buildings were
considered so there was no higher mode effect. D Moroder commented that higher
mode was observed even in three story buildings but perhaps not in those made of
CLT.
M Follesa asked about the bonding connection to the foundation. A Polastri
responded that they did not consider the difference of the behaviour of hold down
between wood elements and between wood wall and concrete foundation elements.
P Zarnani asked about the coupling issues between CLT panels at the corner of a
building. A Polastri agreed that this would be important; here, crossed screws would
be used to connect the corner panels to achieve high stiffness.
S Winter commented about the use of FEM for tall buildings. This type of study
would not only be needed for earthquake, but wind issues would also need
consideration. Also non-standardized hold downs should be considered. X Rad
connectors have to be considered beyond strength including issues with fire, sound,
airtightness etc.
A Polastri responded that they are looking at different covers to address these issues.
In traditional steel plate solutions with the steel exposed, poor fire resistance would
also be expected.
15
A Salenikovich received clarification about comparison between balloon and platform
construction.
I Smith commented in Canada, a related study was done with taller buildings and
wind forces tended to govern the design above 8 stories.

48 ‐ 15 ‐ 4 Dissipative Connections for Squat or Scarcely Jointed CLT Buildings ‐

Experimental Tests and Numerical Validation –R Scotta, L Pozza, D Trutalli,
L Marchi, A Ceccotti
Presented by R Scotta
F Sarti asked whether the differential movement between the wall and the connector
was considered. R Scotta said the connector was assumed to be well connected to
the wall and the differential movement was not considered. F Sarti discussed about
vertical load that could help the self-centring in rocking systems. R Scotta responded
that in squat systems, shear deformation is more important and has to be considered
first.
T Tannert asked about the fancy shape of the connector. R Scotta said that this was
made for production efficiency and consideration to use less material.
H Blass asked whether a steel angle type connector would be needed for the wall to
floor connection. R Scotta said yes especially in wall to concrete foundation some
steel plates would be needed to help transfer the high forces.
F Sarti asked about buckling and low cycle fatigue issues. R Scotta responded that
when designing the connector, limiting the maximum strain would be needed to take
low cycle fatigue into consideration. Buckling was observed and modelled, but did
not influence the strength much.
B Dujič and R Scotta discussed that friction needed to be considered.
Z Li commented that the buckling would change the force distribution and one might
need to avoid this failure mode. He received confirmation that mild steel with 275
MPa yield strength was used.
W Seim commented that this component test was a steel to steel connection but in
buildings this would be steel to wood connection. R Scotta agreed and will use model
to take this aspect into consideration.
U Kuhlmann stated that when using mild steel one must consider the over strength in
the steel.
I Smith commented about practicality and tolerance, and whether the potential users
have any opinions. R Scotta stated that the study mainly focused on the structural
performance issues first.

16
FIRE

48 ‐ 16 ‐ 1 Analysis of Fire Resistance Tests on Timber Members in Compression with

Respect to the Reduced Cross‐Section Method –J Schmid, M Klippel, A
Liew, A Just, A Frangi
Presented by J Schmid
H Blass commented that if these values could be trusted would you use the mean
value or 5th percentile values. J Schmid responded that he would use the mean
values and that there would be many other areas where safety could be added.
K Ranasinghe commented that in the UK they have huge difficulties in accepting the
concept that 7 mm is too high. He asked if J Schmid would be insisting to increase 7
mm in the next EC5 revision. J Schmid responded that they are aiming to build a
model to reduce the variability in the data to establish a better value. He stated that
the starting value should be ~14 mm and then one can reduce it later as variability in
the data base is reduced.
S Winter agreed that more investigation would be needed and that the old data base
has limitations. He stated that the problem of zero strength layer concept as the
actual reduction curve is questionable. Comparison with FEM is very difficult as
exact properties with increase in temperature are not available. The use of 7 mm has
worked well for more than 20 years and worked well all over the world. S Winter
commented that he has never heard of a collapse before expected time of fire
resistance in a real fire. We need to oppose making changes to the current rules
now. J Schmid related to the example of weakness of steel connection in fire design
and that the old standard was built on one unreliable database. The main problem is
missing information on influence of temperature on timber properties if we want to
replace fire testing with models.
BJ Yeh commented on the use of 7 mm for CLT and increase of char rate to account
for falling off of the char layer. J Schmid responded that CLT is more complicated
with many producers and issues with loading directions.
FRACTURE MECHANICS

48 ‐ 19 ‐ 1 A Beam Theory Fracture Mechanics Approach for Strength Analysis of

Beams with a Hole –H Danielsson, P J Gustafsson
Presented by H Danielsson
H Blass commented that the stress distribution shown in slide 8 close to the hole is
very different from reality. He questioned why the results are so good. H Danielsson
agreed but stated that the beam theory is exact for normal stresses but might be
different for shear stresses. PJ Gustafsson added that the solution should be exact in

17
terms of energy release rate. As the crack propagated, the normal forces due to the
moment should be exact but shear would not be true.
BJ Yeh asked how close the hole can be to the support and if there were two holes,
how close can they be. H Danielsson responded the distances should be such that
the support would not influence the hole. This would also apply to the multiple holes
cases.
K Malo asked if there would be a limitation to the hole size. H Danielsson said that
there would be no limit and the analysis would also work for notched beams.
I Smith commented that good results need accurate analysis, real structure, and
accurate material properties; when results do not agree, maybe some of these are
not working.
P Dietsch and H Danielsson discussed the cases of round and square holes and
limitations to the model.
R Jockwer received clarifications that at this moment there is no recommendation for
the size of the hole without reinforcement.

48 ‐ 19 ‐ 2 A Strongest Link Model Applied to Fracture Propagating Along Grain –P J

Gustafsson, R Jockwer, E Serrano, R Steiger
Presented by R Jockwer
F Lam and R Jockwer discussed the factor Kknot as being established based on
judgement with LN(2,0.8). F Lam commented that the fitting at the tail of the
distribution was not very good. R Jockwer responded that there could be other
factors such as influence of slope of grain involved.
I Smith commented about the load control testing situation versus displacement
control and received information that this case related to crack arrest. He
commented that how a dynamically developing crack growth could be handled with
Weibull approach where Weibull did not consider stability issues. P Gustafsson
responded that quasi static for dynamic loading was considered here and one could
integrate the energy release rate as the crack propagates.
P Dietsch questioned if one applied the information from 48-7-4 in this situation what
would happen. R Jockwer responded that one would expect little difference if the
height adjustment model was applied as it might fit better than the volume
adjustment model. With strongest link approach better fit could result.
T Tannert questioned the situation where the predefined failure plane in the FEM
was the weak plane but failure could occur in the strong plane and how strongest link
approach could consider the case with failure in the strong plane. R Jockwer
responded that one could lower the fracture energy of a predefined strong crack
plane so that the crack could develop along the strong plane.
18
STRUCTURAL DESIGN CODES

48 ‐ 102 ‐ 1 A proposal for a new Background Document of Chapter 8 of Eurocode 8 –

M Follesa, M Fragiacomo, D Vassallo, M Piazza, R Tomasi, S Rossi, D
Casagrande
Presented by M Follesa
G Schickhofer stated that many structural types were considered here and that some
should be considered only in the national annex such as log houses. M Follesa said
that log houses would be interesting for Italy.
T Toratti commented that the two floor diaphragm cases where the floor went
through the whole building would result in acoustic issues
F Lam commented that the ductility factor for screws connected light wood frame
walls is too high. M Follesa agreed.
D Moroder discussed the information in slide 28 where β to reduce over-strength
also has a β reduction; it seemed double reduction.
A Ceccotti commented that slide 7 stated that the current provisions are not always
met. He asked whether the provisions are on the safe side. M Follesa said no and
commented that for example some Xlam connectors have lower ductility ratio.
G Schickhofer and M Follesa discussed the definition of ductility as some of these are
not consistent.
S Winter stated that the production of background documents is good; he asked
about time schedule and finalization. M Follesa said that the proposal would be
finalized by 2016. S Winter commented that the document should be circulated as
soon as possible even as a draft.

48 ‐ 102 ‐ 2 Aspects of Code Based Design of Timber Structures –J Köhler, G Fink

Presented by G Fink
K Ranasinghe commented that simplicity would be needed by practicing engineers
and asked the group whether we know who the target audience are for the
standards. G Fink responded that the user of the standards is our aim; however, they
needed to look deep into the problem to get the basic understanding to come up
with provisions. H Blass commented that one would have to find the tools to
translate these findings to the engineering world and practicing engineers would
have deadlines and budget constraints and would use simple tools.
I Smith commented that sizing members would have easy ways but structural system
considerations would be more challenging.

19
F Lam commented that the most needed area for reliability based design
consideration would be connection design.
S Franke agreed that one needed a deeper understanding to transfer the information
into simple equations; however, PhD students could be involved in the useful step of
documentation of the steps from the deep understanding to the simple equation. G
Fink responded that PhD students needed to be more focused and work on a topic in
depth rather than looking into too many areas.
G Schickhofer agreed that COV should be considered but how to do this in a standard
would be the question. G Fink agreed that it would not make sense to have different
γm values to consider COV. F Lam added that in Canada, a reliability normalization
procedure is available to consider different COVs in a reliability based approach.
U Kuhlmann suggested only one γm and provide calibration factor for γm and
document clearly where these factors come from.
P Dietsch and G Fink discussed the aim of Eurocode 5 and the target audience. The
fact that practicing engineers not familiar with wood might complain that the code
would be too complicated.

48 ‐ 102 ‐ 3 The Reality of Seismic Engineering in a Modern Timber World –T Smith, D

Moroder, F Sarti, S Pampanin, A H Buchanan
Presented by D Moroder
M Popovski commented that N America is moving away from ductility factor as yield
point of connections is difficult to define. The focus would be on system and the use
of nonlinear deformation level would be the new approaches. D Moroder agreed but
it would be difficult to translate such information to code provisions.
P Dietsch commented that in some cases the steel braces that were designed for
wood are different from what is used on site. He questioned about timber
thicknesses when we design for ductility. D Moroder agreed and stated that it would
be more a problem for the quality of steel in connection and perhaps some testing
should be performed for important and special buildings.
A Ceccotti asked how to define ductility in NZ. D Moroder said that the 1994 RILEM
committee has some information. In NZ, actually no information is available in this
area in standards. They would define ultimate capacity and yield capacity as the
base.
W Seim asked how to deal with low cycle fatigue in NZ code for connection design. D
Moroder said that in N American light frame system this is primarily based on
prescriptive code. In NZ special elements are usually tested to give guidance and
would take into consideration low cyclic fatigue.

20
H Stamatopoulos and D Moroder discussed performance based design principles and
how this would be taken into consideration in NZ.
M Follesa stated that in Eurocode we have serviceability limit state and ultimate limit
state. Capacity based design deals with ultimate limit state but we need to consider
both limit states.
I Smith stated in the last 5 years we have experienced extreme wind and snow load
and commented that it is very tricky to handle the load side and he commented
about proper design of structural system would be needed. D Moroder agreed and
commented that this would be an education issue including builders, engineers to get
global understanding. Educating insurance agencies would also be needed.
F Lam and D Moroder discussed issues related to performance of the connections in
service where perhaps drying cracks could negatively impact their performance and
change their failure mode. Reinforcement can be considered but not currently
available in NZ standards.

48 ‐ 102 ‐ 4 An Execution Standard Initiative for Timber Construction –T Toratti

Presented by T Toratti
U Kuhlmann commented and explained in steel there is an independent executive
standard EN1090 which depends on executive class. The executive class and
consequence classes are critical. U Kuhlmann recommended similar approach to be
considered for timber.
S Winter commented that we need to learn from other material and people doing
execution on site always fear too much regulation even though this is needed. He
stated that national specification document not national standard would be
interesting for industry. S Winter and T Toratti discussed responsibility of different
parties in a building project as this is a grey area.
R Jockwer and T Toratti discussed issues related to tolerances in relationship to
workability and practicality. Design equations have to consider these tolerances
including length of members etc.
M Follesa stated in Italy foundation tolerances might not be compatible with timber
tolerance. There are discussions that tolerance is an issue for all materials. In steel,
basic tolerances are available with the consideration of additional tolerance. If basic
tolerance is not followed, redesign might be needed to check if this is ok.

NOTES
Three notes were presented.

21
ANY OTHER BUSINESS
University of Zagreb Civil Engineering provided an overview of their timber
engineering research program.
Chairman restated the importance to relate the findings to codes or standards and
suggested that the abstract review process need to take this into consideration.
E Serrano suggested that the authors should be provided with feedback from the
review process. Chairman agreed that overall scores and acceptance level would be
made available to the authors. Options to provide more detailed comment would be
available to the reviewers but not mandatory for the reviewers.
Chairman will continue to chair the meeting for two more years and suggest to the
group to find a successor and options for transition should be considered. Secretariat
could still stay with KIT if the new chair would agree.
Discussion relating to the review process and mini-session for WCTE Vienna took
place.

VENUE AND PROGRAMME FOR NEXT MEETING

G Schickhofer invited the group to INTER 2016 in Graz Austria August 16 to 19, 2016.
The WCTE 2016 in Vienna will follow from August 22 to 25, 2016.
Tentative venues for INTER are Shizuoka, Japan 2017, Seattle, USA 2018, Tallinn,
Estonia 2019, Munich, Germany 2020 and Biel, Switzerland 2021.

CLOSE
Chairman thanked the host university and the organizing group for organizing INTER
2015. Chairman thanked F Lam for the minutes and R Görlacher for his work.

22
3 INTER Papers, Šibenik, Croatia 2015
48 - 2 - 1 Proposal of a Eurocode-based Method for the Buckling Design of
Timber Log-walls - C Bedon, M Fragiacomo, C Amadio
48 - 2 - 2 Design of Timber Members Subjected to Axial Compression or
Combined Axial Compression and Bending Based on 2nd Order Theory -
A Frangi, M Theiler, R Steiger
48 - 6 - 1 Rolling Shear Properties of some European Timber Species with Focus
on Cross Laminated Timber (CLT): Test Configuration and Parameter
Study - T Ehrhart, R Brandner, G Schickhofer, A Frangi
48 - 7 - 1 A Universal Approach for Withdrawal Properties of Self-Tapping
Screws in Solid Timber and Laminated Timber Products - A Ringhofer, R
Brandner, G Schickhofer
48 - 7 - 2 Characteristic Withdrawal Capacity and Stiffness of Threaded Rods - H
Stamatopoulos, K A Malo
48 - 7 - 3 Load-carrying Capacity of Dowelled Connections - H J Blass, F Colling
48 - 7 - 4 Evaluation of the Reliability of Design Approaches for Connections
Perpendicular to the Grain - R Jockwer, R Steiger, A Frangi
48 - 7 - 5 Simplified Fatigue Design of Typical Timber-Concrete Composite Road
Bridges - K Kudla, U Kuhlmann
48 - 12 - 1 Concentrated Load Introduction in CLT Elements Perpendicular to
Plane - T Bogensperger, R A Jöbstl
48 - 12 - 2 Shear Properties of Cross Laminated Timber (CLT) under in-plane Load:
Test Configuration and Experimental Study - R Brandner, P Dietsch, J
Dröscher, M Schulte-Wrede, H Kreuzinger, M Sieder, G Schickhofer, S
Winter
48 - 12 - 3 Advanced Modelling for Design Helping of Heterogeneous CLT Panels
in Bending - L Franzoni, A Lebée, F Lyon, G Foret
48 - 12 - 4 Performance of Canadian Glulam Columns with New Laminae E
Requirements - F Lam, Jung Kwon Oh, BJ Yeh, Jun-Jae Lee
48 - 12 - 5 Design of CLT Beams with Large Finger Joints at Different Angles - M
Flaig
48 - 15 - 1 Performance-Based Seismic Design of Light-frame Structures –
Proposed Values for Equivalent Damping - J Hummel, W Seim
48 - 15 - 2 Simplified Wall Bracing Method Using Wood Structural Panel
Continuous Sheathing - T Skaggs, B J Yeh, E Keith
48 - 15 - 3 Structural Characterization of Multi-storey CLT Buildings Braced with
Cores and Additional Shear Walls. - A Polastri, L Pozza, C Loss, I Smith

23
48 - 15 - 4 Dissipative Connections for Squat or Scarcely Jointed CLT Buildings -
Experimental Tests and Numerical Validation - R Scotta, L Pozza, D
Trutalli, L Marchi, A Ceccotti
48 - 16 - 1 Analysis of Fire Resistance Tests on Timber Members in Compression
with Respect to the Reduced Cross-Section Method - J Schmid, M
Klippel, A Liew, A Just, A Frangi
48 - 19 - 1 A Beam Theory Fracture Mechanics Approach for Strength Analysis of
Beams with a Hole - H Danielsson, P J Gustafsson
48 - 19 - 2 A Strongest Link Model Applied to Fracture Propagating Along Grain -
P-J Gustafsson, R Jockwer, E Serrano, R Steiger
48 - 102 - 1 A proposal for a new Background Document of Chapter 8 of Eurocode
8 - M Follesa, M Fragiacomo, D Vassallo, M Piazza, R Tomasi, S Rossi, D
Casagrande
48 - 102 - 2 Aspects of Code Based Design of Timber Structures - J Köhler, G Fink
48 - 102 - 3 The Reality of Seismic Engineering in a Modern Timber World - T
Smith, D Moroder, F Sarti, S Pampanin, A H Buchanan
48 - 102 - 4 An Execution Standard Initiative for Timber Construction - T Toratti

24
25
26
 INTER / 48 - 02 - 01

Proposal of a Eurocode-based method

for the buckling design of timber log-walls

Chiara Bedon, Dept. of Engineering and Architecture, University of Trieste, Italy

Massimo Fragiacomo, Dept. of Architecture, Design and Urban Planning, University of
Sassari, Alghero (SS), Italy
Claudio Amadio, Dept. of Engineering and Architecture, University of Trieste, Italy

Keywords: timber log-walls, buckling, design method, buckling curves

1 Introduction
Blockhaus structural systems are commonly obtained by assembling multiple timber
logs, which are stacked horizontally on the top of one another. Although based on
simple resisting mechanisms, the structural behaviour of Blockhaus systems is rather
complex to predict, and few design recommendations are available in current stand-
ards for timber structures. In this context, the paper focuses on the assessment of
the typical buckling behaviour of vertically compressed timber log-walls. The effects
of various mechanical and geometrical variables such as possible load eccentricities
and/or initial curvatures, openings (e.g. doors or windows), fully flexible or in-plane
rigid inter-storey floors are investigated by means of detailed finite-element (FE) nu-
merical models validated on buckling test results available in literature (Heimeshoff
and Kneidl 1992; Bedon et al. 2015). By taking into account a wide set of geometrical
configurations of practical interest, the effects of the main input parameters on the
observed compressive buckling responses are first highlighted. In order to provide
buckling design recommendations of practical use, normalized design curves derived
from standards (e.g. the buckling design approach provided in the Eurocode 5 for
timber members in compression) are calibrated and validated for log-wall assemblies.

2 Blockhaus structural systems

In current practice (e.g. www.rubnerhaus.com), the traditional Blockhaus log-wall
with height H and length L is obtained by stacking horizontally a series of spruce logs
with strength class C24, according to (EN 338: 2009). These logs typically have cross-
sectional dimensions of height h by width b, with the h/b ratio generally in the range
1.6 to 2.4 but, in some cases, also down to ≈0.8, and are characterized by small pro-

27
 INTER / 48 - 02 - 01

trusions and tongues providing interlocking with the upper and lower logs (Fig.1a).
Several log-wall profiles are available on the market (see for example Figs.1b-1c). In-
dependently of the cross-sectional properties of logs, however, the design concept
and structural assembly of log-wall structural systems is strictly related to interlocking
of multiple timber elements.
In Blockhaus buildings, the structural interaction between the main perpendicular
walls composing the full structural assembly is in fact provided by appropriate corner
joints (Figs.1d-1e). Each log-wall is connected to the RC foundation slab by means of
steel angular brackets. The permanent gravity loads are thus transferred onto each
main wall by the inter-storey floors. Depending on their assembly, these inter-storey
floors can realize an in-plane rigid diaphragm (e.g. by using OSB panels, timber joists
and blocking, with the OSB sheathing properly nailed along the entire perimeter, or
glulam panels arranged on their edges with proper connection between adjacent
panels), hence resulting in a further lateral restraint able to avoid possible out-of
plane deflections of the wall top logs (Fig.2).

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

Fig. 1 Examples of timber log cross-sections: (a) www.rubnerhaus.com;
(b): www.lincolnlogs.com; (c): www.polarlifehaus.com. Typical corner joints: (d) ‘Standard’ and
(e) ‘Tirolerschloss’ (www.rubnerhaus.com)

(a) Fully flexible inter-storey floor (FF) (b) In-plane rigid inter-storey floor (RF)
Fig. 2. Qualitative fundamental buckling shape for log-walls under in-plane compressive loads (Bedon and
Fragiacomo 2015)

As also highlighted in (Bedon and Fragiacomo 2015), since metal connectors are gen-
erally minimized, the typical Blockhaus wall can sustain the vertical loads as far as a

28
 INTER / 48 - 02 - 01

minimum level of contact among the logs is ensured. At the same time, the low mod-
ulus of elasticity (MOE) of timber in the direction perpendicular to the grain makes
the usually slender (high H/b ratio) Blockhaus walls susceptible to flexural buckling.

3 Existing design buckling methods for timber

structural members
3.1 Log-wall assemblies
The current Eurocode 5 for timber structures does not provide formulations and rec-
ommendations for the prediction of the critical load of log-walls under in-plane com-
pressive loads. In order to provide practical design recommendations, an analytical
and FE investigation was proposed in (Bedon and Fragiacomo 2015), where a prelim-
inary assessment and calibration of closed-form formulations derived from classical
buckling theories was presented. Validation of FE models was then carried out
against full-scale buckling experiments discussed in (Bedon et al. 2015). Based on the
performed parametric studies, it was shown that the critical buckling load of a given
log-wall with top rigid lateral restraint (RF) can be calculated as (see also Fig. 3):

(E) π 2 b3 E⊥ π 2 b3 E⊥
= kσ = kσ
( )
N cr , 0
12 L 1 − ν 2
12 L   E 2
 (1)
1 −  ⊥ − 1 
  2G  
 
with kσ= 6.97 for log-walls without openings and kσ= 1.277 for log-walls with a single
door/window opening (with L= Lef the reference length in Eq.(1), in accordance with
Fig.3a). In Eq.(1), E ⊥ and G denote respectively the MOE in the direction perpendicu-
lar to the grain and the average shear modulus of C24 spruce. If double door/window
openings are present, the theoretical critical buckling load is given by:

(E)
π 2 (EI ef )
N cr , 0 = 2 , (2)
H
where (see Fig.3):
H = 0.7 H d (3)
and
b 3 Li
EI ef = E⊥ + 2 E steel I steel (4)
12
represents the equivalent bending stiffness of the ‘composite’ portion of wall with
metal stiffeners (see Fig.3b).

29
 INTER / 48 - 02 - 01

As shown in (Bedon and Fragiacomo 2015) by means of extended comparative inves-

tigations, Eqs.(1) and (2) can predict with close accuracy the Euler’s critical load of
timber log-walls of various geometrical properties, leading – especially for single or
double door/window openings – to conservative estimations of the expected theo-
retical resistance for the same log-walls. A further imperfection factor was also pro-
posed, so that the actual load-carrying capacity of a given log-wall affected by possi-
ble initial geometrical curvatures and eccentricities could be estimated based on the
expected value of the theoretical compressive buckling resistance Ncr,0(E) as:
N cr = χ imp N cr( E, 0) (5)
where
 e tot 
χ imp = 1 −  (6)
 b 
and etot≡ (u0,max + eload), with u0,max ≥ H/400 the minimum initial curvature amplitude.

(a) (b)
Fig. 3. (a) Example of log-wall configuration and (b) detail of the steel hollow stiffeners introduced along
the vertical edges of door and window openings (www.rubnerhaus.com)

3.2 Timber members in compression

In accordance with the Eurocode 5 (point 6.3. “Stability of members”), a more gen-
eral, standardized buckling design method for log-walls under in-plane compression
could take the form of non-dimensional curves, properly calibrated so that the ef-
fects of a multitude of mechanical and geometrical parameters (e.g. material me-
chanical properties, initial curvatures, load and boundary eccentricities, inter-storey
floor typology, etc.) could be conservatively taken into account. Based on the Euro-
code 5, for example, the flexural buckling verification of a timber column is usually
performed by taking into account its minimum relative slenderness ratio:

30
 INTER / 48 - 02 - 01

λ f c ,0,k
λ rel = , (5)
π E 0.05
where
L0 A
λ= = L0 , with A the cross-sectional area, (6)
ρ min I min

f c , 0 , k is the characteristic compressive strength parallel to the grain, and

E0.05 represents the fifth percentile value of the MOE parallel to the grain.
The stability check of the member in pure compression requires that the design com-
pressive stress σc,0,d along the member would satisfy the condition:
σ c ,0,d ≤ k c ⋅ f c ,0,d , (7)
with fc,0,d the design compressive strength in the direction parallel to the gain and kc a
buckling reduction factor given by:
1
kc = , (8)
k + k 2 − λrel
with
• λrel defined by Eq.(5)
• k a coefficient taking into account the effects of initial imperfections, given by:
(
k = 0.5 ⋅ 1 + β c (λ rel − β 0 ) + λ 2rel , ) (9)
• βc an imperfection factor equal to 0.2 or 0.1 for solid timber or glued laminated
timber and LVL, respectively, while
• the coefficient β0= 0.3 in Eq.(9) represents the minimum relative slenderness ra-
tio λ rel necessary to induce column buckling phenomena.
In accordance with Equation (7), in this work a standardized design method consist-
ing of non-dimensional buckling curves is proposed for timber log-walls under in-
plane compression. The novelty of the current proposal is given by the application of
the existing normalized buckling design procedure to the studied structural system,
as well as by the accurate calibration and validation of the method. As a result, the
actual buckling resistance of a generic timber log-wall can be calculated – for each
specific configuration – as a kc fraction of the corresponding Euler’s critical resistance.

31
INTER / 48 - 02 - 01

4 Extended Finite-Element investigations

4.1 General numerical approach
The typical Finite-Element (FE) model used in this investigation consisted of 8-node,
linear brick, solid elements with reduced integration (C3D8R), available in the
ABAQUS element library. In each numerical simulation, a single log-wall laterally re-
strained by two portions of orthogonal log-walls working as outriggers was analysed
(Fig.4). Each timber log was described with a regular b×h cross section. While the
characteristic small protrusions and tongues along the top and bottom surfaces of
logs (Figs. 1(a), 1(b)) were reasonably neglected, the nominal geometry of logs near
their end restraints, as well as near the doors or windows openings, was correctly re-
produced (Fig.4). In accordance with (Bedon and Fragiacomo 2015), moreover, the
presence of the lateral outriggers or fully rigid inter-storey floors was taken into ac-
count in the form of equivalent nodal restraints for the main log nodes (Fig.4) .

(a) FE-model with outriggers (b) FE-model with equivalent boundaries along the vertical
edges
Fig .4. Example of the typical FE-numerical model of a timber log-wall with single door opening
(ABAQUS/Standard)
The mechanical interaction between the logs composing the main tested walls was
reproduced by means of suitable surface contact algorithms. Possible tangential slid-
ing was allowed between the logs (tangential behaviour), with µ= 0.5 being the static
friction coefficient (Bedon and Fragiacomo 2015). The detachment of logs in the di-
rection perpendicular to the contact surfaces was also taken into account (normal
behaviour), so that the influence of partial uplift and overturning of logs on the over-
all bending deformations of the examined log-walls could be investigated. Due to the
presence of geometrical global curvatures and/or load eccentricities, this modelling
feature was a key parameter for the development of extended parametric studies for
the derivation of practical design rules, as also discussed in Section 4.
Concerning the mechanical characterization of timber, C24 spruce was first described
in the form of an equivalent elasto-plastic, isotropic material having density ρ=
420kg/m3, nominal average MOE E= 370MPa, average shear modulus G= 500MPa

32
 INTER / 48 - 02 - 01

and Poisson’s ratio calculated in accordance with Eq.(1). The main effect of this sim-
plified modelling assumption – in conjunction with the equivalent nodal restraints in-
troduced along the lateral edges of each log-wall – resulted in a computationally effi-
cient FE model able to properly describe the full structural response of the examined
log-walls, with out-of-plane stiffness and load bearing capacity predictions in close
agreement with both full-scale and small-scale buckling test results (Bedon and
Fragiacomo 2015). The occurrence of possible compressive damage and local failure
mechanisms in the timber logs (e.g. localized crushing mechanisms along protrusions
and grooves of main logs) was also taken into account by limiting the compression
stresses to the average compressive strength fc,90,m= 3.57MPa of timber in the direc-
tion perpendicular to the grain.
Throughout the full parametric investigations, linear bifurcation analyses (LBA) were
first carried out on each FE model, so that the fundamental buckling shape could be
separately collected for each log-wall configuration. The same FE models were then
analyzed under linearly increasing in-plane compressive loads N, by means of incre-
mental nonlinear (INL) static simulations able to describe the full structural response
of the examined log walls up to collapse. FE investigations were carried out to log-
walls with both in-plane rigid (RF configuration) and fully flexible (FF) inter-storey
floors. In both the cases, additional boundary restraints along the top log of each wall
were properly implemented or removed, respectively (e.g. Fig.4a).

4.2. Parametric study

Throughout the parametric FE-investigations, log-walls characterized by specific ge-
ometrical properties were assembled as described in Section 4.1, by varying the:
a) Total length L, with H= 2.9m the reference nominal height for a single-storey
building. In the parametric investigations, a total length ranging from 3m to 6m
with incremental steps of 0.5m for each log wall configuration was considered;
b) Nominal width b of each timber log, with h= 16cm and 19cm respectively. Based
on the cross-sections of timber log walls currently available on the market
(www.rubnerhaus.com, www.perr-blockhaus.de, www.rusticasa.pt,
www.lincolnlogs.com; www.polarlifehaus.com; www.loghomescotland.co.uk,
etc.), the total width of each log was varied from 8cm to 24cm, with an incre-
mental step of 2cm, for each one of the examined log wall configurations;
c) Number and position of openings (single or double door and window openings);
d) Metal stiffeners for the openings (see Fig.3b): for a given log-wall with
door/window openings, INL simulations were carried out both with and without
metal stiffeners along their vertical edges;
e) Restraint on the top log of the wall: either an in-plane fully rigid (RF) or flexible
(FF) inter-storey floor was considered;

33
INTER / 48 - 02 - 01

f) Initial geometrical curvature (u0,max =H/400 or H/300) and/or load eccentricity

eload (with b/6 ≤ eload ≤ b/3).
A total number of 890 combinations of the variables (a) to (f) were analyzed through
the INL parametric study.

5 Discussion of FE results
The main results of this extended numerical investigation were properly analyzed and
compared, so that practical information related to the structural effects of the main
input parameters could be separately identified.
5.1. Initial curvature
Log-walls of several geometrical properties were first analysed considering an initial
global curvature with different maximum amplitude u0,max, defined as the scaled LBA
fundamental configuration. As shown in (Bedon and Fragiacomo 2015), initial geo-
metrical imperfections typically result – depending on their maximum amplitude – in
moderate decrease of the initial elastic stiffness for the examined log-walls. This
leads to a premature overturning and detachment of few top logs only, with a pro-
gressive decrease of the global log-wall compressive resistance with respect to its
corresponding Euler’s critical load (e.g. Eqs.(1) and (2)).
In this work, additional INL calculations were carried out, and the ultimate compres-
sive load (Nmax)INL attained by each log-wall was separately collected. The so obtained
FE parametric results are displayed in dimensionless format in Figs.5 and 6, for log-
walls with RF or FF restraints, respectively (with u0,max= H/400 and H/300). In these
figures, the numerically predicted buckling coefficient:
(N max )INL (N max )INL
(k c )* = k c = = (10)
N res f c ,90,m ⋅ bL
is plotted versus the log-wall normalized slenderness ratio:
f c ,90, m ⋅ bL λ f c ,90, m
(λ )
*
crit , c = λcrit ,c = = (11)
N (E)
cr ,0 π E⊥
In Eqs.(10) and (11), fc,90,m denotes the mean compressive resistance of C24 spruce in
the direction perpendicular to the grain, while Ncr,0(E) is calculated – depending on the
presence of none/single/double openings – in accordance with Eqs.(1) and (2).
FE data are compared in each plot with additional analytical approximating curves,
calculated in accordance with Eq.(8) and obtained by means of an appropriate nu-
merical calibration of the βc coefficient in Eq.(9), with β0= 0.3. In Fig.5a, FE data are
also separately collected for log-walls without openings or single/double openings,
respectively.

34
 INTER / 48 - 02 - 01

(a) (b)
Fig 5. Effect of initial curvature u0,max on the buckling response of timber log-walls under in-plane
compression (ABAQUS/Standard). Hypothesis of in-plane rigid inter-storey floor (RF)
As shown, the presence of door and window openings typically results in a marked
increase of the normalized slenderness ratio (Eq.(11)) and decrease of the overall
buckling resistance (Eq.(10)), compared to log-walls without openings. However, a
sufficiently regular behaviour was found for all the investigated geometrical configu-
rations, hence suggesting the assumption of the proposed approximating curves for
log-walls of general geometrical properties.

(a) (b)
Fig. 6. Effect of initial curvature u0,max on the buckling response of timber log-walls under in-plane
compression (ABAQUS/Standard). Hypothesis of fully flexible inter-storey floor (FF)

35
INTER / 48 - 02 - 01

5.2. Load eccentricity

All the tested log-walls also showed a high sensitivity to possible load eccentricities
eload (with b/6 ≤ eload ≤ b/3, in the current parametric study), due to the intrinsic de-
sign concept of timber log-wall structural systems. Due to the implicit null tensile re-
sistance of the examined structural system, the application of compressive loads N
with a small eccentricity eload, typically resulted in fact in the premature, partial over-
turning of the top logs (Fig.7), with progressive detachment of the resisting surfaces
in contact ad hence an abrupt decrease of the overall compressive load-carrying ca-
pacity. Comparative examples are proposed in Fig.7 for a selected geometrical con-
figuration (b=12cm; H= 2.9m; L= 4m, RF configuration), by changing the amplitude of
the assigned load eccentricity eload.
Additional comparative FE parametric results are also collected in dimensionless for-
mat (e.g. see Eqs.(10) and (11)) in Fig.8. In general, the FE collected data highlighted
that for all the examined geometrical configurations, an almost exponential relation-
ship can be found between the log-wall slenderness ratio and the corresponding
buckling reduction coefficient kc, for a given load eccentricity, given by:
( λ
k c = A1 ⋅ A2 crit ,c ) (12)

(a) (b)
Fig. 7. Effect of load eccentricity eload on the buckling response of timber log-walls under in-plane
compression (ABAQUS/Standard). Hypothesis of in-plane rigid inter-storey floor (RF). (a)
Compressive load vs. non-dimensional out-of-plane deformations; (b) detail of the qualitative
deformed shape (ABAQUS/Standard)

36
 INTER / 48 - 02 - 01

(a) (b)
Fig. 8. Effect of load eccentricity eload on the compressive buckling resistance of timber log-walls.
Hypothesis of in-plane rigid (RF) or fully flexible (FF) inter-storey floor. FE numerical data
(ABAQUS/Standard) and corresponding approximating curves (Eq.(12))
In Eq.(12), the constants A1 and A2 depend on the sensitivity of each log-wall configu-
ration and top restraint condition (e.g. RF or FF restraint) on the assigned load eccen-
tricity eload, and can approximately be estimated as suggested in Fig.8 (with 0.3 ≤
λcrit,c ≤ 4). The parametric study highlighted that for A2 an almost stable value of 0.4
can be used, independently of the assigned eccentricity amplitude.

In the same Fig.8, it can be seen that the absence of the top lateral restraints (e.g. FF
configuration) markedly affects the actual compressive resistance of the log-walls,
and in the current study typically resulted in overestimations up to ≈30% of the actu-
al compressive resistance of the same log-walls with in-plane rigid inter-storey floors
(e.g. RF restraint). Largest sensitivity and scatter of the collected FE data was found
especially for large eccentricities (Fig.8b). However, a fairly regular structural behav-
iour was generally observed for all the examined configurations, hence suggesting
the application of a standardized buckling design procedure for in-plane compressed
log-walls. In the same Fig.8b, the experimental data obtained from full-scale (RF)
buckling experiments discussed in (Bedon et al. 2015) are also shown, as a further as-
sessment of the calibrated approximating curves.

5.3. Combined initial curvature and load eccentricity

The application of combined initial geometrical curvature u0,max and load eccentricity
eload was also investigated, being this case of relevance for the development of gen-
eral design recommendations.

37
INTER / 48 - 02 - 01

(a) (b)
Fig. 9. Effect of combined initial curvature and load eccentricity eload on the compressive buckling
resistance of timber log-walls. Hypothesis of in-plane rigid inter-storey floor (RF). (a) Compressive
load vs. non-dimensional out-of-plane deformations; (b) FE numerical data (ABAQUS/Standard)
and corresponding fitting curves (Eq.(12))
This last series of parametric analyses typically resulted – compared to Sections 5.1
and 5.2 – in a further amplification of the destabilizing effects deriving from curva-
tures or eccentricities only, hence leading to a further decrease of the log-wall initial
stiffness (see for example Fig.9a), a premature detachment of the top logs and con-
sequently a reduction of the actual load-carrying capacity for the studied log-walls.
For all the examined geometrical configurations, the application of an initial global
curvature of maximum amplitude u0,max=H/400 and a simultaneous load eccentricity
eload generally led to an average ≈5% decrease of the buckling resistances calculated,
for the same set of log-wall configurations, in presence of a load eccentricity eload on-
ly. This phenomenon is displayed in Fig.9b, for example, for log-walls with fully rigid
top lateral restraint (RF), in dimensionless format of INL parametric results (Eqs.(10)
and (11)) and corresponding analytical fitting curves (Eq.(12)).

6 Proposal of a standardized buckling design

method for timber log-walls
In order to implement a standardized buckling design method of practical use, fur-
ther INL parametric investigations were finally carried out on the same log-wall con-
figurations described in Section 4.2 (e.g. 890 combinations of (a) to (f) variables), by
replacing in all the FE models the mean mechanical properties of C24 spruce with the
corresponding design values:

38
 INTER / 48 - 02 - 01

E⊥ G k mod f c ,90,k
E ⊥,d = , Gd = and f c ,90,d = (13a) (13b) (13c)
γM γM γM
with γM= 1.3 the partial safety factor of wood, fc,90,k= 2.5MPa the characteristic value
of compressive strength perpendicular to the grain (EN 338:2009) and kmod= 0.7 the
partial modification factor for moisture and load duration influence (service class 1
and 2, and imposed load of long-term duration). A practical example is proposed in
Fig.10, where the INL results obtained for a same log-wall (L=4.5m, H= 2.95m, b=
8cm, with u0,max= H/400) by taking into account the mean (‘ABAQUS, mean proper-
ties’) or design (‘ABAQUS, design properties’, e.g. Eqs.(13a)-(13c)) mechanical proper-
ties of timber are shown. In both cases, the INL data are normalized by means of
Eqs.(10) and (11), with the corresponding mean or design elastic properties for tim-
ber. For the same reason, the critical load Ncr,0(E) in Eq.(11) is calculated by means of
Eq.(1), with ‘mean’ or ‘design’ elastic moduli respectively. As shown by Fig.10a,
where ‘A’, ‘B’ and ‘C’ denote three different log-wall geometries, the main effect de-
riving from the use of the design elastic moduli and compressive strength for timber
is a shift of the predicted INL data discussed in Section 5 towards lowest slenderness
ratios λcrit,c and higher buckling reduction coefficients kc. The first effect is strictly re-
lated to the used normalization approach, e.g. Eq.(11) and Eqs.(13a)-(13c). The ap-
parent higher resistance of log-walls with design mechanical properties (e.g. the in-
crease of kc) – compared to the same log-walls with mean mechanical properties – al-
so depends on the normalization of the collected FE data (Eq.(10)), e.g. on the de-
crease of the ultimate design compressive resistance ((Nmax)INL)dand the correspond-
ing design compressive resistance (Nres)d. Due to the appropriate calibration of the β0
and βc imperfection factors for the approximating curve (Eq.(8)), however, being
these values representative of the effects deriving from a possible initial curvature
u0,max, the same values 0.3 and 0.25 proposed in Section 5.1 can be used also for de-
sign purposes (see Fig.10b). Similarly, in the case of load eccentricity eload or com-
bined curvature / eccentricity, the constants A1 and A2 collected in Section 5.2 and
5.3 for the approximating curve (Eq.(12)) are proposed for the buckling design of the
examined log-walls.
In conclusion, in accordance with the existing Eurocode 5 approach for timber mem-
bers in compression (e.g. Eq.(7)) and on the discussed parametric results, the buck-
ling design verification of log-walls subjected to in-plane compressive design stresses
per unit length of wall σsd can take the form of:
σ sd ≤ σ b,Rd = kc ⋅ (σ cr( E,0) )d , (14)
with kc the buckling reduction factor, σ b,Rd the design buckling strength and
(σ ) = (N )
(E)
cr , 0 d L = f (E⊥ ,d , Gd ) the Euler’s critical stress calculated as given in
(E)
cr , 0 d

Table 1, depending on the log-wall geometry (e.g. imperfection, inter-storey floor ty-
pology, number of openings). In presence of double openings, specifically, the possi-
ble presence of metal stiffeners along the vertical edges of doors and windows (e.g.

39
 INTER / 48 - 02 - 01

Fig.3) can be taken into account by means of Eq.(4). When steel stiffeners are not
used, the introduction at least of solid hardwood elements along the openings edges
is recommended, in order to ensure a minimum structural continuity and interaction
between the interrupted logs, as well as a certain stiffness against out-of-plane de-
formations. In the same Table, the calibrated imperfection factors for the approxi-
mating curves are also provided. For intermediate values of initial imperfection u0,max
and load eccentricity eload, a linear interpolation between the imperfection factors of
Table 1 can be used to estimate the corresponding load-carrying capacity of a given
in-plane loaded log-wall, hence suggesting the general applicability of the proposed
method.

(a) (b)
Fig. 10. Validation of the buckling design approach for log-walls with initial geometrical curvature.
(a) Qualitative effect of design mechanical properties; (b) parametric FE study with design properties

Table 1. Reference approximating curves of kc, imperfection factors and design critical loads for
timber log-walls under in-plane compression subjected to initial curvature and/or load eccentricity
No openings / Single opening / (Ncr,0(E))d
Double openings
Imperfection / Approximating Fully rigid floor Fully flexible floor No Single Double
Eccentricity curve (RF) (FF) openings opening openings
u0,max= H/400 Eq.(8), βc= 0.25 βc= 0.3
u0,max= H/300 with β 0= 0.3 βc= 0.5 βc= 0.6
eload= b/4 A1= 0.92 A1= 0.66
Eq.(1) Eq.(1) Eq.(2)
eload= b/3 Eq.(12), A1= 0.80 A1= 0.55 with L= Lef
u0,max= H/400 with A2= 0.4
A1= 0.85 A1= 0.6
+ eload= b/4

40
 INTER / 48 - 02 - 01

6. Conclusions
Based on extended FE parametric investigations validated towards past analytical and
experimental data, in the paper the buckling response of timber log-walls under in-
plane compression has been investigated. Careful consideration has been given both
to the influence of geometrical aspects (e.g. log-wall dimensions, timber log cross-
section, presence and position of door / window openings) and to further influencing
parameters, like initial curvatures and / or load eccentricities of variable amplitude.
Approximating curves have been properly calibrated, in order to provide normalized
design curves for all the examined log-walls. Finally, a standardized buckling design
method is proposed. This method is conceptually similar to that used for the design
of timber columns, and could be easily implemented in the ongoing revision of the
Eurocode 5 – Part 1-1.

7. References
Bedon, C, Fragiacomo, M (2015): Numerical and analytical assessment of the buckling
behaviour of Blockhaus log-walls under in-plane compression. Engineering Struc-
tures, 82: 134-150.
Bedon, C, Rinaldin, G, Izzi, M, Fragiacomo, M, Amadio, C (2015): Assessment of the
structural stability of Blockhaus timber log-walls under in-plane compression via
full-scale experiments. Construction and Building Materials, 78: 474-490.
EN 338:2009: Structural timber-strength classes.
Eurocode 5 (2004): Design of timber structures - Part 1-1: General and rules for build-
ings. CEN. (EN 1995-1-1).
Heimeshoff, B, Kneidl, R (1992): Zur Abtragung vertikaler Lasten in Blockwänden –
Experimentelle Untersuchungen. Holz als Roh – und Werkstoff, 50: 173-180.
Springer-Verlag.
Heimeshoff, B, Kneidl, R (1992): Bemessungsverfahren zur Abtragung vertikaler
Lasten in Blockwänded. Holz als Roh – und Werkstoff, 50: 441-448. Springer-
Verlag.
LogHomeScotland (United Kingdom), www.loghomescotland.co.uk
PERR Blockhäuser (Germany), www.perr-blockhaus.de
Polar Life Haus® (Finland), www.polarlifehaus.com
Rubner Haus AG Spa (Italy), www.rubnerhaus.com
Rusticasa Construções Lda (Portugal), www.rusticasa.pt
Simulia (2012). ABAQUS v.6.12 Computer Software, Dassault Systems, Providence, RI,
USA.
The Original Lincoln Logs© (NY, USA), www.lincolnlogs.com

41
INTER / 48 - 02 - 01

Discussion

The paper was presented by C Bedon

H Blass asked about the G value of 500 MPa in Slide 10 and suggested that shear in
buckling would be rolling shear in these walls. C Bedon responded that the chosen
value was based on calibration with test results. Preliminary study showed that using
a lower G value did not agree with data. Also they did not notice a lot of shear de-
formation as the rotation was almost a rigid body motion.
G Schickhofer commented that it would be better to consider a minimum requirement
rather than such a complicated approach for a simple block wall house.
F Lam asked why the cases with design properties were higher than the cases with
mean properties. C Bedon agreed and responded that the process also involved a
normalization process.
A Ceccotti asked if the authors checked how far the designs are away from capacities
in terms of safety. C Bedon answered that cases where stiffness of walls with large
openings can be important and cases where top restraints were not available can also
be important.
U Kuhlmann commented that compressive strength versus critical loading is confus-
ing.

42
43
44
 INTER / 48 - 02 - 02

Design of timber members subjected to

axial compression or combined axial
compression and bending based on 2nd
order theory
Andrea Frangi, ETH Zurich, Institute of Structural Engineering IBK, Zurich, Switzerland
René Steiger, Empa, Materials Science and Technology, Structural Engineering
Research Laboratory, Dübendorf, Switzerland
Matthias Theiler, dsp Ingenieure & Planer AG, Greifensee, Switzerland

Keywords: Timber structures, columns, compression parallel to the grain, stability,

global buckling, P‐delta effect, 2nd order structural analysis

1 Introduction
Axial compression or combined axial compression and bending are encountered in
many types of timber members such as columns, frame structures or compression
members of truss girders. The behaviour of these structural members is primarily
characterised 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 investigated first by Tetmajer (1896). Tetmajer’s studies set up the
basis for the design of timber members subjected to axial compression for a long
time. Tests performed by Larsen and Pedersen (1975) confirmed the results obtained
by Tetmajer. The experimental investigation showed the great influence of varying
material properties on the load‐bearing capacity. In order to account for these
variations and hence, to estimate the resistance of glued laminated timber members
subjected to compression more accurately, Blaß (1987 and 1991) performed Monte
Carlo simulations. The buckling curves given in different design codes (SIA 265:2012,
EN 1995‐1‐1:2004 and DIN 1052:2008) were derived from these investigations.

45
INTER / 48 - 02 - 02

For timber members subjected to combined axial compression and bending,

Buchanan (1984 and 1985) 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 such as Eurocode 5 (EN 1995‐1‐1:2004), the Swiss
national code for the design of timber structures (SIA 265:2012) or the withdrawn
German code (DIN 1052:2008) 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 reduced to that of an
equivalent simply supported (pinned) column. The 2nd order analysis of the structure
is a method which takes into account the non‐linearity by studying the equilibrium of
the deformed structural system. In general, non‐linearity 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 considered.
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 (SIA 265:2012, EN 1995‐1‐1:2004 and DIN
1052:2008) only provide rules for this 2nd order linear elastic analysis of the structure.
In this paper, a clear distinction 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 (Kessel et al. 2005 and 2006;
Möller 2007; Köhler et al. 2008). The discussion in particular showed that there are
inconsistencies concerning the consideration 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 (Kessel
et al. 2005 and 2006; Möller 2007, Becker and Rautenstrauch 2001, Hartnack et al.
2002) this paper 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 differences between 2nd order linear elastic analysis and the
Effective Length Method. 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 compression and bending
(Hartnack 2004, Hartnack and Rautenstrauch 2005, Becker 2002).

46
 INTER / 48 - 02 - 02

2 Design of timber members in compression

parallel to the grain
In general, the Effective Length Method is used for simple design situations (e.g.
verification of stability of single members) while the 2nd order linear elastic analysis of
the structure provides some advantages for more complex design situations (e.g.
impact of stiffness of members and connections on the force distribution in truss and
frame structures, design of bracing).
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) column (Blass, 1995). For the design, the internal forces and
moments are calculated based on a simple 1st order analysis and the non‐linear P‐
delta effect is taken into account by means of a buckling factor kc. This factor
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 .
The buckling factor kc as given in different design codes (EN 1995‐1‐1:2004, SIA
265:2012 and DIN 1052:2008) is based on extensive investigations performed by Blaß
(1987). In order to determine the characteristic value (i.e. 5th percentile) of the load‐
bearing capacity of timber columns Blass performed Monte Carlo simulations. 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 timber 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 (EN 1995‐1‐1:2004, SIA
265:2012 and DIN 1052:2008) recommend using a linear interaction model for
combined axial compression and bending. 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 buckling.
2.2 2nd order linear elastic analysis
As an alternative to the calculation model based on the ELM, timber members
subjected to axial compression or combined axial compression and bending can be
designed by performing a 2nd order linear elastic analysis. The 2nd order linear elastic
analysis is a method which takes into account the geometric non‐linearity by studying
the equilibrium of the deformed structural member. An initial deformation is
introduced into the calculation in order to account for the geometric imperfection of
the member as e.g. deviation from a perfectly straight shape.

47
INTER / 48 - 02 - 02

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 combination with the axial load leads to an initial bending moment
MI. The P‐delta effect causes a magnified moment MII. MII can be calculated by
multiplying the initial bending moment MI with a magnification factor  (Bazant and
Cedolin 1991):
M II    M I (1)
1  2  EI
 and N Euler  (2)
N  2cr
1
N Euler

With
MII: magnified bending moment (2nd order linear elastic theory, deformed structure)
MI: initial bending moment (1st order theory, undeformed structure)
: magnification factor
N: normal force acting on the column
NEuler: Euler buckling load
E: modulus of elasticity (MOE)
I: 2nd moment of inertia
cr: effective length
Timber members subjected to combined axial compression and bending tend to
develop non‐linear deformations of the compression zone before failure occurs. This
non‐linearity leads to a curved shape of the moment – axial force interaction diagram
depending on the ratio between the tensile strength ft,m,0 and the compressive
strength fc,m,0 parallel to the grain (Buchanan 1985, Steiger and Fontana 2005). For
the ultimate limit state analysis of timber columns the design codes (SIA 265:2012,
EN 1995‐1‐1:2004 and DIN 1052:2008) consider this non‐linear interaction behaviour
by squaring the compression part in the interaction model (Eq. (3)). However, the
non‐linear material behaviour also influences the deformations of the structural
system, and as a consequence, also the magnified moment MII is influenced by the
non‐linear material behaviour. However, these effects are neglected when
performing a 2nd order linear elastic analysis.
2
  c , 0, d   m , II ,d
    1.0 (3)
 f  f m,d
 c , 0, d 
With
c,0,d: design value of the acting compressive stress parallel to the grain
fc,0,d: design compressive strength parallel to the grain
m,II,d: design value of the acting bending stress from a 2nd order structural analysis
fm,d: design bending strength

48
 INTER / 48 - 02 - 02

3 Strain‐based model
In order to predict the global buckling behaviour of timber columns, a numerical
strain‐based model has been implemented. Strain‐based models are widely used in
the design of structural members made from other construction materials than
timber. E.g. for reinforced concrete columns, a strain‐based model is suggested in
CEB/FIP Manual (1978). Due to the failure mechanism in timber being influenced by
the distinct nonlinear stress‐strain relationship and leading to a more complex
calculation procedure, up to now, only a few applications of these models to timber
structures are reported in literature (Buchanan 1984, Hörsting 2008).
Figure 1 shows the calculation procedure of the strain‐based model. On the left hand
side, the calculation of the internal force Ni and bending moment Mi is illustrated.
The calculation starts with selecting values for the strain 0 at the mass centre of the
cross‐section and for the curvature y. These two parameters define the strain
distribution within the whole cross‐section, when assuming that plane sections
remain plane. Based on the strain distribution, the stress distribution is calculated
using the relationship given by the stress‐strain curve. Any shape of stress‐strain
curve can be applied in the calculation. Finally, the internal force Ni and moment Mi
are estimated by integrating the stresses over the whole cross‐section. The right
hand side of Figure 1 shows the calculation of the external force Ne and bending
moment Me. The external bending moment Me depends on the external force Ne as
well as on the deformation of the column due to the initial imperfections and the P‐
delta effect. Since the curvature is equal to the 2nd derivation of the deflection curve,
the maximal deflection eII of the column due to the P‐delta effect can be calculated as
a function of the curvature y. Both the internal moment Mi and the external
moment Me depend on the curvature y. Hence, equilibrium between internal and
external forces and moments can be obtained iteratively.

Figure 1. Calculation procedure in the strain‐based model (Theiler et al. 2013)

49
INTER / 48 - 02 - 02

The strain‐based model has been used for studying the influence of various
parameters. E.g. it can be shown that the plastic behaviour of timber when subjected
to compression parallel to the grain considerably influences the buckling behaviour of
columns (Theiler et al. 2013). Therefore, the application of an adequate material
model (stress‐strain relation) is essential when modelling the behaviour of timber
members subjected to compression. In the present study, the model proposed by
Glos (1978) has been used, since it appears to be more suitable than other material
models because it is based on extensive experimental investigations on solid timber
boards. In addition, Glos (1981) developed the model for timber members subjected
to compression and bending while other models are mainly focused on timber
members subjected to pure bending. Glos’s model accounts for the reduction of
stiffness before reaching the ultimate compression strength as well as for the
subsequent softening. Figure 2 qualitatively shows the stress‐strain relationship
proposed by Glos. The description of the full curve in a mathematical form asks for six
parameters (Figure 2 right).

Figure 2. Qualitative representation of the stress‐strain relationship in the material model proposed
by Glos (1978).

4 Experiments
An experimental campaign on glued laminated timber members subjected to
eccentric compression has been performed at ETH Zurich (Theiler and Frangi 2015).
The aim of the experimental investigations was to create a data base, which could be
used to validate theoretical calculation models and to assess the accurateness of the
design approaches given in codes for the design of timber structures. The specimens
were produced using lamellas made of Norway spruce (picea abies) grown in
Switzerland. A total of 336 lamellas were available. In the first step, non‐destructive
tests on the lamellas were performed. These tests aimed at collecting data in order to
characterise the raw material. In the second step, the lamellas were strength graded.
The aim of the grading process was to select two classes of lamellas for the
production of the test specimens. The lamellas were selected so that they were
suitable to produce glued laminated timber of strength classes GL24h and GL32h.

50
 INTER / 48 - 02 - 02

Within the grading process, visual grading criteria as well as machine grading criteria
(dynamic MOE) were used. Specimens for five tests series were produced, three
series of glued laminated timber GL24h and two series of glued laminated timber
GL32h (Table 1). Each of the test series consisted of ten specimens. The length of the
timber members was varied between the different test series: L = 1’400 mm, L =
2’300 mm and L = 3’200 mm. The cross‐section was 140 mm x 160 mm.
Table 1. Overview of test series on glued laminated timber members subjected to eccentric
compression performed at ETH Zurich (Theiler and Frangi 2015)
Test Number Strength Cross‐section Length L Mean value of axial stress
series of tests class [mm] (Slenderness ) at failure (COV)
1 10 GL24h 140 x 160 1'400 mm (30.3) 25.6 N/mm2 (0.07)
2 10 GL24h 140 x 160 3'200 mm (69.3) 15.3 N/mm2 (0.11)
3 10 GL24h 140 x 160 2'300 mm (49.8) 20.3 N/mm2 (0.12)
4 10 GL32h 140 x 160 1'400 mm (30.3) 31.1 N/mm2 (0.09)
5 10 GL32h 140 x 160 3'200 mm (69.3) 18.1 N/mm2 (0.10)

During the glulam production, the setup of the test specimens was recorded. Hence,
the position and the orientation of every lamella within each test specimen were
documented. Finally, the glued laminated timber members were subjected to
buckling tests. The test specimens were loaded with an eccentric (15 mm)
compression force up to failure. During the tests, the applied loads as well as
horizontal and vertical deformations were recorded. For a subsample of 20 test
specimens, additional local deformation measurements were performed using an
optical measurement system.

Figure 3. Measured load‐bearing capacity for all tests performed in comparison with the assumed
lognormal distribution estimated from the test results (Theiler 2014).

51
INTER / 48 - 02 - 02

The graphs in Figure 3 show the results of all tests performed in comparison with the
assumed lognormal distribution estimated from the tests. A good agreement
between test results and estimated lognormal distribution can be seen. The
coefficient of variation (COV) was in the range of 10% for all test series (Table 1). All
details of the tests are presented in a test report (Theiler and Frangi 2015). The
results of the tests have been used for the validation of the strain‐based model.

5 Numerical simulations
In order to account for the variation in material properties, Monte Carlo simulations
were performed. Columns of different slenderness and different strength grades
were modelled with the strain‐based model by assigning them randomly selected
material properties. Six different parameters are needed to describe the full stress‐
strain relationship (Figure 2). The distributions of the properties in terms of
probability density function PDF as well as the correlation between the different
properties have to be taken into account. The study was performed for two different
grades of solid timber (C24 and C30) and glued laminated timber (GL24h and GL32h).
Characteristic values given in EN 338:2009 and EN 14080:2013 were considered.
However, the characteristic values are not sufficient for the stochastic modelling.
Further information on the variability of the mechanical properties is required that
for example can be found in the JCSS Probabilistic Model Code (2007). In addition,
Glos (1978) investigated variability and correlation of the model parameters. Using
these investigations and the characteristic values as a basis, for each material
property a mean value, a standard deviation and a probabilistic density function was
estimated (Table 2).
Table 2. Mean value, standard deviation and probability distribution function PDF used for the
numerical simulations (see figure 2 for the definition of the material properties) (Theiler 2014)
Material property C24 C30 GL24h GL32h
Mean value 11’000 12’000 11’500 14’200
Et,0 , Ec,0 [N/mm2] Standard deviation 2’200 2’400 1’500 1’850
PDF Lognormal Lognormal
Mean value 38.8 48.6 34.0 45.5
ft,m,0 [N/mm2] Standard deviation 9.7 12.2 5.1 6.8
PDF Lognormal Lognormal
Mean value 29.9 32.6 30.4 40.2
fc,m,0 [N/mm2] Standard deviation 5.3 5.6 3.9 5.2
PDF Lognormal Lognormal
Mean value 3.40*10‐3 3.40*10‐3 3.27*10‐3 3.51*10‐3
‐4
εc,0 [‐] Standard deviation 6.80*10 6.80*10‐4 4.25*10 ‐4
4.57*10‐4
PDF Lognormal Lognormal
Mean value 25.4 27.7 25.6 33.9
fc,m,u,0 [N/mm2] Standard deviation 3.8 4.2 2.6 3.4
PDF Lognormal Lognormal

52
 INTER / 48 - 02 - 02

Starting from the stochastically modelled material properties, 2nd order simulations
were carried out with the strain‐based model. For each timber grade and slenderness
ratio 10’000 simulations were performed allowing an accurate estimation of the
mean value and the 5% fractile value of the load‐carrying capacity. Altogether, about
two million simulations were performed in this research project.

Figure 4. Comparison between experimental data and numerical simulations for glued laminated
timber of strength classes GL24 and GL32h (Theiler 2014).
Figure 4 shows the results of the numerical simulations performed for the analysis of
the test results for glued laminated timber of strength classes GL24h and GL32h. It
can be seen that the variation in the numerical prediction is larger for stocky columns
than for slender ones. This can be explained by the variation of the input parameters.
The load‐bearing capacity of stocky columns is governed by the compression strength
parallel to the grain while the load‐bearing capacity of slender columns is governed
by the modulus of elasticity MOE. Therefore, the variation in the numerical prediction
is a direct consequence of the variation of these parameters. For columns of
intermediate slenderness various material properties as well as the initial deflection
influence the load‐bearing capacity.

6 Assessment and improvement of the Eurocode 5

design approach
The Monte Carlo simulations performed allow checking the accurateness of the
design approaches given in Eurocode 5. In Figure 5 the results of the numerical
simulations using as input parameters the values given in Table 2 are compared to
analytical calculations by means of ELM and 2nd order linear elastic analysis for glued

53
INTER / 48 - 02 - 02

laminated timber GL24h and GL32h. Results of the numerical simulations for solid
timber C24 and C30 can be found in Theiler (2014).

Figure 5. Comparison between ELM, 2nd order linear elastic analysis and numerical simulations for
glued laminated timber GL 24h and GL 32h. Design equations taken from Eurocode 5.
It can be seen that the 2nd order linear elastic analysis may lead to an overestimation
of the load‐bearing capacity especially for columns of intermediate and high
slenderness ( > 50). For slender columns ( > 100), the characteristic values
obtained from 2nd order linear elastic analysis are in the range of the 25th percentile
rather than in the range of the 5th percentile. This indicates that the design rules for
the 2nd order linear elastic analysis given in Eurocode 5 do not ensure an accurate
design of timber members subjected to compression and therefore should be
modified. However, this conclusion is only valid for simply supported columns.
In Eurocode 5 and other design codes the 2nd order linear elastic analysis is based on
the assumption of linear elastic material behaviour, which means that the reduction
in stiffness caused by the non‐linear material behaviour is neglected and the load‐
bearing capacity is overestimated. 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 to enlarge the initial deformations or to reduce the design stiffness.
In particular, the reduction of the design stiffness seems to be a practicable solution,
since it describes the physical phenomena accurately. In 1889 Engesser introduced
the design concept of reduction in stiffness for steel columns and suggested to use a
tangent modulus instead of the MOE. Engesser’s theory was first questioned by other
scientists but Shanley showed in 1947, that Engesser’s method was a valuable
possibility to account for non‐linear deformations in the compression zone.

54
 INTER / 48 - 02 - 02

Based on the results of the research project (Theiler 2014) a very good agreement
between 2nd order linear elastic analysis and numerical simulations is obtained when
using a buckling modulus defined as follows:
E0,05
Tk , d  for  c ,0, d f c ,0 d  0.5 (5)
M
T
E0,05    c ,0, d  
Tk ,d   1   2   1  for  c ,0, d f c ,0 d  0.5 (6)
 M   f c ,0, d  

With T  3.0 for solid timber and T  4.0 for glued laminated timber.

Figure 6. Comparison between 2nd order linear elastic analysis based on a buckling moduslus Tk,d and
numerical simulations for glued laminated timber of strength classes GL24h and GL32h.
A similar approach was already proposed by Roš and Brunner (1931) and used in the
previous standard SIA 164:1981 (Dubas 1981). The investigations have shown that
the strength‐dependent reduction of the stiffness (Eq. (5) and (6)) leads to a very
good agreement between the 2nd order linear elastic analysis and the numerical
simulations (Figure 6). On the other hand, the strength‐dependent reduction of the
stiffness for high load levels leads to more laborious design procedure. The buckling
modulus has to be determined by means of iteration and can be different for
different design situation. Additional calculations have shown that for practical design
the estimation of the buckling modulus Tk,d according to Equation 5 is a reasonable
solution even for high load levels (  c,0,d f c,0d  0.5) making the design easier for the
engineers as no iteration is needed.
This study concentrates on the behaviour of simply supported timber columns. For
structural systems such as frame structures the behaviour is different due to the
distribution of the axial load in the single members. Since the axial load influences the

55
INTER / 48 - 02 - 02

reduction of the stiffness due to the plastic deformations, the buckling behaviour
depends on the distribution of the axial load and, as a consequence, the results
obtained with the ELM or the adjusted 2nd order linear elastic analysis (Eq. (5) and
(6)) would be too safe as shown in a reliability assessment performed in a previous
analysis (Köhler et al. 2008).

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 design
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 structure). However, the two design approaches are not consistent
and can lead to different results. Based on the investigations performed, the
following conclusions can be drawn:
 The load‐bearing capacity of stocky columns ( < 20) is governed by the
compression strength parallel to the grain. For slender columns ( > 100) the
modulus of elasticity (MOE) is the dominant material property. For columns of
intermediate slenderness ratio (50< < 100), the compression strength parallel to
the grain, the MOE and the non‐linear material behaviour impact the load‐bearing
capacity.
 When performing a 2nd order linear elastic analysis, 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
design stiffness of the structural member. The use of a buckling modulus for
columns appears to be an appropriate solution.
 When designing single columns or beam‐columns using a 2nd order linear elastic
analysis, the calculation of the design value of the buckling modulus Tk,d should be
based on 5th percentile values of the modulus of elasticity E0,05. The investigations
have shown that the strength‐dependent reduction of the stiffness (Eq. (5) and (6))
leads to a very good agreement between the 2nd order linear elastic analysis and
the numerical simulations. However, the estimation of the buckling modulus Tk,d
according to Equation 5 is a reasonable solution even for high load levels
(  c,0,d f c,0d  0.5) making the design easier for the engineers as no iteration is
needed.
 When designing structural systems using a 2nd order linear elastic analysis, the
calculation of the design value of the buckling modulus Tk,d should be based on
mean values of the modulus of elasticity E0,mean, as the use of 5th percentile values
would lead to too safe results.

56
 INTER / 48 - 02 - 02

8 References
Bazant Z.P., Cedolin L. (1991). Stability of structures: elastic, inelastic, fracture, and damage
theories. Oxford University Press, New York.
Becker P., Rautenstrauch K. (2001). Time‐dependent material behavior applied to timber
columns under combined loading. Part II: Creep‐buckling. Holz als Roh‐ und Werkstoff
59(6):491–5.
Becker P. (2002). Modellierung des zeit‐ und feuchteabhängigen Materialverhaltens zur
Untersuchung des Langzeittragverhaltens von Druckstäben aus Holz. Dissertation.
Bauhaus‐Universität Weimar, Weimar.
Blaß H.J. (1987). Tragfähigkeit von Druckstäben aus Brettschichtholz unter Berücksichtigung
streuender Einflussgrössen. Dissertation, Universität Fridericiana Karlsruhe, Karlsruhe.
Blaß H.J. (1991). Design of columns. In: Proceedings of the 1991 international timber
engineering conference, vol. 1. London: TRADA. p. 1.75–1.81.
Blaß H.J. (1995). Buckling length. In: Timber engineering, STEP 1. Almere, Netherlands:
Centrum Hout.
Buchanan AH. (1984). Strength model and design methods for bending and axial load
interaction in timber members. Dissertation, University of British Columbia, Vancouver.
Buchanan AH., Johns KC., Madsen B. (1985). Column design methods for timber
engineering. Can J Civil Eng 12(4):731–44.
CEB/FIP (1978). Manual of buckling and instability. Comité Euro‐International du Béton and
Fédération Internationale de la Précontrainte; The Construction Press Ltd; Lancaster.
DIN 1052 (2008). Entwurf, Berechnung und Bemessung von Holzbauwerken – Allgemeine
Bemessungsregeln und Bemessungsregeln für den Hochbau. Deutsches Institut für
Normung e.V., Berlin.
Dubas P. (1981). Stabilitätsprobleme. In: Einführung in die Norm SIA 164 (1981) – Holzbau,
Lehrstuhl für Baustatik und Stahlbau, ETH Zurich, Zurich. 119–154.
EN 338 (2009). Bauholz für tragende Zwecke ‐ Festigkeitsklassen. Europäisches Komitee für
Normung, Brüssel.
EN 1995‐1‐1 (2004) + AC:2006 + A1:2008 + A2:2014. Eurocode 5: Bemessung und
Konstruktion von Holzbauten – Teil 1‐1: Allgemeines – Allgemeine Regeln und Regeln für
den Hochbau. Europäisches Komitee für Normung, Brüssel.
EN 14080 (2013). Holzbauwerke ‐ Brettschichtholz und Balkenschichtholz ‐ Anforderungen.
Europäisches Komitee für Normung, Brüssel.
Engesser F. (1889). Über die Knickfestigkeit gerader Stäbe. Zeitschrift des Architekten‐ und
Ingenieur‐Vereins zu Hannover, 35. 455–462.
Glos P. (1978). Zur Bestimmung des Festigkeitsverhaltens von Brettschichtholz bei
Druckbeanspruchung aus Werkstoff‐ und Einwirkungskenngrössen, Dissertation, TU
München. Munich.
Glos P. (1981). Zur Modellierung des Festigkeitsverhaltens von Bauholz bei Druck‐, Zug‐ und
Biegebeanspruchung. Sonderforschungsbereich 96 & Laboratorium für den konstruktiven
Ingenieurbau – TU München, Munich.

57
INTER / 48 - 02 - 02

Hartnack R., Schober K‐U., Rautenstrauch K. (2002). 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.
Hartnack R. (2004). Langzeitverhalten von druckbeanspruchten Bauteilen aus Holz.
Dissertation. Bauhaus‐Universität Weimar, Weimar.
Hartnack R., Rautenstrauch K. (2005). 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.
Hörsting O.P. (2008). Zum Tragverhalten druck‐ und biegebeanspruchter Holzbauteile.
Dissertation; TU Braunschweig; Braunschweig.
Joint Committee on Structural Safety (2007). Probabilistic Model Code; JCSS Joint
Committee on Structural Safety; www.jcss.byg.dtu.dk.
Kessel MH., Schönhoff T., Hörsting P. (2005). Zum Nachweis von druckbeanspruchten
Bauteilen nach DIN 1052:2004–08, Teil 1. Bauen mit Holz 107(12):88–96.
Kessel MH., Schönhoff T., Hörsting P. (2006). Zum Nachweis von druckbeanspruchten
Bauteilen nach DIN 1052:2004–08, Teil 2. Bauen mit Holz 108(1):41–4.
Köhler J., Frangi A., Steiger R. (2008). 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.
Larsen HJ., Pedersen SS. (1975). Tests with centrally loaded timber columns. In: Proceedings
of CIB‐W18 meeting 4: Paper No. 4‐2‐1. Paris.
Möller G. (2007). Zur Traglastermittlung von Druckstäben im Holzbau. Bautechnik
84(5):329–34.
Roš M., Brunner J. (1931). Die Knickfestigkeit der Bauhölzer. In: Kongress des
internationalen Verbandes für Materialprüfung, Zurich.
Shanley F. (1947). Inelastic column theory. Journal of the Aeronautical Sciences, 14(5). 261–
268.
SIA 164 (1981). Holzbau. Schweizerischer Ingenieur‐ und Architektenverein, Zurich.
SIA 265 (2012). Holzbau. Schweizerischer Ingenieur‐ und Architektenverein, Zurich.
Steiger R., Fontana M. (2005). Bending moment and axial force interacting on solid timber
beams. Materials and Structures, 38 (279). 507–513.
Tetmajer L. (1896). Die Gesetze der Knickungs‐ und der zusammengesetzten Druckfestigkeit
der technisch wichtigsten Baustoffe. Materialprüfungs‐Anstalt am Schweiz.
Polytechnikum Zurich. Zurich.
Theiler M., Frangi A., Steiger R. (2013). Strain‐based calculation model for centrically and
eccentrically loaded timber columns. Engineering Structures, 56. 1103–1116.
Theiler M. (2014). Stabilität von axial auf Druck beanspruchten Bauteilen aus Vollholz und
Brettschichtholz. Dissertation No. 22062, ETH Zurich, Zurich.
Theiler M., Frangi A. (2015). Knickversuche mit Brettschichtholzstützen unter exzentrischer
Normalkraftbeanspruchung. IBK‐Bericht No. 361, Institute of Structural Engineering IBK,
ETH Zurich, Zurich.

58
 INTER / 48 - 02 - 02

Discussion

The paper was presented by R Jockwer

S Winter asked why machine grading was not used to select the laminates. R Jockwer
responded that no machine grading was available at the time of the project and visual
grading is more common.
R Harris commented that the text of the paper seemed to suggest used of Emean for
single member design but the slides clarified that Emean is used for system approach.
P Dietsch commented that there did not seem to be a difference between Eurocode 5
and the approach proposed by the paper where γm is already used.
I Smith asked how often the failures occurred at mid height. R Jockwer responded
that the columns quite often failed at mid height. I Smith commented that failure lo-
cations in timber can be more random.

59
60
 INTER / 48 - 06 - 01

Rolling Shear Properties of some

European Timber Species with Focus on
Cross Laminated Timber (CLT):
Test Configuration and Parameter Study

Thomas Ehrhart*); ETH Zurich, Institute of Structural Engineering, www.ibk.ethz.ch,

Stefano-Franscini-Platz 5, 8093 Zurich, Switzerland
Reinhard Brandner; Graz University of Technology, Institute of Timber Engineering
and Wood Technology, www.lignum.at, Inffeldgasse 24, 8010 Graz, Austria
Gerhard Schickhofer; Graz University of Technology, Institute of Timber Engineering
and Wood Technology, www.lignum.at, Inffeldgasse 24, 8010 Graz, Austria
Andrea Frangi; ETH Zurich, Institute of Structural Engineering, www.ibk.ethz.ch,
Stefano-Franscini-Platz 5, 8093 Zurich, Switzerland
*) phone +41 44 633-3178, fax -1093, e-mail [email protected]

Keywords: rolling shear strength & modulus, European timber species, cross lami-
nated timber, CLT, test configuration, parameter study

1 Introduction
Cross laminated timber (CLT) has gained popularity and relevance in the construction
industry during the past decade. Its versatile applicability, economic competitiveness
as well as an increasing social consciousness for sustainable constructions have been
main reasons for this positive development. Its laminar composition enables CLT to
withstand in- and out-of-plane loads. Due to its structure featuring orthogonally ori-
ented adjacent layers, in CLT loaded out-of-plane, shear and more specific rolling
shear has to be considered in ultimate (ULS) as well as serviceability limit state (SLS)
design. This is because rolling shear constitutes a potential failure mechanism and
contributes a noticeable amount to the overall deflection. Comprehensive knowledge
on rolling shear modulus (GR) and strength (fR) is therefore of utmost importance for

61
INTER / 48 - 06 - 01

an adequate design of CLT structures. Previous investigations on rolling shear proper-

ties and their influential parameters have primarily been performed numerically and
using Norway spruce (Picea abies).
The main goal of our contribution, based on investigations detailed in Ehrhart (2014),
was to identify the most important parameters for rolling shear characteristics and to
quantify their influence. Furthermore, information about the rolling shear perfor-
mance of several timber species was analysed to investigate their potential for use in
CLT-products. In view of upcoming new timber species increasingly pushed into the
market, investigations on rolling shear comprised also some hardwood and other
softwood species with a potential to be used for (cross) laminated timber products.
1.1 Available test configurations
Robust and replicable test configurations are essential to assess mechanical proper-
ties of timber. For most timber properties, such configurations are already imple-
mented in international standards. Regarding rolling shear, standardised configura-
tions and methods are not available yet. Studies reflect a variety of diverse testing
methods to determine rolling shear modulus and / or strength by using quite differ-
ent approaches (Figure 1.1).
F Fβ Fα
F

Fβ
F Fα

Figure 1.1 Some test configurations for determining rolling shear properties proposed by Hassel et
al. (2009), Xavier et al. (2009) and Dumail et al. (2000) (left to right)
Neuhaus (1981) determined GR using torsion bars, Dumail et al. (2000) investigated
the applicability of the Iosipescu method for spruce wood, Blass & Görlacher (2001)
used a tension-shear configuration proposed by Gahl (not published). Görlacher
(2002) calculated GR by means of eigenfrequency, Hassel et al. (2009) presented the
single-cube apparatus (SCA) for determination of (rolling-) shear, Xavier et al. (2009)
proposed the Arcan shear test for the measurement of shear properties of clear
wood. Mestek (2011) used and optimised a sandwich configuration – similar to the
one given in EN 408 (2010) for testing shear parallel to the grain – to determine roll-
ing shear properties of Norway spruce. Bendsten (1976) had already used this
method to investigate twelve structural softwood species about 40 years ago.

62
 INTER / 48 - 06 - 01

Beside these configurations, which focus on testing clear wood and board segments,
there are additional configurations in discussion emphasizing the determination of
rolling shear properties of the product CLT. Mestek (2011) adapted the sandwich con-
figuration to allow also testing of CLT-elements. The draft version of prEN 16351
(2011) provides bending configurations with reduced spans from which rolling shear
stiffness and strength can be derived. Gehri (2011) suggested a bending test of a five-
layer CLT-plate with only the outermost layers oriented parallel to the supporting di-
rection. This to enlarge the potential area for shear field measurements in the area
exposed to rolling shear (Figure 1.2).
F τ(z) G(z) F τ(z) G(z)
z z

Figure 1.2 Left: setup suggested by Gehri (2011); Right: standard CLT structure
1.2 Potential influencing parameters on rolling shear properties
Beside others, the timber species itself is one major “parameter” for mechanical
properties. Regarding strength and stiffness at all grain angles, most European hard-
wood timber species show higher characteristics than softwoods, Kollmann (1936).
Bendtsen (1976) reported also significant differences between the rolling shear char-
acteristics found by analysing nine structural softwoods.
Numerous studies have proven significant correlation between density and most me-
chanical properties of timber, at least for clear wood. Görlacher (2002) carried out
bending vibration tests using board segments of Norway spruce and found positive
(but low) correlation between density and rolling shear modulus.
Past investigations outline sawing pattern as one major parameter for the rolling
shear modulus GR. This was concluded in Aicher & Dill-Langer (2001), Jakobs (2005)
and Feichter (2013) by numerical studies and in Görlacher (2002) based on eigenfre-
quency measurements.
Theoretical analysis on the load bearing behaviour of CLT in shear by Kreuzinger &
Scholz (2001) showed that rolling shear strength depends on the ratio of undisturbed
board’s width – which can be either the actual width or the distance between stress
reliefs – and its thickness. To consider this geometric effect, a reduction of strength
by means of the factor kRed is proposed. Experimental tests by Mestek (2011) con-
firmed the dependency and the results agreed well with values predicted using kRed.

63
INTER / 48 - 06 - 01

2 Material and methods

2.1 Material
2.1.1 Timber species
Because of its major importance to the construction industry and data availability
from previous experimental and numerical studies on rolling shear, Norway spruce
(Picea abies (L.) Karst.) was chosen as the reference timber species. This comprises
also all parameter variations, which are described hereafter. For comparison a sec-
ond coniferous wood, pine (Pinus sylvestris L.), was investigated.
Regarding the microstructure of wood, a general distinction between coniferous- and
deciduous wood is given (Figure 2.1). Due to the concentrated clustering of vessels in
the earlywood of ringporous hardwood species with subsequent potential to influ-
ence at least the failure mechanism, species featuring these characteristics have to
be distinguished from diffuse porous species. Birch (Betula pendula Roth), beech
(Fagus sylvatica L.) and poplar (Populus spp.) were investigated as representatives of
the group of diffuse-porous deciduous timber species. Ash (Fraxinus excelsior L.) was
the only ringporous timber species in this study.

Figure 2.1 Tangential cuts (above) and microscopic cross cut (below) of spruce, pine, poplar, beech,
birch and ash (from left to right) (from Grosser & Teez 1985)
2.1.2 Adjustment factors for physical properties in regard to moisture content
According to EN 408, all specimens were conditioned at 20 °C and 65 % relative hu-
midity. Density, rolling shear modulus and strength were related to the reference
moisture content of uref = 12 % by considering 0.5 % (EN 384), 2 % (Neuhaus 1981)
and 3 % (Brandner et al. 2012; Ringhofer et al. 2014), respectively, per percent mois-
ture difference. Density of all specimens was determined to investigate the influence
on rolling shear properties for different timber species, sawing patterns and boards’
geometries.
2.1.3 Geometry of the boards
The cross sectional geometry of CLT-products has not been standardised yet. Dimen-
sions of boards and plates vary depending on the manufacturer, although – at least in
Europe – board or layer thicknesses of 20, 30 and 40 mm are more and more com-
mon. Furthermore, specific production-techniques requiring stress reliefs influence
the final setup. To cover common geometries, boards with a constant thickness (tl) of

64
 INTER / 48 - 06 - 01

30 mm and a varying width wl of 120 (reference), 60 and 180 mm were investigated

(Figure 2.2). Thus, the tested ratios wl / tl were two, four and six. To separate the pa-
rameter “board geometry” from other parameters, like density, annual ring width,
sawing pattern, reaction wood and other timber characteristics, specimens for both
wl / tl ratios were cut subsequently from the boards. Consequently, in both series
comparable densities were achieved (Table 2).
2.1.4 Sawing pattern
Depending on the former location in the log, the radial distance between specimen’s
cross section centre and the pith (r), different grain patterns are generated. We in-
vestigated radial distances r of 30 (± 10) (rift and half-rift grain), 60 and 100 mm (flat
grain) by a realized but negligible eccentricity e of ± 5 mm (Figure 2.2).
r b [mm]
60 120 180
100
r [mm]

60
e 30

Figure 2.2 Investigated sawing patterns and board geometries

2.1.5 Test plan
The test plan is outlined in Table 1. For getting significant statistics on a reference ba-
sis, the so-called “reference sample” comprised 40 specimens of Norway spruce, with
a radial distance r of 60 mm and a width wl of 120 mm. All other samples comprised
20 specimens each and a setting where only one parameter was changed while the
others remained according to the reference sample. In each sample, only one seg-
ment per board was used to assure representative results and to maintain full varia-
bility.
Table 1 Test plan
group no. species r wl / t l
variation of [-] [-] [-] [mm] [-]
reference 1 40 spruce 60 4
species 2.1 20 birch 60 4
2.2 20 ash 60 4
2.3 20 poplar 60 4
2.4 20 beech 60 4
2.5 20 pine 60 4
r 3.1 20 spruce 30 4
3.2 20 spruce 100 4
wl / t l 4.1 20 spruce 60 2
4.2 20 spruce 60 6

65
INTER / 48 - 06 - 01

2.2 Methods
2.2.1 Test configuration
Purity of shear stress, variability of specimen dimensions and necessary effort were
taken into account as most important decision parameters in choosing the method
used in this work (Figure 2.3). It is based on the shear configuration given in EN 408
(2010) and considers the modifications suggested by Mestek (2011). Beech wood
loading plates were used for softwood timber species and poplar. All other hardwood
timber species were tested using steel plates. Additionally conducted numerical finite
element (FE) analysis confirmed the suitability of the configuration and a uniform dis-
tribution of shear stresses in the main field of interest. Stresses perpendicular to the
shear plane were relatively low, but exceeded transverse tensile- respectively com-
pression strength locally in areas very close to the edges (Figure 2.3, middle and
right).
Rolling shear strength was calculated according to EN 408 (2010) using Eq. (1). By
means of displacement transducers, the relative displacement of the loading plates
(x) was measured and GR could be calculated using Eq. (2).
α F
20 kN 20 kN

C T … area of high
T tensile stress perp.
to the grain
wl
C … area of high
compressive stress
perp. to the grain
x T
C
tl tp
τR N/mm2 σ90 N/mm2
-4 -3 -2 -1 0 1 -12 -8 -4 0 4

Figure 2.3 Test configuration, distribution of shear- and normal stresses (left to right)
The angle between shear plane and force direction (α) was 14° for all tests and the
length of the board segment (l) was 100 mm.
ி೘ೌೣ ή௖௢௦ ఈ
݂ோ ൌ (1)
௟ή௪೗
οఛೃ οிήୡ୭ୱ ఈή௧೗
‫ܩ‬ோ ൌ ൌ (2)
οఊೃ ௟ή௪೗ ήο௫

According to EN 408 (2010), GR was calculated between 0.1 Fmax and 0.4 Fmax. Using
vertical advancing rates of 0.4 up to 0.8 mm/min, Fmax was almost always reached
within 300 ± 120 s.

66
 INTER / 48 - 06 - 01

3 Results and discussion

Main results of this study are summarised in Table 2. It contains values for rolling
shear modulus and strength as well as determined density and moisture content for
all samples.
Table 2 Rolling shear properties by groups: main statistics
GR,12 [N/mm2] fR,12 [N/mm2] ρ12 [kg/m3]
u [%]
group n range ത COV [%]
‫ݔ‬ range ത COV [%]
‫ݔ‬ 05 range ത
‫ݔ‬ ത
‫ݔ‬
1 36 52:158 100 27 1.41:2.34 1.88 13 1.51 367:544 439 11.0
2.1 19 138:247 188 19 2.92:4.07 3.45 10 2.91 548:668 612 11.5
2.2 14 299:508 401 14 4.46:6.35 5.57 12 4.54 708:875 798 11.0
2.3 16 93:159 127 15 2.49:3.23 2.88 8 2.51 393:523 463 10.2
2.4 16 258:410 357 12 4.61:6.13 5.37 9 4.64 663:766 720 10.0
2.5 19 101:210 158 19 1.94:2.69 2.29 10 1.95 441:570 521 10.7
3.1 25 89:207 143 19 1.24:2.53 1.88 22 1.30 361:573 442 11.0
3.2 23 30:87 56 26 1.34:2.45 1.84 16 1.40 408:568 480 11.6
4.1 23 43:96 65 21 0.83:1.70 1.16 20 0.82 382:556 459 11.4
4.2 20 100:224 148 21 1.66:2.99 2.28 14 1.80 401:527 459 11.4

3.1 Correlation between density and GR resp. fR

Considering all investigated timber species in the regression and correlation analysis
on the relationship density vs. rolling shear properties (Figure 3.1), high correlation
exists.
7

* spruce R < 0.20 * spruce R < 0.20

400 500

▫ pine R = 0.37 Δ ▫ pine R < 0.20

ΔΔ ΔΔ
▪
6

● poplar R < 0.20 Δ ● poplar R = 0.77 ΔΔ

Δ
Δ ▪▪
Δ ▪ ▪▪ ▪Δ
Δ
▪ beech R = 0.58 Δ Δ
Δ ΔΔ ▪ beech R < 0.20 ▪
▪▪ ▪ ▪▪▪▪
GR,12 [N/mm2]

Δ▪▪▪Δ
fR,12 [N/mm2]
5

Δ ash R = 0.71 Δ ash R = 0.59 Δ

▪ ▪▪▪ Δ Δ Δ ▪ ▪
◦ birch R = 0.65 ▪ ▪ ▪ ◦ birch R = 0.75 ▪ ▪ ΔΔΔ
300

▪ ▪ Δ ◦◦◦
4

◦◦◦ ◦ ◦
◦ ◦▪ ◦ ◦◦
●●●
●●● ● ◦◦ ◦
200

◦ ◦ ◦
3

▫ ◦ ◦ ● ● ◦ ◦
▫▫▫ ▫ ▫ ◦◦◦ ◦ ◦
● ●
● ●● ●▫▫ ▫ ▫
▫ ● ●▫▫▫ ▫ ◦ ◦ * ▫▫▫▫ ▫▫
*▫ ◦◦◦
● ● ●**●●● ●
▫▫▫*●▫▫▫◦ G **** ** *** ▫***** *▫*▫*▫▫*▫ ▫ f
2
100

R,12 = 0.76∙ρ12 - 241 R,12 = 0.01∙ρ12 - 2.38

* ● *●*●●
* *
* *
* * * ● ** * ▫ **** *** ** * **
******* * * * * * R = 0.92 * R = 0.92
1

** *
400 500 600 700 800 900 400 500 600 700 800 900
ρ12 [kg/m3] ρ12 [kg/m3]

Figure 3.1 Density vs. rolling shear modulus (left) and strength (right)
Doing the same analysis for each timber species separately, for softwoods no signifi-
cant correlation between density and the rolling shear properties is observed. A pos-
sible explanation for this might be found in the anatomy of coniferous wood species:

67
INTER / 48 - 06 - 01

for example in Norway spruce, both, early- and latewood have quite constant densi-
ties of about 300 kg/m3 and 900 to 1,000 kg/m3, respectively. Since the thickness of
latewood is relatively constant over the entire life of such a tree, the thickness of the
earlywood zones varies with the yearly growth conditions and thus decisively influ-
ences the average density of a certain piece of wood. Following the circumstance,
that rolling shear failure takes commonly place in the interface zone between early-
and latewood of two subsequent years, the earlywood density may indicate the mag-
nitude of shear properties.
By analysing the relationship between rolling shear properties and the density spe-
cific for each hardwood species high correlations are found (Figure 3.1). Compared to
softwoods, in case of ring porous species the interface between the early- and late-
wood of two subsequent years may again act as the primary failure domain, however,
because of a nearly constant thickness of earlywood over the years of tree’s life, the
average wood density becomes decisively affected by the thickness of the latewood.
Considering diffuse porous species, a more or less homogeneous density profile over
the entire thickness of annual rings can be assumed.
3.2 Correlation between GR and fR
The regression analysis on rolling shear
fR,12 = 0.0125∙GR + 0.723 Δ strength vs. modulus, done by compris-
Δ Δ
Δ
6

R = 0.92 ▪ ΔΔ
ing all investigated timber species, con-
▪ ▪ Δ ▪▪ΔΔ
▪
▪Δ ▪▪ firms the generally observed positive
▪
fR,12 [N/mm2]
5

▪ Δ
▪▪▪ correlation between these two mechan-
▪ Δ Δ▪ Δ
◦ ical properties (Figure 3.2). By analysing
4

◦◦
◦◦ ◦ ◦◦◦ ◦ * spruce R < 0.20 the same relationship separately for
◦ ◦
● ● ◦◦●◦ ●◦
●● ▫ pine R = 0.37
3

● ◦● ● each timber species, high correlation is

● ● ●● ▫ ▫ ● poplar R = 0.28
▫ ▫
* *▫*▫*▫▫▫▫▫▫ ▫ ▫▫ ▪ beech R = 0.61 found for ash, birch and beech, while
2

* * *
*** * ▫ ▫ ▫ ▫
* * *
****** ****** * Δ ash R = 0.79 low and debateable values are observed
* ** * * ◦ birch R = 0.74 for spruce, pine and poplar.
1

100 200 300 400 500

2
GR,12 [N/mm ]

Figure 3.2 Rolling shear modulus vs. strength

3.3 Influence of the sawing pattern on rolling shear properties
This and the next section discuss outcomes from a parameter study, which only com-
prises Norway spruce.
Boards with large distance to the pith generally tend to own lower ring widths and in-
creasing amount of mature wood which both consequence higher densities. There-
fore, given a positive correlation between rolling shear properties and density, higher
strengths would be expected when increasing the radial distance to the pith r. How-
ever, a corresponding positive correlation between r and the rolling shear strength fR
was not observed (Figure 3.3). The mean values of 1.88, 1.88 and 1.84 N/mm2 given

68
 INTER / 48 - 06 - 01

for r = 30, 60 and 100 mm, respectively, are on equal basis. One possible explanation
could be found in different sawing patterns causing different rolling shear modulus.
Assuming that shear failure occurs under a certain distortion, lower GR would cause
lower strengths for outer boards. Thus, these two effects just seem to compensate
each other.

3.0
GR,12 = -1.22∙r + 170 R = 0.07
200

R = 0.79

2.5
GR,12 [N/mm2]

fR,12 [N/mm2]
150

2.0
100

1.5
50

1.0
20 40 60 80 100 120 20 40 60 80 100 120
r [mm] r [mm]

Figure 3.3 Relationship between the radial distance to the pith, indirectly causing changes in the
sawing pattern and the rolling shear modulus (left) and strength (right); for Norway spruce
However and conform to previous studies by Aicher & Dill-Langer (2001), Jakobs
(2005), Görlacher (2002) and Feichter (2013), a distinct relationship between the ra-
dial distance to the pith, indirectly causing changes in the sawing pattern from rift or
half-rift gain to flat grain, and GR is given. Although previous studies report a non-lin-
ear relationship between GR and r, within the investigated range of sawing patterns,
an almost linear reduction of GR for increasing r is observed (Figure 3.3, left).
3.4 Influence of board geometry on rolling shear properties
The geometry of a board shows to have a strong influence on both, rolling shear
strength and modulus.
Compared to the reference width of wl = 120 mm, the average strength and modulus
decrease by 40 % and 30 %, respectively, when wl is reduced to 60 mm. However,
they raise by approximately 20 % and 50 %, respectively, when wl becomes 180 mm
(Figure 3.4).
Distribution of shear stresses is not constant along the segment’s cross section. In ar-
eas close to the edges, tensile stresses perpendicular to grain arise increasingly. The
actual shear stress in these areas is lower than calculated using Eq. (1). However, in
inner zones, actual stresses exceed those calculated. The lower the ratio wl / tl, the
higher the stress peaks and the larger the gap between the actual and calculated
stress (Figure 3.5). This causes lower determined rolling shear strengths for lower
wl / tl ratios.

69
INTER / 48 - 06 - 01

In our study only boards with negligible eccentricity to the pith, e, were used. Conse-
quently, changes in width of the boards also lead to changes in the grain orientation
at the board’s edges, i.e. wider boards show an increasing amount of half-rift and flat
grain oriented annual rings in their peripheral zones. As already discussed before
(3.3), this leads to higher values of rolling shear modulus, which demonstrates the
difficulty in separating the influences caused by the ratio wl / tl from that dedicated to
the sawing pattern.
GR,12 [N/mm2] fR,12 [N/mm2]

250 3.5

200 3.0

2.5
150
2.0
100
1.5
50 1.0

0 0.5

r = 100
r = 100

w/t = 6
w/t = 6

w/t = 2
w/t = 2

w/t = 4
w/t = 4

r = 60
r = 60

r = 30
r = 30

Figure 3.4 Rolling shear properties of Norway spruce: variation of the radial distance to the pith
(indirectly sawing pattern) and wl / tl

τmax,w/t=2
= τmax,w/t=4
τ [N/mm2]

= τcalc,w/t=all

0 60
0 20 40 60 80 100 120 [mm]

Figure 3.5 Qualitative distribution of shear stresses in boards with a w/t-ratio of two and four
3.5 Rolling shear properties of different timber species
The rolling shear properties of all tested timber species are overall very promising
(Figure 3.6).
Strength and modulus determined for Norway spruce confirm and partly exceed val-
ues reported in previous studies. Bendtsen (1976) for example reported fR,mean of
1.79, 1.88 and 1.79 N/mm2 for red, black and white spruce, respectively. Mean rolling
shear modulus found was 68, 73 and 58 N/mm2, however, no information about saw-
ing pattern is provided. Nevertheless, in our comparative study for Norway spruce

70
 INTER / 48 - 06 - 01

the lowest properties are observed. Properties of pine and poplar surpass those of
spruce significantly. Pine particularly shows high rolling shear modulus and poplar a
remarkable strength. Birch performs very well too with strength and modulus about
double as high than for spruce. Beech and ash show outstanding rolling shear proper-
ties and reach values about three times higher than found for spruce (Figure 3.6).
GR,12 [N/mm2] fR,12 [N/mm2]
500
6
400
5

300
4

200 3

100 2

0 1
poplar

pine
birch

ash
spruce

beech

poplar

pine
spruce

ash
birch

beech
Figure 3.6 Rolling shear properties of different timber species: (left) modulus, (right) strength

4 Conclusion
4.1 Test configuration
Experiences made regarding the test configuration are very promising. Rolling shear
failure along one or few annual rings was observed for most species (Figure 4.1),
board-geometries and sawing patterns. Finite element (FE) analysis of the test config-
uration already showed local areas of stresses perpendicular to the grain. Small pri-
mary cracks were indeed observed during several tests in areas close to the edges
(Figure 4.2, left). As they occurred exclusively after removing the displacement trans-
ducers, influence on measured rolling shear modulus can be excluded. Independent
fracture pattern after failure indicates that the influence of small primary cracks is
also negligible in calculating strengths.
In-depth investigations on the interaction between rolling shear and stresses perpen-
dicular to the grain done by Mestek (2011) showed that compression tends to affect
rolling shear strength positively. Following his model and due to the low angle be-
tween force direction and shear plane (α), negligible influence on the rolling shear
behaviour can be assumed.

71
INTER / 48 - 06 - 01

Investigation of the bonding surface after failure indicated that the bonded connec-
tion between the loading plates and specimen was sufficiently strong (Figure 4.2).
However, the surfaces of a few birch and beech specimens were covered with wood
fibres by less than 50 %.

Figure 4.1 Typical failures of specimens: spruce, pine, poplar, birch, beech and ash (from top left to
bottom right)
Efforts and costs for preparation and conducting the tests were relatively low and ge-
ometrical variations of specimen easy to perform. Loading plates out of beech appear
adequate for softwood species and poplar. We therefore propose to record the con-
figuration described for the determination of rolling shear strength and modulus in
EN 408 (2010).

Figure 4.2 Primary crack ① and from that independent failure (left); bonding surface after failure of
a spruce (middle) and birch (right) specimen
In a follow-up of this study, three-point bending tests using Norway spruce boards of
the same population as base material were carried out, Wilding et al. (2014). Shear-
field measurements on 5-layer beams featuring a standard CLT structure, i.e. subse-
quent orthogonal layering, and the structure suggested by Gehri (2011) – with only
the outermost layers oriented parallel to the supporting direction (Figure 1.2) – led to
GR,app,mean of 188 (apparent value because of the orthogonal middle layer) and
GR,mean = 110 N/mm2, respectively. Thus, results of bending tests using the structure
suggested by Gehri agree very well with those obtained from single segment testing.

72
 INTER / 48 - 06 - 01

4.2 Rolling shear properties

Results of in total more than 200 tests confirm previous findings, extend knowledge
on rolling shear behaviour and allow identifying sawing pattern and board geometry
as the main parameters influencing the rolling shear properties GR and fR, respec-
tively.
For Norway spruce, overall GR,mean = 100 N/mm2 and fR,k = 1.4 N/mm2 are found. Cur-
rent values for GR,mean and fR,k, given in European technical approvals of different CLT
producers (e.g. ETA-06/0009, ETA-06/0138 and ETA-10/0241), lie at about 50 N/mm2
and 0.85 : 1.5 N/mm2, respectively. As a minimal wl / tl-ratio of four is defined for
boards in transverse layers in the same ETAs, it appears that the potential of timber is
partly underestimated in the design process and for product characterisation. A cir-
cumstance which also frequently arises in comparisons of global (over the entire
span) and local (only within area of constant moment) bending modulus, gained from
four-point bending tests on CLT plates. Depending on the actual sawing patterns of
the material used for CLT production, especially the values regulated currently for
rolling shear modulus seem to be on a very conservative basis. This also by consider-
ing that boards within transverse layers commonly show varying sawing patterns.
Table 3 Proposed rolling shear characteristics for wl / tl ≥ 4
fR,k [N/mm2] GR,mean [N/mm2]
spruce 1.4 100
pine 1.7 150
poplar 2.2 120
birch 2.7 180
ash, beech 4.0 350

However, decreasing ratios of wl / tl distinctly decrease GR and fR. Comparable effects

are assumed in boards with stress reliefs and comparable relief-distance vs. thickness
ratios. As further differentiation in sawing patterns in CLT-production processes ap-
pears practicable questionable, representative values for all locations in the log – by
taking into account preferable used board geometries and their relationship to radial
positions due to the sawing process – are needed. A proposal for boards with a ratio
wl / tl ≥ 4 is presented in Table 3. Eq. (3) and (4) are proposed to determine proper-
ties of spruce boards with a lower ratio.
௪ ௪
ͲǤʹ ൅ ͲǤ͵ ή ౢ ͵Ͳ ൅ ͳ͹Ǥͷ ή ౢ
݂ோǡ୩ ൌ ݉݅݊ ቊ ௧ౢ (3) ‫ܩ‬ோǡ୫ୣୟ୬ ൌ ݉݅݊ ቊ ௧ౢ (4)
ͳǤͶ ͳͲͲ
In comparison to Norway spruce and the range of investigated timber species, the
outstanding rolling shear properties of beech and ash are outlined which impose
these species to be used in CLT. Pine, poplar and birch have also a great potential as
base material for CLT.

73
INTER / 48 - 06 - 01

Available information on rolling shear properties in international standards has yet

been very limited. Results of this study therefore seem in particularly relevant for
EN 338 (2009), where only values for shear strength and modulus parallel to the grain
are listed. Findings of this study could also contribute to EN 14080 (2013) and
prEN 16351 and have potential to supplement important points to these standards.

5 Acknowledgement
The holz.bau forschungs gmbh and the FFG COMET K-Project ‘focus_sts’ made this
work possible and their support, including that of the funding agencies and company
partners, is thankfully acknowledged. We would also like to express our gratitude to
the entire team of the Institute of Timber Engineering and Wood Technology at Graz
University of Technology for the valuable support during all stages of the project.

6 References
Aicher, S & Dill-Langer, G (2001): Basic Considerations to Rolling Shear Modulus in
Wooden Boards. Otto-Graf-Journal, No. 11.
Bendsten, B A (1976): Rolling Shear Characteristics of Nine Structural Softwoods. For-
est Products Journal, No. 11/1976.
Brandner, R et al. (2012): Determination of Shear Strength of Structural and Glued
Laminated Timber. University of Technology, Graz.
Blass, H J & Görlacher, R (2001): Zum Trag- und Verformungsverhalten von LIGNOT-
REND Decken- und Wandsystemen aus Nadelschnittholz (in German).
Dumail, J F et al. (2000): An Analysis of Rolling Shear of Spruce Wood by the Iosipescu
Method. Holzforschung, No. 4/2000.
Ehrhart, T (2014): Material-Related Influential Parameters on Rolling Shear Behaviour
Regarding Cross Laminated Timber (in German). University of Technology, Graz
(Master Thesis).
ETA-06/0009 (2013): Multilayered timber elements for walls, ceilings, roofs and spe-
cial construction components; Binderholz Brettsperrholz BBS.
ETA-06/0138 (2011): Solid wood slab elements to be used as structural elements in
buildings; KLH-Massivholzplatten.
ETA-10/0241 (2013): Solid wood slab elements to be used as structural elements in
buildings; Leno Brettsperrholz.
EN 338 (2009): Structural timber – Strength classes. European Committee for Stand-
ardization (CEN).
EN 384 (2010) Structural timber – Determination of characteristic values of mechani-
cal properties and density. European Committee for Standardization (CEN).
EN 408 (2010): Timber structures – Structural timber and glued laminated timber -
Determination of some physical and mechanical properties. European Committee
for Standardization (CEN).

74
 INTER / 48 - 06 - 01

EN 14080 (2013): Timber structures – Glued laminated timber and glued solid timber
– Requirements; European Committee for Standardization (CEN).
EN 13183-1 (2002): Moisture content of a piece of sawn timber – Part 1: Determina-
tion by oven dry method. European Committee for Standardization (CEN).
Feichter, I (2013): Stress – and ultimate load calculations for selected problems in
solid timber constructions with cross laminated timber (CLT) (in German). Univer-
sity of Technology, Graz (Master Thesis).
Gehri, E (2011): Personal Correspondence on Rolling Shear Test Configurations and
Properties.
Görlacher, R (2002): Ein Verfahren zur Ermittlung des Rollschubmoduls von Holz (in
German). Holz als Roh und Werkstoff, No. 60.
Grosser, D & Teez, W (1985): Einheimische Nutzhölzer (in German). Infor-
mationsdienst Holz.
Hassel, B P et al. (2009): The Single Cube Apparatus for Shear Testing – Full-Field
Strain and Finite Element Analysis of Wood in Transverse Shear. Composites Sci-
ence and Technology, No. 69/2009.
Jakobs, A (2005): Zur Berechnung von Brettlagenholz mit starrem und nachgiebigem
Verbund unter plattenartiger Belastung unter besonderer Berücksichtigung des
Rollschubes und der Drillweichheit (in German). Universität der Bundeswehr, Mu-
nich (Doctoral Thesis).
Kollmann, F (1936): Technologie des Holzes und er Holzwerkstoffe (in German). Julius
Springer Verlag, Berin.
Kreuzinger, H & Scholz, S (2001): Schubtragverhalten von Brettsperrholz (in German).
Forschungsvorhaben-Schlussbericht. University of Technology, Munich.
Mestek, P (2011): Punktgestützte Flächentragwerke aus Brettsperrholz – Schubbe-
messung unter Berücksichtigung von Schubverstärkungen (in German). Technical
University, Munich (Doctoral Thesis).
Neuhaus, F H (1981): Elastizitätszahlen von Fichtenholz in Abhängigkeit der Holz-
feuchtigkeit (in German). Ruhr Universität, Bochum (Doctoral Thesis).
prEN 16351 (2012): Timber structures – Cross laminated timber – Requirements. Eu-
ropean Committee for Standardization (CEN). Draft version.
Ringhofer, A et al. (2014): The influence of moisture content variation on the with-
drawal capacity of self-tapping screws. Holztechnologie 55(3):33–40.
Wilding, B et al. (2014): Materialbezogene Einflussparameter auf die Rollschubeigen-
schaften in Hinblick auf Brettsperrholz. Präsentation im Rahmen der Beiratssitzung
des COMET K-Projektes „focus_sts“ (in German). holz.bau forschungs gmbh, Graz.
Xavier, J et al. (2009): Measurement of the Shear Properties of Clear Wood by the Ar-
can Shear Test. Holzforschung, No. 63/2009.

75
INTER / 48 - 06 - 01

Discussion

The paper was presented by T Ehrhart

S Franke commented that the test did not consider different laminate thicknesses but
only focused on the increase of board width. He stated that Beech CLT showed influ-
ence of laminate thickness rather than the laminate thickness to width ratio. T
Ehrhart asked whether the ratio was kept constant in the Beech CLT tests so that
thickness effect could be isolated. S Franke said no. G Schickhofer added that in fu-
ture standardized procedures for LCT, 20 30 and 40 mm thickness will be available;
therefore, in this study 30 mm thickness was used.
H Blass asked about the low G value. G Schickhofer responded that higher G values
up to 100 MPa can be used.
P Zarnani commented that the rolling shear test results using the two plate configura-
tion are not pure shear as tension can play a role.
M Flaig commented that the work was well done and agreed with higher values of
~100 MPa for G values. The rolling shear strength of spruce of 1.4 MPa cannot be
confirmed with bending test results. G Schickhofer discussed 3 layer CLT has lower
strength values possibly due to rolling shear volume effect. Also rolling shear strength
should be between 1.2 and 1.5 MPa.
F Lam commented that in CLT bending tests for rolling shear strength, a high speed
camera was used to capture the actual failure mode confirming existing tension per-
pendicular to grain stresses causing cracks that preceded rolling shear like failure.
The interaction is complicated and requires more attention. He also commented that
the current study for basic rolling strength is valid.

76
77
78
 INTER / 48 - 07 - 01

A Universal Approach for Withdrawal

Properties of Self‐Tapping Screws in Solid
Timber and Laminated Timber Products
Andreas Ringhofer*), Graz University of Technology, Institute of Timber Engineering
and Wood Technology, www.lignum.at, Inffeldgasse 24, 8010 Graz, Austria
Reinhard Brandner, Graz University of Technology, Institute of Timber Engineering
and Wood Technology, www.lignum.at, Inffeldgasse 24, 8010 Graz, Austria
Gerhard Schickhofer, Graz University of Technology, Institute of Timber Engineering
and Wood Technology, www.lignum.at, Inffeldgasse 24, 8010 Graz, Austria
*)
phone +43 316 873 4614, e‐mail [email protected]

Keywords: axially loaded self‐tapping screws; withdrawal strength; stiffness; solid

timber; glued laminated timber; cross laminated timber; universal approach

1 Introduction
Since 25 years modern self‐tapping screws are frequently applied in timber engi‐
neered structures. Main reasons for their success are a flexible geometry enabling
simple and economic installation without pre‐drilling as well as a high load‐bearing
potential in terms of resistance and stiffness if stressed in axial direction. Conse‐
quently, they represent an efficient alternative to conventional laterally loaded fas‐
teners such as nails, dowels and bolts.
In general, self‐tapping screws are predominately used in solid timber (ST) and lami‐
nated timber products such as glued laminated timber (GLT, layers unidirectionally
oriented) and cross laminated timber (CLT, layers orthogonally oriented). Their appli‐
cation can be classified as (a) fastener, i.e. transmitting loads in connections between
elements, and (b) reinforcement of timber members, i.e. persisting exceedance of in‐
ternal resistances in timber’s weak directions, e. g. stresses perpendicular to grain or
shear. In the design of primary axially loaded screws following failure scenarios have
to be distinguished: (i) steel failure in tension and compression, (ii) pull‐through fail‐
ure of the screw head, (iii) withdrawal failure of the threaded part of the screw and
(iv) block shear failure of a group of screws. Concentrating on single screw perfor‐
mance, we focus further on (iii) the withdrawal behaviour of self‐tapping screws as
design criteria influenced by the composite action between screw and timber and
hence defined by timber properties and geometrical conditions.

79
INTER / 48 - 07 - 01

Withdrawal properties were analysed so far e. g. in Blass et al. (2006),

Frese and Blass (2009), Pirnbacher et al. (2009) or Hübner (2013). However, their
main focus was rather on withdrawal strength fax than on stiffness, whereby investi‐
gations comprised primary solid timber and derived models constitute empirical re‐
gression functions considering a large number of influencing parameters. Even
though their high predictive quality, these models are in general only applicable for
the timber product tested and within the overall scope of their investigations. Trans‐
ferring them to other timber products may cause unreliable results. Especially the
complex lay‐up of CLT as given in Fig. 1, including different layer orientations and the
possibility of gaps, requires a comprehensive treatment and thus additional model
parameters.
α
d side face

wgap,0° wgap,90°
side face

α
lef

narrow face

narrow face

Figure 1. Essential influencing parameters of axially loaded screwed connections in CLT.

Uibel and Blass (2007) were the first who developed a prediction model for determin‐
ing the withdrawal capacity Rax of axially loaded screws situated in the side and nar‐
row faces of CLT elements. Their function covers implicitly a reduction in strength
caused by the possibility of screw insertion in gaps (gap width wgap; average, unsys‐
tematic investigated range 0.5 to 2.0 mm) and intermediate zones of two layers with
different axis‐to‐grain angles α. Although again a high agreement between model
predictions and test results is given, two aspects concerning their approach are worth
being discussed: first, α was limited to αi = {0 °; 90 ° and 0|90 ° for intermediate
zones} restricting the scope in applying the model to some extent. Second and as
consequence of the aforementioned unsystematic variation in gap width, a direct re‐
lationship between fax and the type, number and width of gaps cannot be deter‐
mined.
With regard to Kser herein defined as the screw’s withdrawal stiffness, only the pre‐
diction model published by Blass et al. (2006) and a further approach given e. g. in
ETA‐12/0062 (2012) have been found so far. In fact, both models differ regarding
type and manipulation of input parameters. Based on them significantly different
predictions for Kser are observed. Furthermore, the aforementioned characteristics of
laminated timber products with possible influence on Kser are not considered.

80
 INTER / 48 - 07 - 01

Motivated by derived open questions and restricted parameter variation in previous

studies, several additional investigations on the withdrawal behaviour of axially load‐
ed self‐tapping screws in laminated timber products within the last years, e.g. by
Reichelt (2012), Bratulić (2012), Grabner (2013), Ringhofer et al. (2013,2014a,2015),
Silva et al. (2014) and Plüss (2014) were made. These studies aimed additionally to
explicitly determine and describe the specific impact of parameters on withdrawal
properties in form of a systematic variation of parameter characteristics. Amongst
others, the main parameters were the number of penetrated layers N and their ori‐
entation, the timber member’s moisture content u, the axis‐to‐grain angle α as well
as the number (ngap), width and type of gaps (c. f. Fig. 5) in CLT side and narrow faces.
Based on copious findings related, we propose determining withdrawal properties
X = {fax; kser} of self‐tapping screws placed in layered timber products (and solid tim‐
ber) with a new and universal approach conceptually presented in Ringhofer et al.
(2014b), see Eq. (1) and (2).

∙ ∙ ∙ , with (1)

1,00 ,
, , , , , , , (2)

with ρref and Xref as reference values of density and the specific withdrawal property,
the latter determined in one timber layer (N = 1, ST, with ρ = ρref, α = 90 ° and
u = 12 %), kρ as power factor including the density influence on X, kax as function con‐
sidering different α and gap insertion (kgap) and ksys as system value covering the ef‐
fect of penetrated layers N > 1. Due to the withdrawal property’s adaptability for the
specific situation, Eq. (1) can be universally applied for axially loaded self‐tapping
screws, irrespective the timber product used and the position the screw is inserted.
In the frame of this paper, we show and discuss the background and derivation of
these k‐values (or k‐functions). Consequently, we verify them in form of Eq. (1) with
results from selected test series usually not applied for calibration of model parame‐
ters. Beside its general suitability for modelling of connections and reinforcements as
well as for verification of experimental results, we apply Eq. (1) also for determining
an equation for the characteristic withdrawal strength with the potential to be im‐
plemented in design standards.

2 Materials and Methods

2.1 Materials
Tab. 1 shows a brief overview of test series used for calibrating model parameters
and verifying of the approach in Eq. (1). The whole database, classified into the varied
timber products (all in Norway spruce, Picea abies) and parameters, comprises about
8,000 datasets (82 single test series) for withdrawal strength fax and about 5,500 da‐
tasets (107 single test series) for withdrawal stiffness kser. In addition, Tab. 2 shows
the investigated parameters and their ranges.

81
INTER / 48 - 07 - 01

Table 1. Overview of investigated timber products and parameters.

Material Use invest. Parameters rounded Number of Tests n
(before outlier treatment)
strength fax stiffness kser
Solid timber (ST) modelling d, lef, α, ρ, u, lemb 5,200 3,400
GLT verification d, lef, α, ρ, u, N 1,700 1,300
CLT side face verification d, lef, α, ρ, u, N, wgap, ngap 800 400
CLT narrow face verification d, α, ρ, wgap 500 500
d … outer thread diameter of the screw, lef … length of inserted threaded part of the screw,
lemb … embedment depth of the inserted threaded part of the screw
Further parameters possibly influencing withdrawal properties and not listed in
Tab. 1, such as pre‐drilling, different screw test configurations as well as deviating
geometries of the fully and partially threaded screws applied, have also been varied
within the experimental programme. Investigations done in Pirnbacher and Schickho‐
fer (2007), Pirnbacher et al. (2009), Frese and Blaß (2009) and Ringhofer and Schick‐
hofer (2014) outlined that there is no remarkable influence of these parameters on
fax. Thus, in the analysis these data sets were combined. In case of withdrawal stiff‐
ness, results gained by Reichelt (2012) indicate also a negligible effect of pre‐drilling
on kser, while a possible influence caused by applying different test configurations has
not been investigated so far. Consequently, only tests carried out with a standard
push‐pull configuration, as shown in Fig. 2, were considered for modelling this prop‐
erty.
Table 2. Investigated parameters (see Tab. 1) and their range.
Param. Range
strength fax stiffness kser
d [mm] 4, 6, 8, 10, 12 6, 8, 10, 12
lef [mm] 2.5d ÷ 15d 2.5d ÷ 39d
α [°] 0, 12.5, 25, 30, 37.5, 45, 60, 72.5, 90, 45/45, 0/90 0, 30, 45, 60, 90, 45/45, 0/90
ρ12* [kg/m³] 310 ÷ 621 310 ÷ 621
u [%] 8.20 ÷ 20.0 8.20 ÷ 20.0
N [‐] 1 ÷ 20 1 ÷ 20
wgap [mm] 0 | 2 | 4 | 6 0|2|4|6
ngap [‐] 0|1|2|3 0|1|2|3
* density at u = 12 %
2.2 Methods
All experiments summarised in Tab. 1 were performed on two test rigs Dyna Z‐25FS
(concrete adhesion tester) and LIGNUM‐UNI‐275 (universal testing device, Zwick
GmbH & Co. KG) in accordance to EN 1382 (1999). From each specimen and after
testing (about) 4d x 4d x lef clear wood samples were cut centrically around the screw
hole determining density and moisture content. After this specimen were split centri‐
cally to evaluate possible influences on the withdrawal properties caused by knots or
other growth characteristics.

82
 INTER / 48 - 07 - 01

Deviating from EN 26891 (1991), withdrawal stiffness Kser, as inclination coefficient

shown in Fig. 2 (middle), has been determined by a simple regression analysis consid‐
ering load and displacement recorded in the linear elastic part of the test curve. Fol‐
lowing Brandner et al. (2015) and given in Fig. 2 (right), we localised this area by plot‐
ting the load increments Fi+1 – Fi against their related time steps i of our way‐
controlled loading protocol. Corresponding local displacement measurement has
been determined according to Eq. (3):
∙ , (3)
with δ1 and δ2 as displacements recorded by two LVDTs clamped on the screw shank,
lsh as distance between their clamping points and the timber surface, E as the screw’s
elastic modulus and σax as axial stress in the screw’s shank.
Fax
linear‐elastic zone
Fax/2 δ1 δ2 F /2
ax
30 Fmax 100

Fi+1 ‐ Fi [N]
75
Fax [kN]

20
lsh

50
10 Kser 25
lef

0
0 1 2 3 4 0 100 200 300 400
w [mm] time step i [‐]

Figure 2. (left) push‐pull test configuration including local displacment measurement;

(middle) typical test curve of an axially loaded screw, including Kser;
(right) load increment vs. time step in accordance to Brandner et al. (2015).
Similar to withdrawal strength fax = Fax / (U∙lef) = Fax / (d∙π∙lef), with U as outer thread
perimeter we determine kser as withdrawal stiffness per mm displacement and the ef‐
fective lateral area, see Eq. (4):
/ ³. (4)
∙ ∙

Thereby and deviating from the regulations given in EN 1995‐1‐1 (2008), lef is defined
as inserted threaded part of the screw without its tip, c. f. Fig. 2, i.e. only the effective
anchoring length is considered. Withdrawal properties gained from tests with screw
tips situated in timber specimen have thus been determined with lef = lp – 1.17d ac‐
cording to Pirnbacher et al. (2009) with lp as the total inserted threaded part of the
screw. As introduced in Section 1, our reference withdrawal property Xref is referred
to the reference moisture content u = 12 %. Consequently, test data Xu with u differ‐
ing from 12 % has been corrected by an approach similar to that proposed in Ring‐
hofer et al. (2014a), see Eq. (5):
1.00 8% 12 %
, for , with (5)
1.00 ∙ 12 12 %

83
INTER / 48 - 07 - 01

kmc = {0.034; 0.016} for {fax; kser} as inclination coefficient describing the decrease of
withdrawal properties per increasing u. Pirnbacher et al. (2009) observed a positive
effect of screw thread embedment on withdrawal strength, i.e. an increase of fax at
α = 90 ° in cases lemb exceeds 2d. Since further investigations carried out by Burg‐
schwaiger (2010) indicate a minor benefit for screws when inserted parallel to grain,
we assume that this effect decreases by decreasing α. Results from tests with em‐
bedded screw threads (all 2d) were thus corrected according to Eq. (6):
, 1.00 0
, for . (6)
, 1.05 1.11 ∙ 10 ∙ 2
As investigations on the effect of embedment depth comprised only the withdrawal
strength, for kser a quantitative approach is missing. Consequently, related datasets of
kser were left without correction and used only for relative comparisons in series with
the same embedment configuration. Final outlier treatment was done in two steps:
after excluding tests on screws penetrating or touching knots, we performed Tukey’s
criteria for statistical outliers (values outside the inter‐quartil‐range (IQR) ± 1.5‐times
the IQR) on logarithmised data sets and by means of box‐plots.

3 Discussion of model components

3.1 General notes
Data analysis in Section 3 bases on the general hypothesis of lognormal‐distributed
(2pLND) densities and withdrawal properties. In fact, there are two reasons for this
assumption: (i) 2pLND constraints only positive data values as they are common for
physical and mechanical properties such as strength and stiffness (c. f. Brand‐
ner, 2012), (ii) standard EN 14358 (2006) as well as recently published investigations
on screw withdrawal properties (c. f. Pirnbacher et al., 2009; Frese and Blaß, 2009
and Hübner, 2013) also assume 2pLND for data assessment. Consequently, statistical
test procedures and characteristics such as the correlation coefficient according to
Pearson, t‐Test or confidence intervals of selected statistics were carried out (or de‐
termined) for logarithmised data and if required transformed to the linear domain
(c. f. Olsson, 2005).
3.2 System factor ksys considering the number of penetrated layers N
Currently, all mentioned regression models determining withdrawal strength consid‐
er the density as only material property indicating the fastener’s anchoring capacity
in timber members. Due to more homogeneous properties and higher 5 % values of
density ρk in GLT compared to structural timber (ST, N = 1), according to EN 1995‐1‐
1 (2008) and for screws penetrating more than one layer (N > 1), higher 5 %‐
characteristic withdrawal strengths fax,k can be applied. Increasing homogenisation
with increasing N is expressed by a reduced variability in density while expectation
E[ρ] remains constant (c.f. Brandner, 2012; and Ringhofer et al.,2015).

84
 INTER / 48 - 07 - 01

Focusing on this topic, Reichelt (2012) carried out several test series in GLT and CLT
specimen with identical lay‐ups, varying N but constant E[ρ]. Therein, she not only
observed the mentioned increase of fax,0.05 but also a similar behaviour for E[fax] at
N = {3; 6; 20}. Since current regression models are not able to cover this effect, also
proved later by Bratulić (2012), we apply a stochastic approach derived in Ring‐
hofer et al. (2015) considering the increase of E[fax] with increasing N. Corresponding
values ksys,mean and ksys,k (for 5 % quantiles) in dependence of N are given in Tab. 3.
Table 3. Values for ksys,mean and ksys,k in dependence of N; according to Ringhofer et al. (2015).
N 1 2 3 4 5 6 7 8 9 10
ksys,mean 1.00 1.05 1.07 1.09 1.10 1.11 1.12 1.12 1.13 1.13
ksys,k 1.00 1.06 1.10 1.12 1.13 1.14 1.15 1.16 1.17 1.17

Although the withdrawal stiffness is not part of the model in Ringhofer et al. (2015),
we propose to apply the same values (ksys,mean in Tab. 1) for this property. Main rea‐
sons are: (i) the referenced stochastic approach only depends on density without ex‐
clusive calibration to withdrawal strength and (ii) mean values of withdrawal stiffness
determined by Reichelt (2012) show a relationship with N similar to that observed for
withdrawal strength.
3.3 Power value kρ considering the influence of density in case of N = 1
As declared in Eq. (1), we consider the influence of densities differing from ρref on
withdrawal properties by a power function. Thereby, kρ serves as power value deter‐
mined as gradient of the linear regression model between ln(fax) vs. ln(ρ), see Eq. (7).
∙ → ∙ . (7)
Current model approaches discussed in Section 1 consider kρ as constant value vary‐
ing from 0.75 to 1.60. This limits their ability covering explicitly influences from input
parameters others than density on the relationship between ρ and withdrawal prop‐
erties fax and kser. Ringhofer et al. (2014b) observed a significant impact of α and d on
kρ for withdrawal strength fax. Following that we aim on deriving kρ as steady function
in dependence of these two parameters. This was done by doing linear regression
analysis according to Eq. (7) for all withdrawal tests in solid timber, see Tab. 1. After
outlier treatment about 5,000 results (56 single test series) for withdrawal strength
and about 3,000 results (71 single test series) for withdrawal stiffness remained.
Blaß et al. (2006) and own data show, that withdrawal stiffness kser is also significantly
governed by the screw’s effective thread length. Consequently, test series were
grouped according to their lef. Results of the regression analysis for withdrawal
strength analysing kρ vs. α (left, for d = {8; 12 mm}) and d (right for α = {0; 90 °}) are

85
INTER / 48 - 07 - 01

shown in Fig. 3.
1.8
1,8 1.8
1,8
1.6
1,6 1.6
1,6
1.4
1,4 1.4
1,4
1,2
1.2 1,2
1.2
kρ [‐]

1,0
1.0 1,0
1.0

kρ [‐]
0,8
0.8 0,8
0.8
0,6
0.6 0,6
0.6 MS 0 °
MS Ø 12 mm MS Ø 8 mm MS 90 °
0,4
0.4 0,4
0.4
SubS Ø 12 mm SubS Ø 8 mm SubS 0 °
0,2
0.2 Mod. Ø 12 mm Mod. Ø 8 mm
0,2
0.2 SubS 90 °
0,0
0.0 0,0
0.0
0 15 30 45 60 75 90 4 6 8 10 12
α [°] d [mm]
Figure 3. Power value kρ vs. insertion angle (left) and outer thread diameter (right) for withdrawal
strength fax; MS = main dataset; SubS = single test series.
Values of kρ for main data set (MS) show a comparatively high variability. Neverthe‐
less, decreasing kρ with decreasing α as well as a clear negative trend with increasing
d for α = 0 ° can be observed. By considering the position of error bars, corresponding
to the bandwidths of kρ and determined for related single test series (SubS), for sim‐
plification we conclude a constant behaviour of kρ between 30 ° ≤ α ≤ 90 ° followed
by a decrease for α < 30 °. This is in fact quite similar to the behaviour of withdrawal
strength vs. α as e. g. described in Hübner (2013). In analogy to that,
Müller et al. (2015) observed decreasing correlation and inclination coefficient be‐
tween shear modulus (as a potential indicator for withdrawal strength) and density
with decreasing number of annual rings exposed to shear. With regard to the behav‐
iour of kρ in Fig. 3, for modelling the relationship between density and withdrawal
strength in dependence of α and d we apply Eq. (8),
∙
fax: , , and {a; b; kρ,90} = {–0.05; 0.15; 1.10}, (8)
/

with kρ,90 as reference value at α = 90 ° and a and b as model parameters, determined

by calibrating with least‐squares method. For withdrawal stiffness kser and applying
the same procedure, no clear trend between kρ and α, d or lef, but unexpectedly a
high variability of the power factor were observed. Possible reasoning is the lower
accuracy in measuring and examining kser compared to withdrawal strength. Conse‐
quently we treat kρ for withdrawal stiffness as constant value and ignore variation of
the aforementioned parameters, see
kser: 0.75 for 0 ° ≤ α ≤ 90 °, 6 mm ≤ d ≤ 12 mm and 10 mm ≤ lef ≤ 310 mm. (9)
3.4 Function kax considering the influence of angle and gap variation
By inserting screws in CLT narrow faces, as illustrated in Fig. 1, a significant relation‐
ship between axis‐to‐grain angle α and the positioning relative to gaps is given. To ac‐
count for both, we formulate kax = f(α,kgap). Consequently, we analysed the depend‐
encies of withdrawal strength and stiffness on α for screw application in solid timber

86
 INTER / 48 - 07 - 01

(kgap = 1.00 and N = 1). Therefore, both properties were normalised by applying
Eq. (8) and (9), as discussed in Section 3.3. As local displacement measurements are
not available for all tests, withdrawal stiffness of test series with varying angles were
additionally referred to each kser,mean at α = 90 °. Fig. 4 shows the relationship of both
withdrawal properties and α, as combined scatter‐ and boxplot‐graph (notches illus‐
trate the 95 % confidence interval of each median). Values of fax,mean are widely con‐
stant between 15 °≤ α ≤ 90 ° while at α = 0 °, fax,mean is significantly lower. In analysing
the same relationship but for the 5 %‐quantiles of fax (empirically determined and
with 2pLND assumption) already at α ≤ 45 ° a distinctive linear decrease in fax,05 is giv‐
en. In case of withdrawal stiffness, the same but inverse relationship is observed for
mean values, while 5 %‐quantiles show a slight but steady increasing behaviour for
α < 90 °.
1.75
1756

90

85
62
81
378

65

81

2462

774

61

162

63

924
2.50
1.50 2.25 kser,norm,mean
2.00 kser,norm,emp,05
kser,norm [‐]
1.25
fax,norm [‐]

kser,norm,2pLND,05
1.75 Mod. mean (R² = 0,90)
1.00 1.50
0.75 1.25
fax,norm,mean Mod. mean (R² = 0,59)
1.00
0.50 fax,norm,emp,05 Mod. 05 (R² = 0,93) 0.75
fax,norm,2pLND,05
0.25 0.50
0 15 30 45 60 75 90 0 15 30 45 60 75 90
α [°] α [°]

Figure 4. Normalised withdrawal properties (left: strength, right: stiffness) vs. insertion angle α.
Basing on these circumstances and similar to Hübner (2013), we decided to model
the relationship between α and both withdrawal properties with a bilinear approach
given in Eq. (10) and illustrated in Fig. (4):
1.00 45 ° 90 °
, (10)
∙ 0° 45 °
with c = k90‐1 = X0 / X90 and k90 = {1.35; 1.56 and 0.75} for {fax; fax,05 and kser}. Differ‐
ences of k90 between withdrawal strength mean and 5 %‐values can be explained by
an increasing variability of fax with decreasing α.
3.5 Parameter kgap considering the influence of gap insertion
As analysed in Brandner (2013), current technical assessments of CLT panels include
gaps between two boards of one layer with wgap up to 6 mm. In addition, withdrawal
properties are also influenced by the gap type (butt joint, BuJ; T‐joint, TJ; bed joint,
BeJ; c. f. Fig. 5). Brandner et al. (2015) show that an observed loss of withdrawal
strength and stiffness for screws inserted in gaps can be simply modelled by the cor‐
responding reduction of the screw’s effective lateral area Uef ∙ lef. One possibility to
quantify the gap related influence on withdrawal properties of screws randomly situ‐

87
INTER / 48 - 07 - 01

ated in CLT narrow faces (same approach also applicable on side face) is now intro‐
duced briefly.
We use a multi‐modal density function fX,CLT(x) of the property X, defined as sum of
single density functions for specific axis‐to‐grain angles, gap types and widths. As
shown in Eq. (11), they are weighted by their specific probabilities of occurrence pi
which correspond to area ratios Ai, see Fig. 5.
, | , ∙ , ∙ , | ∙ |
, | ∙ | , | ∙ | , with ∑ 1. (11)

Figure 5. Definition of gap types and illustration of areas related to different axis‐to‐grain directions
and gap types exemplarily for screw insertion normal to the CLT narrow face.
The possible cross‐sectional area for screw insertion in CLT narrow face considers a
thickness of the CLT‐element which is on both sides reduced by the minimum spacing
perpendicular to the panel’s axis a2,c as determined in Uibel and Blaß (2007), see
Fig. 5. All in all, related analysis comprised 16 different lay‐ups (N = {3; 5}) of four
leading European CLT manufacturers with standard board thicknesses
tl = {20; 30; 40 mm}, board widths wl = {80; 160; 240 mm}, gap widths
wgap = {0; 2; 4; 6 mm} and screw diameters d = {8; 10; 12 mm}. Screw insertion in
closed gaps (wgap = 0 mm) was modelled without influence on withdrawal properties,
c. f. Brandner et al. (2015). Since screw insertion in T‐joints is hardly conceivable in
practice, corresponding Ai are assigned to those of butt joints. In dependence of
board and gap width, area ratios determined for screws positioned in open gaps
(wgap > 0 mm) with d = 8 to 12 mm vary between 12 to 21 % in case of wl = 80 mm
and between 4 to 7 % in case of wl = 240 mm. Maxima were found for comparatively
thin boards with high wgap and d. Enabling practical applicability we thus consider a
constant and conservative probability for gap insertion pBuJ & TJ|wgap of 25 %. With

, | , ∙ 0.75 , & | ∙ 0.25, and , 16 %,(12)

88
 INTER / 48 - 07 - 01

and a deterministic gap width, it can be shown that both withdrawal properties X
should be reduced about 10 – 15 % if α = 0 ° and wgap > 0 mm. For α > 0 °, the influ‐
ence of kgap on X soon converges to 1.00. Consequently, we propose to modify the
parameter c introduced in Section 3.4 for CLT narrow face application as follows:
∙ , , with kgap,0 = {0.85; 0.90; 0.85} for {fax,mean; fax,05; kser,mean}. (13)

4 Model verification
Verification of our model with test results is done in two steps. In the first step, we
analyse the suitability of both isolated derived parameters kρ and kax, examined in
Section 3 for solid timber, by using the complete model approach in Eq. (1) to esti‐
mate with the test results used for parameter determination. Applying Eq. (1), refer‐
ence values for withdrawal properties and density (ρref = {427; 428 kg/m³} for
{fax; kser}) have thus to be specified. While withdrawal strength fax,ref can in principle
be determined according to one of the regression functions discussed in Section 1,
measured withdrawal stiffness significantly deviates from both therein mentioned
approaches. Consequently, we decided to use own regression models for kser,ref and
also for fax,ref in order to minimise inaccuracies caused by reference value determina‐
tion as far as possible. Justification of this procedure is also argued by conducting a
second verification of our models on independent data sets, i.e. data sets not used
for calibrating model parameters. Data sets (solid timber, α = 90 °) used for these re‐
gression models comprise about 2,500 results for withdrawal strength and about 600
results for withdrawal stiffness. For kser only tests with local displacement measure‐
ments were considered. In case of fax,ref, nonlinear regression analysis was carried out
applying an approach similar to that presented in Frese and Blaß (2009), see Eq. (14):
, ∙ ∙ , R² = 0.57, (14)
with {e; f; g} = {0.014; 1.11; ‐0.33}. In case of kser,ref, the length of the inserted thread‐
ed part of the screw has a significant influence and had thus to be additionally con‐
sidered, see Eq. (15):
, ∙ ∙ ∙ , R² = 0.85, (15)
with {h; i; j; m} = {24.7; 0.75; –1.70; –0.60}. Comparison of test data with values pre‐
dicted by Eq. (1) and (2) (whole dataset for solid timber) is shown in Fig. 6. Following
conclusions are made: Locations of data points as well as the course of the regression
lines given in dependence of α indicate a high conformity (R² = {0.65; 0.73} for
{fax; kser}) of test results with model predictions, especially for withdrawal strength fax.
In case of withdrawal stiffness, we also observe a high accuracy for α = {0; 90 °}, while
experimentally determined kser for α = 45 ° are slightly underestimated. This is in fact
caused by the bilinear model considering angle influence on X, which estimates a
smaller value for α = 45 °, c. f. Fig. 4.

89
INTER / 48 - 07 - 01

12.0 n ≈ 5,000 n ≈ 1,000

15.0
10.0

kser,exp [N/mm³]
fax,exp [N/mm²]

8.0 10.0

6.0
all 5.0
4.0 α = 90° α = 90°
α = 45° α = 0°
α = 0° other
2.0 0.0
2.0 4.0 6.0 8.0 10.0 12.0 0.0 5.0 10.0 15.0
fax,pred [N/mm²] kser,pred [N/mm³]

Figure 6. Comparison of test results with predicted values; left: strength; right: stiffness.
In the second step, we apply our universal approach given in Eq. (1) and (2) for esti‐
mating withdrawal strength and stiffness of self‐tapping screws situated in both
products GLT and CLT. Data used for comparison with experimental results was taken
from the investigations mentioned in Section 1 and comprises after outlier treatment
about 2,800 tests for withdrawal strength and about 1,300 tests for withdrawal stiff‐
ness. Since we know the number of layers penetrated by the screws as well as their
position in gaps, we consider ksys according to Tab. 3 and kgap in form of the ratio be‐
tween the effective and the total lateral area of the screw. Following the assumptions
made in Section 3.5, T‐joints were treated as butt joints while in case of bed joints
the corresponding stiffness was averaged considering differences in interacting
thread‐grain angles.
Fig. 7 consequently compares model predictions with tests results for both properties
fax and kser. Although the data sets applied for this verification were not used for cali‐
brating the model parameters for Eq. (1), a high correspondence between observed
and predicted withdrawal strengths is given (R² = 0.66). Furthermore, specific regres‐
sion lines for α = {0; 90 °} indicate high conformity irrespective the position screws
were inserted. With regard to withdrawal stiffness, we observe comparatively high
deviations of test results from model predictions (R² = 0.55), especially for data sets
with α = 0 ° and gap variation. This is mainly caused by underestimating the size of
experimental values related (declared as “gap tests” in Fig. 7, right). Further devia‐
tions can be explained by inaccuracies in determining kser as well as by the constant
value for kρ (c. f. Eq. (9)), aimed to be adapted in the frame of further considerations.
In case of α = 90 °, our approach slightly overestimates measured stiffness but shows
higher conformity if compared to α = 0 °.

90
 INTER / 48 - 07 - 01

10.0 n ≈ 2,800 n ≈ 1,300

15.0
8.0

kser,exp [N/mm³]
fax,exp [N/mm²]

6.0 10.0

4.0
5.0 α = 90°
2.0 all α = 0°
α = 90° gap tests
α = 0° other
0.0 0.0

0.0 2.0 4.0 6.0 8.0 10.0 0.0 5.0 10.0 15.0
fax,pred [N/mm²] kser,pred [N/mm³]

Figure 7. Comparison of test results in GLT and CLT with predicted values; left: strength; right:
stiffness.

5 Derivation of a characteristic approach for

withdrawal strength
According to the current semi‐probabilistic design approach, the ultimate limit state
(ULS) design considers characteristic properties of action and resistance. With regard
to axially loaded screws, we thus apply Eq. (1) for deriving an approach determining
the characteristic (5 %‐) value of withdrawal strength in unidirectionally and orthogo‐
nally layered laminated timber products and solid timber. Eq. (16) includes the rela‐
tionship between mean value, variability and 5 %‐quantile of a 2pLND variable:
. exp Φ 0.05 ∙ ∙ , (16)
with μY and σY as mean value and standard deviation of Y = ln(X), Ф‐1(0.05) as inverse
of standard ND for a given quantile p = 5 % and ξ as ratio between the mean value
and 5 %‐quantile of X. Assuming [fax, ρ] ~ 2pLND and {CV[ρ]; CV[fax]} =
{8 %; 1.5∙CV[ρ] = 12 %}, ξ results to {0.874; 0.816}. Consequently, the characteristic
withdrawal strength (with model parameters for 5 %‐quantile derived in Section 3) is
determined according to Eq. (17) and (18):

, 0.816 ∙ , ∙ , ∙ , , ∙
,

, ∙ , ∙ , , ∙ , with (17)
,

. .
, , 0.013 ∙ , ∙ ,
1.00 45 ° 90 °
, . ∙ , (18)
0.64 ∙ ∙ 0° 45 °

91
INTER / 48 - 07 - 01

and kρ and ksys,k according to Eq. (8) and Tab. 3. Fig. 8 compares characteristic with‐
drawal strengths fax,k (kgap was determined as described in Section 4) estimated by
Eq. (17) with empirical 5 %‐values of fax of all single test series the whole dataset
(n ≈ 8,000) consists of. Grey coloured data points illustrate test series in solid timber,
black ones (symbols differ if N was known or estimated) those in GLT and CLT exclu‐
sively used for model verification. With regard to their locations and the course of
corresponding regression lines, we can conclude high agreement between model
predictions and test data (R² = 0.78), independent from the material used and further
influencing parameters such as the axis‐to‐grain angle. Sole exceptions are four test
series marked as outliers, which are significantly overestimated by our approach. In
fact, all of them show unexpected high variabilities of withdrawal strength and densi‐
ty (CV[X] ≥ 20 %). Since we considered usual variation of density by CV[ρ] = 8 % for
determining ρk,i of all test series, given difference between fax,05,exp and fax,k can be
quantified.
6.0 all
GLT|CLT
ST
fax, 05,exp [N/mm²]

5.0

4.0

3.0 GLT|CLT (ksys assum.)

GLT|CLT (ksys exact)
solid timber
2.0 „outlier“

2.0 3.0 4.0 5.0 6.0

fax,k [N/mm²]
Figure 8. Comparison of empirical 5 %‐quantiles of all test series with characteristic (5 %) withdrawal
strengths determined according to Eq. (17).

6 Summary and conclusion

In the frame of this paper and basing on copious investigations concerning the axial
load bearing behaviour of self‐tapping screws placed in laminated timber products,
we derived, discussed and verified a new universal approach estimating related with‐
drawal properties. Therein, multiplicative k‐factors (or functions) enable the consid‐
eration of varying influencing parameters such as the density, the screw’s outer
thread diameter, the axis‐to‐grain angle, the number of penetrated layers as well as
gap width and type. Consequently, we can estimate withdrawal strength and stiffness
irrespective the timber product used and the position the screw is inserted. In addi‐
tion, we used this model to derive a characteristic approach for withdrawal strength
and verified it again with experimental data concluding high conformity. For practical

92
 INTER / 48 - 07 - 01

application with random gap insertion we propose to determine fax,k as given in

Eq. (19) in simplified form:

, , ∙ , ∙ , , ∙ , with (19)
,

1.00 45 ° 90 °
, . ∙ , (20)
0.64 ∙ ∙ 0° 45 °

0.90 1.10 0° 90 °
, , , (21)
1.00 1.25 0.05 ∙ 0°
1.00
, 1.10 if N ≥ 3, (22)
1.13
and fax,ref,k for α = 90 °, either determined according to Eq. (18), with one of the cur‐
rent existing models discussed in Section 1 or as given in screw manufacturers’ Euro‐
pean Technical Assessments (ETAs). Although, we implicitly consider gap influence in
CLT narrow faces we still recommend to avoid corresponding screw insertion parallel
to grain because of a questionable long‐time behaviour which has not been conclu‐
sively investigated so far.
With regard to axial stiffness kser considering displacements of both timber and steel
(the inserted threaded part) components, the specific approach derived also covers
the effects mentioned before but shows more inaccuracy in prediction, especially in
cases of α = 0 °. Consequently, we aim to increase its predictability in the frame of
further investigations. Nevertheless, general suitability is given and thus also recom‐
mended for consideration in standardisation and assessments.

7 Acknowledgement
The research work within the project ‘focus_sts’ was financed by the competence
centre holz.bau forschungs gmbh and performed in cooperation with the Graz Uni‐
versity of Technology, Institute of Timber Engineering and Wood Technology. The
project was financed by funds from the Federal Ministry of Economics, Family and
Youth, the Federal Ministry of Transport, Innovation and Technology, the Styrian
Business Promotion Agency Association and the Province of Styria (A12), the
Carinthian Economic Promotion Fund (KWF), the Province of Lower Austria Depart‐
ment of Economy, Tourism and Technology as well as the Business Location Tirol.

8 References
Blaß, H‐J, Bejtka, I, Uibel, T (2006) Tragfähigkeit von Verbindungen mit selbstbohren‐
den Holzschrauben (in German). Bd. 4, Karlsruher Berichte zum Ingenieurholzbau,
Karlsruhe: KIT Scientific Publishing.

93
INTER / 48 - 07 - 01

Brandner, R (2012) Stochastic system actions and effects in engineered timber prod‐
ucts and structures. Dissertation, Graz University of Technology, Graz.
Brander, R (2013) Production and Technology of Cross Laminated Timber (CLT): A
state‐of‐the‐art report. In: Harris, R, Ringhofer, A, Schickhofer, G (Eds.) Focus Solid
Timber Solutions – European Conference on Cross Laminated Timber (CLT), Univer‐
sity of Bath, COST Action FP1004, ISBN 1‐85790‐181‐9.
Brandner, R, Ringhofer, A, Grabner, M (2015) Probabilistic Models for the Withdrawal
Behaviour of Single Self‐Tapping Screws in the Narrow Face of Cross Laminated
Timber. Wood Science and Technology (submitted).
Bratulić, K (2012) Alteration of the withdrawal strength of self‐tapping screws along
the board and over the varying GLT cross section. Master Thesis, Graz University of
Technology, Graz.
Burgschwaiger, M (2010) Einfluss der Einbindelänge auf die Ausziehfestigkeit von
Teilgewindeschrauben (in German). Bachelor Thesis, Graz University of Technology,
Graz.
EN 26891 (1991) Timber structures – Joints made with mechanical fasteners ‐ Gen‐
eral principles for the determination of strength and deformation characteristics.
(CEN).
EN 1382 (1999) Timber structures – Test methods – Withdrawal capacity of timber
fasteners. (CEN).
EN 14358 (2006) Timber structures – Calculation of characteristic 5‐percentile values
and acceptance criteria for a sample. (CEN).
EN 1995‐1‐1+A1 (2008) Eurocode 5: Design of timber structures – Part 1‐1: General –
Common rules and rules for buildings. (CEN).
ETA‐12/0062 (2012): SFS self‐tapping screws WR. Austrian Institute of Construction
Engineering (OIB).
Frese, M, Blaß, H‐J (2009) Models for the calculation of the withdrawal capacity of
self‐tapping screws. Meeting 42 of the Working Commission W18‐Timber Struc‐
tures, CIB, Dübendorf (Switzerland), paper CIB‐W18/42‐7‐3.
Grabner, M (2013) Einflussparameter auf den Ausziehwiderstand selbstbohrender
Holzschrauben in BSP‐Schmalflächen (in German). Master Thesis, Graz University of
Technology, Graz.
Hübner, U (2013) Withdrawal strength of self‐tapping screws in hardwoods. Meeting
46 of the Working Commission W18‐Timber Structures, CIB, Vancouver (Canada),
paper CIB‐W18/46‐7‐4.
Müller, U, Ringhofer, A, Brandner, R, Schickhofer, G (2015) Homogeneous shear
stress field of wood in an Arcan shear test configuration measured by means of
electronic speckle pattern interferometry: description of the test setup. Wood Sci‐
ence and Technology, DOI 10.1007/s00226‐015‐0755‐3, (in press).

94
 INTER / 48 - 07 - 01

Olsson, U (2005) Confidence Intervals for the Mean of a Log‐Normal Distribution.

Journal Statistics Education, Volume 13, Issue (1), pp. 1‐8.
Pirnbacher, G, Schickhofer, G (2007) Schrauben im Vergleich – eine empirische Be‐
trachtung (in German). 6. Grazer Holzbau‐Fachtagung (6. GraHFT’07), Graz (Aus‐
tria).
Pirnbacher, G, Brandner, R, Schickhofer, G (2009) Base parameters of self‐tapping
screws. Meeting 42 of the Working Commission W18‐Timber Structures, CIB,
Dübendorf (Switzerland), paper CIB‐W18/42‐7‐1.
Plüss, Y (2014) Prüftechnische Ermittlung des Tragverhaltens von Schraubengruppen
in der BSP‐Schmalfläche (in German). Master Thesis, Graz University of Technology,
Graz.
Reichelt, B (2012) Einfluss der Sperrwirkung auf den Ausziehwiderstand selbstboh‐
render Holzschrauben – Eine vergleichende Betrachtung zwischen BSP und BSH (in
German). Master Thesis, Graz University of Technology, Graz.
Ringhofer, A, Ehrhart, T, Brandner, R, Schickhofer, G (2013) Prüftechnische Ermittlung
weiterer Einflussparameter auf das Tragverhalten der Einzelschraube in der BSP‐
Seitenfläche (in German). Research Report, holz.bau forschungs gmbh, Graz Uni‐
versity of Technology, Graz.
Ringhofer, A, Grabner, M, Silva, C, Branco, J, Schickhofer, G (2014a) The influence of
moisture content variation on the withdrawal capacity of self‐tapping screws. Holz‐
technologie, Volume 55, Issue (3), pp. 33‐40.
Ringhofer, A, Grabner, M, Brandner, R, Schickhofer, G (2014b) Die Ausziehfestigkeit
selbstbohrender Holzschrauben in geschichteten Holzprodukten (in German).
Doktorandenkolloquium Holzbau “Forschung und Praxis”, Stuttgart (Germany).
Ringhofer, A, Schickhofer, G (2014) Influencing parameters on the experimental de‐
termination of the withdrawal capacity of self‐tapping screws. 13th World Confer‐
ence on Timber Engineering (WCTE), Quebec (Canada).
Ringhofer, A, Brandner, R, Schickhofer, G (2015) Withdrawal resistance of self‐
tapping screws in unidirectional and orthogonal layered timber products. Materials
and Structures, Volume 48, Issue (5), pp. 1435‐1447.
Silva, C, Ringhofer, A, Branco, J, Lourenco, P‐B, Schickhofer, G (2014) Influence of
moisture content and gaps on the withdrawal resistance of self‐tapping screws in
CLT. 9th Congresso Nacional de Mecânica Experimental, Aveiro (Portugal).
Uibel, T, Blaß, H‐J (2007) Edge joints with dowel type fasteners in cross laminated
timber. Meeting 40 of the Working Commission W18‐Timber Structures, CIB, Bled
(Slovenia), paper CIB‐W18/40‐7‐2.

95
INTER / 48 - 07 - 01

Discussion

The paper was presented by G Schickhofer

H Blass commented and discussed about the usefulness of Kser (screw’s withdrawal
stiffness) as Kser is very much dependent on test configuration. G Schickhofer agreed
as the test configuration dependency of stiffness is different from that of strength. R
Jockwer commented that there are discussions in EC5 about test configuration. G
Schickhofer agreed that the availability of a standardized test configuration would be
good.
S Winter commented that he did not doubt the accuracy of the equations but they
should be further simplified for practising engineers. G Schickhofer disagreed as there
are so many products but the equation has three main variables for consideration.
H Stamatopoulos commented that withdrawal stiffness had more factors affecting
the values.
P Zarnani asked about different screw manufacturers. G Schickhofer answered that
different screws should have little effect in terms of the values. T Tannert commented
that some believe in Canada that the design method for self-tapping wood screws
should be extended to all types of screws. He asked whether the model fits to other
data or other types of screws. G Schickhofer answered that this would be a good idea.
A Salenikovich agreed that diameter mattered and not the product type and that the
difference between self-tapping and non self-tapping screws is small. H Blass stated
that this would depend on density of the wood as some species would need predrill-
ing.
K Malo supported that withdrawal stiffness would be important for vibration cases.
M Flaig commented that using a probabilistic approach about gap influence would be
valuable. F Lam commented that angle application of screws would lessen the influ-
ence of gaps. G Schickhofer agreed.

96
97
98
 INTER / 48 - 07 - 02

Characteristic withdrawal capacity and

stiffness of threaded rods

Haris Stamatopoulos and Kjell Arne Malo

Norwegian University of Science and Technology (NTNU), Trondheim, Norway

Keywords: Threaded rod, withdrawal capacity, withdrawal stiffness, embedment

length, rod-to-grain angle, withdrawal strength parameter

1 Introduction

Long threaded rods show high withdrawal capacity and stiffness and thus they may
be used in order to realize strong and stiff connections for timber structures. In com-
parison to dowel-type connectors, they have no initial soft response and no initial
slip. In comparison to glued-in-rods they are less prone to construction quality issues,
less brittle and offer greater protection against high temperatures (Mischler and
Frangi 2001). Due to their length, their withdrawal capacity and stiffness are not sig-
nificantly affected by local defects. Furthermore, a high degree of pre-fabrication is
possible and hence easy and fast erection on site may be achieved.
Over the last years, the vast majority of the research effort has been devoted to the
withdrawal capacity of screws with diameters up to 12 mm. The influence of parame-
ters such as the embedment length and the angle between the screw axis and the
grain direction has been investigated; see for example (Pirnbacher, Brandner and
Schickhofer 2009, Frese and Blaß 2009). On the other hand, the research effort on
the withdrawal capacity and also stiffness of threaded rods with diameters up to 20-
25mm has not been so intensive and mostly it is limited to rods installed parallel and
perpendicular to the grain (Jensen et al. 2011, Jensen et al. 2012, Nakatani and
Komatsu 2004, Mori et al. 2008).
Eurocode 5, EC5 (CEN 2004) do not provide guidelines for the estimation of the with-
drawal stiffness which is required for the evaluation of the stiffness of connections
with threaded connectors (Tomasi, Crosatti and Piazza 2010, Malo and Ellingsbø

99
INTER / 48 - 07 - 02

2010). Some expressions may be found in technical approvals of screws, but mostly
these expressions are valid for screws with relatively small diameters. Moreover, EC5
does not allow the installation of rods in an angle to the grain less than 30° in order
to eliminate the risk of splitting failure. However, in practice, it may be desired to in-
stall threaded rods in an angle to the grain smaller than 30° (in combination with
some sort of reinforcement to prevent splitting failure).
In the present paper, an experimental study on withdrawal of threaded rods embed-
ded in glue-laminated timber (abbr. glulam) elements is presented. The parameters
of this study were the embedment length and the angle between the rod axis and the
grain direction (with emphasis on angles which are smaller than 30°). Moreover, ana-
lytical expressions for the estimation of withdrawal capacity and stiffness are provid-
ed. The characteristic withdrawal capacity and the mean withdrawal stiffness were
obtained by the experimental results and compared to the analytical estimations.

2 Experimental methods

2.1 Experimental set-up

The experimental set-up for the withdrawal tests is presented in Figure 1. As shown,
the loading condition of the specimens was a ‘remote’ pull-push (i.e. the support was
provided in the same plane surface as the entrance of the rod, but at a distance to
the rod). A thin steel plate, as shown in Fig. 1d, was placed between the supports and
the specimen. The plate was used to counteract bending stresses and prevent tensile
splitting failure, while allowing local deformation on the surface of the specimen in
the vicinity of the rod. Two displacement transducers were placed next to the sup-
ports of the specimen, measuring the relative displacement between the rod and
support as shown in Figures 1a, 1c and 1e. The average of these two measurements
was used for the displacement. Testing was performed using the loading protocol
given in EN 26891:1991 (ISO6891:1983) (CEN 1991).

2.2 Materials
The specimens were cut from glulam beams of Scandinavian class L40c which corre-
sponds to European strength class GL30c (CEN 2013). This type of glulam is fabricat-
ed with 45 mm thick lamellas, made of Norwegian spruce (Picea Abies). The mean
and characteristic density of L40c is ρmean = 470 kg/m3 and ρk = 400 kg/m3 respective-
ly. The mean moduli of elasticity, parallel and perpendicular to the grain, are E0.mean =
13000 MPa and E90.mean =410 MPa respectively, and the shear modulus is G= 760
MPa.

100
 INTER / 48 - 07 - 02

For increased homogeneity, all specimens were manufactured such that the rods
were inserted in the inner, weaker lamellas of the beams. SFS WB-T-20 (DIBt 2010)
steel threaded rods were used. These rods are made according to DIN7998 (DIN
1975). The outer-thread diameter d of the rods is 20 mm and the core diameter, dc, is
15 mm. According to the manufacturer, the steel grade of the rods is 8.8 and their
characteristic tensile capacity is 145 kN.

2.3 Specimens
Prior to rod installation, all specimens were pre-drilled with a diameter equal to dc. All
specimens were conditioned to standard temperature and relative humidity condi-
tions (20°C / 65% R.H.), leading to approximately 12% moisture content in the wood.
The parameters of the experimental investigation were the rod-to-grain angle, α, and
the embedment length of the rod, lef. Specimens with 6 different rod-to-grain angles
(α = 0, 10, 20, 30, 60 and 90°) and 4 different embedment lengths (lef = 100, 300, 450,
600 mm) were tested. The series of specimens are denoted Sα-lef, based on their rod-
to-grain angle and embedment length. The width, b, of the glulam beams and conse-
quently of all specimens was equal to 140 mm. A full description of the specimens’
dimensions can be found in (Stamatopoulos and Malo 2015b).

Figure 1. Experimental set-up: (a) 3D representation, (b) plan view, (c) side view, (d) steel plate and
(e) photo

101
INTER / 48 - 07 - 02

3 Eurocode 5

According to EC5 (for screws with d > 12 mm) the characteristic withdrawal capacity,
Fax.Rk, is given by (the expression is re-arranged):

𝐹𝑎𝑎.𝑅𝑅 = 𝑛𝑒𝑒 ∙ 𝑓𝑎𝑎 .α.𝑘 ∙ 𝑑 ∙ 𝑙𝑒𝑒 (1)

The parameter nef is the effective number of screws and equal to nef = n0.9, where n is
the number of screws acting together in a connection. The withdrawal strength pa-
rameter, fax.α.k, is given by:

𝑓𝑎𝑎 .90.𝑘 𝜌𝑘 0.8

𝑓𝑎𝑎 .α.𝑘 = ∙� � (𝛼 ≥ 30°) (2)
1.2 ∙ cos 2 𝛼 + sin2 𝛼 𝜌a
where fax.90.k is the withdrawal strength parameter perpendicular to the grain which
must be experimentally determined, for the associated density ρa. EC5 provides no
guidelines for the estimation of withdrawal stiffness.
In the technical approval of WB-T-20 rods, Z-9.1-777 (DIBt 2010), the following ex-
pression is provided for the withdrawal strength parameter (unit MPa and kg/m3):

𝑓𝑎𝑎 .𝑘 = 70 ∙ 10−6 ∙ 𝜌𝑘 2 (45° ≤ 𝛼 ≤ 90°) (3)

4 Analytical model

Analytical estimations can be obtained by use of the concept of the classical

Volkersen theory (Volkersen 1938), applied for axially loaded connectors (Jensen et
al. 2001). This model has initially been developed assuming that all shear defor-
mation occurs in an infinitely thin shear layer, while the connector and surrounding
wood are assumed to be in states of pure axial stress. The shear stress-displacement
behaviour (τ - δ) of the shear layer is approximated by a linear constitutive law, which
is a reasonable approximation for glued-in-connectors.
In the case of screwed-in connectors, however, it is more convenient to assume a bi-
linear constitutive law, because these connectors are by far less brittle than glued-in
connectors and their post-elastic behaviour should not be omitted. The bi-linear con-
stitutive law is presented in Figure 2. The bi-linear idealization separates the curve in
two distinct domains; the linear elastic domain and the fracture domain. These do-
mains are characterized by the equivalent shear stiffness parameters Γe and Γf, which
are the slopes of the two branches of the bi-linear constitutive law. The advantage of
this method is that, apart from the withdrawal capacity and stiffness, it also allows

102
 INTER / 48 - 07 - 02

the estimation stress and displacement distributions for any given withdrawal force
level. Thus, an analytical estimation of the force-displacement curve can be obtained.
Note that all shear deformation is assumed to occur in a shear zone of finite dimen-
sions. A full description of this method is given in (Stamatopoulos and Malo 2015a).

Figure 2. Bi-linear approximation of τ-δ curve

The withdrawal stiffness, Kw, and the characteristic withdrawal capacity, Fax.Rk, are
provided by the following expressions (Stamatopoulos and Malo 2015a, Jensen et al.
2001):

tanh𝜔 (4)
𝐾𝑤 = π ∙ 𝑑 ∙ 𝑙𝑒𝑒 ∙ 𝛤𝑒 ∙
𝜔

𝐹𝑎𝑎.α.𝑅𝑅 sin(𝑚 ∙ 𝜔 ∙ 𝜆𝑢 ) tanh{(1 − 𝜆𝑢 ) ∙ 𝜔} ∙ cos (𝑚 ∙ 𝜔 ∙ 𝜆𝑢)

= + (5)
𝑑 ∙ 𝑙𝑒𝑒 ∙ 𝑓𝑎𝑎 .𝛼.𝑘 𝜔∙𝑚 𝜔
Note that these expressions are valid for pull-push or pull-shear loading conditions,
but not for the pull-pull loading condition. The parameter m has been introduced as:

𝑚 = �𝛤𝑓⁄𝛤𝑒 (6)

This parameter is a measure of the brittleness of the shear zone. In the limits, m→ 0
indicates perfect plastic post-elastic behaviour, while m→ ∞ indicates totally brittle
behaviour. The parameters ω and β have been defined as follows:

𝜔 = �𝜋 ∙ 𝑑 ∙ 𝛤𝑒 ∙ 𝛽 · 𝑙𝑒𝑒 2 (7)

1 1
𝛽= + (8)
𝐴𝑠 ∙ 𝐸𝑠 𝐴𝑤 ∙ 𝐸𝑤.𝛼

103
INTER / 48 - 07 - 02

where Es and Ew.α are the moduli of elasticity of steel and wood (as function of α), re-
spectively. Τhe core cross-sectional area of the rod is As = π∙dc2/4 and Aw is the area of
wood subjected to axial stress. Ew.α may be estimated by the Hankinson formula and
Aw by an effective area, confer (Stamatopoulos and Malo 2015b). The parameter λu is
a dimensionless length parameter which expresses the percentage of the embed-
ment length (at failure), in which post-elastic behaviour takes place and it can be de-
termined by the diagram in Figure 3.

Figure 3. Diagram for the determination of parameter λu

The parameters Γe (in MPa/mm) and m are provided as functions of α, by the follow-
ing expressions(Stamatopoulos and Malo 2015a):

9.35
𝛤𝑒.𝛼 = (9)
1.5 ∙ sin2.2 𝛼 + cos 2.2 𝛼

𝑚0 0.332
𝑚𝛼 = = (10)
(𝑚0⁄𝑚90) ∙ sin𝛼 + cos𝛼 1.73 ∙ sin𝛼 + cos𝛼
Finally, fax.α.k can be calculated by Equation (2).

5 Results and discussion

5.1 Withdrawal stiffness

The experimentally derived mean values of Kw and the coefficient of variation (abbr.
C.o.V.) for all embedment lengths and rod-to-grain angles are summarized in Table 1.
The sample size for each sub-set of parameters (lef and α) was 5 tests. The analytical

104
 INTER / 48 - 07 - 02

estimations are compared to the experimental results in Figure 4, where Kw is plotted

as function of lef for all rod-to-grain angles. Results from finite element simulations
are also provided in Figure 4. The finite element model has been presented in detail
in (Stamatopoulos and Malo 2015b).

Table 1. Experimentally recorded mean withdrawal stiffness (units kN/mm) and C.o.V.
lef =100 mm lef =300 mm lef = 450 mm lef = 600 mm
Kw.mean /C.o.V. Kw.mean /C.o.V. Kw.mean /C.o.V. Kw.mean /C.o.V.
α = 0° 54.6 / 0.16 121.0 / 0.30 121.8 / 0.13 128.6 / 0.17
α = 10° 56.0 / 0.27 137.3 / 0.19 132.8 / 0.22 131.1 / 0.05
α = 20° 53.8 / 0.23 125.9 / 0.20 121.7 / 0.16 128.0 / 0.14
α = 30° 42.6 / 0.27 111.2 / 0.11 100.3 / 0.10 114.8 / 0.11
α = 60° 36.6 / 0.33 73.5 / 0.17 90.1 / 0.09 (-)1
α = 90° 29.0 / 0.31 61.4 / 0.11 66.6 / 0.16 (-)1
1
Experiments were not performed for lef = 600mm and α = 60°, 90°

Figure 4. Withdrawal stiffness as function of l ef

105
INTER / 48 - 07 - 02

It is clear from the experimental results that the specimens exhibited high stiffness,
especially for small rod-to-grain angles. As shown in Figure 4, the increase of with-
drawal stiffness due to increasing embedment length becomes gradually smaller as
the embedment length increases. This is estimated both analytically and by numerical
results and validated experimentally. In fact, the experimental results for these
threaded rods suggest that Kw has no correlation with the embedment length if lef ≥
300 mm. This is especially true for small rod-to-grain angles. Finally, according to ex-
perimental observations, no initial slip occurred if the threaded steel coupling parts
of the set-up were tightly fastened.

5.2 Withdrawal strength parameter

The withdrawal strength parameter was calculated for all angles from the experi-
mental results for all specimens. The mean values, the C.o.V., the median and the
5%-fractile characteristic values are provided in Table 2. It should be noted that the
requirements of EN1382 (CEN 1999) for the determination of fax.α have not been met
with respect to the embedment length and the edge distances. The characteristic
values are calculated according to EN14358 (CEN 2006). In comparison to the exper-
imental results presented in the previous Section, some additional experimental re-
sults have been used in Sections 5.2 and 5.3.
Table 2. Values of the withdrawal strength parameter fax.α
fax.α = Fmax / d·lef (MPa)
Number of tests Mean C.o.V. Median 5% - fractile
α = 0° 25 13.81 0.152 13.79 10.19
α = 10° 22 14.14 0.168 13.90 10.06
α = 20° 20 15.70 0.145 16.05 11.46
α = 30° 20 15.16 0.136 15.52 11.47
α = 60° 16 15.17 0.124 15.75 11.50
α = 90° 20 14.88 0.108 15.04 11.92
* Note: the requirements of EN 1382 with respect to lef and the edge distances were not met for all speci-
mens

The variability decreases with increasing angle. The ratio fax.90.k / fax.0.k is equal to 1.17
which is very close to the ratio 1.20 according to Equation (2). Moreover, the with-
drawal strength for rod-to grain angles 0° and 10° is significantly smaller than the
withdrawal strength for greater angles. The experimental results together with the
estimations by Equations (2) and (3) are presented in Figure 5.

106
 INTER / 48 - 07 - 02

Figure 5. Withdrawal strength parameter as function of α

5.3 Withdrawal capacity

All specimens with lef ≤ 450 mm failed due to withdrawal of the rod. In a few speci-
mens with lef = 450 mm yielding of the rod was observed, however the increasing
force due to steel hardening led to withdrawal failure prior to steel fracture. In the
vast majority of the specimens with lef = 600 mm yielding of the rod was observed. All
5 specimens in S20-600 and S30-600 series and 3 out of 5 specimens in S10-600 se-
ries failed due to steel fracture (none in the S0-600 series). These values have been
excluded from the calculation of fax.α in the previous Section. Yielding and steel frac-
ture of the rods occurred at load levels which were significantly higher than those
predicted by the nominal yield and ultimate strength properties of steel. The ob-
served increase in strength of the steel can probably be attributed to steel hardening
due to thread rolling.
The mean experimentally recorded capacities and their C.o.V. as well as the charac-
teristic capacity for all embedment lengths and rod-to-grain angles are summarized
in Table 3. The characteristic capacities have also been calculated according to EN
14358. A minimum C.o.V equal to 0.05 was used to calculate the characteristic capac-
ities, in cases where C.o.V. was smaller.
The experimentally recorded capacities, together with the EC5 and the analytical es-
timations are plotted as function of the embedment length for all rod-to-grain angles
in Figure 6. The withdrawal strength parameter was determined by Equation (2) and

107
INTER / 48 - 07 - 02

by setting fax.90.k = 11.92 MPa (from Table 2). Note that Equation (2) has been used al-
so outside its valid range for α.

Table 3. Experimentally recorded withdrawal capacity for all specimens (in kN)
lef =100 mm lef =300 mm lef = 450 mm lef = 600 mm
(10 tests) (5 tests) (5 tests) (5 tests)
Fax.Rm / C.o.V. / Fax.Rk Fax.Rm / C.o.V. / Fax.Rk Fax.Rm / C.o.V. / Fax.Rk Fax.Rm / C.o.V. / Fax.Rk
α = 0° 26.2 / 0.14 / 19.6 89.7 / 0.12 / 66.8 130.2 / 0.24 / 66.7 161.6 / 0.05 / 141.8
α = 10° 25.8 / 0.18 / 17.9 99.8 / 0.10 / 76.9 127.5 / 0.14 / 88.7 173.11a / (-) / (-)
α = 20° 30.2 / 0.19 / 19.5 98.7 / 0.11 / 74.3 145.8 / 0.06 / 124.7 175.7/0.01/155.31b
α = 30° 27.9 / 0.13 / 20.9 99.9 / 0.11 / 77.4 144.6 / 0.09 / 115.5 176.7/0.01/156.21b
2
α = 60° 28.7 / 0.17 / 18.3 93.6 / 0.12 / 66.9 141.7 / 0.03 / 125.2 (-) 3
α = 90° 28.0 / 0.12 / 21.7 96.5 / 0.07 / 80.8 139.2 / 0.05 / 121.9 (-) 3
1a
Steel and withdrawal failures were observed and thus no characteristic capacity was calculated, 1b Steel
failure, characteristic value
3
calculated with C.o.V = 0.05, 2 6 tests (instead of 10), have been performed for lef
= 100mm and α = 60°, No experiments performed for lef = 600mm and α = 60°, 90.

Figure 6. Withdrawal capacity as function of l ef

108
 INTER / 48 - 07 - 02

As shown in Figure 6, Equation (5) results in a nearly linear relation between the ca-
pacity and the embedment length and thus the difference between Equations (1) and
(5) is small. The estimations by Equations (1) and (5) are generally conservative, es-
pecially for lef ≥ 300 mm and for α ≥ 20°. According to the experimental results, the
withdrawal capacity of specimens with α = 20° was equally reliable as the capacity of
specimens with greater angles. On the other hand, for α < 20° the capacity may be
less reliable like in series S0-450 where the evaluated from experiments characteristic
capacity was smaller than the analytical prediction.
Finally, it has been reported (Ringhofer and Schickhofer 2014) that the long-term be-
haviour of axially loaded screws inserted parallel to the grain is very poor. It follows
that the long-term behaviour of threaded rods (as function of the rod-to-grain angle
and the embedment length) should be further explored.

6 Conclusions

The withdrawal of axially loaded threaded rods with a diameter of 20 mm, screwed
into glulam was studied using experimental and analytical methods. The following
main conclusions are drawn:
• The withdrawal stiffness and capacity can be estimated by use of a simple analyti-
cal procedure, based on the principle of Volkersen model.
• The characteristic withdrawal strength, as estimated by EC5 expression, is on the
safe side especially for rod-to grain angles 20° and 30°.
• The characteristic withdrawal strengths for rod-to grain angles 0° and 10° are sig-
nificantly smaller than the strengths for greater angles.
• The capacity of specimens with a rod-to-grain angle equal to 20° was equally relia-
ble as the capacity of specimens with greater angles.
• Experimental, analytical and numerical results suggest that the increase of with-
drawal stiffness due to increasing embedment length becomes gradually smaller as
the embedment length increases.
• According to experimental observation, initial slip did not occur when the steel
coupling parts of the set-up were tightly fastened.
• Steel fracture of the rods occurred at load levels which were significantly higher
than those predicted by the nominal yield and ultimate strength properties of
steel.

109
INTER / 48 - 07 - 02

7 Acknowledgements

The support by The Research Council of Norway (208052) and The Association of
Norwegian Glulam Producers, Skogtiltaksfondet and the Norwegian Public Road Ad-
ministration is gratefully acknowledged. The authors would like to acknowledge the
contribution of students Joakim Troller and Roland Falk in the preparation of the ex-
periments.

8 References

CEN, European committee for standardization. 1991. EN 26891:1991 (ISO

6891:1983): Timber structures- Joints made with mechanical fasteners-General
principles for the determination of strength and deformation characteristics.
Brussels, Belgium.
CEN, European committee for standardization. 1999. EN 1382-1999: Timber
structures- - Tests methods - Withdrawal capacity of timber fasteners.
Brussels, Belgium.
CEN, European committee for standardization. 2004. EN 1995-1-1:2004: Design of
timber structures. Part 1-1: General-Common rules and rules for buildings.
Brussels, Belgium.
CEN, European committee for standardization. 2006. EN 14358-2006: Timber
structures- Calculation of characteristic 5-percentile values and acceptance
criteria of a samle. Brussels, Belgium.
CEN, European committee for standardization. 2013. EN 14080-2013: Timber
structures- Glued laminated timber and glued solid timber - Requirements.
Brussels, Belgium.
DIBt, Deutsches Institut für Bautechnik. 2010. SFS intec, GmbH., Gewindestangen mit
Holzgewinde als Holzverbindungsmittel, Allgemeine bauaufsichtliche Zulassung
Z-9.1-777.
DIN, Deutsches Institut für Normung. 1975. DIN 7998:Gewinde und Schraubenenden
für Holz-schrauben. Berlin, Germany.
Frese, M. & H. J. Blaß. 2009. Models for the calculation of the withdrawal capacity of
self-tapping screws. In Proceedings of the 42nd CIB-W18 meeting Dübendorf,
Switzerland.
Jensen, J. L., A. Koizumi, T. Sasaki, Y. Tamura & Y. Iijima (2001) Axially loaded glued-in
hardwood dowels. Wood Science and Technology, 35, 73-83.
Jensen, J. L., M. Nakatani, P. Quenneville & B. Walford (2011) A simple unified model
for withdrawal of lag screws and glued-in rods. European Journal of Wood and
Wood Products, 69, 537-544.

110
 INTER / 48 - 07 - 02

Jensen, J. L., M. Nakatani, P. Quenneville & B. Walford (2012) A simplified model for
withdrawal of screws from end-grain of timber. Construction and Building
Materials, 29, 557-563.
Malo, K. A. & P. Ellingsbø. 2010. On connections for timber bridges. In Proceedings of
the International Conference Timber Bridges (ICTB), 297-312. Lillehammer,
Norway.
Mischler, A. & A. Frangi. 2001. Pull-out tests on glued-in-rods at high temperatures.
In Proceedings of the 34th CIB-W18 meeting Venice, Italy.
Mori, T., M. Nakatani, S. Kawahara, T. Shimizu & K. Komatsu. 2008. Influence of the
number of fastener on tensile strength of lagscrewbolted glulam joint. In 10th
World Conference on Timber Engineering, 1100-1107.
Nakatani, M. & K. Komatsu. 2004. Development and verification of theory on pull-out
properties of Lagscrewbolted timber joints. In Proceedings of the 8th World
Conference on Timber Engineering, 95-99. Lahti, Finland.
Pirnbacher, G., R. Brandner & G. Schickhofer. 2009. Base parameters of self-tapping
screws. In Proceedings of the 42nd CIB-W18 meeting Dübendorf, Switzerland.
Ringhofer, A. & G. Schickhofer. 2014. Influencing parameters on the experimental
determination of the withdrawal capacity of self-tapping screws. In
Proceedings of the 13th World Conference on Timber Engineering. Quebec City,
Canada.
Stamatopoulos, H. & K. A. Malo (2015a) Withdrawal capacity of threaded rods
embedded in timber elements. Construction and Building Materials, 94, 387-
397.
Stamatopoulos, H. & K. A. Malo (2015b) Withdrawal stiffness of threaded rods
embedded in timber elements. Submitted to Construction and Building
Materials.
Tomasi, R., A. Crosatti & M. Piazza (2010) Theoretical and experimental analysis of
timber-to-timber joints connected with inclined screws. Construction and
Building Materials, 24, 1560-1571.
Volkersen, O. (1938) Die nietkraftverteilung in zugbeanspruchten nietverbindungen
mit konstanten laschenquerschnitten. Luftfahrtforschung, 15, 41-47.

111
INTER / 48 - 07 - 02

Discussion

The paper was presented by H Stamatopoulos

H Blass commented about the stiffness dependence on the embedment and asked
why FEM predictions were higher than experimental values except for the 90 degrees
case. H Stamatopoulos answered that the rod slipped with the interface more at 0
degrees than with 90 degrees. Therefore, with less relative slip at 90 degrees, there
was better agreement.
S Franke asked how was the load slip evaluated in relation to the strength. H Sta-
matopoulos answered that small embedment length was used only. FEM calculation
for strength would be more difficult as crack formation could be issues that needed to
be considered; therefore, only used for stiffness prediction. S Franke stated that he
has a student working on strength prediction.
P Zarnani and H Stamatopoulos discussed local shear failure mode such as block shear
failure issues.
W Seim asked about the definition of fracture energy and from where these values
were obtained. H Stamatopoulos clarified that he did not use fracture energy and just
used a bilinear constitutive law. E Serrano commented that you needed fw and two
slopes for the bilinear constitutive law; therefore, you have defined the fracture ener-
gy. H Stamatopoulos agreed.
I Smith and H Stamatopoulos discussed progressive collapse and the use of a long rod
to get steel yielding rather than withdrawal.
R Jockwer asked about pull-pull rather than push-pull test configuration. H Stamato-
poulos responded that there was work done and support conditions played an im-
portant role in withdrawal stiffness.

112
113
114
 INTER / 48 - 07 - 03

Load‐carrying capacity of
dowelled connections

H.J. Blass, Karlsruhe Institute of Technology

F. Colling, Augsburg University of Applied Sciences

Keywords: Dowel, yield moment, connection

1 Introduction
The load‐carrying capacity of joints with dowel‐type fasteners in Eurocode 5 (2010) is
mainly based on the Johansen theory (Johansen, 1949), later extended by Meyer
(1957). Even though Johansen’s model is based on plastic hinge formation in the
dowel‐type fasteners for some of the failure modes considered, the elastic bending
moment capacity of the fasteners is used. Eurocode 5 contains an empirical equation
to calculate the fastener yield moment which in many cases results in values between
the elastic and full plastic fastener bending capacity. However, Sandhaas (2012)
showed that for large diameter dowels of high steel grades the predicted yield mo‐
ment according to Eurocode 5 is even lower than the elastic moment capacity.
The introduction of the Johansen theory in the German design code DIN 1052:2004,
being very similar to Eurocode 5, in many cases led to a significant decrease of the
calculated load‐carrying capacity of dowelled joints with drift pins compared to the
design according to the former version DIN 1052:1988. The reason for this apparent
decrease in load‐carrying‐capacity is mainly due to the much more stringent consid‐
eration of the group effect in DIN 1052:2004 using nef for dowels in line with load and
grain direction. A comparison between the 1988 and 2004 versions of DIN 1052 also
revealed that the difference in calculated load‐carrying‐capacity increases with in‐
creasing dowel diameter. These differences motivated the studies described in the
following.
In order to find a more realistic bending moment capacity of dowel‐type fasteners,
the load‐carrying capacity of dowelled joints with drift pins was comprehensively
studied and evaluated, based on 1588 tests with dowelled connections reported in
seven different research studies (Brühl, 2010; Ehlbeck & Werner, 1989; Jorissen,

115
INTER / 48 - 07 - 03

1998; Kneidl, 2009; Mischler, 1998; Sandhaas, 2012; Schmid, 2002). Additionally,
bending and tensile tests with dowels sampled in companies during third party quali‐
ty control visits formed the basis for a more realistic equation for the calculation of
the yield moment My,k.

2 Eurocode 5 versus DIN 1052:1988

Calculated load‐carrying capacities of dowelled connections with drift pins according
to EN 1995‐1‐1:2010 (Eurocode 5, 2010) are in many cases significantly lower than
the corresponding values according to the former German DIN 1052:1988. Conse‐
quently, structures comprising connections designed according to DIN 1052:1988
might be unsafe or the design according to Eurocode 5 might be overly conservative.
Figures 2.1 to 2.4 exemplarily show a comparison between the load‐carrying‐
capacities according to Eurocode 5 and DIN 1052:1988, respectively. The two design
codes are based on different safety concepts: Eurocode 5 uses partial safety factors
for both, actions and resistances while DIN 1052:1988 uses permissible loads for
connections. In order to compare the load‐carrying‐capacities, the following assump‐
tions were made:
 Design actions are calculated by multiplying characteristic actions with a partial
factor of 1.4;
 The design load‐carrying‐capacity of a dowelled joint is calculated for service class
1 or 2 and load‐duration class medium‐term.
Using these assumptions, the permissible load according to DIN 1052:1988 was com‐
pared with the design resistance of the connection, divided by the partial action fac‐
tor of 1.4:
kmod Fv ,Rk
Rcomp    0.44  Fv ,Rk  zul N (1)
 M  G /Q
Here, kmod = 0.8, M = 1.3, G/Q = 1.4, Fv,Rk is the characteristic load‐carrying‐capacity
and zul N is the permissible load of a dowelled connection. In Figures 2.1 to 2.4, the
dowel diameter d, the side and middle member’s slenderness ratios sm and mm
(timber member thickness over dowel diameter) as well as the number of fasteners
nh arranged parallel to the load and grain direction were varied.
If a single dowel is considered, Eurocode 5 results in higher load‐carrying‐capacities
for small diameter dowels and low slenderness ratios (see Fig. 2.1). For larger diame‐
ters and slenderness ratios, Eurocode 5 shows lower load‐carrying‐capacities (see Fig.
2.2). For several dowels arranged in line with the load and grain direction (nh > 1), the
difference between Eurocode 5 and DIN 1052:1988 increases, especially for large di‐
ameter dowels and large slenderness ratios (see Fig. 2.3 and Fig. 2.4).

116
 INTER / 48 - 07 - 03

2,5 2,3
2,11
2,1
2,0 1,9
Rcomp or zul N [kN]

1,7
1,5 1,5

Rcomp / zul N
1,355 1,3
1,0 1,1
0,9
nh = 1
0,5 d = 8 mm EC 5 0,7
mm = 3,0 DIN 1052
0,5
0,0 0,3
0 1 2 3 4 5 6 7 8
sm
Figure 2.1. Rcomp versus zul N; nh = 1, d = 8 mm, middle member slenderness ratio mm = 3,0
35 1,9
1,74
30 1,7
Rcomp or zul N [kN]

25 1,5
1,3 Rcomp / zul N
20
1,1
15
0,9
10 nh = 1 0,7
d = 24 mm 0,721 EC 5
5 0,5
mm = 6,0 DIN 1052
0 0,3
0 1 2 3 4 5 6 7 8
sm
Figure 2.2. Rcomp versus zul N; nh = 1, d = 24 mm, middle member slenderness ratio mm = 6,0
12 1,5
1,39 1,4
10 1,3
Rcomp or zul N [kN]

1,2
8 1,1
1,0
Rcomp / zul N

6 0,9
0,892 0,8
4 0,7
nh = 6 0,6
2 d = 8 mm EC 5 0,5
mm = 3,0 DIN 1052 0,4
0 0,3
0 1 2 3 4 5 6 7 8
sm
Figure 2.3. Rcomp versus zul N; nh = 6, d = 8 mm, middle member slenderness ratio mm = 3,0

117
INTER / 48 - 07 - 03

200 1,15 1,2

180 1,1
160 1,0
Rcomp or zul N [kN]

140 0,9
120

Rcomp / zul N
0,8
100
0,7
80
60 0,6
nh = 6 0,5
40 d = 24 mm
0,474 EC 5
0,4
20 mm = 6,0
DIN 1052
0 0,3
0 1 2 3 4 5 6 7 8
sm
Figure 2.4. Rcomp versus zul N; nh = 6, d = 24 mm, middle member slenderness ratio mm = 6,0

3 Connection test results

3.1 Test specimens
Altogether 1588 tests were evaluated, 1045 timber‐to‐timber and 543 steel‐to‐tim‐
ber connections. The different sources yield the following tests with sufficient infor‐
mation regarding the test configuration, the timber members and the steel proper‐
ties:
 Jorissen: 919 timber‐to‐timber connections, tension and compression;
 Ehlbeck and Werner: 126 timber‐to‐timber connections, tension and compression;
 Kneidl: 58 steel‐to‐timber connections, tension;
 Brühl: 22 steel‐to‐timber connections, tension;
 Mischler: 190 steel‐to‐timber connections, tension;
 Sandhaas: 179 steel‐to‐timber connections, tension;
 Schmid: 94 steel‐to‐timber connections, tension;
The side member slenderness ratios sm varied between 1.0 and 7.5, for timber‐to‐
timber connections most tests were performed with sm < 5. Dowel spacing a1 paral‐
lel to the grain ranged from 3 d to 11 d with most test specimens between 5 d and
7 d.
The predominant dowel diameter used in the tests was 12 mm (see Fig. 3.1 left). The
majority of the dowels were made of steel with lower grades (see Fig. 3.1 right).

118
 INTER / 48 - 07 - 03

Figure 3.1. Used dowel diameters in the test specimens (left) and dowel steel tensile strength in
N/mm² (right)
The arrangement of the dowels parallel (nh) and perpendicular (nn) to the load and
grain direction is given in Fig. 3.2.

Figure 3.2. Number of dowels parallel (nh) and perpendicular (nn) to load and grain direction
Some of the tests were performed with parameters either outside the requirements
of the design codes Eurocode 5 and DIN 1052:1988 or the parameters were quite ex‐
ceptional for practical applications like side member slenderness ratio sm < 2. Con‐
nection tests with hardwood showed significantly higher load‐carrying‐capacities
compared to the expected values from the design codes. Therefore, test specimens
fulfilling one of the following conditions were excluded from the evaluation:
• Spacing parallel to the grain a1 < 5 d,
• Loaded end distance a3,t < 6 d,
• Unloaded edge distance a4,c < 3 d,
• Side member slenderness ratio sm < 2,
• Density  > 600 kg/m³ (hardwood).
Discounting the excluded values, 561 test results with timber‐to‐timber and 325 with
steel‐to‐timber connections remain for the following evaluation.

3.2 Test results versus calculated permissible load according to DIN 1052
This evaluation shows the ratio of the ultimate test load Fv,R versus the calculated
permissible load zul N according to DIN 1052:1988. In the calculation of zul N the
dowel steel strength is not considered, only a minimum steel grade of S235 for dow‐

119
INTER / 48 - 07 - 03

els and 3.6 for bolts is required. Similarly, the strength class of solid or glued laminat‐
ed softwood timber is not accounted for in the calculation. Only for more than six
fasteners parallel to the load and grain direction a reduction of the effective number
of fasteners, nef < nh is taken into account. Figures 3.3 and 3.4 show the ratios
Fv,R/zul N for timber‐to‐timber and steel‐to‐timber connections, respectively. The ra‐
tios were calculated for every single test, the dark red triangles show the ratios for
the excluded test results.

Figure 3.3. Ratios of the ultimate test load Fv,R versus the calculated permissible load zul N
according to DIN 1052:1988 for 1045 timber‐to‐timber connections

Figure 3.4. Ratios of the ultimate test load Fv,R versus the calculated permissible load zul N
according to DIN 1052:1988 for 543 steel‐to‐timber connections
The characteristic ratio calculated according to EN 14358 is 1.79 for timber‐to‐timber
and 1.77 for steel‐to‐timber connections only calculated for the test results not ex‐
cluded. The characteristic ratio corresponds to a global safety factor. Depending on
the service and load‐duration classes, it should be between 2.0 and 2.3. The results
hence show a deficiency of 10% to 25% in the global safety factor for dowelled con‐
nections designed according to DIN 1052:1988.

120
 INTER / 48 - 07 - 03

3.3 Test results versus calculated characteristic load‐carrying‐capacity according

to Eurocode 5
The second evaluation compares the ultimate loads Fv,R in the connection tests to the
calculated characteristic load‐carrying‐capacities Fv,Rk according to Eurocode 5. When
calculating the load‐carrying‐capacities, the factors 1.05 and 1.15 in sections 8.2.2
and 8.2.3 of Eurocode 5 are disregarded, since they only compensate the lower re‐
quired partial factor and the use of the modification factor kmod for the fastener’s
yield moment.
In order to determine the characteristic timber density, the mean density was deter‐
mined for each test series and the associated characteristic density according to EN
338 or EN 14080 was assumed for the test series.
Similarly, a characteristic dowel tensile strength was assumed for dowels, where the
tensile strengths were given in the test report. If only a steel grade of the dowels was
given, corresponding characteristic tensile strength was used. If no information re‐
garding the dowel steel grade was available, the characteristic tensile strength of
steel grade S235 was assumed (fu,k = 360 N/mm²).
An effective number of dowels according to equation (8.34) of Eurocode 5 was used
for connections with several dowels arranged parallel to load and grain direction.
Figures 3.5 and 3.6 show the ratios Fv,R/Fv,Rk for timber‐to‐timber and steel‐to‐timber
connections, respectively. The ratios were calculated for every single test, the dark
red triangles show the ratios for the excluded test results. The characteristic ratio cal‐
culated according to EN 14358 is 1.074 for timber‐to‐timber and 1.073 for steel‐to‐
timber connections only for the test results not excluded. Ideally, the characteristic
ratio would be 1.0 for both cases. The calculation model according to Eurocode 5 is
hence slightly conservative.

Figure 3.5. Ratios of the ultimate test load Fv,R versus the calculated characteristic load‐carrying‐
capacity Fv,Rk according to Eurocode 5 for 1045 timber‐to‐timber connections

121
INTER / 48 - 07 - 03

The ultimate test loads for timber‐to‐timber connections published by Ehlbeck and
Werner (1989) are significantly higher than the calculated characteristic load‐carry‐
ing‐capacities (test series No. 59 and higher). The dowel slenderness ratios in the
tests by Ehlbeck and Werner were significantly larger than those used by Jorissen
(1998).

Figure 3.6. Ratios of the ultimate test load Fv,R versus the calculated characteristic load‐carrying‐
capacity Fv,Rk according to Eurocode 5 for 543 steel‐to‐timber connections
Another tendency observed during the evaluation was that the difference between
the ultimate test load and the calculated characteristic load‐carrying‐capacities in‐
creases with increasing dowel diameter. Obviously, the load‐carrying‐capacity of
connections with large diameter dowels is underestimated by Eurocode 5.

4 Dowel test results

4.1 General
The evaluation of the connection tests in section 3 shows an increasing underestima‐
tion of the characteristic load‐carrying‐capacities for larger dowel diameters. The
same holds for higher dowel slenderness ratios where failure modes including dowel
bending occur and the yield moment of the dowel more and more influences the
load‐carrying‐capacity. Therefore, dowel yield moments were experimentally deter‐
mined for different dowel diameters and different steel grades. In order to check
equation (8.30) of Eurocode 5, dowels were sampled in different timber construction
companies as well as ordered from different suppliers. Altogether 159 dowel tensile
tests in 31 series and 122 dowel bending tests in 38 series were carried out. If possi‐
ble, part of each sample was tested in tension and another part in bending. Long
dowels were cut in half and one half was tested in tension and the other in bending.
Since the variation of test results within a test series was very low, the yield moments
according to EN 409 could be compared to the calculated yield moments according

122
 INTER / 48 - 07 - 03

to equation (8.30) of Eurocode 5 by directly using the tensile strength from the test.
Figure 4.1 exemplarily shows dowels after tensile or bending tests.

Figure 4.1. 16 mm dowels after tensile tests (left) and 8 mm dowels after bending tests (right)

4.2 Yield moment My

The tests showed different moment‐rotation behaviour of steel grades with low and
high tensile strengths, respectively. For mild steel the bending moment still increased
significantly after plastic deformation started. This increase is less pronounced for
higher steel grades (see Fig. 4.2).

Figure 4.2. Bending moment – angle relation for 16 mm dowels made of mild steel (dotted line) and
higher grade steel (solid line)
The yield moment was determined according to EN 409 at a bending angle :
0,44
 2,78 k 
  1    (2)
 fu 
Here, k is the characteristic timber density and fu the dowel tensile strength. Since
the timber density is not known, k = 350 kg/m³ is assumed. Table 1 shows the yield
moments My determined according to EN 409, and the steel tensile strengths fu from
the tensile tests. For comparison the yield moments My according to equation (8.30)

123
INTER / 48 - 07 - 03

of Eurocode 5 on the one hand using the tensile strength Rm,mean of each test series
and on the other hand on the basis of the nominal tensile strength of the dowel fu,k. If
fu,k was unknown, the tensile strength of S235 of 360 N/mm² was assumed.

Table 1. Results of dowel bending tests compared to calculated yield moments according to Euro‐
code 5.
Source Diameter/Length Rm,mean My,EN409,mean My,EC5,Rm,mean My,EC5,fuk
[mm] [N/mm²] [Nm] [Nm] [Nm]
SFS 7/233 584 32 28 26
GH 8/200 593 52 40 24
RB 8/140 662 59 44 24
Rög 8/160 634 56 42 24
Würth 8/115 687 60 46 24
Alberts 10/140 622 106 74 45
Murr 10/210 603 101 72 43
Rie 10/140 607 107 72 43
Würth 10/140 604 102 72 43
AHH 12/180 641 193 123 69
Alberts 12/220 631 184 121 73
Bsch 12/320 712 198 137 69
D 12/400 652 184 125 69
Gei 12/160 717 196 138 69
Gei 12/200 591 136 113 69
Gei 12/240 440 95 84 69
GH 12/200 604 174 116 69
RB 12/200 567 166 109 69
San 12/140 752 202 144 69
Würth 12/200 697 201 134 69
DX 16/200 397 198 161 146
Gei 16/240 535 377 217 146
GH 16/300 540 377 219 146
HO 16/140 446 257 181 146
RB 16/240 742 494 301 146
SF 16/220 542 349 220 146
VK 16/200 414 213 168 146
B 20/420 564 776 408 261
GH 20/300 572 759 414 261
RB 20/240 628 825 455 261
Rie 20/390 483 696 350 261

124
 INTER / 48 - 07 - 03

In order to enable a more realistic calculation of dowel yield moments, an alternative

to equation (8.30) of Eurocode 5 is determined. Here, the different behaviour of steel
dowels made of low or high grade steel (see Fig. 4.2) is taken into account. Those test
results are used to derive an equation to determine the yield moment, where both
tensile and bending tests were carried out with dowels from the same batch. The
best agreement between test results and calculated values was found for the follow‐
ing expression, representing the mechanically correct full plastic bending moment of
a circular cross‐section:
fy ,ef  d 3
My  (3)
6
 0 ,9  ( fy  fu )
 for fu  450 MPa
fy ,ef  2 (4)
0 ,9  fu for fu  450 MPa
Here, d is the dowel diameter, fy is the fastener yield strength and fu is the fastener
tensile strength.
Fig. 4.3 left shows the ratio between My according to EN 409 and the calculated value
according to equation (3) for the 122 bending tests, on the one hand based on the
mean tensile strength from the tests (diamonds) and on the other hand based on the
nominal dowel tensile strength (squares). The ratio is independent of the dowel di‐
ameter. The average ratio for test based tensile strengths is 1.09, the characteristic
ratio is 1.00. Equation (3) hence provides an excellent description of the dowel yield
moments according to EN 409. Since in a real design situation nominal rather than
real tensile strength values are applied, the proposed equation (3) is conservative in
most cases due to the over‐strength of the steel dowels.

Figure 4.3. Ratio between My according to EN 409 and My according to equation (3) (left) or My
according to Eurocode 5 (right)
For comparison the ratio between My according to EN 409 and the calculated value
according to equation (8.30) of Eurocode 5 is shown in Figure 4.3 (right). It is obvious
that Eurocode 5 is increasingly conservative for larger dowel diameters.

125
INTER / 48 - 07 - 03

4.3 Influence of yield moment My on calculated results

Figures 4.4 and 4.5 again show the ratios Fv,R/Fv,Rk for timber‐to‐timber and steel‐to‐
timber connections, respectively. The ratios were calculated using equation (3) in‐
stead of equation (8.30) of Eurocode 5 to calculate the characteristic yield moment of
the dowels. The characteristic ratio calculated according to EN 14358 decreases from
1.074 to 1.048 for timber‐to‐timber and from 1.073 to 1,001 for steel‐to‐timber con‐
nections, again only for the test results not excluded. The calculation model accord‐
ing to Eurocode 5 with the modified yield moment My hence is still slightly conserva‐
tive for the tested timber‐to‐timber connections and appropriate for the tested steel‐
to‐timber connections.

Figure 4.4. Ratios of the ultimate test load Fv,R versus the calculated characteristic load‐carrying‐
capacity Fv,Rk taking into account My according to equation (3) for 1045 timber‐to‐
timber connections

Figure 4.5. Ratios of the ultimate test load Fv,R versus the calculated characteristic load‐carrying‐
capacity Fv,Rk taking into account My according to equation (3) for 543 steel‐to‐timber
connections
In the average, the timber‐to‐timber connections tested by Ehlbeck and Werner
(1989) still show higher ratios Fv,R/Fv,Rk even with the modified equation for the dowel
yield moment My (see test series 59 through 118 in Fig. 4.4). Apart from the plastic
dowel bending capacity there seem to exist further causes for higher ratios with in‐
creasing dowel slenderness ratios. If a slenderness effect is taken into account for
dowelled connections similar to the rope effect in Eurocode 5, leading to an increase
of 25 % of the lateral load‐carrying‐capacity of dowels with a failure mode showing

126
 INTER / 48 - 07 - 03

two plastic hinges per shear plane, the characteristic ratio for timber‐to‐timber con‐
nections would only drop to from 1,048 to 1,037, for steel‐to‐timber connections
from 1,001 to 0,978.
Reasons for the additional safety margin for slender dowels could be friction between
the dowel and the surrounding timber along the length of the dowel, especially in ar‐
eas where the embedding strength is reached. This friction would create a withdraw‐
al capacity leading to a twofold rope effect: friction between the timber or steel
members and the fastener tensile component parallel to the shear plane. Further re‐
search is required to quantify this possible rope effect in dowelled connections with
drift pins.

5 Conclusions
The load‐carrying capacity of dowelled joints with drift pins was comprehensively
studied and evaluated, based on 1588 tests with dowelled connections reported in
seven different research studies (Brühl, 2010; Ehlbeck & Werner, 1989; Jorissen,
1998; Kneidl, 2009; Mischler, 1998; Sandhaas, 2012; Schmid, 2002).
The analysis of the short‐term tests shows an overestimation of the load‐carrying ca‐
pacity according to DIN 1052:1988 by 10 – 25 %. Consequently, connections designed
according to DIN 1052:1988 are below the reliability level required today. The evalu‐
ations also show that some load‐carrying capacities according to Eurocode 5 are con‐
servative and hence could be increased accordingly.
Based on bending and tensile tests with dowels sampled in companies during third
party quality control visits, a modified equation for the calculation of the yield mo‐
ment My,k was derived, leading to higher calculated load‐carrying capacities especially
for large diameter dowels or higher steel grades. The dowel bending and tensile tests
also revealed that actual steel strength values often show significant over‐strength.
For dowelled connections with a failure mode showing two plastic hinges per shear
plane, an additional slenderness effect was observed, increasing the load‐carrying
capacity of these connections in the order of 25 % compared to calculated values
based on the Johansen model. This is surprising, since drift pins so far show no signif‐
icant withdrawal capacity and hence a rope effect is hardly to be expected.
The design rules in DIN 1052:1988 were originally derived based on tests, where the
dowel steel strength was not determined. This means that both effects mentioned
above, namely the surplus strength of the steel dowels and the slenderness effect,
were implicitly included in the permissible loads according to DIN 1052:1988.
Considering the consequences of these findings (modified equation for My, slender‐
ness effect and steel over strength), the existing differences between the calculated
load‐carrying capacities according to DIN 1052:1988 and Eurocode 5, respectively,
may be explained to a large extent.

127
INTER / 48 - 07 - 03

A new equation for Eurocode 5 for calculating the characteristic yield moment of
bolts and dowels is proposed.

6 References
Brühl, F (2010): Ductile timber connections (in German). Research report, Universität
Stuttgart, Stuttgart.
Colling, F & Blass, HJ (2014): Load‐carrying‐capacity of dowelled connections (in Ger‐
man). In: Proceedings, Karlsruher Tage 2014 ‐ Holzbau: Forschung für die Praxis.
KIT Scientific Publishing, Karlsruhe.
Ehlbeck, J & Werner, H (1989): Load‐slip behaviour of dowels in glued laminated tim‐
ber and solid timber of different species considering different dowel arrangements
(in German). Research report, Universität Fridericiana Karlsruhe, Karlsruhe.
Eurocode 5 (2004): Design of timber structures ‐ Part 1‐1: General and rules for build‐
ings. CEN. (EN 1995‐1‐1).
Johansen, KW (1949): Theory of Timber Connections. IABSE publications 9 (1949), Zü‐
rich, Switzerland.
Jorissen, A (1998): Double shear timber connections with dowel type fasteners. Dis‐
sertation, Delft University Press, Delft.
Kneidl, R (2009): Final report regarding experimental studies of dowelled connections
(in German). Bayrische Ingenieurkammer Bau, München.
Meyer, A (1957): Load‐carrying‐capacity of nailed joints under static load (in Ger‐
man). Holz als Roh‐ und Werkstoff 15 Heft 2.
Mischler, A (1998): Relevance of ductility for the load‐slip behaviour of bolted steel‐
to‐timber joints. Dissertation, ETH Zürich, Zürich.
Sandhaas, C (2012): Mechanical behaviour of timber joints with slotted‐in steel
plates. Dissertation, Delft University Press, Delft.
Schmid, M (2002): Application of fracture mechanics on timber connections. Disser‐
tation (in German), Universität Fridericiana Karlsruhe, Karlsruhe.

128
 INTER / 48 - 07 - 03

Discussion

The paper was presented by H Blass

K Malo asked whether the approach is valid for stainless steel. H Blass responded yes.
S Franke and H Blass discussed about the fitting process for screws are more difficult.
A Salenikovich asked for comments for multiple fasteners in a row. H Blass responded
that EC5 equations were used.
S Franke commented that the attempt was to justify changes to EC5. H Blass re-
sponded that the old allowable values were not based on tests of steel strength,
therefore over-strength situations were not correctly considered. Here the old code is
still non-conservative by ~ 10% but not 25% as previously thought.
V Rajčić and H Blass discussed the lack of conservatism of the old code when different
failure cases were considered.
R Jockwer commented the yield strength of the dowels were very important. H Blass
responded that high strength steel dowel compared to mild steel would still be more
beneficial although it would be dependent on cost and economics.
I Smith commented that this is a manifestation of system effect.
U Kuhlmann commented about target failure mode in relationship to the type steel
used.

129
130
 INTER / 48 - 07 - 04

Evaluation of the reliability of design

approaches for connections
perpendicular to the grain
Robert Jockwer
ETH Zürich, Institute of Structural Engineering, Zurich, Switzerland
René Steiger
Empa, Swiss Federal Laboratories for Materials Science and Technology,
Dübendorf, Switzerland
Andrea Frangi
ETH Zürich, Institute of Structural Engineering, Zurich, Switzerland

KEYWORDS: Connection perpendicular to grain, brittle failure, fasteners, partial safety factor
ABSTRACT: In this paper different design approaches for connections loaded perpendic-
ular to the grain are evaluated with regard to their reliability. The structural behaviour of
connections loaded perpendicular to the grain is described based on existing experimental
and theoretical studies. The detailed failure behaviour of dowel connections loaded per-
pendicular to grain with slotted in steel plates is analysed in a recent test series. Based on
these observations and on a large number of test results from literature different design
approaches are benchmarked. The design and characteristic values of the relevant material
parameters and the partial safety factors for the design of connections loaded perpendicular
to grain are determined in a reliability analysis. As a result a much lower design value of the
material parameter used in the EC5 design equation for connections loaded perpendicular
to grain is proposed.

1 Introduction
1.1 General

In connections global failure can occur either due to local failure of the fastening elements
and the timber surrounding the fasteners or due to failure in the timber next to the joints.
Failure of the fasteners can be accounted for by designing them according to appropriate
design rules, e.g. the so called European yield model (for dowel type fasteners). Design to
prevent this failure mechanism includes prevention of embedment timber failure or of fail-
ure of the metallic fastener. However, sometimes such a design is not sufficient because
splitting of the timber next to the joint might occur and hence design has also account for
this failure. For timber members loaded parallel to the grain spacing requirements are spec-
ified in Eurocode 5 (2004) (EC5). In EC5 there is also a design approach for situations where
connections are loaded perpendicular to the grain. It is independent from the load-carrying
capacity of the fasteners. Alternative design approaches to be applied for the design of con-
nections loaded perpendicular to grain are described in Chapter 2. The aim of this paper is
to evaluate different approaches for the design of single connections in the span of beams
loaded perpendicular to the grain with regard to their reliability and simplicity.

131
INTER / 48 - 07 - 04

1.2 Geometrical properties of connections loaded perpendicular to the grain

b b

ar t t t t

n h
hm
αh
m

l1 F90 l2 F90/2 F90/2 F90

l

Figure 1: Definition at connections loaded perpendicular to grain.

The geometrical properties and definitions of a connection loaded perpendicular to the

grain are depicted in Fig. 1. The level of tensile stresses perpendicular to the grain depends
amongst others on the distance between loaded edge of the member and the top row of
fasteners of the joint. This distance, expressed as a portion α of the full member height is
one of the parameters with the highest impact on the reduction of the load-carrying capac-
ity in connections loaded perpendicular to the grain. From fracture mechanics analysis (e.g.
Leicester (1973)) and from Weibull’s weakest link theory applied to tension perpendicular to
grain (e.g. Mistler (1998)) it is known that the strength of the timber volume surrounding a
connection loaded perpendicular to the grain decreases with increasing beam height h. The
impact of the beam width b on the load-carrying capacity is often assumed to be linear. Ad-
ditional geometrical parameters are hn the distance between the nth -row of fasteners to the
loaded edge and ll the distance between multiple connections along the beam axis.

1.3 Types of connections

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

Figure 2: Examples for connections loaded perpendicular to grain: rafter to purlin connection (a), joist
hanger (b), bolted steel-to-timber connection (c), glued in rods (d) and punched metal plate fasteners (e).

Due to the low strength and brittle failure mechanism of the timber, as a general rule situa-
tions where timber is subjected to tension perpendicular to the grain should be avoided in
general. Hence, it is often poor design when connections transfer high tensile stresses per-
pendicular to the grain into timber elements. However, there are certain situations where
this cannot be avoided. Examples of situations where connections introduce loads perpen-
dicular to the grain are given in Fig. 2:

132
 INTER / 48 - 07 - 04

- Nailed connections with sheet metal plates for hold downs between rafter and perlins
(Fig. 2 (a)) or for joist hangers for main and secondary beam connections (Fig. 2 (b)).
These connections are often single sided with a large number of fasteners of relatively
small diameter.
- Doweled or bolted connections between timber-to-timber or steel-to-timber connec-
tions (Fig. 2 (c)). Bolts are often used for external steel plates and dowels are used for
internal slotted-in steel plates.
- Screwed or glued-in rod connections are used for hanging loads (Fig. 2 (d)). Screws are
also used for steel-to-timber connections.
- Other examples are shear connections by means of ring connectors, punched in metal
steel plates (Fig. 2 (e)) or special products like e.g. Sherpa R connectors.

2 Design approaches
Failure of connections loaded perpendicular to the grain due to splitting has often been dis-
cussed in literature. A comprehensive study with a review of various design approaches and
test results was made by Schoenmakers (2010). Existing design approaches are based either
on stress criteria like the approach in the former DIN 1052 (2008) as proposed by Ehlbeck
et al. (1989) and the approach in Lignum (1990) or on fracture mechanics theory like e.g.
the design approach in EC5 developed by van der Put (1990). The linear elastic fracture me-
chanics (LEFM) approach developed by van der Put (1990) was extended and adapted e.g.
based on quasi-nonlinear fracture mechanics (Jensen et al., 2003) or semi-empirical theories
(Ballerini, 2004). Mixed mode failure in tension perpendicular to the grain and shear was
accounted for in the studies by Franke and Quenneville (2011). Empirical based design ap-
proaches are presented by Quenneville and Mohammad (2001) and Lehoux and Quenneville
(2004) based on the design of rivet connections.
The impact of the parameters beam height h, relative connection height α , number of fas-
tener columns m and number of fastener rows n is graphed in Fig. 3. Overall the approaches
show a similar behaviour with different intensity of the specific impact of the parameters.
The beam height h, the relative connection height α and the beam width b are accounted
for by all approaches but not the geometry of the connections. The respective terms for the
number of rows n and number of columns m of fasteners in a connection are summarized
in Table 1.

Table 1: Terms accounting for the geometry of the connection.

Parameter Ehlbeck et al. (1989) Lignum (1990) nBallerini (2004)

 o Franke & Quennev. 2011
max 1; 0.7 + 1.6 ahr (1 + ar /(αh))0.1 min 2.2; 1 + 0.75 ar h+ll

m (width) X

1
 2
h1 0.4
κ
1 + 1.75 1+κ 0.1 + arctan(n)0.6
n (height) n ∑ni=1 hi
hm nhm
with κ = 1000 for n > 1

As can be seen from Tab. 2 the type and diameter of the fasteners is only accounted for
by the design approaches in former DIN 1052 based on Ehlbeck et al. (1989) and in Lignum
(1990). The effective beam width be f is reduced in these approaches in order to account for
early splitting in the vicinity of fasteners with small penetration depth t or high slenderness.
Also Zarnani and Quenneville (2013) proposed to differentiate between full and partial split-
ting of the beam and account for the effective embedment depth of fasteners. For partial
splitting the load-carrying capacity of the fasteners is relevant for the design whereas split-
ting failure of the surrounding timber is decisive for full splitting.

133
INTER / 48 - 07 - 04

1.8 2

1.6 1.8
F90,R,i/F90,R,i(h = 600mm) [-]

1.6

F90,R,i/F90,R,i(α = 0.5) [-]

1.4
1.4
1.2
1.2
1
1
0.8
0.8 van der Put
0.6 van der Put Ehlbeck et al.
0.6
Ehlbeck et al. Lignum
0.4 0.4
Lignum Jensen et al.
0.2 Jensen et al. 0.2 Ballerini
Franke & Quenneville Franke & Quenneville
0 0
0 200 400 600 800 1000 1200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
h [mm] α [-]
(a) (b)

2 4
van der Put
1.8 3.5 Ehlbeck et al.
1.6 Lignum
F90,R,i/F90,R,i(m = 1) [-]

F90,R,i/F90,R,i(n = 1) [-]

3 Ballerini
1.4 Franke & Quenneville
1.2 2.5

1 2
0.8 1.5
0.6 van der Put
Ehlbeck et al. 1
0.4 Lignum
0.2 Ballerini 0.5
Franke & Quenneville
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
m [-] n [-]
(c) (d)

Figure 3: Impact of beam height h (a), relative connection height α (b), number of fastener columns m (c)
and number of fastener rows n (d) on the relative load-carrying capacities F90,R,i /F90,R,i (re f ).

Table 2: Impact of the type and diameter of the fastener on the effective beam width be f .

two-sided connections
DIN 1052 (2008) Lignum (1990)
Nails be f = min{ b ; 2t ; 24d (Timber / Timber) } be f = min{ b ; 2t ; 24d }
30d (Steel / Timber)
Screws be f = min{ b ; 2t ; 24d (Timber / Timber) }
Dowels be f = min{ b ; 2t ; 12d } be f = min{ b ; 6d }
Shear connectors be f = min{ b; 100 mm } be f = 0.6d (ring connectors)
d (toothed plate connector)
Glued in rods be f = min{ b; 6d }

one-sided connections
DIN 1052 (2008) Lignum (1990)
Nails be f = min{ b;t ; 12d (Timber / Timber) } be f = min{ b ; t ; 12d }
15d (Steel / Timber) }
Screws be f = min{ b;t ; 12d (Timber / Timber) }
Dowels be f = min{ b;t ; 6d }
Shear connectors be f = min{ b; 50 mm } be f = 0.3d (ring connectors)
0.5d (toothed plate connector)

134
 INTER / 48 - 07 - 04

3 Experimental data
3.1 Experiments reported in literature

Test results from literature which were used in this paper to evaluate the design approaches
for connections loaded perpendicular to the grain are summarized in Tab. 3. Only tests
on glulam beams and with a single connection in the span were used in this evaluation.
The most common type of connection exposed to experiments was a bolted or doweled
connection with external steel plates.

Table 3: Experimental data from literature used for the evaluation of design approaches for connections
loaded perpendicular to the grain

Literatur Number Species Fastener type

Möhler and Siebert (1981) 28 Softwood Nails, dowels, ring connectors
Ehlbeck and Görlacher (1983) 41 Softwood Nails
Ballerini (1999) 49 Norway Spruce Dowels
Ballerini and Giovanella (2003) 72 Norway Spruce Dowels
Reshke (1999) 138 Spruce/Pine, GL 20f-E Bolts
Kasim (2002) 90 Spruce/Pine & Douglas Fir Bolts
Habkirk (2006) 50 Spruce/Pine & Douglas Fir Bolts
Jensen and Quenneville (2011) 18 Douglas Fir, GL8 + GL15 Dowels
Schoenmakers (2010) 59 Spruce Nails, dowels

120 120 120

188 164 197 125 140
100 100 100
Number of tests

80
Number of tests

Number of tests

80 80

60 60 60

40 40 40

20 20 20

0 0 0
0 200 400 600 800 1000 1200 0 25 50 75 100 125 150 175 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
h [mm] b [mm] α [-]

Figure 4: Histograms of beam and connection parameters height h, width b and relative connection height
α as studied in the publications listed in Tab. 3.

Dimensions of the beams used in the tests and the geometry and relative connection height
as studied in the publications listed in Tab. 3 are shown in Fig. 4. The suitability of the existing
studies for the evaluation of the design of connections loaded perpendicular to the grain in
practice can be summarized:
- The majority of the tests has been carried out on small sized specimen with small
height and width (Fig. 4). This has to be accounted for when evaluating the size ef-
fects.
- A large number of tests has been performed on bolted steel-wood-steel connections.
The reason for the choice of these connections might be the convenience in the exper-
imental assembly.
- In many of the tests the diameter of the fastener was relatively large compared to
the width of the beams. This very stiff configuration with thick fasteners allows for an
uniform loading over the entire beam width. However, the load-carrying capacity of
the fasteners was often much higher compared to the load observed at brittle failure
of the timber. Due to economic reasons it should be aimed for an equal load-carrying
capacity of the fasteners and the timber in practice .

135
INTER / 48 - 07 - 04

3.2 Recent tests at ETH Zurich

A test series recently carried out at ETH Zurich was focused on the evaluation of the failure
behaviour of connections loaded perpendicular to the grain and on the assessment of the
crack growth. Doweled connections with slotted-in steel plates inserted at midspan position
in glulam beams of grade GL 24h with h = 440 mm, b = 140 mm and a span of l = 2 m were
tested. The thickness of the slotted-in steel plate was 10 mm. Smooth dowels with diameter
d = 12 mm were used. The yield moment of the dowels was determined in 3-point bending
tests with My,mean ≈ 190 Nm. The estimated load-carrying capacity of connections loaded
perpendicular to the grain consisting of 8 dowels is Fv,mean ≈ 114 kN for the failure mode 2
according to the EYM.
In the test series three different relative connection heights α = 0.6, 0.7 and 0.8 were studied.
Width and height of the connection was chosen as m × n = 4 × 2 and 2 × 4. Six different
configurations with 4 replicates each result in a total of 24 individual tests. 4 specimens with
α = 0.6 were reinforced against splitting by means of a total of 4 SFS WR-T 13 self-tapping
fully threaded screws inserted 50 mm besides the external columns of fasteners.
The applied load was measured by means of oil pressure at the manually powered pump.
The deflection in midspan of the beam, pull-out deformation of the steel plate and crack
opening were measured by means of inductive measurements and LVDT. The deformation
of the timber surface was determined by means of digital image correlation measurements.

240 unreinforced 240 240

reinforced
220 220 220
crack initiation load [kN]

200
ultimate load [kN]

200 200
failure load [kN]

180 180 180

160 160 160

140 140 140

120 120 120

100 100 unreinforced 100 unreinforced

reinforced reinforced
80 80 80
0.6 0.7 0.8 0.6 0.7 0.8 0.6 0.7 0.8
α [-] α [-] α [-]
(a) (b) (c)

Figure 5: Loads at crack initiation (a), failure with unstable crack growth (b) and ultimate load (c) in the
tests.
250 250 250
225 225 225
200 200 200
175 175 175
Load [kN]

Load [kN]

Load [kN]

150 150 150

125 125 125
100 100 100
75 75 75
50 50 50
25 α = 0.6, m = 4, n = 2 25 α = 0.7, m = 4, n = 2 25 α = 0.8, m = 4, n = 2
α = 0.6, m = 2, n = 4 α = 0.7, m = 2, n = 4 α = 0.8, m = 2, n = 4
0 0 0
0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25
Steel pull out deformation [mm] Steel pull out deformation [mm] Steel pull out deformation [mm]
(a) (b) (c)

Figure 6: Pull-out deformation of the slotted-in steel plates for relative connection heights α = 0.6 (a),
α = 0.7 (b) and α = 0.8 (c).

In Fig. 5 the ultimate loads for specific failure events observed in the tests are given. It can
be distinguished between load at crack initiation, unstable crack growth and ultimate load.
Connections with a small relative connection height failed in a very brittle way. With increas-

136
 INTER / 48 - 07 - 04

ing relative connection height further loading after crack initiation was possible. Beams with
α = 0.8 showed a relatively stable crack growth and reached even higher capacities after
separation of the upper and lower part of the beam.
The deformation of the slotted-in steel plates in relation to the timber is shown in Fig. 6. It
can be seen that with increasing relative connection height a pronounced deformation of the
connection occurs. Such ductile behaviour should be aimed at when designing connections
loaded perpendicular to the grain for applications in practice. For α = 0.6 failure occurs
already at small relative deformations with a brittle failure behaviour. It can be concluded,
that the load-carrying capacity of the fasteners is too high compared to the load-carrying
capacity perpendicular to the grain of the timber in the vicinity of the connection.

250 250 250

225 225 225
200 200 200
175 175 175
Load [kN]

Load [kN]

Load [kN]
150 150 150
125 125 125
100 100 100
75 75 75
50 50 50
25 α = 0.6, m = 4, n = 2 25 α = 0.7, m = 4, n = 2 25 α = 0.8, m = 4, n = 2
α = 0.6, m = 2, n = 4 α = 0.7, m = 2, n = 4 α = 0.8, m = 2, n = 4
0 0 0
−1 0 1 2 3 4 5 6 7 8 9 10 −1 0 1 2 3 4 5 6 7 8 9 10 −1 0 1 2 3 4 5 6 7 8 9 10
Crack opening [mm] Crack opening [mm] Crack opening [mm]
(a) (b) (c)

Figure 7: Crack opening in the center of the connection for relative connection heights α = 0.6 (a), α = 0.7
(b) and α = 0.8 (c).
200 200 200
180 180 180
160 160 160
140 140 140
Load [kN]

Load [kN]

Load [kN]

120 120 120

100 100 100
80 80 80
60 60 60
40 40 40
20 α = 0.6, m = 4, n = 2 20 α = 0.7, m = 4, n = 2 20 α = 0.8, m = 4, n = 2
α = 0.6, m = 2, n = 4 α = 0.7, m = 2, n = 4 α = 0.8, m = 2, n = 4
0 0 0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Crack length [mm] Crack length [mm] Crack length [mm]
(a) (b) (c)

Figure 8: Development of the crack length in dependency of the applied load for relative connection heights
α = 0.6 (a), α = 0.7 (b) and α = 0.8 (c).

By evaluating the crack opening and crack length development (Fig. 7 and 8) a distinction
between stable and unstable crack growth is possible. The crack opening was measured by
means of LVDTs in the centre of the connection between two points approximately 25 mm
above and below the top row of fasteners. The crack length was computed from the defor-
mation measurements on the surface of the beam by means of 2D digital image correlation.
It can be seen that crack initiation starts at a similar load level irrespective of the relative
connection height. This load level corresponds approximately to the estimated load-carrying
capacity of the fasteners according to EYM. Load levels for further crack growth and crack
opening differ between the test series. For α = 0.6 unstable crack growth occurs already
at crack initiation with no potential of an increase in load. For α = 0.7 and 0.8 continued
loading is possible and a more stable crack growth can be observed before ultimate failure.
The observed different degree in brittleness of the failure mechanism could serve as a basis
for defining partial safety factors for design.

137
INTER / 48 - 07 - 04

4 Evaluation of selected design approaches

Best fit between design approaches (Chapter 2) and test data (Tab. 3) is achieved when ap-
plying the approach by Ballerini (2004). The approach by Ehlbeck et al. (1989) shows a good
match with test data as well. The relatively simple approach by van der Put (1990) exhibits a
worse agreement with experimental data.
For a more detailed evalution of the design approaches the value of the material strength
parameter C1,i can be back-calculated from the test results according to Eq. 1 and 2 for the
design approaches by van der Put and Ballerini, respectively.

F
C1,vanderPut = q90 (1)
αh
2b (1−α)

F90
C1,Ballerini = r (2)
αh
2b fw fr
(1−α 3 )

The theoretical value of C1 can be calculated from the mode 1 fracture energy G f ,I and the
shear modulus Gv according to Eq. 3.

r
3G f ,I Gv
C1,T heory = (3)
5
In this study the following values were used: mode 1 fracture energy G f ,I,mean = 0.3 N/mm
with CoV = 30%, shear modulus Gv,mean = 650 N/mm2 with CoV = 13%.
The approach by Ehlbeck et al. is evaluated with regard to the tensile strength perpendicular
to the grain back-calculated according to Eq. 4.

F90
ft,90,Ehlbeck =
0.8 (4)
ks kr (6.5 + 18α 2 ) be f h

35 35

30 30
C 1,vanderP ut [N/mm1.5]

C 1,Ballerini [N/mm1.5]

25 25

20 20

15 15

10 10

5 5

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
d/b [-] d/b [-]

(a) (b)

Figure 9: Dependency of C1,i on the relative fastener diameter d/b.

138
 INTER / 48 - 07 - 04

The mean coefficient of variation of the empirical value of material strength parameter C1,i
and the tensile strength perpendicular to the grain ft,90,Ehlbeck , respectively, are in the order
of CoV ≈ 24% (Ballerini), CoV ≈ 30% (van der Put) and CoV ≈ 27% (Ehlbeck et al.). The C1
value back-calculated from the approach proposed by van der Put markedly depends on the
relative fastener diameter d/b (Fig. 9 (a)). This approach is based on a situation represented
by a beam with a crack over the entire beam width starting at a single dowel connection. For
a larger number of dowels or nails with small diameter this approach leads to a conservative
estimate. For the approach by Ballerini and by Ehlbeck et al. no such dependency can be
found (Fig. 9 (b)). These approaches account for the effect of the geometry of the connection
(Fig. 3 (c) and (d)).
The material parameter C1,T heory has a higher mean value and shows a lower variation than
the values C1,i defined by Eqs. 1 and 2. The bias and increased variation of C1,i defined by
Eqs. 1 and 2 can be accounted for by multiplying C1,T heory in Eq. 3 with a model uncertainty X
with the distribution characteristics given in Tab. 4. It can be seen, that the model uncertainty
XBallerini according to the approach suggested by Ballerini is larger but the resulting coeffi-
cient of variation (CoV ) of the model uncertainty is lower than according to the approach
suggested by van der Put.

Table 4: Distribution characteristics of the model uncertainty X for multiplication with the theoretical value
of material parameter C1,T heory to fit the back-calculated values from the approaches proposed by van der
Put and Ballerini for the test data in Tab. 3.

Model uncertainty Distribution Mean CoV

Xvan der Put Lognormal 0.94 29%
XBallerini Lognormal 0.85 18%

5 Reliability of the evaluated design approaches

5.1 Actions and resistances

The reliability of the approaches for the design of connections loaded perpendicular to the
grain can be evaluated as desribed in Kohler et al. (2012) or Jockwer et al. (2011). In the
general design equation (Eq. 5) the characteristic values of resistance R and permanent (G)
and variable (Q) actions are factored by partial safety factors γi in order to guarantee for the
desired reliability of the design.

kmod · Rk
z − γG Gk − γQ Qk = 0 (5)
γm
A modification factor kmod = 1 is chosen in this study which should represent a short term
action and a connection applied in service class 1 condition. Information about duration of
load effects and influence of variations in moisture content can be found e.g. in Gustafsson
and Larsen (2001) and Vesa and Kevarinmäki (2001). The parameter z is chosen during the
design of the structure in order to fulfill Eq. 5.
The partial safety factors for permanent and variable loads are set as defined in EN 1990
(2005) γG = 1.35 and γQ = 1.5, respectively. The partial safety factor for the resistance of
connections is suggested γm = 1.3 in EC5 (2004). This values is generally valid for all types
of connections with brittle or ductile failure. The calibration of this factor has partly been
based on studies of the bending capacity beams as described in (Sørensen, 2002). Other
design situations, especially when being represented by different distribution functions or

139
INTER / 48 - 07 - 04

exhibiting different variation, result in different partial safety factors. Kohler et al. (2012)
suggest a considerably higher partial safety factors γm ≈ 3 for the design situation of tension
perpendicular to the grain. This supports the need for an evaluation of the reliability of
the design of connections loaded perpendicular to the grain where high tensile stresses
perpendicular to the grain occur. A careful evaluation of the adequacy of the 5th -percentile
value as the characteristic value of material strength or the proposal of new partial safety
factors γm are necessary.
For the assessment of the reliability of the design approaches evaluated above the resis-
tance R is represented by the material parameters C1 and the tensile strength perpendicular
to the grain ft,90 , respectively. In Fig. 10 (a) cumulative distributions of C1,i are given accord-
ing to Eq. 1 and 2 based on the test results summarized in Tab. 3. In addition the tensile
strength perpendicular to the grain according to Eq. 4 is graphed in Fig. 10 (b). A Lognor-
mal distribution shows the best fit with the test results. However, the Weibull distribution
is suggested by JCSS (2001) for the representation of tensile strength perpendicular to the
grain. For the reliability study of the design approaches the parameters C1,i and ft,90 were
represented by distribution functions as given in Table 5. The distribution characteristics of
permanent and variable loads are summarized in Tab. 6.

1 1
0.9 0.9
Cumulative distribution [-]
Cumulative distribution [-]

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
van der Put
0.2 Lognormal 0.2
Ballerini Ehlbeck et al.
0.1 Lognormal 0.1 Lognormal
Theory Weibull
0 0
0 5 10 15 20 25 30 35 0 0.2 0.4 0.6 0.8 1 1.2 1.4
C 1 [N/mm1.5] (a)
ft,90 [N/mm2 ] (b)

Figure 10: Cumulative distribution of the material parameters C1,i according to van der Put (Eq. 1) and
Ballerini (Eq. 2) with the corresponding fitted Lognormal distributions and according to theory (Eq. 3) (a)
and tensile strength perpendicular to the grain according to the approach by Ehlbeck et al. (Eq. 4) with
fitted Lognormal and Weibull distribution (b).

Table 5: Distribution parameters and functions for material property value C1 and ft,90 determined from
test data for the approaches by van der Put, Ballerini and Ehlbeck et al.

Approach Parameter Unit Distr. function Mean CoV

van der Put C1 [N/mm1.5 ] Lognormal 16.7 30.6%
Ballerini C1 [N/mm1.5 ] Lognormal 15.2 21.8%
Ehlbeck et al. ft,90 [N/mm2 ] Lognormal 0.60 26.5%
Ehlbeck et al. ft,90 [N/mm2 ] Weibull 0.60 28.3%

Table 6: Distribution characteristics for loading types according to JCSS (2001) and partial factors.

Load type Distribution CoV char. level γ

Self weight Normal 10% 50% 1.35
Live load Gamma 53% 98% 1.5

140
 INTER / 48 - 07 - 04

5.2 Reliability analysis

The limit state function can be written as:

g = z·R−G−Q (6)

In this function the resistance R can be represented by the random variables of the material
parameters C1 and ft,90 , respectively, with the distribution characteristics summarised in
Tab. 5. The parameter z will be chosen in the design procedure such that the desired failure
probability of Pf ≤ 10−5 ( βr = 4.2, for general situations) is not exceeded.

Pf (g ≤ 0) = Pf (z · R − G − Q ≤ 0) ≤ 10−5 (7)

The fraction of variable load Qmean / (Gmean + Qmean ) can cover a range between 0 and 1. For
a value of 0.8 as proposed in (Kohler et al., 2012) the resulting design values of the material
strength parameter C1,d and ft,90,d , respectively, and the corresponding partial safety factors
γm are listed in Tab. 7. Other values of the fraction of variable load lead to different reliability
indices β f (Fig. 11). Especially the design approach by Ballerini leads to considerably higher
reliability indices for smaller fractions of variable loads compared to the desired value βr =
4.2 corresponding to a probability of failure of p f = 10−5 .

8
7.5
7
6.5
6
5.5
βr [-]

5
4.5
4
3.5 van der Put
Ballerini
3 Ehlbeck et al (Logn)
Ehlbeck et al (Wbl)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Q/(G + Q) [-]

Figure 11: Reliability indices βr in dependency of the fraction of variable load for the approaches suggested
by van der Put, Ballerini and Ehlbeck.

Table 7: Mean, 5%-, and design values of the material parameter C1 and ft,90 , respectively, and the cor-
responding partial safety factors γm for the design approaches suggested by van der Put, Ballerini and
Ehlbeck for a fraction of variable load Qmean / (Gmean + Qmean ) = 0.8.

Approach Parameter Unit Mean 5%-value design value γm

van der Put C1 [N/mm1.5 ] 16.7 9.8 7.19 1.36
Ballerini C1 [N/mm1.5 ] 15.2 10.4 8.39 1.25
Ehlbeck et al. (Logn) ft,90 [N/mm2 ] 0.60 0.38 0.292 1.30
Ehlbeck et al. (Wbl) ft,90 [N/mm2 ] 0.60 0.31 0.12 3.27

141
INTER / 48 - 07 - 04

5.3 Interpretation of the reliability study

When proposing design recommendations for connections loaded perpendicular to the grain,
it has to be accounted for the different concepts behind design equations. A conservative
design philosophy is to avoid design situations with loading in tension perpendicular to the
grain and to avoid any possibility of crack initiation. Cracking is often considered as a diminu-
tion of esthetics though the load-carrying capacity might not be affected. The Swiss standard
for timber structures SIA 265 (2012) is formulated based on the philosophy to give only low
values of tensile strength perpendicular to the grain in order to make the designer not to cal-
culate and verify the tensile stresses but either not to apply critical connections or to foresee
strengthening elements.
The design value for tensile strength perpendicular to the grain ft,90,d = 0.15 N/mm2 for glu-
lam and ft,90,d = 0.1 N/mm2 for solid timber given in SIA 265 (2012) is on the same level as the
design value ft,90,d = 0.12 N/mm2 for connections loaded perpendicular to the grain deter-
mined for the approach by Ehlbeck et al. when using a Weibull distribution (Tab. 7). In con-
trast to that the characteristic value ft,90,k in EN 14080 (CEN, 2013) is even higher compared
to the 5%-value determined for the approach by Ehlbeck et al. when using a Lognormal dis-
tribution. An increase of the partial safety factor γm for the tensile strength perpendicular to
the grain or a reduction of the characteristic value ft,90,k seems necessary.
The material parameter C1 in the approach by van der Put for the implementation in EC5
should be based on the design value C1,d = 7.2 N/mm1.5 . This design values leads in combi-
nation with a general partial safety factor γm = 1.3 to a characteristic value C1,k = 9.3 N/mm1.5 .
A further reduction of the characteristic value based on the above discussion regarding ten-
sile strength perpendicular to the grain or the implementation of the more sophisticated
approach by Ballerini can be reasonable.

6 Reinforcement of connections
In the tests carried out at ETH Zurich also the effect of the reinforcement on the failure
behaviour and the load-carrying capacity of connections loaded perpendicular to the grain
was studied. A beneficial impact of the reinforcement on the load-carrying capacity could
be confirmed, though crack initiation could not be avoided as shown in Fig. 12. After crack
initiation the crack was stabilized approximately at the position of the reinforcement. The
load at ultimate failure reaches the level of the bending capacity of the full cross-section.

250 250
225 225
200 200
175 175
Load [kN]

Load [kN]

150 150
125 125
100 100
75 75
50 50
25 α = 0.6, m = 4, n = 2 25 α = 0.6, m = 4, n = 2
α = 0.6, m = 2, n = 4 α = 0.6, m = 2, n = 4
0 0
−1 0 1 2 3 4 5 6 7 8 9 10 0 100 200 300 400 500 600 700 800
Crack opening [mm] Crack length [mm]
(a) (b)

Figure 12: Crack opening in the center of the connection (a) and development of the crack length (b) in
dependency of the applied load for reinforced connections loaded perpendicular to the grain with α = 0.6.

142
 INTER / 48 - 07 - 04

7 Conclusion
From the study presented in this paper, the following conclusions can be drawn:

- The efficiency of the design approaches evaluated in this study depends on the degree
of detailing of the approach. The relatively simple approach by van der Put forming the
basis of the EC5 design equation requires a higher partial safety factor compared to the
more detailed design approach by Ballerini. A characteristic value C1,k = 9.3 N/mm1.5
can be used in combination with a general partial safety factor γm = 1.3 for the ap-
proach by van der Put which is considerably lower than the current value in EC5.
- The design approach suggested by Ehlbeck et al. based on the tensile strength per-
pendicular to the grain requires a very low design value ft,90,d or a considerable high
value of the partial safety factor γm = 3.3 compared to the current general value in or-
der to achieve the desired reliability (βr = 4.2). Duration of load effects and effects from
moisture variations have to be accounted separately.
- Cracking at connections loaded perpendicular to the grain due to excessive tensile
stresses perpendicular to the grain can occur also for relative connection heights α >
0.7 exceeding the threshold currently given in EC5. Though crack growth becomes more
stable with increasing relative connection height the separation of the beam into indi-
vidual parts of small cross-section has to be accounted for in the design. Bending
failure of the lower cross-section occurred in the tests carried out at ETH Zurich for
α = 0.8.
- Reinforcement is an easy and efficient measure to restore the load-carrying capacity
of beams with connections loaded perpendicular to the grain. Nevertheless, initial
cracking at a load level similar to the unreinforced beam will occur also when using
reinforcement.

References
Ballerini M. (1999): A new set of experimental tests on beams loaded perpendicular-to-grain by dowel-
type joints. In: Proc. of the CIB-W18 Meeting 32, Graz, Austria, Paper No. CIB-W18/32-7-2
Ballerini M. (2004): A new prediction formula for the splitting strength of beams loaded by dowel-type
connections. In: Proc. of the CIB-W18 Meeting 37, Edingburgh, Scotland, Paper No. CIB-W18/37-7-5
Ballerini M., Giovanella A. (2003): Beams transversally loaded by dowel-type joints: influence on split-
ting strength of beam thickness and dowel size. In: Proc. of the CIB-W18 Meeting 36, Colorado,
USA, Paper No. CIB-W18/36-7-7
CEN (2004): EN 1995-1-1: Eurocode 5: Design of timber structures - Part 1-1: General - Common rules
and rules for buildings. European Committee for Standardization CEN, Bruxelles, Belgium
CEN (2005): EN 1990/A1: Eurocode - Basis of structural design. European Committee for Standardiza-
tion CEN, Bruxelles, Belgium
CEN (2013): EN 14080: Timber structures - Glued laminated timber and glued solid timber - Require-
ment. European Committee for Standardization CEN, Bruxelles, Belgium
DIN (2008): DIN 1052: Entwurf, Berechnung und Bemessung von Holzbauwerken - Allgemeine Be-
messungsregeln und Bemessungsregeln für den Hochbau. DIN 1052 (2008-12), DIN Deutsche In-
stitut für Normung e.V., Berlin, Germany
Ehlbeck J., Görlacher R. (1983): Tragverhalten von Queranschlüssen mittels Stahlformteilen, insbeson-
dere Balkenschuhen, im Holzbau. Tech. rep., Versuchsanstalt für Stahl, Holz und Steine, Abteilung
Ingenieurholzbau, Universität Fridericiana Karlsruhe, Germany
Ehlbeck J., Görlacher R., Werner H. (1989): Determination of perpendicular to grain stresses in joints
with dowel-type fasteners - a draft proposal for design rules. In: Proc. of the CIB-W18 Meeting 22,
Berlin, Germany, Paper No. CIB-W18/22-7-2

143
INTER / 48 - 07 - 04

Franke B., Quenneville P. (2011): Design approach for the splitting failure of dowel-type connections
loaded perpendicular to grain. In: Proc. of the CIB-W18 Meeting 44, Vancouver, Canada, Paper No.
CIB-W18/44-7-5
Gustafsson P., Larsen H. (2001): Dowel joints loaded perpendicular to grain. In: Proceedings of the
Int. RILEM Symposium, Stuttgart, Germany, pp 577–586
Habkirk R. (2006): Bolted wood connections loaded perpendicular-to-grain - effect of wood species.
Tech. rep., Department of Civil Engineering, Royal Military College of Canada, Kingston, Ontario,
Canada, master thesis
JCSS (2001): Probabilistic Model Code. Joint Committee of Structural Safety, available online at:
http://www.jcss.byg.dtu.dk/
Jensen J.L., Quenneville P. (2011): Splitting of beams loaded perpendicular to grain by connections–
some issues with EC 5. In: Proc. of the CIB-W18 Meeting 44, Alghero, Italy, research note
Jensen J.L., Gustafsson P.J., Larsen H.J. (2003): A tension fracture model for joints with rods or dowels
loaded perpendicular to grain. In: Proc. of the CIB-W18 Meeting 36, Colorado, USA, Paper No. CIB-
W18/36-7-9
Jockwer R., Steiger R., Frangi A., Kohler J. (2011): Impact of material properties on the fracture me-
chanics design approach for notched beams in Eurocode 5. In: Proc. of the CIB-W18 Meeting 44,
Alghero, Italy, Paper No. CIB-W18/44-6-1
Kasim M.H. (2002): Bolted timber connections loaded perpendicular-to-grain - effect of row spac-
ing on resistance. Tech. rep., Department of Civil Engineering, Royal Military College of Canada,
Kingston, Ontario, Canada, master thesis
Kohler J., Steiger R., Fink G., Jockwer R. (2012): Assessment of selected Eurocode based design equa-
tions in regard to structural reliability. In: Proc. of the CIB-W18 Meeting 45, Växjö, Sweden, Paper
No. CIB-W18/45-102-1
Lehoux M.C.G., Quenneville J.H.P. (2004): Bolted wood connections loaded perpendicular-to-grain, a
proposed design approach. In: Proc. of the CIB-W18 Meeting 37, Edinburgh, Schottland, Paper No.
CIB-W18/37-7-4
Leicester R.H. (1973): Effect of size on the strength of structures. Commonwealth scientific and indus-
trial research organization CSIRO, Melbourne, Australia
Lignum (1990): Holzbau-Tabellen HBT 2. Lignum, Zürich
Mistler H.L. (1998): Design of glulam beams according to EC 5 with regard to perpendicular-to-grain
tensile strength - A comparison with research results. Holz als Roh - und Werkstoff 56(1):51–60
Möhler K., Siebert W. (1981): Queranschlüsse bei Brettschichtträgern oder Vollholzbalken. Bauen mit
Holz 83:84–89
van der Put T.A.C.M. (1990): Tension perpendicular to the grain at notches and joints. In: Proc. of the
CIB-W18 Meeting 23, Lisbon, Portugal, Paper No. CIB-W18/23-10-1
Quenneville J., Mohammad M. (2001): A proposed canadian design approach for bolted connections
loaded perpendicular-to-grain. In: Proceedings of the Int. RILEM symposium on joints in timber
structures, pp 61–70
Reshke R. (1999): Bolted timber connections loaded perpendicularto grain: influence of joint config-
uration parameters on strength. Tech. rep., Department of Civil Engineering, Royal Military College
of Canada, Kingston, Ontario, master thesis
Schoenmakers J.C.M. (2010): Fracture and failure mechanisms in timber loaded perpendicular to the
grain by mechanical connections. PhD thesis, Technische Universiteit Eindhoven
SIA (2012): Standard SIA 265 - Timber Structures. SIA Swiss Society of Engineers and Architects, Zurich,
Switzerland
Sørensen J.D. (2002): Calibration of partial safety factors in Danish structural codes. In: Proc. of the
JCSS Workshop on Reliability Based Code Calibration, Zurich, Switzerland
Vesa J., Kevarinmäki A. (2001): Long-term capacity of bolted joints loaded perpendicular to grain. In:
Proceedings of the Int. RILEM Symposium, Stuttgart, Germany, pp 587–596
Zarnani P., Quenneville P. (2013): Wood splitting capacity in timber connections loaded transversely:
Riveted joint strength for full and partial width failure modes. In: Proc. of the CIB-W18 Meeting 46,
Vancouver, Kanada, Paper No. CIB-W18/46-7-5

144
 INTER / 48 - 07 - 04

Discussion

The paper was presented by R Jockwer

F Lam commented that 3 P Weibull distribution would provide a better fit for the dis-
tribution tails. R Jockwer agreed.
I Smith suggested use of reinforcement rather than depending on design approaches
that rely on stable crack growth.
H Blass asked about reinforcement between the dowels. R Jockwer responded that
they thought about this but did not use it because of the arrangement of steel plates
and dowels were rather close. H Blass and R Jockwer further discussed the diameter
of the dowel in relationship to the size of the potential reinforcement screws.
P Zarnani and R Jockwer discussed issues related to recording the crack opening.
P Gustafsson asked if this is only valid for symmetric test set up. R Jockwer answered
that they had considered one sided loading and it seemed to be okay. However the
experimental base for unsymmetrical loading is quite weak.

145
146
 INTER / 48 - 07 - 05

Simplified Fatigue Design of Typical Tim-

ber-Concrete Composite Road Bridges
Dipl.-Ing. Katrin Kudla, Scientific Researcher
Prof. Dr.-Ing. Ulrike Kuhlmann, Head of Institute
Institute of Structural Design, University of Stuttgart, Germany

Keywords: Timber-concrete composite, road bridge, notch connection, fatigue

1 Introduction
Timber-concrete composite (TCC) road bridges are an economical solution for one- or
two-lane single span bridges with spans between 10 and 30 m. Approximately 50 % of
all existing road bridges in Germany have spans ranging from 5 to 30 m (Schmitt et al.,
2003), which shows the big market. Because traffic is representing a variable cyclic
load, fatigue verifications of all structural members including the shear connectors are
required for the design.
Design rules for fatigue verifications of TCC bridges and of appropriate connectors are
not given in the standards. That is why, based on an extensive parameter study and
traffic simulations, fatigue verifications and possible simplifications for typical types of
TCC bridges with notched shear connections have been investigated (Kuhlmann &
Kudla, 2015). Notched connections are especially appropriate for bridges due to high
stiffness, high strength values and the simplicity of construction. Thus a practice-ori-
ented design proposal for simplified fatigue verifications of the structural members
and the notched connection has been developed. The results are directly applicable
also for the connector type described in Müller (2014). Other possible connection
types for TCC road bridges, e.g. hbv system (Bathon & Bletz-Mühldorfer, 2014) or X-
connectors (Kuhlmann & Aldi, 2009), were not considered, but studies could easily be
extended.

2 Important parameters of the bridges

2.1 Geometry and material properties
For the improvement of the fatigue verifications given in EN 1995-2 (2010) regarding
TCC road bridges, simulations with two different fatigue load models according to
EN 1991-2 (2010) were carried out. Herein typical geometries, span lengths and mate-
rial properties, which are relevant for practice, were taken into account. In Table 1 the

147
INTER / 48 - 07 - 05

most important conditions and parameters, the simulations were based on, are sum-
marised.
In total 15 different TCC road bridges were investigated. The cross-sections of the
bridges, given in Table 2, were calculated using two-dimensional framework models,
as e.g. in Figure 5, under static loading conditions and with all relevant load cases. Time
steps t = 0, t = 3-7 years and t = ∞ were considered according to Schänzlin & Fragia-
como (2006).

Table 1. Parameters for the simulations with TCC bridges.

Parameter Property
Structural system Single span bridges
Span length 10 m, 15 m, 20 m, 25 m and 30 m
Bridge Type 1 (Figure 1) Total width 4.5 m (single lane bridge), one block timber beam
Bridge Type 2 (Figure 2) Total width 10.5 m (two lanes), two block timber beams
Bridge Type 3 (Figure 3) Total width 10.5 m (two lanes), six timber beams
Shear connector Notch, slip modulus Kser = Ku = 1,600 kN/mm per m beam width
Timber strength class Glulam GL28c according to EN 14080 (2013)
Concrete strength class C 40/50 according to EN 1992-1-1 (2011)

Table 2. Cross-sections for all bridge Types.

Bridge Span Width of tim- Height of tim- Effective width of con- Thickness of con-
Type length [m] ber beam [m] ber beam [m] crete slab per beam [m] crete slab [cm]
Type 1 10 2.52 0.6 4.5 22
15 2.70 0.8 4.5 25
20 3.06 0.9 4.5 28
25 3.24 1.1 4.5 34
30 3.60 1.2 4.5 40
Type 2 10 2 x 2.52 0.6 4.19 28
15 2 x 2.70 0.8 4.76 28
20 2 x 3.24 0.9 5.25 28
25 2 x 3.42 1.1 5.25 32
30 2 x 3.60 1.2 5.25 38
Type 3 10 6 x 0.90 0.7 1.60 22
15 6 x 0.90 0.8 1.70 25
20 6 x 1.08 0.9 1.78 28
25 6 x 1.26 1.1 1.76 32
30 6 x 1.26 1.2 1.76 36

148
 INTER / 48 - 07 - 05

For a span length of 20 m, bridge Types 1 to 3 are shown in Figures 1 to 3. Bridge Type 1
with only one lane could be built for agricultural or forestry use and low flow rates of
lorries (see Table 4.5 in EN 1991-2 (2010), traffic categories 3 or 4). In Germany, for
this kind of bridges a total width of 4.5 m is recommended. Bridge Types 2 and 3 could
be used for roads and motorways with up to 300 lorries per day. This restriction is not
related to the fatigue failure, but to geometry specifications given in the German
guideline for road construction (Forschungsgesellschaft für Straßen- und Verkehrs-
wesen, 1996). Here the number of heavy vehicles per day is decisive for the choice of
a standard cross-section for the bridge. For the project standard cross-section types
RQ 7.5 (width of one lane 2.75 m) or RQ 9.5 (width of one lane 3.0 m) were chosen for
bridge Types 2 and 3. Even though type RQ 7.5 should be used for a maximum of 60 lor-
ries per day and type RQ 9.5 for a maximum of 300 lorries per day, for investigation of
the fatigue verifications, traffic categories 3 to 1 according to Table 4.5 in EN 1991-2
(2010) were considered. Traffic categories given in EN 1991-2 (2010) indicate the num-
ber of heavy vehicles expected per year and per lane. Traffic category 3 stands for
343 lorries per day and traffic category 1 stands for 5480 lorries per day.
Standard cross-section type RQ 7.5 with a total width of 10.0 m was chosen for bridge
Type 2 with span lengths of 10 m and 15 m due to the long cantilever of the concrete
slab. Also the thickness of the concrete slab was increased in these cases, see Table 2.
In practice it might be a better solution to support the cantilevers with additional tim-
ber beams. In any other case of bridge Types 2 and 3 standard cross-section type
RQ 9.5 was used.

Figure 1. Type 1 - Cross-section for single lane bridge (span length 20 m), dimensions in cm.

149
INTER / 48 - 07 - 05

Figure 2. Type 2 - Standard cross-section for two lane bridge (span length 20 m), dimensions in cm.

Figure 3. Type 3 - Standard cross-section for two lane bridge (span length 20 m), dimensions in cm.

2.2 Properties of the notched shear connection

2.2.1 Dimensions and slip modulus
In Figure 4 the functionality of a notched connection, where the shear force is trans-
ferred directly through the compression edge of the notch, is shown. Additionally a
screw (not shown) is placed vertically in the notch for transferring lifting forces be-
tween timber and concrete. For the shear connection between timber and concrete
50 mm deep and 200 mm long notches were considered. The longitudinal distance
between the notches was varied according to the shear force distribution for a uni-
formly distributed load.
The slip modulus of the connection of Ku = Kser = 1,600 kN/mm per m beam width has
been estimated according to former research works and experimental tests. In
Kuhlmann & Kudla (2015) a summary of different values for the slip modulus based on
various push-out tests is included. A relatively high slip modulus of 1,600 kN/mm repre-
sents a conservative assumption for
the fatigue verification. A lower slip
modulus would lead to a better load
distribution between the notches and
Figure 4. Notched connection within a TCC beam. to a decreased shear force of the maxi-
mum loaded notch.

150
 INTER / 48 - 07 - 05

2.2.2 Fatigue strength

For an application to traffic loading conditions, the fatigue behaviour of the notch is
needed. At the University of Stuttgart (Germany) experimental tests on push-out and
beam specimens with notches have been carried out under static and fatigue loading
(Kuhlmann & Aldi, 2010). All the tests showed a shear failure mechanism in the timber
beam in front of the notch, which is more critical than a relatively ductile compression
fatigue failure. Based on the test results from Kuhlmann & Aldi (2010) it was identified
that the longitudinal shear strength of the notch and the corresponding number of
load cycles to failure match the specifications given in EN 1995-2 (2010), Annex A,
§ A3 (3) for timber elements under shear. At the University of Stuttgart the fatigue
tests were carried out for a load ratio of R = 0.1. Based on the good accordance of tests
and design code in this case, the values for strength reduction according to EN 1995-2
(2010) were used for the traffic simulations in Kuhlmann & Kudla (2015). The reduction
of strength with number of load cycles may be calculated depending on the type of
action with Equation (1).
1 R
k fat  1   log(   N obs  t ) (1)
a  (b  R)

where R = Fmin / Fmax, Nobs t = number of constant amplitude load cycles in total, β = de-
sign fatigue factor based on the damage consequence for the actual structural compo-
nent, a, b = coefficients representing the type of fatigue action. The value β has been
chosen to 3 for substantial consequences in case of damage (EN 1995-2, 2010, An-
nex A, § A3 (3)). Corresponding to the dominating failure mechanism in the tests, the
values of coefficients a and b have been chosen to a = 6.7 and b = 1.3 for shear, see
Table A.1 in EN 1995-2 (2010).
Using the factor representing the reduction of strength, the characteristic strength for
static loading may be modified for the fatigue verification. The design fatigue strength
for shear may be calculated according to Equation (2).
Fult
F fat,d  k fat  (2)
 M , fat

where Fult = characteristic shear strength for static loading, γM,fat = 1.0 = material safety
factor for fatigue.

3 Traffic simulations
3.1 Fatigue load models (FLM) for the fatigue design
For the traffic simulations, two different fatigue load models (FLM 3 and FLM 4 accord-
ing to EN 1991-2, 2010) were used. FLM 3 consists of a single vehicle load, whereas
FLM 4 represents five different types of lorries. FLM 3 is a simplified model, which orig-
inally was derived through a comparison of the damage of different internal forces in

151
INTER / 48 - 07 - 05

a main girder due to FLM 3 and due to a more detailed load model. As common for
other construction materials, the resulting internal forces due to FLM 3 may be modi-
fied with damage equivalent λ-factors corresponding to the geometry of the bridge
and the real traffic load in order to simplify the fatigue verification. Without such a
modification, only the total number of constant amplitude load cycles ΔFmax per year
can be taken into account, see EN 1995-2 (2010), Annex A, § A3 (1). The consequence
is an overall conservative estimation of the damage due to traffic loads. A more precise
and detailed possibility is the fatigue verification with FLM 4 and a damage accumula-
tion according to Palmgren-Miner rule (Miner, 1945). It consists of five types of lorries
that are typical for the European traffic. The standard lorries are defined by the num-
ber of axles, axle distances, axle load and wheel types. For FLM 4, the percentages of
different lorry types in the sequence depend on the traffic type (long distance, medium
distance or local traffic). For FLM 4 as well as for FLM 3, the total number of lorries
expected per year and per lane depends on the traffic category (EN 1991-2, 2010).
How the damages due to FLM 3 and FLM 4 have been calculated is shown in Stephan
& Kuhlmann (2014).

3.2 Determination of internal forces due to fatigue loading with two-dimensional

and three-dimensional models
To determine internal forces of a TCC bridge with notches due to fatigue loading, a
two-dimensional framework model according to Grosse et al. (2003) may be used, see
Figure 5. In the model, properties of the concrete slab are assigned to the upper chord
and properties of the timber beam are assigned to the lower chord. The slip modulus
of the connection is transformed into an effective bending stiffness of the vertical
struts with hinges at the edges, connecting upper and lower chord and representing
the notches.

Figure 5. Framework model for a TCC bridge with notches (side view and cross-section Type 3).

152
 INTER / 48 - 07 - 05

With two-dimensional models, influence lines for a moving load of 1 kN in longitudinal

direction were developed for all bridges and the governing internal forces. These influ-
ence lines were the basis for the calculation of load ranges due to the crossing of mov-
ing axle loads from fatigue load models FLM 3 and FLM 4.
For bridge Type 3, also influence lines for the transverse direction were calculated us-
ing three-dimensional models. With three-dimensional models the influence of the
load distribution between the timber beams were analysed. Especially the influence of
the load distribution on the maximum shear force at the notch was investigated with
regard to the fatigue design. One of the results was that if the span length increases,
the load distribution between the notches in longitudinal direction and also the distri-
bution between adjacent beams in transverse direction are improved, because the
overall stiffness of the system decreases. For the fatigue verification of the notch for
bridge Type 3, the maximum load range could be calculated with a reduction factor
depending on the span length (see Table 3), Table 3. Reduction factor considering load dis-
if the load distribution between the beams tribution for bridge Type 3 (calculation of the
should be taken into account. The reduction shear force at the notch).
factor corresponds to the maximum shear Span length [m] Reduction factor
force at the outer beam, calculated with a 10 0.74
three-dimensional model. To derive the 15 0.74
shear force from a two-dimensional model, 20 0.71
the maximum force from this model can be 25 0.67
multiplied by the reduction factor given in 30 0.66
Table 3.

4 Summary of the numerical parameter studies

4.1 Conditions for the parameter studies
Within the parameter studies, three different bridge types with five different span
lengths (see Table 2), four traffic categories, three traffic types and service lives be-
tween 80 and 140 years were considered. Damage effects of fatigue load models FLM 3
and the more detailed FLM 4 were compared to each other in order to identify critical
cases for which fatigue verifications are required. A detailed description of the param-
eter studies and calculated damages may be found in Kuhlmann & Kudla (2015).

4.2 Results
 For a bridge with a span length of 25 m or more and a service life less than 140 years,
there is no risk of a fatigue failure of any timber member or notch.
 A high traffic volume (traffic category 1 with 2 million lorries per year) leads to a
greater fatigue damage than a low traffic volume (traffic category 4 with 50,000 lor-
ries per year). This correlation is linked to the total number of load cycles during

153
INTER / 48 - 07 - 05

service life. Therefore the fatigue damage also rises with longer service lives, but the
influence of the traffic category is higher.
 Damages due to FLM 4 are significantly less than the damages due to FLM 3 with
only a few exceptions.
 Based on investigations with the more detailed FLM 4, fatigue verifications of the
notches are relevant for the design if all of the following conditions apply:
o Design for traffic categories 1 or 2 according to EN 1991-2 (2010) and span
length smaller than 25 m
o Traffic types of medium or long distance (a minimum distance of 50 km) ac-
cording to EN 1991-2 (2010), which is decisive for the amount of lorries
o Bridge Type 3 (without consideration of the load distribution in transverse
direction)
 In any other case strength verifications of the notch under static loading conditions
are governing the design compared to fatigue verifications. Especially for local traffic
and traffic categories 3 or 4 there is no risk of fatigue failure.
 Fatigue failures of the timber beams under bending moment and tension force at
mid-span and under shear at the supports are not relevant.
 A fatigue verification of the concrete slab was not required in any case, because the
concrete compression stresses for the characteristic combinations of actions were
smaller than 60 % of the characteristic compression strengths.

5 Proposal for a simplified fatigue verification of TCC

bridges with notches
5.1 Design steps
5.1.1 Overview
For the combination of bending moment and tension force in the timber beam as well
as for the shear at the support fatigue verifications are not required, if the ratio κ be-
tween the stress range for fatigue loading Δσ = |σmax - σmin| and the corresponding
design strength fk/γM,fat is not higher than the following values (EN 1995-2, 2010, An-
nex A, § A1 (3)):
 Members in bending and tension: 0.2

 Members in shear: 0.15

Based on investigations in Kuhlmann & Kudla (2015) it is proposed that these threshold
values should be checked, if the bridge is designed for traffic categories 1 or 2 and a
span length smaller than 25 m. For any other case the fatigue verifications of these
internal forces have proved to be not relevant. If the given threshold value for the

154
 INTER / 48 - 07 - 05

combination of bending moment and tension force or shear at the support is ex-
ceeded, the fatigued design should follow the procedure given in EN 1995-2 (2010),
Annex A.3.
For the notched connection, the new design procedure for the fatigue verification is
summarised in Figure 6. The chosen standard cross-sections of the bridges are primar-
ily applicable for traffic categories 3 and 4 (see Chapter 2.1). For these traffic categories
no fatigue verifications are required, because the strength verifications govern the de-
sign. In the following, results of further investigations with traffic categories 1 and 2
are given which show the chances in TCC road bridge construction regarding fatigue.

Figure 6. Proposed design steps for the simplified fatigue verification of the notched connection.

The threshold value for ratio κ given in EN 1995-2 (2010) indicates that a fatigue ver-
ification of the notch (timber member in shear) is not required, if the stress due to
FLM 3 is smaller than 15 % of the characteristic shear strength. It is not known on
which tests or publications this assumption is based on. Using a threshold value for
ratio κ implies the existence of a constant amplitude fatigue limit for each loading
direction. If the maximum stress from fatigue loading is lower than this fatigue limit,
an infinite number of load cycles would be bearable and no fatigue failure would oc-
cur. But the existence of constant amplitude fatigue limits similar to steel is not
proven for timber. According to Mohr (2001) a bilinear S-N-curve with a smaller gra-
dient of the curve for load cycles of more than 2 million might be possible for the
shear fatigue strength. But the specifications for the threshold values given in
EN 1995-2 (2010), where no fatigue limit has to be considered, proved to be on the
safe side for the cases considered in Kuhlmann & Kudla (2015) and should be applied.
Verifications with a threshold value for ratio κ do not include influences from the num-
ber of load cycles per year, the service life and load ratio R. There is no relation to the
governing S-N-line or the reduction of strength with number of load cycles. Therefore
a new design procedure is proposed to consider these influences within a simplified
fatigue verification of the notched connection.

155
INTER / 48 - 07 - 05

5.1.2 Step 1: Verification of the notch with a threshold value to check whether a de-
tailed fatigue verification is needed
For traffic categories 1 and 2 and span lengths smaller than 25 m, a verification of ratio
kv,max, which is the ratio between the maximum shear force under fatigue loading and
the shear strength, is recommended. Ratio kv,max is calculated according to Equa-
tion (3).
Fd ,min  FFLM 3 Fd ,max
k v ,max  
Fult Fult (3)
 M , fat  M , fat

where Fult = characteristic shear strength, γM,fat = 1.0, Fd,min = shear force due to per-
manent load and ΔFFLM3 = shear force due to FLM 3. The characteristic shear strength
of the notch is determined according to Equation (4).
Fult  8  tv  b  kcr  f v,k (4)

where tv = notch depth, b = beam width, kcr = 2.5 / fv,k according to German NA for
EN 1995-1-1 (2010) and fv,k = 3.5 N/mm² according to EN 14080 (2013).
The verification for ratio kv,max is shown in Equation (5).
k v,max  k fat (5)

The threshold value kfat (reduction of strength with number of load cycles according to
Equation (1)) for ratio kv,max can be calculated for traffic categories 1 and 2 separately.
A diagram that shows the reduction of strength depending on the number of load cy-
cles according to the traffic category (TC 2 traffic category 2 and TC 1 traffic category
1) and load ratio R = Fd,min / Fd,max allows for a simple estimation of the fatigue strength
of the notched connection,
see Figure 7.
In Table 4 and Table 5 values
for ratio kv,max and corre-
sponding parameters for the
considered bridges are sum-
kfat

marised. In Table 5 it is distin-

guished whether the load dis-
tribution between the timber
beams of bridge Type 3 is con-
sidered or not.
For example, a TCC bridge of R

Type 2 with a span length of Figure 7. Reduction of shear strength for a service life of 100
20 m gives a load ratio R of years depending on traffic category (TC 2 traffic category 2
0.691 (see Table 4). Looking at and TC 1 traffic category 1 according to EN 1991-2 (2010)).

156
 INTER / 48 - 07 - 05

the diagram in Figure 7, ratio kv,max for the notch has to be smaller than 0.4 (for R = 0.7),
so that the road bridge may be built for traffic category 2 (TC 2). In this case
kv,max = 0.266 < kfat =0.381 (see Table 4) , so the fatigue resistance is verified.

Table 4. Shear forces at the notch due to permanent load and FLM 3 and values kv,max and kfat for
bridge Types 1 and 2.
Bridge Type 1 2 1 2 1 2 1 2 1 2
Span length [m] 10 10 15 15 20 20 25 25 30 30
Fult [kN] 2520 2520 2700 2700 3060 3240 3240 3420 3600 3600
Fd,min [kN] 288 340 364 443 515 595 615 689 772 852
ΔFFLM3 [kN] 238 238 260 260 266 266 234 234 223 223
Fd,max [kN] 526 578 624 703 781 861 849 923 995 1075
R [-] 0.548 0.588 0.583 0.630 0.659 0.691 0.724 0.746 0.776 0.793
kv,max [-] 0.209 0.229 0.231 0.260 0.255 0.266 0.262 0.270 0.276 0.299
kfat,TC 1 [-] 0.212 0.242 0.238 0.277 0.303 0.335 0.373 0.400 0.440 0.464
kfat,TC 2 [-] 0.266 0.294 0.291 0.326 0.351 0.381 0.416 0.441 0.478 0.501

Table 5. Shear forces at the notch due to permanent load and FLM 3 and values kv,max and kfat for
bridge Type 3.
Bridge Type Type 3 without load distribution Type 3 with load distribution
Span length [m] 10 15 20 25 30 10 15 20 25 30
Fult [kN] 900 900 1080 1260 1260 900 900 1080 1260 1260
Fd,min [kN] 67 109 154 176 225 72 107 144 181 204
ΔFFLM3 [kN] 103 130 133 117 112 76 91 95 79 74
Fd,max [kN] 170 239 287 293 337 148 198 239 260 278
R [-] 0.394 0.456 0.537 0.601 0.669 0.486 0.540 0.603 0.696 0.734
kv,max [-] 0.189 0.266 0.266 0.233 0.267 0.164 0.220 0.221 0.206 0.221
kfat,TC 1 [-] 0.124 0.156 0.205 0.252 0.312 0.173 0.207 0.253 0.341 0.384
kfat,TC 2 [-] 0.184 0.213 0.259 0.303 0.360 0.230 0.262 0.305 0.386 0.426

In Table 5, cases for which a detailed fatigue verification is recommended (as the ra-
tio kv,max exceeds kfat) are printed in bold. If the load distribution is taken into account
for bridge Type 3, values for ratio kv,max become smaller. Here only a fatigue verification
for traffic category 1 and span lengths of 10 m and 15 m is necessary.

5.1.3 Step 2: Fatigue verification of the notch using a damage equivalent load
If ratio kv,max, calculated in step 1, exceeds the threshold value kfat, a more detailed
fatigue verification is recommended. For the investigated types of TCC bridges, a fa-
tigue verification with a damage equivalent force has been derived. This concept is
applicable for the following cases:

157
INTER / 48 - 07 - 05

 Traffic category 1 and long distance traffic: Bridge Type 2 with span length 10 m and
Type 3 with span length 10 m, 15 m, 20 m and 25 m
 Traffic category 2 and medium or long distance traffic: Bridge Type 3 with span
length 10 m, 15 m and 20 m
A verification with a damage equivalent force avoids a complex analysis with FLM 4
and damage accumulation according to Palmgren-Miner rule, but allows for the con-
sideration of parameters like span length, service life, traffic volume or bridge type.
Therefore damage equivalent factors (λ-factors) were derived for the modification of
the shear force caused by FLM 3 in view of damage caused by FLM 4 (Stephan &
Kuhlmann, 2014). For the verification, the maximum shear force at the notch has to be
calculated with the sum of the force due to permanent load and the modified force
due to FLM 3, see Equation (6).
Fd , max  Fd ,min    FFLM 3 (6)

The maximum damage equivalent force according to Equation (6) has to be smaller
than an equivalent fatigue strength given in Equations (7) and (8). The definition of an
equivalent fatigue strength is similar to the simplified fatigue verification in steel con-
struction, where a reference value of the fatigue strength at 2 million load cycles (= de-
tail category) is used. The reduction factor kequ is calculated according to Equation (1)
with certain values for R, Nobs and t. Further information on the chosen parameters,
which were applied to determine the equivalent shear strength reduction kequ = 0.385,
and on the calculation of λ-factors with the damage due to FLM 4 may be found in
Stephan & Kuhlmann (2014) or Kuhlmann & Kudla (2015).
Fult
Fd , max  k equ  (7)
 M , fat

Fult
Fd , max  0.385  (8)
 M , fat

Using a reference value for the fatigue strength means that a strength reduction kfat
does not have to be calculated for each case separately. Here the consideration of the
effects caused by load ratio R, number of load cycles Nobs and service life t are included
in the λ-factors. The calculated λ-factors, given in Table 6, are only applicable in com-
bination with kequ = 0.385 and for the chosen cross-sections. For other geometries or
another number of beams for bridge Type 3, different λ- factors may occur.
For example, for bridge Type 3 and 15 m span length (load distribution in transverse
direction is neglected), the simplified verification in step 1 is exceeded for traffic cate-
gory 2, which means kv,max > kfat (0.266 > 0.213, see Table 5). In this case, the λ-factor
is 2.3 for a service life of 100 years and traffic type “medium distance”, see Table 6. In
general, λ > 1.0 means that the force due to FLM 3 is increased and 0 < λ < 1.0 means

158
 INTER / 48 - 07 - 05

that it is decreased. Values λ < 0 are not permitted, because that would cause a reduc-
tion of the permanent load. Further λ-factors for traffic category 1 (medium and long
distance traffic) are included in Kuhlmann & Kudla (2015).

Table 6. λ-factors for bridge Type 3 and traffic category 2 for medium and long distance traffic.
Traffic Type Medium distance Long distance
Span length [m] 10 15 20 10 15 20
80 a 0.9 1.6 - 1.3 3.0 0.4
100 a 1.2 2.3 0.1 1.8 3.9 0.8
120 a 1.6 2.9 0.3 2.3 4.9 1.2
140 a 2.0 3.5 0.6 2.8 5.8 1.6

6 Conclusions
Even though the preferred field of application for pilot projects with TCC bridges are
road bridges for a low flow rate of heavy vehicles, the research project showed that
high traffic categories and service lives of more than 100 years are realistic regarding
fatigue. The dimensions of typical TCC bridges with spans up to 30 m are mainly defined
by short-term and long-term strength verifications.
Fatigue verifications of the timber elements for the combination of bending moment
and tension force at mid-span and shear force at the support can be calculated accord-
ing to EN 1995-2 (2010). In general, the parameter studies showed that fatigue verifi-
cations of TCC bridges with notches are not required for traffic categories 3 and 4 and
for a span length of more than 25 m. In any other case, a simplified fatigue design
procedure modifying the verification according to EN 1995-2 (2010), is proposed for
the verification of the notch. Thereby, a complex fatigue verification with FLM 4 and
the linear damage accumulation method (Palmgren-Miner rule) is not necessary for
typical TCC bridges.
Further parameter studies with other types of cross-sections and shear connectors are
recommended in order to identify the importance and the verification of fatigue fail-
ures for these situations.

7 Acknowledgement
The research was financially supported by the international Association for Technical
Issues related to Wood (iVTH) and by the German Federation of Industrial Research
Associations (AiF). This support is gratefully acknowledged.

159
INTER / 48 - 07 - 05

8 References
Bathon, L. & Bletz-Mühldorfer, O. (2014): Fatigue Design of Wood-Concrete-Composite System.
In: World Conference on Timber Engineering (Québec City, Canada).
EN 1991-2 (2010): Eurocode 1: Actions on structures – Part 2: Traffic loads on bridges. With
German NA (2012). DIN – Deutsches Institut für Normung e.V.
EN 1992-1-1 (2011): Eurocode 2: Design of concrete structures – Part 1-1: General rules and
rules for buildings. With German NA (2013). DIN – Deutsches Institut für Normung e.V.
EN 1995-1-1 (2010): Design of timber structures – Part 1-1: General – Common rules and rules
for buildings. With German NA (2013). DIN – Deutsches Institut für Normung e.V.
EN 1995-2 (2010): Eurocode 5: Design of timber structures – Part 2: Bridges. With German NA
(2011). DIN – Deutsches Institut für Normung e.V.
EN 14080 (2013): Timber structures – Glued laminated timber and glued solid timber – Require-
ments. DIN – Deutsches Institut für Normung e.V.
Forschungsgesellschaft für Straßen- und Verkehrswesen – Arbeitsgruppe Straßenentwurf (1996):
Richtlinien für die Anlage von Straßen RAS Teil Querschnitte RAS-Q 96 (in German) / Bundes-
ministerium für Verkehr.
Grosse, M. et al. (2003): Modellierung von diskontinuierlich verbundenen Holz-Beton-Verbund-
konstruktionen, Teil 1: Kurzzeittragverhalten (in German). Bautechnik 80 (8), p. 534-541.
Kuhlmann, U. & Aldi, P. (2009): Prediction of the Fatigue Resistance of Timber-Concrete Compo-
site Connections. In: International Council For Research And Innovation In Building And Con-
struction. Meeting 42, CIB-W18/42-7-11, Dübendorf, Switzerland.
Kuhlmann, U. & Aldi, P. (2010): Research project DGfH/iVTH 15052 N Ermüdungfestigkeit von
Holz-Beton-Verbundträgern im Straßenbrückenbau (in German). Universität Stuttgart, Insti-
tut für Konstruktion und Entwurf, No. 2010-60X, final report.
Kuhlmann, U. & Kudla, K. (2015): Research project AiF/iVTH 17070 N Vereinfachter Ermüdungs-
nachweis von Holzbauteilen in Holz- und Holz-Beton-Verbundstraßenbrücken (in German).
Universität Stuttgart, Institut für Konstruktion und Entwurf, No. 2015-5X, final report.
Miner, A. (1945): Cumulative Damage in Fatigue. In: Journal of Applied Mechanics 12 p.159-164.
Mohr, B. (2001): Zur Interaktion der Einflüsse aus Dauerstands- und Ermüdungsbeanspruchung
im Ingenieurholzbau (in German). Berichte aus dem Konstruktiven Ingenieurbau, Technische
Universität München, Diss.
Müller, J. (2014): Trag- und Verformungsverhalten spezieller Verbundelemente für Holz-Beton-
Verbundstraßenbrücken unter Kurzzeit-, Ermüdungs- und Langzeitbeanspruchung (in Ger-
man), Institut für Konstruktiven Ingenieurbau, Bauhaus-Universität Weimar, Diss.
Schänzlin, J. & Fragiacomo, M. (2006): Extension of EC5 Annex B formulas for the design of tim-
ber-concrete composite structures. In: International Council For Research And Innovation In
Building And Construction. Meeting 40, CIB W-18, Bled, Slovenia, pages 178-188.
Schmitt, V. et al. (2003): Statistische Grundlage zum Forschungsprojekt: Untersuchungen zum
verstärkten Einsatz von Stahlverbundkonstruktionen bei Brücken kleiner und mittlerer Stütz-
weiten (in German). Research project FOSTA P 629.
Stephan, K. & Kuhlmann, U. (2014): Determination of damage equivalent factors for the fatigue
design of timber-concrete composite road bridges with notched connections. In: World Con-
ference on Timber Engineering (Québec City, Canada).

160
 INTER / 48 - 07 - 05

Discussion

The paper was presented by K Kudla

I Smith commented that we do not really know what the damage accumulation rule
should be. Creep rupture rather than cyclic load effect would be more important with
the concept of killer load. He asked about the experimental verification of the work. K
Kudla answered that experimental verification was not done. Miner’s rule should be
on the safe side.
W Seim received clarification that the S-N line results are conservative. U Kuhlmann
commented that this is a typical approach.
P Zarnani asked about the geometry of the concrete slab compared to the beam. K
Kudla answered that it would not make sense to change the geometry as they had
aimed for optimal conditions.
K Malo asked if temperature rise was experienced during testing. U Kuhlmann re-
sponded yes, but that it should not be a big issue.

161
162
 INTER / 48 - 12 - 01

Concentrated load introduction in CLT

elements perpendicular to plane
Bogensperger Thomas, Competence Center for Timber Engineering – holz.bau for‐
schungs gmbh, A‐8010 Graz, [email protected]
Jöbstl Robert August, Haas Fertigbau Holzbauwerk GmbH & Co KG/Haas Holzpro‐
dukte GmbH, Radersdorf 62, A‐8263 Großwilfersdorf /Industriestraße 8, D‐84326
Falkenberg, robert.joebstl@haas‐group.com
Keywords:
Cross laminated timber (CLT), concentrated load introduction, column supported
CLT slab, punching strength

1 Introduction
Cross laminated Timber (CLT) – also known as X‐lam – can be utilised as a plate‐like
element for loads in and/or out of plane. Under loads out of plane bending moment
and shear force develop often in one direction. For this cases several methods for
computation of internal section forces are available. For example the well known
Timoshenko beam theory, which takes into account the high weakness of layers un‐
der rolling shear stresses is one commonly used representative. Other methods are
e. g. the modified γ‐method and the shear analogy method [8, 9]. A recently work
out study [1] showed that differences in the calculation of stresses determined with
higher theories and simple beam theories like Timoshenko's are negligible.
Modern challenging architecture is demanding solutions where CLT slabs activate
two dimensional load carrying behaviour. Support by columns (e.g. see Fig. 1.1) is
often needed for such solutions. For this purpose plate theories under consideration
of shear effects (better known as Reissner‐Mindlin‐plate theory) are a de‐facto
standard for the computation of stresses. The mechanical stiffness properties for
the Reissner‐Mindlin plate theory can be found in [2]. Usually five stiffness values –
two bending, one twist and two shear stiffness values – are necessary. Typically the
distribution of internal section forces and moments of CLT elements under concen‐
trated loads show high values in a small zone around the load. It this context it has
to be mentioned that strength verifications for CLT elements are widely clear for in‐
ternal section forces and moments except twisting moments [3].
Reissner‐Mindlin plate theory corresponds to Timoshenko beam theory, but compa‐
rably 2D plate‐theories on comparable level of modified γ‐method or shear analogy

163
INTER / 48 - 12 - 01

method, implemented e.g. in FEM software for engineering applications, are still
missing.

Figure 1.1. examples for CLT plates with column support.

This paper deals with bending and shear failure of cross laminated timber (CLT)
plates under concentrated loads generated by e.g. columns as supports. Several ver‐
ifications have to be performed for this cases: One is the verification of compression
perpendicular to plane (presented by Bogensperger et. al at CIB meeting in 2011
[10]). In addition the bending and shear stresses in the CLT plate under high local
loads (punching) have to be done which shall be presented in this paper.

2 State of the art

Bending verifications of CLT panels have to consider effective cross section proper‐
ties due to the orthogonally layered cross‐section the of CLT panels. A typical CLT
cross section is shown in Fig. 2.1.(a) and has a typical stress distribution like a "saw
blade". Strength properties are based on tension strength values of the graded tim‐
ber boards used for the production. A model, which links tension strength values of
graded timber boards to the CLT bending strength values was presented in [6]. For
the application in practise the proposed formula is often simplified in a form
fmk,CLT=α∙fmk,g,0 with α = 1,1 to 1,2 and fmk,g,0 as bending strength of comparable glu‐
lam. This formula can be found in most of the approvals of relevant CLT producers.
Typical shear stress distribution can be seen in Fig. 2.1.b. The maximum shear stress
occurs at the centre of gravity of the underlying cross section. Shear verification has
to consider the corresponding strength values of each layer. In layers where rolling
shear stresses develop shear stress distribution is almost constant and the crucial
location of the rolling shear verification can be found in that layer, which is next to
the centre of gravity. The relevant rolling shear strength fv,90,k,CLT can be assumed
with 1,25 N/mm² if grooves in the lamellas are absent and geometric ratio of used
lamellas fulfils ratio of b/t ≥ 4 (b ... width and t ... thickness of the board).
P. Mestek [5, 11] investigated shear respectively punching resistance of CLT ele‐
ments under uniformly distributed and high concentrated loads. The used test con‐

164
 INTER / 48 - 12 - 01

figuration is shown in Fig. 2.2. The measurements of the tested CLT elements was
chosen with 1380/1020 mm which were simple supported at all four sides. The CLT
section was a 7‐layered element with an overall thickness of 189 mm. Based on
these geometrics a span to thickness ration of
L t CLT  a  b t CLT  1380 1020 189  6, 28 can be estimated. The work of Mestek fo‐
cused on the improvement of shear and punching resistance by the application of
self tapping screws drilled into the slab under 45° closed to the local load introduc‐
tion. CLT panels were investigated which were built up with grooves in the lamellas.
As a disadvantage of these grooves the rolling shear strength is significantly re‐
duced. Therefore a reinforcement with self tapping screws is a meaningful approach
for such CLT panels whereas the research presented in the following is concentrated
on the resistance of CLT panels without screws in CLT panels (except for the local
load introduction with screws under 90° of the layers).

(a) (b)
Figure 2.1. (a) typical bending stress distibution in CLT; (b) typical shear stress distibution in CLT

Figure 2.2. Test configuration used by Mestek [5, 11]

Although Mestek investigated primarily the effect of a reinforcement with screws in
CLT elements, some tests without 45° screws were conducted (Fig. 2.3.a). CLT ele‐
ments with grooves in lamellas were used. Additional tests were carried out which
compare the rolling shear strength of lamellas with and without grooves. Based on
these results a rolling shear strength for CLT panels built‐up with lamellas without
grooves could be estimated between 2,06 to 2,25 N/mm². It has to be mentioned
that the verification of the rolling shear was performed along a controlling perime‐
ter line or − written in terms of Mestek − along the intersection line (Fig. 2.3.b).

165
INTER / 48 - 12 - 01

(a) (b)
Figure 2.3. (a) Test results of Mestek without reinforcements with selftapping screws
(b) controlling perimeter for shear stresses [5, 11]

3 Numerical studies
The dominating failure mechanism (bending, shear) of CLT plates under a concen‐
trated load depends primarily on the geometric dimensions of the CLT plate (mean
span to thickness √(a∙b)/tCLT, ratio a/b of spans in x and y direction and the number
of layers and the relation of thickness of adjacent layers).
These parameters were investigated numerically by means of FEM in combination
with strength verifications based on the findings mentioned in Cap. 2. In a first step
the optimum a/b ratio (Fig. 3.1.a) was evaluated under the constraint condition that
the total area of the CLT plate should remain a∙b=1 m² and under consideration that
the rolling shear stresses should be the same for both directions x and y. The rolling
shear verification was carried out in the controlling perimeter line around the load
introduction. This controlling perimeter line was computed under an assumed load
spreading angle of 35° according to Mestek (Fig 3.1).

Figure 3.1. controlling perimeter line based on the findings of Mestek

This study was carried out with a 5‐ and 7‐ layered CLT cross section. The thickness
ratio of adjacent layers was 1:1 and 2:1 respectively. Based on symmetry conditions
only ¼ of the system was modelled as illustrated in Fig. 3.1 and Fig. 3.2.a (blue shad‐
owed area). The overall thickness of the CLT panel was chosen with 160 mm (Fig.
3.2.b).

166
 INTER / 48 - 12 - 01

Figure 3.2. (a) variation of a/b (b) investigated CLT sections worked‐out FEM study
The results with optimised 2∙a and 2∙b values are summarized in Table 1. One con‐
clusion of the done study is that 5‐layered CLT element with equal lamella thickness
does not behave well under two dimensional load transfer as an extraordinary a/b
ratio of 50,7/197,3 is the optimum with same rolling shear stress level in both direc‐
tions. 7‐layered CLT element behaves much better with the optimum a/b ratio of
105,6/94,7=1,11 which is close to a quadratic shape.
Table 1. optimum span length 2∙a and 2∙b
2∙a [cm] 2∙b [cm]
BSP5s_160mm_1:1 50,7 197,3
BSP5s_160mm_2:1 106,1 94,3
BSP7s_160mm_1:1 105,6 94,7
BSP7s_160mm_2:1 161,1 62,1

The a/b ratio (Mestek chose 138/102=1,35 in his work) differs from results of this
study. One important reason lies in the differences of the numerical model: In the
present study a 2D shell model was used while Mestek used a 3D solid model.
An other approach for a further study is to investigate differences of expected load
carrying capacities under deviating a/b ratios. Proposed a/b ratios for a second
study are shown in Fig. 3.3.

Figure 3.3. investigated a/b ratios

Results of computed FEM‐load carrying capacities are illustrated in Fig. 3.4. Based on
these it can be concluded, that the a/b ratio of the CLT elements is not as dominant
as expected.

167
INTER / 48 - 12 - 01

Figure 3.4. expected load carrying capacities for different a/b ratios based on a FEM‐study

4 Development of test configuration

Test specimens for the tests presented in the following should contain 5‐ and 7‐
layered CLT elements. Furthermore the chosen cross sections should be available as
industrially produced elements. The geometric as wells as stiffness values of the se‐
lected cross sections are summarized in table 2. Specimens denoted as “CLT 5s” and
“CLT 7s” were delivered from producer 1 while the specimen of series “CLT 7s*” orig‐
inated from a different producer 2.
Table 2. CLT cross sections used for experimental investigations
E0 E90 G0 G90 # t1 t2 t3 t4 t5 t6 t7
N/mm² N/mm² N/mm² N/mm² mm mm mm mm mm Mm mm

CLT 5s 11600 ≈0 690 50 5 34 30 34 30 34

CLT 7s 11600 ≈0 690 50 7 19 30 19 30 19 30 19
CLT 7s* 11600 ≈0 690 50 7 20 30 20 30 20 30 20
A graphical illustration of “CLT 5s” and “CLT 7s” and associated stress distributions
for unit section moments and shear forces in longitudinal (0°) and in transverse di‐
rection (90°) (mx=my=1 kN∙cm/cm, qx=qy=1 kN/cm) is given in Fig 4.1.

Figure 4.1. CLT cross sections used for experimental investigations and shear stress distributions

168
 INTER / 48 - 12 - 01

A sketch of the test configuration “punching” (test configuration 1) can be found in

Fig. 4.2. Ratio of a/b was chosen with 1,0 (a = b = 1,0 m). The supports were imple‐
mented by means of tension rods which could be constraint within an existing grid
of load introduction points with a distance of 50/50 cm in the laboratory. The span
to thickness ratio varied between 6,17 and 5,88 depending on the cross‐section. The
geometric boundary conditions were thereby similar to the mean span to thickness
ratio of Mestek configuration with 6,28. Three images of test configuration 1 are
shown in Fig. 4.3.

Figure 4.2. Test configuration for expected punching shear failure (test configuration 1)

Figure 4.3. Images of test configuration for expected punching shear failure (test configuration 1)
The test configuration “bending under concentrated loads” were designed with sig‐
nificant higher length to thickness ratio in order to achieve bending failure modes
(test configuration 2). Restricted to the grid of load introduction points of the test
field the longer span a was chosen with 4,0 m, the shorter span b with 2,5 m. Mean
length to thickness ratios were between 19,5 and 18,6 depending on the used cross
section. As the parameter study already showed that 5‐layered CLT elements be‐
have worse than 7‐layered ones under all side support with global bidirectional
bending behaviour. Therefore tests with one directional bending behaviour with
support only along both shorter sides was proposed for 5‐layered CLT elements
while a two directional bending loading (with simple supports on all four sides) of
the CLT elements was implemented for the 7‐layered CLT element (Fig. 4.4).

169
INTER / 48 - 12 - 01

Test specimens denoted with “A” are build up in the laboratory for series "CLT 5s"
and "CLT 7s". The lamellas used were graded in respect to density of the boards
within a range of 400 kg/m³ ± 20 kg/m³. Annual ring pattern was the second grading
parameter. The allowance of knots was reduced in comparison to industrially graded
boards. Test specimens with CLT elements from producer 1 are denoted with “B”. If
they came from producer 2 they are denoted as series “C”.

Figure 4.4. Test specimens for expected bending failure (test configuration 2)
Two images of test configuration 2 can be found in Fig. 4.5.

Figure 4.5. Images of test configuration 2 for test series with expected bending failure mode

5 Test results
The load‐displacement curves of the tested showed a relative ductile behaviour for
test configuration 1 (Fig 5.1) Different failure modes were observed. The non linear
characteristic indicates an increasing loss of shear strength which is more or less a
relative steady process. In some cases failure modes of lamellas in tension in con‐
junction with unsteady load‐displacement curves was observed.
For specimens of test configuration 2 bending failure always occurred first. Subse‐
quently rolling shear failures were observed. A ductile behaviour could be recog‐
nised also for test configuration 2 (Fig 5.2). As failure of several lamellas in tension
leaded to an immediate load decrease discontinuities in the load displacement curve
were developed more significant in configuration 2 than 1, especially in case of the
tests with 5‐layered CLT elements.

170
 INTER / 48 - 12 - 01

(a) (b)
Figure 5.1. load displacement curves (a) 5‐layered CLT; (b) 7‐layered CLT in test configuration 1

(a) (b)
Figure 5.2. load displacement curves (a) 5‐layered CLT; (b) 7‐layered CLT in test configuration 2
Failure of lamellas under tension was also observed in configuration 1. But the al‐
ready established non linear character of the load displacement curve indicates al‐
ready existing failures due to shear, whereas in configuration 2 unsteadiness always
finishes the linear load displacement behaviour.
Typical shear failures are shown in figure 5.3. These failures became visible by cuting
the test elements along their axis of symmetry after the tests. In the area of the load
introduction could be easily detected due to indentations of the steel plate and the
used reinforcement screws.

Figure 5.2. observed typical failures in test configuration 1

Figure 5.4 shows typical failures observed in test configuration 2. In the first row
failures of 5‐layered bending under more or less one dimensional bending could be

171
INTER / 48 - 12 - 01

detected whereas the images in the second row show impressive destruction under
pure two dimensional bending for the 7‐layered CLT specimens.

Figure 5.3. observed typical failures in test configuration 2

All numerical results concerning test configuration 1 can be found in table 3, in an
analogous manner table 4 contains the test data of test configuration 2.
Table 3. Test results of test configuration 1
Typ of CLT panel 5‐layered CLT panels 7layered CLT panels
Identification of test series A_D‐5s B_D‐5s A_D‐7s B_D‐7s C_D‐7s

Width of Longitudinal lay. [mm] 235 230 235 230 181 (90)1)
lamellas b Cross layer [mm] 120 190/120 120 190 184 (90)1)
Thickness Longitudinal lay. [mm] 34 34 19 19 20
of boards h Cross layer [mm] 30 30 30 30 30
Longitudinal lay. [‐] 6,91 6,76 12,4 12,1 9,05 (4,50)1)
ratio b/h
Cross layer [‐] 4,00 6,33/4,00 4,00 6,33 6,13 (3,00)1)
Number of test specimens
[‐] 6 4 6 3 4
(in evaluation)
Moisture Mean value [%] 11,7 11,3 12,1 10,0 11,0
content u COV [%] 6,40 9,90 5,88 1,33 3,06
Mean value [kg/m³] 414 435 412 460 431
density ρ
COV [%] 4,13 5,03 5,46 5,20 6,60

172
 INTER / 48 - 12 - 01

Mean value [kN/m 35,4 32,2 39,0 42,8 42,6

stiffness m]
COV [%] 4,28 3,80 4,40 7,29 6,90
Mean value [kN] 328 289 350 (400)2) 351
Force level COV [%] 6,90 5,76 4,17 (7,08)2) 9,42
at first
crack 5%‐Quantile NV [kN] 291 262 326 (353)2) 296
F1. crack charact. value
[kN] 279 247 311 (320)2) 273
acc. to EN 14358
Mean value [kN] 358 339 372 (408)2) 366
COV [%] 3,87 3,72 2,04 (4,70)2) 5,85
Maximum
of load Fmax 5%‐Quantile NV [kN] 335 318 360 (376)2) 331
charact. value
[kN] 318 296 331 (348)2) 313
acc. to EN 14358
Remarks:
1)
Values in brackets represent ratios of b/t under consideration of grooves.
2)
Appropriate values of series B_D‐7s represent bending failure.
Table 4. Test results of test configuration 2
5‐layered CLT 7‐layered CLT panels, support at all
Typ of CLT panel
panels sides
Identification of test series B_B‐5s B_B‐7s C_B‐7s
Number of test specimens [‐] 3 3 3
Moisture Mean value [%] 10,7 9,70 10,8
content u COV [%] 4,07 9,54 2,45
Mean value [kg/m³] 442 440 445
density ρ
COV [%] 10,8 4,48 2,72
Mean value [kN/m 4,78 11,0 11,2
stiffness m]
COV [%] 1,85 5,12 2,15
Mean value [kN] 238 312 361
Force level COV [%] 3,03 9,18 8,09
at first
crack 5%‐Quantile NV [kN] 226 265 313
F1.crack charact. value
[kN] 203 232 280
acc. to EN 14358
Mean value [kN] 253 347 369

Maximum COV [%] 5,10 4,25 5,84

of load 5%‐Quantile NV [kN] 232 323 333
Fmax
charact. value
[kN] 216 296 314
acc. to EN 14358

173
INTER / 48 - 12 - 01

6 Recommendations for standardisation

The results of test configuration 1 serve as basis for punching shear strength values.
Rolling shear strength values are evaluated with the maximum of elastic shear forc‐
es qx and qy along the controlling perimeter line (see Fig. 3.1). The test results of se‐
ries C (producer 2) are exceptional as lamellas of this series contains grooves. These
grooves are essentially smaller than those in [5]. A graphical comparison shows the
differences of the grooves in [5] and the present study (Fig. 6.1).

Figure 6.1. differences between grooves in Mestek [5] and test specimens of the presented tests
Based on forces defined at first crack level (see table 3) the associated stress values
for rolling shear could be identified and values are summarized in table 5.
Table 5. rolling shear strength values (based on results of configuration 1)
Typ of panel 5‐layered CLT panel 7‐layered CLT panel
Identification of test series A_D_5s B_D_5s A_D_7s B_D_7s C_D_7s
charact. value (EN 14358) 279 247 311 320 273
qx,max [kN/cm] 2,74 2,43 2,37 2,44 2,08
qy,max [kN/cm] 1,59 1,41 2,40 2,47 2,10
rolling shear str. τ90,x [N/mm²] 2,07 1,83 1,92 1,98 1,69
rolling shear str. τ90,y [N/mm²] 2,21 1,95*) 2,31 2,38*) 2,03
*) elements with a width of lamellas b=230 mm in the longitudinal layers and b=115 mm in the
cross layers; influence has to be evaluated
A shear strength value of 2,21 N/mm² could be computed (mean strength value:
2,62 N/mm²; COV: 6%, see table 3) with bold values in table 5 for series A and B.
Shear in y‐direction was dominating, whereas a mean stress level of about 85% de‐
veloped in x‐direction. Test specimens acc. to series C had grooves similar to test
specimens of Mestek, but in a less dominant implementation (see Fig. 6.1). As a con‐
sequence the strength reduction is quite moderate (9%).
The proposed strength value of 2,21 N/mm² ‐ a similar value can be calculated with
data from Mestek ‐ is only valid if stress computation is based on a pure elastic
method and rolling shear stresses have to be verified in the vicinity of concentrated
load introduction. In bending test according to the usual standards (e.g. EN 16351)
the characteristic rolling shear strength is significantly lower (for spruce: fv,k,90=1,25

174
 INTER / 48 - 12 - 01

N/mm²). The differences can be explained by the definitely non‐linear stress strain
behaviour (see following chapter). Therefore the proposed strength values in com‐
bination with elastic stress calculation is a simplification for practice.
Bending strength can be checked similarly with forces defined at first crack level (see
table 4). Results are summarized in table 6. Values for the specimens of series A are
missing as only industrial fabricated panels were tested with test configuration 2.
Table 6. bending strength values (based on results of configuration 2)
Typ of panel 5‐layered CLT panel 7‐layered CLT panel
Identification of test series A_B_5s B_B_5s A_B_7s B_B_7s C_B_7s
charact. value (EN 14358) — 203 — 232 280
mx,max [kN∙cm/cm] 110,18 61,54 74,27
my,max [kN∙cm/cm] 31,21 69,37 83,72
bending stress σx [N/mm²] — 30,95 — 22,17 24,99
bending stress σy [N/mm²] — 22,25 — 28,52 34,72
Grooves do not influence bending strength significantly. Therefore no differences
were made between series B and C. In case of 5‐layered CLT element the dominating
direction is longitudinal (x) whereas for 7‐layered elements the crucial direction is
the direction transverse (y). The test results showed that the limiting direction in
case of the 7‐layered CLT panels does not fit always with the numeric prediction.
Mean value over these test results (bold values in table 6) delivers 31,4 N/mm²
which is 9% higher than comparable bending strength values of CLT panels
(fm,k,CLT=1,2∙24=28,8 N/mm²). An increase of bending strength can be expected be‐
cause under a concentrated load only a small area is exposed to high bending mo‐
ments (volume effect).

7 FE‐model for punching shear configuration

It was already mentioned, that a numerical analysis with non‐linear behaviour de‐
scribes mechanical behaviour more precisely than a pure linear‐elastic model can
do. Such an analysis was carried out with the FE‐Software ABAQUS. A solid 3D‐
model for the 5‐layered punching shear configuration '1' was established for one
quarter of the test specimen taking into account two planes of symmetry, as shown
in fig. 7.1 (a). The numerical model was meshed with linear 3D solid elements
'C3D8'. The first axis of the material orientation denotes the orientation of the la‐
mellas (longitudinal direction), whereas third axis is parallel to the normal direction
of the CLT element. Axis 2 is perpendicular to axis 1 and 3 (see fig 7.1 (b)).
The orthotropic MOE parallel to grain was chosen with 12000 N/mm², perpendicular
to grain with 370 N/mm². The value of the shear modulus was 690 N/mm² and the
rolling shear value 50 N/mm². Stress‐strain curves based on data in [12] are shown
in fig 7.2 (a). The observed rolling shear behaviour is not ideal elastic‐plastic, but this

175
INTER / 48 - 12 - 01

assumption can be seen as a simplification for the FE‐model. The plastic plateau was
assumed to be 1,5 N/mm² (see fig 7.2 (a)).

(a)

(b)
Figure 7.1 (a) FEM Model for punching shear configuration (b) orientations of layers
For simplification all stress‐interactions between all stress components were ne‐
glected. The orthotropic elastic behaviour and the elastic‐plastic behaviour for roll‐
ing shear were to be implemented in a user subroutine for ABAQUS. A comparison
between computed and experimentally founded load displacement curves can be
seen in fig 7.2 (b).

(a) (b)

Figure 7.2 (a) stress strain behaviour for rolling shear, based on data of master thesis [12]
(b) FEM results with elastic‐plastic behaviour for rolling shear
The agreement is quite pretty good at the beginning where linear elastic behaviour
dominates. In the non‐linear part (load above 250 kN) the computational model de‐
scribes behaviour correctly in principal, but overestimates the mechanical strength
and resistance of the test specimen. This leads to the conclusion, that another fail‐
ure mechanism has to be implemented in the model. At the mentioned load level of
250 kN and above significant cracks can be seen in the experimental load displace‐
ment curves, which indicates, that softening due to tension in the outer lamellas is
missing in the model. This softening behaviour was supplemented in a small domain
of the numerical model in two outer lamellas. This domain is highlighted in a red
colour and shown in fig 7.1 (a). The size of this selected domain is equal to the
height of the outer lamella (34 mm), the width is set to 150 mm. Due to symmetry
conditions two lamellas can fail now due to tension in the numeric model in both di‐
rections. The location of failure was chosen in the middle of the test specimen,
where maximum bending stresses occur. The maximum tension stress, when soften‐

176
 INTER / 48 - 12 - 01

ing starts, was introduced with ft,mean =39,0 N/mm². This value bases upon various
CLT‐bending tests in our laboratory as a mean value. Descend of stress‐strain func‐
tion during softening depends on size of mesh. The area under the stress‐
displacement curve equals to the fracture energy release rate for Mode I, which was
introduced with a value of 1650 J/m², taken from [13]. The complete stress‐strain
curve with softening is shown in fig 7.3 (a) and added to the user subroutine for
ABAQUS. The comparison between computed and experimentally measured load
displacement curves is now much better than without softening. Two significant
steps in the computed load displacement curve indicates failure in y (my) and x (mx)
direction (see fig. 7.3 (b)).
(a)

(b)

Figure 7.3 (a) stress strain behaviour with softening for outer lamellas under tension (b) FEM
results with elastic‐plastic behaviour for rolling shear and tension softening in outer lamella
Shown computational results show, that the proposed high strength value of 2,21
N/mm² for rolling shear can be explained by elastic plastic behaviour of rolling shear
with some stress redistribution. Further improvements of the numerical simulation
could be an implementation of a more precise stress‐strain behaviour for rolling
shear as well as an implementation of stress interactions, in particular rolling shear
with compression stresses perpendicular to grain.

8 Acknowledgement
The research work within the project ‘focus_sts’ was financed by the competence
centre holz.bau forschungs gmbh. The project was financed by funds from the Fed‐
eral Ministry of Economics, Family and Youth, the Federal Ministry of Transport, In‐
novation and Technology, the Styrian Business Promotion Agency Association and
the Province of Styria (A12), the Carinthian Economic Promotion Fund (KWF), the
Province of Lower Austria Department of Economy, Tourism and Technology as well
as the Business Location Tirol.

9 References
[1] Bogensperger T. , Silly G., Schickhofer G. (2012): Comparison of Methods of
Approximate Verification Procedures for Cross Laminated Timber, Research Report,

177
INTER / 48 - 12 - 01

holz.bau forschungs gmbh, Institute for Timber Engineering and Wood Technology,
Graz University of Technology.
[2] Bogensperger T. , Silly G. (2014): Zweiachsige Lastabtragung von Brettsperr‐
holzplatten, Ernst & Sohn publisher for architecture and technical science GmbH &
Co. KG, Berlin. Bautechnik 91 issue 10.
[3] Silly G. (2010): Numerische Studien zur Drill‐ und Schubsteifigkeit von Brett‐
sperrholz (BSP), diploma thesis, Graz University of Technology.
[4] Bogensperger T. , Wallner B., Augustin M. (2015): Durchstanzen von Brett‐
sperrholzplatten, Research Report to project focus_sts 2.2.3_2, holz.bau forschungs
gmbh Graz, (report in progress).
[5] Mestek P. (2011): Punktgestützte Flächentragwerke aus Brettsperrholz (BSP) –
Schubbemessung unter Berücksichtigung von Schubverstärkungen, Ph.D., Munich
University of Technology.
[6] Jöbstl R. A., Bogensperger T., Moosbrugger T., Schickhofer G. (2006): A Contri‐
bution to the Design of Cross Laminated Timber, CIB W18, 39th Meeting, Florence (I)
[7] Schickhofer G., Bogensperger T. , Moosbrugger T. (2009): BSPhandbuch Holz‐
Massivbauweise in Brettsperrholz Nachweise auf Basis des neuen europäischen
Normenkonzepts, Graz University of Technology, 2011, ISBN 978‐3‐85125‐109‐8
[8] Peterson, L. A. (2008): Zum Tragverhalten nachgiebig verbundener Biegeträger
aus Holz, Berichte des Instituts für Bauphysik der Leibniz Universität Hannover Her‐
ausgeber: Univ.‐Prof. Dr.‐Ing. Nabil A. Fouad; Leibniz Universität Hannover ‐ Institut
für Bauphysik Heft 1
[9] Schelling, W. (1982): Zur Berechnung nachgiebig zusammengesetzter Biegeträ‐
ger aus beliebig vielen Einzelquerschnitten In: Ehlbeck, J. (publisher), Steck, G. (pub‐
lisher): Ingenieurholzbau in Forschung und Praxis. Bruderverlag Karlsruhe.
[10] Bogensperger T, Augustin M, Schickhofer G. (2011): Properties of CLT‐Panels
Exposed to Compression Perpendicular to their Plane, CIB W18, 44th Meeting, Al‐
ghero , Italy
[11] Mestek P, Kreuzinger H, Winter S (2011): Design concept for CLT ‐ reinforced
with self tapping screws, CIB W18, 44th Meeting, Alghero, Italy
[12] Ehrhart T (2014): Materialbezogene Einflussparameter auf die Rollschubeigen‐
schaften in Hinblick auf Brettsperrholz, Masterarbeit, Graz University of Technology.
[13] Mackenzie‐Helnwein P. et al (2005): Analysis of layered wooden shells using an
orthotropic elasto‐plastic model for multi‐axial loading of clear spruce wood, Com‐
puter methods in applied mechanics and engineering 194, p. 2661‐2685

178
 INTER / 48 - 12 - 01

Discussion

The paper was presented by T Bogensperger

F Lam received confirmation that the equation involved fr,90mean and the characteristic
strength was incorrect and should be modified. F Lam asked about the softening pro-
cedure used in FEM analysis. T Bogensperger provided some explanation and agreed
to add information in text of paper.
H Blass received clarification for the justification of punching factor k of 1.75 as possi-
ble localized effect where rolling shear failure was assumed not to be brittle. F Lam
asked about the deflection in compression perpendicular to grain. T Bogensperger
stated that the deflection was small because reinforcement screws were used. These
screws did not add to the shear strength.

179
180
 INTER / 48 - 12 - 02

Shear Properties of Cross Laminated

Timber (CLT) under in‐plane load: Test
Configuration and Experimental Study

Reinhard Brandner, Graz University of Technology *)

Philipp Dietsch, Technische Universität München *)
Julia Dröscher, Graz University of Technology
Michael Schulte‐Wrede, Technische Universität München
Heinrich Kreuzinger, Technische Universität München
Mike Sieder, Technische Universität Braunschweig
Gerhard Schickhofer, Graz University of Technology
Stefan Winter, Technische Universität München
*) joint first authorship

Keywords: cross laminated timber; CLT; shear in‐plane; diaphragm; test configura‐
tion; experimental study; parameter study; shear strength; shear modulus; failure
mechanisms; characteristic properties; design concept

1 Introduction
Cross laminated timber (CLT) is a two‐dimensional laminated engineered timber
product, commonly composed of an uneven number of orthogonally and rigidly con‐
nected layers. High resistances in‐ and out‐of‐plane predestines it for numerous ap‐
plications, e.g. for floor and wall elements, shear walls, folded panels and beams.
With respect to its resistances and properties as a structural product, it is differenti‐
ated between out‐of‐plane and in‐plane loading. For CLT under out‐of‐plane loading,
test configurations and characteristic values are well agreed. For CLT under in‐plane
loading, some properties are still under discussion, presently resulting in conservative
regulations, e.g. tension and compression in direction of the top layers. The same is
valid for CLT under in‐plane shear. To fully profit from the high capacities of CLT
in‐plane, a detailed knowledge of all relevant mechanical properties, which are

181
INTER / 48 - 12 - 02

dependent on the geometrical layup of the elements, as well as the development

and verification of practicable test configurations to determine these properties are
indispensable.
Consolidated knowledge of CLT properties under in‐plane shear is crucial for typical
structural applications such as wall and floor diaphragms, cantilevered CLT walls and
CLT used as (deep) beams, in all cases potentially featuring holes or notches. The cur‐
rent technical approvals for CLT products contain differing regulations to determine
their load‐carrying capacities in‐plane. Generally they imply a verification of the tor‐
sional stresses in the cross‐section of the cross‐wise glued elements as well as a veri‐
fication of the shear stresses proportionally assigned to the boards of the top and
cross layers. The basis of theoretical and practical considerations are the following
three basic failure scenarios for a CLT‐element under in‐plane shear: (i) gross‐shear
(longitudinal shearing in all layers), (ii) net‐shear (transverse shearing in all layers in
weak direction), and (iii) torsion failure in the gluing interfaces between the layers
(Bogensperger et al. 2007, 2010; Flaig and Blaß 2013; Brandner et al. 2013). All failure
mechanisms can be achieved if a corresponding test configuration is applied.
Properties for the mechanism (iii) “torsion”, based on Blaß and Görlacher (2002), Jeit‐
ler (2004) and Jöbstl et al. (2004) are well accepted (DIN EN 1995‐1‐1/NA). In con‐
trast, the determination of the properties (i) gross‐shear and (ii) net‐shear by testing
is challenging, as it is practically impossible to secure larger fields of pure shear. Up to
now, the properties for in‐plane shear provided in technical approvals are based on
testing single nodes. The resulting strength values are partly seemingly high and fea‐
ture a higher variability than expected for diaphragms. Associated investigations in‐
clude Wallner (2004), Jöbstl et al. (2008) and Hirschmann (2011). After re‐evaluating
and summarizing previous findings Brandner et al. (2013) propose fv,net,05 = 5.5 N/mm²
as 5 %‐quantile of net‐shear strength for a reference CLT node in conjunction with
the test configuration “EN” of Hirschmann (2011). Board thickness, gap width and
annual ring pattern were identified as parameters with significant influence on shear
resistance. Tests on single‐nodes are able to produce separated stress conditions,
hence all test configurations on single‐nodes can represent and lead to separate fail‐
ure mechanisms in CLT under in‐plane shear. The full stress state within a full‐scale
CLT‐element under in‐plane shear, however, cannot be represented by them.
Several efforts were made to determine shear properties on full‐scale CLT dia‐
phragms, e.g. Bosl (2002), Bogensperger et al. (2007) and Andreolli et al. (2014). The
main challenges within the tested configurations were – apart from their rather
costly implementation – (i) to realize a continuous load introduction, (ii) to receive a
field of pure shear and (iii) to achieve failure under in‐plane shear. It is expected that
these challenges are also encountered when applying the standardized test configu‐
ration which is used to determine the racking strength and stiffness of timber frame
wall panels, see EN 594 (2011). The determination of shear strength based on four‐
point bending tests, e.g. given in FprEN 16351 (2015) (based on CUAP 03.04/06 2005)

182
 INTER / 48 - 12 - 02

has to be critically analysed as well. Here, the determination of shear strength is

based on beam theory considering the total thickness of all cross layers in the evalua‐
tion. The typical stress states within CLT diaphragms under in‐plane shear are not
represented by this approach.
In the context of an approval in the individual case, Kreuzinger and Sieder (2013)
published a proposal for a test configuration and evaluation procedure for CLT
diaphragms. The principle approach to determine shear strength from a combined
stress state with transverse stresses can already be found in Szalai (1992). The ap‐
proach proposed by Kreuzinger and Sieder (2013) is based on a simple compression
test, the test results are evaluated using theoretical approaches from plate theory
(in‐plane stresses). The evaluation procedure is partly extended and specified in the
frame of this paper.

2 Test configuration and evaluation procedure

2.1 Description of test configuration
In this configuration, column‐shaped rectangular specimen, which are cut out under
45° rotated to the main orientation of CLT elements, are tested in compression, see
Fig. 1.
with:
xM direction of CLT top and middle layer (TL; ML)
yM direction of CLT cross layer (CL)
x x‐direction of the column
y y‐direction of the column
F load
A column cross‐section
α 45°
Figure 1. System.
2.2 Determination of in‐plane shear strength
The stresses on the column as well as on a differential CLT section are given in Fig. 2.
Based on the Cartesian coordinate system of the column cross‐section (x, y), the
principal stresses are:
0; ; ,

The shear stress at maximum load is determined according to Eq. (1), see Fig. 3.
1
, ∙ (1)
2 2∙ ∙

183
INTER / 48 - 12 - 02

Figure 2. Stress states in column (left) and in differential CLT‐section Figure 3. Internal stresses
(right) and external loading.
In general, the shear resistance is influenced by stresses perpendicular to the grain,
see Spengler (1982) and Hemmer (1984). Compressive stresses perpendicular to the
grain result in an increase of shear resistance. In the given test setup (see Fig. 1), the
obtained shear stresses τxM,yM are higher than the actual shear strength fv. The test
setup leads to compressive stresses σxM and σyM, which equal the shear stresses
τxM,yM, see Fig. 3. In cases of gross shear failure, the compressive stresses σyM are
primarily transferred by the layers featuring a board direction yM. The compressive
stresses perpendicular to the grain on the layers with a board direction xM will fea‐
ture a magnitude, which is reduced by the relationship
⁄ with:
EyM weighted modulus of elasticity in yM‐direction, CLT cross layer
(value standardized or determined by testing (preferred))
E90 modulus of elasticity perpendicular to the grain of the base material
top layers (value standardized or determined by testing)
The approach
∙ ∑ , ∙ ∑ , ∙ with ∑ , ∑ , (2)
and ∑ , ∑ ,

and E0 modulus of elasticity

parallel to the grain of the
leads to base material
∑ , ∙ ∑ , ∙ (3)

with the relationship

(4)
∙ , ∙

184
 INTER / 48 - 12 - 02

Assuming softwood of typical strength classes according to EN 338 with (C16 to C30)
and layup parameters (ratios between the sum of layer thicknesses in weak direc‐
tion, ∑tℓ,T, to that in the strong direction, ∑tℓ,L) of 0.25 ≤ ∑tℓ,T / ∑tℓ,L ≤ 1.0, this leads to
values σ90 = τxM,yM (0.06 to 0.25) and to σ90 = τxM,yM ∙ (0.07 to 0.17) for C24.
Using the test results reported in Spengler (1982), an attempt to estimate this influ‐
ence is given by the approach taken by Blaß & Krüger (2012), based on results of
Spengler (1982), which can be modified as follows:
, , 1.15 ∙ 0.13 ∙ (5)
whereby σ90 is negative if representing compression stresses.
To determine the shear strength fv,gross, the obtained shear stresses τxM,yM should be
reduced in the range of fv,gross = τxM,yM ∙ (0.75 to 0.94) (C16 to C30) and fv,gross = τxM,yM
(0.83 to 0.93) (C24 only). The higher the layup parameters, ∑tℓ,T / ∑tℓ,L, the smaller
the reduction.
In case of a net‐shear failure in principle the same considerations can be made. In
doing so the layers relevant for transferring compression perpendicular to grain
stresses change and the number of layers which fail in transverse shear is equal to
the number of layers in the weak direction of the CLT element. Consequently, the
following relationships apply:
∑ , ∙ ∑ , ∙ with:
ExM weighted modulus of
elasticity in xM‐direction, CLT (6)
top layer (value standardized
or determined by testing)

∙ , ∙ (7)

, , ∙ 1.15 ∙ 0.13 ∙ with: ∑ , (8)

The resistances in net‐ and gross‐shear in case of gross‐ and net‐shear failure, re‐
spectively, can be calculated by considering the relevant ratio between ∑ , and
.

2.3 Determination of in‐plane shear stiffness

The shear modulus G can be determined using the flexibility matrix and its transfor‐
mation. Using the constitutive Eq. ∙ , the flexibility matrix describing the state
of plane stress

185
INTER / 48 - 12 - 02

1
0 0
1
, 0 0 (9)
1
0 0
,

can be transformed from the coordinates xM, yM to x, y by the angle 360° – α = 315°,
see Eq. (10).
0.25 0.25 0.25 0.25 0.25 0.25 0.5 0.5
, ,
0.25 0.25 0.25 0.5 0.5
, (10)
,
1 1
.

From the load‐deformation characteristics of the column‐section, the load F and the
modulus of elasticity Ey can be determined.
For a discrete stress state σy with associated strain ε and using the constitutive Eqs.
∙ and ∙ , the following relationship, Eq. (11), can be found:
1 1 1 1 with: Ey modulus of elasticity in
0.25 ∙ y‐direction of the column‐
,
section (determined by test)
(11)
ExM, EyM weighted modulus of elasticity
in xM‐ or yM‐direction,
CLT top or cross layer
The shear modulus GxM,yM can then be determined according to Eq. (12).
1
, 4 1 1 (12)

First tests at the Technische Universität München (TUM) and Graz University of
Technology (TU Graz) in 2013 indicated the functional and operational efficiency of
the test configuration. Motivated by these promising results, a joint research project
between TUM and TU Graz was initiated with the aim
 to prove the applicability and suitability of the test configuration for a wider range
of parameter settings,
 to investigate and quantify possible influences on the shear properties, and
 to answer the open question on a possible transfer from single‐node outcomes
to CLT diaphragms.

186
 INTER / 48 - 12 - 02

3 Materials and methods

3.1 Test programme
The test programme was developed in consideration of all relevant product parame‐
ters and their range found in current European Technical Approvals (ETAs) of CLT
products. Only CLT elements with glued surfaces were investigated. Tab. 1 contains
an overview of the tested parameters and their range of values. The parameters of
each series are given in Tab. 2. Fig. 4 shows the scheme of a specimen featuring 5
layers including a notation of some parameters used throughout the text.
tℓ,TL

t  ,L   t  ,TL  t  ,ML
tℓ,ML tℓ,CL

t   t  ,CL
tCLT

 ,T

t ,fail  t ,CL

wℓ wgap
Fig. 4. Schematic drawing of a 5‐layer CLT‐element cross section and notation of some parameters.

Table 1. Overview of tested parameters and their values (range).

Parameter [‐] Values [‐]
Gap execution edge bonded (EB); not edge bonded, gap width wgap = {0; 5} mm
Board width wℓ = {80; 160; (230) 240} mm
Layer thickness tℓ = {20; 30; 40} mm
Number of layers {3; 5; 7} layers
Stress reliefs {Yes; No}
Layup parameter ∑tℓ,T / ∑ℓ,L = {0.32; 0.35; 0.46; 0.50; 0.68; 0.75; 0.86},
with ∑tℓ,T ≤ ∑tℓ,L
Producer {A; B; C}

Only CLT from Norway spruce (Picea abies) was used which was provided by three
producers, leading to three groups of specimen, A, B and C. For the boards used for
group A, strength class C24 according to EN 338 was agreed. The boards for all series
within group A were delivered in one stack with the exception of the boards of series
A4 and A5, which were delivered at a later stage. Due to production limits at the
producer, series A1 and A3 were produced at the laboratories at TU Graz, see
Dröscher (2014) for further details. All specimen within groups B and C were
produced according to the specific Technical Approvals of the producers. These
allow the use of boards of strength class C16 according to EN 338 at a share ≤ 10 %.

187
INTER / 48 - 12 - 02

Table 2. Test programme; overview of test series including all necessary parameters

188
 INTER / 48 - 12 - 02

The series in groups A and B consisted of 6, the series in group C of 7 specimen.

The specimen were generally retrieved consecutively from one CLT plate. Thus it is
expected that the variability of the parameters within one series is reduced due to
partly the same base material within this series. To evaluate this influence, a stochas‐
tic simulation was conducted, see Section 4.1.
3.2 Test configuration
The test configuration was realized according to the configuration described in
Section 2. The geometric relationship was set to hCLT / wCLT = 3 / 1, more specifically
to hCLT / wCLT = 1,500 mm / 500 mm. This effectuated a field of pure shear outside the
quadratic area potentially influenced by the support conditions while eliminating the
potential for stability failure in most configurations. The assumption of a field of
constant shear was verified by means of a Finite‐Element (FE) study, in which
geometric and stiffness parameters were varied in a practical range, see Silly (2014)
for further details. The potential influence of friction between the support (surface
of load application) and the test specimen was investigated by using (i) lubricated
edges, (ii) teflon intermediate layers, (iii) roller bearing and (iv) blank steel to wood
contact. The differences in determined transverse strains were evaluated by
measurements of the horizontal deformation near the load application and found
to be not of practical relevance. All tests within group A were realized using Teflon
intermediate layers, all tests within groups B and C were conducted with a roller‐
bearing at the bottom support and steel plate to wood contact at the load introduc‐
tion. In all cases, the load was applied at a constant rate to achieve failure within
300 ± 120 s according to EN 408 (2010). The tests within group A were realized in the
4 MN four‐column test frame of the Laboratory for Structural Engineering (LKI) at TU
Graz. All tests of groups B and C were conducted in the Zwick Z‐600 testing machine
at the MPA BAU at TUM. In case of very slender test specimen, one horizontal
support was added to each side face of the specimen to prevent premature buckling.
The deformation was determined on both side faces of the specimen using centrically
placed measurement crosses featuring a measuring distance of h0 = 400 mm. For this,
the specimen of group A were equipped with DD1 strain transducers, which were
removed at approximately 50 % of Fmax. The specimen of groups B and C were
equipped with rope extensometers on one side face. On the other side face, the con‐
tact‐free optical measurement system GOM with software Pontos (2007) was used.
3.3 Determination of parameters
3.3.1 Moisture content and density
For each specimen, the mean density as well as the mean moisture content (group A:
kiln drying, groups B and C: resistance method) were determined. In the case of mois‐
ture contents differing from the reference moisture content uref = 12 %, the mean
density at 12 % moisture content, ρ12, was determined according to EN 384 (2010).

189
INTER / 48 - 12 - 02

3.3.2 Shear strength and torsional stresses

The shear strength in case of gross‐ and net‐shear failure was determined according
to Eq. (5) and (8), respectively. The (low) influence of compressive stresses perpen‐
dicular to the grain on shear strength was taken into account using the regression
formula from Blaß & Krüger (2012), applying compressive stresses perpendicular to
the grain determined with Eqs. (4) and (7).
The torsional stresses ∗ , at the time of failure in gross‐ or net‐shear were deter‐
mined on the basis of polar torsion, considering a finite number of layers N and a
heterogeneous layup with thicknesses , by establishing ideal layer thicknesses ∗,
to take into account bonded areas in the outer and core region of the CLT‐elements,
with
∗ ∗ ∗
, 2∙ , ; , resp. , , ; 2∙ , and , (13)
, ; ,
with: , as thickness of the layer i = 1, 2, …, N, and the relationship
∗
∗ ,
, 3∙ , ∙ , (14)
see Bogensperger et al. (2010). To determine the shear strength at the reference
moisture content uref = 12 %, a relationship of 3 % per percent change in moisture
content was applied.
3.3.3 In‐plane shear stiffness of the CLT elements
The shear modulus G090,CLT of the CLT elements under in‐plane shear was determined
with two approaches. The first approach, described in Section 2, is based on the
measured vertical deformation in the local measurement field and moduli of
elasticity E0,mean and E90,mean, standardized according to the strength class of the
boards, considering a strength class of C24 according to EN 338 (2009), with
E0,mean = 11,000 N/mm² and E90,mean = 370 N/mm². The shear modulus
G090,CLT,xM,yM,KS = G090,KS is determined according to Eq. (12).
The second approach applied is standardized in EN 408 (2010) with
, , , , , with: (15)
∆ ⁄
∙ h0 measurement length
∙ ∆
∆F/∆wG relationship between load and shear
deformation, determined in the linear
elastic range between 0.1 and 0.4 Fmax
With aid of a Finite‐Element study it could be shown, that the differences between
ideal and real stress distribution are negligible for given geometric and stiffness
relationships (< 1 %), hence no correction factor αG was applied, see Dröscher (2014)
for further details. To determine the shear moduli at reference moisture content
uref = 12 % a relationship of 2 % per percent change in moisture content was applied.

190
 INTER / 48 - 12 - 02

4 Results and Discussion

4.1 General
Three groups with a total of 18 series featuring different product configurations were
tested. The statistics of the main parameters in each series are illustrated in Tab. 3.
The statistical analysis as well as stochastic simulations were carried out in R (2015).
The moisture content u of all specimen was in the range of 12 ± 2 %. Regarding the
density ρ12 it can be noted that it decreases from group A to C (A: 463, B: 437,
C: 419 kg/m³). Only series B5 exhibited a density, which is below the expected range
for series within group B. Deriving all specimen from the same CLT element was
concluded to be the reason for low CVs in density.
The low CVs of the shear strength (2 % ≤ CV[fv,net,12] ≤ 8 %) in combination with the
very reliable failure in gross‐shear respectively net‐shear, independent of the multi‐
tude of parameters and their range, affirmed the very robust test configuration.
No differences in results were identified between both test institutes as well as the
utilized testing machines. However, as mentioned above, these low CVs are biased
by the applied sampling approach. Based on a stochastic simulation, conducted by
considering parallel and serial interaction of nodes and sections of lamellas in the test
area, CV[fv,net] is estimated to be approximately 6 %.
In contrast to common expectation, shear moduli feature higher CVs than the shear
strength. This is attributed to the known difficulties in deriving distinct values from
deformation curves, which are the result of measurements of very low deformations.
4.2 Shear modulus
A comparison of the shear moduli determined with above given approach Eq. (12)
and the approach given in EN 408 (2010), Eq. (15), shows that the values determined
with latter approach are on average about 10 % higher. The reason is the considera‐
bly higher vertical deformation in comparison to the horizontal deformation. The
approach by Kreuzinger und Sieder (2013), Eq. (12), only takes into account the
vertical deformation. Furthermore, the application of standardized values for E0,mean
and E90,mean leads to higher CVs for shear moduli compared to shear moduli deter‐
mined according to EN 408 (2010). It should be discussed how both approaches could
be adapted to better eliminate the influence of deformations from other stresses
than shear stresses. For the time being, the approach according to EN 408 (2010)
is preferred as it returns more stable results.

191
INTER / 48 - 12 - 02

Table 3. Statistics of tested series: moisture content, density, maximum load, apparent fracture
deformation, shear strength, shear moduli, torsional stresses

192
 INTER / 48 - 12 - 02

Table 3 contains also values G090,CLT,mean,est calculated with the formalism given in Bo‐
gensperger et al. (2010), see Eq. (16)
,, ,
, , , , with ∙ and , , (16)
,
∙ ∙

with G0,l,mean as average shear modulus of the lamellas, p and q as parameters of func‐
tion αT, see Tab. 4. Compared to G090,EN,12,mean overall congruent shear module, with
deviations within ± 10 % and only for some series of ± 20 %, are found, apart from A3.
Table 4. Parameters p and q for αT from Dröscher (2014).
No. of layers N [‐] p [‐] q [‐]
3 0.53 –0.79
5 0.43 –0.79
7 0.39 –0.79

Figure 5. (left) load‐displacement curves of series A1 (with) & A2 (without edge bonding);
(right) typical impressions of net‐ and gross‐shear failure mode.
4.3 Shear strength
All specimen within series A1, featuring edge bonded boards, failed in gross‐shear.
All specimen without edge bonding failed in net‐shear. The failure in gross‐shear was
followed by a failure in net‐shear and corresponding softening to a plateau of about
30 – 60 % of net‐shear strength, see Fig. 5. In contrast to gross‐shear failure, net‐
shear failure exhibited a considerable proportion of non‐linear deformation. All
series without edge bonded boards failed due to a net‐shear failure in the cross
layer(s) with the exception of series A4 in which most specimen exhibited a failure in
direction of the top layer. The mean vertical deformations at time of failure feature,
independent of the type of failure, a low range (7.7 mm ≤ wf,app,mean ≤ 9.5 mm). Two
specimen within series B1 experienced a stability failure (second eigenmode due to
horizontal support) before net‐shear failure. A comparison to the strength values of

193
INTER / 48 - 12 - 02

the other specimen within that series did not show any influence of stability failure
on shear properties. Fig. 6 shows the net‐shear strength of individual series arranged
by certain parameters to enable examination of parameters relevant for shear
strength. In the following sub‐sections, these parameters will be discussed with
respect to their influence on shear strength.
gap exec. wℓ tℓ layers stress relief ∑tℓ,T / ∑tℓ,L producer
param. EB 0 5 80 160 240 20 30 40 3 5 7 N Y N Y N Y .32 .35 .46 .50 .68 .75.86 A B B C
wℓ 160 80 160 240 160 240 160 230 160 230 160 160 240230
a 1) 160 80 160 120 160 120 160 230 115 160 80 230115 160 160 120115
tℓ,fail 2) 30 20 40 20 30 40 30 30 30 40 20 30 20 30 40
tℓ,fail/a .18 .25 .13 .33 .13 .19 .33 .19 .13 .26 .19 .38 .17 .35 .13 .19 .13 .19 .17
1) a = min (wℓ; trelief), trelief as edge distance of relief
2) thickness of failed layer(s), rounded to 10
fv,net,12,05,LND
fv,net,12 [N/mm²]

12
10
8
6
A3

A5
A9

A8
A6
A7

A4
A9

B3
A1
A2

A2

A2
A5

A2
B1
B2
B5
B2
B3
B5

B4

B5
B3
C1
C2

C3
C4

C4
Series [‐]
Figure 6. Box‐plots of net‐shear strength for identification of relevant parameters.
4.3.1 Gap execution
Series A1, A2 and A3 were used to analyse the influence of gap execution. The edge
bonded specimen A1 exhibited increased stiffness and a failure in gross‐shear, fol‐
lowed by failure in net‐shear. The parameters determined for series A1 fv,gross,12,05 =
3.8 N/mm2 and G090,EN,12,mean = 650 N/mm2 are comparable to those of glulam GL24h
with, according to EN 14080 (2013), fv,g,k = 3.5 N/mm2 and Gg,mean = 650 N/mm2. Se‐
ries A2 and A3, without edge bonding, like all other remaining series, failed in net‐
shear. The shear strength of series A2 and A3, compared to the edge bonded series
A1, is almost halved. The lower shear parameters of series A3 in comparison to series
A2 can mainly be attributed to the unintended but common edge bonding of CLT
with closed gaps due to the penetration of glue from the side faces into the gaps be‐
tween the boards during the production process. Another effect is the activation of
friction between the boards in contact. Current technical approvals allow for gaps be‐
tween 4 and 6 mm. The resulting reduction in cross‐section is, however, negligible for
practical applications (< 5 %). Higher shear properties could be attributed to CLT ele‐
ments with closed gaps and/or edge bonding. This implies however that the closed
gap is preserved throughout the lifetime of the structure. Cracks due to climatic
changes are at least to be expected in the top layers.

194
 INTER / 48 - 12 - 02

4.3.2 Board width

To analyse the parameter board width wℓ respectively gap or relief distance a, the
results of series B1 and B2 were used directly; series B5, due to its board thickness
and stress relief, could be used to a limited extent. Taking into account the pro‐
nounced influence of the parameter board thickness (see Section 4.3.3), the results
given in Fig. 6 and Tab. 3 indicate a regressive relation between board width and
shear parameters.
Jöbstl et al. (2008), Hirschmann (2013) and Brandner et al. (2013) state that the fail‐
ure in net‐shear happens as a result of a local interaction of torsional and longitudinal
shear failure at the board edges. From this it can be expected that increasing board
width and hence decreasing torsional stresses due to a decreasing relation (tℓ / a) has
a positive effect on shear strength. On the other hand, wider boards are usually cut
close to the core of the log, leading to an increased proportion of rift or half‐rift cuts.
The shear strength in the longitudinal‐tangential plane, fv,LR, is lower than in the longi‐
tudinal‐radial plane, fv,LT (see e.g. Keenan et al. 1985, Denzler & Glos 2007, Brandner
et al. 2012). With respect to knots and checks, a reciprocal relation is expected. Due
to the very local formation of failure, the influence of these timber characteristics is
expected to be low. Taking into account the very heterogeneous densities of the
series compared and the comparable outcomes of C1 vs. C3 and C2 vs. C4, both
pairs without and with reliefs, no clear influence of the board width can be derived.
In accordance with the results from tests on single CLT nodes (Brandner et al. 2013),
the influence of board width on the shear parameters is evaluated as low, thus it is
proposed to disregard this parameter for practical applications.
4.3.3 Board (layer) thickness
This parameter was evaluated by comparison of series B2 & B3 (to a limited extent
also B5) as well as C1 & C3 and C2 & C4. With increasing layer thickness, a distinct
decrease in net‐shear strength could be identified. This is in accordance with results
from tests on single CLT nodes (Brandner et al. 2013). This result can be attributed to
the locking effect due to the orthogonal arrangement of layers as well as the tenden‐
cy of thicker boards to feature an increased proportion of wood prone to fail in the
longitudinal‐tangential plane, featuring a lower shear strength, fv,LR, see Section 4.3.2.
Another potential effect is the size effect of wood under shear, i.e. area available in
which shear failure (e.g. cracking) can develop. The shear properties of the series
within group A were lower compared to the results of series within groups B and C.
However, the relative differences between series featuring board thicknesses
tℓ = 20 mm and 30 mm were comparable. A comparison of the series within group B
showed that the shear strength of series B5 is unexpectedly low, accompanied by
very low densities and unexpectedly high CV. This series is therefore disregarded
when determining characteristic shear strength.

195
INTER / 48 - 12 - 02

4.3.4 Number of layers

A comparison of series A2 (3 layers), A5 (5 layers) and A9 (7 layers) showed, inversed
to the density, a slightly concave relationship between the shear strength, fv,net,12, and
the number of layers N. It should be noted that the boards used within series A5
were delivered at a later stage, a corresponding influence cannot be excluded. Due to
the relative small differences between the series, the parameter number of layers is
evaluated negligible for practical applications.
4.3.5 Stress relief
For an assessment of this parameter, three pairs of series with / without stress relief
were available. Due to the local interaction of shear and torsional stresses in the case
of net‐shear failure, it was expected that higher relationships of (tℓ / a) lead to lower
shear properties. Apart from one exception, only small differences could be found in
this comparison. With respect to building practice and regarding the potential ques‐
tion of how to define individual shear parameters for CLT with stress reliefs, it is
proposed to disregard this parameter.
4.3.6 Layup parameter
The layup parameter (ratio between the sum of layer thicknesses in weak direction to
that in the strong direction) of all tested series featuring layer thicknesses tℓ = 20, 30
and 40 mm was in the range of 0.25 to 1.00. The results of the series of group A,
grouped according to the thickness of the failing layer, tℓ,fail, show a progressive trend
of gross‐shear strength fv,gross,12 while the net‐shear strengths, fv,net,12, were rather
constant for given layer thickness, tℓ,fail. Series A4 exhibited comparatively low net‐
shear strengths. In this series, not the cross but the top and middle layers failed. CLT
elements with ratios close to 1.0 can exhibit failure of the top and middle layers,
series A4 featured a comparatively high ratio of 0.86. It is expected that the missing
locking effect at the outer side of the top layers leads to a decreasing shear strength
in the magnitude of about one thickness class.

5 Design proposal
The results of the test series described in the preceding sections show that the main
parameters influencing the shear properties are the layer thickness (decreasing
properties with increasing thickness) and the gap execution (edge bonded, not edge
bonded and without / with gaps, with decreasing properties in mentioned order).
The distinct relation between layer thickness and net‐shear strength leads to a
dependency of the gross‐shear strength on the layup parameter (ratio between
sums of layer thickness). Therefore the most practical approach would be to define
a verification concept based on the net‐shear strength and the associated layers
prone to fail. Such a concept would allow for a design independent of the above
mentioned layup parameter. In addition, it would mirror the approach applied for the
verification of longitudinal stresses in CLT elements under in‐plane loads. In case of

196
 INTER / 48 - 12 - 02

CLT‐elements with a layup parameter ≥ 0.8, indicating a potential failure of the

top and middle layer(s), verification of net‐shear has to be met for both diaphragm
directions. In doing so, a reduced shear strength of the top layers, following the
approach in Section 4.3.6, shall be considered.
For CLT elements with expected gross‐shear failure, a verification on the basis of
gross‐shear strength and assuming the full element width is feasible. Due to the
longitudinal shear failure of all layers in edge bonded CLT elements, the dependency
on the layup parameter is expected to be low and not of practical relevance. This
implies however that the closed gap is preserved throughout the lifetime of the
structure. Cracks due to climatic changes are at least to be expected in the top layers.
The approach given in EN 1995‐1‐1+A1 (2008), implying the reduction factor kcr to
take into account shrinkage cracks in glulam, could be translated to edge bonded CLT
elements. Following this approach, the cross‐section utilized for verification would be
reduced by a certain proportion of the top layer thickness, hence by considering only
30 to 50 % of tℓ,TL. However, additional investigations to better quantify this approach
are required. Securing the full potential utilization of the core layers over the lifetime
of the structure implies as well, that the load‐carrying capacity of the edge bond, i.e.
the certified applicability of the utilized glue and the correct execution of the bond,
is ensured and controlled.
For CLT‐elements that are expected to fail in net‐shear, the verification of torsional
stresses, i.e. the potential failure between two layers in the vicinity of the glued
bond, has to be met, in addition to the verification of net‐shear. Following Schickho‐
fer et al. (2010), i.e. considering a characteristic torsional strength fv,tor,k = 2.5 N/mm²,
in combination with the values for fv,net,k presented in this paper, it can be concluded
that the torsional failure mechanism can potentially govern only in cases of CLT
diaphragms featuring a ratio between board thickness to board width / distance of
reliefs, tl / a or tl / wl, exceeding 0.25.

6 Conclusions
The new shear test configuration was successfully applied to the full spectrum of
tested configurations, demonstrating its functional and operational efficiency and
reliable shear failures of all tested CLT diaphragms. Consequently, we propose this
test configuration for implementation in EN 16351. Regarding the investigated pa‐
rameters, qualitatively congruent results to experiences made on single node tests
were achieved. This comprises the influence of gap width and board or layer
thickness. All specimen without layers of edge bonded boards failed in net‐shear.
For CLT‐elements that are expected to fail in net‐shear, a design concept based on
the net‐shear strength of the layers in the weaker direction is proposed in combina‐
tion with a net‐shear strength fv,net,k,ref = 5.5 N/mm2. Here, layer thicknesses up to
40 mm and gap widths up to 6 mm are taken into account. For layers in weak direc‐
tion with thicknesses between 20 mm ≤ tl,fail < 40 mm and without gaps or reliefs

197
INTER / 48 - 12 - 02

higher strength values are expected. Also taking into account the results from single
CLT nodes (Brandner et al. 2013), a relationship fv,net,k = fv,net,k,ref ∙ min{(40 / tℓ,fail) 0.30; 1.20}
is proposed. The shear modulus can be determined according to Eq. (16) (Bogensper‐
ger et al. 2010). For simplification a value of G090,mean = 450 N/mm2 is proposed. In
case of CLT elements with a layup parameter ≥ 0.8, the net‐shear strength of both
directions of layers has to be verified. The reason is the potential failure of the
weaker top layers. The lower shear strength of the top layers can be taken into
account using the approach given in Section 4.3.6.
In case of edge bonded specimen, gross‐shear failure, followed by net‐shear failure
was observed together with significantly higher resistances and shear moduli. For
such elements, the shear properties known from glulam, fv,gross,k = 3.5 N/mm2 and
G0mean = 650 N/mm2 are proposed. This necessitates, however, the consideration of
potential influences during the lifetime of the structure, e.g. crack formation and
delamination, in the design and production process, see Section 5. Further research
could include a comparison of shear properties of intact edge bonded specimen to
edge bonded specimen featuring pronounced shrinkage cracking. In addition to the
verification of CLT diaphragms in gross‐ or net‐shear, the verification of the torsional
stresses, as third potential failure mechanism, is required in cases of CLT diaphragms
prone to fail in net‐shear and featuring a ratio tl / a or tl / wl, exceeding 0.25.

7 Acknowledgement
This research project originated from a cooperation project between the holz.bau
forschungs gmbh, in the frame of the FFG COMET K‐Project „focus_sts“, the Graz
University of Technology, Institute of Timber Engineering and Wood Technology and
the Technische Universität München (TUM), Chair of Timber Structures and Building
Construction, in cooperation with the Glued Laminated Timber Research Association
inc., Wuppertal. The support by the funding bodies and project partners as well as
the funding of a short‐term scientific mission in the frame of COST Action FP1402 is
gratefully acknowledged.

8 References
Andreolli, M, Rigamonti, M, A Tomasi, R (2014) Diagonal compression test on cross
laminated timber panels. WCTE, Quebec, Canada.
Blaß, H‐J, Görlacher, R (2002) Zum Trag‐ und Verformungsverhalten von Brettsperr‐
holz‐Elementen bei Beanspruchung in Plattenebene: Teil 2 (in German). Bauen mit
Holz, 12:30–34.
Blaß, H‐J, Krüger, O (2010) Schubverstärkung von Holz mit Holzschrauben und Ge‐
windestangen, Karlsruher Berichte zum Ingenieurholzbau, Band 15, Universitäts‐
verlag Karlsruhe.
Bogensperger, T, Moosbrugger, T, Schickhofer, G (2007) New test configuration for
CLT‐wall‐elements under shear load. CIBW18/40‐21‐2, Bled, Slovenia.

198
 INTER / 48 - 12 - 02

Bogensperger, T (2008) A contribution to the characteristic shear strength of a CLT

wall under shear. 3rd Workshop, COST E55, Espoo, Finland.
Bogensperger, T, Moosbrugger, T, Silly, G (2010) Verification of CLT‐plates under
loads in plane. WCTE, Riva del Garda, Italy.
Bosl, R (2002) Zum Nachweis des Trag‐ und Verformungsverhaltens von Wandschei‐
ben aus Brettsperrholz (in German). Military University Munich, Munich.
Brandner R, Gatternig W, Schickhofer G (2012) Determination of Shear Strength of
Structural and Glued Laminated Timber. CIB‐W18/45‐12‐2, Växjö, Sweden.
Brandner, R, Bogensperger, T, Schickhofer, G (2013) In plane Shear Strength of Cross
Laminated Timber (CLT): Test Configuration, Quantification and influencing Param‐
eters. CIB‐W18/46‐12‐2, Vancouver, Canada.
CUAP 03.04/06 (2005) Common Understanding of Assessment Procedure: Solid wood
slab element to be used as a structural element in buildings. OIB, Wien.
DIN EN 1995‐1‐1/NA (2013) National Annex – Nationally determined parameters –
Eurocode 5: Design of timber structures – Part 1‐1: General – Common rules and
rules for buildings. (DIN).
Dröscher, J (2014) Prüftechnische Ermittlung der Schubkenngrößen von BSP‐
Scheibenelementen und Studie ausgewählter Parameter (in German). Master The‐
sis, Graz University of Technology, Graz.
EN 338 (2009) Structural timber – Strength classes. (CEN).
EN 384 (2010) Structural timber – Determination of characteristic values of mechani‐
cal properties and density. (CEN).
EN 408+A1 (2010) Timber structures – Structural timber and glued laminated timber
– Determination of some physical and mechanical properties. (CEN).
EN 594 (2011) Timber structures – Test methods – Racking strength and stiffness of
timber frame wall panels. (CEN).
EN 1995‐1‐1+A1 (2008) Eurocode 5: Design of timber structures — Part 1‐1: General
— Common rules and rules for buildings. (CEN)
EN 14080 (2013) Timber structures – Glued laminated timber and glued solid timber
– Requirements. (CEN).
Flaig, M, Blaß, H J (2013) Shear strength and shear stiffness of CLT‐beams loaded in
plane. CIB‐W18/46‐12‐3, Vancouver, Canada.
FprEN 16351 (2015) Timber structures ‐ Cross laminated timber – Requirements.
CEN.
Hemmer, K (1984) Versagensarten des Holzes der Weißtanne (Abies Alba) unter
mehrachsiger Beanspruchung, Dissertation, TH Karlsruhe.
Hirschmann, B (2011) Ein Beitrag zur Bestimmung der Scheibenschubfestigkeit von
Brettsperrholz (in German). Master Thesis, Graz University of Technology, Graz.

199
INTER / 48 - 12 - 02

Jeitler, G (2004) Versuchstechnische Ermittlung der Verdrehungskenngrößen von or‐

thogonal verklebten Brettlamellen (in German). Master Thesis, Graz University of
Technology, Graz.
Jöbstl, R A, Bogensperger, T, Schickhofer, G, Jeitler, G (2004) Mechanical Behaviour of
Two Orthogonally Glued Boards. WCTE, Lahti, Finland.
Jöbstl, R A, Bogensperger, T, Schickhofer, G (2008) In‐plane shear strength of cross
laminated timber. CIB‐W18/41‐12‐3, St. Andrews, Canada.
Kreuzinger, H, Sieder, M (2013) Einfaches Prüfverfahren zur Bewertung der Schubfes‐
tigkeit von Kreuzlagenholz / Brettsperrholz (in German). Bautechnik, Volume 90,
Issue (5), pp. 314–316.
PONTOS (2007) Benutzerhandbuch. GOM – Gesellschaft für optische Messtechnik,
Braun‐schweig.
R CORE TEAM (2015) R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria, http://www.R‐project.org.
Schickhofer, G, Bogensperger, T, Moosbrugger, T (eds., 2010) BSPhandbuch: Holz‐
Massivbauweise in Brettsperrholz – Nachweise auf Basis des neuen europäischen
Normenkonzepts. Verlag der Technischen Universität Graz, ISBN 978‐3‐85125‐109‐8.
Silly, G (2014) Schubfestigkeit der BSP‐Scheibe – numerische Untersuchung einer
Prüfkonfiguration (in German). Research Report, holz.bau forschungs gmbh, Graz
University of Technology, Graz.
Spengler, R (1982) Festigkeitsverhalten von Brettschichtholz unter zweiachsiger Be‐
anspruchung, Teil 1, Ermittlung des Festigkeitsverhaltens von Brettlamellen aus
Fichte durch Versuche (in German). Technische Universität München, München
(Berichte zur Zuverlässigkeitstheorie der Bauwerke, Heft 62).
Szalai, J (1992) Indirekte Bestimmung der Scherfestigkeit des Holzes mit Hilfe der
anisotropen Festigkeitstheorie. Holz als Roh‐ und Werkstoff, Volume 50, Issue 6,
pp. 233 238.
Wallner, G (2004) Versuchstechnische Ermittlung der Verschiebungskenngrößen von
orthogonal verklebten Brettlamellen (in German). Master Thesis, Graz University of
Technology, Graz.

200
 INTER / 48 - 12 - 02

Discussion

The paper was presented by P Dietsch

F Lam received clarification that the effect of compressive stresses on rolling shear
strength was taken into considered using adjustment factors.
H Blass and P Dietsch discussed the roller bearing details and member under pure
shear such as the case of shear wall. They also discussed the case of CLT as bending
members on edge where standard bending test would be available but difficult to
achieve shear failure of the members. Here torsional shear and rolling shear would
coexist. A Ceccotti stated that in a 3x3 m wall in shear connection design would be
more critical.
I Smith asked how the specimen size was selected. P Dietsch responded that they
were chosen to avoid localized stresses therefore a ratio of 1:3 for width to height was
chosen. G Schickhofer added that the tests are good for wall and diaphragm cases
but beam solutions would need more work.
P Zarnani asked whether mechanistic modelling approach would be used. P Dietsch
answered that the model is already mechanics based.

201
202
 INTER / 48 - 12 - 03

Advanced modelling for design helping of

heterogeneous CLT panels in bending

L. Franzoni – Laboratoire Navier (ENPC/IFSTTAR/CNRS) – Université Paris Est

A. Lebée – Laboratoire Navier (ENPC/IFSTTAR/CNRS) – Université Paris Est
F. Lyon – Centre Scientifique et Technique du Bâtiment (CSTB)
G. Foret – Laboratoire Navier (ENPC/IFSTTAR/CNRS) – Université Paris Est

Keywords: Cross Laminated Timber, bending, modelling, heterogeneities, bending

tests, innovative floors

1 Introduction
Cross Laminated Timber panels are gaining importance in timber construction, due
to their advantages coming from their crosswise lay-up. However the heterogenei-
ties affecting their bending behaviour still require advanced tools for being taken
into account.
In the first part of the present paper the “low” heterogeneities are mainly repre-
sented by gaps between narrow boards of each layer and CLT transverse shear
weakness. The suggested equivalent-layer model at the layer scale combined with
the 3D exact structure solution (Pagano, 1969 – 1970) and a failure criterion for
wood (van der Put, 1982) provide a good comparison with a reference test (Ho-
chreiner, 2013). Then, parameter studies are performed with the validated model in
order to quantify shear effects and the influence of varying layers’ number and ori-
entation.
The second part of this paper deals with stronger CLT heterogeneities, namely regu-
larly spaced voids within the panel. The results of an experimental campaign are
presented. The experimental behaviour is therefore compared with the behaviour
predicted with the equivalent-layer model and design methods for CLT (Eurocode 5,
2004, Kreuzinger, 1999). The input mechanical properties are reduced by wood vol-
ume fractions. The results of this comparison shows the need of a more accurate
modelling which is currently in development. The first results of such a new model-
ling procedure are presented as well.

203
INTER / 48 - 12 - 03

2 Low heterogeneities
A previous study (Franzoni et al, 2015) identified the gaps between lateral boards of
each layer and CLT transverse shear weakness as the main “low heterogeneities” af-
fecting their bending behaviour. In the following, the modelling procedure and main
results are briefly summarized.
2.1 Modelling
CLT heterogeneities are taken into account with a combined equivalent layer-based
mechanical model at the layer scale (Fig. 1) and the exact 3D solution at the struc-
ture scale (Pagano, 1969 – 1970). The equivalent layer model takes into account the
edge-gluing or not of lateral boards by means of simplified hypotheses on layer’s
mechanical behaviour (Tab. 1). A failure criterion for wood (van der Put, 1982) is im-
plemented in order to include a failure analysis. The reference bending behaviour
(Hochreiner, 2013) documents well the elastic and failure response of CLT, highlight-
ing the edge-gluing detachment as a firs failure mode. The comparison is made in
terms of panel’s global stiffness and failure stages within the apparent elastic re-
gime.

Figure 1 Schematic representation of continuous (a) and discontinuous (b) equivalent layer with the
layer’s reference frame

Table1 Elastic and strength properties of continuous and discontinuous equivalent layers [Mpa]
Elasticity (Keunecke et al, 2008) EL EN EZ GLZ GLN GZN νLZ νLN νZN
Continuous 12800 511 511 602 602 53 0.41 0.41 0.21
Discontinuous 12800 0.0 511 602 602 53 0.41 0.0 0.0

Failure (Dahl, 2009) fL,t fL,c fN,t fN,c fZ,t fZ,c fLZ fLN fZN
Continuous 63.4 28.9 2.8 3.6 2.8 3.6 4.8 4.8 2.0
Discontinuous 63.4 28.9 - - 2.8 3.6 4.8 4.8 2.0

2.2 Results
Comparison / edge-gluing. Each equivalent-layer model turns out to fit well the ref-
erence behaviour within the corresponding edge-gluing regime. It appears that
edge-glued layers increase CLT panel’s stiffness of about 8% but introduces also an
additional failure mode. Indeed, the edge-gluing detachment is one of the first fail-
ure modes, and therefore the discontinuous model gives a better prediction of

204
 INTER / 48 - 12 - 03

global load-carrying capacity in terms of failure modes (see Franzoni et al, 2015 for
further details). The discontinuous model is then used to perform parameter studies
on CLT properties. For all parameter studies the bending configuration is a panel
supported on two sides and submitted to an evenly distributed load.
Shear effects. The slope variation of failure load trend as a function of panel’s slen-
derness ratio (Fig. 1a) clearly separates the bending failure regime and the rolling-
shear one. This leads to the identification of a transition slenderness of 15 for a 5-ply
CLT and 19 for a 3-ply. The normalized difference between the predicted mid-span
deflection and the one using the thin-plates theory quantifies the shear part in de-
flection as a function of slenderness ratio (Fig. 1b).

Figure 2 Shear effects on CLT in bending. Failure load trend (a) and shear contribution to mid-span
deflection (b) as a function of panel’s slenderness ratio

Number of layers. Figure 2 shows that, from a deterministic point of view, increasing
layers’ number for a fixed CLT total thickness yields lower failure load (Fig. 2a) and
higher mid-span deflection (Fig. 2b). Both cases when the panel is thick and slender
are presented. The oscillations in shear failure load trend (blue line in Fig. 2a) derive
from the position of shear-compliant cross layers, which change with the lay-up.

Figure 3 Decreasing CLT bending performance while enlarging the number of layers. Failure load
(a) and normalized mid-span deflection (b)

205
INTER / 48 - 12 - 03

Transverse layers’ orientation. Intuitively, intermediate orientations of transverse

layers between 90° and 0° may mitigate CLT shear weakness. Figure 3 presents the
variation of mid-span deflection and failure load as a function of the varying orienta-
tion of 5-ply CLT transverse layers. Similar results are obtained for three or seven
layers configurations. The favourable effects of rotating transverse layers are signifi-
cant only for thick CLT panels, while for slender ones the gains are lower. Moreover,
for both cases, the failure load trend shows a drastic drop at several lamination an-
gles due to high in-plane shear stresses related to the torsion moment coming from
the non-orthotropic configuration of the plate.

Figure 4 Mid-span deflection (a) and failure load (b) as a function of the varying orientation of
transverse layers

3 Strong heterogeneities
Cross Laminated Timber panels having periodic spacing within each layer are already
in production. The challenge is to assess the bending efficiency of these lighter and
more acoustically efficient floors. Since they are innovative products and the knowl-
edge about them is limited, an experimental campaign has been carried out. A mod-
elling procedure in order to predict their bending behaviour, and especially trans-
verse shear effects, is in development. In this section, the main results of the ex-
periments and the first results of the modelling are presented and discussed.
3.1 Experimental campaign
3.1.1 Four-points bending tests on classical and innovative timber floors
The experimental campaign was based on four-points bending tests on classic and
aerated CLT floors. The distance between the supports and the loading points was
L/3 (where L is the span). The spaced floors had the same regular spacing between

206
 INTER / 48 - 12 - 03

boards along both direction and for all layers. The voids were filled with an insulat-
ing material (glass wool), having no mechanical properties. The wood species was
Norway spruce of strength class S10 (DIN 4074) for the CLT panels and C24 (EN 338)
for the spaced floors. Figure 5 and Table 2 present the geometry of the tested
specimens as well as the main results.

Figure 5 and Table 2 Geometry of the tested panels in four-points bending and main results
Configuration CLT Panel-1 Panel-2
Number of specimens 2 2 2
L (m) 4.65 5.88 5.88
b (m) 1.25 1.31 1.26
e (mm) 20 30 30
Number of layers 5 7 7
h (mm) 100 210 210
Voids length - lv (mm) - 150 300
Wood length - lw (mm) - 100 100
Wood volume (%) 100 40 25
Failure load (kN) 75 68 34
Failure mode CL TL RS
Global stiffness (kN/m) 470 765 390
Bending stiffness (kN·m2) 870 3100 1700
Shear / bending deflection (%) 3.5 15 25
Failure modes: RS = rolling shear; CL = longitudinal compressive; TL = longitudinal tensile

The measuring system was mainly based on linear variable differential transformers
(LVDTs) for displacement measurement. For all specimens the global mid-span de-
flection (U) and the bending deflection (u) were measured. For some specimen, the
absolute rotation (φ) of the plate’s cross section was also measured. Then, the ef-
fective plate bending stiffness can be computed using the following equations:
∙ ∙ ∙ ∙
( ) = (1) ( ) = (2)
48 ∙ 18
where F is the load and Lb is the distance between the LVDTs used to measure the
bending deflection u. Once the bending stiffness is determined, panel’s shear stiff-

207
INTER / 48 - 12 - 03

ness and shear contribution to deflection can be estimated by means of equations

(3) and (4):
216 ∙
= ∙ ( ) ∙ (3)
1296 ∙ ( ) − 23 ∙ ∙

ℎ!"# $!%&!'()*+ 216 ∙ ( )

= (4)
,!+$)+- $!%&!'()*+ 23 ∙ ∙

Additionally, the panel’s failure load and failure mode were monitored. Table 2
shows the main results of the bending tests
Due to its high slenderness, the classical CLT floor failed in bending, with several
ductile compressive cracks appearing in the upper layer before the brittle tensile
failure of bottom layer. Both specimens of Panel-2 failed in rolling shear, with a
significant rotation of transverse boards (Fig. 6), while the other configuration of
spaced floors failed in tension in the bottom layer.

Figure 6 Rolling shear failure of transverse boards of spaced timber floors (Panel-2) in bending

3.1.2 Small-scale tests on clear raw timber

In order to mitigate the uncertainity on the mechanical properties for the
subsequent modelling procedure, tests for the identification of elastic and strength
properties of the raw timber are currently in progress. Axial-parallel to the grain (EL,
fL,t, fL,c) and rolling shear (GZN, fZN) tests have been carried out. The previous study
(Franzoni et al, 2015) identified such mechanical properties as the dominant ones
when dealing with bending behaviour of crosslams. Moreover, the mechanical
properties of clear spruce have been found to be adequate to reproduce CLT
bending response. Fig. 7-Table 4 and Fig. 8-Table 5 show the axial and shear tests on
clear specimens of timber. The remaining elastic and strength properties for the

208
 INTER / 48 - 12 - 03

modelling are taken from Table 1. Classic crosslam floors and spaced ones have
been supplied from two different producers; hence all results on the respective
lumber boards are presented separately.

Axial Tests – Parallel to grain (L) n .̅ (MPa) COV (%)

Spaced floors-Norway spruce C24 (EN 338)
EL (tension + compression) 21 12700 17.2
fL,t 10 85 17.3
fL,c 8 51 6.4
CLT floors-Norway spruce S10 (DIN 4074)
EL (tension + compression) 17 9850 11.8
fL,t 7 77 13.7
fL,c 8 38 4.1
Figure 7 and Table 4 Geometry and results of the axial tests on clear wood. Dimensions in mm

Rolling Shear Tests (ZN) n .̅ (MPa) COV (%)

Spaced floors-Norway spruce C24 (EN 338)
GZN 10 110 25
fZN 12 1.6 18
CLT floors-Norway spruce S10 (DIN 4074)
GZN 8 70 25
fZN 8 1.3 9
Figure 8 and Table 5 Geometry and results of the shear tests on clear wood. Dimensions in mm

3.2 Advanced modelling and existing design methods

The bending tests on CLT panels allowed a further validation of the previously de-
scribed equivalent-layer model. The deflections predicted at the same points of
LVDTs lead to the identification of structure’s elastic moduli using equations (1), (3)

209
INTER / 48 - 12 - 03

and (4). Two existing design methods for CLT which allow the direct estimation of
the bending stiffness are applied: the gamma-method (Eurocode 5, 2004) and the
shear analogy method (Kreuzinger, 1999). Both of them are implemented following
the instructions found in Gagnon & Pirvu, 2013. Table 6 shows the relative differ-
ence between the actual and predicted bending behaviour f CLT panels.

Table 6 Relative distances between the actual and predicted bending behaviour of CLT
Comparison - CLT Gamma Shear Analogy Equivalent layer -
Method discontinuous
Failure load - - +5%
Failure mode - - CL
Global stiffness -5% -4% -4%
Bending stiffness -8% -4.5% -4.5%
Shear / bending deflection -8.5% +8% -5.5%

From Table 6 it is clear how both the suggested advanced model of equivalent-layer
and design methods can reproduce well CLT bending response.
A starting point for analysing more heterogeneous CLT floors could be using the
methods presented above and reducing wood mechanical properties by the wood
volume fractions. This approach has been already used in Blass & Gorelacher, 2000
for the rolling shear modulus and is common in engineering practice.
A more accurate model for periodically spaced CLT panels is currently in develop-
ment. This model is based on a homogenization scheme handled by a high-order
plate theory (Lebée & Sab, 2012). Basically it enforces membrane, bending and
shear strains on an elementary unit cell of the spaced crosslam panel (Fig. 9) and
equalizes the elastic energy of such an unit cell to the elastic energy of the whole
panel. This homogenization approach leads to the identification of panel’s bending
and shear moduli, which will be used by the plate theory for computing the stresses
and displacements.

Figure 9 Unit cell of spaced CLT floors and imposed membrane (e), bending (χ) and shear (γ) strains

210
 INTER / 48 - 12 - 03

At present, only the membrane-bending step was conducted and therefore only the
effective bending stiffness can be computed. Tables 7 and 8 present the distances
between the actual and predicted behaviour of spaced floors. The equivalent-layer
model predicts plate’s moduli by means of equations (1), (3) and (4) using the pre-
dicted deflections, while the shear analogy method and the homogenization scheme
compute directly the plate bending and shear stiffnesses. Due to the significant dis-
tance from the reference and from the other models, the gamma method is not pre-
sented and it is replaced by the periodic homogenization model.

Table 7 Relative distances between the actual and predicted bending behaviour of Panel-1
Comparison – Panel-1: wood Shear Analogy* Equivalent layer Periodic homog-
volume = 40% - discontinuous* enization
Failure load - +40% in progress
Failure mode - RS in progress
Global stiffness +25% +23% in progress
Bending stiffness +18% +16% -9.5%
Shear / bending deflection -73% -60% in progress
* Mechanical properties reduced by wood volume fraction

Table 8 Relative distances between actual and predicted bending behaviour of Panel-2
Comparison – Panel-2: wood Shear Analogy* Equivalent layer Periodic homog-
volume = 25% - discontinuous* enization
Failure load - +34% in progress
Failure mode - RS in progress
Global stiffness +32% +28% in progress
Bending stiffness +31% +28% +2%
Shear / bending deflection -84% -76% in progress
* Mechanical properties reduced by wood volume fraction

Table 7 and 8 shows that the wood volume fractions approach fails the more the
panel becomes heterogeneous. The equivalent-layer model returns less margin of
error than the design method. The advanced model based on periodic homogeniza-
tion gives a good prediction of the effective bending stiffness of strongly heteroge-
neous panels.

4 Conclusion and outlooks

Low heterogeneities. The developed equivalent-layer model turned out to be appro-
priate to reproduce elastic and strength bending response of CLT panels. The edge-
gluing of lateral boards does not contribute very much to the bending performance,

211
INTER / 48 - 12 - 03

introducing an additional failure mode. The discontinuous layer model gives a better
prediction of CLT load-carrying capacity than the continuous layer. Moreover, me-
chanical properties of clear wood lead to a good comparison with a reference
specimen having knots. This suggests the “system effect” when assembling lumber
boards in a CLT configuration, which effect is to regularize the presence of knots and
increase boards stiffness and strength. The parameter studies performed with the
validated model quantified shear effects in CLT in bending and showed the loss and
low gains while enlarging the number of layers or the orientation of transverse lay-
ers.
Stronger heterogeneities. Reducing wood mechanical properties by wood volume
fractions is not sufficient for a reliable design of spaced timber floors, especially with
respect to transverse shear effects. This means that such a simplified approach can-
not take into account the complexity of stress and strains distribution within these
strongly heterogeneous panels. Therefore a more accurate model based on a ho-
mogenization scheme is currently in development. First results show that such an
advanced model can precisely predict the bending stiffness of spaced floors. There-
fore the following steps of this advanced modelling are the prediction of the panel’s
shear modulus, deflection and failure load/mode. The final aim is to develop a sim-
plified calculation tool for the design of heterogeneous CLT floors in order to en-
courage their safe application in timber construction.

5 References
Blass, H. Gorlacher, R. (2000) Rolling shear in structural bonded timber elements. In-
ternational conference on wood and wood fiber composites, Stuttgart, Germany.
Dahl, K. (2009) Mechanical properties of clear wood from Norway spruce. PhD the-
sis, Norwegian University of science and technology
DIN 4074-1:2012-06 Sortierung von Holz nach der Tragfahigkeit, Nadelschnittholz.
EN 338:2010 Structural timber—strength classes
Eurocode 5 (2004): Design of timber structures - Part 1-1: General and rules for
buildings, CEN (EN 1995-1-1)
Franzoni, L. Lebée A., Lyon, F. Foret, G. (2015) Cross Laminated Timber panels in
bending: an equivalent-layer approach (submitted)
Gagnon, S. Pirvu, C. (2013) CLT handbook: Cross Laminated Timber. FPInnovations,
Quebec, Canada
Hochreiner, G. Fussl, J. Eberhardsteiner, J. (2013) Cross Laminated Timber plates
subjected to concentrated loading. Experimental identification of failure mechan-
isms. Strain, 50(1):68-71

212
 INTER / 48 - 12 - 03

Keunecke, D. Hering, S. Niemz, P. (2008) Three-dimensional elastic behaviour of

common yew and Norway spruce. Wood Science and Technology, 42: 633-647
Kreuzinger, H. (1999) Platten, Scheiben und Schalen. Ein Berechnungsmodell fur
gangige Statikprogramme (German). Bauen mit Holz, 1:34-39
Lebée, A. Sab, K (2012) Homogenization of thick periodic plates: application of the
Bending-Gradient plate theory to a folded core sandwich panel. International
journal of solids and structures, 49: 2778-2792
Pagano, N.J. (1969) Exact solutions for rectangular bidirectional composites and
sandwich plates. Journal of composite materials, 4:20-34
Pagano, N.J. (1970) Influence of shear coupling in cylindrical bending of anisotropic
laminates. Journal of composite materials, 4:330-343
Van der Put, TACM (1982) A general failure criterion for wood. In Proceedings of
15th Union of Forest Research Organizations Meeting, Boras, Sweden

213
INTER / 48 - 12 - 03

Discussion

The paper was presented by L Franzoni

H Blass asked if the effect of board width was considered. L Franzoni said not yet and
3.5 is the ratio between boards. H Blass asked why were tension and compression
properties established using small clear tests. L Franzoni answered that they tried to
be consistent with modelling approach using clear properties and agreed that the ma-
terial in reality has defects.
K Malo asked why there were sharp changes shown in the curves in slide 9 where with
in-plane shear failure one would expected smooth rather than sharp changes. L Fran-
zoni said that this could be caused by numerical issues
G Fink received clarification that the model was based on small clear properties.
K Malo and L Franzoni further discussed sharp changes were due to change in failure
mode because the model could not determine mixed failure mode and therefore sharp
changes were seen.
G Schickhofer asked whether the end goal should be aimed towards design code. L
Franzoni said that this was not their goal at this moment.

214
215
216
 INTER / 48 - 12 - 04

Performance of Canadian glulam columns

with new laminae E requirements
Frank Lam, University British Columbia
Jung Kwon Oh, Seoul National University (formerly Post Doc University of B.C.)
B.J. Yeh, APA – The Engineered Wood Association
Jun-Jae Lee, Seoul National University

Keywords: Glulam column, Reliability index, Cubic Rankine Gordon curve

1 Introduction
It is well-known that glulam column performance depends on its length and the
strength properties of the laminae (Blass 1987, John 1991). When the column height
is short, its design is governed by the compression strength parallel to grain. When
the column height is long, its design is governed by Euler buckling (hence modulus of
elasticity (MOE)). As intermediate columns, the design is governed by both compres-
sion strength parallel to grain and MOE. The traditional approach recognized three
different categories of columns: short, intermediate and long, representing a transi-
tion from short column plastic failure to elastic buckling failure modes. In Canada, the
Code on Engineering Design in Wood CSA O86 adopted a column design equation
based on the Cubic Rankine Gordon (CRG) curve which was originally proposed by
Neubauer and Tekinel (1966). The CRG formula is continuous over all slenderness ra-
tios, and is more conservative than the traditional column formula in the intermedi-
ate slenderness range. In the US, Ylinen’s buckling equation is used in column design.
The equation is based on an empirically fitted fourth power parabolic function (New-
lin and Gahagan, 1930) and a nonlinear function representing the compressive stress-
strain relation of wood (Ylinen 1956). Zahn and Rammer (1995) and Rammer and
Zahn (1997) provided a detailed database and an evaluation of the Ylinen equation
for glulam and parallel strand lumber columns, respectively.

The strength properties of the laminae are influenced by the grading rules of the lam-
inae that control the quality of the wood. For Canadian 16c-E grade glulam, CSA
O122 requires that Douglas fir laminae have a minimum MOE of 12,400 MPa and
minimum visual C-grade. These rules were established several decades ago based
primarily on common practice of the time, industry experience, and product perfor-
mance record. As the characteristics of the timber resources evolve through time,

217
INTER / 48 - 12 - 04

these rules may no longer be suitable in terms of resource utilization. The competi-
tive position of glulam can be improved if modifications to the grading rules of glulam
laminae can be considered to fit closer to the characteristics of the current resource
with technical evidence that the final product can still meet the target level of per-
formance of Canadian wood products.

This paper presents the results of an experimental program to address the impact of
modifying the laminae MOE requirements on the strength properties of glulam col-
umns. Glulam specimens manufactured with laminae having a MOE range of 11,000
MPa and 13,100 MPa were considered. The performance of the glulam columns was
quantified by reliability analysis to study their safety level per Canadian Design code
provisions.

2 Materials and Methods

2.1 Test set-up for full-size intermediate-length column
Full size glulam column testing is challenging because it involves high axial forces.
Since the behaviour of columns depends on the boundary condition, such high forces
can create problems because there could be unintended restrains at the column
ends. Neubauer (1972) carried out column tests with pointed end steel supports and
reported that these supports appeared to be unsatisfactory for stronger and short
column under heavy load. Zahn and Rammer (1995) used roller supports (pinned
end) for full-size intermediate-length glulam column tests. They also indicated that
rollers in the end supports sometimes locked up under the heavy axial load. This is
because rolling friction could be high enough to make the test behave like a test of
members with square ends on rigid platens rather than a test of members with sim-
ple supports. Therefore, as a solution he applied dithers (vibrators that supply the
energy needed to break the static friction) to make “pinned end” boundary condition.
If rotation is restrained by friction, the effective length would decrease significantly,
which would cause overestimation of the column performance. In practice, it is very
difficult to create pure frictionless pinned end supports or to estimate the amount of
rotational restraints contributed by friction. In this study, an alternative conservative
approach using fixed support conditions was chosen.

A total of 90 glulam specimens were prepared for the full-size glulam test. All glulam
specimens were made with four C grade Douglas fir laminae with a minimum and
maximum MOE of 11,000 MPa and 13,100 MPa, respectively. The test material was
produced by Western Archrib following CSA O122 and shipped to the Timber Engi-
neering and Applied Mechanics (TEAM) Laboratory at the University of British Co-
lumbia for testing. The cross section of the glulam test columns was 79 mm by 150
mm. The size of the test columns was chosen based on the capacity of the test actu-
ators in the TEAM laboratory.

218
 INTER / 48 - 12 - 04

Three slenderness ratios (Sr=Le/d) were chosen in the range of intermediate length.
For the full-size compressive test, both ends of the specimens were carefully sawn off
square. Table 1 shows the final length of the three slenderness ratio groups.

Table 1. Test specimens.

Slenderness Effective Visual
No. of Cross Length
Group ratio *1 length*1 Grade of Species
specimen section (mm)
(Sr=Le/d) (Le in mm) lamination
A 25.1 30 3,657 1,986
79mm x
B 20.1 30 3,048 1,590 C Douglas-fir
150mm
C 15.1 30 2,439 1,194
*1 Effective factor = 0.65 (CSA O86); Total reaction support length (602 mm); Le=0.65*(Length-602mm)

MOE of each specimen was measured by the Metriguard E computer. Hand-tapping

initiated vibration of the simple-supported glulam specimen to make it vibrated in
the same direction as it would be buckled in full-size column test.

In a companion test program, APA-The Engineered Wood Association carried out

short column tests (381 mm in length). The glulam specimens were made with the
same C-grade laminae and cross section as the UBC test specimen. The test results
show that the compressive strength in current Canadian design code seemed to be
slightly conservative. Table 2 shows the statistics of short column compressive
strength parallel to grain.

Table 2. Short column test results performed by APA.

Compressive strength parallel to grain (MPa)

Average 42.9
COV (%) 5.1
N 30
Max 47.5
Min 40.0
5% tolerance 38.8
* NOTE - The laminating lumbers of this glulam were visual C grade satisfying the requirements for 16c-E
grade of CSA O122.

Fig. 1 shows a schematic drawing for the full-scale compression test. The fixed sup-
ports were designed and custom built as shown in Figs. 2a and 2b. At one end, the
fixed support composed of a steel shoe and a set of steel roller guide to allow axial
movement while preventing rotation. At the other end, the support did not have to
move axially; hence, it consisted of a steel shoe only to restrain rotation.

219
INTER / 48 - 12 - 04

Glulam specimens were placed into these supports with the narrow face of the spec-
imen oriented parallel to the ground. At one end, the MTS hydraulic actuator pushed
the steel shoe in the fixed support transferring the axial force to the glulam column
with rotational restraint about the vertical axis (Fig. 2a). The other end was also
placed in a custom-made fixed support (Fig. 2b) with the fixed-support attached to a
strong steel column. To prevent unexpected buckling in strong axis, two lateral sup-
ports were installed between a fixed support and mid-point. In this test setup, buck-
ling occurred only in weak axis and horizontal direction (Fig. 3). To create a small load
eccentricity to initiate lateral deformation, a lateral load of 68 kg (150lb, Group B and
C) and 45 kg (100 lb, Group A) was applied at the mid-point (Fig. 4b). This small lat-
eral load was less than 0.022% of peak axial load.

Axial force was applied at the displacement controlled speed of 2.5 mm/min for this
testing and all specimens reached to maximum load in 10 min. The test continued
until the load dropped below 50% of the maximum load. Axial load and test machine
stroke were recorded. Lateral displacements resulting from column buckling were
measured at mid-span by linear voltage displacement transducers (LVDT).

Figure 1. Schematic drawing for full size column test set-up (Top view)

Figure 2. Picture of custom-made fixed support. (a: Support at the hydraulic actuator side. b:
Support at the other side.)

220
 INTER / 48 - 12 - 04

(a) (b)

Figure 3. Picture of full-size column test set-up. (a) Vertical movement were restrained by two
supports to prevent buckling in strong axis. (b) Pulley system applied small lateral load.

2.2 Test observations and results

In this study, transverse vibration technique was applied to measure the vibration
MOE for each glulam pieces. Table 3 shows the statistics of the MOE measurements.
Normal distribution was used to estimate the 5th percentile MOE. The 5 % parametric
tolerance limit with a 75% confidence level (E05) was 11,561 MPa.

The glulam specimens of this study were manufactured with laminae having a mini-
mum MOE of 11,000 MPa and a maximum MOE of 13,100 MPa. For 16c-E grade glu-
lam, CSA O122 requires that laminating lumber to have a minimum MOE of 12,400
MPa and minimum visual C-grade. For the 16c-E grade glulam, the CSA O86 specifies
12,400 MPa as average MOE value and 10,788 MPa as 5th percentile value.

Although the glulam in this study was made with lower MOE rated laminations than
the CSA requirements, the average and 5th percentile MOE of the glulam were higher
than the values specified by CSA O86 (Table 4). This comparison means that the spec-
ified MOE value for glulam columns in CSA O86 may be conservative.

Table 3. Statistics of vibration MOE for each group

Length MOE (MPa)
Group
(mm) Average St. dev. COV (%)
A 3,657 12,299 748.8 6.1
B 3,048 13,126 579.8 4.4
C 2,439 13,259 543.7 4.1
Total 12,895 755.7 5.9

221
INTER / 48 - 12 - 04

Three different glulam column lengths were considered in full scale column tests. For
the three groups, non-parametric 5th percentile tolerance of maximum axial load was
calculated. Table 5 shows the statistics of maximum axial load for each group.

Table 4. Comparison between test specimen and 16c-E grade glulam. (Unit : MPa)
Test CSA
Lamination
- Min. Visual grade C-grade C-grade
- Min. MOE 11,000 12,400
Glulam
- Compression parallel, fc 38.8 30.2 (Specified Strength)
- Average MOE, E 12,895 12,400
th
- 5 percentile MOE, E05 11,561 10,788

Table 5. Statistics of maximum axial load for each group.

Group A Group B Group C
Average (N) 230,992 314,211 414,823
St. dev. (N) 10,150 20,650 27,463
COV (%) 4.4 6.6 6.6
th
5 Percentile tolerance (N) 208,274 275,563 373,465

2.2.1 Determination of actual effective length factor

The custom-made support was intended to restrain rotation to create a sufficiently
strong and stiff steel assembly for this support. However, it is practically difficult to
make a pure fixed support because the steel can deform elastically and there can be
some free play in the system (albeit very small). Therefore, for the purpose of the
analysis the custom-made support was assumed to be represented by a very stiff
elastic rotational spring. The spring constant was calibrated by measurements in the
test. To calibrate the spring constant, the graph of PΔ versus angle (Ɵ, 2∆/LBS in Fig. 1)
was drawn. The slope in this graph is governed by spring constant as shown in Fig. 4.

Figure 4. Estimated slope change for different spring constants predicted by SATA.

222
 INTER / 48 - 12 - 04

All specimens of Group A did not show any compressive failure until peak load. But in
Groups B and C, compressive wrinkles were found before peak load. Therefore, only
Group A specimens were used to calibrate the spring constant. The spring constant of
other groups was assumed to be the same as the constant of Group A, because the
same support was used for all test groups.

The spring constant was calibrated by least square method where the spring constant
yielding the minimized sum of square error (Eq. 1) was determined as the spring con-
stant of the support.

 
n
S =  Ki  Ki* (k)
2

i 1 (1)
*
where Ki is the actual slope from the test and Ki (k) is the predicted slope from the
FEM program SATA (Song 2009), of the ith specimen with a specific spring constant, k.

Here Ki=∂(P∆)/∂Ɵ and Ki *=∂(P∆)*/∂Ɵ*. The spring constant to minimize the sum of
square error was determined by Eq. 2.

S
0
k (2)

In SATA, a polynomial model was assumed to represent the non-linear parallel-to-

wood-grain stress-strain relationship (Eq. 3).

E  f t / E0    0
 0
  
  r  2 f c ( )3  3  2r  f c ( ) 2  E0 0  p
 p p
 E      f u     p
 d p c
(3)

where σ and ε are the stress and strain, respectively; ft and fc are the tensile and
compressive strengths, respectively; E0 and Ed are the initial modulus of elasticity and
the slope of the falling branch of the stress strain curve, respectively; εp is the strain
corresponding to the compression strength, fc; r= εpE0/ fc defines the nonlinearity of
the model.; and εu and ft/E0 are the maximum compressive and tensile strains, re-
spectively. The model is shown schematically in Fig. 5.

In the spring constant calibration, nonlinearity was assumed to be represented by the

equation εp = 1.250 fc/E (Blass, 1987). Because the other properties were not sensi-
tive in this spring constant calibration, specified design values in CSA O86 for 16c-E
glulam were used for the other properties.

223
INTER / 48 - 12 - 04

Before calibrating the spring constant, the effect of initial lateral deflection on this
slope was investigated. In this analysis, initial deflected shape was assumed to be si-
nusoidal with maximum values at the mid-span. From the comparison of slope with
changing initial deflection up to 3.0 mm (l/1000), it was found that maximum differ-
ence from the case of l/2000 (1.5 mm) was only 2.6% and it was small enough to dis-
regard. The measured initial deflection was less than 3 mm.

Figure 5. Polynomial model of the parallel-to-wood-grain stress-strain relationship (Song, 2009).

From the linear elastic range of the test data, the slope (Ki) in the graph of PΔ versus
angle was calculated. In SATA, the assumed sinusoidal initial lateral deflection was as-
sumed to be l/2000 at mid-point. Also for all specimens, average compressive
strength from APA and as-measure MOE for each specimen were used as input fc and
E0 to SATA. Based on these inputs the slope (Ki *) was obtained from SATA analysis
with spring constant varying from 1.0×108 Nmm/rad to 5.0×108 Nmm/rad. The best-
fit spring constant was calibrated as 2.3×108 Nmm/rad.

Effective length factor can be determined from the best-fit spring constant at the
end. Effective length can be expressed by

Le  L (4)

where β is the effective length (or buckling length) factor and L=LBS is length of the
member.
Exact elastic effective length factors were determined by the zero determinant
condition of the stiffness matrix of the restrained compression member, as expressed
by the well-known transcendental equations (Eq. 6, Hellesland, 2007).

 
2
     
    tan  
   2     2 
 1   1
H2 H    
 tan   
   2
(5)

224
 INTER / 48 - 12 - 04

kL
H
EI (6)

where, k is spring constant, E is modulus of elasticity and I is moment of inertia. By

this equation, the effective length factor was calculated for the spring constant of
2.3×108 Nmm/rad where the length of specimen is different according to the length
group. Therefore, actual effective length factor was calculated for each length group.
The actual effective length factor for each length group was 0.60 (Group A), 0.62
(Group B) and 0.65 (Group C). This value is between the theoretical effective length
factor (0.5) and the factor for fixed support specified by CSA O86 (0.65). For subse-
quent analysis 0.65 was used.

3 Reliability Analysis
The CSA O86 code provides the CRG formula (Eq. 8) as the slenderness factor the for
column design.

1
 FcCc 
3
Kc  1.0   (7)
 NE05 

where, Kc is slenderness factor and Cc is slenderness ratio as the effective length di-
vided by the member’s smaller cross sectional dimension. Fc and E05 are the short-
column compressive strength and fifth percentile MOE, respectively (see Table 4). N
is a calibration factor taken as 35 based on compression behaviour of dimension
lumber.

The factored column resistance Pr can be estimated as:

Pr   Fc A Kc (8)

where A is the cross section area (mm)and  is the performance factor. It is noted
that other adjustment factors for volume, treatment, and serviceability effects are
taken as unity. This is appropriate considering the size of the columns tested and the
test condition.

Furthermore to allow direct comparison with test results, the specified compressive
strength Fc needs to be increased by a factor of 1.25 to convert load duration adjust-
ed specified strength from the standard load term of 3 months to short term test du-
ration as FC’ = 30.2*1.25 MPa =37.75 MPa.

225
INTER / 48 - 12 - 04

Reliability analyses were conducted to evaluate the performance of the glulam mem-
bers under dead and snow load conditions for six Canadian cities (Arvida, Halifax, Ot-
tawa, Saskatoon, Quebec City and Vancouver) following the procedures outlined by
Foschi et al. (1989).

The limit state design equation for glulam column short term compressive resistance
parallel to grain can be expressed as

D GD DN  L GL LN   Fc' A Kc (9)

where D and L are the load factors for dead (1.25) and live (1.5) loads, respectively;
GD and GL are the dead and live load geometric factors which convert the applied
loads to compression capacity; DN and LN are the nominal design dead load and nom-
inal design total roof snow and rain load, respectively.

The failure function developed to relate the compression resistance and the effect of
loads for reliability analyses is as follows:

G  P  (GD D  GL L) (10)

where P, D, and L are random variables representing the compression capacity, dead
load, and live snow load, respectively. G=0 => limit state; G>0 => safe and G<0 =>
failure. Statistical distributions and parameters for the snow load for the six Canadian
cities were described in detail in past studies (Foschi et al. 1989).

The failure function can be rewritten as:

1
 Fc' A  Fc ' Cc 
3
G  P- 1.0         (11)
 D     L  NE05 

where =DN/LN; =D/DN; =L/LN; and =GD/GL. The variables  and  were taken as 1.0
and 0.25, respectively. The random variable  was assumed to be normal with mean
of 1.0 and standard deviation of 0.1.

The compressive resistance P was considered using two approaches: 1) log- normal
distributions fitted to the test data and 2) calibrated CRG curve considering compres-
sion strength c of short glulam column and modulus of elasticity E* as correlated
random variables.

226
 INTER / 48 - 12 - 04

The lognormal distribution parameters fitted to the three sets of intermediate col-
umn compressive capacity data are given in Table 6 and the comparison between fit-
ted distribution and test data is shown in Figure 6.

Table 6. Lognormal distribution parameters of column capacity for each group.

Group A Group B Group C
Average (kN) 231.001 314.240 414.848
St. dev. (kN) 10.182 20.912 27.258
COV (%) 4.41 6.65 6.57

1.0
0.9 Test Data Sr=25.1
Cumulative Probability

0.8
0.7 Test Data Sr=20.1
0.6 Test Data Sr=15.1
0.5
0.4 Lognormal
0.3
0.2
0.1
0.0
0 100 200 300 400 500
Column Compression Capacity (kN)

Figure 6. Comparison between lognormal fitted distribution and test data.

Alternatively the input strength properties parameters of the CRG Curve were con-
sidered as random. The modulus of elasticity E* was assumed to be normally distrib-
uted with mean of 12895 MPa and coefficient of variation of 5.8% (based on modulus
of elasticity test data). The short column compressive strength c was assumed log-
normally distributed and calibrated to have a mean of 46.5 MPa and a coefficient of
variation of 7.8%. The calibrated c is higher than the measured results from APA.
However the “short” column strength in the CRG curve is referenced to columns with
very short length whereas the APA short column specimens had lengths of 0.381 m;
hence, the higher calibrated c is justifiable.

The short column compressive strength and the modulus of elasticity were also as-
sumed correlated with a coefficient of correlation of 0.525. This relatively weak cor-
relation can be justified because of the narrow range of modulus of elasticity values.
Finally the parameter N was also calibrated to be 38.

1
  cCc 3 
P  cA1.0   (12)
 38E * 

227
INTER / 48 - 12 - 04

Comparisons between simulation results from the fitted random Cubic Rankine Gor-
don Curve and test data are shown in Figure 7. It can be seen that the fit was quite
reasonable. The errors of the mean values were -4.3%, -0.6 % and 0.6% for the three
slenderness ratios of 25.1, 20.1 and 15.1, respectively. The errors of the fifth percen-
tile values were -6.0%, 1.5 % and -0.9% for the three slenderness ratios of 25.1, 20.1
and 15.1, respectively. Negative values imply conservatism. It may be possible to
achieve better fit but in this process we tried to keep the positive errors as small as
possible without excessive conservatism on the negative errors.

1.0
0.9 Simulated: Sr = 25.1
Cumulative Probability

0.8 Simulated: Sr = 20.1

0.7
0.6 Simulated : Sr = 15.1
0.5 Test Data: Sr = 25.1
0.4
0.3 Test Data: Sr = 20.1
0.2 Test Data : Sr = 15.1
0.1
0.0
0 100 200 300 400 500

Column Compression Capacity (kN)

Figure 7. Comparison between Cubic Rankine Gordon Curves with random parameters and test data.

First-order second-moment reliability analyses were performed considering both ap-

proaches to represent P for dead load and live load of snow plus rain for the six Ca-
nadian cities. A 30 year return period of the live load is considered with a probability
of non-exceedance of 29/30. Results of the reliability analyses are shown in Figs 8
and 9.

Performance of Glulam Column

Fitted Capacity Distribution
6

5
Reliability Index ()

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Performance Factor ()

Sr = 25.1 Sr=20.1 Sr=15.1 Average Minimum

Figure 8. Reliability index versus performance factor relationship of column resistance (based on
fitted column resistance distribution).

228
 INTER / 48 - 12 - 04

Performance of Glulam Column

CRG Analysis
6

5
Reliability Index ()

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Performance Factor ()

Sr = 25.1 Sr=20.1 Sr=15.1 Average Minimum

Figure 9. Reliability index versus performance factor relationship of column resistance (based on
fitted random Cubic Rankine Gordon Curve).

In the Canadian Code CSA O86 a performance factor  of 0.8 was used in the design
provision of compressive resistance parallel to grain for glulam columns. At  =0.8,
the associated mean reliability index is 3.267 and 3.195 for the fitted capacity distri-
bution and the fitted random CRG Curve approach, respectively. Similarly at  =0.8,
the associated minimum reliability index is 3.019 and 3.020 for the fitted capacity dis-
tribution and the fitted random CRG Curve approach, respectively. Considering the
target mean reliability index adopted in the Canadian Code CSA O86 is between 2.6
and 2.7 for a  =0.8, the new laminae of glulam columns evaluated in this study ex-
ceeded the target safety level. A higher  value in the range of 0.9 for the new glu-
lam column can be justified.

4 Conclusions
This paper presented the results of an experimental program to address the impact
of modifying the laminae MOE requirements on the strength properties of glulam
columns. Glulam specimens manufactured with laminae having a MOE range of
11,000 MPa and 13,100 MPa were considered. The measured MOE of the glulam
specimens was slightly higher than the code specified value even though the tested
glulam was made with laminate having lower MOE than the requirements of CSA
O122.

A framework for reliability analysis of Canadian glulam columns has been formulated.
The procedures take into consideration the randomness of the effects of applied
loads and the column resistance. Two approaches were used to model the random-

229
INTER / 48 - 12 - 04

ness of the column capacity. The calibrated CRG approach with random representa-
tion of c and E* is more flexible than the direct fitting of the column capacity distri-
bution because it can be applied to various slenderness ratios. Evaluations of the
performance of glulam made with new laminae grading rules show that these col-
umns exceeded the target reliability for structural wood products in Canada.

5 References
Canadian Standards Association (2014). Engineering design in wood. CAN/CSA-086-
14, Canadian Standards Association, Rexdale, Ontario, Canada.
Canadian Standards Association (2011). Structural glued-laminated timber. CAN/CSA-
O122-06, Canadian Standards Association, Rexdale, Ontario, Canada.
Blass, H.J. (1987). Design of Timber Columns. CIB W18A – Timber Structures Meeting
20, Dublin, Ireland.
Foschi, R.O., Folz, B.R., and Yao, F.Z. (1989). Reliability-based Design of Wood Struc-
tures. Structural Research Series, Report No. 34, Dept. of Civil Engrg., University of
British Columbia, Vancouver, Canada.
Hellesland, J. (2007). Mechanics and effective lengths of columns with positive and
negative end restraints. Engrg. Structures 29: 3464-3474.
Johns, K.C. (1991). A continuous design formula for timber column. Canadian J.of Civil
Engrg. 18:617-623.
Neubauer L.W. and Tekinel O. (1966). A more efficient column formula for the design
of wooden posts and studs. Transactions of the ASAE 9(6):816-817.
Neubauer L.W. (1972). Full-size stud tests confirm superior strength of square-end
wood columns. Transactions of the ASAE 15(2):346-349.
Newlin J.A. and Gahagan J.M. (1930). Test o large timber columns and prsenation of
the forest products laboratory column formula. Tech. Bull. No.167. USDA FPL.
Washington, D.C.
Rammer D.R. and Zahn J.J. (1997) Determination of Ylinen’s parameter for parallel
strand lumber. J. of Struct. Engrg. 123(10):1409-1411.
Song, X. (2009). Stability and reliability analysis of metal plate connected wood truss
assemblies. Ph.D. thesis, University of British Columbia, Vancouver, Canada.
Ylinen A. (1956). A method of determining the buckling stress and the required cross-
sectional area for centrally loaded straight columns in elastic and inelastic range.
Int. Assoc. for Bridge and Struct. Engrg. Zurich Switzerland. 16:529-550.
Zahn J.J., and Rammer D.R. (1995). Design of glued laminated timber columns. J. of
Struct. Engrg. 121(12):1789-1974.

230
 INTER / 48 - 12 - 04

Discussion

The paper was presented by F Lam

T Bogensperger received clarification of the definition of column length being the total
length of the column minus the length of the supporting shoes.
H Blass asked about the orientation of the column in relation to the buckling direction.
Since the columns buckled in the direction parallel to the glue-line of the laminae,
would the bending strength of the column be different if the columns buckled in the
other direction. F Lam responded the “bending” strength is governed by the compres-
sion strength of the laminae when maximum load was reached. The final failure of
tension occurred at a lower load level; therefore, orientation should not matter.
I Smith asked whether pin-pin conditions were tried. F Lam said that it was consid-
ered but not adopted given that the literature reported friction issues from pin-ended
cases. H Blass commented that in reality pin-ended conditions do not exist so the
code would be conservative if one assumed pin-ended case. F Lam agreed especially
at high load levels.
A Salenikovich received clarification that the laminae were sampled from a mixed pool
with MOE ranged of 11 to 13.1 GPa.

231
232
 INTER / 48 - 12 - 05

Design of CLT beams with large finger

joints at different angles
M. Flaig, Blaß & Eberhart GmbH, Germany

Keywords: long span CLT beams, double pitched CLT beams, CLT frame corners, large
finger joints, bending strength under an angle to the grain

1 Introduction
Despite the good mechanical properties and the increasing availability of the materi-
al, CLT beams are used for a limited number of applications only. Two of the main
reasons for the rare use are the limitation of the overall length of CLT members to
about 18 m and the lack of production methods for CLT beams with special shape.
Large finger joints (LFJ) offer the LFJ LFJ

possibility to connect prismatic max L = 18 m

or tapered CLT members to
prod

max L > 18 m
ges

form, e.g. double pitched CLT LFJ

beams or CLT frame corners

(Figure 1). Compared to glulam
members with similar shapes, LFJ LFJ

finger jointed CLT beams with

kinked axis provide appreciable
advantages, since stresses per- LFJ LFJ

pendicular to the beam axis can

be absorbed by the transverse LFJ LFJ

layers included in the material.

In the present paper the results
of tests with different types of
finger jointed CLT beams loaded
in plane direction and a proposal
for the design of LFJ in CLT
Figure 1. Possible dimensions and shapes of CLT
members are presented. members with large finger joints (LFJ)

233
INTER / 48 - 12 - 05

2 Large finger joints in CLT

2.1 Finger joint profile
In finger jointed CLT elements, the position of fingers in individual layers and hence, al-
so the number of finger tips within a layer depends on the ratio between the pitch p of
the finger joint profile and the layer thicknesses. To ensure an equal reduction of the
cross section in all layers of finger jointed CLT elements, the finger joint profile shown
in Figure 2 was developed. The geometry of the new profile is based on the profile 50 –
12 – 2 which is specified as a suitable profile for LFJ in EN 387. It features a finger
length j of 40 mm, a pitch p of 10 mm and a tip width bt of 1.4 mm. In combination
with layer thicknesses that are integer multiples of 5 mm, the pitch p of 10 mm entails
a constant nominal reduction νnom of 0.14 for all layers within a CLT element.
p = 10 bt = 1.4 j finger length

p pitch
j = 40

4°
5,1 bt tip width
α=
α finger slope

Figure 2. Finger joint profile 40 – 10 – 1.4

2.2 Effective reduction of the cross section

Besides the chosen profile, the design of edge fingers and the production process
have significant influence on the residual cross section of a finger-jointed connection.
To avoid fraying at the flat ends of the outermost fingers, LFJ are often produced with
a shoulder perpendicular to the beam surface. In the tested CLT beams, the tip of the
outermost fingers was enlarged by e = 2.5 mm resulting in shoulders with a width of
3.2 mm (Figure 3, left). The tip gap s (see Figure 3, right) also influences the residual
cross section of a finger jointed connection. The size of tip gap mainly depends on the
type of adhesive, the applied pressure and the pressing time.

2,5 mm Δt
3,2 mm 0,7 mm
s

Δt

Δt

Figure 3. left: broadened finger tip of edge finger

right: production tolerance in thickness direction ∆t and tip gap s

234
 INTER / 48 - 12 - 05

A further reduction of the cross section can be caused by production tolerances ∆ in

thickness direction resulting from a deviation of the finger joint profile from the
planned position. In CLT beams the thickness of every block of longitudinal layers, ar-
ranged at the surface or between two cross layers, is reduced by the production tol-
erance ∆ in thickness direction. Consequently, the total reduction of the load carrying
cross section amounts to (ncross + 1) times the tolerance ∆. Taking into account the
above-mentioned factors - the nominal reduction νnom of the finger joint profile, the
enlarged tip of edge fingers, the tip gap s and the production tolerance ∆ in thickness
direction – the effective reduction of the cross section in finger-jointed connections
in CLT members can be calculated according Eq. (1).
(n cross + 1) ⋅ ∆ + e
ν eff ,CLT =
(1 −ν eff ) ⋅ +ν eff (1)
t net,long
b t + 2 ⋅ s ⋅ tanα
with ν eff =
p
ncross number of cross layers within the element thickness
tnet,long sum of thicknesses of longitudinal layers

3 Experimental work
3.1 General
CLT beams with LFJ were tested to determine the bending capacity of the finger-
jointed connections under varying angles. The test series comprised tests with
straight CLT beams, double pitched CLT beams and CLT corner frames. In addition,
tests with straight CLT beams without LFJ were performed to determine the in-plane
bending strength of CLT as a reference value.
Throughout all series, the tested CLT beams had a 200
total thickness tgross of 200 mm and a height h of
600 mm. The layup of the beams was symmetric
to the centre plane and consisted of four longitu-
dinal layers with a thickness of 40 mm and two
transversal layers with a thickness 20 mm (Figure
600

4). The individual layers consisted of softwood la-

mellae with a width of 150 mm made of strength
graded boards of strength class C24 / T14 accord-
ing to EN 338 / EN 14080. Wood species of longi-
tudinal layers was exclusively northern spruce
20

20

(Picea abies). The transversal layers, too, mainly

40
40
40
40

consisted of northern spruce, but also contained Figure 4. Cross section of

some lamellae made of pine wood. tested CLT beams

235
INTER / 48 - 12 - 05

Before the tests, the production tolerance ∆ in thickness direction of each finger jointed
connection was measured at four different points near the corners of the cross section.
The tip gap s was also measured, but only in two points per connection in the middle of
the beam height. After the tests, the density and the moisture content were determined
from 20 to 30 mm thick samples taken over the complete cross section. An overview of
the test series and the dimensions and the properties of the specimens are given in Ta-
ble 1.

Table 1. CLT beams with and without LFJ; Test series and selected properties
joint moisture thickness
beam type LFJ series number span density tip gap
angle content tolerance
n  α ρmean umean ∆mean smean
in mm in ° in kg/m³ in % in mm in mm
none S-REF 15 10,800 - 442 10.5 - -
straight
1 S-LFJ 15 9,000 0 436 10.7 2.2 4.9
1 PI-05 5 9,000 5 451 11.0 0.9 5.0
double
1 PI-15 5 9,000 15 447 10.7 1.5 6.0
pitched
1 PI-25 5 9,000 25 448 11.1 1.4 5.5
2 CO-90. 5 4,500 22.5 458 11.8 3.0 6.0
frame
2 CO-105 5 4,500 18.75 463 11.6 3.2 5.4
corner
2 CO-115 5 4,500 16.25 469 11.2 2.9 5.3

3.2 Testing
In the first two test series, 30 four-point bending tests were performed to determine
the in-plane bending strength and the MOE of straight CLT beams with and without
LFJ. The beams without LFJ had a span  of 18 times the beam height. For the beams
with finger joints, the span was reduced to 15 times the beam height. During the
tests, the local deflection between the load application points was measured within a
length of (/3 – h) to determine the MOE of the beams. In Figure 5 and Figure 6 the
test setup and the dimensions of the straight CLT beams are illustrated.
F F

uglob
600

ulok

3000
F F
3600 3600 3600

10800

Figure 5. Test setup for straight CLT beams without LFJ (series S-REF)

236
 INTER / 48 - 12 - 05

F F

uglob LFJ

600
ulok

2400
F F
3000 3000 3000

9000

Figure 6. Test setup for straight CLT beams with LFJ (series S-LFJ)

Tests with double pitched CLT beams were performed to determine the in-plane
bending strength of LFJ in kinked CLT-beams subjected to opening moments. The
tested beams had three different slopes of 5°, 15° and 25°. The specimens were
symmetric about the plane of the finger joints, resulting in cutting angles α of the fin-
ger joints that were equal to the slope of the beams. Like the straight beams, the
double pitched beams were tested in four-point bending tests. The vertical compo-
nent of the local deflection between the load application points was measured within
a length of 2.4 m. The dimensions of the tested double-pitched CLT beams and the
test setup are shown in Figure 7.

F F
α = 5°, 15°, 25° uglob LFJ

α
ulok
600

2400

F F

3000 3000 3000

Figure 7. Test setup for double pitched CLT beams (series PI-05, PI-15 and PI-25)

In three test series CLT frame corners with opening angles β of 90°, 105° and 115° were
tested. Each corner frame contained two finger-jointed connections with cutting angles
α of 22.5°, 18.75° and 16.25°. To induce a closing moment in the finger-jointed connec-
tions in the corner a compressive force was applied to the ends of the legs of the spec-
imens. During the tests the overall deformation in the line of action of the compressive
forces and the local deformation at the finger-jointed connections were measured. The
dimensions of the corner frames and the test setup are illustrated in Figure 8. The ar-
rangement of displacement transducers in the corners is shown in Figure 9.

237
INTER / 48 - 12 - 05

e = 1926 e = 1495 e = 1250

F F F

0
60 60
0

0
60
uges LFJ uges LFJ uges LFJ

β = 105°

β = 115°
β = 90°
4500

400

400
400 LFJ LFJ LFJ

F F F

Figure 8. Test setup and dimensions of CLT frame corners (series CO-90, CO-105 and CO-115)

ulok,1
40

ulok,2
40

ulok,4
ulok,3
290
290

Figure 9. Measurement of local deformation in LFJ of tested frame corners

3.3 Results
3.3.1 Load carrying capacity of LFJ in CLT
In all test specimens with LFJ failure occurred due to bending stresses in the tension
zone of the finger jointed connection. In the test series with straight beams and dou-
ble pitched beams the failure was brittle and occurred abruptly without prior indica-
tion. In the test series with frame corners compressive wrinkles were observed in

238
 INTER / 48 - 12 - 05

nearly all specimens before the ultimate load was reached. Figure 10 through Figure
12 show examples of the observed failures.

Figure 10. Typical failure in straight CLT beams with LFJ (left) and in CLT beams without LFJ (right)

Figure 11. Typical failure in double pitched CLT beams

Figure 12. Typical failure in CLT frame corners

From the ultimate loads of the individual tests the normal stresses parallel to the
grain in the extreme fibre of longitudinal layers were calculated for all tests series.
For straight beams and double pitched beams the stresses were calculated as
M/Wnet,long, for the corner frames the normal stresses at the inward corner resulting
from combined bending and compression were calculated as (M/Wnet,long + N/Anet,long).
The section modulus Wnet,long and the area Anet,long used for the calculation of stresses
are the section properties of longitudinal layers in a cross sections perpendicular to
the beam axis. For all test series the obtained values are summarized in Table 2. In
addition to the stresses the effective reduction of the cross section νeff was calculated

239
INTER / 48 - 12 - 05

for every failed LFJ according Eq. (1) using the measured values of the tip gap s and
the production tolerance ∆ in thickness direction. In the last row of Table 2 the values
of (1-νeff) are given, that represent the proportion of the remaining cross section in
the tested finger-jointed connections.
For the tested straight CLT beams the ratio 20.3/29.2 = 0.698 between the mean values
of the bending strength determined for beams with and without LFJ is almost equal to
the value of (1-νeff,mean) = 0.726 that was determined for LFJ in straight CLT beams in test
series S-LFJ. Consequently, the bending strength of LFJ can be calculated by multiplica-
tion of the in-plane bending strength of the CLT member with the factor (1-νeff).

Table 2. Bending strength of LFJ related to the net cross section of longitudinal layers

series n α LFJ
f m,mean LFJ 1)
f m,k 1-νeff,mean

- in ° in N/mm² in N/mm² -
S-REF 15 0 29.2 23.0
S-LFJ 15 0 20.3 17.1 0.726
PI-05 5 5 17.7 14.3 0.747
PI-15 5 15 11.8 9.8 0.719
PI-25 5 25 10.3 9.0 0.730
CO-90. 5 16.25 -24.8 -21.9 0.694
CO-105 5 18.75 -24.1 -19.1 0.697
CO-115 5 22.50 -23.7 -22.2 0.693
1)
determined according EN 14358

The test results for the double pitched CLT beams show that for LFJ subjected to opening
moments the bending strength strongly decreases with increasing angle α. For LFJ sub-
jected to closing moments, in contrast, the bending strength is significantly larger than
the bending strength of LFJ in straight beams and more or less independent of the angle
α. To find an analytical solution for the calculation of the bending strength of LFJ in de-
pendence of the cutting angle α the approaches of Norris Eq. (2) and Hankinson Eq. (3)
for the calculation of the strength reduction factor kα were used. In both equations fac-
tors a and b were introduced to be able to adapt the analytical solutions to test results.

1
k αLFJ = 2 2
 (1 - ν eff ) ⋅ f m,k
CLT
  (1 - ν eff ) ⋅ f m,k
CLT
 (2)
 ⋅ sin 2
α +
  ⋅ sin α ⋅ cos α  + cos α
4

 a ⋅ f t,90,k   b ⋅ f v,k
CLT BSP


240
 INTER / 48 - 12 - 05

1
k αLFJ =
 (1 - ν eff ) ⋅ f
CLT
  (1 - ν eff ) ⋅ f m,k
CLT
 (3)

m,k
⋅ sin α  + 
2
⋅ sin α ⋅ cos α  + cos 2 α
 a ⋅ f t,90,k   b ⋅ f v,k
CLT CLT


The strength reduction factors for the tested LFJ were calculated with the bending
= (1 - ν eff ) ⋅ f m,k
LFJ
strength f m,k CLT
= 17.1 N/mm² obtained from test series S600LFJ for
straight CLT beams with LFJ. The effective tensile strength perpendicular to the grain
and the effective shear strength were calculated according Eq. (4) and Eq. (5), given
in Blaß and Flaig (2013) and in Blaß and Flaig (2014), respectively.

 t net,cross lam 
t ⋅ f t,0,k 
 net,long 
CLT
f t,90,k = min   (4)
 n CA ⋅ b ⋅ f lam 
 t net,long R,k 

 t gross 
t ⋅ f lam
v,k 


net,long

f v,kCLT = min  n CA ⋅ b 1  (5)
2 ⋅ t ⋅
1  1  2  1 1 
 net,long
lam
⋅  1 − 2  + lam ⋅  − 2  
 f v,tor,k  m  f R,k  m m  

For the calculation of the effective strength properties of the tested double pitched beams
with m = 4 lamellae in direction of the beam height and nCA = 4 crossing areas in thickness
direction the strength properties given in Table 3 were assumed for the lamellae.

Table 3. Assumed strength properties of lamellae

lam
f t,0,k in N/mm² f v,klam in N/mm² f R,klam in N/mm² lam
f v,tor,k in N/mm²
14.0 4.0 1.1 2.75

By substituting the given values in Eq. (4) and Eq. (5) an effective tensile strength
CLT
perpendicular to the beam axis of f t,90,k = 3,5 N/mm² and an effective shear strength
CLT
of f v,k = 2,75 N/mm² were calculated.
The strength properties were then used to calculate strength reduction factors accord-
ing Eq. (2) and Eq. (3). The best agreement with the test results was found for the modi-
LFJ
fied Hankinson’s equation with factors a = 2.5 and b = 2.5. The bending strength f m,α
calculated from Eq. (3) and the respective values obtained from test results are illus-
trated in Figure 13. Since the experimentally obtained values of f m,LFJα include different

241
INTER / 48 - 12 - 05

values of (1-νeff) the bending strengths in the diagram are related to the residual thick-
ness in the finger jointed connection teff,LFJ = tnet,long ∙ (1-νeff).
36
S-LFJ
S600KZV
32
PI-05
SD05
PI-15
SD15
28
- ν-νeffeff)) ininN/mm²
N/mm²

SD25
PI-25
24 mean
meanvalues
values
k α ⋅ f m,mean
kalpha
CLT
* fm,CLT,mean
20
k α ⋅ f m,k
*CLTfm,CLT,k
/ (1

kalpha
m,α /(1
ffm,α,LFJ

16
LFJ

12

8
-5 0 5 10 15 20 25 30
in °°
ααin
Figure 13. Experimentally and analyticlly obtained bending strength fm,α,LFJ of LFJ in CLT beams
subjected to opening moments

3.3.2 Stiffness of LFJ in CLT

From the measured deflection, the local MOE of the tested straight and double
pitched CLT beams was evaluated according to Eq. (6). In the evaluation of the tests
with double pitched CLT beams, the influence of the larger length and the direction
of the measured deflection were taken into account by the factor cosα in the de-
nominator. For the straight beams, in addition, the dynamic MOE was determined
from the eigenfrequencies of the longitudinal vibration of the beams according
Eq.(7). The obtained mean values of the MOE for the different test series are summa-
rized in Table 4. The values are related to the net cross section of longitudinal layers.

2 ΔM10−40  2m
E loc,net = ⋅ ⋅
16 Δu 10−40 ⋅ cos α I net,long
(6)
where  m is the measuring length (3,000 mm or 2,400 mm)

t gross
E dyn,net = 4 ⋅ f 2 ⋅ L2 ⋅ ρ ⋅ (7)
t net,long

242
 INTER / 48 - 12 - 05

Table 4. MOE of the tested straight and double pitched CLT beams
Series Eloc,net,mean COV Edyn,net,mean COV Eloc/Edyn
in N/mm² - in N/mm² - -
S-REF 12,385 0.085 12,203 0.054 1.015
S-LFJ 11,283 0.045 11,884 0.035 0.949
PI-05 11,488 0.050 - - -
PI-15 10,642 0.031 - - -
PI-25 9,432 0.030 - - -

The decreasing values in the second column of Table 4 indicate that LFJs have signifi-
cant influence on the overall deformation of finger jointed CLT beams. For straight
beams with LFJ the additional deformation in the LFJ results in about 9% decreased
values of the MOE compared to CLT beams without LFJ. In the tested double pitched
CLT beams, the influence of the LFJ on the deflection strongly increases with increas-
ing cutting angle α. To determine the stiffness of LFJ in CLT beams in dependence of
the cutting angle α, the mutual displacement between the joined parts was meas-
ured in the test series with CLT frame corners (cf. Figure 9). Figure 14 shows an ex-
ample of the obtained load displacement curves. The two curves in the third quad-
rant of the diagram clearly show that in the compression zone of the tested LFJ the
strength was reached well before the ultimate load and that considerable non-elastic
deformation occurred.
200
ulok,1 (upper LFJ, outward corner)
150
ulok,2 (upper LFJ, inward corner)

ulok,3 (lower LFJ, outward corner)

100
ulok,4 (lower LFJ, inward corner) outward corner
(tension)
50
force in kN

0
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0
-50
inward corner
(compression)
-100

-150

-200
displacement in mm

Figure 14. local displacements u lok,i in the LFJ in tested CLT frame corner CO105-05

243
INTER / 48 - 12 - 05

The stiffness of the tested LFJ was evaluated on the assumption that the measured
displacements are composed of two parts resulting from the bending moment and
the normal force acting in the corner.

= u N,i + u M,i
u loc,i (8)

The two components u N and u M of the displacement were calculated according to

Eq. (9) and Eq. (10) where k is a bedding factor describing the stiffness of the LFJ in
dependence of the cross sectional area Anet,LFJ and the second moment of inertia
Inet,LFJ. The cross sectional values Anet,LFJ and Inet,LFJ were calculated using the net thick-
ness of longitudinal layers and the height of the LFJ hnet,LFJ = h/cos α.
In Eq. (10) e is the distance between the line of action of the applied force F and the
center line of the vertical part of CLT frame corner (cf. Figure 8) and 290 mm is the
distance of the displacement transducers from the center line of the CLT beams
measured in direction of the LFJ (cf. Figure 9).

F ⋅ cos α
uN = (9)
k ⋅ Anet,LFJ

F ⋅ e ⋅ 290mm
uM = ± (10)
k ⋅ I y,net,LFJ

Substituting Eq. (9) and Eq. (10) into Eq. (8) and solving the equation for the bedding
factor k yields the expression in Eq. (11).

F  cos α e ⋅ 290mm 
k = ⋅ ±  (11)
u lok  Anet,LFJ I y,net,LFJ 

In Table 5 minimum, mean and maximum values of the bedding factor k evaluated for
the three test series with CLT frame corners are summarized.

Table 5. Bedding factors k of LFJ evaluated from local displacements in CLT frame corners
series kmin kmean kmax
in N/mm³ in N/mm³ in N/mm³
CO-90 54,4 65,2 71,5
CO-105 64,6 76,3 88,1
CO-115 75,1 84,2 108

By means of a linear regression the relationship given in Eq. (12) was found between
the bedding factor k and the cutting angle α.

244
 INTER / 48 - 12 - 05

k =133 − 3,01 ⋅ α with k in N/mm³ and α in °

(12)
= =
r 0,425 =
s R 9,38 n 15

In Figure 15 the bedding factor k evaluated from the test results are plotted against
the cutting angle α. In the diagram, the linear regression according Eq. (12) is illus-
trated by a dotted line that was extrapolated to α = 0°.
140
130
bedding factor k in N/mm³

120
110
100
90
80 individual test
70 results
mean values
60
of test series
50
0 5 10 15 20 25
cutting angle α in °

Figure 15. Bedding factor k of LFJ plotted against the cutting angle α
Bedding factors k calculated according Eq. (12) were used to determine the MOE of
the tested straight and double pitched CLT beams in account of the stiffness of the
k
LFJ. The resulting values of E loc,mean are summarized in Table 6. Although, the bedding
factors used for the new evaluation had to be extrapolated far beyond the tested cut-
k
ting angles, the values E loc,mean determined for straight CLT beams with LFJ differ only
REF
slightly from the value E loc,mean = 12,385 N/mm² that was determined for straight CLT
beams without LFJ. For the double pitched CLT beams of series, PI-05 and PI-15 with
cutting angles of 5° and 15° the agreement is also very good. For the double pitched
beams of series PI-25, the obtained values of the MOE still differ significantly.

Table 6. MOE of the tested straight and double pitched CLT beams evaluated with and without con-
sideration of the bedding factor k of LFJ

Reihe k E loc,mean k
E loc,mean REF
/ E loc,mean

in N/mm³ in N/mm² -
S-REF - 12385 1
S-LFJ 133 12.104 0.98
PI-05 118 12.417 1.00
PI-15 88 11.665 0.94
PI-25 58 10.529 0.85

245
INTER / 48 - 12 - 05

4 Design proposal
4.1 LFJ in straight CLT beams
The comparison of the values in the first two lines of Table 2 shows that there is good
agreement between the value (1-νeff) = 0.726 for straight beams and the ratio of
20.3/29.2 = 0.70 between the bending strength of straight beams with and without
LFJ. It is therefore suggested to calculate the bending strength of LFJ in straight CLT
beams loaded in plane direction by multiplying bending strength of CLT member with
the value (1-νeff). In general cases it is recommended to use a value of νeff = 0.30 to
take into account the effective reduction of the cross section.

= LFJ
f m,k (1 - ν eff ) ⋅ f m,k
CLT
(13)

For the design in the ultimate limit state, the bending stresses should be calculated
according Eq. (14) using the section modulus Wnet,long of longitudinal layers and satisfy
the condition given in Eq. (15).

M
σ m,net = (14)
Wnet,long

σ m,net,d
LFJ
≤1 (15)
f m,d

4.2 LFJ in CLT beams with kinked axis

For LFJ in CLT beams subjected to opening moments the bending strength can be cal-
culated from the modified Hankinson’s relation given in Eq. (16) that was derived
from the test results by customizing the factors for shear strength and the tensile
strength perpendicular to beam axis. The equation is valid for angles 0 < α ≤ 25°.

(1 - ν eff ) ⋅ f m,k
CLT
f LFJ
m,α ,k =
 (1 - ν eff ) ⋅ f m,k
CLT
  (1 - ν eff ) ⋅ f m,kCLT
 (16)
 ⋅ sin α  + 
2
⋅ sin α ⋅ cos α  + cos α
2

 2,5 ⋅ f t,90,k   2,5 ⋅ f v,k

CLT CLT


For the design of LFJ in CLT beams subjected to closing moments, in general, the
bending strength of straight CLT beams with LFJ can be used.

=
f m,LFJα ,k (1 - ν eff ) ⋅ f m,k
CLT
(17)

For the design in service class 1 according to EC 5, the bending strength according
Eq. (17) can be increased by 50 %. Due to the strong influence of the moisture con-
tent on the compressive strength of softwood, the higher bending strength should
not be used in service class 2.

246
 INTER / 48 - 12 - 05

Stresses in LFJ in kinked CLT beams should be calculated using the section modulus
Wnet,long and the area Anet,long of longitudinal layers in a cross section perpendicular to
the beam axis. For the verification of stresses resulting from combined bending and
tension or from combined bending and compression, the following procedure is sug-
gested.
• Calculation of normal stresses σm = M/Wnet,long and σN = N/Anet,long related to the
net cross section of longitudinal layers.
• Calculation of the maximum normal stress by linear superposition of σm and σN
• Calculation of the bending strength f m,LFJα according Eq. (16) or Eq. (17) in depend-
ence of the cutting angle and verification by comparison of the maximum normal
stress and the bending strength
In the diagram in Figure 16 the characteristic values f m,LFJα ,k according Eq. (16) and
Eq. (17) are depicted by the red curve. The values were calculated using a character-
istic bending strength f m,CLTα ,k of 24 N/mm² and an effective reduction of the cross sec-
tion of νeff = 0.30. The results of all test series are also given in the diagram. On the
abscissa the cutting angle α is given. Positive values of α indicate opening moments
whereas negative angles stand for closing moments.

30
compressive stress at inward corner
27 tensile stress at outward corner
24
in N/mm²
in N/mm²

21

18
m,α,LFJ
ffm,
LFJ

15
test results
12
mean values
9 adaption to mean values
design proposal (char. values)
6
-30 -20 -10 0 10 20 30
in°°
α in
LFJ
Figure 16. Bending strength f m,α in dependence of the angle α

247
INTER / 48 - 12 - 05

5 Summary and Conclusions

A new finger joint profile, developed for LFJ in CLT beams, is presented. To determine
the reduction of the bending strength resulting from LFJ, tests with straight CLT beams
with and without LFJ were performed. Based on the test results a reduction factor of (1
– 0.30) = 0.70 is proposed for the bending strength of LFJ in straight beams. The factor
includes the nominal reduction resulting from the finger joint profile and further re-
duction due to production tolerances and manufacturing processes.
For LFJ in CLT beams jointed under varying angles the influence of the cutting angle α
on the bending strength was determined from tests with double pitched CLT beams
and CLT frame corners. The results of tests show that the bending strength of LFJ in
CLT is far less dependent on the cutting angle than the bending strength of LFJ in glu-
lam. For LFJ subjected to opening moments the reduction of the bending strength for
angles up to 25° is less than 50 %. For LFJ subjected to closing moments even a greater
bending strength was found than for straight beams. Based on the test results a design
proposal for LFJ in CLT beams is made, taking into account the strength reduction re-
sulting from the LFJ and the influence of the cutting angle α.

6 References
Aicher S (2003): Structural Adhesive Joints Including Glued-In Bolts. In: Timber Engi-
neering, Chapter 18, pp 333 – 363, Jon Wiley and Son LTD, Chichester
Flaig M, Blaß H J (2015): Keilgezinkte Rahmenecken und Satteldachträger aus Brett-
sperrholz. Karlsruher Berichte zum Ingenieurholzbau Band 29, KIT Scientific Pub-
lishing, Karlsruhe
Flaig M., Blaß H.J. (2014): Tapered beams made of cross laminated timber. In: Mate-
rials and Joints in Timber Structures. Springer, Berlin
Flaig M., Blaß H.J. (2013): Shear Strength and Shear Stiffness of CLT beams Loaded in
Plane. In: Proceedings. CIB-W18 Meeting 46, Vancouver, Canada, Paper 46-12-3
Hankinson R L (1921): Investigation of crushing strength of spruce at varying angles of
grain. Air service Information Circular III, No. 259, US Air Service
Heimeshoff B (1976): Berechnung von Rahmenecken mit Keilzinkenverbindungen. In:
Holzbau Statik Aktuell, Folge 1, S. 7-8
Heimeshoff B, Seuß R (1982) Berechnung von Rahmenecken mit Keilzinkenverbin-
dungen. Forschungsbericht, Universität München
Kolb H (1966): Versuche zur Ermittlung der Tragfähigkeit geleimter Rahmenecken. In:
Bauen mit Holz, Heft 8, S. 363-369, Bruderverlag, Karlsruhe
Komatsu K et. al (2001): Moment-resisting performance of glulam beam to column
joints composed of various types of large finger joints. In: Proceedings of the Inter-
national RILEM Symposium, Stuttgart, Germany, pp. 520-530

248
 INTER / 48 - 12 - 05

Discussion

The paper was presented by M Flaig

G Schickhofer asked about the 0.3 reduction factor as a constant and suggested it
would be better to have this as a function.
S Winter asked why there was a nominal reduction of 14% but found a 30% reduction
for the profile. M Flaig responded that these specimens were from industry and they
did not have control over the production accuracy. Here, optimized production of CLT
as beams was not yet available.
W Seim received clarification that Elocal as shear free MOE. He asked for examples of
where these members could replace glulam. M Flaig responded that CLT beam has
higher strengths in the perpendicular to grain direction therefore, these members
could be suitable for cases where higher stress in perpendicular to grain directions as
well as cases where cracks might be important.
L Franzoni asked given a straight beam why were those dimensions given. M Flaig
said that bending strength of straight beams would depend on the specimen size; so
standardized test span to depth ratio of 18 to 25 are typically used. In the current
tests finger joint governed so the specimen length did not matter.
G Fink commented that based a sample size of 5 to get a characteristic strength might
be too low. M Flaig agreed but stated that it would be too expensive to test many
beams and that they relied on prior knowledge about CLT as beam element and used
this information as guidance. G Fink commented they could consider different models
to estimate characteristic properties.
T Bogensperger was surprised by the large stiffness degradation and asked if the
strength depended on the glue. M Flaig stated that they did not know whether differ-
ent glue would affect the strength but did not expect so. He also did not expect the
large stiffness degradation at first but the test results clearly showed this.
T Bogensperger also asked about gluing large finger joint on site. M Flaig stated that
it could be done but would not be recommended.

249
250
 INTER / 48 - 15 - 01

Performance-based seismic design of

light-frame structures – Proposed values
for equivalent damping

Johannes Hummel
Werner Seim
University of Kassel, Department of Structural Engineering,
Kurt-Wolters-Straße 3, 34125 Kassel, Germany.

Keywords: light-frame, cyclic behaviour, dynamic behaviour, seismic design

1 Summary
This paper discusses different methods for performance-based design. In this
context, damped and inelastic spectra will be addressed as demand spectra. How to
create demand spectra within the capacity spectrum method and the N2 method will
be explained. The challenges which occur for timber structures in performance-based
design with the capacity spectrum method and the N2 method will be identified.
A constant approach for equivalent damping will be proposed for light-frame
structures for application within the capacity spectrum method. Values for equivalent
damping can be derived directly from cyclic testing. A fairly constant progression of
the values was observed here over increasing wall drift.
The constant approach for equivalent damping is applied to validate results from
dynamic testing. Test results will be presented for two dynamic tests. The capacity
spectrum method, according to ATC 40, and the N2 method, according to Eurocode 8
(2010), will also be applied for comparison.

251
INTER / 48 - 15 - 01

2 Introduction
There are basically three design methods for buildings under earthquake impact
which are introduced with Eurocode 8 (2010). The most common method for
engineers in practice is the so-called force-based design. This method uses elastic
response spectra scaled by a behaviour factor q to calculate the base shear, which is
equivalent to the total of seismic forces. In addition, the time history analysis is used
as the most comprehensive calculation method. The time history analysis needs
accelerograms of real earthquakes or synthetically generated accelerograms, a
definition of nonlinear hysteretic material behaviour and a powerful computer
programme to determine displacements and stresses for the structure over time
history. Performance-based design procedures can be classified more or less
between force-based design and time history analyses.
Different performance-based design procedures have been developed in the last five
decades. The most relevant procedures in Europe are the capacity spectrum method
(Freeman, 1998) and the N2 method (Fajfar, 1999), which have been critically
discussed by Chopra & Goel (1999) and Freeman (2004). The direct-displacement
method (Priestley & Kowalsky, 2000) is also common in North America. These
methods differ slightly in the analysis steps and mainly in the definition of seismic
action. Whereas the capacity spectrum method and the direct-displacement method
use elastic response spectra, which are scaled by equivalent damping ξ’eq, the N2
method makes use of inelastic spectra.
Two formats of the response spectrum are commonly applied within performance-
based design procedures: acceleration over displacement (Sa-Sd format) and
displacement over period (Sd-T format). The spectrum in the Sa-Sd format is especially
convenient for a graphical solution (see Figure 1a). Since the various codes usually
provide elastic response spectra in the format elastic spectral acceleration Se over
period T, the spectral values for the displacement should be determined with
equation (1).
T2
Sd   Sa with Sa  Se (T ) (1)
4  π2
Code spectra usually consider a proportion of 5 % viscous damping. For other values
of viscous damping, the spectra can be adapted by the damping correction factor η,
see equation (2), according to Eurocode 8.
10
η  0.55 (2)
5ξ
Inelastic spectra can be generated according to equation (3)
Se (T ) μ T2
Sa  and Sd    Se (T ) (3)
Rµ Rµ 4  π2

252
 INTER / 48 - 15 - 01

with the reduction factor Rµ depending on ductility µ.

A bilinear Rµ-µ-T relationship was proposed for the N2 method by Vidic et al. (1994).
Here, the reduction factor Rµ increases linearly with period T up to the “corner
period” TC (see Eurocode 8 and Figure 1a), and from that point the reduction factor
remains constant and is equal to the ductility µ:
T
Rµ  (µ  1)  1 for T  TC
TC (4)
Rµ  µ for T  TC
Here, ductility µ is defined as the ratio between ultimate displacement uu and the
yield displacement uy (see Figure 1b). The Rµ-µ-T relationship according to equation
(4) is based on a hysteretic model – the Q-model – which is representative for
reinforced concrete frame constructions (Saidi & Sozen, 1979).

(a) (b)
Figure 1. (a) Comparison of inelastic and damped spectra; (b) bilinear idealisation of the load-
diplacement curve of the structure.
It must be noted that the Q-model does not necessarily match the hysteretic
behaviour of timber structures, since it does not capture pinching effects. This issue
was also addressed by Fragiacomo et al. (2011) with reference to CLT buildings.
Consideration of the realistic hysteretic behaviour seems to be indispensable,
because it accounts for the energy dissipation.
Another way to consider energy dissipation is to describe the hysteretic behaviour by
equivalent damping. ATC 40 (1996) provides an approach for equivalent damping
depending on the associated displacement, respectively, the related ductility, and
hysteretic behaviour based on a bilinear idealisation of the load-displacement curve
(see Figure 1b). Furthermore, Filiatrault et al. (2002) proposed an equivalent
damping approach for light-frame structures focused on an application with the
direct-displacement method. A bilinear relationship for equivalent damping over

253
INTER / 48 - 15 - 01

building drift was found by applying the CASHEW hysteretic model (Folz & Filiatrault,
2001).
In contrast to the inelastic spectrum, which can be determined directly with equation
(3), there are always series of damped spectra for different values of ξ’eq if equivalent
damping depends on the building drift. A transition curve (see Figure 1a) might be
found if a formulation for damping as a function of displacement is available (Kawai,
1999).
Applying the N2 method, according to Eurocode 8, to timber structures without
further considerations seems to be questionable. There are certainly a number of
other Rµ-µ-T relationships (e.g. Miranda & Bertero, 1994), but none of these
approaches were validated for timber structures. The hysteretic behaviour can differ
significantly between the various construction types, even for timber structures, such
as light-frame, cross laminated timber, moment resisting frames and others (see Seim
& Hummel, 2013; Seim & Vogt, 2013; Seim & Schick, 2013).
On the other hand, different hysteretic behaviour can be considered comparatively
simply by equivalent damping. The hysteretic behaviour can be determined by
standardised experimental investigations (see section 3) and approaches for
equivalent damping can be derived directly from test results (see section 4). Anyway,
experimental investigations – cyclic and dynamic tests – are necessary for any
structural material or construction type to get basic information about the
characteristics under earthquake impact and to validate design procedures.
Thus, in the author’s opinion, it seems to be more promising to utilise equivalent
damping to determine demand spectra with reference to performance-based design
of timber structures. The adverse iteration process, which comes with the capacity
spectrum method to obtain the effective damping value for defining the demand
spectrum, can be avoided by means of a simplified formulation of equivalent
damping (see section 4).

3 Experimental Investigations
Extensive experimental investigations on wall elements – light-frame and cross
laminated timber – were carried out at the University of Kassel within the research
project Optimberquake. The wall elements were tested under cyclic and dynamic
loading.
3.1 Programme and Test Set-up
A total of 18 cyclic tests were performed with light-frame wall elements (see Table 1).
Each of these elements had the dimensions 2.50 m × 2.50 m. A horizontal load was
applied by a horizontal actuator on the top of the wall element. The vertical load was
applied by two vertical actuators (see Figure 2a). The horizontal load and the
corresponding horizontal displacement were measured on the top of the wall. Details

254
 INTER / 48 - 15 - 01

concerning the cyclic tests can be found in the documentation (Seim & Hummel,
2013; Seim & Vogt, 2013).

(a) (b)
Figure 2. Test set-ups: (a) cyclic testing; (b) uniaxial shaking table.

Table 1. Cyclic tests on light-frame wall elements.

Fasteners Loading Vert. Load

Test Sheathing1) Anchoring
Type ø - ℓ [mm] Spacing [mm] protocol
3) [kN/m]
WL-1.2
2×OSB ISO21581
WL-1.3 Nail 2.8 - 65 75/150 symmetric 10
18 mm
WL-1.4 CUREE
WL-2.2
2×GFB ISO21581
WL-2.3 Staple 1.53 – 552) 75/150 symmetric 10
18 mm
WL-2.4 CUREE
WL-3.1 1×OSB ISO21581
WL-3.2 10 mm CUREE
Nail 2.8 - 65 75/150 symmetric 10
WL-3.3 1×OSB ISO21581
WL-3.4 18 mm CUREE
WL-4.1 2×OSB Nail 2.8 - 65
18 mm

WL-4.2 2×GFB Staple 1.53 – 552)

75/150 ISO21581 asymmetric 10
WL-4.3 1×OSB Nail 2.8 - 65
WL-4.4 1×GFB Staple 1.53 – 552)
WL-5.1 1×GFB ISO21581
WL-5.2 10 mm CUREE
Staple 1.53 – 552) 75/150 symmetric 10
WL-5.3 1×GFB ISO21581
WL-5.4 18 mm CUREE
1)
“1” indicates sheathing at one side, “2” indicates sheathing at two sides
2)
The thickness of the wire is 1.53 mm and length of the staple from the crown to the tip is 55 mm.
3)
First value along the edge of the sheathing board, second value along the intermediate studs.

255
INTER / 48 - 15 - 01

Furthermore, five light-frame elements were tested under dynamic loading (see
Table 2) on the uniaxial shaking table of the University of Kassel (see Figure 2b).
The configuration of the light-frame wall elements for dynamic testing was chosen
according to the test set-up of the cyclic tests. Two different materials – OSB and GFB
(gypsum fibre board) – were used for the sheathing with thicknesses of 18 mm and
10 mm, respectively. Dimensions of the sheathing boards are 1.25 m by 2.50 m. The
spacing of the studs (C24, 140 mm × 60 mm) was 625 mm. Four specimens had the
same dimensions as in the cyclic tests. The half wall length (ℓ = 1.25 m) was chosen
for one element. The wall elements were anchored with hold-downs at both ends
against uplift. Shear steel plates were used as end stops (block shear connectors) to
take the horizontal force. Hold-downs and steel plates were fixed to the base plate of
the shaking table. Each hold-down (Type HTT 22 SIMPSON, 2011) was connected to
the outer stud by means of 18 annular ring shank nails (ø4 - 60 mm). The load-bearing
capacity of the anchoring amounts to 50 kN at the minimum which was obtained by
monotonic and cyclic testing (Seim, Hummel & Vogt, 2013).
Table 2. Specimen and loading for dynamic tests.

Dimensions Fasteners Time Vert. Load

Test Sheathing1)
[m] × [m] Type ø - ℓ [mm] Spacing3) [mm] history [kN/m]
1×OSB synthetic
WL-dyn-1 2.50 × 2.50 Nail 2.8 - 65 75/150 9.5
18 mm earthquake
1×OSB El Centro,
WL-dyn-2 2.50 × 2.50 Nail 2.8 - 65 75/150 9.5
18 mm 1940
1×OSB synthetic
WL-dyn-3 2.50 × 2.50 Nail 2.8 - 65 75/150 9.5
10 mm earthquake
1×OSB synthetic
WL-dyn-4 1.25 × 2.50 Nail 2.8 - 65 75/150 9.5
10 mm earthquake
1×GFB synthetic
WL-dyn-7 2.50 × 2.50 Staple 1.53 – 552) 75/150 9.5
18 mm earthquake
1)
“1” indicates sheathing at one side
2)
The thickness of the wire is 1.53 mm and length of the staple from the crown to the tip is 55 mm.
3)
First value along the edge of the sheathing board, second value along the intermediate studs.

The vertical load was applied by means of steel plates fixed on the top of the wall
element. The shaking table uses a displacement-controlled regulation where the
displacement capacity is ± 200 mm. The signal for the displacement is calculated from
the acceleration time history (see Figure 3). The target acceleration is measured
directly on the shaking table. In addition, reaction force (H-Load), top displacement,
relative displacement between the top and bottom of the wall element, and the
acceleration on the top of the wall were recorded.
An artificial time history compatible with the elastic response spectrum with ground
conditions A-R, according to Eurocode 8/NA (see Figure 3), and the north-south
component of the El Centro earthquake (1940) were used as input data for the target

256
 INTER / 48 - 15 - 01

signal (acceleration). The time histories were scaled to 1 g in order to reach the load-
bearing capacity of the wall elements and to remain in the displacement range of the
shaking table.

Figure 3. Artificial time history compatible to the elastic response spectrum with ground conditions
A-R, according to Eurocode 8/NA (synthetic earthquake).

3.2 Test Results

Test results of the cyclic tests have already been published (Seim et al., 2014) and are
fully documented in test reports (Seim & Hummel, 2013; Seim & Vogt, 2013).
The dynamic test was initially started with a smaller earthquake, where the peak
ground acceleration was 0.5 g to 0.6 g. This step was necessary to figure out the
interaction between the shaking table and wall element. The peak ground
acceleration was then increased to 1 g in order to reach the load-bearing capacity of
the element.

(a) (b)
Figure 4. Results of test WL-dyn-1: (a) relative displacement over time and (b) top acceleration over
time.

257
INTER / 48 - 15 - 01

Figures 4 and 5 present the test results of the relative displacement and top
acceleration for the two tests WL-dyn-1 and WL-dyn-3. The relative displacement is
the difference between the top displacement and the displacement of the shaking
table (“table displacement”), see Figure 2b.

(a) (b)
Figure 5. Results of test WL-dyn-3: (a) relative displacement over time and (b) top acceleration over
time.

WL-dyn-1 and WL-dyn-3 show a similar behaviour with respect to the progression of
the top acceleration (see Figures 4b and 5b). However, the relative displacements
exhibit a slightly different characteristic (see Figures 4a and 5a). A completely
different behaviour was obtained in test WL-dyn-2, where a real earthquake (El
Centro, see Table 2) was applied. Obviously, significantly higher relative
displacements occurred. The maximum acceleration is also higher.
Table 3. Results of dynamic tests, absolute values
H-Load Table disp. Top disp. Top acc.
Test
[kN] [mm] [mm] [g]
WL-dyn-1 40.3 87.7 106.2 1.3
WL-dyn-2 45.5 153.8 236.4 1.8
WL-dyn-3 37.9 88.2 121.4 1.2
WL-dyn-4 25.3 86.4 125.7 1.4
WL-dyn-7 50.5 103.8 138.5 1.8

The absolute values of top acceleration, top displacement, table displacement and H-
Load (reaction force) for the dynamic test are summarised in Table 3.

258
 INTER / 48 - 15 - 01

An amplification of the acceleration from the bottom to the top can be noticed, since
the maximum acceleration of the shaking table is 1 g (see section 3.1) and the top
acceleration goes up to 1.8 g (see Table 3).

4 Equivalent damping approach

4.1 Definition
Equivalent damping ξ’eq is defined in the following as a superposition of viscous
damping ξ0 and equivalent viscous damping ξeq:
'
ξeq  ξ0  ξeq (5)
The nominal damping ratio ξ0 accounts for energy dissipation within linear elastic
behaviour and considers, for example, damping effects which are caused by friction.
The damping value ξ0 is typically determined by means of experimental tests and
approaches are available in the literature (e.g. Petersen, 1996). Nominal damping
ratios between 2% and 3% of critical damping are given for timber structures.
The equivalent viscous damping value ξeq represents energy dissipation which
appears due to nonlinear hysteretic behaviour. Equivalent viscous damping was first
defined by Chopra (1995) as an approximation to be compared to linear viscous
damping. The values of equivalent viscous damping can be derived from data for load
and displacement from cyclic testing (see section 3.2).
Values of equivalent viscous damping can be determined as
Ed
ξ eq  νeq  , (6)
4 π  Ep
where Ed is the dissipated energy for one cycle – enclosed area of one hysteresis loop
– and Ep is the potential energy determined from the current stiffness corresponding
to the maximum amplitude u per cycle (see Figure 6a).

(a) (b)
Figure 6. Definition of equivalent viscous damping: (a) according to Chopra (1995) and (b) according
to ATC (1996).

259
INTER / 48 - 15 - 01

4.2 Test Results

Data from a total of 18 cyclic tests on light-frame wall elements were evaluated for
equivalent viscous damping. The damping values ξeq calculated from the first and the
third cycle of the hysteresis are summarised in Table 4. Minimum, maximum, mean
and median values are specifically reported. The damping values for the second cycle
differ only slightly from those of the third cycle, therefore, the results of the second
cycle are omitted.

Table 4. Values of equivalent viscous damping for light-frame elements.

First Cycle Third Cycle
Test
Min Max Mean Median Min Max Mean Median
WL-1.2 0.106 0.143 0.127 0.132 0.094 0.132 0.112 0.111
WL-1.3 0.126 0.168 0.146 0.143 0.115 0.131 0.122 0.123
WL-1.4 0.149 0.167 0.161 0.162 0.131 0.227 0.160 0.142
WL-2.2 0.161 0.207 0.185 0.185 0.153 0.174 0.162 0.161
WL-2.3 0.155 0.263 0.196 0.181 0.161 0.197 0.176 0.169
WL-2.4 0.161 0.188 0.174 0.173 0.167 0.254 0.203 0.191
WL-3.1 0.172 0.240 0.199 0.194 0.165 0.200 0.176 0.169
WL-3.2 0.172 0.198 0.187 0.188 0.178 0.246 0.214 0.215
WL-3.3 0.195 0.278 0.226 0.220 0.164 0.209 0.189 0.190
WL-3.4 0.209 0.248 0.219 0.215 0.186 0.297 0.236 0.233
WL-4.1 0.153 0.167 0.160 0.160 0.122 0.159 0.135 0.130
WL-4.2 0.143 0.292 0.185 0.164 0.095 0.189 0.135 0.131
WL-4.3 0.161 0.174 0.168 0.168 0.124 0.168 0.138 0.131
WL-4.4 0.137 0.216 0.164 0.148 0.123 0.164 0.134 0.127
WL-5.1 0.194 0.332 0.256 0.241 0.144 0.175 0.159 0.159
WL-5.2 0.144 0.162 0.150 0.145 0.113 0.133 0.122 0.122
WL-5.3 0.149 0.228 0.182 0.165 0.160 0.185 0.168 0.163
WL-5.4 0.153 0.307 0.202 0.190 0.134 0.273 0.203 0.207

It becomes clear from Table 4 that the damping values are scattered, even for wall
elements with the same assembly. The influences are the loading protocol, sheathing
material, and anchoring and support conditions (see Table 1). However, if one looks
at the progression of equivalent viscous damping with the increasing deformation,
the damping value remains approximately constant until the drift limit is reached (see
Figure 7). This applies especially to the damping values of the third cycle. The drift
limit was defined as 0.02⋅h with a wall height of 2.50 m (see Table 2).

Nevertheless, in some cases, outliers also exist for specific amplitudes and certain
cycles. For that reason, the median value seems to be more descriptive than the
mean value. The median value mostly coincides with the mean value.

260
 INTER / 48 - 15 - 01

(a) (b)
Figure 7. Equivalent viscous damping over displacement amplitude for light-frame wall elements:
(a) WL-1.3, (b) WL-4.1.

4.3 Proposal
As a result of the findings in the previous section, constant values for equivalent
viscous damping are regarded as suitable for transferring elastic spectra into damped
spectra for light-frame timber structures. In order to derive constant damping values,
all equivalent viscous damping values were summarised in one statistical population,
separately for the first and the third cycle. The values derived are reported in Table 5
and the summary of the equivalent viscous damping values is shown in Figure 8.
Table 5. Summary of equivalent viscous damping for light-frame timber structures.
First Cycle Third Cycle
Min Max Mean Median Min Max Mean Median
0.106 0.332 0.182 0.172 0.094 0.297 0.167 0.164

The histogram in Figure 8 emphasises a strong concentration of the equivalent

viscous damping values in the range of the constant values derived.
The use of constant values for equivalent damping will simplify the capacity spectrum
method significantly, since a relationship between equivalent damping and
displacement is no longer required. The design procedure will be demonstrated in
section 5.
Based on Table 5 and Figure 8, a constant value between 0.16 (third cycle) and 0.18
(first cycle) for equivalent viscous damping might be suitable. However, the more
conservative value 0.16 is proposed for light-frame constructions, since recurrent
loading can lead to a reduction of equivalent viscous damping. Equivalent viscous
damping can be affected by the assembly (e.g. sheathing material, sheathing
thickness). If the assembly of the light-frame element is not clear, 10% equivalent
viscous damping can be taken into account in any case.
A constant equivalent damping value of 0.21 can be derived for light-frame structures
from the equivalent viscous damping values (0.16) proposed and 5% viscous
damping, as considered in code spectra.

261
INTER / 48 - 15 - 01

Figure 8. Summary of equivalent viscous damping for light-frame wall elements and overall
distribution of equivalent viscous damping for cycle 1 and cycle 3.

5 Comparative Study and Validation

The application of performance-based design by means of the capacity spectrum
method in combination with the values for equivalent damping proposed is
illustrated in this section. The N2 method is used for comparison. Furthermore, the
approaches for equivalent damping according to ATC 40 (1996) will be considered.
Exemplarily, two load displacement curves obtained by cyclic testing are used to
carry out performance-based design. The backbone curves of WL-3.1 and WL-3.3 (see
Table 1) were chosen here, since these cyclic tests are comparable to the dynamic
tests WL-dyn-3 and WL-dyn-1. The results of the calculation will be compared with
the results of the dynamic test.
The load displacement curve – also called the capacity curve – has to be transformed
into the Sa-Sd format for the sake of consistency. The acceleration Sa is calculated with
equation (7):
F
Sa  (7)
m
The cyclic tests were performed with a vertical of 10 kN/m (see Table 1). That leads to
a mass of 2.5 t for a wall length of 2.5 m. It must be noted that the light-frame wall
elements tested are understood as a single degree-of-freedom system. For that
reason, the relative displacements of the wall element are the same for the spectral
displacements Sd.
The spectrum according to Eurocode 8/NA ground type A-R was chosen as the elastic
response spectrum, which was scaled by 1.2 to match the spectrum of the time

262
 INTER / 48 - 15 - 01

history applied (see Figure 9). The demand spectra were gained from the elastic
response spectrum. The value 0.21 is used for the constant damping approach, as
proposed in section 4.3.
The elastic response spectrum was transferred with equation (2) to generate the
damped spectrum. Similarly, the demand spectrum, according to ATC 40 (1996), was
created. Therefore, the effective damping for several steps of spectral displacement
of the capacity spectrum has to be calculated following the procedure proposed in
ATC 40. Consequently, effective damping can be determined with equation (8), see
Figure 6b.

2 Sa ,y  Sd ,pi  Sd ,y  Sa ,pi
ξeff  ξ0  κ  ξeq with ξeq   (8)
π Sa ,pi  Sd ,pi
The modification factor κ was set to 0.33 in order to consider the hysteretic
behaviour for light-frame elements (pinched shape). The definition of ISO 21581
(2010) was used as initial stiffness. Equal energy under the nonlinear and the bilinear
capacity curve was considered.
The inelastic spectrum was created by following the N2 method proposed in annex B
of Eurocode 8 (2010). A bilinear idealisation of the nonlinear capacity curve is
required to determine the reduction factor Rµ, respectively, the ductility µ (see
section 2). The capacity curve of WL-3.1 or WL-3.3 (backbone) was taken here and
was idealised into a bilinear curve. In order to define the bilinear capacity curve,
firstly, the maximum displacement was chosen and then the yield displacement was
calculated by considering the energy equivalency between the nonlinear and bilinear
curve. Regarding the bilinear capacity curve, the reduction factor Rµ and the ductility
µ were determined, and the spectral values for the inelastic spectrum could be
calculated with equation (3).
The spectra and capacity curves obtained were drawn into the Sa-Sd diagram (see
Figure 9 and 10). The intersection between the demand spectrum and the capacity
curve defines the performance point.
Figures 9 and 10 depict that reasonable performance points could be found for the
two different wall configurations. The displacement demand ranges from 23 mm to
31 mm for WL 3.3 and from 27 mm to 41 mm for WL 3.1. In both cases, the
performance points using the constantly damped spectrum provide the lower values
and the performance points with the inelastic spectrum define the upper limit, while
the performance point with the spectrum according to ATC 40 (1996) lies in between
these two limits.
The hysteretic behaviour under realistic dynamic impact was used for a direct
evaluation of inelastic and damped spectra for both wall configurations. In both
cases, the Sa values and the related Sd value, which could be interpreted as one point
of the damped or inelastic spectra, confirm basically the level of the transformation.

263
INTER / 48 - 15 - 01

Figure 9. Seimic performance of dynamic test WL-dyn-1 compared with the results of performance-
based design from different methods for WL-3.3 (graphical solution).

Figure 10. Seimic performance of dynamic test WL-dyn-3 compared with the results of performance-
based design from different methods for WL-3.1 (graphical solution).

264
 INTER / 48 - 15 - 01

6 Conclusion
It was found that the values of equivalent viscous damping which are obtained from
cyclic testing of light-frame elements do not depend on the displacement amplitude,
since they remain fairly constant with increasing deformations. A constant value of
equivalent viscous damping for light-frame elements of 0.21 was proposed. The use
of constant values for equivalent damping to create damped spectra simplifies the
transformation significantly.
Different performance-based design methods were applied to light-frame elements
and the results were compared. It was confirmed that performance-based design
with constant equivalent damping can lead to suitable results. Different results,
especially in respect of the maximum relative displacement, were obtained with the
N2 method. The comparison of results from performance-based design with results
from dynamic tests exhibits a good agreement with the maximum acceleration, but
sometimes deviations in the maximum relative displacement occurred.
It must be noted that the performance-based design with constant equivalent
damping seems to underestimate the seismic performance slightly. A better
agreement with the performance from dynamic testing can be achieved if a value of
2 % for viscous damping – as proposed by Filiatrault et al. (2002) – is considered for
the constant damping approach (see section 4.3).
The N2 method is also possibly applicable to timber structures. An adaption of the N2
method is needed in order to consider the typical hysteretic for timber structures.

7 References
ATC-40 (1996): Seismic evaluation and retrofit of concrete buildings. Applied
Technology Council. California Seismic Safety Commission. Volume 1.
Chopra, AK (1995): Dynamics of structures (Vol. 3). New Jersey: Prentice Hall.
Chopra, AK & Goel, RK (1999): Capacity-demand-diagram methods for estimating
seismic deformation of inelastic structures: SDF systems. Civil and Environmental
Engineering, 53.
Eurocode 5 (2004): Design of timber structures - Part 1-1: General and rules for
buildings. CEN. (EN 1995-1-1).
Eurocode 8 (2010): Design of structures for earthquake resistance – Part 1: General
rules, seismic actions and rules for buildings. CEN. (DIN EN 1998-1).
Eurocode 8/NA (2011): National Annex – Nationally determined parameters –
Eurocode 8: Design of structures for earthquake resistance – Part 1: General rules,
Seismic actions and rules for buildings CEN. (DIN EN 1998-1/NA).

265
INTER / 48 - 15 - 01

Fajfar, P (1999): Capacity spectrum method based on inelastic demand spectra.

Earthquake Engineering and Structural Dynamics, 28, 979-993.
Fajfar, P (2000): A nonlinear analysis method for performance-based seismic design.
Earthquake Spectra, 16, 573-592.
Filiatrault, A & Folz, B (2002): Performance-based seismic design of wood framed
buildings. Journal of Structural Engineering, 128(1), 39-47.
Filiatrault, A, Isoda, H & Folz, B (2003): Hysteretic damping of wood framed buildings.
Engineering Structures, 25, 461-471.
Fragiacomo, M, Dujic, B & Sustersic, I (2011): Elastic and ductile design of multi-
storey crosslam massive wooden buildings under seismic actions, Engineering
Structures, 33, 3043-3053.
Freeman, SA (1998): The capacity spectrum method as a tool for seismic design
Proceedings of the 11th European Conference on Earthquake Engineering, pp. 6-
11.
Freeman, SA (2004): Review of the development of the capacity spectrum method.
ISET Journal of Earthquake Technology, 41, 1-13.
ISO 21581 (2010): Timber structures – Static and cyclic lateral load test method for
shear walls.
Kawai, N (1999): Application of capacity spectrum method to timber houses. CIB-W18
Graz, Austria.
Miranda, E & Bertero, VV (1994): Evaluation of strength reduction factors for
earthquake-resistant design. Earthquake Spectra, 10, 357-379.
Petersen, C (1996): Dynamik der Baukonstruktionen, Vieweg Verlagsgesellschaft.
Priestley, MJN & Kowalsky, MJ (2000): Direct displacement-based seismic design of
concrete buildings. Bulletin of the New Zealand National Society for Earthquake
Engineering, 33(4), 421-444.
Saiidi, M & Sozen, MA (1979): Simple and complex models for nonlinear seismic
response of reinforced concrete structures. University of Illinois Engineering
Experiment Station. College of Engineering. University of Illinois at Urbana-
Champaign.
Seim, W, Hummel, J & Vogt, T (2014): Earthquake design of timber structures –
Remarks on force-based design procedures for different wall systems. Engineering
Structures, 76, 124-137.
Seim, W & Hummel, J (2103): Deliverable 2D: CLT wall elements – monotonic and
cyclic testing. Technical Report, Department of Structural Engineering, Building
Rehabilitation and Timber Engineering, University of Kassel.

266
 INTER / 48 - 15 - 01

Seim, W, Hummel, J & Vogt, T (2013): Deliverable 2C: Anchoring units – monotonic
and cyclic testing. Technical Report, Department of Structural Engineering, Building
Rehabilitation and Timber Engineering, University of Kassel.
Seim, W & Vogt, T (2013): Deliverable 2B: Timber framed wall elements – monotonic
and cyclic testing. Technical Report, Department of Structural Engineering, Building
Rehabilitation and Timber Engineering, University of Kassel.
Seim, W & Schick, M (2013): Deliverable 2E: Moment resisting frames – monotonic
and cyclic testing. Technical Report, Department of Structural Engineering, Building
Rehabilitation and Timber Engineering, University of Kassel.
SIMPSON STRONG-TIE (2011): Qualitätsverbinder für Holzkonstruktionen –
Charakteristische Werte nach EC5 und DIN1052.
Vidic, T, Fajfar, P & Fischinger, M (1994): Consistent inelastic design spectra: strength
and displacement. Earthquake Engineering and Structural Dynamics, 23, 507-521.

267
INTER / 48 - 15 - 01

Discussion

The paper was presented by J Hummel

M Popovski received confirmation that light frame and not CLT was shown in slide 7;
also 2.8 mm diameter nails 65 mm in length and 75 mm spacing were used in the test.
M Popovski commented that µ=1.64 from the data seemed too low.
D Moroder and J Hummel discussed the use of elastic spectrum and equal displace-
ment approach. J Hummel also clarified issues related to rocking in the steel plate
and the provision of ball bearing guides.
M Yasumura commented that in Japan a similar approach is available via the BSL us-
ing equivalent single mass. Here, long wall would depend on the joint at the end of
the panel. Also in multi-story cases, different rotations of wall exist in each story. He
asked how one would handle these cases. J Hummel answered that long walls were
not considered yet. This could be considered in a model. Also multi-story cases need
hold down to confine rotation such that shear would govern. M Yasumura comment-
ed that one would need a load displacement relationship and one should model the
structure not just the wall. J Hummel agreed.
A Ceccotti stated that the CLT hysteresis loops would depend on different boundary
conditions and the vertical load and paper reference on the topic is available.
B Dujič asked if this method would be applicable to timber structures. J Hummel re-
sponded that it has the potential to, but they would still need to work on inelastic
spectrum for timber structures.
S Winter commented on editorial issues where details about the test specimens
should be added. He also questioned the test with gypsum fibre boards and what
would be the difference. J Hummel has the gypsum results where higher but similar
load characteristics were observed. Also fatigue failures of staples and nails were ob-
served.
WS Chang and J Hummel discussed about the flag shape hysteresis loops for CLT and
the real building would behaviour differently because of centring effect. Also shear
walls have relatively high equivalent viscous damping and the issue of building infor-
mation from a wall to the whole structure.
F Lam received clarification the PGA used was 1g and it was chosen to drive the walls
into the inelastic range. He commented that the spectrum of time histories looked
strange. I Smith discussed about missing wood based material with other material.
B Dujič commented about viscous damping issues. D Moroder and J Hummel dis-
cussed the spectrum, time history shifting and scaling procedures.

268
269
270
 INTER / 48 - 15 - 02

Simplified Wall Bracing Using Wood

Structural Panel Continuous Sheathing

Thomas D. Skaggs, APA – The Engineered Wood Association

Borjen Yeh, APA – The Engineered Wood Association
Edward L. Keith, APA – The Engineered Wood Association (Retired)

Keywords: International Residential Code, Prescriptive Lateral Design, Wall Bracing

1 Introduction
Before the introduction to the inaugural version of the International Residential Code
(IRC) in 2000, code conforming wall bracing was poorly understood with marginal
code compliance in some areas of the United States. The first version of the IRC pro-
vided a comprehensive methodology for providing prescriptive wall bracing, including
expanding the methods of bracing, and expanding the applicability of the prescriptive
provisions. With each triannual update of the IRC, the prescriptive bracing provisions
have continued to evolve into very complicated solutions to prescriptively brace walls
for wind and seismic loads.
The IRC provisions have raised the awareness to code officials; thus strict compliance
to the building code is now being required by jurisdictions throughout the United
States. However, the complexity of the residential code makes it difficult for design-
ers and builders to navigate the various options of the IRC. The IRC attempted to ad-
dress this issue by introducing a simplified wall bracing method in 2012, though this
simplified method was still fairly complicated.
Based on a multi-year research and testing initiative carried out by APA, the APA Sim-
plified Wall Bracing Method expands on the IRC Simplified Bracing Method to provide
an approach to bracing that is even more valuable to builders and building officials
by, in many cases, decreasing the amount of required wall bracing and the minimum
length of braced wall panels. In addition, the APA Simplified Wall Bracing Method in-
creases the applicability of the IRC simplified wall bracing provisions to as much as 4

271
INTER / 48 - 15 - 02

times as many house plans, including those with multiple window and door openings
on the front and rear elevations.
The Simplified Wall Bracing Method described in this paper is based on APA System
Report SR-102 (2015). This methodology provides building officials, builders and de-
signers with an approach and the supporting technical information to meet the re-
quirements of the 2015 IRC Wall Bracing (Section R602.10). The IRC Simplified Wall
Bracing has been modified to increase its applicability to a greater percentage of
home designs. To achieve broad applicability and acceptance, the system uses the
most common type of wall sheathing, wood structural panels, based on their superior
structural performance. To provide the user with the greatest possible architectural
latitude, SR-102 only covers continuously sheathed wood structural panel bracing
(IRC Method CS-WSP) with an increased sheathing thickness (called “Performance
Category” in product standards) and a closer nailing schedule on the first story of a
two-story structure. This approach increases the performance of the bracing panels
on the first story due to the additional restraint provided by the mass and stiffness of
the structure above, through strength from increased fastening and with the use of
thicker wood structural panel continuous sheathing. This enhanced performance on
the bottom story of multi-story structures leads to reduced length of required brac-
ing in these areas, allowing for the method to be used on homes with abundant win-
dow and door openings typically found on the front and back elevations. These de-
creases in the required bracing of multi-story structures are reflected in Table 3.
Design simplification and flexibility are achieved through the enhanced sheathing
thickness and nailing described in this paper. Intermittent wood structural panel
(Method WSP) and other bracing methods, except as discussed in Section 2.1 of this
paper, are outside the scope. Like the IRC simplified bracing method, the APA Simpli-
fied Wall Bracing Method may be used for houses located in areas of low to moder-
ate wind and seismicity. To increase the usability of the method, SR-102 includes ad-
ditional details to augment IRC simplified bracing provisions. Also included are refer-
ences to specific areas of the IRC and other publications when additional information
is required.
Buildings meeting the requirements of SR-102 meet all of the bracing requirements
of the 2015 Section R602.10 Wall Bracing with the enhancements discussed in Sec-
tion 2 of this paper.

2 Methodology
2.1 Applicability
Residential structures must meet all of the following conditions when using this
method:

272
 INTER / 48 - 15 - 02

1. The entire building should be continuously sheathed with wood structural

panels in accordance with the requirements specified in this section.
2. Other bracing provisions of the 2015 IRC Section R602.10, except as specified
herein, are outside the scope of this method.
3. The foundation or basement walls are concrete or masonry, or concrete slab;
and the structure above should be 3 stories or less. Permanent wood founda-
tions are beyond the scope of this methodology.
4. Floor cantilevers are not more than 0.61 m (24 inches) beyond the foundation
or bearing wall below.
5. Stud wall height is 3.0 m (10 feet) or less when using the minimum required
bracing lengths specified in Table 3 of this paper unless adjustments are made
for other wall heights up to 3.7 m (12 feet) in accordance with Footnote c to
Table 3 in this paper.
6. Roof eave-to-ridge height is 4.6 m (15 feet) or less.
7. Interior finish of exterior walls consist of minimum 12 mm (1/2-inch) gypsum
boards installed on the interior side fastened in accordance with IRC Table
R702.3.5. Interior gypsum finish is not required on continuously sheathed
wood structural panels adjacent to garage openings (Method CS-G) and con-
tinuously sheathed portal frame (Method CS-PF) bracing panels (see Section
2.5 of this paper).
8. Design wind speed is 58 m/s (130 mph, Ultimate Design Wind Speed in the
2015 IRC) or less and the Wind Exposure Category is either B or C.
9. Seismic Design Category is either A, B or C for detached one- and two-family
dwellings or Seismic Design Category A or B for townhouses.
10.Cripple walls, if present, is considered as the first story of the structure when
using this method unless they are designed in accordance with 2015 Section
R301.1.2. When the foundation has been engineered/designed to support all
of the loads from the structure above, the method described in this paper is
appropriate. Such foundation systems may include cripple walls, daylight and
pile foundations, and permanent-wood and insulated-concrete-form founda-
tions.
11.Horizontal joint blocking of the bracing panels may be omitted if the amount of
bracing on a given wall is 2 times or more than the minimum required amount
of bracing derived from Table 3 of this paper after adjustment by the relevant
footnotes.

273
INTER / 48 - 15 - 02

2.2 Circumscribed Rectangle

Traditional wall bracing following the IRC must consider many wall lines, wall line
spacing, off-sets in wall lines, lengths of each wall line, and many various multipliers
and bracing methods. Figure 1a demonstrates the plan view of a traditional home in
North America with the initial layout for resolving the lateral loads. Figure 1b
demonstrates the methodology described in this paper using a circumscribed rectan-
gle. The rectangle surrounds all enclosed offsets and projections, such as sunrooms
and attached garages, unless an attached garage or portion of the building is to be
designed as a separate structure in accordance with IRC Section R301.1.3 or a sepa-
rate element. SR-102 shows a detailed design example for designing a home in parts
of circumscribed rectangles. Open structures, such as attached carports and decks,
may be excluded from the rectangle. The rectangle should have no side longer than
18 m (60 feet) and the ratio between the long side and the short side should not ex-
ceed 3:1.

(a) (b)
Figure 1. a.) Traditional wall bracing, b.) simplifed wall bracing.

2.3 Wood Structural Panel Materials

The wood structural panel sheathing is either Rated Sheathing or siding with a mini-
mum 11 mm (7/16 Performance Category), meeting the requirements of DOC PS1 or
PS2.

2.4 Wood Structural Panel Attachment

The wood structural panel sheathing is attached to framing in accordance with the
following requirements:
1. The wood structural panels should be installed with minimum 8d common
nails, 3.3 x 65 mm (0.131 x 2-1/2 inches), spaced at 100 mm (4 inches) on cen-
ter at panel edges and at 300 mm (12 inches) on center over intermediate

274
 INTER / 48 - 15 - 02

supports. For single-story or the top story of two- or three-story buildings, the
panels may be installed with 8d common nails spaced at 150 mm (6 inches) on
center at panel edges and 300 mm (12 inches) at intermediate supports.
2. The wood structural panels are applied continuously over all areas of the exte-
rior walls except windows and doors, and including gable ends; and may be in-
stalled either vertically or horizontally.
3. All horizontal panel joints should occur over and be nailed to common framing
or blocking with an appropriate panel edge-nailing schedule in accordance with
IRC Section R602.10.10.
4. Each end of a continuously sheathed braced wall line should have a 0.61 m
(24-inch) return corner as defined in IRC Section R602.10.7 or a 3.6-kN (800-
lbf) hold-down attached to the end stud of the braced wall panel closest to the
corner.
• If a continuously sheathed braced wall line contains an opening greater
than 6 m (20 feet), each end of each resulting braced wall line seg-
ment/section should have one of the conditions described above.
• If a continuously sheathed braced wall line contains two or more offset
braced wall line segments/sections as permitted in Section R602.10.1.2,
each end of each braced wall line segment/section should have one of
the conditions described above.
5. Gypsum wallboard is installed on the opposite side of wall bracing panels. Gyp-
sum wallboard is a minimum of 12 mm (1/2-inch) thick and is fastened with
nails or screws in accordance with IRC Table R702.3.5. Exception: Gypsum
wallboard may be omitted if the amount of bracing on a given wall is equal to
or greater than 1.4 times the minimum required amount of bracing derived
from Table 3 of this paper after adjustment by the relevant footnotes.

2.5 “Qualified” Bracing Panel

A single “qualified” bracing panel consist of a full-height portion of an exterior wall
continuously sheathed with wood structural panels with a minimum length as shown
in Tables 1 and 2 of this paper. The bracing panel should have no openings, except
that small drilled holes in the wall sheathing and not penetrating the wall framing up
to 38 mm (1-1/2 inches) for the passage of wiring and utilities. When using narrow
wall bracing methods, CS-G and CS-PF, the minimum permissible lengths and con-
tributing lengths for computing available bracing is shown in Table 1 of this paper.
When using Method CS-WSP, Table 1 provides the minimum permissible lengths and
contributing lengths based on both the wall height and the adjacent clear opening
height. If an 2.4- or 2.7- m (8-, or 9-foot) tall wall line is present, Method CS-WSP
braced wall segments less than the Table 1 minimum length may be used but with a

275
INTER / 48 - 15 - 02

corresponding reduction in contributing lengths for computing available bracing in

accordance with Table 2 of this paper.

Table 1. Minimum length of braced wall lines (excerpt from 2015 IRC Table R602.10.5, modified in
accordance with R602.12.3).

Adjacent clear Minimum length (m)

opening height Wall height Contributing length (m)
Method (m) 2.4 m 2.7 m 3.1 m
CS-G -- 0.61 0.69 0.76 actual length(a)
CS-PF(c) -- 0.41(b) 0.46(b) 0.51(b) 1.5 x actual length(a)
≤ 1.63 0.61 0.69 0.76
1.73 0.66 0.69 0.76
1.83 0.69 0.69 0.76
1.93 0.76 0.74 0.76
2.03 0.81 0.76 0.76
CS-WSP 2.13 0.89 0.81 0.81 Actual length(a)
2.24 0.91 0.89 0.84
2.34 0.91 0.91 0.89
2.44 0.91 0.91 0.91
2.54 – 2.74 -- 0.91 0.91
2.84 – 3.05 -- -- 0.91
(39.4 in. = 1 m)
(a) Use the actual length when it is greater than or equal to the minimum length
(b) The wall height for CS-PF is based on the height of the portal frame, as documented in Keith (2014). The
height of the portal frame is measured from the bottom plate to the top of the portal frame header.
(c) See IRC Figure R602.10.6.4.

2.5.1 Partial Credit for CS-WSP Panels

CS-WSP panels in 2.4- or 2.7-m (8- or 9-foot) tall walls between 0.51 and 0.61 m (20
and 24 inches) in length that do not meet the minimum length requirements of Table
1 may be used as bracing units at a full or reduced contributing length (depending on
the adjacent opening height), as shown in Table 2 of this paper based on the latest
APA research results, as documented in Keith (2012a and 2012b).

276
 INTER / 48 - 15 - 02

Table 2. Partial credit for CS-WSP less than full length with 2.4- or 2.7-m (8- or 9-foot) tall walls(a).
Length of full height Adjacent to a clear
Wall Height Method CS-WSP opening height Contributing length of
(m) Panel (m) (m) or less braced wall panels (m)
≤ 1.55 0.61
1.63 0.56
1.73 0.51
0.61
1.83 0.46
1.93 0.41
2.03 0.36
2.4 or 2.7
≤ 1.55 0.51
1.63 0.46
1.73 0.41
0.51
1.83 0.38
1.93 0.33
2.03 0.28
(39.4 in. = 1 m)
(a) Linear interpolation may be used

2.6 Computing “Qualified” Wall Bracing Length

Within an exterior wall, only those full-height wall panels with a length greater than
or equal to the lengths specified in Tables 1 and 2 of this paper is deemed to contrib-
ute to resisting lateral load, and counted toward the required bracing length. The to-
tal bracing length contributing to the side of a rectangle is equal to the sum of the
contributing lengths of each “qualified” wall panel. Any length of a “qualified” bracing
panel over the minimum bracing length required in Table 1 of this paper may be used
toward the total bracing length required for that side of the rectangle. Thus, if the
minimum requirement for a specific method is 0.61 m (24 inches) in accordance with
Table 1 of this paper and two such panels with lengths of 0.66 and 0.86 m (26 and 34
inches) are present, (0.66 + 0.86 =) 1.5m (60 inches) of bracing are present and
should be used in determining the total bracing length for that wall.
For Methods CS-G and CS-PF, the bracing length on either side of the opening is con-
sidered a “qualified” bracing panel and contributes to bracing lengths for meeting the
minimum length requirements of Table 1 of this paper. Examples of utilizing this sim-
plified approach are demonstrated in the Appendix to SR-102 (APA, 2015).

2.7 Length of Bracing Required

Determining the minimum bracing length required is relatively straightforward:
1. Circumscribe the building with a rectangle. The rectangle encloses the maxi-
mum building length and width dimensions as described in Section 2.2.

277
INTER / 48 - 15 - 02

2. Ensure that the long side of the rectangle is not greater than 3 times the short
side of the rectangle or greater than 18 m (60 feet). If it is greater, consider us-
ing the multiple rectangle method covered in Appendix A of SR-102. The alter-
natives are to:
• Use the “legacy” bracing provisions of IRC Section R602.10,
• Use the multiple rectangle method in conjunction with the APA Simpli-
fied Wall Bracing Method (see Appendix A or SR-102), or
• Have the structure designed in accordance with IRC Section R301.1.3
and the International Building Code (IBC).
3. With the dimensions of this circumscribed rectangle, use Table 3 of this paper
to determine the bracing length that is required on each rectangle side per-
pendicular to the side used to enter the table. Note that interpolation may be
used. Either value, the rounded or interpolated value, is multiplied by a wall
height adjustment factor in accordance with Footnotes (c) and (d) to Table 3 of
this paper, as applicable.
4. Parallel wall lines within 1.2 m (4-feet) of each other are considered the same
wall line when following the Distribution Rules of Section 2.8 of this paper.

2.8 “Distribution Rules” for Bracing Panels

Once the required minimum bracing length has been determined for each side of the
circumscribed rectangle using Table 3 of this paper, this bracing length is distributed
along the actual exterior walls of the structure. In distributing these bracing panels,
all of the following “Distribution Rules” should be met:
1. The first “qualified” bracing panel on each side of the rectangle begins within
3.7 m (12 feet) of the wall corner. The 3.7 m is measured between the wall
corner and closest edge of the first full-height “qualified” bracing panel.
2. The distance between the closest edges of adjacent full-height “qualified”
bracing panels is 6.1 m (20 feet) or less.
3. Any exterior wall with a length of 2.4 m (8 feet) or greater, when used as brac-
ing, should have a minimum of 0.91 m (3 feet) of bracing.
In some cases, a greater bracing length is required to meet the Distribution Rules
than is required by Table 3. In this case, the greater bracing length required by the
Distribution Rules governs. In any cases, the bracing length required by Table 3 or the
Distribution Rules, whichever is greater, should be met.
If the upper and lower stories share common exterior wall lines and the amount of
bracing on the second floor equals or exceeds the amount of bracing located on the
story immediately below, and the distribution rules of Section 2.8 for all such stories
are met, only the bracing in the bottom story must be checked. If the bottom story
checks out, the upper stories will be acceptable as well.

278
 INTER / 48 - 15 - 02

Table 3. Minimum required bracing length on each side of the circumscribed rectangle for wind ex-
posure B (a)(b)(c)(d). (Table provides required bracing amount for walls perpendicular to the maximum
bracing length used to enter the table.)

Eave-to Minimum required bracing

Ridge length on each long/short side (m)
Wind Height Length of short/long side (m)
Speed Story Level (m) 3.0 6.1 9.1 12.2 15.2 18.3
0.61 1.07 1.52 1.83 2.29 2.74

3.0 0.88 1.65 2.26 3.02 3.66 4.27

51 m/s
(115 mph) 1.25 2.41 3.41 4.42 5.43 6.40
ultimate
based on 0.79 1.40 1.98 2.38 2.99 3.57
2015 IRC
4.6 1.01 1.89 2.59 3.47 4.21 4.91

1.37 2.65 3.75 4.88 5.97 7.04

0.76 1.22 1.83 2.29 2.90 3.35

3.0 1.13 2.01 2.77 3.66 4.54 5.33

58 m/s
(130 mph) 1.52 2.90 4.15 5.43 6.68 7.80
ultimate
based on 1.01 1.58 2.38 2.99 3.78 4.36
2015 IRC
4.6 1.31 2.32 3.20 4.21 5.21 6.13

1.68 3.20 4.57 5.97 7.35 8.60

(39.4 in. = 1 m)
(a) Based on IRC Table R602.10.3(1) and modified in accordance with Keith (2011).
(b) Interpolation may be used.
(c) The Wall Height Adjustment Factor, as shown below, is used to multiply the minimum bracing lengths
listed in the table above to accommodate wall heights from 2.4 to 3.7 m (8 to 12) feet based on IRC Table
R602.10.3(2). Interpolation may be used.
Wall Height (m) Wall Height Adjustment factor
2.4 0.90
2.7 0.95
Any Story 3.0 1.00
3.4 1.05
3.7 1.10

(d) For Wind Exposure Category C, multiply length required from table above by 1.2 for single-story build-
ings, 1.3 for two-story buildings and 1.4 for three-story structures.
(e) The first story of two stories and the first and second of three stories should be continuously sheathed
with wood structural panels attached with 8d common nails, 3.3 x 65 mm, (0.131 x 2-1/2 inches) spaced 100
mm (4 inches) on center around the panel perimeter and at 300 mm (12 inches) on center over intermediate
supports.

279
INTER / 48 - 15 - 02

3 Lateral Support
For bracing panels in exterior walls located along eaves where the distance between
the top of the top plates to the underside of the roof sheathing is 0.23 m (9-1/4 inch-
es) or less, blocking between the rafters or trusses is not required. When the distance
between the top of the top plates to the underside of the roof sheathing above
braced walls is greater than 0.24 m (9-1/4 inches) and less than 0.39 m (15-1/4 inch-
es), attachment should be in accordance with IRC Section R602.10.8.2, item 1. These
details are not duplicated here because they vary slightly between different editions
of the IRC and because the 0.39 m (15-1/4 inches) limitation is not commonly ex-
ceeded.
If the vertical distance between the underside of the roof sheathing and the top of
the top plate is greater than 0.39 m (15-1/4 inches), or if the user wants to use the
wall sheathing to block raised-heel trusses to meet the wind uplift and lateral load
requirements of IRC sections R602.3.5 and R602.10.2.1, see APA System Report SR-
103, for more information.

4 Limitations
Recommendations provided in this paper are subject to the following conditions:
1. The exterior walls of the structure is continuously sheathed with a minimum 11
mm (7/16 Performance Category) wood structural panel sheathing or siding
meeting the requirements of DOC PS1 or PS2 and is attached to framing with
8d common nails, 3.3 x 65 mm, (0.131 x 2-1/2 inches) at 100 mm (4 inches) on
center around the panel perimeter and at 300 mm (12 inches) on center over
intermediate supports. For exterior walls in single story structures or in the top
story of multi-story structures the 8d common nails are spaced at 150 mm (6
inches) on center around the panel perimeter and at 300 mm (12 inches) on
center over intermediate supports.
2. The APA Simplified Wall Bracing Method is applicable to buildings of no more
than three stories, subject to the applicability listed in Section 2.1 of this paper.
3. When placed over masonry or concrete stem walls, wall bracing panels used in
the APA Simplified Wall Bracing Method must meet the requirements of IRC
Section R602.10.9.
4. While the APA Simplified Wall Bracing Method is not part of the code, it is
based on the code and other modifications permitted by IRC Sections R301.1.3
Engineering Design. Further modifications to the APA Simplified Wall Bracing
Method by the user of this paper are beyond the scope of this paper.
5. The basis for this paper, APA System Report SR-102 is subject to periodic re-
view. The latest copy of SR-102, in imperial units, is available for download at
www.apawood.org/resource-library.

280
 INTER / 48 - 15 - 02

5 References
APA. 2015. APA simplified wall bracing method using wood structural panel continu-
ous sheathing, APA System Report – SR-102C, APA – The Engineered Wood Associ-
ation, Tacoma, WA
APA. 2014. Use of wood structural panels for energy-hell trusses, APA System Report
– SR-103A, APA – The Engineered Wood Association, Tacoma, WA
AWC. 2015. Wood frame construction manual for one and two family dwellings
(WFCM), American Wood Council, Leesburg, VA.
ICC. 2015a. International building code (IBC), International Code Council, Falls
Church, VA.
ICC. 2015b. International residential code (IRC), International Code Council, Falls
Church, VA.
Keith, E. L. 2011. Wall bracing capacity enhanced due to partial restraint, APA Report
T2011L-33, APA – The Engineered Wood Association, Tacoma, WA
Keith, E. L. 2012a. Narrow wall bracing. APA Report T2012L-16, APA – The Engineered
Wood Association, Tacoma, WA
Keith, E. L. 2012b. Narrow wall bracing – eight-foot tall walls, APA Report T2012L-16,
APA – The Engineered Wood Association, Tacoma, WA
Keith, E. L. 2014. Portal frame aspect ratio, APA Report T2014L-39, APA – The Engi-
neered Wood Association, Tacoma, WA

281
INTER / 48 - 15 - 02

Discussion

The paper was presented by T Skaggs

I Smith received confirmation that there are plan limitations for diaphragm and wall
to match each other.
G Schickhofer asked if it would be possible to introduce CLT with this approach. T
Skaggs responded yes and that it could be possible although the market demand in N
America for CLT in residential construction is low.
A Salenikovich asked if calculation tools would be available. T Skaggs said yes but the
calculations are so simple that the tools would not really be needed.
W Seim and T Skaggs discussed how to deal with symmetrical and asymmetrical cases
where location of the bracing elements would have to be within certain distance of
the centre of the building plan.
F Lam and T Skaggs discussed consideration of CLT for tornado resistance structures
although tornado forces are not considered in N American codes. Here, CLT for safe
room tornado has been considered.

282
283
284
 INTER / 48 - 15 - 03

Structural characterization of multi-storey

CLT buildings braced with cores and addi-
tional shear walls

Andrea Polastri, Research Associate, Trees and Timber Institute - National Research
Council of Italy (CNR IVALSA), Via Biasi 75, 38010 San Michele all’Adige (TN), Italy
Luca Pozza, Ph.D., Department of Civil, Environmental and Architectural Engineering,
University of Padova, via Marzolo, 9, 35131 Padova, Italy
Cristiano Loss, Research Fellow, Department of Civil, Environmental and Mechanical
Engineering, University of Trento, via Mesiano 77, I-38123 Trento
Ian Smith, Professor Emeritus, Faculty of Forestry & Environmental Management,
University of New Brunswick, Canada

Keywords: CLT structures, core structures, seismic design, shear walls

1 Introduction
In last twenty years the CLT panels have become widely employed to build multi-
storey residential and mercantile buildings. These buildings are often characterised
by the presence of many internal and perimeter shear walls. Such structures have
been widely studied through experimental and numerical simulation methods. The
most comprehensive experimental investigation to date on seismic behaviour of CLT
buildings was carried out by CNR–IVALSA, Italy, under the SOFIE Project (Ceccotti
2008, Ceccotti et al. 2013). Other important investigations have been conducted at
the University of Trento, Italy (Tomasi and Smith 2015). European seismic perfor-
mance related tests have also been conducted at the University of Ljubljana, Slovenia
where the behaviour of 2-D CLT shear walls with various load and boundary condi-
tions were assessed (Dujic et al. 2005). FPInnovations in Canada has undertaken tests
to determine the structural properties and seismic resistance of CLT shear walls and
small-scale 3-D structures (Popovski et al. 2014). Those and other studies have ena-
bled characterisation failure mechanisms in large shear wall systems (Pozza et al.
2013). Multi-storey building superstructures in which beam-and-column frameworks
resist effects of gravity loads and cross-braced or core substructures and exterior CLT

285
INTER / 48 - 15 - 03

shear walls resist effects of lateral forces from earthquakes or wind have been found
structurally effective, and fail in predictable stable ways if overloaded (Smith et al.
2009). Advantages of such systems can include creation of large open interior spaces,
high structural efficiency, and material economies.
Recently innovative connection solutions that create discrete panel-to-panel, or pan-
el-to-other material joints have been developed in Italy (Polastri and Angeli 2014).
The method results in point-to-point mechanical connections that only connect cor-
ners of individual CLT panels in ways that fulfil hold-down and lateral shear resistance
functions (Gavric et al. 2013). This has the advantage of making the load paths within
superstructures unambiguous. Different connectors have also been tested to find the
best ways to make point-to-point connections between CLT panels and steel struc-
tures (Loss et al. 2014).
During recent years connector designs had evolved considerably making them suita-
ble for much large systems that place high capacity demands on connections, with
emphasis on requirements for high seismicity areas (Polastri et al. 2014). During such
development attention was paid to avoiding the possibility of brittle behaviour of
joints to CLT panels having many nails.
Structural performance issues not fully studied are those related to using CLT building
cores as replacements for one constructed from reinforced concrete or masonry.
Pertinent issues relate to vertical continuity between storeys, connections between
building core elements and elevated floors, and building core-to-foundation connec-
tions.

2 Mechanical characterization of the connectors

2.1 Experimental studies

The mechanical behaviour of connection systems for CLT structures that employ thin
metal elements fastened to panels with nails or other slender metal fasteners is well
known, as demonstrated by numerous scientific papers (Ceccotti et al. 2008, Pozza et
al. 2013). The behaviour of such connectors is determined largely by the elastoplastic
response of the fasteners, and to a lesser extent by the response of steel elements.
Stiffness and capacity values implemented into the numerical models described in
Section 4 were calculated directly from experimental data.
The first study was carried out at CNR-IVALSA (Gavric et al 2011), the second study at
the University of Trento (Tomasi and Smith 2015). In both cases tests were conduct-
ed according to the European standard EN 12512 (CEN 2006). The CEN 2006 protocol
provides a load history characterized by load cycles of increasing intensity and is in-
tended to apply to structures in seismic regions. As suggested by the standard, a pre-

286
 INTER / 48 - 15 - 03

liminary monotonic test was undertaken to define the magnitudes of cyclic load ex-
cursions, Figure 1.
120 80

100 60
80 40
60
20

Force (kN)
Force (kN)

40
0
20 -30 -20 -10 0 10 20 30
-20
0
-5 0 5 10 15 20 25 30 -40
-20
-40 -60

-60 -80
Displacemets (mm) Displacemets (mm)

Figure 1. Typical tests results: hold-down (left) and angle bracket (right)

Initial stiffness was calculated according to ‘method b’ specified by EN 12512 that

permits description of the mechanical behaviour of trends representing elastic phase
and post-elastic phase responses (Piazza et al. 2011). However, as this paper deals
with Linear Dynamics Analysis of superstructure systems only the parameters that
characterize the elastic properties of connections (ktest) and the maximum load at
failure (Fmax) are reported here, Table 1.

2.2 Analytical definition of stiffness according to Eurocode 5

The Finite Element (FE) model presented in Section 4 implements hold-down Rotho-
blaas WHT 620 (EOTA 2011) and angle brackets TITAN TTF200 (EOTA 2012) connect-
ors joined to CLT panels manufactured from class C24 wood boards using 32 4x60 or
30 4x60 Anker nails. The initial stiffness of connectors was calculated taking into ac-
count the stiffness of the steel-to-timber nailed joints in shear and hold-down con-
nections. Deformation of steel parts within the connections are very small, compared
deformation of nailed joints, and was therefore neglected. Characteristic load-
carrying capacities, Fv,Rk, and slip moduli, kser, were calculated based on Eurocode 5
(CEN 2014), Table 1.

Table 1. Experimental and Eurocode 5 derived connection properties

Connection type Elastic stiffness (kN/mm) Capacity (kN)
Test (ktest) EC5 (kser) Test (Fmax) EC5 (Fy,Rk)
TITAN TTF 200 8.2 23.1 70.1 35.5
WHT 620 12.1 24.8 100.1 85.2

287
INTER / 48 - 15 - 03

3 Estimation of T1 and design of connections

A crucial issue in the design of a CLT building under horizontal seismic action, is the
definition of the principal elastic vibration period (T1) of an entire superstructure
(CEN 2013). Such vibration period depends on the mass distribution and on the glob-
al stiffness of the buildings. In a CLT structure the global stiffness of the buildings is
highly sensitive to deformability of the connection elements (Pozza et al. 2013). Con-
sequently for a precise control of the vibration period of the building it is crucial to
define the stiffness of each connections used to assemble a superstructure. During
design engineers are required to solve iteratively to find the principal natural fre-
quency ( f1 = 1/T1) using a scheme such as that in Figure 2. Under the shown scheme:
(1) the stiffness of the connections influences the global stiffness of the building and
therefore its principal elastic period; (2) the external force induced by earthquake in
each connection is a function of the principal vibration period; (3) the load bearing
capacity of the connection must be compatible with the external force; (4) the
strength and the stiffness of the connection are linked through the effective number
of fasteners.

PRELIMINARY DESIGN VIA LINEAR STATIC ANALYSES

elastic period T1 and horizontal force distribution according to EC8 formulation

BUILDING DEFINITION OF 1st (ith) ATTEMPT CONNECTION DISTRIBUTION

PRINCIPAL
ELASTIC
PERIOD BUILDING MODELING ADOPTING THE CONNECTION ELASTIC
STIFFNESS (kser – EC5 formula) RELATED TO ith DISTRIBUTION
EARTHQUAKE
FORCE ON
CONNECTION NATURAL FREQUENCY ANALYSES TO DEFINE THE ACTUAL
ELEMENTS VIBRATION MODE OF THE STRUCTURE

CONNECTION MODAL ANALYSES TO DEFINE THE ACTUAL FORCE ACTING IN

S ELASTIC EACH CONNECTION ELEMENTS
STIFFNESS

CONNECTION CALCULATED FORCE CALCULATED FORCE

> (greater than) < (lower than)
LOAD
CONNECTION STRENGTH CONNECTION STRENGTH
BEARING
CAPACITY
UPDATE CONNECTION
DISTRIBUTION (i+1) STOP

Figure 2. Calculation process for design of connections

An efficient approach to design a CLT structure is to start from a preliminary defini-

tion of the external force induced by earthquake in each wall panel according to the
common equivalent static force linear static analysis approach (CEN 2013). This does
not involve the definition of T1 accounting for effects of connection stiffness. Once
static forces on each CLT wall panel are defined connection capacities can be de-
signed to be compatible with external static forces. This allows estimation of the con-

288
 INTER / 48 - 15 - 03

nection elastic stiffness (ktest or kser), and therefore realistic preliminary estimation of
T1. Then T1 can be in modal analyses to calculate the effective forces induced in con-
nections by earthquakes. Obtained connection forces may or may not be compatible
with the connection strength, and if not it is necessary to redesign connections. Af-
terward it is possible to perform a more iterative precise frequency analyses until so-
lutions, including connection designs, are convergent.

4 Numerical analysis of core tall buildings

The behaviour of multi-storey buildings braced with cores and CLT shear walls is ex-
amined using numerical modal response spectrum analyses, with connection proper-
ties calibrated based Eurocode 5 (CEN 2014) and experimental test discussed in Sec-
tion 2. Analyses followed the scheme in Figure 2 and are presented in terms of prin-
cipal elastic periods, base shear and up-lift forces, and inter-storey drift.

4.1 Case study buildings

The aim is to characterize behaviour of multi-storey CLT buildings braced with cores
and additional shear walls from the seismic design perspective based on effects of
varying design parameters. Varied design parameters are: number of storey (3-5-8),
lateral shear wall panels width, construction methodology, and regularity of connect-
ors as a function of the height within a superstructure, Table 2.

Table 2. Examined building configuration

Case study ID 3(5-8) A R 3(5-8) A I 3(5-8) B R 3(5-8) B I 3(5-8) C R

Graphical schema-
tization of building
cores (ex. 3-storey
case)

Panel assembly Joint free wall panels Joint free wall

Jointed wall panels
panels
Elevation regularity Regular Irregular Regular Irregular Regular

Construction
methodology Platform System -

289
INTER / 48 - 15 - 03

4.1.1 Geometric configurations

Examined case-study building superstructures have footprint dimensions of 17.1m
(direction X) by 15.5m (direction Y). Seismic Force Resistant Systems (SFRS) include a
building core that is 5.5m by 5.5m on plan, and partial perimeter shear walls con-
structed from CLT panels with a total base length of 6m, Figure 3. Storey height is 3m
in all cases, resulting in total superstructure heights of 9m, 15m and 24m respective-
ly. All CLT panels in the core walls have a thickness of 200mm. CLT panels in perime-
ter shear walls are 154mm thick, except for those in the lowest four storeys of the 8-
storey SFRS which are 170mm thick. Floor diaphragms are composed of 154mm CLT
panels in all cases.

Figure 3. SFRS wall configurations of case study buildings (left) and typical FE model (right)

4.1.2 Design method

The earthquake action for these case study buildings was calculated according to Eu-
rocode 8 (CEN 2013) and the associated Italian regulations (MIT 2008) using design
response spectra for building foundations resting on ground type C*, assuming the
PGA equal to 0.35g (the highest value for Italy) with a building importance factor of
= 0.85. [*Deep deposits of dense or medium dense sand, gravel or stiff clay with
thickness from several tens to many hundreds of meters]. The seismic action was cal-
culated starting from the elastic spectra and applying an initial q-reduction factor of 2
(CEN 2014). The coefficient kr was taken equal to 1.0 for regular configurations and
0.8 for non-regular configurations. Figure 4 shows the adopted design spectra and
other relevant design parameters. The figure shows T1 values determined by simpli-
fied formula and numerical frequency analyses methods for configurations A R 3-5- 8.
Connections were first designed using the force pattern obtained applying linear
elastic static analysis (CEN 2013) and the seismic action defined by taking T1 = T1_EC8.
Connection designs were then refined using the rotation and translation force
equilibrium approach described by Gavric et al. (2011) and Pozza and Scotta (2014)
and the iterative design process in Figure 2.

290
 INTER / 48 - 15 - 03

seismic area zone 1 0.5

Design Spectra - soil C - q=2
soil type C
Peak Ground Acceleration 0.35 g 3 A R - EC8
0.4
q0 (behaviour factor) 2
5 A R - EC8
kR (regular configuration) 1
0.3 8 A R - EC8

Sd [ag/g]
λ (importance factor) 0.85
3 A R - ANALYTICAL
3AR 5AR 8AR
0.2
H [m] 9 15 24 5 A R - ANALYTICAL
3/4
T1_EC8 = 0.05 H [s] 0.26 0.38 0.54 8 A R - ANALYTICAL
0.1
M [t] 276 482 800
3 A R - TEST
Sd-el (T1) [ag/g] * 0.82 0.82 0.78
q = k R q0 2 2 2 0.0 5 A R - TEST
Sd (T1) [ag/g] * 0.41 0.41 0.39 0.0 0.5 1.0 1.5 2.0
8 A R - TEST
*using elastic period evaluated via code (CEN 2013) T [s]

Figure 4. Input data for seismic analysy (left); design spectra and calculated periods (right)

4.1.3 Finite element (FE) models

Numerical models of the investigated building were realized using the finite-element
code Strand 7 (2005). The illustrative FE model in Figure 3 (right) uses linear elastic
shell elements to represent CLT panels and link elements to simulate the elastic stiff-
nesses of connectors. Beam elements with pinned end conditions were used to rep-
resent beam members interconnecting perimeter shear walls and shear walls in the
building core at the top of each storey. Horizontal slabs elements in floor and roof di-
aphragms were assumed to be rigid in-plane.
All the 15 building configurations have been modelled respecting the geometrical
features and connection stiffnesses in Tables 1 and 2.

4.2 Analysis results

Results presented here were obtained by modal response spectrum analyses of case
study buildings, Tables 3 to 6 and Figure 5. Those tables and figure show calculated
building principal elastic periods (T1), base shear forces (v) on angle brackets at the
Ultimate Limit State (ULS), uplift forces on base hold-down anchors at ULS (N), and
the maximum inter-storey drift values () at Damage Limitation State (DLS). The al-
ternative values given represent effects of taking connection stiffnesses (kconn) equal
to values derived from Eurocode 5 (kser) versus values derived from experiments
(ktest). Plus in the case of T1, the simple formula value T1_EC8 is included (CEN 2013). In-
ter-storey drift was calculated for each case study building using the Modal Response
Spectrum Analyses and the DLS design spectrum.

291
INTER / 48 - 15 - 03

Table 3. Predicted principal elastic periods (T1)

[s] 3AR 3AI 3BR 3BI 3CR 5AR 5AI 5BR 5BI 5CR 8AR 8AI 8BR 8BI 8CR

T1_EC8 0.26 0.26 0.26 0.26 0.26 0.38 0.38 0.38 0.38 0.38 0.54 0.54 0.54 0.54 0.54

T1_kconn. = kser 0.22 0.19 0.22 0.20 0.18 0.37 0.34 0.38 0.35 0.30 0.73 0.69 0.73 0.68 0.50

T1_kconn. = ktest 0.47 0.39 0.44 0.37 0.33 0.80 0.61 0.73 0.57 0.45 1.40 1.14 1.30 1.08 0.66

Table 4. Predicted base shear per unit of length (v)

[kN/m] 3AR 3AI 3BR 3BI 3CR 5AR 5AI 5BR 5BI 5CR 8AR 8AI 8BR 8BI 8CR

v_kconn. = kser 28.2 33.2 32.7 40.3 35.8 42.3 51.7 51.8 64.2 50.4 50.2 64.6 61.6 82.6 79.7

v_kconn. = ktest 28.1 34.9 33.4 36.4 29.2 29.7 49.3 38.4 58.4 50.5 34.0 47.4 42.7 76.3 61.6

Table 5. Predicted free edge base uplift forces (N)

[kN] 3AR 3AI 3BR 3BI 3CR 5AR 5AI 5BR 5BI 5CR 8AR 8AI 8BR 8BI 8CR

N_kconn. = kser 128.1 170.9 152.0 178.1 161.3 316.0 420.6 355.4 447.3 373.2 511.6 884.1 559.5 759.7 897.0

N_kconn. = ktest 138.5 185.7 149.4 232.1 138.5 246.2 403.8 260.1 455.7 366.9 334.4 527.3 339.5 516.1 678.8

Table 6. Predicted maximum inter-storey drift ()

[mm] 3AR 3AI 3BR 3BI 3CR 5AR 5AI 5BR 5BI 5CR 8AR 8AI 8BR 8BI 8CR

_kconn. = kser 2.9 1.8 2.5 1.9 1.4 5.6 4.4 5.1 4.3 2.6 9.9 8.8 8.8 7.8 4.7

_kconn. = ktest 12.3 7.8 10.9 6.8 3.8 15.6 9.7 14.0 8.8 5.1 21.3 13.9 18.5 12.4 5.6

1.50 1000
principal elastic periods [s]

holdown tensile force [kN]

1.25 800
1.00
600
0.75
400
0.50
0.25 200

0.00 0
3AR

3BR

5AR

5BR

8AR
8AI
8BR
3AI

3BI

5AI

8BI
3CR

5BI
5CR

8CR
3AR
3AI
3BR

5AR
5AI
5BR

8AR
8AI

8BI
3BI

5BI

8BR
3CR

5CR

8CR

T1_EC8 T1_kconn. = kser T1_kconn. = ktest N_kconn. = kser N_kconn. = ktest

Figure 5. Comparison of estimates of principal elastic periods and base free edge uplift forces
Observing Figure 5 it is apparent that can be seen that in most cases use of experi-
mental connection stiffnesses (kconn = ktest) leads to much larger T1 values than those
predicted based Eurocode 5 based estimates of connection stiffnesses (kconn = kser).
Similarly using the simple formula given by Eurocode 8 leads to low estimates of T1
values. Interestingly use of Eurocode 5 based estimates of kconn results is estimates of

292
 INTER / 48 - 15 - 03

T1 relatively close to simple formula values. However results suggest that neither of
those approaches are reliable ways of estimating principal natural periods of build-
ings having SFRS consisting of CLT cores and perimeter shear walls. Consequences of
discrepancies in kconn values from those found by testing varied in their effects on v, N
and  values, but in general results show that how connection stiffnesses are esti-
mated can alter design force and lateral drift estimates by substantial amounts. For
example, estimates of  were especially sensitive for eight-storey buildings.
It is important to underline that the adopted FE model is a limiting condition repre-
senting the maximum deformability of the system since the interaction between the
orthogonal walls and the out of plane stiffness, provided by the interposed floor
slabs, are neglected. On the other hand, FE models did not take into account nonlin-
ear deformability or large displacements effects.
It is possible to achieve large vertical reaction forces using a group of hold-down an-
chors working “in parallel”. To obtain the required uplift force resistances, that can
be greater than 600kN (configuration 8CR), it is necessary to use more than eight
hold-down anchors; however it is not demonstrated that the hold downs, disposed in
the aforementioned group configuration, are able to spread the total force between
the different reaction elements.

5 Connection solutions for innovative diaphragms

Diaphragms are an integral part of the building any SFRS and if they have high stiff-
ness and capacity any non-linear behaviour of the entire structure is primarily de-
fined by the response of the vertical bracing elements and complexity of the seismic
analysis is reduced. CLT multi-storey buildings are erected using panels with limited
dimensions because of production and transportation limitations (FPInnovations
2011). In floors and roofs, the different CLT panels are commonly joined together at
the edges using dowel-type mechanical fasteners like self-tapping screws, Figure 6
(left).

Figure 6. Typically floor-floor panel connections (left) and new X-RAD connector (right)

293
INTER / 48 - 15 - 03

The in-plane behaviour of the horizontal floors constructed from CLT panels and con-
nections is mainly affected by the response of panel-to-panel connections, with the
overall length-to-width ratio of the floor and aspect ratio of the CLT panels playing
primary role (Ashtari et al. 2014). More generally, the in-plane behaviour of CLT
floors depends on the building system, the location of the bracing walls and their
stiffnesses (Loss et al. 2015). For multi-storey CLT buildings with cores and additional
perimeter shear walls the construction system varies significantly compared to other
common CLT structures. In such cases the mechanism of deformation of the floors
under in-plane actions can increase the level of shear forces in the connections, due
to the distance between the supports and the number and placement of shear walls
around the perimeter of the building. Consequently standard connection techniques
for CLT elements can be inadequate in terms of capacity and special high perfor-
mance connections are then required, e.g. Figure 6 (right). Discussion here addresses
use of two innovative high-capacity connection technologies suitable for the purpose.
H q
Figure 7 (left) shows a slab made of CLT panels joined together by steel beams, with
the beam-to-panel connections designed and engineered 1
from the perspectives of
mechanical behaviours of the materials, installation tolerances,
2 2 feasibility of on-site
assembly, and cost. The load-slip curve (F-) for these connections measured by tests
is shown in Figure 7 (right) based on Loss et al. (2014). The operating principle of such
floors is similar to a truss system in which each pair of steel beams is braced by the
CLT panel and related to characteristics of the beam-to-panel connections.1
In Figure
7 it is presented a beam-to-panel connection solution obtained by the use of steel
2

plates welded to the beam and glued to the CLT panel.

2

Beam-to-panel joint
250
qH
(9.3;209.9)
(15;204.3)
200
1 (0.4;190.1)
Force F (kN)

2 2 150

100 k=961.4 kN/mm

1
Δ
50
1 t
2 0
0 10 20 30 40
Slip δ (mm)
2
Actual behaviour Numerical model
Beam-to-panel joint
Figure
2507. Innovative hybrid floor system (left) and tests results (right)
The second (9.3;209.9)
innovative connection method discussed here employs X-RAD connect-
(15;204.3)
200
ors, Figure 6 (right) that create discrete panel-to-panel joints. This method results in
(0.4;190.1)
Force F (kN)

point-to-point
150 mechanical connections in ways that fulfil hold-down and lateral shear

100 k=961.4 kN/mm

1 294
Δ
50
t
 INTER / 48 - 15 - 03

resistance functions (Polastri and Angeli 2014). As for shear walls, making point-to-
point interconnections lessens the chances that structural systems will fail in unin-
tended ways if overloaded. Figure 8 (left) shows use of X-RAD connectors in a floor
diaphragm with the result being ability to transfer very large forces and achieve very
high stiffness (Polastri and Angeli 2014). As shown in Figure 8 (right) the load capacity
was 171kN and the elastic stiffness 23.6kN/mm.

Figure 8. Innovative hybrid floor system (left); tests setup and test results (right)
Although results are not reported here it is to be mentioned that the authors are cur-
rently studying use on the described innovative connectors as ways of creating next
generation of CLT floor diaphragms. It is anticipated this will enable new applications
of CLT like construction of tall building having large footprints and braced by one or
more building cores and perimeter shear walls.

6 Discussion and Implications for Design Practice

As the case studies demonstrate, hold-down and shear connections at the bases of
CLT wall panels largely determine the behaviors of SFRS. It is therefore crucial to
properly represent the stiffnesses of connections during structural analyses from
which T1, peak dynamic forces flowing through wall and connection elements and in-
ter-storey drift are estimated.
For buildings having three to eight storeys T1 estimates, shear and uplift forces at ba-
ses of wall panels, and inter-storey drift can all be miscalculated by substantial mar-
gins, Tables 3 to 6. The remainder of this discussion assumes connection stiffnesses
derived from test data (kconn = ktest) are the most accurate and therefore correct esti-
mates of how connections embedded within SFRS actually behave. Although design-
ers could also estimate connection stiffnesses in many other ways, the authors be-
lieve it reasonable to suppose that estimating stiffnesses will often be based on in-
formation in Eurocode 5 and similar international codes (i.e. kconn = kEC5 in case stud-
ies).

295
INTER / 48 - 15 - 03

Case studies suggest T1 values being underestimated by up to 50 percent is a realistic

scenario unless designers use test data to estimate connection stiffnesses. Large er-
rors occurring during subsequent calculation of shear and hold-down forces and in-
ter-storey drift is also highly feasible. In capacities terms estimation of design forces
and sizing shear and hold-down connection the likely consequences of how connec-
tion stiffnesses are characterized are lesser, with connections being somewhat over-
designed being normal (i.e. based on assuming kconn = kEC5). However, errors in esti-
mation of inter-storey drift are likely to be much greater. As results in Table 6 show,
interstorey-drift was estimated to be up to four times larger assuming kconn = ktest than
assuming kconn = kEC5.
Based on findings here it suggested that design standards require testing of all con-
nections intended to be used in SFRS constructed partially or completely from CLT
wall panels. Furthermore, it is recommended that testing be required to characterize
both initial stiffness and capacities of such connections. Also highly desirable is that
engineers be given explicit guidance about what constitute appropriate structural
model representations of SFRS and appropriate Modal Response Spectrum Analyses.
The calculation process for design of connections in Figure 2 is believed suitable as
the basis for such guidance.

7 Conclusions
The primary finding of work reported here is that estimates of the principal vibration
periods of buildings with Seismic Force Resisting Systems containing CLT wall panels
can be grossly inaccurate if proper attention is not paid to accurate representation of
connection stiffnesses. Estimates of T1 obtained using the simple formula in Eurocode
8 can deviate greatly from values found using finite element models employing con-
nection stiffnesses test data. Similarly finite element model predictions of T1 in which
connection stiffnesses are estimated from information in Eurocode 5 can differ great-
ly from values obtained using connection test data. Inaccurate representation of
connection stiffnesses can also result in incorrect sizing of elements in SFRS, and
gross inaccurate in predictions of inter-storey drift. For these reasons it is important
that design standards give specific guidance related to determination of initial stiff-
nesses as well as capacities of connections. A suitable calculation process for design
of connections is required based on proposals here dealing with the structural anal-
yses of CLT shear wall systems.

8 Acknowledgement
The authors wish to thank Eng. Davide Trutalli, student Matteo Pasin of University of
Padua and Arch. Francesca Paoloni trainee at CNR-IVALSA.

296
 INTER / 48 - 15 - 03

9 References
Ashtari S., Haukaas T. and Lam F. (2014): In-plane stiffness of cross-laminated timber
floors, Proceedings 13th World Conference on Timber Engineering, Quebec City, Can-
ada.
Ceccotti A. (2008): New technologies for construction of medium-rise buildings in
seismic regions: the XLAM case. Structural Engineering International, 18(2):156-165.
Ceccotti A., Sandhaas C., Okabe M., Yasumura M., Minowa C. and Kawai, N. (2013):
SOFIE project – 3D shaking table test on a seven-storey full-scale cross-laminated
timber building. Earthquake EngStruct. Dyn.,42: 2003-2021.
Comité Européen de Normalisation (CEN) (2006): EN 12512 - Timber structures – test
methods – cyclic testing of joints made with mechanical fasteners, CEN, Brussels, Bel-
gium.
Comité Européen de Normalisation (CEN) (2013): Eurocode 8 - design of structures
for earthquake resistance, part 1: General rules, seismic actions and rules, CEN, Brus-
sels, Belgium.
Comité Européen de Normalisation (CEN) (2014): Eurocode 5 - design of timber
structures, Part 1-1, General - Common rules and rules for buildings, CEN, Brussels,
Belgium.
Dujic B., Aicher S. and Zarnic R. (2005): Investigation on in-plane loaded wooden ele-
ments – influence of loading on boundary conditions, Otto Graf Journal, Materi-
alprüfungsanstaltUniversität, Otto-Graf-Institut, Stuttgart, Germany, Vol. 16.
European Organisation for Technical Assessment (EOTA) (2011), Rotho Blaas WHT
hold-downs, European Technical Approval ETA-11/0086, Charlottenlund, Denmark.
European Organisation for Technical Assessment (EOTA) (2011), Rotho Blaas TITAN
Angle Brackets, European Technical Approval ETA-11/0496, Charlottenlund, Den-
mark.
Gavric I., Ceccotti A. and Fragiacomo M. (2011): Experimental cyclic tests on cross-
laminated timber panels and typical connections. In: Proceeding of ANIDIS, Bari, Italy.
Loss C., Piazza M. and Zandonini A. (2014): Experimental tests of cross-laminated
timber Floors to be used in Timber-Steel Hybrid Structures, Proceedings 13th World
Conference on Timber Engineering, Quebec City, Canada.
Loss C., Piazza M. and Zandonini R. (2015-in press): Innovative construction system
for sustainable buildings, IABSE Conference on Providing Solutions to Global Chal-
lenges, September 23-25, 2015, Geneva, Switzerland.
Ministero delle Infrastrutture e dei Trasporti (MIT) (2008): DM Infrastrutture 14 gen-
naio 2008 - Norme tecniche per le costruzioni - NTC (Italian national regulation for
construction), MIT, Rome, Italy.

297
INTER / 48 - 15 - 03

Piazza M., Polastri A., Tomasi R., (2011): Ductility of Joints in Timber Structures, Spe-
cial Issue in Timber Engineering, Proceedings of the Institution of Civil Engineers:
Structures and Buildings, 164 (2): 79-90.
Polastri A. and Angeli A. (2014): An innovative connection system for CL T structures:
experimental - numerical analysis, 13th World Conference on Timber Engineering
2014, WCTE 2014, Quebek City, Canada.
Polastri A., Pozza L., Trutalli D., Scotta R., Smith I., (2014): Structural characterization
of multistory buildings with CLT cores, Proceedings 13th World Conference on Timber
Engineering, Quebec City, Canada.
Popovski M., Pei S., van de Lindt J.W. and Karacabeyli E. (2014): Force modification
factors for CLT structures for NBCC, Materials and Joints in Timber Structures, RILEM
Book Series, 9:543-553, RILEM, Bagneux, France.
Pozza L. and Scotta R. (2014): Influence of wall assembly on q-factor of XLam build-
ings, Proceedings of the Institution of Civil Engineers Journal Structures and Buildings,
ISSN: 0965-0911 E-ISSN: 1751-7702 – DOI:10.1680/stbu.13.00081.
Pozza L., Trutalli D., Polastri A. and Ceccotti A., (2013): Seismic design of CLT Build-
ings: Definition of the suitable q-factor by numerical and experimental procedures,
Proceedings 2nd International conference on Structures and architecture, Guimarães,
Portugal: 90 – 97.
Smith T., Fragiacomo M., Pampanin S. and Buchanan A. H. (2009): Construction time
and cost for post-tensioned timber buildings. Proceedings of the Institution of Civil
Engineers: Construction Materials, 162:141-149.
Strand 7 (2005): Theoretical Manual – Theoretical background to the Strand 7 finite
element analysis system, http//www.strand7.com/html/docu_theoretical.htm.
FPInnovations (2011): CLT Handbook: cross-laminated timber, Eds. Gagnon S. and
Pirvu C., FPInnovation, Quebec City, Canada.
Tomasi R. and Smith I. (2015): Experimental characterization of monotonic and cyclic
loading responses of CLT panel-to-foundation and angle bracket connections, Journal
of Materials in Civil Engineering, 27(6), 04014189.

298
 INTER / 48 - 15 - 03

Discussion

The paper was presented by A Polastri

D Moroder commented that the proposed interactive effort for design included FEM
model for tall building but no practicing engineers are doing this. A Polastri agreed
but stated that the FEM model is important and needed for tall buildings. D Moroder
asked about higher mode effect. A Polastri said that only regular buildings were con-
sidered so there was no higher mode effect. D Moroder commented that higher mode
was observed even in three story buildings but perhaps not in those made of CLT.
M Follesa asked about the bonding connection to the foundation. A Polastri respond-
ed that they did not consider the difference of the behaviour of hold down between
wood elements and between wood wall and concrete foundation elements.
P Zarnani asked about the coupling issues between CLT panels at the corner of a
building. A Polastri agreed that this would be important; here, crossed screws would
be used to connect the corner panels to achieve high stiffness.
S Winter commented about the use of FEM for tall buildings. This type of study would
not only be needed for earthquake, but wind issues would also need consideration.
Also non-standardized hold downs should be considered. X Rad connectors have to be
considered beyond strength including issues with fire, sound, airtightness etc.
A Polastri responded that they are looking at different covers to address these issues.
In traditional steel plate solutions with the steel exposed, poor fire resistance would
also be expected.
A Salenikovich received clarification about comparison between balloon and platform
construction.
I Smith commented in Canada, a related study was done with taller buildings and
wind forces tended to govern the design above 8 stories.

299
300
 INTER / 48 - 15 - 04

Dissipative connections for squat or

scarcely jointed CLT buildings. Experi‐
mental tests and numerical validation

Roberto Scotta, Luca Pozza, Davide Trutalli, Luca Marchi, University of Padova, Italy
Ario Ceccotti, University IUAV of Venezia, Italy

Keywords: CLT; X‐Lam; dissipative connection; numerical model; experimental tests.

1 Introduction
The seismic behaviour of CLT system has been studied by numerous researchers from
various countries. Quasi‐static tests on shear‐wall systems and shake‐table tests on
full‐scale buildings showed that CLT structures are characterized by high strength and
stiffness with respect to seismic actions but might have low ductility and dissipative
capacity if not correctly designed (e.g., Ceccotti et al., 2013; Gavric et al., 2015;
Popovski & Gavric, 2015).
In particular, squat and scarcely jointed buildings realized with large horizontal panels
show limited dissipative capacity due to their prevailing sliding behaviour. However,
in the constructive practice these structures are largely preferred, because adoption
of large panels with few joints allows the reduction of time and cost for on‐site as‐
sembling, despite more shear‐resistant connectors are needed to resist to earth‐
quake forces because of the low dissipative capacity of such buildings. According to
such practice, Eurocode 8 (2013) suggests a quite prudential behaviour factor for
seismic design of CLT buildings: q=2. Recent studies demonstrated that the seismic
response of CLT system is affected by the building geometry (e.g., slenderness) and
by the number, type, arrangement and design of joints used to assemble the timber
panels. Slender or highly jointed buildings have much more dissipative and ductility
capacities than squat and scarcely jointed buildings. Such dependency was demon‐
strated by experimental tests on different shear walls (e.g., Gavric et al., 2015) and by
numerical simulations on different building systems (Pozza et al., 2013; Pozza & Scot‐
ta, 2014). According to these researchers more generous values of the behaviour q‐

301
INTER / 48 - 15 - 04

factor have been suggested, depending on building geometry and panel arrange‐
ment.
Seismic performance of CLT buildings is mainly related to the capability of connec‐
tions to do plastic work if loaded over yielding, whereas timber elements have limited
capability to deform inelastically (Smith et al., 2003). Nowadays in CLT buildings,
hold‐down and angle‐bracket connections, which were developed for platform‐frame
constructions, are typically employed. However, the dissipative capacity of light‐
frame buildings is mainly concentrated in nailing between frames and panels, while in
CLT buildings dissipative contribution is demanded exclusively to the connections be‐
tween wall panels, floor panels and foundation. Actually, hold‐downs and angle
brackets subjected to cyclic loading show a marked pinching behaviour due to wood‐
embedment phenomenon, which reduces the energy dissipation capability of con‐
nections. Moreover, they are optimized and certified for uniaxial loading (i.e., only
tension or only shear) while they may show undesired brittle behaviour if subjected
to combined loading and a rigorous capacity design is not applied (Gavric et al.,
2013).
The aim of this work is to demonstrate how the adoption of innovative dissipative
connections, specifically developed to be adopted in CLT buildings, can improve the
intrinsic ductility of stocky and scarcely jointed CLT buildings and the cyclic behaviour
assured by standard hold‐downs and angle brackets when used in CLT buildings.
Several of such innovative connections have already been proposed and tested
(Latour & Rizzano, 2015; Loo et al., 2014; Polastri et al., 2014; Sarti et al., 2015). In
this work first the design phase of a newly developed joint is proposed. The mechani‐
cal behaviour of this innovative connection element has been validated by means of
experimental tests and numerical models. Then numerical simulations of quasi‐static
cyclic‐loading tests were used to demonstrate the increased seismic performances of
shear walls using the proposed connection.

2 Design process
2.1 Design criteria
Currently adopted connectors for CLT panels are differentiated to prevent sliding
(angle brackets) or rocking movements (hold‐downs). Instead, the connection pro‐
posed in this work operates properly in both circumstances and can be subjected to
mixed axial and shear forces. It assures high ductility before failure and has a negligi‐
ble pinching behaviour to empathise dissipative capacity under cyclic loading. Criteria
fixed in the design of the connection were: displacement capacity not less than 30
mm according to EN 12512 (2006a); high ductility class according to Eurocode 8
(2013); resistance comparable to the traditional connectors (Gavric et al., 2011); op‐
timized shape to reduce production costs with minimum scarf production.

302
 INTER / 48 - 15 - 04

The core of the proposed connection is constituted by a “X” shaped plate obtained by
cutting of a flat steel sheet. Grade S275 steel was found to be the more appropriate
to achieve adequate ductile hysteretic behaviour without excessive decay of the
strength capacity. The “X” shape is optimized in order to prevent localized yielding of
material and then to assure diffuse yielding and emphasize ductility and energy dissi‐
pation capacity. The chosen shape assures low production costs and minimal scarf
production.

2.2 Parametric design assisted by numerical modelling

Once decided the tentative “X” shape, a parametric FE model was used to find the
optimal one fulfilling the design criteria. A 2D FE model of the “X” shaped plate using
shell elements was implemented into ANSYS Workbench (ANSYS Inc., 2012). An elas‐
to‐plastic constitutive law combined with a kinematic hardening model was adopted
for steel. Non‐linear geometrical analysis option was activated to account for possible
buckling phenomenon under high displacements.
To minimize failure risk due to oligocyclic fatigue, a limit to the maximum strain of
steel was imposed (Priestley et al., 1996). Accordingly maximum axial deformation of
the “X” plate was limited between ‐2% and +10% (small compression was accounted
for), while its allowable shear strain was set in the range ±6%.
For the shape optimization of the “X” plate, a parametric FE model with modifiable
geometries was employed. The geometrical parameters chosen as variable in the
model are evidenced with letters in Figure 1a. To enable a single‐cut production as
shown in Figure 1b other geometrical parameters are dependent on these.
Numerical simulations were conducted for pure tension and pure shear cyclic load‐
ing. A total of 70 different combinations of the variable parameters led to the defini‐
tion of the optimal final shape. Figure 2 shows the deformation at maximum imposed
displacements, in pure tension and pure shear loading. The darkest contour shows
plastic regions in which the yield strength (275 MPa) has been exceeded. Position and
extension of yielded zone vary with the loading type, but spreading of yielding is well
evident in both tests. Figure 3 shows the force‐displacement hysteresis cycles ob‐
tained with the optimal combination of parameters, compared with experimental re‐
sults.
2.3 Anchoring to the timber panel
In order to ensure the localization of deformations in the dissipative X plate, a proper
design of fixing details to the timber panel is required. Connection has to guarantee a
suitable over‐resistance with respect to the X plate, while remaining elastic and rigid
with limitation of wood embedment. An effective semi‐rigid solution could be the use
of dowel‐type fasteners coupled with punched metal plates fixed to the timber panel
(Blass et al., 2000). An alternative method to reduce the wood embedment is the use
of toothed‐plate connectors (e.g., Bulldog or Geka).

303
INTER / 48 - 15 - 04

(a) (b)
Figure 1. (a) Geometrical parameters of connectors; (b) demonstration of single‐cut production.

(a) (b)
Figure 2. Numerical output: deformation and plastic zones. (a) Tension test; (b) Shear test.

3 Experimental tests
Once concluded the design and optimization phases, experimental tests on proto‐
types have been conducted to obtain actual cyclic behaviour of the X brackets. Three
pairs of X brackets were tested in pure tension and as many in pure shear, for a total
of 12 specimens. Tests were performed at the Laboratory of Construction and Mate‐
rials of the University of Padova.
3.1 Test setup
Two different setups were designed for tension and shear tests. In order to evaluate
exclusively the behaviour of the X plate, suitable rigid steel frames were realized to
transmit load from actuators. The couple of X brackets were fixed externally on both
sides of the supporting frame without any buckling restraining elements.
In the tension test, the two lowest fixing points were connected to a 20‐mm thick
steel plate rigidly fixed to the portal. The two upper fixing points were connected to
another 20‐mm thick plate fixed to the hydraulic jack with an eyebolt mechanism.
The pure shear loading was obtained with an unbraced steel truss, in which the X

304
 INTER / 48 - 15 - 04

brackets operated as cross‐bracing element. 15‐mm thick steel plates were used for
the steel truss. PTFE sheets were interposed between each contact surface to mini‐
mize friction. The whole assembly was positioned in a rotated configuration, in order
to keep loading direction as close as possible to the virtual diagonal line. It was fixed
at the bottom to the portal with 2 M20 steel bolts and at the top to the hydraulic jack
with an M45 eyebolt.
3.2 Test procedure
Tests were performed according to quasi‐static cyclic‐loading protocol recommended
by EN 12512 (2006a). The cyclic procedure was stopped after reaching a relative dis‐
placement of 30 mm, then the specimens were loaded until their failure. Tests were
conducted under displacement control with a deformation rate of 0,02 mm/s.
3.3 Test results
Figure 3 illustrates the results of the experimental tests and the comparison with
those from numerical analyses. The good matching of the experimental curves with
the numerical models is clearly shown. It was possible thanks to the well predictable
behaviour of the steel used for the realization of the X brackets.
On 30‐mm cycle of the tension test (Figure 3a), the reloading path decreased gradu‐
ally due to instability phenomenon. For the same reason the maximum compression
force measured during unloading was lower than the tension one, but still maintain‐
ing a wide area and, consequently, an appropriate dissipative capacity. Lastly, the
numerical model tolerably underestimated force and stiffness for the unloading se‐
quences. For the shear test the progressive rotation of the steel frame was taken into
account to correctly evaluate shear force plotted in the force‐displacement curves in
Figure 3b. The experimental hysteresis loops are perfectly centred on the origin of
the axis thus demonstrating the suitability of the setup configuration. The experi‐
mental results are in good agreement with expectation, even if the numerical predic‐
tion slightly over estimates shear force at high displacements. In general, no noticea‐
ble strength degradation was observed in experimental tests and no cracks induced
by fatigue were observed.

(a) (b)
Figure 3. Experimental cycles in comparison with FEM results. (a) Tension tests; (b) Shear tests.

305
INTER / 48 - 15 - 04

Figure 4 shows the tested specimens at maximum deformation. The main evidence is
that X brackets are able to experience large plastic deformations before failure, in
both loading configurations. Instability phenomena of limited parts of the connection
occurred during both shear and axial tests without impairing significantly the me‐
chanical performance of the connections. Failure occurred for very large displace‐
ments due to stress concentration for a not correct realization of fillet “j” in Figure
1.a. Therefore, ductility of X brackets could be further improved by modifying this de‐
tail.

(a) (b)
Figure 4. Deformed specimens: (a) Axial test (b) Shear test. In dashed white line the original shape.
3.3.1 Analysis of test results
The performed cyclic‐loading tests allow to define the main mechanical parameters
for a proper characterization of the tested elements according to EN 12512 (2006a):
elastic and post‐elastic stiffness (kel, kpl), yielding point (Vy, Fy), failure condition (Vu,
Fu), and ductility ratio . From these data it is possible to classify the proposed con‐
nection into the appropriate ductility class, according to Eurocode 8 (2013), i.e., low
(L), medium (M) or high (H) ductility class. Various methods were proposed to com‐
pute these parameters (EN 12512; Muñoz et al., 2008). In this work, the envelope of
the hysteresis curves was fitted using the analytical formulation proposed by Foschi &
Bonac (1977). Then, method (a) of EN 12512 was chosen for both axial and shear
tests, in order to obtain the best linear fitting of the envelope curve. This method is
adequate for curves with two well‐defined linear branches and the yielding point is
defined by the intersection of these two lines. Moreover, also the equivalent Elastic–
Plastic Energy (EEEP) method (Foliente, 1996) was used to analyse the results of the
shear test, because of the elastic perfectly‐plastic behaviour shown. Tables 1 and 2
list the obtained values from the three tests referred to a single plate of the couple.
Average value, standard deviation (SD) and 5% characteristic value according to EN
1990 (2006b), were computed considering a sample of six elements.

306
 INTER / 48 - 15 - 04

Table 1. Analysis of tension test (EN 12512 method).

Test 1 Test 2 Test 3 Average SD
Fy [kN] 17.55 18.37 17.99 17.97 0.36
Vy [mm] 1.89 2.01 1.98 1.96 0.06
Fu [kN] 37.18 37.84 38.25 37.76 0.48
Vu [mm] 44.30 47.30 47.00 46.20 1.48
kel [kN/mm] 9.31 9.12 9.08 9.17 0.11
kpl [kN/mm] 0.46 0.43 0.45 0.45 0.01
(Vu) [‐] 23.49 23.49 23.72 23.57 0.12
Ductility Class H H H H ‐
Table 2. Analysis of shear test (EN 12512 method and EEEP method).
Test 1 Test 2 Test 3 Average SD
EN EEEP EN EEEP EN EEEP EN EEEP EN EEEP
Fy [kN] 26.71 27.41 29.41 28.88 28.14 27.83 28.09 28.04 1.21 0.68
Vy [mm] 2.38 2.60 4.00 4.45 4.02 4.53 3.46 3.86 0.84 0.98
Fu [kN] 29.00 27.41 29.70 28.88 28.40 27.83 29.03 28.04 0.58 0.68
Vu [mm] 50.00* 50.00* 58.00* 58.00* 80.00 80.00 ‐ ‐ ‐ ‐
kel [kN/mm] 11.24 10.55 7.36 6.49 7.00 6.14 8.53 7.73 2.10 2.19
kpl [kN/mm] 0.05 0.00 0.01 0.00 0.00 0.00 0.02 0.00 0.02 0.00
 (Vu=50mm) 21.04 19.24 12.51 11.24 12.44 11.03 15.33 13.84 4.42 4.19
Ductility Class H H H H H H H H ‐ ‐
* Tests 1 and 2 were stopped before the ultimate displacement.
Results show that the proposed connection is characterized by an high initial stiffness
and adequate resistance both for tension and shear loads. However, the most valua‐
ble result is the very high value of ductility obtained, coupled with almost null
strength degradation and pinching effect. The highest values of ductility were ob‐
tained for the axial configuration. However, ductility for the shear configuration was
computed assuming Vu as 50 mm, even if in test 3 failure of the specimen was
reached for a displacement equal to 80 mm, whereas tests 1 and 2 were stopped be‐
fore failure. If the ultimate displacements of 80 mm were assumed, ductility values in
shear tests would become higher and comparable with those from axial tests. Com‐
paring values of initial stiffness and of yielding and ultimate forces from the two con‐
figurations, it can be observed that the connection shows an almost uniform re‐
sponse when subjected to shear or axial loads.
Table 3 lists the 5th and the 95th percentile of the ultimate force (F0.05 and F0.95). Ac‐
cording to Fragiacomo et al. (2011) the ratio F0.95 / F0.05 is fundamental for the esti‐
mation of the overstrength factor to be used in a capacity design approach. Since on‐
ly a single type of steel has been used the obtained values should be amplified to ac‐
count for the typical variation of steel yielding stress. The limited values F0.95 / F0.05 as‐

307
INTER / 48 - 15 - 04

sured by the proposed connection, which are much lower than those shown by
standard ductile connections failing on the timber side, suggest that the adoption of
capacity design rules would become feasible if the proposed connection were em‐
ployed.
Table 3. Characteristic values and overstrength ratio.
Tension test Shear test (EN method) Shear test (EEEP method)
F0.05 F0.95 ov F0.05 F0.95 ov F0.05 F0.95 ov
Fy [kN] 17.18 18.76 1.09 25.46 30.71 1.21 26.56 29.52 1.11
Fu [kN] 36.70 38.81 1.06 27.76 30.30 1.09 26.56 29.52 1.11

3.3.2 Comparison with typical connections for CLT walls

Values of mechanical parameters obtained for the proposed connection can be com‐
pared with analogous quantities assured by angle brackets and hold‐downs typically
used in CLT buildings. Comparative values are reported in Gavric et al. (2011) for
hold‐downs and angle brackets loaded respectively in tension and in shear.
In comparison with a commercial hold‐down having almost the same strength and
stiffness, the proposed connection assures, on average, an approximately twice ulti‐
mate displacement and a ductility value about eight times larger. In comparison with
a commercial angle bracket having similar strength, the proposed X shaped connec‐
tion assures, on average, an ultimate displacement two times larger, a ductility value
nine times larger and an elastic stiffness about 4 times larger.
The proposed connection element shows performances much higher than traditional
ones, in particular when loaded in shear. This evidence testifies that this element can
be properly used to improve the ductility and seismic performances of intrinsically
fragile buildings, such as squat CLT buildings realised with large horizontal panels.
Moreover, the proposed connection makes more feasible the introduction of capaci‐
ty criteria in design of timber buildings limiting value of overstrength factor.

4 Numerical modelling of CLT shear walls

The FE models of three CLT shear walls (wall A, wall B and wall C) were implemented
into ANSYS Workbench (ANSYS Inc. 2012). Quasi‐static cyclic‐loading tests were simu‐
lated according to EN 12512 (2006a) protocol. Wall A and wall B have the same ge‐
ometry (dimension of 2.95 x 2.95 m, aspect ratio 1:1) and vertical distributed load
(18.5 kN/m) of CLT specimens I.1 and I.2, tested by Gavric et al. (2015). These config‐
urations were chosen in order to allow a direct comparison with true panels an‐
chored using traditional connection system. Wall I.1 is anchored with two hold‐
downs and two angle brackets; wall I.2 with two hold‐downs and four angle brackets.
Wall C, representing a squat and scarcely jointed CLT wall, has dimensions equal to
5.90 x 2.95 m (aspect ratio 2:1) and the same vertical distributed load of the other
two walls. Figure 5 shows the geometry and fastener arrangement of the modelled

308
 INTER / 48 - 15 - 04

CLT shear walls. Wall A is anchored with four X brackets (two per each side), whereas
wall B and C with six X brackets (three per side).

(a) (b) (c)

Figure 5. Shear walls: (a)Wall A; (b) Wall B; (c) Wall C. (dimensions are in cm)
4.1 Numerical models of timber walls
Linear elastic shell was used to simulate the timber panels (Figure 6), whereas the
connectors were modelled with the same FE non‐linear models previously described
in section 2. Coupling constraint equations were applied in correspondence of the fix‐
ing points to avoid relative displacements between panels and X brackets and permit
exclusively the relative rotation (hinge connections). The supporting curb was mod‐
elled with only‐compression frictionless rigid springs. Large deformations were ena‐
bled to allow out‐of‐plane buckling of the X brackets, well simulated by the model, as
shown in Figure 7. It has to be stressed that in these analyses X brackets are subject‐
ed to mixed loadings.

(a) (b)
Figure 6. Deformed geometries at maximum top displacement: (a) Wall B ; (b) Wall C .

(a) (b)
Figure 7. Plate buckling under shear loading. (a) Experimental evidence; (b) Numerical prediction.

309
INTER / 48 - 15 - 04

4.2 Analysis of results

This section reports a comparison between numerical results on walls A and B with
experimental ones on walls I.1 and I.2 respectively. Moreover, the predicted behav‐
iour of wall C is presented.
4.2.1 Comparison with CLT walls anchored with traditional connections
Figure 8 shows the predicted base shear force vs. applied top displacement (i.e., hys‐
teresis cycles) for wall A and wall B specimens. The main evidence is the different be‐
haviour of these walls in terms of strength, displacement and cycle amplitude (i.e.,
dissipated energy capacity). It can be seen that wall B (with 6 X brackets) reaches an
higher base shear force than wall A (with 4 X brackets), even if they both fail with a
combined rocking‐sliding behaviour. In wall A connectors are subjected to combined
shear‐tension load. However wall A shows a good seismic response: it reaches an ul‐
timate displacement of 60 mm (drift 2%) without exhibiting strength degradation and
demonstrating good dissipative capability. The two additional connectors placed in
the middle, whose shear resistance is less weakened by contemporary traction due
to rocking, are responsible of the increased ultimate load of wall B. Wall B fails at 60
mm after three fully reversed cycles and with slight strength degradation due to
buckling of X brackets, which however does not compromise the overall behaviour of
the wall.
200.00 200.00
150.00 150.00
100.00 100.00
50.00 50.00
Force (kN)
Force (kN)

0.00 0.00
‐100.00 ‐50.00
‐50.00 0.00 50.00 100.00 ‐100.00 ‐50.00
‐50.00 0.00 50.00 100.00

‐100.00 ‐100.00
‐150.00 ‐150.00
‐200.00 ‐200.00
Displacement (mm) (a) Displacement (mm) (b)
Figure 8. Hysteresis cycles: (a) Wall A; (b) Wall B.
The similitude in terms of geometry, test configuration and loading protocol between
walls A and B and walls I.1 and I.2 tested by Gavric et al. (2015), allows a direct com‐
parison in terms of ductility (), viscous damping ratio (), strength degradation (F)
and strength (Fy, Fmax). These parameters were evaluated according to EN 12512 pro‐
visions. Figure 9 resumes such comparison. It can be seen that the yielding loads for
the CLT walls with the proposed connection system remains similar to that of the
walls with traditional connections, whereas ductility, strength degradation and pinch‐
ing effect are strongly improved. The limited strength degradation and the higher and
stable equivalent viscous damping ratio lead to a strongly increased dissipative capac‐
ity of the walls.

310
 INTER / 48 - 15 - 04

120.00 120.00
100.00 100.00
Resistance [kN]

Resistance [kN]
80.00 80.00
60.00 60.00
40.00 40.00
20.00 20.00
0.00 0.00
Fy Fmax Fy Fmax
Wall I.1 Wall A Wall I.2 Wall B

14.00 14.00
12.00 12.00

Ductility [‐]
Ductility [‐]

10.00 10.00
8.00 8.00
6.00 6.00
4.00 4.00
2.00 2.00
0.00 0.00
20 40 60 20 40 60
Displacement [mm] Displacement [mm]
Wall I.1 Wall A Trend Wall I.2 Wall B Trend

30.00% 30.00%
Strength degradation

Strength degradation

25.00% 25.00%
20.00% 20.00%
15.00% 15.00%
10.00% 10.00%
5.00% 5.00%
0.00% 0.00%
20 40 20 40 60
Displacement [mm] Displacement [mm]
Wall I.1 Wall A Wall I.2 Wall B

30.00% 30.00%
25.00% 25.00%
Viscous Damping
Viscous Damping

20.00% 20.00%
15.00% 15.00%
10.00% 10.00%
5.00% 5.00%
0.00% 0.00%
1st cycle 3rd cycle 1st cycle 3rd cycle
Cycle number Cycle number
Wall I.2 (d=40mm) Wall B (d=40mm)
Wall I.1 Wall A Wall B (d=60mm)

Figure 9. Comparison of seismic performace parameters between CLT walls anchored with
traditional (I.1 and I.2) and proposed (A and B) connections.

311
INTER / 48 - 15 - 04

4.2.2 Analysis of squat CLT wall

The analysis of wall C was performed in order to provide a comparative application
test on a squat wall realized with a unique horizontal CLT panel. Such constructive
methodology is normally adopted in practice and ductility and dissipation capacity of
such walls are mainly demanded to the shear behaviour of traditional angle brackets.
The use of the proposed connection assures adequate ductile seismic response also
for these buildings, as illustrated in Figure 10.
200.00 30.00%
150.00 25.00%

Viscous Damping
100.00 20.00%
50.00 15.00%
Force (kN)

0.00 10.00%
‐100.00 ‐50.00
‐50.00 0.00 50.00 100.00 5.00%
‐100.00 0.00%
1st cycle 3rd cycle
‐150.00
Cycle number
‐200.00 Wall C (d=40mm) Wall C (d=60mm)
Displacement (mm)
14.00 30.00%
Strength degradation

12.00 25.00%
Ductility [‐]

10.00 20.00%
8.00
15.00%
6.00
4.00 10.00%
2.00 5.00%
0.00 0.00%
20 40 60 20 40 60
Displacement [mm] Displacement [mm]

Figure 10. Hysteresis cycles and analysis of wall C.

It can be seen that wall C reaches the highest value of resistance (166.0 kN) and of
viscous damping ratio (about 30.0%). This is mainly due to the pure shear behaviour
of the connection element (i.e., shear strength is not impaired by the traction forces).
At the 40‐mm cycles the out‐of‐plane buckling of X brackets causes a slight strength
degradation, but the connection still maintains its capacity and the strength degrada‐
tion decreases for higher displacements (60 mm).

5 Conclusions
The adoption of newly developed dissipative connections in substitution of traditional
anchoring systems allows to strongly increase the seismic response of CLT buildings,
in terms of ductility and energy dissipation capacity.
The design and optimization phases of a newly developed X shaped dissipative brack‐
et has been described in the paper, together with following experimental validation
and interpretation of obtained results. It has been demonstrated that such innovative

312
 INTER / 48 - 15 - 04

devices, specifically developed to be used in CLT structures, assure several times

larger ductility and dissipative capacity than standard connections of comparable
yielding strength and elastic stiffness. The assessment of the mixed shear‐tension be‐
haviour (i.e., definition of the strength domain), keeping into account for local buck‐
ling phenomenon, is fundamental to correctly assess the actual dissipative properties
of such devices. The procedure adopted in this paper could serve as a reference for
the due certification of dissipative connections by means of testing and/or numerical
modelling.
Effects of utilization of the proposed X brackets to fasten CLT shear walls have been
simulated numerically. Obtained results confirm that obtained CLT shear walls are
classifiable into High Ductility Class according to Eurocode 8. Even shear walls real‐
ized with large horizontal CLT panels show a favourable highly dissipative behaviour.
In view of a code implementation, increased q‐factor should be allowed when dissi‐
pative connection elements are employed for the realization of massive shear walls
or shear frames. It has to be highlighted also that the utilization of dissipative connec‐
tions with limited controlled overstrength makes the utilization of a capacity design
approach more realistically affordable in designing of timber structures.

6 References
ANSYS Mechanical Workbench R14 (2012).
Blass, HJ, Schmid, M, Litze, H, Wagner, B (2000): Nail plate reinforced joints with
dowel‐type fasteners. In Proceedings of the 6th World Conference on Timber Engi‐
neering
Ceccotti, A, Sandhaas, C, Okabe, M, Yasumura, M, Minowa, C, Kawai, N (2013): SOFIE
project – 3D shaking table test on a seven‐storey full‐scale cross‐laminated timber
building. Earthquake Engineering & Structural Dynamics, 42(13):2003‐2021.
EN 12512 (2006a): Timber structures—test methods—cyclic testing of joints made
with mechanical fasteners. CEN. Brussels, Belgium.
EN 1990 – Eurocode (2006b): Basis of structural design. CEN. Brussels, Belgium.
Eurocode 8 (2013): Design of structures for earthquake resistance, part 1: general
rules, seismic actions and rules for buildings. CEN. (EN 1998‐1‐1). Brussels, Belgium.
Foliente, GC (1996): Issues in seismic performance testing and evaluation of timber
structural systems. Proceedings of the International Wood Engineering Conference,
New Orleans, USA.
Foschi, RO, Bonac, T (1977): Load slip characteristic for connections with common
nails. Wood Science and Technologly, 9(3): 118‐123.
Fragiacomo, M, Dujic, B, Sustersic, I (2011): Elastic and ductile design of multi‐storey
crosslam massive wooden buildings under seismic actions. Engineering Structures,
33(11):3043‐3053.

313
INTER / 48 - 15 - 04

Gavric, I, Ceccotti, A, Fragiacomo, M (2011): Experimental cyclic tests on cross‐

laminated timber panels and typical connections. Proceedings of ANIDIS, Bari, Italy.
Gavric, I, Fragiacomo, M, Ceccotti, A (2013): Capacity seismic design of X‐LAM wall
system based on connection mechanical properties. Meeting 46 of the Working
Commission W18‐Timber Structures, CIB, Vancouver, Canada, paper CIB‐W18/46‐
15‐2.
Gavric, I, Fragiacomo, M, Ceccotti, A (2015): Cyclic behaviour of CLT wall systems: ex‐
perimental tests and analytical prediction models. ASCE Journal of Structural Engi‐
neering, DOI: 10.1061/(ASCE)ST.1943‐541X.0001246.
Latour, M, Rizzano, G (2015): Cyclic Behavior and Modeling of a Dissipative Connector
for Cross‐Laminated Timber Panel Buildings. Journal of Earthquake Engineering,
19(1): 137‐171, DOI: 10.1080/13632469.2014.948645.
Priestley, M J N, Seible, F, Calvi, G M (1996): Seismic design and retrofit of bridges,
Wiley, New York.
Loo, WY, Kun, C, Quenneville, P, Chouw, N (2014): Experimental testing of a rocking
timber shear wall with slip‐friction connectors. Earthquake Engineering and Struc‐
tural Dynamics, 43(11):1621‐1639.
Muñoz, W, Mohammad, M, Salenikovich, A, Quenneville, P (2008): Need for a har‐
monized approach for calculations of ductility of timber assemblies. Proceedings of
the Meeting 41 of the Working Commission W18‐Timber Structures, CIB, Saint An‐
drews, Canada.
Polastri, A, Angeli, A, Dal Ri, G (2014): A new construction system for CLT structures.
Proceedings of World Conference on Timber Engineering WCTE, Quebec City, Cana‐
da.
Popovski, M, Gavric, I (2015): Performance of a 2‐story CLT house subjected to lateral
loads. ASCE Journal of Structural Engineering, DOI: 10.1061/(ASCE)ST.1943‐
541X.0001315.
Pozza, L, Scotta, R, Trutalli, D, Ceccotti, A, Polastri, A (2013): Analytical formulation
based on extensive numerical simulations of behavior factor q for CLT buildings.
Meeting 46 of the Working Commission W18‐Timber Structures, CIB, Vancouver,
Canada, paper CIB‐W18/46‐15‐5.
Pozza, L, Scotta, R (2014): Influence of wall assembly on q‐factor of XLam buildings.
Institution of Civil Engineers Journal Structures and Buildings.
DOI:10.1680/stbu.13.00081.
Sarti, F, Palermo, A, Pampanin, S (2015): Quasi‐Static Cyclic Testing of Two‐Thirds
Scale Unbonded Posttensioned Rocking Dissipative Timber Walls. ASCE Journal of
Structural Engineering, DOI: 10.1061/(ASCE)ST.1943‐541X.0001291.
Smith, I, Landis, E, Gong, M (2003): Fracture and fatigue in wood. John Wiley and
Sons, Chichester, UK.

314
 INTER / 48 - 15 - 04

Discussion

The paper was presented by R Scotta

F Sarti asked whether the differential movement between the wall and the connector
was considered. R Scotta said the connector was assumed to be well connected to the
wall and the differential movement was not considered. F Sarti discussed about verti-
cal load that could help the self-centring in rocking systems. R Scotta responded that
in squat systems, shear deformation is more important and has to be considered first.
T Tannert asked about the fancy shape of the connector. R Scotta said that this was
made for production efficiency and consideration to use less material.
H Blass asked whether a steel angle type connector would be needed for the wall to
floor connection. R Scotta said yes especially in wall to concrete foundation some
steel plates would be needed to help transfer the high forces.
F Sarti asked about buckling and low cycle fatigue issues. R Scotta responded that
when designing the connector, limiting the maximum strain would be needed to take
low cycle fatigue into consideration. Buckling was observed and modelled, but did not
influence the strength much.
B Dujič and R Scotta discussed that friction needed to be considered.
Z Li commented that the buckling would change the force distribution and one might
need to avoid this failure mode. He received confirmation that mild steel with 275
MPa yield strength was used.
W Seim commented that this component test was a steel to steel connection but in
buildings this would be steel to wood connection. R Scotta agreed and will use model
to take this aspect into consideration.
U Kuhlmann stated that when using mild steel one must consider the over strength in
the steel.
I Smith commented about practicality and tolerance, and whether the potential users
have any opinions. R Scotta stated that the study mainly focused on the structural
performance issues first.

315
316
 INTER / 48 - 16 - 01

Analysis of fire resistance tests on timber

column buckling with respect to the Re-
duced Cross-Section Method
Joachim Schmid, SP Wood Building Technology and LTU, Sweden
Michael Klippel, ETH Zurich, Institute of Structural Engineering, Switzerland
Andrew Liew, ETH Zurich, Institute of Structural Engineering, Switzerland
Alar Just, SP Wood Building Technology and TUT, Sweden and Estonia
Andrea Frangi, ETH Zurich, Institute of Structural Engineering, Switzerland
Keywords: Fire design model, Reduced Cross-Section Method, Buckling, Timber,
Standard fire, Fire resistance, Fire testing

1 Abstract
The Reduced Cross-Section Method (RCSM) is a popular method for the design of
timber members exposed to fire, which uses an effective cross-section and mechani-
cal properties at normal temperature. The RCSM was adopted from publication that
was originally developed for single-span beams subjected to bending (Schaffer
1984).It has been introduced in EN 1995-1-2 (Eurocode 5) (CEN 2004) for a large
range of timber members, including columns under compression and members under
tension. Recently, the applicability of the method and its extended applications has
been questioned with respect to the basis of limitations and contradictions found by
advanced simulations (Klippel et al. 2012) and a comprehensive review (Schmid et al.
2014) of fire resistance test results for members in bending, tension and compres-
sion. In the present paper, results of a review of buckling members exposed to fire
are presented. This paper analyses a total of 126 full-scale fire resistance tests of
members subjected to buckling and using the RCSM. The analysis shows that (i) most
of the literatures are of too poor quality, or incomplete, to be used for developing or
verifying a design model, and (ii) considering the variability of input data, a significant
scatter implies that the RCSM in its present form is not able to describe sufficiently
the behaviour of timber columns exposed to fire. It is therefore recommended that
the existing design approach in Eurocode 5 should be revised and further work initi-
ated.

317
INTER / 48 - 16 - 01

2 Introduction
The RCSM in EN 1995-1-2 can be used to verify the load-bearing capacity of a large
number of timber member in a fire situation, e.g. bending, compression buckling or
lateral torsional buckling. Although no limitations are explicitly given on the use of
the RCSM, it is assumed that any verification calculation described in EN 1995-1-1
(CEN 2004) can be performed using the effective cross-section. This is in contradic-
tion to the different properties governing the mechanical response of structural tim-
ber members and the different reduction curves for specific properties of timber with
increasing temperature also given in Annex B of EN 1995-1-2. The RCSM was original-
ly proposed for simply-supported glued-laminated (glulam) beams by Schaffer (1984).
For its implementation in EN 1995-1-2, the method was further simplified and its use
extended to other members, e.g. members influenced by buckling. A systematic
comparison of the RCSM and fire test results was never published when drafting
EN 1995-1-2. Recently, this RCSM specifying a general zero-strength layer d0 of 7 mm
was discussed in some papers (Klippel, Schmid et al. (2012), Schmid et al. (2014)). A
comparison of fire test results with the RCSM of members in bending, tension and
compression was presented by Schmid, Klippel et al. (2014). It was shown that the
scatter of results is significant and that most reports are of such poor quality that it is
not recommended to use them to verify a calculation model. Simulation results from
Klippel, Schmid et al. (2012) showed that the RCSM appears to be appropriate for
members in tension, but my lead to non-conservative results for members in com-
pression.
The aim of this paper is to compare the RCSM specified in EN 1995-1-2 with actual
fire test results. Test results presented here range from fire tests performed in the
1960s to recent tests performed in 2005. This paper compares fire test results in
terms of the determined zero-strength layer d0. It was focused on analysing fire tests
for members in buckling, excluding members with failure in pure compression. Since
the main design procedure in EN 1995-1-2 is based on the standard fire exposure
EN 1363-1 (CEN 2012), only fire tests following the standard time-temperature fire
curve were evaluated.

3 Analysis procedure
The large variety of available literature from different authors, fire test standards and
technical setup possibilities results in a large spectrum of reports and papers to ana-
lyse. To overcome a subjective judgement of data, a procedure to classify results was
followed, as described in this section. The procedure for the suitability, classification
and extension of data is described in detail by Schmid, Klippel et al. (2014).
3.1 Suitability and classification of references
In the first step, the available literature was analysed, reports and papers had been
classified as primary or secondary. Secondary references were considered as those

318
 INTER / 48 - 16 - 01

where test reports were used to determine or improve a design method instead of
reporting test results. These are often referred to when the reliability of today’s de-
sign models is discussed (Lie 1977, Scheer 1994). Only primary references give details
of conducted tests, but do not contain necessary design recommendations and de-
sign rules. For the quantitative analysis of the zero-strength layer, mainly primary
references were used.
In the second step, the suitability of the references was evaluated before further
analysis. Although very ambitious and costly, some test series are not suitable for fur-
ther analysis with respect to the assessment of the RCSM and determination of the
zero-strength layer respectively, since serious mistakes and wrong assumptions were
made. Sometimes the results of such studies are misinterpreted to prove agreement
with today’s calculation methods. Although these references were summarised, and
errors highlighted in this study, in order to point out the unsuitability of the refer-
ences with respect to the RCSM and to identify the mistakes which can be avoided in
future test series. The data from these references were not analysed further.
When a reference was found to be suitable for further analysis, the reported data
sets were classified as (i) certain data, (ii) uncertain data, and (iii) very uncertain data,
depending on the availability of fire test details and the characterisation of the mate-
rials tested.
To classify the references, the test characteristics reported as data sets were evaluat-
ed in terms of the feasibility of performing backwards calculation of a corresponding
zero-strength layer. The requirements to consider a data set as complete, and the
calculated results for the zero-strength layer as certain, call for the following test de-
tails in the references: (i) standard fire exposure during loaded tests with (ii) well-
defined support conditions, (iii) documentation of the failure time, (iv) appropriate
definition of the timber at normal temperature to allow a prediction of the load-
bearing capacity, (v) adequate initial moisture content (MC) of the timber member
before the fire test and (vi) documentation of the residual cross-section by means of
appropriate measurements. Only determined values for the zero-strength layer
based on data sets fulfilling all these obligatory requirements are classified as certain.
If one requirement was not fulfilled, the result was classified as uncertain. If more
than one compulsory requirement was not fulfilled, the result was classed as very un-
certain.
3.2 Extension of the data sets
Available reports describe fire tests with varying degree of completeness. To extend
incomplete data sets, different procedures were used to allow further analysis of the
test reports. Two major problems were observed when analysing the available litera-
ture. (i) the poor quality of the material determination allowing a prediction of the
load-bearing capacity at normal temperature and (b) the inadequate description of
the boundary conditions for the buckling tests considering the support conditions.

319
INTER / 48 - 16 - 01

3.2.1 Estimation of the support conditions

3.2.1.1 General
In general, the buckling behaviour of structural members is complex. The mechanical
performance of timber members was investigated in many studies at normal tem-
perature (Theiler et al. 2012). For the fire situation, different models are available,
e.g. the model of (Lie 1977, APA 2009) where the residual cross-section is considered
alongside a reduction of the strength and stiffness by 20%.
In EN 1995-1-1 (CEN 2004), the concept of the Effective Length Method (ELM) is giv-
en for the calculation of the load-bearing capacity of timber columns at normal tem-
perature. Using this method the buckling problem can be reduced to the structural
system of a simply supported column. Non-linear effects are taken into account by
means of a buckling factor kc. Alternative to the ELM, according to EN 1995-1-1 a
second order analysis may be performed which tends to deliver non-conservative re-
sults compared to the ELM when using mean values for the MOE in bending (Theiler,
Frangi et al. 2012). For the fire design, EN 1995-1-2 specifies the RCSM which is also
valid for buckling members.
In the present investigation, the zero-strength layer for the buckling situation was de-
termined using the ELM: (i) due to the intention to use mean values for strength and
stiffness properties, and (ii) due to the fact that the RCSM should be capable to be
used together with the ELM as intended by designers.
To estimate the buckling conditions typical for fire testing frames a Finite Element
Analysis (FEA) was conducted. Results were used to perform appropriate backwards
calculation of the load-bearing performance of the columns tested in fire.
3.2.1.2 Buckling lengths in fire testing frames
The support conditions are important for the analysis of tests at normal temperature
as well as for fire resistance tests. In some of the literature, timber columns were
placed between horizontal supports at both ends, e.g. a horizontal moving but not
rotating top support (e.g. a spreader beam).
The shape influence of the column end surface in buckling was analytically and exper-
imentally investigated by König (1988).The effective length is dependent on the radi-
us r of the contact surface. The two limiting cases can be described by the Euler
mode 2 with pin jointed ends and an effective length lk = l (buckling factor 1.0) and at
the other extreme where r =  corresponding to Euler mode 4 with fixity (clamped
support) at both ends and an effective length lk = 0,5·l (buckling factor 0.5), see Fig-
ure 1. Timber specimens exposed to fire keep often the curvature after the test de-
veloped during the fire test under loaded conditions. This can be used to evaluate
support conditions with respect to their mechanical boundary conditions, see Figure
2. The cross laminated timber (CLT) wall element was placed on the bottom to a hori-
zontal CLT to reflect building practice. The top support was free to rotate. It can be

320
 INTER / 48 - 16 - 01

observed that the horizontal support at the bottom end (r = ) corresponds to re-
straint conditions (clamped support) while for the top end the pinned support is
clearly visible in Figure 2.
1.0
=lk/l

0.8

0.5
0 0.5 1 1.5
r/l

Figure 1. Effective length, ex- Figure 2. Buckling shape of a CLT element after
pressed as lk = ·l as function of the fire resistance test with an elastic clamping
the end surface radius (König at the bottom end and pinned at the top end
1988). (Schmid et al. 2015).

3.2.1.3 Finite Element analysis to investigate the support conditions

In the past, many fire testing frames did not use well defined support conditions. The
influence of the actual support conditions on the load-bearing behaviour of timber
columns in buckling subjected to compression was studied with non-linear Finite El-
ement Analyses (FEA). These simulations used test data obtained from the tests at
normal temperature (Theiler, Frangi et al. 2012) to verify the model. A three-
dimensional FE model was implemented in Abaqus using a phyton script to control a
variety of input parameters. The analyses were performed to show the influence of a
non-rotating and horizontal support on the buckling behaviour of timber columns
loaded in compression. Three-dimensional solid brick elements were used with the
“Concrete Damaged Plasticity” (CDP)-model provided by the Abaqus material library
to describe the material behaviour. The CDP-model is able to model both the linear
elastic brittle behaviour of timber in tension and the linear elastic plastic behaviour
under compression. The developed Abaqus model was first verified by comparing the
results from the simulations with the test results of Theiler, Frangi et al. (2012), who
investigated the behaviour of 50 columns (constant 𝑤 = 140 mm , ℎ = 160 mm,
and varying 𝑙 =1400, 2300, 3200 mm) under compression load.
Prior to testing, the lamellas of each column were thoroughly inspected for knots and
the dynamic modulus of elasticity was measured. Based on this information, Theiler,
Frangi et al. (2012) estimated the strength and stiffness of each lamella. This infor-
mation for the material properties of each lamella was introduced in the Abaqus

321
INTER / 48 - 16 - 01

model. Simulations and tests were in good agreement with regard to both the initial
stiffness and ultimate strength of the columns. Both pinned and fixed support condi-
tions were modelled and showed that a horizontal support, i.e. a stiff steel beam, co-
incides with the behaviour of fixed support conditions for a timber column. Further,
mixed support conditions (FP) typically used in fire resistance tests were investigated.
Results for a comparison of the support conditions are given in EN 1995-1-2. The FEA
results were used to determine appropriate buckling lengths for the buckling model
given in EN 1995-1-2. This assumption is supported by König (1988). For mixed sup-
port conditions (FP) with a pinned support and a horizontal support, a conservative
buckling factor (slightly lower value than following the FEA) of 0.7 was used in the
analysis. For support on horizontal and non-rotating surfaces (elastic clamping) the
buckling factor 0.5 was used.
3.2.2 Estimation of the material properties
It is self-evident that the timber quality of members tested in fire has a large influ-
ence on the load-bearing capacity. In most of the literature, timber is graded using
different philosophies either related to the bending strength at normal temperature,
visual grading or distinction in solid or glulam timber only. The lack of thorough inves-
tigations into the material properties, and the less than satisfactory prediction meth-
ods used by many authors in the references, lead to significantly high uncertainties in
the conclusions regarding the load-bearing capacity and subsequently in the results
of the zero-strength layer depth based on these tests. Any underestimation of the
strength at normal temperature leads obviously to unacceptable general conclusions
and to unusable data for the development of any calculation method. Results in this
paper that show negative values for the zero-strength layer imply either (i) a contri-
bution of the char layer to the load-bearing capacity, (ii) an improvement in strengths
of the heated cross-section or (iii) an underestimation of the strengths at normal
temperature. Since (i) and (ii) can be excluded, the error source is the limited infor-
mation describing the material tested in the fire situation.
Reference tests at normal temperature are considered as a reliable method for the
prediction of the load-bearing capacity of the specimens. As no buckling tests were
performed in the literature analysed, the determined material properties, e.g. densi-
ty, were used to predict the load-bearing capacity of the columns and determine the
input data for a backwards calculation of the zero-strength layer. Details of the pro-
cedure are given in Section 4 and in (Schmid, Klippel et al. 2014).

4 Fire test results and analysis

In total, 126 fire test results were collected and reviewed, with 84 tests suitable for
further analysis to perform a backwards calculation to determine the zero strength
layer depth. Details are given in Table 1 specifying the test program, standard and pa-
rameters as well as the number of performed tests. Report summaries in Sections 4.1

322
 INTER / 48 - 16 - 01

to 4.7 are needed, since the testing procedures vary significantly and make it impos-
sible to compare all details in the tables. Furthermore, many of the references are lit-
erature published in languages other than English.
Table 2. Overview of the references available and further analysed.

Title Test program and test parameters

Author and and

No. of fire
Classification of results tests* Determined resistance

Essais de résistance au feu

5 Geometry
Fackler (1961) (Fire resistance tests)
(2)
Not Suitable 41 to 72 min

Grade, adhesive, geometry, species, load level, fire retardant

Fire-Resistance of Laminated Timber Columns
Malhotra and 26 treatment, fire protection system
Rogowski (1967) (16)
Very Uncertain 23 to 77 min

Das Brandverhalten von Holzstützen unter Druck-

beanspruchung Monolithic and segmented, grade, solid and glulam timber,
Klement et al. 72
(The fire behaviour of timber columns exposed to resorcinol and urea adhesives, geometry, load ratio
(1972) (68)
compression load)

Uncertain 20 to 114 min

Untersuchungsbericht zum Brandverhalten von

brettschichtverleimten Holzstützen mit Rechteck-
Glulam columns, geometry, support condition (elastic clamped
querschnitt 17
Haksever (1976) and simply supported)
(Report on the fire behaviour of glue laminated (15)
timber columns with rectangular cross-section)

Uncertain 22 to 96 min

Ali and Kavanagh Fire resistance of timber columns 6 Geometry

(2005) Not Suitable (6) 6 to 20 min

* number in brackets specifies the no. of fire tests with rectangular cross-sections which are considered in this study if found Suitable

4.1 ......Fackler, 1961

The oldest reference available presents results from a larger study on the fire re-
sistance including different kind of building products. Two rectangular loaded glulam
columns were tested with a fire resistance of about 48 minutes. Since neither the
support conditions nor the timber grade is specified nor reference tests reported, no
further evaluation of these fire tests was performed in this study.

323
INTER / 48 - 16 - 01

4.2 ......Malhotra and Rogowski (1967)

26 full-scale tests with glued laminated tim- 1.0 40

Ffi/F20 [mm]
Series Urea
Series Casein
ber columns were performed of different Series Resorcinol
35

squared and rectangular cross-sections of 0.8 Series Phenolic

Zero-strength layer
30

approximate constant area equal to the 25

y = e-0.04x
squared 230 mm x 230 mm. The grade was 0.6
R² = 0.78
20
UK grade LB and four different adhesives 15
were used to produce the columns. Tests 0.4 10
with different load-levels were performed in 5
a loading frame with restrained support 0.2 0
conditions and with constant loading. Some -5
tests were performed with applied fire pro- 0.0 -10
tection or fire-retardant treatments, 16 0 15 30 45 60 75
t [min]
tests were evaluated further in the present
investigation. For the analysis, the specific Figure 3. The load-ratio and failure time
strength and MOE in bending was deter- of the test results (coloured markers) in-
mined by means of the density. Thus, the cluding an exponential regression line
results are classed as very uncertain alt- and the corresponding zero-strength
hough an exponential regression show lim- layer depths (white squares).
ited deviations, see Figure 3. Further results
are given in Table 2. The exponential behaviour of timber members in fire was earlier
observed (König 1995).

4.3 Klement et al. (1972), Stanke et al. (1973), Seekamp (1969), and Rudolphi (1979)

A comprehensive test program was performed to investigate the fire performance of

timber members in buckling by Stanke et al. (1973). A different selection of tests and
different details are given in different reports with the intention to develop national
verification models, e.g. the German F60 criterion (Seekamp and Stanke 1969). Thus,
a large number of parameters was varied, see Table 2. Segmented cross-sections
were not further analysed in this review. The fire tests came along with an investiga-
tion of the source material comprising the moisture content, the density (all col-
umns), the compression strength and stiffness (cubes 20 mm x 20 mm x 60 or
120 mm; ca. every third column), the MOE in bending (static; three columns) and the
dynamic MOE using ultrasonic sound transmission.
The length of the tested members was 3650 mm while the fire exposed length was
3000 mm (non-centric). In some fire tests the oxygen content of the exhaust gases
was measured (around 5%) to control the completeness of the combustion. Fire tests
lasted from 10 min to 114 min and were terminated after buckling failure of the
member.

324
 INTER / 48 - 16 - 01

20 000

fc,prediction (Thunell, rmean) [MPa]

MOEprediction (rmean) [MPa]
45 Glulam timber
Solid timber

15 000 exakt match

15% deviation

30
10 000 2 tests

Glulam timber 15
5 000
Solid timber
exakt match
15% deviation
0 0
0 5 000 10 000 15 000 20 000 0 15 30 45
MOEc,reference test [MPa] fc,reference test [MPa]

Figure 4. The predicted values for the modu- Figure 5. The predicted values for the
lus of elasticity in compression based on the compression strength using the model
reference tests. of Thunell (1971) based on the density
of the specimens.

The charring depth and the geometry of the residual cross-section were evaluated on
the basis of unloaded, 500 mm long reference specimens exposed in the same fire
test. One residual cross-section was used to determine the notional charring depth
(rectangle with corresponding area) which is in-line with the actual design model
(CEN 2004). The moisture content of the columns at the beginning of the fire test was
not specified. Deviating from the verification model presented in (Stanke, Klement et
al. 1973), the buckling length for the actual testing frame with a horizontal, non-
rotating top support and a hydraulic jack was estimated to 0,75𝑙 following Section
3.2.1.3. As the density is the only material characteristic available for all columns it
was used to estimate the compression strength 𝑓𝑐 and the modulus of elasticity in
compression for the analysis presented this study. To use the buckling model given in
EN 1995-1-2 the MOE in bending was determined by means of the density using the
technique presented by Schmid, Klippel et al. (2014) based on the relationship be-
tween compression strength and MOE in bending given in EN 1194 (CEN 1999),
EN 338 (CEN 2009) and JCSS (JCSS 2007).
A comparison of the prediction of the MOE by using the density and results of the
reference tests is given in Figure 4. A comparison of the prediction of the compres-
sion strength using the density and results of reference tests prediction is given in
Figure 5. In both figures it is shown that the error of prediction is about 15%. The
span of ±15% for the strength and stiffness was later used to estimate the influence
on the zero-strength layer depth. When the strength or the MOE in compression was
available, this value was used for further calculations.

325
INTER / 48 - 16 - 01

The residual cross-section for glulam mem- 28

d0 [mm]
26
bers was determined for this study using 24
the specified linear equation by Stanke, 22
20
Klement et al. (1973) while for solid mem- 18
16
bers no equation was available a 10% in- 14
12
creased value was used for further calcula- 10
tions in this study. Using the assumptions 8
6
for the estimation of the material proper- 4
2
ties the results are evaluated as Uncertain, 0
-2
results are given in Figure 6 and Table 2. -4 Glulam timber
The variation of the strength and stiffness -6 Solid timber
-8
would result in 2,7 mm lower and 2 mm -10
0 15 30 45 60 75 90 105 120
higher values respectively (maximum devi- t [min]
ation values).
Figure 6. Calculated zero-strength layer
for glulam and solid timber columns.

4.4 ......Haksever (1976)

The report contained the imprecise support 18

d0 [mm]

conditions of earlier fire tests (Stanke, 16

Klement et al. 1973), thus simply supported 14
12
(PP) soft wood (GK1, German grading class 10
1) and glulam columns (glued with resorcinol 8
adhesive) were tested and compared to the 6
elastic climbed condition (four tests). To de- 4
fine the material properties, the strength fc 2
0
and the stiffness Ec under compression (par- -2 glulam
allel to grain) were determined by means of -4
eight specimens (areas 140 mm x 140 mm solid timber
-6
and 280 mm x 280 mm; lengths 600 or -8
0 15 30 45 60 75 90
900 mm) produced from the same materials t [min]
as the columns. The charring rate was de-
termined to 0,7 mm/min by means of Figure 7. Calculated zero-strength layer
for glulam and solid timber columns.
unloaded specimens placed at the bottom of the furnace. The residual cross-sections
and the charring depths specified in the report were estimated using the middle sec-
tion of the specimens.
In the analysis for this study, the results for the reference tests were used for all col-
umns with the particular cross-section. The MOE in bending was determined by
means of the compression strength fc using the technique presented in (Schmid,

326
 INTER / 48 - 16 - 01

Klippel et al. 2014) based on the relationship between compression strength and
MOE in bending given in EN 1194, EN 338 and JCSS. Due to the weak material predic-
tion as well as the limited information on the residual cross-sections the results are
considered as Very Uncertain. Results are shown in Figure 7 and given in Table 2.
4.5 ......Lie (1977)

The work is used in literature as reference for design models for timber columns in
buckling although it is a secondary source where 67 fire tests in buckling and bending
were evaluated. However, it is not further specified which fire test results of columns
were evaluated. References lead back to fire tests with members in bending and
buckling reporting beam tests and 88 column tests (Malhotra and Rogowski 1967,
Klement, Rudolphi et al. 1972). It remains unknown which selection criteria were
used and which tests were further used to develop the model. Further, in (Lie 1977)
test loads are described using the indication “allowable load” which indicates nation-
al regulation systems that are not further specified. It is mentioned that the actual
safety factors was not known. In this study, this reference was not used any further.
4.6 ......Rudolphi (1978)

The doctoral thesis deals with the fire design of steel and timber members in general
and the validity of fire resistance tests performed at different labs in specific. No fire
resistance tests were conducted in this study but results reported by Klement,
Rudolphi et al. (1972) leading back to tests of Stanke, Klement et al. (1973). This sec-
ondary source is referenced in literature to verify calculation models, e.g. Schnabl et
al. (2011) gives interesting details about the fire tests performed earlier. Whereas for
steel columns tested in the same fire testing frame the buckling factor was assumed
to be 0,7l the buckling length of timber columns were assumed equal to the column
length. The support conditions and effects of an elastic clamping were discussed but
rejected later. The author estimated the charring depth at the failure time for all fire
tests and presented linear regression functions for all results of Klement, Rudolphi et
al. (1972). Since the curves for a reduced buckling length of 0,7l and 0,5l would inter-
sect at about 10 and 20 mm at t = 0 min respectively it was concluded that the sup-
port conditions were pinned on both ends (PP) instead of ascribing the increase to
losses in strength and stiffness. With this observation, it can be concluded that the
author was the first observing and quantifying a zero-strength layer. Based on his ob-
servations a very rough estimation of the value can be estimated up to 20 mm. As the
report is a secondary source, this study was not further evaluated in this paper.
4.7 ......Ali and Kavanagh (2005)

The authors performed six fire resistance tests with timber columns of 1800 mm
length to determine failure temperatures depending on the load level and slender-
ness as available for steel structures. Two different cross-sections (100 mm x 100 mm

327
INTER / 48 - 16 - 01

and 75 mm x 75 mm) of solid timber columns (C24) were tested as buckling members
in normal temperature (reference test) and in the fire situation. No information on
the support condition is found in the reference but according to personal author in-
formation the columns were simply supported on the top and bottom using half
spherical steel supports (Ali, Faris, personal communication, 20th May, 2015) repre-
senting pinned supports on both ends. Fire tests were terminated at buckling failure
between 6 and 20 minutes. The charring layer depth was evaluated but the shape of
the residual cross-section was not further investigated. Failure loads are given in ratio
to the design model of EN 1995-1-1 (402 kN and 106 kN) and reference tests (446 kN
and 127 kN). All values deviate considerably from a correct estimation of the buckling
failure load at normal condition (164 kN and 68 kN disregarding safety factors) and
were not considered for further estimation in this paper.

5 Results and discussion

The evaluation of buckling members required specific effort since the support condi-
tions are crucial and standard testing has experienced significant changes since be-
ginning of fire testing. Early design models assumed simply supported columns while
later tests showed significant influence of the support conditions even in fire tests.
The results of the comprehensive analysis of performed fire tests support the con-
cerns raised recently. In this paper, on the basis of extensive experimental investiga-
tions supported by FEA it could be shown that the zero-strength layer d0 for timber
members subjected to compression is different to the constant value currently used
in EN 1995-1-2. This paper extends the analysis of the zero-strength layer significantly
by a comprehensive evaluation of 126 fire tests under ISO-fire exposure. The mean
value for the zero-strength layer of all results is about 11 mm for members in buck-
ling (S.D. about 8 mm). Further details are given in Table 2. In general, results show
that no significant difference between glulam and solid timber members was found
but a considerable scatter of results indicate severe problems to describe the com-
plex processes of fire exposed timber members in buckling with a single value.
Table 2. Minimum, mean and maximum values for the zero-strength layer of the analysed fire test
results for buckling members (values in mm), number or results and classification of sources.
Malhotra, 1967 Haksever, 1976 Haksever, 1976 Stanke et. al, 1973 Stanke et. al, 1973
(glulam timber) (glulam timber) (solid timber) (glulam timber) (solid timber)
min -9.4 0.7 4.9 -8.2 2.0
mean 10.0 12.5 6.6 9.9 14.1
max 23.6 27.2 9.2 25.4 25.9
No. of results 16 11 4 53 12
Classification Very Uncertain Very Uncertain Very Uncertain Uncertain Uncertain

328
 INTER / 48 - 16 - 01

6 Conclusions
In general, the analysis of fire tests is complex since the references give only limited
information regarding the material tested. Where information was lacking, the data
sets were extended and results classified in Certain, Uncertain and Very Uncertain.
None of the tests performed meet today’s requirements of verification tests and
none of the reports meet the requirements to perform a backwards calculation of
the zero-strength layer depth without appropriate assumptions. Thus, not a single
reference is classified Certain.
Today, in EN 1995-1-2 (Eurocode 5) (CEN 2004) the complex behaviour of timber
members in fire is described by a single constant value of d0 = 7 mm developed origi-
nally for bending members. When the design model was presented a further verifica-
tion of the design model using the zero-strength layer was proposed (Schaffer 1984),
however this was not performed. In this paper it is shown that the available full-scale
fire tests cannot be used to verify the design model introduced also for buckling
members. The scatter of results indicates either that the uncertainties regarding the
material tested cannot be controlled easily or that the zero-strength layer may be
depending on several parameters.
As the EN 1995-1-2 is under revision, it is recommended to consider the modification
of the existing design model with respect to the different mechanical models for
bending and buckling since earlier investigations showed different results for mem-
bers in tension, compression and bending (Klippel, Schmid et al. 2012, Schmid, Just et
al. 2014). Based on the analysis of all test results presented in this paper a mean val-
ue for the zero-strength layer depth for buckling members in the range of about
11 mm has been obtained.
For further investigations a combined approach of simulations and verification tests
in model and full-scale is recommended. Simulations should cover the variation of
the material properties as well as the change of stiffness when the member is re-
duced by char.

7 References
Ali, F. and S. Kavanagh (2005). "Fire resistance of timber columns." Journal of the
Institute of Wood Science 17(2): 85-93.
APA (2009) "Calculating Fire Resistance of Glulam Beams and Columns."
CEN (1999). EN 1194: Timber structures — Glued laminated timber — Strength
classes and determination of characteristic values. European Standard. Brussels,
European Committee for Standardization.
CEN (2004). EN 1995-1-1: Design of timber structures - Part 1-1: General - Common
rules and rules for buildings. Brussels, European Committee for Standardization.
CEN (2004). EN 1995-1-2 Design of timber structures – Part 1-2: General – Structural
fire design. Brussels, European Committee for Standardization.

329
INTER / 48 - 16 - 01

CEN (2009). EN 338: Structural timber – Strength classes. Brussels, European

Committee for Standardization.
CEN (2012). EN 1363-1: Fire resistance tests - Part 1: General requirements. Brussels,
European Committee for Standardization.
Fackler, J. P. (1961). Essais de résistance au feu. Paris, Centre Scintifique et Technique
du Bâtiment, Paris, France.
Haksever, A. (1976). Untersuchungsbericht zum Brandverhalten von
brettschichtverleimten Holzstützen mit Rechteckquerschnitt. Technische
Universität Braunschweig, Berlin, Germany.
JCSS (2007). Probabilistic Model Code. (http://www.jcss.byg.dtu.dk), Joint Committee
on Structural Safety, Danmark.
Klement, E., R. Rudolphi and J. Stanke (1972). Das Brandverhalten von Holzstützen
unter Druckbeanspruchung (in German), Germany.
Klippel, M., J. Schmid and A. Frangi (2012). The Reduced Cross-section Method for
timber members subjected to compression, tension and bending in fire.
International Council for Research and Innovation in Building and Construction,
Working Commission W18 – Timber Structures, Meeting 45 (CIB-W18 Meeting
2012). Växjö, Sweden, CIB-W18 Meeting 2012. 45: CIB-W18/45-16-11.
König, J. (1988). The structural behaviour of axially Loaded Wood Studs Exposed to
Fire on one Side. Report I8808057. Stockholm, Sweden, TräteknikCentrum.
König, J. (1995). Fire Resistance of Timber Joists and Load Bearing Wall Frames.
Stockholm, Trätek, Institutet för träteknisk forskning: 102.
Lie, T. T. (1977). "A method for assessing the fire resistance of laminated timber
beams and columns." Canadian Journal of Civil Engineering 4(2): 161-169.
Malhotra, H. L. and B. F. W. Rogowski (1967). Fire Resistance of Laminated Timber
Columns. Fire and Structural Use of Timber in Buildings, HMSO.
Schaffer, E. (1984). "Structural Fire Design: Wood, Research Paper FPL 450." US
Department of Agriculture, Forest Service, Forests Products Laboratory, Madison,
WI.
Schaffer, E. L. (1984). Structural fire design: wood. Madison, Wis., U.S. Dept. of
Agriculture.
Scheer, C. K., Thorsten (1994). "Brandschutz unbekleideter Holzbauteile -
Mindestquerschnitte, die einer Feuerwiderstandsklasse F30 genügen." Bautechnik
71(4): 6.
Schmid, J., A. Just, M. Klippel and M. Fragiacomo (2014). "The Reduced Cross-Section
Method for Evaluation of the Fire Resistance of Timber Members: Discussion and
Determination of the Zero-Strength Layer." Fire Technology.
Schmid, J., M. Klippel, A. Just and A. Frangi (2014). "Review and analysis of fire
resistance tests of timber members in bending, tension and compression with
respect to the Reduced Cross-Section Method." Fire Safety Journal Fire Safety
Journal, 68, 81 - 99.

330
 INTER / 48 - 16 - 01

Schmid, J., A. Menis, M. Fragiacomo and G. Bochicchio (2015). "Behaviour of loaded

cross-lamnated timber wall elements in fire conditions." Fire Technology.
Schnabl, S., G. Turk and I. Planinc (2011). "Buckling of timber columns exposed to
fire." Fire Safety Journal 46(7): 431-439.
Seekamp, H. and J. Stanke (1969). "Das Brandverhalten von belasteten Holzstützen.
(In German)." Bauen mit Holz 5, Germany.
Stanke, J., E. Klement and R. Rudolphi (1973). Das Brandverhalten von Holzstützen
unter Druckbeanspruchung. Bundesanstalt für Materialprüfung: 37, BAM-BR 024,
Berlin, Germany.
Theiler, M., A. Frangi and R. Steiger (2012). Design of timber columns based on 2nd
order structural analysis. International Council for Research and Innovation in
Building and Construction, Working Commission W18 – Timber Structures,
Meeting 45 (CIB-W18 Meeting 2012). Växjö, Sweden.
Thunell (1971). "Grading and Strength of Timber." Holz als Roh- und Werkstoff 9(29):
4.

331
INTER / 48 - 16 - 01

Discussion

The paper was presented by J Schmid

H Blass commented that if these values could be trusted would you use the mean val-
ue or 5th percentile values. J Schmid responded that he would use the mean values
and that there would be many other areas where safety could be added.
K Ranasinghe commented that in the UK they have huge difficulties in accepting the
concept that 7 mm is too high. He asked if J Schmid would be insisting to increase 7
mm in the next EC5 revision. J Schmid responded that they are aiming to build a
model to reduce the variability in the data to establish a better value. He stated that
the starting value should be ~14 mm and then one can reduce it later as variability in
the data base is reduced.
S Winter agreed that more investigation would be needed and that the old data base
has limitations. He stated that the problem of zero strength layer concept as the ac-
tual reduction curve is questionable. Comparison with FEM is very difficult as exact
properties with increase in temperature are not available. The use of 7 mm has
worked well for more than 20 years and worked well all over the world. S Winter
commented that he has never heard of a collapse before expected time of fire re-
sistance in a real fire. We need to oppose making changes to the current rules now. J
Schmid related to the example of weakness of steel connection in fire design and that
the old standard was built on one unreliable database. The main problem is missing
information on influence of temperature on timber properties if we want to replace
fire testing with models.
BJ Yeh commented on the use of 7 mm for CLT and increase of char rate to account
for falling off of the char layer. J Schmid responded that CLT is more complicated with
many producers and issues with loading directions.

332
333
334
 INTER / 48 - 19 - 01

A beam theory fracture mechanics

approach for strength analysis of
beams with a hole
Henrik Danielsson, Division of Structural Mechanics, Lund University, Sweden
Per Johan Gustafsson, Division of Structural Mechanics, Lund University, Sweden

Keywords: Glulam, Hole, Fracture Mechanics, Strength Analysis.

1 Introduction
A hole in a member constitutes a sudden change in the cross section and concen‐
trated perpendicular to grain tensile stress and shear stress appear in the vicinity of
the hole, see Figure 1.1. This stress situation may for relatively low external loads
lead to crack propagation in the fiber direction. Looking at the design approaches for
beams with a hole in timber engineering design codes over the last decades, it can be
seen that the issue has been treated in many different ways. The theoretical back‐
grounds on which the design approaches are based show fundamental differences
and there are also major discrepancies between the strength predictions according to
the different codes as well as between tests and predictions according to codes
(Aicher & Höfflin 2004, Höfflin 2005, Danielsson & Gustafsson 2008, Danielsson &
Gustafsson 2011).
There is at the moment no fully accepted design method based on a completely ra‐
tional mechanical background. There are for example no design equations for beams
with a hole in the contemporary version of Eurocode 5 (2004). However, design
equations are found in the German and Austrian National Annexes to EC 5. This de‐
sign approach originates from the work presented by Kolb and Epple (1985), although
simplifications and empirical modifications have been added over time.
+

- -

Figure 1.1. Schematic illustration of perpendicular to grain tensile stress distribution; hole placed in
shear force dominated region (left), pure bending (middle) and axially loaded member (right).

335
INTER / 48 - 19 - 01

For applications with concentrated perpendicular to grain tensile stress and shear
stress, conventional maximum stress failure criteria are seldom of any use but in‐
stead strength analysis by means of fracture mechanics is more relevant. The EC 5
design equations for end‐notched beams and dowel‐type connections loaded at an
angle to grain are examples of fracture mechanics based design equations.
Concerning the case of end‐notched beams, an equation for the energy release rate
during crack extension and the corresponding beam strength was derived by compli‐
ance analysis using beam theory almost 30 years ago (Gustafsson 1988). In its EC 5
implementation, the Gustafsson design criterion is expressed as a comparison of a
nominal shear stress and the reduced shear strength, although the decisive material
properties from the fracture mechanics approach are fracture energy and stiffness.
Inspired by the relative simplicity of the design equation for end‐notched beams, a
similar but generalized method was outlined by Gustafsson (2005). This generalized
method is in the present paper in detail developed and presented for the case of
beams with a hole. The application of a beam theory fracture mechanics approach
for beams with a hole is more complex than for end‐notched beams due to:
 the significant influence of shear makes it necessary to consider the respective
contributions of modes I and II to the total energy release rate,
 the normal force acting on the cross section parts above and below the hole must
be considered and
 the cross section forces and moments acting on the parts above and below the
hole are statically indeterminate.
In this paper, a beam theory fracture mechanics approach for strength analysis of
beams with a hole is presented. This approach is based on assumptions of an ortho‐
tropic material behavior, a beam model according to Timoshenko beam theory and a
mixed mode fracture criterion based on Linear Elastic Fracture Mechanics (LEFM).
The strength analysis approach is evaluated by means of a comparison between the‐
oretically predicted beam strengths and strengths found from experimental tests. For
this purpose, test results for beams with circular holes (Höfflin 2005, Aicher & Höfflin
2006) are used as well as test results of beams with square holes (Danielsson 2008).
The tests of beams with square holes are further also presented and discussed in
(Danielsson & Gustafsson 2008) and (Danielsson & Gustafsson 2011).
Further evaluation is carried out by means of comparison to other strength analysis
methods, including both code type methods and more general methods based on lin‐
ear and nonlinear fracture mechanics approaches carried out using 2D and 3D finite
element (FE) analysis.

336
 INTER / 48 - 19 - 01

2 Theory and model formulation

The basic concept of the present approach for analysis of beams with a hole relates
to rectangular/square holes and by assumption also circular holes are analyzed. Cross
section forces and moments in the beam parts above and below the hole are deter‐
mined by equilibrium considering kinematic assumptions according to Timoshenko
beam theory. An infinitesimally short part of the beam at the end of the hole is then
considered. The forces and moment acting across a horizontal section of the infinites‐
imally short part of the beam, dividing it into one part above and one part below an
assumed crack, are determined by equilibrium. The energy release rates for modes I
and II are then obtained by using the method of work of crack closure with considera‐
tion to the deformations of the infinitesimal parts above and below the assumed
crack. The beam strength with respect to cracking is then found by using a mixed
mode fracture criterion.
In order to facilitate a convenient formulation, which is consistent for circular and
rectangular/square holes with or without rounded corners, a number of assumptions
and simplifications are introduced. These are partly based on engineering approxima‐
tions and partly based on experience from experimental tests.
2.1 Basic assumptions and definition of geometry
Definitions and notation for the parameters used to describe the beam and the hole
geometry are found in Figure 2.1. A circular hole with diameter 2r may formally be
regarded as a rectangular/square hole with a = hd = 2r. The parameter s defines the
hole placement with respect to beam height and is given by the y‐coordinate of the
centre of the hole. N, V and M refer to values at the edge of the hole on its right hand
side, although the figure below may seem to indicate another location.
The formulation of the present approach is general in the sense that it allows for any
combination of cross sectional forces N, V and M. Depending on loading conditions
and the signs of N, V and M, different areas in the hole vicinity are exposed to per‐
pendicular to grain tensile stress and may experience cracking, see Figure 1.1. The
most common case for practical engineering purposes is however believed to be
transversal loading giving a combination of shear force and bending moment only.
a

hu
h/2 r M
y
hd
x s
h
N
h/2 hl V

b
Figure 2.1. Definition of beam and hole geometry.

337
INTER / 48 - 19 - 01

The location of the point of crack initiation is assumed to be known a priori and crack
propagation is assumed to occur in the parallel to grain direction, which is assumed
to be aligned with the beam length direction (the x‐direction). The chosen predefined
crack location is based on results from the experimental tests mentioned above of
beams loaded in bending, having either a circular or a square hole. The tests with
square holes having rounded corners (r ≈ 0.12hd) showed that cracking commonly
started at the top right corner of the hole. Crack initiation commonly occurred within
the rounded part of the hole corner. The tests with circular holes showed crack initia‐
tion at the hole periphery at an angle of approximately 45° with respect to the beam
central axis, for holes placed centrically with respect to the beam height. Based on
these findings from experimental tests, the location of the crack is in the present ap‐
proach assumed to be at an angle of 45° from the beam length direction according to
Figure 2.2 a), yielding a distance from the upper horizontal hole edge to the crack
plane of 0.3r.
The geometry of the beam parts above and below the hole are in the calculations
simplified in the sense that these are assumed to be prismatic. As illustrated in Figure
2.2 b)‐e), there are a number of feasible interpretations of the hole geometry which
result in this simplification. The interpretation used for all analyses presented here is
according to Figure 2.2 b). Interpretation of hole geometry according to Figures 2.2
c)‐d) or some other simplification may however possibly yield equal or better agree‐
ment between model strength predictions and experimentally found beam strength.

a) b)

0.3r
r r
45o assumed 45o
crack locaon

c) d) e)

Figure 2.2. Location of crack plane and possible simplifications of the hole geometry.

338
 INTER / 48 - 19 - 01

2.2 Element cross sectional forces and moments

The cross sectional forces Ni, Vi and Mi (i = 1, 2) in the beam parts above (1) and be‐
low (2) the hole are determined by equilibrium considerations of the statically inde‐
terminate system shown in Figure 2.3. Kinematic assumptions according to Timo‐
shenko beam theory are used for beam elements 1 and 2. The vertical cross sections
on the two sides of the hole are assumed to remain plane at loading. With the simpli‐
fication given in Figure 2.2 b), the beam parts above and below the hole are assumed
to have constant heights according to
0.3 (1)
0.3 (2)
To account for the elastic clamping of the beam elements in an approximate way,
these are in the calculations given a length slightly longer than the actual length of
the hole, L = a + 0.5h1, based on numerical results presented by Petersson (1974).
L

1 h1
M
e1

e2 N
2 h2 V

Figure 2.3. Definition of subelements 1 and 2 for beam parts above and below the hole respectively.
According to the kinematic assumptions, the infinitesimally thin cross section to the
right of the hole remains plane prior to crack initiation and propagation. At crack
propagation however, it is split into two cross sections according to Figure 2.4. The
cross section sectional forces Ns, Vs and Ms – forcing the cross section to remain
plane in the uncracked state – may be determined from equilibrium giving
(3)
(4)
(5)
where Vres, Nres and Mres are the resulting forces and the moment from the normal‐
and shear stresses acting on the right side of the cross section part above the crack
plane according to

(6)

6 6 (7)

(8)

339
INTER / 48 - 19 - 01

M1 V1
h1 N1 crack locaon
Vs
Ms
N Ns
h2 N2 V M central axis
M2 V2

Figure 2.4. Illustration of cross section sectional forces Ns, Vs and Ms.
2.3 Relative displacements at crack propagation
At crack initiation and propagation in the beam length direction, the original beam
cross section is split into two separate cross sections. The two respective cross sec‐
tions are still assumed to remain plane but since they are separated, they may be de‐
formed in different ways. The relative displacements u and v and the relative rota‐
tion θ of the two cross sections, above and below the crack plane, are illustrated in
Figure 2.5. The relative displacement v, related to Ns, corresponds to mode I crack
deformation. The relative displacement u and the relative rotation θ are influenced
by both Vs and Ms with u corresponding to mode II crack deformation and θ assumed
to contribute to mode I crack deformation.
dx aer crack propagaon
before crack propagaon
-θ h1
-v
crack locaon
-u
h-h1

Figure 2.5. Cross section relative displacements at crack propagation.

The relations between the relative displacements u and v and the relative rotation θ
and the cross section sectional forces and moment Ns, Vs and Ms may be expressed as
(9)
where the displacement and force vectors are given by
(10)
(11)
and the compliance matrix is given by

340
 INTER / 48 - 19 - 01

0
0 0 (12)
0
The components of the compliance matrix are determined based on the compliance
of the cross sections of infinitesimal length above and below the crack plane. Consid‐
ering kinematic assumptions according to Timoshenko beam theory gives
(13)

(14)

(15)

(16)

where Ex is the parallel to grain stiffness and Gxy is the shear stiffness.
2.4 Crack closure work, energy release rate and fracture criterion
According to LEFM, see e.g. textbook (Hellan 1985), crack propagation is governed by
crack propagation criteria which may be formulated based on, for example, the con‐
cept of strain energy release rate or stress intensity factors. The strain energy release
rate G can, according to LEFM, also be calculated in terms of crack closure work, i.e.
the work required to completely close a propagated crack. The crack closure work W
may for the present application be expressed as

(17)
, , ,

where WI and WII refer to the crack closure work related to mode I and II respec‐
tively. Indices u, v and θ refer to the relative displacements and the rotation illus‐
trated in Figure 2.5 above. According to LEFM, the stress intensity factors KI and KII
are proportional to the applied load. The principle of superposition may be used for
multiple load cases giving contributions in the same mode of deformation. For the
present application this means that the mode I stress intensity factor may be ex‐
pressed as
, , (18)
The relationship between the mode I and mode II stress intensity factors Ki and the
corresponding energy release rates Gi (i = I, II) is given by
(19)
where Ei (i = I, II) is a measure of the stiffness of the material with respect to the cor‐
responding mode of deformation (Gustafsson 2002).

341
INTER / 48 - 19 - 01

The energy release rate at crack extension over an area dA = bdx is equal to the cor‐
responding crack closure work giving for the present case
, , (20)
, , (21)
, , (22)
and
, ,
, , (23)
,
(24)
In order to determine the beam strength with respect to cracking, a crack propaga‐
tion criterion needs to be chosen. In the present applications, with in general mixed
mode of loading, the following interaction criterion has been used

1.0 (25)

where GI and GII are the energy release rates in mode I and II given above and where
GIc and GIIc are the corresponding critical energy release rates (or fracture energies).
Results presented in Sections 4 and 5, relating to analysis of beams with a hole using
the present approach, are based on m = n = 0.5 and material property parameters ac‐
cording to Table 1. The choice of crack propagation criterion is not obvious, and other
criteria could possibly be more suitable. In case of negative values of GI and/or GII,
the negative contribution is ignored and fracture in pure mode I or II is considered.
Table 1. Material property parameters used for strength analysis.
Parameter Notation Value
Parallel to grain stiffness Ex 12 000 MPa
Shear stiffness Gxy 600 MPa
Fracture energy, mode I GIc 300 Nm/m2
Fracture energy, mode II GIIc 900 Nm/m2

3 Comparison to end‐notch beam equation

The present analysis approach for beams with a hole has many features in common
with the LEFM‐based design approach for notched beams, originally presented by
Gustafsson (1988) and with modifications included in EC 5. The present approach for
strength analysis of beams with a hole may also be used for analysis of notched
beams, since this is essentially only a special case in terms of the general geometry
illustrated in Figure 2.1.
As mentioned in the introduction, there are however some differences in the general
formulation between the two approaches. For the end‐notched beam approach, no

342
 INTER / 48 - 19 - 01

distinction between modes I and II is made in terms of the crack propagation crite‐
rion and the material is characterized by the mode I fracture energy only. Another
difference relates to the effect of elastic clamping of the reduced beam parts at the
section where the notch or hole corner is located, which is considered partly differ‐
ent in the two approaches. The expression for the nominal beam strength with re‐
spect to cracking at a right angled notch derived by Gustafsson (1988) reads
⁄
(26)
. ⁄ ⁄

where Vc is the crack shear force, Anet is the net cross section area at the reduced
cross section and where α and β are defined in Figure 3.1.
Illustrations of the predicted beam strengths as influenced by the normalized beam
height α and the normalized notch length β according to the two approaches are
shown in Figure 3.1. The comparison is based on a beam of height h = 500 mm, with
GIc = GIIc = 300 Nm/m2 and values of the exponent in Equation 25 as m = n = 1. For
this choice, i.e. for GIc = GIIc and m = n = 1, is the mixed mode crack propagation crite‐
rion the same for both calculations. More or less slight differences in predicted
strengths are to be expected due to the differences in the model formulations with
respect to the influence of the elastic clamping of the reduced cross section.

2 2
Shear strength Vc/Anet [MPa]
Shear strength Vc/Anet [MPa]

αh 0.5h
h h
1.5 1.5

1 0.5h 1 βh

0.5 0.5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized beam height α Normalized notch length β
Figure 3.1. Strength according to Equation 26 (dashed black) and present approach (solid red).

4 Comparison to experimental tests

Examples of theoretically predicted strengths for beams with a hole are presented in
Figures 4.1 and 4.2, where also results of experimental tests on beams with circular
holes (Höfflin 2005, Aicher and Höfflin 2006) and with square holes (Danielsson 2008)
are shown. Theoretical beam strengths refer to the crack propagation load predicted
according the theory presented in Section 2, using Equation 25 with m = n = 0.5 and
with material data according to Table 1. The net cross section area is defined as
Anet = b(h‐hd) and the strength from experimental tests corresponds to the load at the
instant of crack propagation across the entire beam width b.

343
INTER / 48 - 19 - 01

Beam and hole geometries are for these illustrations chosen based on availability of
test results, yielding partly different geometries for the case of circular and square
holes respectively. Results presented in Figure 4.1 relate to the influence of beam
height, hole size and bending moment to shear force ratio for holes centrically placed
with respect to beam height (s = 0). Results presented in Figure 4.2 relate to the influ‐
ence of hole corner radius and hole placement with respect to beam height.
3 3
Test results Test results
Shear strength Vc/Anet [MPa]

Shear strength Vc/Anet [MPa]

2.5 a = hd = 2r = 0.3h 2.5 a = hd = h/3
M/(Vh) = 1.65 r = 0.12hd
2 2 M/(Vh) = 2.17 V

1.5 1.5

1 1

0.5 0.5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Beam height h [m] Beam height h [m]

3 3
Test results Test results
Shear strength Vc/Anet [MPa]
Shear strength Vc/Anet [MPa]

2.5 2.5 a = hd = h/3

r = 0.12hd
2 2 h = 630 mm
M(Vh) = 2.17
1.5 1.5

1 h = 450 mm 1
M/(Vh) = 1.65 ± 0.05
0.5 0.5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized hole height hd/h [−] Normalized hole height hd/h [−]

3 3
Test results Test results
Shear strength Vc/Anet [MPa]

Shear strength Vc/Anet [MPa]

2.5 a = hd = 0.3h 2.5 a = hd = h/3

h = 450 mm r = 0.12hd
2 2 h = 630 mm

1.5 1.5

1 1

0.5 0.5

0 0
0 2 4 6 8 0 2 4 6
Moment to shear force ratio M/(Vh) [−] Moment to shear force ratio M/(Vh) [−]

Figure 4.1. Experimentally found strength and theoretically predicted strength as influenced by
beam height, hole size and bending moment to shear force ratio. Left (in red): Circular holes (Höfflin
2005, Aicher & Höfflin 2006). Right (in blue): Square holes (Danielsson 2008).

344
 INTER / 48 - 19 - 01

3 3
Test results Test results, h = 180 mm
Shear strength Vc/Anet [MPa]

Shear strength Vc/Anet [Pa]

2.5 2.5 Test results, h = 630 mm

2 2

1.5 1.5

1 a = hd = 0.3h 1
h = 450 mm a = hd = h/3
0.5 M/(Vh) = 1.65 0.5 r = 0.12hd
s=0 M/(Vh) = 2.17
0 0
0 0.1 0.2 0.3 0.4 0.5 −0.2 −0.1 0 0.1 0.2
Normalized hole corner radius r/hd [−] Hole placement w.r.t beam height s/h [−]

Figure 4.2. Experimentally found strength and theoretically predicted strength as influenced by hole
corner radius and hole placement with respect to beam height. Left (in red): Circular holes (Höfflin
2005, Aicher & Höfflin 2006). Right (in green/blue): Square holes (Danielsson 2008).

5 Comparison of strength analysis approaches

An overview of the ratio between experimentally found beam capacity and theoreti‐
cally predicted capacity according to some different approaches for strength analysis
is given in Figure 5.1. The experimental tests used for this comparison consist of
beams with either circular holes (Höfflin 2005, Aicher & Höfflin 2006) or square holes
(Danielsson 2008) also used for comparison in Section 4.
The strength analysis methods included are:

1. PFM – A Probabilistic Fracture Mechanics method based on 2D FE‐analysis and

consideration of fracture ductility according to a generalized LEFM approach in
combination with consideration of material variability according to Weibull theory
(Danielsson & Gustafsson 2011).
2. Weibull – Classical Weibull theory based on 2D FE‐analysis (Danielsson & Gus‐
tafsson 2011).
3. NLFM – A nonlinear fracture mechanics approach (a cohesive zone model) based
on 3D FE‐analysis (Danielsson & Gustafsson 2014).
4. Present approach – see previous sections.
5. Glulam Handbook – end‐notched beam analogy approach for beams with a hole
found in the old version of the Swedish Glulam Handbook (Carling 2001) and also
included in a draft version of EC 5 (2002).
6. DIN EN 1995‐1‐1/NA – Semi‐empirical approach found in German and Austrian
National Annexes to EC5.
7. Aicher et al – Design approach based on Weibull theory presented by Aicher, Höf‐
flin and Reinhardt (2007). Originally suggested to be used for circular holes only
but here used also for square holes assuming a = hd = 2r.

345
INTER / 48 - 19 - 01

Theoretical capacity / Experimental capacity [−]

3.00 ( )
2.50
2.00

1.50 ( ) ( )
( )
( )
( )
1.00

0.67

0.50
0.40
FE−based methods code type methods
0.33
1. PFM 2. Weibull 3. NLFM 4. Present 5. Glulam 6. DIN EN 7. Aicher
approach Handbook 1995−1−1/NA et al

Figure 5.1. Comparison of experimentally found strengths and theoretically predicted strengths
according to some different approaches for strenth analysis of beams with a hole. The test result
indicated as (o) appears abnormal without known cause; see further comments in (Danielsson &
Gustafsson 2011).
For the general approaches (items 1‐4 above), experimentally found mean values are
compared to predicted strength based on assumed mean values of material property
parameters. For the NLFM approach (item 3 above), only results for the beams with
square holes are presented since analysis of beams with circular holes were not in‐
cluded in the work presented in (Danielsson and Gustafsson 2014).
The strength according to the three code type approaches (items 5‐7 above) are
based on characteristic material strength values fvk = 3.5 MPa and ft90k = 0.5 MPa,
valid for strength class GL 32h according to SS‐EN 14080:2013. The predicted
strengths are for these methods compared to characteristic values of the experimen‐
tally found strengths; see (Danielsson & Gustafsson 2011) for further information.
All test results included in the comparison relate to holes with larger height hd than
allowed to be used without reinforcement according to the German and Austrian Na‐
tional Annexes to EC 5.

6 Discussion
The presented approach is general in the sense that it allows for strength analysis in‐
cluding not only loading in terms of shear force and bending moment, but also axial
force. Depending on relative load levels and the sign of the axial force, the assumed
location of the crack (see Figure 2.2) may however need to be adjusted.

346
 INTER / 48 - 19 - 01

Beams with holes are in practical design situations often reinforced. The reason for
this is likely related to two different factors: (i) the actual strength reduction due to
the hole and (ii) the uncertainty and lack of knowledge related to strength analysis
and design of beams with a hole. The cross section sectional forces Ns, Vs and Ms (see
Equations 3‐5) could possibly be of interest in relation to design of internal or exter‐
nal beam reinforcement in terms of fully threaded screws, glued‐in rods or glued‐on
panels.
A new version of the Swedish Glulam Handbook is planned to be released during
2015, within which the end‐notch beam analogy approach for beams with a hole will
be removed in favour of the design approached found in the German and Austrian
National Annexes to EC 5.

7 Conclusions
From the work presented in this paper, the following conclusions are drawn:
 The present approach is consistent with the LEFM‐based design approach for end‐
notched beams found in EC 5.
 The present approach appears to be able to capture the strong beam size influ‐
ence found from experimental tests of beams with a hole fairly well.
 The present approach appears fairly accurate in capturing the absolute values of
the beam strength for both circular and square holes of various sizes and locations.
Although based on beam theory analysis, the present approach is not suitable for di‐
rect incorporation into timber engineering design codes of practice in its current
form, because of rather complex equations. For certain applications and load configu‐
rations, more user‐friendly design expressions may possibly be derived and could
then serve as a base for improved design recommendations. Further work regarding
verification and calibration is however needed before this can be realized.

8 References
Aicher S, Höfflin L (2004): New design model for round holes in glulam beams. In:
Proceedings of 8th World Conference on Timber Engineering, vol 1, Lahti, Finland.
Aicher S, Höfflin L (2006): Tragfähigkeit und Bemessung von Brettschichtholzträgern
mit runden Durchbrüchen – Sicherheitsrelevante Modifikationen der Bemes‐
sungsverfahren nach Eurocode 5 und DIN 1052. Materialprüfungsanstalt (Otto‐
Graf‐Institute), University of Stuttgart, Germany.
Aicher S, Höfflin L, Reinhardt HW (2007): Runde Durchbrüche in Biegeträgern aus
Brettschichtholz. Teil 2. Tragfähigkeit und Bemessung. Bautechnik 84(12):867‐880.
Carling O (2001): Limträhandbok (Glulam handbook). Svenskt Limträ AB, Print & Me‐
dia Center i Sundsvall AB, Sweden.

347
INTER / 48 - 19 - 01

Danielsson H (2008): Strength tests of glulam beams with quadratic holes – Test re‐
port. Report TVSM‐7153, Structural Mechanics, Lund University, Sweden
Danielsson H, Gustafsson PJ (2008): Strength of glulam beams with holes – Test of
quadratic holes and literature test result compilation. In: Proceedings of CIB‐W18
Meeting 41, St Andrews, Canada, Paper no. CIB‐W18/41‐12‐4.
Danielsson H, Gustafsson PJ (2011): A probabilistic fracture mechanics approach and
strength analysis of glulam beams with holes. Eur J Wood Prod 69:407‐419.
Danielsson H, Gustafsson PJ (2014): Fracture analysis of glued laminated timber
beams with a hole using a 3D cohesive zone model. Engng Fract Mech 124‐
125:182‐195.
DIN EN 1995‐1‐1/NA:2013‐08 (2013): German National Annex to EC 5
EN 14080:2013 (2013): Timber structures – Glued laminated timber and glued timber
– Requirements. CEN.
Eurocode 5 (2004): Design of timber structures ‐ Part 1‐1: General – Common rules
and rules for buildings. CEN. (EN 1995‐1‐1).
Eurocode 5 (2002): Design of timber structures ‐ Part 1‐1: General – Common rules
and rules for buildings. Final draft 2002‐10‐09. (prEN 1995‐1‐1).
Gustafsson PJ (1988): A study of strength of notched beams. In: Proceedings of CIB‐
W18 Meeting 21, Parksville, Canada, Paper no. CIB‐W18/21‐10‐1.
Gustafsson PJ (2002): Mean stress approach and initial crack approach. In: Aicher S,
Gustafsson PJ (ed) Haller P, Petersson H: Fracture mechanics models for strength
analysis of timber beams with a hole or a notch – a report of RILEM TC‐133. Report
TVSM‐7134, Structural Mechanics, Lund University, Sweden.
Gustafsson PJ (2005): Mixed mode energy release rate by a beam theory applied to
timber beams with a hole. In: Abstract Book of 11th International Conference on
Fracture, Turin, Italy.
Hellan K (1985): Introduction to fracture mechanics. International Edition. McGraw‐
Hill.
Höfflin L (2005): Runde Durchbrüche in Brettschichtholzträger – Experimentelle und
theoretische Untersuchungen. PhD thesis, Materialprüfungsanstalt (Otto‐Graf‐In‐
stitute), University of Stuttgart, Germany.
Kolb H, Epple A (1985): Verstärkung von durchbrochenen Brettschichtholzbindern.
Vorschungsvorhaben I.4‐34810, Forschungs‐ und Materialprüfungsanstalt Baden‐
Württemberg, Germany.
Petersson H (1974): Analysis of loadbearing walls in multistorey buildings: stresses
and displacements calculated by a continuum method. PhD thesis, Chalmers Uni‐
versity of Technology, Sweden.
ÖNORM B 1995‐1‐1:2014 (2014): Austrian National Annex to EC 5.

348
 INTER / 48 - 19 - 01

Discussion

The paper was presented by H Danielsson

H Blass commented that the stress distribution shown in slide 8 close to the hole is
very different from reality. He questioned why the results are so good. H Danielsson
agreed but stated that the beam theory is exact for normal stresses but might be dif-
ferent for shear stresses. PJ Gustafsson added that the solution should be exact in
terms of energy release rate. As the crack propagated, the normal forces due to the
moment should be exact but shear would not be true.
BJ Yeh asked how close the hole can be to the support and if there were two holes,
how close can they be. H Danielsson responded the distances should be such that the
support would not influence the hole. This would also apply to the multiple holes cas-
es.
K Malo asked if there would be a limitation to the hole size. H Danielsson said that
there would be no limit and the analysis would also work for notched beams.
I Smith commented that good results need accurate analysis, real structure, and accu-
rate material properties; when results do not agree, maybe some of these are not
working.
P Dietsch and H Danielsson discussed the cases of round and square holes and limita-
tions to the model.
R Jockwer received clarifications that at this moment there is no recommendation for
the size of the hole without reinforcement.

349
350
 INTER / 48 - 19 - 02

A strongest link model applied to fracture

propagating along grain
Per Johan Gustafsson, Division of Structural Mechanics, Lund University, Sweden
Robert Jockwer, Institute of Structural Engineering, ETH Zurich, Switzerland
Erik Serrano, Division of Structural Mechanics, Lund University, Sweden
René Steiger, Swiss Federal Laboratories for Materials Science and Technology EMPA,
Dübendorf, Switzerland

Keywords: Fracture, perpendicular to grain, knot, heterogeneity, notch, modelling,

size effect

Abstract: The impact of knots and grain deviations on the perpendicular to grain
strength of wood was studied by computational models and test results from
literature. It was found that knots in general seem to have a positive influence on the
perpendicular to grain fracture strength. A strongest link model for strength analysis
relating to failure due crack propagation along grain is discussed, taking into account
the influence of random strong spots along the crack path.

1 Introduction
Wood is a markedly heterogeneous material containing knots and grain deviations
that strongly affect structural strength (Jockwer, 2014). This heterogeneity is the
reason for the interest in application of strength models like the weakest link model
(Weibull, 1939) and for the discussion in this paper about a strongest link model. One
of the most important implications of the heterogeneities in wood is the size effect in
load-carrying capacity as limited by tensile and/or shear failure perpendicular to
grain. The influence of heterogeneities on this kind of fracture is the topic of this
paper, in particular in relation to fracture starting at a local high stress region.
For cases of uniform or fairly uniform stress distribution in brittle materials it is
reasonable to assume that structural failure is governed by the weakest part or link of
the material. The weakest link concept agrees well with compilations of test results
found in literature showing a significant size effect in strength of wood loaded in
homogenous tension perpendicular to grain, see e.g. (Barret, 1974), (Barett et al.
1975), (Ehlbeck and Kürth, 1991), (Mistler, 1998), (Aicher et al., 2002) and
(Gustafsson, 2003).

351
INTER / 48 - 19 - 02

The situation may however be very different for materials with significant fracture
toughness and for non-homogeneously stressed and/or statically indeterminate
structures. In these cases failure typically involves progressive damage or crack
propagation and one or many stress redistributions during the course of fracture
events before the situation corresponding to maximum load-carrying capacity of the
structure is reached. This may imply a better structural robustness since the ultimate
load-carrying capacity is instead determined by the strongest link in a chain of states
and stress distributions before complete structural damage. This general type of
analysis is in the present study applied to beam failure caused by crack propagation
along grain, taking into account random variation of the fracture toughness in the
wood along the crack propagation path. As an introduction to the analysis of crack
propagation failure, this paper also aims at investigating the impact of some different
growth characteristics and heterogeneities on the fracture behavior of timber.

2 Impact of growth characteristics – some

numerical and experimental results
2.1 Impact of a knot oriented out of the plane of fracture
A knot which is oriented out of the plane of fracture can give a very significant local
contribution to the perpendicular to grain strength and fracture energy. This is
illustrated in Figure 1 by a test result from (Jockwer, 2014) showing recorded load vs
displacement for Single Edge Notched Beam specimens (SENB, (Larsen and
Gustafsson, 1990), (Nordtest, 1993)) without a knot and for a specimen with a knot.
The specimen with a knot has about twice the bending moment capacity and the
fracture energy is increased by a factor of ten or more.

Figure 1. Load vs
displacement for SENB
specimens without and
with a knot.

352
 INTER / 48 - 19 - 02

2.2 Impact of grain deviations around a knot on fracture energy

The deviation in crack propagation path in the vicinity of a knot causes an influence of
a knot also when its axis lies within the fracture plane. Such a grain deviation is
illustrated in Figure 2. The impact of this grain deviation on the load-deflection
behavior and fracture energy was analyzed numerically for a SENB specimen by
means of the FE-software Abaqus, using linear elastic fracture mechanics and an
enriched FE-method, xFEM, described in (Qiu et al., 2014). The calculated apparent
fracture energy versus knot size is shown in Figure 2.

crack r

32 mm

Figure 2. Impact of grain deviation at a knot on fracture energy Gf,I.

2.3 Impact of cross grain

The magnitude of the cross grain (see Figure 3) was quantified by means of the
results from fiber orientation scanning of 450 laminations reported by Oscarsson et
al. (2014). The median of the nominal grain direction on the edges of the laminations
was used as a measure in order not to have the data corrupted by the local grain
deviation close to knots. A median deviation from perfectly aligned grain in the range
of up to 2 degrees was found to be quite common and the 10% fractile values of the
median included deviations of approximately 4-6 degrees.
Figure 3 shows the influence of grain orientation in an end-notched beam as found
by means of linear elastic fracture mechanics applied by the compliance method,
using Matlab for plane stress FE-calculations. From the results it is evident that even
a small grain inclination may strongly affect both the maximum load-carrying capacity
and the development of the load causing crack propagation.

353
INTER / 48 - 19 - 02

400
600
γ
P

150 x [mm]

Figure 3. Beam cross grain and development of load causing crack propagation.

3 End-notched beam with random knots analyzed

by means of Monte Carlo simulations
The strength of end-notched beams with geometrical shaping according to Figure 3
and with horizontal grain orientation was studied with regard to the influence of
random material properties and knots. Monte Carlo simulations using an explicit
strength equation for end-notched beams were carried out. The strength equation is
based on linear elastic fracture mechanics and beam theory (Gustafsson, 1988) and
for the present application is written as (Jockwer, 2014):
3Vf 3.75 G f Gv
 (1)
2bαb h  α  α 2  β 10 Gv /E0 1/α  α 2 
 
The left hand term of the equation represents the shear stress in the net beam cross-
section at start of crack propagation, where b is the beam width and Vf is the shear
force. Gv and Eo is the shear stiffness and the modulus of elasticity parallel to the
grain, respectively, and Gf is the fracture energy for the current mixed mode state of
stress at the tip of the notch. A reasonable approximation is to assign Gf the mode 1
fracture energy. In the Monte Carlo calculations, stochastic values of Gv , Eo and Gf
=Gf,I according to Table 1 were used, unless otherwise specified. It is to be noted that
the fracture energy values indicated in Table 1 relate to clear wood while the stiffness
parameters relate to glulam including growth characteristics and inhomogeneities.
Eq. (1) was used also to evaluate the wood fracture mechanics strength parameter
(3.75 Gf Gv)0.5 for a large number of end-notched beam strength test results compiled
from literature (Jockwer, 2014). The result of this compilation of test results is
indicated by circles in Figure 4. The numbers for (3.75 Gf Gv)0.5 given in the figure are

354
 INTER / 48 - 19 - 02

thus a kind of normalized value of the failure load Vf of the notched beams assuming
that Gv/E0=650/11500≈1/18 and that Eq. (1) is valid.

Table 1. Distribution parameters of material properties relating to glulam equivalent to strength

class GL24h at MC = 12% acc. to EN 14080, (JCSS, 2001) and (Jockwer, 2014).
Parameter Unit Symbol Mean 5th perc. CoV PDF
MOE II to the grain [N/mm2] E0,mean 11’500 9’600 13 % Lognormal
MOE ┴ to the grain [N/mm2] E90,mean 300 250 13 % Lognormal
Shear modulus [N/mm2] Gv,mean 650 540 13 % Lognormal
Fracture energy Mode 1 * [N/mm] Gf,I,mean 0.3 0.218 20 % Lognormal
Fracture energy Mode 2 * [N/mm] Gf,II,mean 1.15 0.695 30 % lognormal
* clear wood

In a first Monte Carlo simulation it was assumed that (1) the stochastic values of the
fracture energy were equal to those for clear wood, see Table 1, and (2) that the
ultimate failure load equaled the load at the beginning of crack propagation. The
result from this simulation is depicted with a dash-dotted curve in Figure 4.
The simulation was then improved by considering the influence of knots and the
variation in fracture resistance along the crack propagation path. In these calculations
clear wood sections along the crack propagation path were assumed to be enclosed
by intermediate knot or knot cluster sections. Values for the lengths of clear wood
parts and knot parts were taken from (Fink et al., 2013). Hence, for Norway spruce
(picea abies) wood the knot parts were assigned a constant length of 150 mm and
the clear wood parts a stochastic length following a Gamma-distribution with a mean
value 530 mm and a standard deviation 250 mm. The end-notch was assigned a
random location along the wood member. Within each clear wood part, the fracture
energy was assigned a constant value taken from a lognormal distribution with CoV
20%. A bias of the mean value was used to fit the test results. Within each knot part,
the fracture energy was assigned a constant value taken as the mean value of the
fracture energy for the clear wood times a lognormal distributed stochastic knot
factor Fknot. The best fit to the test results was found by reducing the mean fracture
energy of the clear wood, as compared to the value used in the reference calculation
from 0.3 N/mm to about 0.2 N/mm (corresponding to a 20% reduction in the
strength parameter (3.75 Gf Gv)0.5 for the clear wood from 27 N/mm3/2 to 22
N/mm3/2) and at the same time setting the mean value and CoV of Fknot equal to 2.0
and 40 %, respectively. The 20% strength reduction is in line with studies of e.g.
Franke (2008) and Jockwer (2014). The resulting failure probability of this Monte
Carlo simulation is depicted with the solid line in Figure 4.
Additional simulations showed that the mean value of the distance between knots
has a strong impact on the distribution and magnitude of beam strength, while the

355
INTER / 48 - 19 - 02

CoV of the inter knot distance is of minor importance. It was also found that the value
of Fknot has a strong impact on beam strength.
The general trend found in the simulations is that the presence of knots leads to an
increase in load-carrying capacity of notched beams. This is in contradiction to codes
and procedures where only weakening effects of knot sections are accounted for, e.g.
with respect to bending strength.

Figure 4. Cumulative strength distribution of test results compared to numerical calculations

accounting for different growth characteristics.

4 End-notched beam with knots – additional

observations
Early experimental research on the impact of knots on the perpendicular to grain
fracture strength of timber was conducted by Larsen and Riberholt (1972). 200 end-
notched solid softwood beams of quality "ungraded" and with varying height of the
reduced cross-section αh were subjected to 3-point bending tests. Beam cross
section height was h = 125 mm and the beams failed due to crack propagation
starting at the notch. The recorded failure loads are summarized in Table 2. The
failure load of the beams with one or more knots in the failure region of the beam
was for all three αh tested found to be higher both at the mean value level and at the
10th– and 5th– percentile levels compared to beams lacking of knots. The CoV in
strength for beams with knots was in general not higher than for those without
knots.
About 20 years later, Riberholt et al. (1991) tested a large series of end-notched
beams in order to study the influence of various geometrical parameters. A crack
retarding effect of knots was again observed and specimens with knots showed

356
 INTER / 48 - 19 - 02

higher load-carrying capacities. To determine fracture energy also small SENB

specimens were tested and it was noted that a specimen with a knot had a
considerably higher fracture energy than the clear wood specimens.

Table 2. Shear stresses at failure τu in the reduced cross-section of beams with

variable notch height tested in 3-point bending, separated into samples with and
without knots (Larsen and Riberholt, 1972).
α τu,mean (CoV) τu,0.05 τu,0.10
[-] [N/mm2] (%) [N/mm2] [N/mm2]
With knots 0.75 3.63 (30.6) 1.78 2.04
0.5 2.41 (20.9) 1.55 1.71
0.25 2.77 (31.6) 1.47 1.70
Without knots 0.75 2.72 (24.1) 1.75 1.92
0.5 2.12 (27.9) 1.32 1.45
0.25 1.78 (31.8) 1.02 1.15
All 0.75 2.93 (29.6) 1.70 1.90
0.5 2.19 (26.5) 1.36 1.51
0.25 2.00 (38.0) 1.02 1.17

Similar impacts of knots on the load-carrying capacity of end-notched beams were

detected in tests reported by Möhler and Mistler (1978). A reduction of the load-
carrying capacity was on the other hand observed for beams with checks along the
crack propagation path.
Jockwer (2014) showed that the large variation in load-carrying capacity of end-
notched beams cannot be explained only by the variation of the elastic stiffness
properties and fracture energy of timber and clear wood, respectively. The high level
of variation could be explained by the presence of knots along the crack path. The
crack retarding effect of knots was studied by means of optical measurement
systems.

5 Strongest link model

5.1 Background
Perpendicular to grain failure of a structural element is initiated at some location and
then followed by crack propagation along a potential fracture surface ending in
ultimate failure. To analyse the load-carrying capacity of the structural element, in
principle all possible fracture initiation locations and their corresponding crack paths
must be investigated with respect to the load which initiates fracture and the load
needed to propagate the crack along the fracture surface. For each initiation point,
the maximum load required for assuring crack propagation until final failure must be
determined. For all these maximum loads their stochastic character has to be
identified and the minimum of these maxima must be determined. A complete

357
INTER / 48 - 19 - 02

analysis is demanding and probably is rarely needed. Instead special cases can be
defined and analysed.
Such a special case is at hand if the fracture initiation load is greater or equal to the
maximum of the load needed for crack propagation along the crack path. In that case
the weakest link analysis of Weibull (1939) illustrated in Figure 5a) can be applied, the
links in the chain representing the different possible locations of the failure initiation
point.
Another special case is the situation where only one fracture initiation point together
with the corresponding crack path needs to be considered. This is typically the case
for failure initiated at a notch and can be analysed by means of a strongest link
model. In this case the links represents the chain of crack tip locations along the crack
propagation path. Strongest link modelling is illustrated in Figure 5b): the links are
not loaded at the same time and they must all fail before global failure occurs. It is
thus neither parallel coupling nor series coupling. One may instead compare the
crack propagation with opening a zipper: the zipper link which provides the highest
resistance is decisive.
In order to enable comparisons first some relations of the weakest link model shall be
summarised. The probability of material failure of a “small” unit volume ΔV or dV of
material is in Weibull weakest link model commonly described by the 2-parameter
Weibull distribution:

pf σ   1  e σ/f c 
m
(2)
pf is the probability that failure will occur at some stress lower than σ. σ is the
magnitude of stress in the homogeneously stressed unit volume and fc and m are
material parameters which define magnitude and scatter in strength. If not uniaxial
stressed, then σ and fc represent an effective stress and strength, respectively, for
the state of stress under consideration. fc can be determined as the 0.63 fractile
value of tested failure values of σ and m is uniquely related to the CoV.
From Eq. 2 it is evident that the probability that the unit volume does not fail is
exp(-(σ/fc)m). The probability that no link in a chain made up of n links fails is then
(exp(-(σ/fc)m))n = exp(-n(σ/fc)m) and accordingly the probability of failure of the chain
is

pf σ   1  e n σ/fc 
m
(3)
This relation also estimates the probably of failure initiation in a structure made up of
n equally stressed unit volumes. If the unit volumes are not equally stressed, then
n
  Pσ 0i /f c m
n
pf P   1  Π e

 Pσ 0i /f c 
m
1 e i 1 (4)
i 1

358
 INTER / 48 - 19 - 02

P is the load and σ0i is the stress in unit volume i when P = 1. The basic strength
distribution Eq. 2 and the corresponding lowest-value extreme value distribution Eq.
3 or 4 exhibit the same curve shape and the same CoV. This is a very convenient
feature of the 2-parameter Weibull distribution. The sum in Eq. 4 can be replaced by
an integral if the “small” unit material volume is taken as dV. Some finite volume Vo
must then be introduced as reference volume for the material parameters fc and m.
The conventional Weibull weakest link theory with volume-integration is not
applicable to structural elements with a sharp notch, crack or other shaping that
reveals a stress singularity since the theory for such situations in general predicts
either zero strength or no failure at the crack tip, no matter the magnitude of load
(Gustafsson and Enquist, 1988). Mistler (1998) did, however, combine a previously
developed wire-model for analysis of crack growth starting from the tip of a notch in
wood (Mistler, 1979) with weakest link analysis in order to analyse the size effect in
the perpendicular to grain tensile strength of structural elements like curved and
tapered beams.

a) b)

Figure 5. a) Weakest link model. b) Strongest link model.

5.2 A strongest link model

A strongest link model for crack propagation along grain and other progressive failure
is considered (Gustafsson, 2014). The model is schematically illustrated in Figure 5b).
If as an example the strength distribution of a single link can be described with a 2-
parameter Weibull distribution according to Eq. 2, then, - since failure of the zipper in
Figure 5 b) requires failure of all links -, the strongest link strength distribution of a
zipper with n links is
n
pf σ   1  e σ/f c  
 m
(5)
 
The weakest link correspondence to Eq. 5 is Eq. 3. Commonly the first link would
correspond to fracture initiation and the following links to crack propagation. If
fracture starts at a crack or a notch, then both initiation and propagation can be
analyzed applying fracture mechanics theory. If there is no initial crack or notch, then
the strength of the first link may be determined by a conventional stress based
criterion.
In crack propagation analysis the strength fc of a link is replaced by the fracture
toughness Kc of the link, being a measure of the magnitude of the fracture toughness
of the material as determined for crack propagation along a “short” reference length

359
INTER / 48 - 19 - 02

Δx. The parameter m is a measure of the scatter in fracture toughness. The numerical
values of the material parameters Kc and m corresponding to reference length Δx can
be determined experimentally by fitting experimental and theoretical strength
distribution curves. σ in the ratio σ/fc represents the stress intensity K. K = PK0(x) can
be determined by conventional linear elastic fracture mechanics analysis, P being the
external load, K0 the stress intensity when P = 1 and x the length of the crack. If K is
constant along the crack path and if the length of the path is L = n Δx, then:
n
pf P   1  e PK 0 /K c  
 m
(6)
 
If the stress intensity PK0 varies with the location x along the crack propagation length
L, then
n 
pf P    1  e (PK 0 )i /K c  
m 
(7)
i 1  
where (PK0)i = PK0(xi) and xi = (i-0.5)Δx.
The fracture toughness of a material can be represented by the square root of a
measure of the elastic stiffness times the critical energy release rate Gc or fracture
energy Gf, e.g. by Kc = (GvGf)0.5. Similarly, PK0 is represented by (GvG)0.5. The
calculation of G is simplified if all variation in Kc along the crack path is attributed to
variation in Gf and thus taking Gv as constant. With fracture toughness replaced by
fracture energy, Eq. 7 can be written
n
pf P    1  e G i /G f 
 m/2 
 (8)
i 1  
In the above quasi-static loading conditions have tacitly been assumed. In more
accurate analysis the possibility of accumulation of kinetic energy during crack
propagation should be considered.

5.3 Application to end-notched beams

Failure probability results from an application of the strongest link model to end-
notched beams are shown in Figure 6, also showing a comparison with test data from
literature (Jockwer, 2014). The calculation of G for different loads P = V and crack
propagation lengths x were carried out by Eq. 1, giving
2
0.6  P h  2   β  x  10Gv 1/α  α 2 

G(P, x)   α  α   (9)
Gv  bα h   h  E0 

Knowing G(P,x), Eq. 8 was used, thus assuming a stochastic distribution of fracture
energy according to the 2-parameter distribution of Weibull. This makes the
calculations simple, but some other distribution that would imply more

360
 INTER / 48 - 19 - 02

comprehensive numerical calculations would probably give a better representation of

the stochastic distribution due to presence of knots. The calculations were carried
out for a beam with notch geometry α = 0.6, βh = 120 mm and h = 600 mm as
defined in Figure 3, while the test results used for comparison relate to beams with
different size and shape, assuming that Eq. 1 properly corrects for different
geometrical shaping of the end-notch. The material parameters from Table 1 were
used with deterministic E0 = 11500 N/mm2 and Gv = 650 N/mm2, and stochastic
Gc = Gf,I with magnitude parameter Gf in Eq. 8 made equal to 0.3 N/mm and with
different values of scatter parameter m. The total crack propagation length studied
was throughout L = 1000 mm.
A good fit of the model with the test data was achieved for m = 3.5 and n = 8 links.
This corresponds to a crack propagation length of each link of Δx = 125 mm. For an
increasing number of links at a constant total crack length both the 5th– and the 50th–
percentile values increase, see Figure 6a). In contrast, the model parameter m has a
major impact on the progress of the upper tail of the probability density, see Figure
6b). The validity of Δx as a reference length for other fracture mechanics problems
has to be evaluated more extensively. Basically Δx should be taken as the reference
length for which the material parameters for stiffness and fracture energy (i.e. the
fracture toughness) were determined.

a) b)

Figure 6. Cumulative probability of failure of end-notched beam strength: numerical predictions and
test results.

5.4 Application to connections loaded perpendicular to grain

A further application example relates to a beam with a dowel loaded perpendicular
to the beam, creating a risk for cracking along the beam, see Figure 7. For this
situation, according to Jensen (2003) the energy release rate during crack
propagation is:

361
INTER / 48 - 19 - 02

0.6  P 2  Gv  x 2 
  (1  α ) 
G P, x   3
  1  α   2.5 (10)
Gv αh  2b   E0  αh  

The result of application of Eq. 8 with G from Eq. 10 and a comparison with test data
from literature (Jockwer et al., 2015) is shown in Figure 7. The same material data as
for the application in Section 5.3 were used. The crack propagation length analysed
was L = 1800 mm. A good fit between the theoretical result and the test data was
found for m = 3 and with n = 3 links, shown by the solid line in Figure 7.
L L (1-α)h
αh
P

a) b)
Figure 7. Cumulative probability of failure of test results for a beam with a connection perpendicular
to the grain compared with numerical results for various numbers of links in a) and for various
scatter in fracture energy in b).

5.5 A size effect comparison between the strongest and weakest link models
The strongest link model predicts increased or constant structural strength with
increased crack propagation length and with increased scatter in material fracture
toughness. The weakest link model predicts on the contrary decreased structural
strength with increased volume and with increased scatter in material strength.
The magnitude of these influences of size and scatter is different for different
probability distribution functions. Here the effects obtained when using the 2-
parameter Weibull distribution are studied. The volume and scatter effects predicted
by conventional Weibull analysis, see e.g. (Danielsson, 2013), is
1/m
f A  ΩA 
 
f B  ΩB  (11)

362
 INTER / 48 - 19 - 02

fA and fB is the stress in some reference point in equally shaped specimens A and B at
loads corresponding to failure, e.g. the median failure loads. Ω indicates volume and
m is the measure of scatter in strength. Theoretically m is related to CoV, but is often
determined by fitting of Eq. 11 to the size effect test results. For wood in
homogeneous tension perpendicular to grain and using size effect tests, m is typically
found to be about 5, theoretically corresponding to CoV of about 20-25 %.
The crack propagation length effect predicted by the strongest link model is more
complex and here studied for the case illustrated in Figure 8. For this geometry and
loading the stress intensity K is constant along the crack path. Parameters Kc, m and
Δx are assumed to be equal for specimens A and B. Eq. 6 gives
1/n 1/m
M A  ln(1  pf A ) 

M B  ln(1  p 1/nB ) 
 f  (12)
The length effect is thus different for different failure fractiles, pf. A numerical
example: for pf = 0.5, nA = 4, nB = 2 and m = 5 is found that MA/MB = 1.08 while for
pf = 0.05, nA = 4, nB = 2 and m = 5 is found that MA/MB = 1.20. If the scatter is
increased to the equivalent of m = 3 all other conditions being unchanged then is
found MA/MB = 1.14 and 1.36 for pf = 0.5 and pf = 0.05, respectively. The weakest link
model gives with ΩA/ΩB = 2 that fA/fB = 0.87 for m = 5 and that fA/fB = 0.79 for m = 3.
With respect to the influence of the height h of the specimen in Figure 8 is an effect
found according to fracture mechanics theory. Analysis of the influence of beam
width does not enter the present 2D analysis. 3D analysis would presumably indicate
some decrease in beam strength with increased width due to less scatter in the
fracture toughness along the beam.

MA MB
A h B h
MA MB
nAΔx nBΔx

Figure 8. DCB-specimens with different crack path length.

6 Summary with remarks relating to design

The impact of some heterogeneities in wood from growth characteristics of the tree
on the strength of timber in structural size was studied by computational models and
test results from literature. It was found that the heterogeneities in timber due to
knots and associated grain orientation disturbances in general seem to have a
positive influence on the perpendicular to grain fracture strength.
The strongest link model discussed relates to fracture propagation along a crack path
starting at a given point or, more generally, to the sequence of events for a given

363
INTER / 48 - 19 - 02

failure mode. The weakest link model relates to analysis of different possible fracture
locations or, more generally, to different possible modes of failure.
Contemporary grading specifications relate primarily to loading situations resulting in
stress parallel to the grain, e.g. bending of a beam. It is then tacitly assumed that the
strength perpendicular to the grain correlates with the strength parallel to grain. This
assumption is questionable. An example: a knot commonly gives a weak spot with
respect to bending, but as discussed in the above, commonly a reinforcing spot with
respect to fracture perpendicular to grain. Another example: a resin pocket or a
check along grain may be devastating with respect to tensile strength perpendicular
to grain, but may hardly have any significant influence on the bending strength. It can
thus be difficult to find a reliable strength grading correlation, neither positive nor
negative. It might therefore be appropriate to consider the development of a 2-
parameter grading system. And the more so as fracture perpendicular to grain seems
to be one of the more frequent causes of structural failure (Frühwald et al., 2007).
Research efforts on size effects have so far mainly been dealing with the influence of
random weak spots in the wood. The more or less random presence of strong spots
such as knots may, however, also be of matter in some cases, both in relation to size
effects and to magnitude of structural strength. A strong spot effect may be found for
crack propagation along grain and analyzed by a strongest link model. Such modelling
might be useful also more generally in relation to analysis of robustness and failure
mechanisms involving a sequence of events.

7 References
Aicher S, Dill-Langer G, Klöck W (2002): Evaluation of different size effect models for tension
perpendicular to grain strength of glulam”. Proc. of CIB-W18 Meeting 35, Paper 35-6-1.
Barrett JD (1974): Effect of size on tension perpendicular-to-grain strength of Douglas-fir.
Wood and Fiber Science, 6(2), 126-143.
Barrett JD, Foschi RO, Fox SP (1975): Perpendicular-to-grain strength of Douglas-fir.
Canadian Journal of Civil Engineering, 2(1), 50-57.
Danielsson H (2013): Perpendicular to grain fracture analysis of wooden structural elements
– models and applications. PhD thesis, TVSM-1024, Div. of Struct. Mech., Lund University,
Sweden.
Ehlbeck J, Kürth J (1991): Influence of perpendicular to grain stressed volume on the load-
carrying capacity of curved and tapered glulam beams. Proc. of CIB-W18 Meeting 24,
Paper 24-12-2.
Fink G, Frangi A, Kohler J (2013): Modelling the bending strength of glued laminated timber
– considering the natural growth characteristics of timber. Proc. of CIBW18 Meeting 46,
Paper 46-12-6.
Franke B (2008): Zur Bewertung der Tragfähigkeit von Trägerausklinkungen in Nadelholz.
PhD thesis, Bauhaus-Universität, Weimar, Germany.

364
 INTER / 48 - 19 - 02

Frühwald E, Serrano E, Toratti T, Emilsson A, Thelandersson S (2007): Design of safe timber

structures – How can we learn from structural failuresin concrete, steel and timber?.
Report TVBK-3053, Div. of Struct. Eng. Lund University, Sweden
Gustafsson PJ (1988): A study of strength of notched beams. Proc. of CIB-W18 Meeting 21,
Paper 21-10-1.
Gustafsson PJ, Enquist B (1988): Träbalks hållfasthet vid rätvinklig urtagning. Report TVSM-
7042, Div. of Struct. Mech., Lund University, Sweden.
Gustafsson PJ (2003): Fracture perpendicular to grain – structural applications. In: Timber
Engineering, ed:s Thelandersson S, Larsen HJ. Wiley.
Gustafsson PJ (2014): Lecture notes on some probabilistic strength calculation models.
Report TVSM-7161, Div. of Struct. Mech., Lund University, Sweden
JCSS (2001): Probabilistic Model Code. Joint Committee on Structural Safety. URL:
http://www.jcss.byg.dtu.dk/.
Jensen JL (2003): Splitting strength of beams loaded by connections. Proc. of CIB-W18
Meeting 36, Paper 36-7-8.
Jockwer R (2014): Structural behaviour of glued laminated timber beams with unreinforced
and reinforced notches. Ph.D. thesis, Inst. of Struct. Eng., ETH Zurich, Switzerland.
Jockwer R, Steiger R, Frangi A (2015): Evaluation of the reliability of design approaches for
connections perpendicular to grain, Submitted to INTER Meeting 2, Sibenik, Croatia.
Larsen H, Gustafsson PJ (1990): The fracture energy of wood in tension perpendicular to the
grain. Proc. of CIB-W18 Meeting 23, Paper 23-19-2.
Larsen, HJ, Riberholt H (1972): Forsøg med uklassificeret konstruktionstræ. Rapport R 31,
Afd. for Bærende Konstruktioner, Danmarks Tekniske Højskole.
Mistler HL (1979): Die Tragfähigkeit des am Endauflager unten rechtwinklig ausgeklinkten
Brettschichtträgers. PhD thesis, Technischen Hochschule Karlsruhe, Karlsruhe, Germany
Mistler HL (1998): Design of glulam beams according to EC 5 with regard to perpendicular-
to-grain tensile strength - A comparison with research results. Holz als Roh – und
Werkstoff, 56(1), 51–60.
Möhler K, Mistler HL (1978): Untersuchungen über den Einfluss von Ausklinkungen im
Auflagerbereich von Holzbiegeträgern auf die Tragfestigkeit, Vol. F 1504.
Nordtest (1993): Wood: Fracture energy in tension perpendicular to the grain, Vol. NT Build
422. Nordtest.
Oscarsson J, Serrano E, Olsson A, Enquist B (2014): Identification of weak sections in gulam
beams using calculated stiffness profiles based on lamination surface scanning. Proc. of
WCTE 2014, Quebec, Canada.
Qiu L, Zhu E, Van De Kuilen J (2014): Modeling crack propagation in wood by extended finite
element method. European J. of Wood and Wood Products, 72(2), 273–283.
Riberholt H, Enquist B, Gustafsson PJ, Jensen RB (1991): Timber beams notched at the
support. Afd. for Baerende Konstruktioner, Danmarks Tekniske Højskole, Denmark.
Weibull W (1939): A statistical theory of the strength of materials. Proceedings no 151,
Royal Swedish Institute for Engineering Research, Sweden.

365
INTER / 48 - 19 - 02

Discussion

The paper was presented by R Jockwer

F Lam and R Jockwer discussed the factor Kknot as being established based on judge-
ment with LN(2,0.8). F Lam commented that the fitting at the tail of the distribution
was not very good. R Jockwer responded that there could be other factors such as in-
fluence of slope of grain involved.
I Smith commented about the load control testing situation versus displacement con-
trol and received information that this case related to crack arrest. He commented
that how a dynamically developing crack growth could be handled with Weibull ap-
proach where Weibull did not consider stability issues. P Gustafsson responded that
quasi static for dynamic loading was considered here and one could integrate the en-
ergy release rate as the crack propagates.
P Dietsch questioned if one applied the information from 48-7-4 in this situation what
would happen. R Jockwer responded that one would expect little difference if the
height adjustment model was applied as it might fit better than the volume adjust-
ment model. With strongest link approach better fit could result.
T Tannert questioned the situation where the predefined failure plane in the FEM was
the weak plane but failure could occur in the strong plane and how strongest link ap-
proach could consider the case with failure in the strong plane. R Jockwer responded
that one could lower the fracture energy of a predefined strong crack plane so that
the crack could develop along the strong plane.

366
367
368
 INTER / 48 - 102 - 01

A proposal for a new Background Docu-

ment of Chapter 8 of Eurocode 8
Maurizio Follesa, Massimo Fragiacomo, Davide Vassallo - Department of Architecture,
Design and Urban Planning, University of Sassari, Palazzo del Pou Salit, Piazza Duomo
6 07041 – Alghero – Italy / dedaLEGNO, Via Masaccio 252, 50132 – Firenze - Italy
Maurizio Piazza, Roberto Tomasi, Simone Rossi, Daniele Casagrande - Department of
Civil, Environmental and Mechanical Engineering, University of Trento, Via Mesiano
77 - I 38123 Trento - Italy

Keywords: Eurocode 8, Multi-storey timber buildings, Background document

1 Introduction
Timber systems are becoming more and more popular in the construction of medium
to high rise buildings in many European regions with different levels of seismicity,
replacing day by day other materials like reinforced concrete, masonry or steel. Only
considering the last decade, more than 20 buildings with a number of storeys ranging
from 6 to 14 have been built in low to high seismicity areas.
Sustainability, energy efficiency and speed of construction together with the
consequent costs reductions, and above all the excellent seismic performance
demonstrated by recent research results based on extensive numerical simulations
and full-scale tests on multi-storey buildings are just some of the advantages which
explain this growing success.
However this progress and the growing diffusion have not been followed by a
corresponding update of the provisions to be used in the seismic design, which for
the case of timber buildings are included within Chapter 8 of Eurocode 8. This
chapter, published in 2004, is very short and incomplete in many parts, especially
when considering design provisions for modern construction systems that are
nowadays widely used, thus causing real difficulties to structural engineers who have
to apply these rules in the seismic design of timber buildings.

2 Seismic design according to Eurocode 8

The seismic design of structures according to Eurocode 8 (CEN, 2004) is founded on a
force-based approach, according to which the energy dissipation capacity of the
whole structure is implicitly taken into account by dividing the seismic forces

369
INTER / 48 - 102 - 01

obtained from a linear (static or dynamic) analysis by the behaviour factor q

associated to the relevant ductility classification, which, according to the definition of
Eurocode 8 “accounts for the non-linear response of a structure, associated with the
material, the structural system and the design procedures”.
Therefore, in order to reach the desired means of ductility and energy dissipation
capacity, (i) the structural system should be clearly identified, (ii) capacity design
provisions should be given so as to address without any possible misinterpretation
the components of the structural system devoted to the ductile behaviour and on the
other hand the parts of the structure which should be over-designed to avoid any
possible anticipated brittle failures and (iii) ductility rules for the dissipative zones
should be given.

This could be achieved by designing the brittle elements for a force equal to the
strength of the ductile elements multiplied by the over-strength factor. Summarizing,
according to the Capacity Based Design philosophy, the design procedure of a
structure should be based on the following method:
1. Clearly identify the structural type and the associated behaviour factor q.
2. Follow the Capacity Based Design rules and the detailing provisions for dissipative
zones defined for the corresponding structural type.
3. Adopt the over-strength factors defined for the design of the brittle elements.

The current version of Chapter 8 of Eurocode 8 (CEN, 2004) is partially or totally

missing the above mentioned conditions for most of the structural types currently
used in the construction practice. Therefore is almost inapplicable for the seismic
design of most of the structural systems nowadays used, forcing the structural
designer to make assumptions which could be not necessarily conservative.

3 A critical analysis of the current version

The current provisions for timber buildings included in Chapter 8 of Eurocode 8
consist of only 5 pages and present some critical aspects partly already addressed in
Follesa et al. (2011), which cause difficulties to structural engineers who have to
apply these rules in the seismic design of timber buildings. More specifically, there
are some clauses which should be improved and some others which deserve further
explanations in order to give a correct guidance to the structural designers, as listed
in the following.
• The structural types should be clearly identified, without any possible
misinterpretation and should be all referred to lateral load resisting systems for
buildings. Furthermore the current list of structural types should be completely
reviewed and updated to the current construction practice.
• Capacity based design rules and detailing provisions for dissipative zones are
totally missing for most of the structural types thus making the choice of the

370
 INTER / 48 - 102 - 01

correct value of the behaviour factor to be applied in the design difficult. Like for
other building materials, these provisions should be given for each structural type
and for each Ductility Class in order to address the choice of the correct behaviour
factor q to be adopted in the seismic design (Casagrande et al., 2014).
• The definition of static ductility should be clarified. Furthermore, the requirements
currently given for DCM and DCH are not always reached by some structural
elements (e.g. nailed wall panels) and connections for timber structures (e.g. hold-
downs and angle brackets in X-Lam structures).
• The ductility rules currently given for dissipative zones, i.e. mechanical joints in the
case of timber structures, are “prescriptive” and not “performance based” as in
the basic philosophy of Eurocodes.
• New provisions for materials and properties of dissipative zones should be added,
including new wood-based materials recently developed such as X-Lam panels, and
the existing provisions for wood-based panels used as sheathing material should
be updated to incorporate, for example, oriented strand board (OSB).
• The values of the over-strength factors γRd, to be used to design the non-dissipative
parts of the structure in order to avoid anticipated brittle failure mechanisms, are
not provided for the different structural types. These values should be included
(Gavric et al. 2014, 2015).
• The partial safety factors γM for fundamental and accidental load combinations for
the ultimate limit state verifications in case of dissipative and non-dissipative
structural behaviour should be reviewed and corrected.
• Inter-storey drift limits for the Damage Limit State verifications should be provided.

4 Background Document for a new version of

Chapter 8 of Eurocode 8
The proposal for the revision of Chapter 8 of Eurocode 8 presented herein is based
both on the results of research projects conducted on the seismic behaviour of
timber buildings through tests and numerical simulations on full-scale buildings,
structural components and single joints, and on the technical development reached
in the field of timber construction in the last years. In the following paragraphs all the
proposed modifications to the code text will be referenced and the motivation as
well as the scientific background behind each proposal will be given, together with
the corresponding literature references.
4.1 Materials and properties of dissipative zones
New types of wood-based and other type of panels largely used in construction
practice should be added to the existing ones. These are:

371
INTER / 48 - 102 - 01

• Oriented Strand Board sheathing (OSB) type 2, 3 or 4 according to EN 300, with a

minimum thickness of 12 mm.
• Gypsum Fibre boards (GF) sheathing according to EN 15283-2, with a minimum
thickness of 12 mm.
• X-Lam panels according to prEN 16351 used in shear walls and diaphragms of solid
construction, with a minimum thickness of 60 mm for shear walls and of 18 mm for
floor and roof diaphragms.
OSB is probably the most used wood-based sheathing material in Light-Frame
construction. There are a large number of experimental results about the good
dissipation properties of Light-Frame shear walls sheathed with OSB panels (e.g.,
Sartori & Tomasi, 2013).
Light-Frame buildings sheathed with Gypsum Fibre boards (GF) sheathing and stapled
connections are becoming more and more used in the current construction practice.
Moreover recent research conducted at the University of Trento, Italy (Sartori and
Tomasi, 2013) and within the SERIES Project (Piazza et al., 2013) have proved the
suitability of Gypsum Fibre Panels (PF-GF) connected to the timber framing with
staples as a sheathing material for shear walls in Light-Frame construction.
Cross Laminated Timber (X-Lam) buildings are widely used all over Europe for the
construction of medium to high-rise buildings in seismic areas. Currently the
qualification process of the structural panels is made by CE marking each panel
according to European Technical Approvals (ETA) which are specific of each single
producer and are generally based on the procedure described in CUAP 03.04/06. At
present a European Standard for X-Lam (prEN 16351 Timber structures — Cross
Laminated Timber — Requirements) is under formal vote and will be published soon.
After the publication, the CE marking of X-Lam panels will be possible only according
to the procedure established in this standard and all the single ETA’s will expire. The
limitation of 60 and 18 mm of panel thickness is given according to current
production specification of most of the European producers.
4.2 Structural types
Similarly to the chapters related to other materials included in Eurocode 8, a clear
definition of the different structural types should be provided, possibly with the
inclusion of graphical sketches in order to ease the understanding. A possible
proposal including new structural types and the re-definition of the existing ones is
given in Table 1.

372
 INTER / 48 - 102 - 01

Table 1. Structural types for timber buildings.

Structural type Example of structure
1. Cross Laminated Timber (X-Lam) system, i.e.
buildings comprised of X-Lam shear walls
according to XX (reference to the Material
Properties section) with the specifications
given in YY (reference to the Capacity Design
Rules section).

2. Light wood-frame system, i.e. structures in

which shear walls are made of timber frames
to which a wood-based panel or other type of
sheathing material according to XX (reference
to the Material Properties section) are
connected according to the specifications
given in YY (reference to the Capacity Design
Rules section).
3. Log House building system, i.e. structures in
which walls are made by the superposition of
rectangular or round solid or glulam timber
elements, prefabricated with carpentry joints
at their ends and with upper and lower
grooves according to specifications given in YY
(reference to the Capacity Design Rules
section).
4. Moment-resisting frames, i.e. frames
composed of timber elements with semi-rigid
joints between the members - made with
mechanical fasteners according to the
specifications given in YY (reference to the
Capacity Design Rules section).

5. Post and beam timber systems, namely

systems of timber columns and beams
pinned-connected, with vertical bracings
made of timber trusses according to the
specifications given in YY (reference to the
Capacity Design Rules section).

6. Mixed structures made of timber framing and

masonry infill resisting to the horizontal
forces.

373
INTER / 48 - 102 - 01

Structural type Example of structure

7. Large span arches with two or three hinged
joints according to the specifications given in
YY (reference to the Capacity Design Rules
section).

8. Large span trusses with nailed, screwed,

doweled and bolted joints according to the
specifications given in YY (reference to the
Capacity Design Rules section).

9. Vertical cantilever systems made with glulam

or X-Lam wall elements according to the
specifications given in YY (reference to the
Capacity Design Rules section).

New structural systems for timber buildings already widely used in seismic regions
such as the Cross Laminated Timber (X-Lam) system and the Log House system are
introduced. With respect to the current 2004 version, all the structural types referred
to structural assemblies for building roofs like trusses with nailed, doweled or bolted
joints or with connectors were removed. The reason for this change is that the
timber trusses were introduced in the 2004 edition probably overlooking the
meaning of timber trusses given in the previous 1995 ENV edition where this system
referred to vertical bracing systems used in buildings (or even large span glulam
roofs, where the timber elements are directly connected to the foundation and resist
vertical and horizontal loads). As this chapter refers to lateral load resisting systems
in timber building, there is no reason to refer to structural assemblies used for roofs.
The structural type referenced in 2004 edition as “Hyperstatic portal frames” is here
referenced with the most common definition of “Moment resisting frames”. Also the
vertical cantilever system is a new structural type not referenced in the 2004 edition
which is nevertheless widely used in seismic regions.
4.3 Behaviour factors
Similarly to the chapters related to other materials included in Eurocode 8, also for
timber structures two different values should be defined, if applicable, for DCM and
DCH ductility classes, according to the ductility provisions given for the dissipative
zones and specified in the Capacity Design Rules which should be provided for each
structural type. For structures designed in accordance with the concept of low-

374
 INTER / 48 - 102 - 01

dissipative structural behaviour (DCL) the behaviour factor q should not be taken
greater than 1.5. A possible proposal is given in Table 2.
Table 2. Structural types and upper limit values of the behaviour factors for buildings regular in
elevation
Structural type DCM DCH
X-Lam buildings 2.0 3.0
Light-Frame buildings 2.5 4.0
Log House buildings 2.0 -
Moment resisting frames 2.5 4.0
Post and beam timber buildings 2.0 -
Mixed structures made of timber framing and masonry infill 2.0 -
resisting to the horizontal forces.
Large span arches with two or three hinged joints - -
Large span trusses with nailed, screwed, doweled and bolted - -
joints
Vertical cantilever systems made with glulam or X-Lam wall 2.0 -
elements

The values given for X-Lam structures are based on research results and numerical
investigations conducted within the Sofie Project and referenced in Ceccotti &
Follesa, (2006), Ceccotti et al. (2007), Fragiacomo et al. (2011), Pozza et al. (2013).
For Light-Frame structures two different values of the behaviour factor q are given
for DCM and DCH. The highest values of 5.0 given in the 2004 edition, even if
provided also in the National Building Code of Canada (NBCC, 2010) is not confirmed
by other international codes (e.g. New Zealand, NZS 3603, 1993) and by numerical
investigations (Follesa, 2015). Therefore a more conservative value of 4.0 is
proposed. For the seismic design according to DCM a value of 2.5, given in Campos
Costa et al., 2013, is proposed in order to include Light-Frame buildings sheathed
with gypsum fibre boards and stapled connections. Unlike the 2004 edition, and
according to the provisions given in the previous 1995 ENV edition, no distinction is
made between glued and nailed diaphragms.
For Log House buildings no reference could be found on behaviour factor proposals;
however studies are in progress (Bedon et al. 2014, 2015). Therefore the proposal is
to use a value of 2.0 for DCM in order to take into account the dissipative
contribution of frictions between logs, anyway further research is needed to confirm
this value.
For moment resisting frames the lower q value (2.5) is confirmed for DCM. For DCH a
value of 4.0 is confirmed, according to recent research results based on extensive

375
INTER / 48 - 102 - 01

numerical investigations on moment resisting frames with densified veneer wood

reinforced joints with expanded tube fasteners (Wrzesniak et al., 2013).
4.4 General rules and capacity design rules
In order to allow the correct choice of the behaviour factor q to be used in the
seismic design for the reference Ductility Class, for each structural type a hierarchy of
resistance among the different structural components should be established. The
whole structure should have an adequate capacity to develop plastic deformations
without substantial reduction in the overall resistance against horizontal and vertical
loads. Furthermore any possible anticipated brittle failure mechanism in the
elements devoted to the energy dissipations, i.e. mechanical joints, should be
avoided.
This could be achieved, as mentioned earlier, by over-designing the brittle elements
with respect to the ductile elements, so as to ensure with a reasonable reliability that
ductile failure modes occur before any possible failure in the brittle structural
elements. Furthermore any possible global instability mechanism or soft-story
mechanism should be avoided at a global level, as well as any possible brittle failure
in the ductile structural elements should be prevented at a local level.
For this reason, capacity design rules should be provided for each structural type
both at building level and at connection level. Seismic design rules should be divided
for the different structural types into General Rules (where the above defined
structural types should be further detailed, even by means of detailing drawings, in
order to avoid any possible misinterpretations) and Capacity Design Rules, both at
global and local level for the different Ductility Classes.
As an example, a proposal of possible General Rules and Capacity Design Rules for X-
Lam buildings is presented. Due to length limitations it is not possible to provide the
same provisions for all the above defined structural types in this paper. For further
details see Follesa, 2015.
1 -Rules for X-Lam buildings
1.1 – General rules
(1) Cross laminated (X-Lam) timber buildings are structures in which walls are composed of cross
laminated timber panels according to XX.
(2) The connection of the walls to the foundation should be made by means of mechanical fasteners
(hold-down anchors, steel brackets, anchoring bolts, nails and screws) and shall adequately restrain
the wall against uplift and sliding. Uplift connections should be placed at wall ends and at opening
ends, while shear connections should be distributed uniformly along the wall length (Figure 1).
(3) Walls shall have heights at least equal to the inter-storey height and may be made of a unique
element up to the maximum transportable length or may be composed of more than one panel
(“segmented wall”). Each segment shall have width not lower than 0,25 h, h signifying the inter-
storey height, and shall be connected to the other segments by means of vertical joints made with
mechanical fasteners such as screws or nails. Individual wall-panels with a width of less than 0,25 h

376
 INTER / 48 - 102 - 01

shall not be regarded as a seismic resistant shear wall. Perpendicular walls are connected by means
of joints made with mechanical fasteners (usually screws). Horizontal joints between walls should be
avoided unless special provisions are taken to ensure adequate out- of- plane restraint (e.g. properly
connected to perpendicular stabilizing walls, timber studs, etc.).
(4) Floor and roof diaphragms are made of X-Lam timber panels connected together by means of
horizontal joints made with mechanical fasteners (screws or nails). The floor panels bear on the wall
panels and on timber beams if present, to which they are connected with mechanical fasteners
(screws or nails).
(5) Other types of horizontal diaphragms may be used, provided that their in-plane rigidity is
ensured by means of wood-based sheathing panels according to XX (reference to the Material
Properties section). Timber-concrete composite floors may be used provided that they are
adequately connected to the lower and upper walls by means of mechanical fasteners. The concrete
topping, in particular, shall be connected to the vertical panels to ensure the in-plane shear due to
the diaphragm action is transferred to the walls and down to the foundations.
(6) The upper walls bear on the floor panels (platform construction), and are connected to the lower
walls using mechanical fasteners similar to those used for the wall-foundation connection. Tie-down
connections nailed to the X-Lam walls may be used for the external walls uplift restraint.
Wall panel

Vertical joint
Single piece wall

Horizontal joint

Hold-down
Steel brackets
Hold-down

Hold-down

Figure 1. Walls and floors in monolithic, left and segmented, right Cross Laminated Timber buildings
(after Follesa et al., 2011).
1.2 – Capacity design rules for DCM
(1) In X-Lam buildings designed for DCM, each wall may be composed either by only one X-Lam
panel or by more than one panel connected with rigid vertical joints.
1.2.1 – Capacity design rules at building level
(1)P X-Lam buildings should be regarded as box-type structures. In order to achieve this behaviour
some structural components should be considered as dissipative zones and some other should be

377
INTER / 48 - 102 - 01

considered as non-dissipative and properly designed with sufficient overstrength to avoid any
possible brittle failure mechanism. The dissipative connections will use dowel-type fasteners inserted
perpendicular to the shear direction. Connections with dowel-type fasteners inclined in the shear
direction (e.g. screws inclined in the shear direction) are allowed only in non-dissipative connections.
The structural elements which should be designed with sufficient overstrength, according to
Equation (1), in order to ensure the development of cyclic yielding in the dissipative zones, shall be
(Figure 2):
• all X-Lam wall and floor panels;
• connections between adjacent floor panels (or connection of other type of sheathing material like
in XX (reference to the Material Properties section)) in order to limit at the greater possible extent
the relative slip and to assure a sufficiently rigid in-plane behaviour;
• connections between floors and the underneath walls thus assuring that at each storey there is a
sufficiently rigid floor to which the walls are rigidly connected;
• connections between perpendicular walls, particularly at the building corners, so that the stability
of the walls themselves and of the structural box is always ensured.

CLT floor panels

Connection between
floors and walls underneath

Connection between
perpendicular walls

CLT wall panels

Figure 2. Connections designed with overstrength in order to fulfil the capacity design criteria in
Cross Laminated Timber buildings in DCM. (after Follesa et al., 2011, modified).
(2)P The connections devoted to the dissipative behaviour in a X-Lam building should be:
• shear connections between walls and the underneath floor, and between walls and foundation
(usually steel brackets or screwed connections);
• anchoring connections against uplift placed at wall ends and at wall openings (usually hold-down
anchors).
1.2.2 – Capacity design rules at connection level

378
 INTER / 48 - 102 - 01

(1)P When designing connections as defined in 1.2.1(2)P a ductile failure mode characterized by
yielding of fasteners (nails or screws) in steel-to-timber or timber-to-timber connections should be
achieved and brittle failure mechanisms should be avoided according to overstrength rules.
Specifically the following failure mechanisms shall be avoided:
• tensile and pull-through failure of anchor bolts and screws;

• steel plate tensile and shear failure in the weakest section of hold-down and angle brackets
connections.
Other brittle failures such as splitting, shear plug, tear-out and tensile fracture of wood in the
connection regions should be always avoided.
1.3 – Capacity design rules for DCH
(1) In X-Lam buildings designed for DCH, in each perpendicular direction, at least 50% in length of
the seismic resistant shear walls shall be segmented. Segmented walls shall be composed by more
than one panel, each one of width not lower than 0,25h, h signifying the inter-storey height, and not
greater than h, connected with joints made with mechanical fasteners (screws or nails) inserted in a
horizontal plane.
1.3.1 – Capacity design rules at building level
(1)P The same provisions of 1.2.1(1)P apply.
(2)P In addition to the provisions of 1.2.1(2)P the following connections should be considered and
designed for dissipative behaviour:
• vertical screwed or nailed step joints between adjacent parallel wall panels within the segmented
shear walls.
1.3.2 – Capacity design rules at connection level
(1)P The provisions of 1.2.2 apply.
4.5 Ductility rules for dissipative zones
A modification of the provisions currently included in Chapter 8 of Eurocode 8 in
order to assign the dissipative zones of timber structures to a Medium or High
Ductility Class is proposed as in the following:
(3)P In order to ensure that the given values of the behaviour factor may be used, the dissipative
zones, specified in the capacity design rules for each structural type, shall be able to deform
plastically for at least three fully reversed cycles at a static ductility ratio reported in Table 3, without
more than a 20% reduction of their resistance between the first and third cycles backbone curve.
Table 3. Static ductility values of dissipative zones tested according to EN12512 without more than a
20% reduction of their resistance between the first and third cycles envelope backbone curve for all
structural types.
Structural type Dissipative sub- DCM DCH
assembly/element
X-Lam buildings Shear wall 3.0 4.0
X-Lam buildings Hold-downs, angle brackets, 3.0* 4.0*
screws

379
INTER / 48 - 102 - 01

Light-Frame buildings Shear wall 3.0 5.0

Light-Frame buildings Fastener (nail/screw)** 4.0* 6.0*
Log House buildings Shear wall 2.0 -
Moment resisting Portal Frame 2.0 3.0
frames
Moment resisting Beam-column joint 6.0 10.0
frames
Post and beam timber Braced Frame 2.0 -
buildings
Mixed structures Shear wall 2.0 -
made of timber
framing and masonry
infill resisting to the
horizontal forces
Vertical cantilever Shear wall 2.5 -
systems made with
glulam or X-Lam wall
elements

(4) The provisions of (3)P of this sub clause may be regarded as satisfied in the dissipative zones of
all structural types classified in DCH if the following provisions are met:
a) in doweled, bolted and nailed timber-to-timber and steel-to-timber joints with fasteners inserted
perpendicular to the shear plane, the minimum thickness of the timber connected members is 10d
and the fastener-diameter d does not exceed 12 mm;
b) In shear walls and diaphragms of Light-Frame construction, the sheathing material is wood-based
with a minimum thickness of 4d, where the nail diameter d inserted perpendicular to the sheathing
plane does not exceed 3,1 mm.

If the above requirements are not met, but the minimum member thickness of 8d and 3d for case a)
and case b), respectively, is assured, the dissipative zones of all structural types can be regarded as
ductility class M.

As an alternative, the above provisions may be regarded as satisfied:

- for the dissipative zones of all DCM structural types and of the ductility class H X-Lam system
with segmented wall according to 1.3., if a ductile failure mechanism characterized by the formation
of at least one plastic hinge in the mechanical fasteners inserted perpendicular to the shear plane is
attained for the seismic design load condition;
- for the dissipative zones of all DCH structural types, if a ductile failure mechanism
characterized by the formation of two plastic hinges in the mechanical fasteners inserted
perpendicular to the shear plane is attained for the seismic design load condition.
Referring to 8.2.2 of EN 1995-1-1 for timber-to-timber and panel-to-timber connections, failure
modes a, b and c for fasteners in single shear, and g and h for fasteners in double shear should be
avoided. Referring to 8.2.3 of EN 1995-1-1 for steel-to-timber connections, failure modes a, c for
fasteners in single shear, and f, j and l for fasteners in double shear should be avoided. Special care
should be taken in avoiding brittle failures characterized by splitting, shear plug, tear out and tensile

380
 INTER / 48 - 102 - 01

fracture of wood in the connection regions. In the case of connections with multiple fasteners in
dissipative zones, adequate reinforcement should be added to avoid the aforementioned brittle
failure mechanisms.

The value marked with asterisk (*) in Table 3 have been calculated according to
Casagrande et al. 2015.
For Light-Frame buildings with nailed/screwed connections (**) different values
should be provided for the case of nailed or screwed sheathing-to-frame
connections. However, since researches on Light-Frame buildings with screwed
sheathing-to-frame connections are still ongoing, conservative values to be
considered valid for both cases are provided.
There are a number of recent scientific experiences which demonstrate that the
current provisions in order to classify timber structures in Medium and High ductility
class (i.e. the dissipative zones shall be able to deform plastically for at least three
fully reversed cycles at a static ductility ratio of 4 for ductility class M structures and
at a static ductility ratio of 6 for ductility class H structures, without more than a 20%
reduction of their resistance) are not always met both for Light-Frame (Boudaud et al,
2010; Vogt et al., 2012; Sartori and Tomasi, 2013; Tomasi and Sartori, 2013) and X-
Lam (Gavric et al., 2011, Gavric et al., 2014, Gavric et al., 2015) shear walls and
connections both in terms of values of static ductility and/or strength degradation.

Also regarding moment-resisting frames, Wrzesniak et al., 2013, by performing

Incremental Dynamic Analysis (IDA) on an industrial portal frame and a three-story,
five-bay frame, found that for a joint ductility of at least 6 and 4, q-factors of
respectively 2 and 1.5 could be suggested for the portal frame, while the
recommended q-factor values could be 2.5 and 2 for the 3-storey, 5-bay frame,
which are markedly different for the q values suggested by the current version of
Eurocode 8.
Thus the code proposal is to specify different minimum values of static ductility in
order to attain a certain behaviour factor q depending upon the scale of the
dissipative component being considered. More specifically, two levels are considered:
(i) connection (single fastener in Light-Frame systems; hold-down/angle bracket and
screw connection in X-Lam systems; beam-column joint in moment-resisting frames)
or (ii) sub-assembly (e.g. shear walls, portal frames or single span trusses according to
the structural type).
Also a new formulation of the detailing rules for dissipative zones is proposed,
integrating the existing "prescriptive" philosophy of 2004 and previous editions with a
new alternative provision towards “performance based” design rules. The new
provisions are intended to require the attainment of a certain ductile failure mode
(one or two plastic hinge formation in the fastener) according to the Johansen theory
as given in Eurocode 5 independently of the fastener diameter and the member
thickness.

381
INTER / 48 - 102 - 01

4.6 Over-strength factors and safety verifications

As already discussed above, in order to achieve the desired ductile behaviour for the
whole building and fulfil the capacity based provisions given for the specified
structural system and material, the overstrength factors to be used in the design of
the brittle elements should be defined.
A possible proposal could be the following.

(2)P In order to achieve the attainment of the capacity design rules defined in the following sub
clauses for all the structural types listed in Table 1, the design strength of the brittle parts FRd,b
should be greater than or equal to the design strength of the ductile parts FRd,d multiplied by an
overstrength factor γRd and divided by a reduction factor for strength degradation βsd due to cyclic
loading according to the following equation:
∙ , ≤ , (1)

where the values of γRd are provided in Table 4, and the value of βsd is defined in XXX (reference to
the Safety Verifications section) and assumes the values given in Table 5.
Lower values of the overstrength factor γRd can be used only if supported by experimental tests.
Table 4. Values of the over-strength factor γRd.
Structural type Overstrength factor γRd
X-Lam buildings 1.3
Light-Frame buildings 1.3
Log House buildings 1.3
Moment resisting frames 1.6
Post and beam timber buildings 1.6
Mixed structures made of timber framing and masonry infill 1.3
resisting to the horizontal forces
Vertical cantilever systems made with glulam or X-Lam wall 1.6
elements

The proposed value of over-strength factors are referenced in Casagrande et al.

(2014), Schick et al. (2013), Jorissen & Fragiacomo, (2011) and Follesa et al., (2011).
The definition and the proposed values of the reduction factor for strength
degradation due to cyclic loading βsd is given later in the Safety Verification section as
follows:

(2)P The partial factors for material properties γM for accidental load combinations in accordance
with EN 1995-1-1 apply.
(3)P For ultimate limit state verifications of structures designed in accordance with the concept of
dissipative structural behaviour (Ductility classes M or H), the strength degradation of the dissipative

382
 INTER / 48 - 102 - 01

zones shall be taken into account by multiplying the characteristic strength in static conditions by a
reduction factor βsd given in Table 5. The strength degradation of the non-dissipative zones may not
be taken into account. Values of the reduction factor βsd higher than in Table 5 may be used if the
actual strength degradation due to cyclic loading is appropriately derived from experimental tests;
the strength degradation shall be evaluated at the static ductility ratio shown in Table 3.
(4)P For ultimate limit state verifications of structures designed in accordance with the concept of
low dissipative structural behaviour (Ductility class L), no strength degradation will be considered.
(5)P In order to ensure the development of cyclic yielding in the dissipative zones, all other structural
members and connections shall be designed with sufficient overstrength.
Table 5 Values of the reduction factor βsd for strength degradation due to cyclic loading
Structural type Reduction factor βsd
X-Lam buildings 0.8
Light-Frame buildings with stapled sheathing-to- 0.7
frame connections
Light-Frame buildings with nailed and screwed 0.8
sheathing-to-frame connections
Log House buildings 0.8
Moment resisting frames 0.8
Post and beam timber buildings 0.8
Mixed structures made of timber framing and 0.8
masonry infill resisting to the horizontal forces
Vertical cantilever systems made with glulam or 0.8
X-Lam wall elements

As already mentioned earlier, the partial safety factors γM for fundamental and
accidental load combinations for the ultimate limit state verifications in case of
dissipative and non-dissipative structural behaviour were inverted in the current
2004 edition with respect to the previous 1995 ENV version, where they were partly
consistent with the same provisions given for other materials, like RC, steel or
masonry.
In seismic design the partial safety factor γM=1.0 given for the accidental load
combination should be used, as seismic actions should be regarded as an accidental
load case. However, the reference strength considered for the dissipative zones,
when designing according to the dissipative behaviour concept, should be adequately
reduced to take into account the strength degradation due to cyclic loading. For
timber structures, the dissipative elements are mechanical connections, whose
strength is calculated according to Eurocode 5 (EN 1995-1, CEN, 2009) using the
Johansen theory. This strength is calculated for monotonic loading and no allowance

383
INTER / 48 - 102 - 01

for degradation due to cyclic loading is made. Therefore the proposal is to consider
the strength degradation by multiplying the characteristic strength in static
conditions by a reduction factor βsd given in Table 5. If the strength degradation is
appropriately accounted for by means of experimental evaluation at the static
ductility values given in Table 3, higher values of βsd may be used. This proposal is in
accordance with the other materials chapters, where it is therefore prescribed that
the partial safety factor γM=1.0 given for the accidental load combination could be
used only if the strength degradation due to cyclic loading is appropriately accounted
for in the evaluation of the connection resistance.
For structures designed in accordance with the concept of low dissipative structural
behaviour (Ductility class L), and for the non-dissipative zones of structures designed
in accordance with the concept of dissipative structural behaviour, no strength
degradation should be taken into account.

5 Conclusions
The current provisions for timber buildings included in Section 8 of Eurocode 8 seem
not sufficient and adequate to guarantee a correct design of timber buildings
according to the general principles of seismic design included in the common
sections of Eurocode 8. This is demonstrated by the absence of specific provisions for
new structural types currently widely used such as the X-Lam system and the lack of
capacity based design criteria and over-strength factors to be used in the design in
order to ensure the desired dissipative behaviour for the whole structure.
For these reasons in this paper a proposal for a new “Background Document” useful
for a revision of Chapter 8 of Eurocode 8 is presented, based on recent research
results and literature review.
The proposal is formulated in terms of revision and addition of some specific parts
such as: (i) materials and properties of dissipative zones, (ii) structural types, (iii)
behaviour factors, (iv) general rules and capacity based design rules and (v) ductility
rules for dissipative zones and safety verification.
Given the length limitation of this paper, it is not possible to provide a comprehensive
overview of the full proposal which includes also some provisions for (i) buildings
with different lateral load resisting systems, (ii) the design of new structural elements
and systems not listed in the structural types included in this proposal, (iii) inter-
storey drift limits, and (iv) other minor changes and additions to the existing
provisions.
Although the suggested additions are not sufficient to define a complete document
according to the general principles of Eurocode 8, the proposal is still under
discussion and provides a fundamental update of the code provisions for the seismic
design of timber structures. This proposal could therefore represent a useful tool,

384
 INTER / 48 - 102 - 01

updated to the recent technical and scientific developments for the future revision of
Chapter 8 of Eurocode 8.

6 References
Becker K. et al. (1993): Eurocode 8 – Part 1.3 – Chapter 5 – Specific rules for timber
buildings in seismic region. Proceedings of the 26th CIB W18, Athens, Georgia.
(paper 26-15-2).
Bedon, C. et al. (2014): Experimental study and numerical investigation of
“Blockhaus” shear walls subjected to in-plane seismic loads. ASCE Journal of
Structural Engineering. (ASCE published online, doi: 10.1061/(ASCE)ST.1943-
541X.0001065).
Bedon, C. et al. (2015): Non-linear modelling of the seismic behaviour of 'Blockhaus'
structures. Engineering Structures, in print.
Bratulic, K. et al. (2014): Monotonic and cyclic behaviour of joints with self-tapping
screws in CLT structures. Cost Action FP 1004. Experimental Research with Timber,
University of Bath, Prague, Czech Republic (pages 1-8).
Boudaud, C. et al. (2010): European seismic design of shear walls: experimental and
numerical tests and observations. Proceedings of the 11th World Conference on
Timber Engineering, Riva del Garda (TN), Italy.
Campos Costa, A. et al. (2013): Seismic performance of multi-storey timber buildings -
RubnerHaus building - Final Report. SERIES. (Work Package [WP9-TA5 LNEC]).
Casagrande, D. et al. (2014): Capacity design approach for multi-storey timber-frame
buildings. Proc. of the International Network on Timber Engineering Research
meeting INTER, Bath, United Kingdom. (paper 47-15-3).
Casagrande, D. et al.: (2015) A predictive analytical model for the elasto-plastic
behaviour of a light timber-frame shear-wall. Construction and Building Materials,
(Special Issue “SHATIS 2013”, 10.1016/j.conbuildmat.2015.06.025)
Ceccotti, A. & Follesa, M. (2006): Seismic Behaviour of Multi-Storey X-Lam Buildings.
Proceedings of 426 COST E29 International Workshop on Earthquake Engineering
on Timber Structures, Coimbra, Portugal. (pages 81-95).
Ceccotti, A. et al. (2007): Quale fattore di struttura per gli edifici multipiano a
struttura di legno con pannelli a strati incrociati? (in Italian). Proceedings of the
12th ANIDIS Conference, Pisa (Italy).
Ceccotti, A. & Larsen, H.J. (1988): Background document for specific rules for timber
structures in Eurocode 8. Report EUR 12266 EN for the Commission of the
European Communities, Brussels, Belgium
EN12512 (2001): Timber structures- Test methods. Cyclic testing of joints made with
mechanical fasteners. CEN.
Eurocode 5 (2009): Design of timber structures - Part 1-1: General and rules for
buildings. CEN. (EN 1995-1-1).

385
INTER / 48 - 102 - 01

Eurocode 8 (2004): Design of structures for earthquake resistance, Part 1: General

rules, seismic actions and rules for buildings. CEN. (EN 1998-1-1).
Follesa, M. (2015): Seismic design of timber structures - A proposal for the revision of
Chapter 8 of Eurocode 8. Phd Thesis, Università degli Studi di Cagliari, Italy.
Follesa, M. et al. (2011): A proposal for revision of the current timber part (Section 8)
of Eurocode 8 Part 1. Proceedings of the 44th CIB W18 Meeting, Alghero, Italy.
(paper 44-15-1).
Flatscher, G. et al. (2014a): Experimental tests on cross-laminated timber joints and
walls. ICE Journal Structures and Buildings.
(http://dx.doi.org/10.1680/stbu.13.00085).
Flatscher, G. et al. (2014b): Screwed joints in Cross-Laminated timber structures.
Proceedings of the 13th World Conference on Timber Engineering WCTE 2014,
Quebec City, Canada.
Flatscher, G. & Schickhofer, G. (2015): Shaking table test on a cross-laminated timber
structure. ICE Journal Structures and Buildings.
(http://dx.doi.org/10.1680/stbu.13.000856).
Fragiacomo, M. et al. (2011): Elastic and ductile design of multi-storey crosslam
massive wooden buildings under seismic actions. Engineering Structures, Special
Issue on Timber Structures. (Vol. 33 No. 11, pp. 3043-3053).
Gavric, I. et al. (2011): Experimental cyclic tests on cross-laminated timber panels and
typical connections. Proceedings of the 14th ANIDIS Conference, Bari (Italy).
Gavric, I. et al. (2014): Cyclic behaviour of typical metal connectors for cross-
laminated (CLT) structures. RILEM Materials and Structures. (Vol. 48 No. 3, pp.
1841-1857, doi: 10.1617/s11527-014-0278-7).
Gavric, I. et al. (2015): Cyclic behaviour of typical screwed connections for cross-
laminated (CLT) structures. European Journal of Wood and Wood Products. (Vol.
73 No. 2, pp. 179-191, doi: 10.1007/s00107-014-0877-6).
Hummel, J. et al. (2013): CLT wall elements under cyclic loading – details of
anchorage and connection. Cost Action FP1004, Focus Solid Timber Solutions –
European Conference on Cross Laminated Timber (CLT), The University of Bath,
Graz. (R.Harris, A. Ringhofer and G. Shickhofer eds., pp. 152-165).
Jorissen, A. & Fragiacomo, M. (2011): General notes on ductility in timber structures.
Engineering Structures. (Vol. 33, N. 11, pp. 2987-2997).
Norme Tecniche delle Costruzioni NTC 2015 (in Italian) – Draft Nov. 4, 2014.
NRC (2010): National Building Code of Canada. Canadian Commission on Building and
Fire Code, National Research Council of Canada, Ottawa, Ont.
NZS 3603 (1993): Timber structures standard. Wellington, New Zealand.

386
 INTER / 48 - 102 - 01

Piazza, M. et al. (2013): Seismic performance of multi-storey timber buildings -

RubnerHaus building – Final Report. Seismic Engineering Research Infrastructures
For European Synergies (SERIES). (Work package [WP9 – TA5 LNEC]).
Pozza, L. et al. (2013): A non-linear numerical model for the assessment of the
seismic behaviour and ductility factor of X-Lam timber structures. Proceedings of
the 46th CIB-W18-Meeting, Vancouver, Canada. (paper 46-15-5).
Sartori, T. & Tomasi, R. (2013): Experimental investigation on sheathing-to-framing
connections in wood shear walls. Engineering Structures. (Vol. 56, pp. 2197- 2205).
Schick, M. et al. (2013): Connection and anchoring for wall and slab elements in
seismic design. Proceedings of the 46th CIB-W18-Meeting, Vancouver, Canada.
(paper 46-15-4).
Tomasi, R. & Sartori, T. (2013): Mechanical behaviour of connections between wood
framed shear walls and foundations under monotonic and cyclic load. Construction
and Building Materials. (Vol. 44, pp. 682-690).
Vogt, T et al., (2012): Timber framed wall elements under cyclic loading. Proceedings
of the 12th World Conference on Timber Engineering, Auckland, New Zealand
Wrzesniak, D. et al. (2013): Proposal for the q-factor of moment-resisting timber
frames with high ductility dowel connectors. Proceedings of the 46th CIB-W18-
Meeting, Vancouver, Canada. (paper 46-15-6).

387
INTER / 48 - 102 - 01

Discussion

The paper was presented by M Follesa

G Schickhofer stated that many structural types were considered here and that some
should be considered only in the national annex such as log houses. M Follesa said
that log houses would be interesting for Italy.
T Toratti commented that the two floor diaphragm cases where the floor went
through the whole building would result in acoustic issues
F Lam commented that the ductility factor for screws connected light wood frame
walls is too high. M Follesa agreed.
D Moroder discussed the information in slide 28 where β to reduce over-strength also
has a β reduction; it seemed double reduction.
A Ceccotti commented that slide 7 stated that the current provisions are not always
met. He asked whether the provisions are on the safe side. M Follesa said no and
commented that for example some Xlam connectors have lower ductility ratio.
G Schickhofer and M Follesa discussed the definition of ductility as some of these are
not consistent.
S Winter stated that the production of background documents is good; he asked
about time schedule and finalization. M Follesa said that the proposal would be final-
ized by 2016. S Winter commented that the document should be circulated as soon as
possible even as a draft.

388
389
390
 INTER / 48 - 102 - 02

Aspects of Code Based Design of

Timber Structures1
Jochen Kohler
Dept. of Structural Engineering, Norwegian University of Science and
Technology (NTNU), Trondheim, Norway
Gerhard Fink
Empa, Swiss Federal Laboratories for Material Science and Technology,
Dübendorf, Switzerland

KEYWORDS: Code calibration, Structural reliability, Load and resistance factor design, Eu-
ropean timber design standards
ABSTRACT: The European timber design standard is under development and a new ver-
sion will be issued at the end of this decade. In this paper the present design standard is
critically assessed in regard to its ability to identify design solutions with a consistent level of
reliability. The main issues to enhance the current standards are identified and discussed.
Thereunder, the influence of different material properties in different load directions, the
quality of the grading process, the application of the current safety concept on non-linear
design equations, duration of load effects, effects due to moisture induced stresses, vol-
ume and length effects, reliable and uniform design equations for joints and the adoption of
consequence classes for associated situations are considered.

1 INTRODUCTION
Sustainable development is the important requirement and goal for modern society and the
international research community is in demand to find solutions that provide the foundation
for this aim. The role of structural engineering research is thereby of significant importance.
The development of methodologies and principles that allows for the optimal allocation of
resources into the structural performance and their implementation into the daily engineer-
ing practice constitute the major challenge for ongoing and future research in the field of
structural engineering.
The broad implementation of newly developed principles requires their proper transi-
tion into rules and regulations that constitute the basis for the daily work of practicing engi-
neers. Thus, rules and regulations as structural design codes constitute the mayor interface
between structural engineering research and practical application and it is of utmost impor-
tance that structural design codes are up to date with the best scientific information available
and, at the same time, are simple enough for straight forward application.
This challenge outlined above is general for the entire structural engineering research
and professional community. Here, timber and timber based materials might be attributed
1 The
work is based on an article published at the 12th International Conference on Applications of Statistics
and Probability in Civil Engineering (Kohler and Fink, 2015) including some adoptions after discussions at the
conference.

391
INTER / 48 - 102 - 02

with a special status since timber as a natural grown material plays an important role in the
safe, cost efficient and sustainable development of our future build environment because of
its beneficial properties. Timber is an efficient building material, not least in regard to its
mechanical properties but also because it is a highly sustainable material considering all
phases of the life cycle of timber structures: production, use and decommissioning.

Timber is a widely available natural resource; e.g. with proper management, there is a po-
tential for a continuous and sustainable supply of raw timber material in the future. Because
of the low energy use and the low level of pollution associated with the manufacturing of
timber structures the environmental impact is much smaller than for structures built in other
materials.
However, besides the beneficial properties of timber the confident use of timber as a load-
bearing material is particularly challenging compared to other common structural materials
as steel and concrete. One of the main reasons for this is that timber is a highly complex
material; the proper use in structures actually requires a significant amount of expertise in
structural detailing.
Another main reason is that any prediction of the structural performance of timber is as-
sociated with large uncertainties. Timber is by nature a very inhomogeneous material. The
material properties depend on the specific wood species, the geographical location and fur-
thermore on the local growing conditions over the entire lifetime of the tree. Timber is an
orthotropic material, i.e. it consists of “high strength” fibres/grains which are predominantly
orientated along the longitudinal axis of a timber log/ tree and packed together within a “low
strength” matrix. After a log is sawn into pieces of structural timber, irregularities, such as
grain direction or knots, become, in addition to the orthotropic characteristics mentioned
above, highly decisive for the load-bearing capacity of a timber structural element. Conse-
quently, the properties of solid timber cannot be designed or produced by means of some
recipe but may be ensured to fulfil given requirements only by quality control procedures
implemented during the production process for sawn timber. Timber material for structural
purpose is generally associated to a certain grade or strength class. However, there are
various different ways how quality control is implemented in the production process and the
properties of timber of a certain strength class are highly sensitive to the quality control
scheme applied to the timber.
Timber is a viscoelastic and hygroscopic material. When using timber as a load-bearing
element in a structure it is of high interest how the load-bearing performance is developing
over time, i.e. how the building environment with its variable loads, temperature and moisture
is influencing the timber structural element.
The high importance of structural timber and timber products for the sustainable devel-
opment of our build infrastructure together with the fact that many features of the structural
behaviour of timber are not known with accurate precision underlines the urgent need for
extensive and coordinated research in this field. Furthermore it is necessary that current
and future knowledge about timber and timber based materials load- bearing behaviour is
represented in the current design standards in a sensible way.

In Europe the design of structures is regulated by the Eurocodes, a suite of consistent stan-
dards for structural design covering all relevant load scenarios and building materials. They
were developed under the supervision of the European Committee of Standardization (CEN)
and regulate to a large extent the performance criteria of the build environment being reliabil-
ity, serviceability and safety of structures. The Eurocodes had been introduced in the 1980s
and are by now compulsory for structural engineering design in most European countries.
Until 2020 a revision and update of the Eurocodes is planned. Thus, this constitutes an

392
 INTER / 48 - 102 - 02

excellent opportunity to critically reflect the design procedures prescribed in the Eurocode 5
– “Design of Timber Structures” in the light of recent scientific developments.

Motivated by the ongoing revision and of the current version of Eurocodes a COST Action,
entitled Basis of Structural Timber Design – from research to standards2 , was initiated.
The presented paper represents a summary of tasks planned to work on during this COST
Action.

2 BASIC PRINCIPLES OF RELIABILITY BASED

CODE CALIBRATION
Modern design codes, such as the Eurocodes (2002), are based on the so-called load and
resistance factor design (LRFD) format. Next, the principle of LRFD is explained for the case
of two loads; one that is constant and one that is variable over time. The LRFD equation is
given in Eq. (1). Here Rk , Gk and Qk are the characteristic values of the resistance R, the
permanent load G, and the time variable load Q. γm , γG and γQ are the corresponding partial
safety factors. z is the so-called design variable, which is defined by the chosen dimensions
of the structural component.
Rk
z − γG Gk − γQ Qk = 0 (1)
γm

The characteristic values for both load and resistance are in general defined as fractile val-
ues of the corresponding probability distributions. In Eurocode 5 (2004) the following charac-
teristic values are defined: Rk is the 5% fractile value of a Lognormal distributed resistance,
Gk is the 50% fractile value (mean value) of the Normal distributed load (constant in time),
and Qk is the 98% fractile value of the Gumbel distributed yearly maxima of the load (variable
in time).
The corresponding partial safety factors can be calibrated to provide design solutions
(z) with an acceptable failure probability Pf (Eq. 2). Here R, G, and Q are resistance and
loads represented as random variables, z∗ = z(γm , γG , γQ ) is the design solution identified with
Eq. (1) as a function of the selected partial safety factors, and X is the model uncertainty.

Pf = P{g(X, R, G, Q) < 0}
(2)
with g(X, R, G, Q) = z∗ XR − G − Q = 0

Often the structural reliability is expressed with the so-called reliability index β (Eq. 3). A
common value for the target reliability index is β ≈ 4.2 which corresponds to a probability of
failure Pf ≈ 10−5 (JCSS, 2001).

β = −Φ−1 (Pf ) (3)

In general, different design situations are relevant; i.e. different ratios between G and Q. This
can be considered using a modification of Eq. (1)–(2) into Eq. (4)–(5). αi might take values
between 0 and 1, representing different ratios of G and Q. R̂, Ĝ, and Q̂ are normalized to a
mean value of 1. For each αi one design equations exists, thus altogether n different design
equations have to be considered.

2 http://www.costfp1402.tum.de

393
INTER / 48 - 102 - 02

Table 1: Chosen representation of the model uncertainty X, the bending strength R, the permanent
load G and the variable load Q.

X R G Q
Mean value 1 1 1 1
Standard deviation 0.1 0.25 0.1 0.4
Distribution type Lognormal Lognormal Normal Gumbel
Fractile - 0.05 0.5 0.98
Characteristic value - 0.647 1 2.037

R̂k
zi − γG αi Ĝk − γQ (1 − αi )Q̂k = 0 (4)
γm

gi (X, R̂, Ĝ, Q̂) = z∗i X R̂ − αi Ĝ − (1 − αi )Q̂ = 0 (5)

Afterwards, the partial safety factors (γm , γG , and γQ ) can be calibrated by solving the optimi-
sation problem give in Eq. (6).
" #
n  2
min ∑ βtarget − β j (6)
γ
j=1

The reliability based code calibration is briefly introduced to illustrate the influence of un-
certainties (load and resistance), in respect to codes. Please find more information in (e.g.
JCSS, 2001; Faber and Sørensen, 2003).
The application of the above sketched framework constitutes the basis for reliability based
calibration of the partial safety factors of a load and resistance factor design format. And it
entirely depends on a realistic representation of loads, resistances and model accuracy by
the random variables R, Q, G, and X.

2.1 Example
The design equation for a structural component can be calibrated according to the proce-
dure described in above. The chosen variables of Eq. (4) and Eq. (5) are summarized in
Table 1. Using this values the situation could represent a solid timber bending beam loaded
by constant (e.g. self-weight of beam and installations) and variable (e.g. live load).
In the presented example, which is explained in more detail in Kohler and Fink (2012),
the range α = [0.1, 0.2, ..., 0.8] is chosen, to exclude rather unrealistic design situations.
The calculations was performed with the software CodeCal (JCSS, 2001). In Figure 1
the chosen target reliability index of β = 4.2 (red line) is compared with the design solu-
tions for the structural component obtained according to the current version of the Eurocode
(γm = 1.3, γG = 1.5, γQ = 1.5); represented by the line with squares. The reliability indices of
the design solutions according to the Eurocode tend to be too low compared to the target
reliability index, especially for small α. The line with the diamonds is obtained when all
partial safety factors are subject to optimization: γm = 1.29, γG = 1.30, γQ = 1.57. However, it
is the philosophy of the Eurocodes that the partial safety factors for the loads are material
independent. Thus, γG and γQ are fixed and the optimization is performed only subject to γm .
The line with the circles in Figure 1 is representing the corresponding result (γm = 1.33).

394
 corresponding result (  m  1.33 ). An 3.1. Different “material properties”
INTER
enhancement in fit to the target reliability can be / 48is- a102
Timber rather- 02
complex buildin
observed for both calibrated solutions. properties are highly variable, sp
time. In structural engineeri
4.4 properties of timber are in general
the stress and stiffness related
standard test specimen under giv
4.3
loading and climate conditions a
density. Test configurations are pre
ISO 8375 and any statement abo
 4.2
stiffness related properties of struc
conditional to the corresp
4.1 configuration. In general it is
between the different loading
“material properties” are given cor
4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
the loading direction relative to t
direction of a beam shaped element

Target  Optimized  m
Eurocode Optimized  m ,  G ,  Q

Figure 1: Reliability Index Figure

over different design
1: Reliability Indexsituations alpha
over different for solid timber in bending. The
design
different lines represent different setsalpha
situations of partial safety
for solid factors
timber (Kohler
in bending. Theand Fink, 2012).
different lines represent different sets of partial safety
factors. Figure 2: Different “material properti
The above example demonstrates the validity of the Eurocode 5 (2004) designmode.
on the loading safety
concept for timber load-bearing
The above elements
example under the assumption
demonstrates the validity that
of the parameters given
in Figure 1 represent the therealEurocode
situation5 with sufficient
design accuracy.
safety concept The “material
In the following
for timber properties” ha
it will be dis-
cussed in which ways theload -bearing
actual loadelements under the
load-bearing assumption
behavior that from
derivate statistical properties and when usi
the assumptions
the parameters given how
in Figure criterion
1 represent the deviation affectsin Eq. 1 and applying th
in Table 1. It is demonstrated and quantified the corresponding the re-
real situation with sufficient accuracy. In the safety factor  m , as it is pra
liability of design situations and it is discussed how recent research results might integrated
in the further developed issue of Eurocode 5 (2004).

3 PARTICULARITIES IN TIMBER MATERIAL

MODELLING
In the following section the main issues to enhance the current standards are identified
and discussed. That includes the influence of different material properties in different load
directions, the quality of the grading process, the application of the current safety concept
on non-linear design equations, duration of load effects, effects due to moisture induced
stresses, volume and length effects, reliable and uniform design equations for joints and the
adoption of consequence classes for associated situations.

3.1 Different “material properties”

Timber is a rather complex building material. Its properties are highly variable, spatially and
in time. In structural engineering, material properties of timber are in general understood as
the stress and stiffness related properties of standard test specimen under given (standard)
loading and climate conditions and the timber density. Test configurations are prescribed
in e.g. ISO 8375 (1985) and any statement about stress and stiffness related properties
of structural timber is conditional to the corresponding test configuration. In general it is
distinguished between the different loading modes and “material properties” are given cor-
responding to the loading direction relative to the main fiber direction of a beam shaped
element (Figure 2).

395
INTER / 48 - 102 - 02

bending

compression tension
perpendicular perpendicular

tension/compression
parallel

Figure 2: Different “material properties” dependent on the loading mode.

The “material properties” have different statistical properties and when using the design
criterion introduced before and applying the same partial safety factor γm , as it is practiced
in the Eurocode, the reliability of the corresponding design solutions differ.
The influence of different “material properties” was investigated in Kohler and Fink (2012).
There, the distribution functions and the associated variability for different types of “material
properties” were chosen, as recommended in the Probabilistic Model Code JCSS (2006),
see also Köhler (2006) for background information. The results are summarized in Table 2.
The obtained scatter in partial safety factors suggests a rather differentiated treatment of the
different design situations in future developments of design codes.
The most extreme deviation from the values proposed in the Eurocode γm = 1.30 is ob-
tained for the load case tension perpendicular to the grain γm = 3.05. This also indicated that
if a structural element for this load case is designed with the current safety factor of γm = 1.30,
very low reliability indices, in the order of magnitude of 3.1 are obtained. However, the re-
sults concerning this particular load case have to be considered with special care. In fact
the material capacity under this loading mode is specified by EN 338 (2010) with a nominal
value that does not correspond to the 5%-fractile value taken from the statistical distribution
that is derived from test data for the same loading mode. It is rather a value well below the
5%-fractile value. Furthermore, in best practice timber engineering design this loading mode
at its limit is avoided due to the high sensibility to aspects that are not directly controlled in
design, as e.g. moisture induced stresses and macro and micro cracks in the timber.
On the other hand the current design for compression strength perpendicular to the grain
seems to be rather conservative γm = 1.20. Here, it has to be considered that the conse-
quences that are resulting from failure of the compression strength perpendicular to the grain
are rather low (see also Chapter 3.7). Thus, the conservative approach has to questioned.

Table 2: Calibrated partial safety factors for the resistance, for constant γG = 1.35 and γQ = 1.50 (from
Kohler and Fink, 2012).

Ultimate limit state γm

Bending strength 1.33
Tension strength parallel to the grain 1.40
Tension strength perp. to the grain 3.05
Compression strength parallel to the grain 1.24
Compression strength perp. to the grain 1.20
Shear strength 1.33

396
 INTER / 48 - 102 - 02

3.2 Timber as a graded material

Due to the special way timber material properties are ensured by means of grading in the
production line, special considerations must be made when modeling their probabilistic char-
acteristics. Previous work on this subject is reported in (e.g. Rouger, 1996; Pöhlmann and
Rackwitz, 1981). Further assessment of the probabilistic modelling on the properties of
graded timber material was presented in Faber et al. (2004); Sandomeer et al. (2008). In the
latter references it is reported that the scatter of strength related material properties is highly
sensitive to the grading procedure applied and to the properties of the original ungraded
material. This observation is confirmed by a large experimental campaign that took place
recently in Europe in connection to the Gradewood project3 . Here a large number of graded
samples have been tested and a large between sample variability has been observed. Fur-
thermore it has been shown that it is highly uncertain whether a sample that is graded to
a specific grade actually meets the corresponding requirements in regard to minimum 5%
fractile values of strength properties.
It is continued along the example introduced above, assuming that the grading accuracy
directly affects the coefficient of variation of the timber bending capacity. The material partial
safety factor are calibrated for different grading schemes that correspond to different coef-
ficients of variation in the range from 0.2 – 0.4. The corresponding partial material safety
factors rage between γm = 1.2 − 1.65 depending on the applied grading procedure. These
results suggest a better differentiation of the grading procedure in future design codes. Alter-
natively, if no information about the accuracy of timber grading is utilized a larger coefficient
of variation for representing the bending capacity should be used.

3.3 Non linear design equations

For common design equations a linear comparison of load effects and component resis-
tance as in Equation (1) is not sufficient. One example is the design of slender columns
where strength and stiffness properties and creep effects play an important role for assess-
ing the stability. For the analysis of single members, standards generally give simplified
calculation models that do not require a 2nd order ultimate limit state analysis. However, for
the analysis of more complex systems like unbraced frame structures, a 2nd order structural
analysis is more appropriate and accurate and an alternative design procedure is given e.g.
in the Eurocode (2004). Compared to the simple design format as presented in Eq. (1),
the design equations for slender columns are more complex containing uncertain proper-
ties as strength, stiffness and load eccentricity in non-linear combination. The problem was
addressed in Köhler et al. (2008) and quite uneven reliabilities for different column slender-
ness have been reported. Figure 3 the reliability index of design solutions with different
slenderness-ratio are presented. The different colors correspond to different design frame-
works; I. EN 1995-1-1 (2004), 2nd order method with the stiffness considered as the mean
modulus of elasticity;. II. DIN 1052 (2004), 2nd order method with the stiffness considered
as the 5% fractile of the modulus of elasticity, and; III. EN 1995-1-1 (2004) / DIN 1052 (2004)
according to the so called simplified equivalent length approach. For more details compare
Köhler et al. (2008).
In future code safety formats design strength and stiffness should be clibrated in order to
obtain consistent reliability levels for different design situations.
3 http://www.woodwisdom.net/wp-content/uploads/2014/08/Gradewood_final_report.pdf

397
 Kohler et al. (2008) and quite uneven reliabilities
ending on the applied for different column slenderness have been
INTER
results suggest / 48 - 102reported
a better - 02 (figure).
ng procedure in future
ly, if no information 5.2
er grading is utilized a
on for representing the 5
used.
4.8

Reliability Index
4.6 I.
II.
III.
4.4

4.2

3.8
0 50 100 150 200 250

λ =  eff i
‘
Figure 3: Reliability Index Figure
over slenderness
(Kohler for
et design solutions according to different design formats
al. 2008)
(Köhler et al., 2008).

rated partial material  CIB paper Köhler/Steiger

qualities. 3.4 Duration of load effect and moisture induced stress
The capacity of a timber structural element is highly dependent on the time duration of
3.4. Duration of load effect and moisture
the load effect to which it is exposed to. As an example the capacity of a bending beam
ations continuously loaded is only induced
60% stress
of that of a similar beam exposed to an instant load (Wood,
n equations a linear
1947).
ects and component The capacity of a timber structural element is
Timber is a hygroscopic material, i.e. it adsorbs and desorbs moisture from the surround-
1 is not sufficient. One
ing air. Variations highly
in moisturedependent
content on the surrounding
in the time duration air of
will,the load
with a corresponding time
lack, lead
lender columns where to effect
variations in to which
moisture it is
content exposed
in the to. E.g.
timber, the
this capacity
affects the mechanical proper-
ties ofeffects
erties and creep the timber butof more importantly
a bending beamit continuously
will induce stresses
loaded dueistoonly
shrinkage and swelling
alongside the moisture gradients in the timber. These moisture induced stresses have been
assessing the stability. 60% of thatinof
a matter of intensive discussion theatimber
similar beam exposed
engineering community to in anthe last years.
e members, standards
Both, duration of instant load, and
load effect as itmoisture
was alreadyinducedobserved
stresses byare(?).highly relevant phenom-
alculation ena
models that
to take into account Timber
in structural
is a hygroscopic material, i.e. phenomena
design. They are also challenging it since the
er ultimateunderlying
limit state physicaladsorbs
mechanisms andaredesorbs
not fully understood
moisture andfrom empirical theevidence is scarce.
However,
he analysis of more in practical
surrounding air. Variations in moisture contentofinmoisture on the du-
design, as in the Eurocode 5 (2004), the effect
ration of load effect is considered with the joint modification factor kmod which is given for
different climate exposures in design codes. Values for this factor are prescribed in a matrix
for three different so-called service classes, i.e. different climate scenarios, and five different
6
load classes, i.e. load scenarios.
This format appears to be oversimplified and further research and enhancement of the
level of detail in structural design should be developed.

3.5 Volume and length effects

One major topic that is continuously discussed within the research community is the appro-
priate representation of size effects on strength in solid timber. For most loading modes as
tension parallel or perpendicular to grain, shear or bending, timber predominately presents
brittle failure behaviour. A (perfect) brittle material is defined as a material that fails if a single
particle fails (see e.g. Bolotin, 1969). The strength of the material is thus governed by the

398
 INTER / 48 - 102 - 02

strength of the ‘weakest’ particle; therefore the model for ideal brittle materials is also called
the weakest link model (Weibull, 1939). This model was applied to the different failure modes
in timber, the model parameters have been calibrated based on experimental evidence on
the different failure modes. A literature review can be found e.g. in Kohler et al. (2013).
There it is concluded that the size effects in timber are better represented with a model that
takes into account the multi scale variability of structural timber and a corresponding model
framework is suggested.
In present code formats size effects are often not completely taken into account or ne-
glected. This is particularly critical when large scale engineered timber sections are used in
modern timber construction. In a revision of the codes this aspect should earn appropriate
attention and current research results should be implemented.

3.6 Joints
For timber structures, the structural performance depends to a considerable part on the
connections or joints between different timber structural members; joints can govern the
overall strength, serviceability and fire resistance. Despite their importance timber joint de-
sign frameworks are not based on a consistent basis compared to the design regulations of
timber structural components.
Explanations for this difference in progress of design provisions for members and joints
can be found in the relative simplicity of characterizing mechanical behaviour of members, as
compared to connections. A diversity of joints types is used in practice and these types have
infinite variety in arrangement. This usually precludes the option of testing large numbers of
replicas for a reliable quantification and verification of statistical and mechanical models. The
main and most important group of joints corresponds to the joints with dowel type fasteners,
i.e. joints with dowels, nails, screws and staples belong to this group.
Different failure modes can be observed for dowel type fastener joints and the modes
are partly captured by a simple mechanical model based on the works of Johansen (1949);
Meyer (1957). These models build the basis for the current European design framework
for dowel type connectors in the Eurocode. However, different failure modes correspond
to different failure behavior and consequences (brittle or ductile). In Köhler (2006) is has
also been observed that model uncertainty and model bias for the different failure modes is
significantly different. This is not considered in the current version of the European design
standard and should be subject for further investigation.

3.7 Consequence classes

In the previous chapter it was mentioned that different failure modes in dowel type fastener
joint lead to different magnitudes of consequences. This is in principle true for all failure
modes in timber structure. In Chapter 3.1 different failure modes of timber components have
been compared to the same target reliability, implying that the consequences for all failure
modes are classified uniformly. However, if a failure scenario for tension or bending failure
is visualized and compared with a typical failure scenario for compression perpendicular to
the grain, it might be agreed that the consequences are quite different and correspondingly
the target reliability should be defined separately for the different cases.

399
INTER / 48 - 102 - 02

4 CONCLUSIONS
Timber will play an important role in the future developments towards a more sustainable
building sector. However, many stakeholders are still skeptical when it comes to the tech-
nological maturity of the material, especially compared to concrete and steel. The structural
design regulations in general can be seen not only as the main interface connecting the
state of knowledge in the engineering research community with the implementation of the
real build environment; design standards are also the precondition for the implementation of
building material on a high technological level.
In the present paper the major challenges for the future development of timber design
standards have been highlighted from a European perspective; i.e. taking the Eurocodes as
references. The challenges are hereby related to both, the further development of the knowl-
edge basis for the behavior of timber in structures and the implementation of this knowledge
into practicable rules in the future standards.

References
Bolotin, V. V. (1969). Statistical Methods in Structural Mechanics. Holden-Day Series in
Mathematical Physics.
DIN 1052 (2004). Bemessungsregeln für Holzkonstruktionen. Normenausschuss Bauwesen
(NABau) im DIN (Deutsches Institut für Normung e.V.), Berlin, Germany.
EN 1990 (2002). Eurocode 0: Basis of structural design. European Committee for Stan-
dardization, Brussels, Belgium.
EN 1995-1-1 (2004). Eurocode 5: Design of timber structures – Part 1-1: General – Common
rules and rules for buildings; German version. European Committee for Standardization,
Brussels, Belgium.
EN 338 (2010). Structural timber – Strength classes; German version. European Committee
for Standardization, Brussels, Belgium.
Faber, M. H., Köhler, J., and Sørensen, J. D. (2004). Probabilistic modelling of graded timber
material properties. Structural safety, 26(3):295–309.
Faber, M. H. and Sørensen, J. D. (2003). Reliability based code calibration – The JCSS
approach. In Applications of statistics and probability in civil engineering : proceedings
of the 9th International Conference on Applications of Statistics and Probability in Civil
Engineering, San Francisco, USA, July 6–9, pages 927–935.
ISO 8375 (1985). Solid timber in structural sizes: Determination of some physical and
mechanical properties. International Standards Organization, Geneva, Switzerland.
JCSS (2001). Probabilistic Model Code Part I - Basis of Design. http://www.jcss.byg.
dtu.dk/Publications/Probabilistic_Model_Code.
JCSS (2006). Probabilistic Model Code Part III - Resistance Models (3.05 Timber). http:
//www.jcss.byg.dtu.dk/Publications/Probabilistic_Model_Code.
Johansen, K. (1949). Theory of timber connections. In International Association of Bridge
and Structural Engineering, Publication No. 9, pp. 249-262, Bern, Switzerland.
Köhler, J. (2006). Reliability of Timber Structures. PhD thesis, ETH Zurich, Zurich, Switzer-
land.
Köhler, J., Andrea, F., and René, S. (2008). On the role of stiffness properties for ultimate
limit state design of slender columns. In Proceedings of the 41th Meeting, International
Council for Research and Innovation in Building and Construction, Working Commission
W18 - Timber Structures, St. Andrews, Canada, CIB-W18, Paper No. 41-1-1.
Kohler, J., Brandner, R., Thiel, A., and Schickhofer, G. (2013). Probabilistic characterisation

400
 INTER / 48 - 102 - 02

of the length effect for parallel to the grain tensile strength of central european spruce.
Engineering Structures, 56:691–697.
Kohler, J. and Fink, G. (2012). Reliability Based Code Calibration of a Typical Eurocode 5
Design Equations. In Quenneville, P., editor, World Conference on Timber Engineering
2012 (WCTE 2012) : The Future of Timber Engineering : Auckland, New Zealand : 15-19
July 2012. Volume 4, pages 99–103, Red Hook, NY. Curran Associates, Inc.
Kohler, J. and Fink, G. (2015). Aspects of code based design of timber structures. In Pro-
ceedings of the 12th International Conference on Applications of Statistics and Probability
in Civil Engineering, Vancouver, Canada on July 12-15, 2015.
Meyer, A. (1957). Die Tragfähigkeit von Nagelverbindungen bei Statischer Belastung. Holz
als Roh- und Werkstoff, 15(2):96–109.
Pöhlmann, S. and Rackwitz, R. (1981). Zur Verteilungsfunktion der Festigkeitseigenschaften
bei kontinuierlich durchgeführter Sortierung. Materialprüfung 23, Hanser, Munich, Ger-
many.
Rouger, F. (1996). Application of a modified statistical segmentation method to timber ma-
chine strength grading. Wood and Fibre Science, 28(4).
Sandomeer, M. K., Köhler, J., and Faber, M. H. (2008). Probabilistic output control for
structural timber - modelling approach. In Proceedings of the 41th Meeting, International
Council for Research and Innovation in Building and Construction, Working Commission
W18 - Timber Structures, St. Andrews, Canada, CIB-W18, Paper No. 41-5-1.
Weibull, W. (1939). A statistical theory of strength of materials. Nr. 151, Royal Swedish
Institute for Engineering Research.
Wood, L. (1947). Behaviour of wood under continued loading. Eng. News-Record,
139(24):108–111.

401
INTER / 48 - 102 - 02

Discussion

The paper was presented by G Fink

K Ranasinghe commented that simplicity would be needed by practicing engineers

and asked the group whether we know who the target audience are for the standards.
G Fink responded that the user of the standards is our aim; however, they needed to
look deep into the problem to get the basic understanding to come up with provisions.
H Blass commented that one would have to find the tools to translate these findings
to the engineering world and practicing engineers would have deadlines and budget
constraints and would use simple tools.
I Smith commented that sizing members would have easy ways but structural system
considerations would be more challenging.
F Lam commented that the most needed area for reliability based design considera-
tion would be connection design.
S Franke agreed that one needed a deeper understanding to transfer the information
into simple equations; however, PhD students could be involved in the useful step of
documentation of the steps from the deep understanding to the simple equation. G
Fink responded that PhD students needed to be more focused and work on a topic in
depth rather than looking into too many areas.
G Schickhofer agreed that COV should be considered but how to do this in a standard
would be the question. G Fink agreed that it would not make sense to have different
γm values to consider COV. F Lam added that in Canada, a reliability normalization
procedure is available to consider different COVs in a reliability based approach.
U Kuhlmann suggested only one γm and provide calibration factor for γm and docu-
ment clearly where these factors come from.
P Dietsch and G Fink discussed the aim of Eurocode 5 and the target audience. The
fact that practicing engineers not familiar with wood might complain that the code
would be too complicated.

402
403
404
 INTER / 48 - 102 - 03

The Reality of Seismic Engineering in a

Modern Timber World

Tobias Smith, University of Canterbury, New Zealand

Daniel Moroder, University of Canterbury, New Zealand
Francesco Sarti, University of Canterbury, New Zealand
Stefano Pampanin, University of Canterbury, New Zealand
Andrew H. Buchanan, University of Canterbury, New Zealand

Keywords: Capacity design, ductile timber, seismic engineering, timber over‐strength

Abstract
While timber structures are rapidly gaining in popularity, modern timber engineer‐
ing codes are struggling to keep up with the rapid advancement of timber tech‐
niques, technologies and philosophies. One of the most evident areas of this strug‐
gle is in seismic design which combines timber engineering with the equally rapidly
developing subject of earthquake resistant design.
This paper critically examines several key aspects of seismic design in the context of
the modern timber world and discusses how to improve the inclusion of basic prin‐
ciples into international design codes. The items presented have been drawn from
the authors’ experience during their contribution to the writing of the new seismic
section of the New Zealand timber design code.

1 Introduction
The use of timber in the construction of large scale structures has increased signifi‐
cantly in recent years. This trend has created structures which extend beyond the
scope of traditional timber design and therefore beyond codified design approaches.
This is particularly true for seismic design as the potential for mass timber construc‐
tion increases, thanks to the recent advances of timber technologies and calculation

405
INTER / 48 - 102 - 03

methods which have principally been developed for non‐seismic applications in low
seismic hazard countries.
Recent innovations in engineered wood products have reduced the variability be‐
tween timber members and have allowed designers to assign higher loads. This reli‐
ance on smaller and higher performing members coupled with the increase in build‐
ing bay length and building height has significantly increased the consequences of
failure which could lead to structural collapse.
This paper summarises the general principles and aims of seismic design of struc‐
tures. The implications of seismic design principles in relation to timber structures
are then discussed. Finally the paper suggests codification of these principles and
identifies gaps still needing to be addressed. Although these principles are not new
(Blass et al. 1994), it is imperative that they are carefully considered in a modern
timber world.

2 The Aim of Seismic Design

Earthquake engineering and seismic design have the principal goal of ensuring a cer‐
tain response when a structure is subjected to a predetermined level of ground mo‐
tion. Traditionally, both the level of ground shaking and the accepted level of re‐
sponse are set by seismic design codes.
Although the estimated likely level of ground shaking is material independent, the
response of a structure to ground motion is dependent on structural form that is of‐
ten dictated by the material used.

2.1 The role of non‐linearity

A very stiff building designed for only elastic response will be subjected to very high
accelerations and therefore high seismic forces. High seismic forces result in expen‐
sive buildings, and high accelerations lead to high levels of damage to the building’s
contents. Engineers have two weapons against these high accelerations: ductility
and damping as shown in Figure 1. Both properties rely on the non‐linear behaviour
of materials or structural elements.
Displacement ductility is normally defined as the ratio of the ultimate displacement
(defined by the fracture of the material or by a defined displacement limit) to the
displacement at yielding, or in more general terms to the point where a distinctive
change in stiffness can be observed. Ductility is therefore linked to non‐linear behav‐
iour, which can either be elastic (i.e. through rocking) or inelastic (i.e. material yield‐
ing). Damping is the ability of a material or structural element to dissipate energy; in
seismic design this is normally in the form of plastic yielding (structural damping),
friction (Coulomb damping) or viscous drag (viscous damping).

406
 INTER / 48 - 102 - 03

Figure 1: Effects of ductility and damping on the lateral load response of a structure.
Non‐linearity also has a further function in seismic design, that of ensuring resilience
by allowing a structure to deform beyond its design capacity. Seismic hazard, as de‐
fined by codified design methods, is based on probability and as such a likelihood of
exceedance exists. This was demonstrated in the recent seismic sequence in Christ‐
church, New Zealand where seismic loading in some areas of the city was over twice
the predicted design values (Bradley et al. 2011).

2.2 Representation of non‐linearity in codified seismic design

All modern design codes for seismic areas use a force‐based design approach. This
approach requires an assumption to be made regarding the initial period of vibra‐
tion of the structure and the amount of non‐linear displacement (normally defined
as ductility) that the structure will possess at its ‘performance point’. The perfor‐
mance point is an idealization which represents the behaviour of the structure when
subjected to the design spectrum acceleration.
Traditional seismic design assumes that the provision of some ductility will also pro‐
vide the structure with hysteretic damping. It is worth noting however that ductility
influences seismic response separately from the influence of damping, as shown in
Figure 1.
2.2.1 Structural classification
The response of a structural system to lateral loading is classified as being either
ductile or brittle. Brittle Structures (µ = 1) are designed for elastic behaviour and ex‐
hibit brittle failure under any loading greater than the Ultimate Limit State (ULS)
loading. Example structures include, but are not limited to, moment‐resisting frames
or portals with glued knee joints, cantilevered timber poles, and all structures where
capacity design principles are not complied with, such that brittle failure of a struc‐
tural element or connection precedes the activation of the potential ductile ele‐
ments.
Ductile structures are those in which non‐linear behaviour is possible. These are
often split into subcategories: Nominally Ductile, Limited Ductile and Fully Ductile.
Nominally Ductile Structures (µ = 1.25) are capable of sustaining small inelastic

407
INTER / 48 - 102 - 03

material deformations in predefined ductile areas. Example structures include but

are not limited to timber arch structures, walls with screw fixed sheathing, joints
which rely on embedment‐only failure modes, fully threaded screws in tension, and
carpentry joints in compression.
Limited Ductile Structures (1.25 < µ ≤ 3) are designed to exhibit a ductile failure un‐
der any seismic loading greater than the ULS load levels. Significant inelastic materi‐
al deformations are required in identified non‐linear elements, to limit the earth‐
quake design forces. Example structures include but are not limited to nailed ply‐
wood shear walls, well designed ductile moment frames, and braced frames with
buckling restrained braces which are designed to yield in both tension and compres‐
sion.
Similar to Limited Ductile Structures, Fully Ductile Structures (µ > 3) are designed to
exhibit an even more ductile failure mode under seismic loading greater than the
ULS load levels. Inelastic material deformations beyond those of a Limited Ductile
system are required.
2.2.2 Seismic reduction factors
Seismic hazard is represented in modern seismic design codes through the presenta‐
tion of a single‐degree‐of‐freedom elastic site hazard acceleration spectrum. As‐
sumed approximations summarized in Figure 2 provide the relationship between the
seismic response of elastic and non‐linear systems which allow for the reduction of
elastic accelerations by using a spectrum representing non‐linear response. It now
becomes possible to determine the required yield strength of the structural system
based on the codified elastic hazard spectrum, the predicted system ductility, and
the estimated period of vibration of the structure to be designed.
It is important to note that all of these modifications are based on the idealization
that the force‐deformation relationship is bi‐linear elastic perfectly‐plastic (Figure
2). When the force is between the positive and negative yield values (‐Fy and Fy) the
stiffness is equal to the initial elastic stiffness, k, but when the force is equal to the
yield force (either ‐Fy or Fy) the stiffness is equal to zero (i.e. perfectly plastic). For
this idealized system, both ductility and damping contribute in the reduction of
seismic forces.
Moderate‐to‐long period structures: Equal displacement approximation
For extremely long period structures (e.g. 30 seconds, as an extreme example) the
maximum system displacement will equal the maximum ground displacement
(Figure 2a). The reason for this is that the system is so flexible that when the ground
moves, the mass of the system essentially stays in its initial position. Hence, this
maximum system displacement will be the same whether the system responds elas‐
tically or in‐elastically. For moderate‐to‐long period structures the above explana‐
tion does not hold strongly, but practically speaking, the inelastic and elastic dis‐
placements are similar giving the equal displacement approximation.

408
 INTER / 48 - 102 - 03

For moderate‐to‐long period structures the equal displacement approximation

means that the reduction in seismic force (Ry = Fu/Fy) is equal to the design dis‐
placement ductility (µ = Δu/Δy).
Short period structures: Equal energy approximation
For structures with short‐vibration periods the peak inelastic displacements tend to
be larger than the peak displacement of the corresponding linear elastic system.
Therefore the equal displacement approximation for moderate‐to‐long period struc‐
tures is inappropriate for short‐period structures.
An alternative approach is to compare the energy in both the elastic and inelastic
structures at their peak displacement. Since at the peak displacement the velocity is
zero, then at this point there is no kinetic energy and only potential strain energy,
which can be computed as the area beneath the force‐displacement relationship.
Based on the assumption that the two areas are equal, it is possible to calculate the
reduction in seismic force as a function of the system displacement ductility (Ry =
2 1). This assumption is independent of the building’s period of vibration
which is not completely accurate. For this reason some loading codes include build‐
ing period in the application of the equal energy approximation.

a)

Figure 2: a) Equal displacement approximation and b) equal energy approximation allowing for non‐
linear system response

2.3 The role of capacity design and over‐strength in ensuring ductility

Capacity design philosophy developed in New Zealand in the nineteen seventies for
the seismic design of reinforced concrete structures is the most common way of en‐
suring structural ductility. Capacity design protects all brittle (or less ductile) parts of
a structure by understanding the load exerted upon them from the selected ductile
elements, even at large displacements.
This is similar to loading a chain where the maximum capacity is dictated by the
weakest link, when the weakest link is ductile, the overall chain is ductile. In order to
ensure the capacity of all the brittle elements within the lateral load resisting system
is adequate, their strength demand is set equal to the “over‐strength” of the ductile

409
INTER / 48 - 102 - 03

elements. Paulay et al. (1975) state that the over‐strength should take into account
all factors that may increase the strength of the weakest element or connection,
however, there is little consensus on a common approach of how to determine
over‐strength values and what factors should be included for the design of ductile
timber structures.

3 Non‐linearity and Timber Design

Wood exhibits ductile behaviour in compression parallel to the grain because even
though the fibres buckle, they still support a limited load. If member buckling can be
excluded, timber in compression can be classified as ductile. Clear wood is strong
but brittle in tension, with no non‐linear behaviour.
Timber in bending will display limited non‐linear capability when loaded monoton‐
ically, but has no ductility under reversed cyclic loading. .

3.1 Sources of non‐linearity in timber structures

Steel is one of the few structural materials capable of sustaining repeated non‐linear
loading cycles without strength degradation or brittle fracture. As in concrete struc‐
tures, the use of steel components is the most common method of providing non‐
linear behaviour in timber structures.
Dowel type fasteners have traditionally been used to obtain ductility in timber struc‐
tures. Small diameter fasteners (nails, screws or rivets, Figure 3a) tend to be more
ductile than large diameter fasteners (bolts, dowels and coach screws, Figure 3b),
because steel yielding is the likely governing failure mode. More recently, specifical‐
ly designed steel elements have been used to provide non‐linearity at beam‐column
and foundation connections (Figure 3c).
The European Yield Model (Johansen 1949) for small and large dowel type fasteners
defines seven potential non‐linear failure modes for timber‐to‐timber or steel‐to‐
timber connections under monotonic loading as shown in Figure 4. These failure
modes are classified as ‘yielding’ due to their ability to provide non‐linear response
and therefore some ductility under monotonic loading. As shown in Figure 4 the
non‐linearity comes from the crushing of timber (a, b, c, g and h) or the combination
of timber crushing and fastener yielding (d, e, f, j and k).
For a group of several fasteners making up a joint, failure of the fastener group may
occur through block tear‐out for small dowels or row shear and group tear out for
large dowels. The fastener group must be designed so that none of these brittle fail‐
ure mechanisms occur before the development of fastener yielding, if any ductility is
to be achieved.

410
 INTER / 48 - 102 - 03

(a) (b) (c)

Figure 3: a) Small (Versuche zur DIN 1052 2010) and b) large dowel type fasteners (Bejtka 2005), c)
specifically designed ductile steel hold downs as part of a wall system

Figure 4: ‘Yielding’ failure modes combining timber crushing (shown in red) and steel yielding (shown
in green)

3.2 Seismic non‐linearity and damping

Seismic loading is cyclic and random in amplitude, therefore the ability of a connec‐
tion to provide ductile seismic response must be evaluated within the context of the
nature of this loading. Figure 4 showed the yielding failure modes of small and large
dowel type connections and how the non‐linear behaviour of these connections

411
INTER / 48 - 102 - 03

consists of timber crushing and in some cases flexural yielding of the fastener.Figure
5: Figure 5 shows schematically the cyclic behaviour of these failure modes. When
non‐linear behaviour is only through timber crushing, the timber can only yield in
the first cycle, creating severely pinched force‐displacement loops and a ‘sloppy’
connection similar to a braced structure with tension‐only yielding. Figure 5 also
shows how the amount of hysteretic energy dissipation in such a connection is sig‐
nificantly reduced. When timber crushing is combined with steel yielding the curves
remain pinched, but the cyclic characteristics of the steel yielding mean that some
ductile capacity always remains, which also creates more favourable hysteretic be‐
haviour.

Figure 5: Comparison of hysteretic behaviour for cyclic behaviour of ‘yielding’ fastener failure modes

3.3 Global and local ductility and the displacement paradox

Design to control displacements is becoming increasingly important in the seismic
design of buildings, to reduce the potential costs of both structural damage and
non‐structural damage. The damage potential of displacement is far beyond the
damage potential of force; however, displacements are too often overlooked
(Buchanan et al. 2015).
Section 2.1 discussed how non‐linear behaviour of a structural system is typically
quantified using the parameter of ductility, the ratio between the ultimate dis‐
placement of a structural system and the point in which yield has been deemed to
occur. Section 3.1 and 3.2 described how this ductility is normally provided by steel
fasteners at beam‐column, wall‐foundation or column‐foundation connections. It is
therefore evident that the ductility of a timber building is governed by the ductility
of its connections. The seismic demand, however, is governed by the ductility of the
entire seismic system. Therefore the elastic displacements of the structure must al‐
so be considered in design creating a displacement paradox.
The displacement paradox occurs because, on the one hand, designers need to con‐
trol lateral displacements to reduce damage and the possibility of structural col‐
lapse, but on the other hand, they must provide enough displacement to ensure
non‐linear response if ductility has been used in the calculation of seismic forces (i.e.
using the reduced design acceleration spectrum). These two objectives sometimes
clash, creating a paradox. The displacement paradox can be a particular problem in

412
 INTER / 48 - 102 - 03

timber structures because they tend to have larger elastic displacements than steel
or concrete structures, giving a smaller displacement window in which to provide
the required ductility. The displacement paradox often creates a severe mismatch
between the intended ductility of a building and its actual response, especially if
there are large linear elastic displacements before any non‐linearity occurs. A large
assumed ductility means that the structure is required to undergo very large dis‐
placements, creating large P‐Delta and other undesired effects.

4 The Future of Seismic Timber Codes

New timber technologies and construction methods are creating new and exciting
opportunities for timber in construction. Practical experience is being added to a
wealth of new research significantly advancing our understanding of the seismic re‐
sponse of timber structures. Design codes are attempting to keep up with these de‐
velopments. This section discusses the relationship between the response of timber
structures as discussed in the previous sections and the intent of the seismic loading
codes.

4.1 Ensuring resilience in timber structures

Non‐linearity is used to reduce the seismic loading on a structure and to increase
the resilience of a structure against collapse under loads which are higher than the
design loading.
In modern seismic design using any structural materials, specific locations of non‐
linearity are identified and protected through capacity design. In well‐designed con‐
crete and steel structures, however, most regions outside the defined non‐linear
zones still possess some non‐linear capacity. This is in contrast to timber buildings
where the timber materials outside the identified non‐linear zones are brittle. In the
existing New Zealand Earthquake Loading Code (NZS 1170.5 2004) a structure of lim‐
ited ductility has a 25% reduction in seismic loading but structural members in gen‐
eral need not adhere to capacity design principles. The resilience of such a lateral
load resisting system made of timber is not adequate and as such it is recommended
that although the connection may be designed for µ = 1.25 the surrounding timber
members (those which would be capacity protected in a ductile structure) should be
designed with some reserve strength, by using demands derived through an elastic
assumption (µ = 1).
The current draft amendment to the New Zealand Earthquake Loading Code will fur‐
ther impose that earthquake actions on brittle structures be increased by 50%. This
will apply to all materials including timber. This means that all brittle structures will
have to be designed to resist Maximum Credible Event (MCE) loading, representa‐
tive of an earthquake with a return period of 1 in 2500 years.

413
INTER / 48 - 102 - 03

4.2 Ensuring desired response from timber connections

Timber is inherently a brittle material and non‐linearity must come from its connec‐
tions. These connections must therefore be designed to provide the required level
of non‐linear response without brittle failure.
It is clear that the pinched hysteretic loops shown in Figure 5 for a timber joint with
combined fastener yielding and timber crushing shown in Figure 3 and Figure 4 does
not match well to the elastic perfectly‐plastic assumption of the equal displacement
approximation or the equal energy approximation. The match is even worse for tim‐
ber crushing with no yielding of steel, so it is recommended that connections with
only timber crushing mechanisms should not be considered as ductile for seismic
design.
Recent research (Franke et al. 2013; Schoenmakers et al. 2013; Zarnani et al. 2014a;
Zarnani et al. 2014b; Jensen et al. 2015; Zarnani et al. 2015) has enabled the accu‐
rate calculation of the brittle failure modes of doweled connections. The formulas
allowing this calculation are also currently being implemented in new design codes,
in some cases replacing minimum edge distances.
Following the occurrence of yield, reached at the yield failure strength Nα,y,y the sys‐
tem will displace until the required level of non‐linear behaviour, defined through
displacement ductility, is reached, as shown in Figure 6. In order for a timber struc‐
ture to continue to display good seismic behaviour, the yielding failure mode must
continue to govern the connection behaviour until it has developed its ultimate
yielding failure mode strength. To ensure this, all brittle failure modes in the con‐
nection must be excluded. Capacity design principles, as used to protect any other
element outside the potential ductile zone, need to be applied for design of any
components of the connection which could have brittle failure modes.

Figure 6: Strength hierarchy within the joint fastener group to ensure non‐linearity

4.3 Ensuring desired response from timber structures

Section 3.3 recommended that a dowel type connection be designed to ensure the
desired amount of non‐linear capacity. These connections however are part of a

414
 INTER / 48 - 102 - 03

wider lateral load resisting system and it is the non‐linearity (ductility) of this entire
system that was used in the determination of the seismic demand. It is therefore
crucial that the structural beams, columns and walls or connections which are not
intended to provide non‐linear response: 1) have adequate capacity to avoid failure
and 2) are adequately stiff to avoid excessive displacement before the required non‐
linearity is achieved.
4.3.1 Ensuring adequate strength of brittle elements which are designed to remain
elastic
As stated in Section 2.3, over‐strength is required to ensure that the intended non‐
linear element is the weakest link in the lateral load resisting system. Although sev‐
eral values for over‐strength are available in literature (BRANZ 1999; Popovski et al.
2005; Jorissen et al. 2011; Sustersic et al. 2011; Schick et al. 2013; Brühl et al. 2014),
the definition of what these values account for is unclear. Definitions of over‐
strength normally include one or more of the following items:
1. The difference between the demand on a joint and the capacity of the joint
(over‐capacity or over‐design);
2. Material safety factors;
3. Strain hardening and the rope effect;
4. Statistical distribution of materials or joint capacities (characteristic strength
values are given as the lower 5th percentile strength);
5. The difference between the real and calculated strength of a joint (i.e. exper‐
imental vs. European Yield Model values);
6. Mechanical effects or hidden reserves (such as friction).
Item 1 is based on the designer’s choice and can easily be quantified. Items 2 and 3
are allowed for by designing the capacity‐protected structural members for the fas‐
tener group’s ultimate yield strength, Nα,y,u. Item 4 is accounted for by calculating
the upper characteristic value of the ultimate yield strength (i.e. calculating the ul‐
timate yield strength using upper 95th percentile fastener yield and embedment
strength).
Items 5 and 6 remain to be considered in the calculation of over‐strength values for
the capacity‐protected members. Values for the partial over‐strength factor regard‐
ing item 5 for dowel type connections were found to be between 0.94 (Brühl et al.
2012) and 1.18 (Jorissen et al. 2011), suggesting that a value of 1.0 could be used
however further research into these partial factors is required.
4.3.2 Respecting displacement limits and the displacement paradox
As discussed in Section 3.3 a paradox can occur in seismic design where in order to
obtain required levels of non‐linear behaviour, excessive displacements may be re‐
quired. Typical design codes place limits on maximum inter‐storey drifts at ULS de‐
sign levels, for example a limit of 2.5% is imposed by the New Zealand Earthquake
Loadings Standard (NZS 1170.5 2004) while a limit of 2.0% generally exists in Ameri‐

415
INTER / 48 - 102 - 03

can codes with some allowance for building importance and period (ASCE 7‐10
2010). Eurocode 8 defines maximum inter‐storey deflection based on a stability co‐
efficient which includes P‐delta actions (Eurocode 8 2004). Often these deflection
limits are reduced even further to limit the possibility of building pounding or to
prevent damage to non‐structural elements, which further reduces the ability of the
designer to provide the required ductility.
The entire structural system, including all elastic deformation contributions, must be
considered in the evaluation of non‐linear behaviour and the definition of ductility.
This means that horizontal deflection calculations must include not only the connec‐
tions providing non‐linearity but also the entire structural system containing the
connections (accounting for the stiffness of all structural members and all other
elastic connections).
For example, a 5 storey, limited ductile, moment resisting frame (moderate‐to‐long
period, µ = 3) having a maximum allowable drift of 2.0%, must yield at 0.67% drift.
This means that all horizontal displacements deriving from beam, column, and elas‐
tic connection deformations must be less than this value, which will require a very
stiff structure to be provided.

4.4 The use of the mean material properties in capacity design

If it can be assumed that a correlation exists between material properties responsi‐
ble for the yielding connection strength (timber density) and the brittle failure
modes of the connection (i.e. fracture energy, tension perpendicular to grain
strength) or the strength of the structural element (i.e. tension, compression or
bending strength), then a similar statistical distribution can be assumed. This fact
means that using the 95% material properties is overly conservative in some areas
of the structure. However, to still guarantee a minimum level of safety, the mean ul‐
timate yield strength Nα,y,u (i.e. calculated using the mean density) could be used in‐
stead (Priestley et al. 2007). As shown in Figure 7 this correlation can be applied to
the brittle connection failure modes, as well as the brittle element where the ductile
connection is located. All elements and connections outside the element where the
ductile connection is located need to be verified against the upper characteristic ul‐
timate yield strength.

Figure 7: Material characteristics to be used in capacity design

416
 INTER / 48 - 102 - 03

4.5 Higher mode effects

Because of the more flexible nature of timber structures and the potentially delayed
onset of yielding, higher modes can significantly affect the dynamic behaviour of
taller timber structures. Especially wall structures can experience much higher sto‐
rey shears and moments as well as peak floor accelerations when compared to val‐
ues determined by methods based on a the first mode approximation (such as the
equivalent static analysis). Until dynamic amplification factors for timber structures
are available, more advanced methods like response spectrum analysis or time his‐
tory analysis should be applied to taller timber buildings.

5 Conclusions
Assisted by the development of better materials and connections, timber buildings
are pushing beyond traditional boundaries of height and size. Timber as a structural
material is moving into structural types which have been traditionally dominated by
reinforced concrete and steel. This paper has studied several of the foundation prin‐
ciples of modern seismic design, showing that some of the basic assumptions in cal‐
culating both seismic demand and structural resilience to seismic load are not as
easily applicable to timber construction as they are to reinforced concrete or struc‐
tural steel.
The following recommendations are therefore made for new codes specifying the
seismic design of modern timber buildings:
Structural design
1. Brittle structures should be designed to resist MCE (1/2500 year) loading;
2. Brittle elements in limited ductility structures (µ = 1.25) should be designed
for elastic seismic demand (µ = 1);
3. When assessing the ductility of a structural system, all components of dis‐
placement must be considered, including all local displacements and rotations
resulting from deformations of the elastically responding elements and all the
connections along the seismic load path;
4. For buildings taller than four stories, higher mode effects should be consid‐
ered through special study, this remains a significant research gap in the im‐
plementation of tall timber buildings.
Connection design
1. The capacity of a joint in an intended ductile zone should be evaluated at the
yielding state (Nα,y,y);
2. Joints governed by crushing only failure modes should not be used in connec‐
tions intended to provide damping;

417
INTER / 48 - 102 - 03

Capacity design
1. No consensus on a unique definition of over‐strength factors is currently
available (Moroder et al. 2014) and therefore values for different connection
types and different structural types are still missing. A partial over‐strength
factor of 1.2 is recommended, allowing for hidden reserves and the difference
between analytical models and real behaviour, until further research is carried
out;
2. The capacity of all brittle elements and elastically responding connections
should exceed the capacity of the ultimate yielding failure mode (Nα,y,u) of the
non‐linear connections, calculated using the upper limit (95%ile) material val‐
ues multiplied by the partial over strength factor;
3. The capacity of all potential brittle failure modes within the connection should
exceed the capacity of the ultimate yielding failure mode (Nα,y,u) of the non‐
linear connections, calculated using the mean (50%ile) material values multi‐
plied by the partial over strength factor.

6 References
ASCE 7‐10 (2010). Minimum design loads for buildings and other structures. Reston, Va,
American Society of Civil Engineers.
Bejtka, I. (2005). Verstärkung von Bauteilen aus Holz mit Vollgewindeschrauben,
Versuchsanstalt für Stahl, Holz und Steine (VAKA)
Blass, H.J., Ceccotti, A., Dyrbye, C., Gnuschke, M., Hansen, K.F., Nielsen, J., Ohlsson, S., Parche,
M., Reyer, E., Stieda, C.K.A., Vergne, A., Vignoli, A., Yasumura, M. and Dolan, J.D. (1994).
Timber structures in seismic regions RILEM state‐of‐the‐art report. Materials and
Structures 27(3): 157‐184.
Bradley, B.A. and Cubrinovski, M. (2011). Near‐Source Strong Motions Observed in the 22
February 2011 Christchurch Earthquake. Bulletin of the New Zealand National Society for
Earthquake Engineering 44(4).
BRANZ (1999). Evaluation and Test Method EM1: Structural Joints ‐ Strength and Stiffness
Evaluation. Porirua City, New Zealand.
Brühl, F. and Kuhlmann, U. (2012). Requirements on ductility in timber structures. CIB Working
Commission W18 ‐ Timber Structures, Växjö, Sweden.
Brühl, F., Schänzlin, J. and Kuhlmann, U. (2014). Ductility in Timber Structures: Investigations on
Over‐Strength Factors. Materials and Joints in Timber Structures. S. Aicher, H. W.
Reinhardt and H. Garrecht, Springer Netherlands. 9: 181‐190.
Buchanan, A.H. and Smith, T. (2015). The Displacement Paradox for Seismic Design of Tall
Timber Buildings. New Zealand Conference of Earthquake Engineering. Rotorua, New
Zealand.
Eurocode 8 (2004). EN 1998‐1:2004/AC:2009 Design of structures for earthquake resistance.
Part 1: General rules, seismic actions and rules for buildings, European Committee for
Standardization.
Franke, B. and Quenneville, P. (2013). Design Approach for the Splitting Failure of Dowel‐type
Connections Loaded Perpendicular to Grain. CIB Working Commission W18 ‐ Timber
Structures, Vancouver, Canada.

418
 INTER / 48 - 102 - 03

Jensen, J.L., Quenneville, P., Girhammar, U.A. and Källsner, B. (2015). Brittle Failures in Timber
Beams Loaded Perpendicular to Grain by Connections. Journal of Materials in Civil
Engineering.
Johansen, K.W. (1949). Theory of timber connections. International Association for Bridge and
Structural Engineering(9): 249‐292.
Jorissen, A. and Fragiacomo, M. (2011). General notes on ductility in timber structures.
Engineering Structures 33(11): 2987‐2997.
Moroder, D., Smith, T., Pampanin, S., Palermo, A. and Buchanan, A.H. (2014). Design of Floor
Diaphragms in Multi‐Storey Timber Buildings. International Network on Timber
Engineering Research. Bath, England.
NZS 1170.5 (2004). Structural Design Actions Part 5: Earthquake Actions ‐ New Zealand.
Wellington, New Zealand, Standards New Zealand.
Paulay, T. and Park, R. (1975). Reinforced concrete structures. New York, Wiley.
Popovski, M. and Karacabeyli, E. (2005). Framework for lateral load design provisions for
engineered wood structures in Canada. CIB Working Commission W18 ‐ Timber
Structures. Karlsruhe, Germany.
Priestley, M.J.N., Calvi, G.M. and Kowalsky, M.J. (2007). Displacement‐Based Seismic Design of
Structures, IUSS Press.
Schick, M., Vogt, T. and Seim, W. (2013). Connections and anchoring for wall and slab elements
in seismic design. CIB Working Commission W18 ‐ Timber Structures. Vancouver, Canada.
Schoenmakers, J.C.M., Leijten, A.J.M. and Jorissen, A.J.M. (2013). Beams loaded perpendicular
to grain by connections. CIB Working Commission W18 ‐ Timber Structures. Vancouver,
Canada.
Sustersic, I. and Dujic, B. (2011). Influence of connection properties on the ductility and seismic
resistance of multi‐storey cross‐lam buildings. CIB Working Commission W18 ‐ Timber
Structures. Alghero, Italy.
Versuche zur DIN 1052 (2010). Teilprojekt II: „Einsatz visueller Medien in der Aus‐, Fort‐ und
Weiterbildung im Zimmerer‐ und Holzbaugewerbe“. Bundesbildungszentrum des
Zimmerer und Ausbaugewerbes GmbH.
Zarnani, P. and Quenneville, P. (2014a). Splitting Strength of Small Dowel‐Type Timber
Connections: Rivet Joint Loaded Perpendicular to Grain. Journal of Structural Engineering
140(10): 4014064.
Zarnani, P. and Quenneville, P. (2014b). Strength of timber connections under potential failure
modes: An improved design procedure. Construction and Building Materials 60: 81‐90.
Zarnani, P. and Quenneville, P. (2015). Group Tear‐Out in Small‐Dowel‐Type Timber
Connections: Brittle and Mixed Failure Modes of Multinail Joints. Journal of Structural
Engineering 141(2).

419
INTER / 48 - 102 - 03

Discussion

The paper was presented by G Fink

K Ranasinghe commented that simplicity would be needed by practicing engineers

and asked the group whether we know who the target audience are for the standards.
G Fink responded that the user of the standards is our aim; however, they needed to
look deep into the problem to get the basic understanding to come up with provisions.
H Blass commented that one would have to find the tools to translate these findings
to the engineering world and practicing engineers would have deadlines and budget
constraints and would use simple tools.
I Smith commented that sizing members would have easy ways but structural system
considerations would be more challenging.
F Lam commented that the most needed area for reliability based design considera-
tion would be connection design.
S Franke agreed that one needed a deeper understanding to transfer the information
into simple equations; however, PhD students could be involved in the useful step of
documentation of the steps from the deep understanding to the simple equation. G
Fink responded that PhD students needed to be more focused and work on a topic in
depth rather than looking into too many areas.
G Schickhofer agreed that COV should be considered but how to do this in a standard
would be the question. G Fink agreed that it would not make sense to have different
γm values to consider COV. F Lam added that in Canada, a reliability normalization
procedure is available to consider different COVs in a reliability based approach.
U Kuhlmann suggested only one γm and provide calibration factor for γm and docu-
ment clearly where these factors come from.
P Dietsch and G Fink discussed the aim of Eurocode 5 and the target audience. The
fact that practicing engineers not familiar with wood might complain that the code
would be too complicated.

420
421
422
 INTER / 48 - 102 - 04

An Execution Standard Initiative for

Timber Construction

Tomi Toratti,
Wood working industries, Finland

Keywords: Execution, building works, quality control, weather protection, tolerances,

construction process,

1 Introduction
At present, execution standards exist at the European level for concrete, steel and
masonry construction. For timber construction such standards do not yet exist. How-
ever, in several European countries as well as at least in Canada and Australia, na-
tional standards or guidelines for the execution of timber structures have developed.
In Finland, the national execution standard for timber construction has recently been
published [SFS 5978:2014 Execution of timber buildings – Rules for load-bearing
structures of buildings]. In this national standard, quality requirements are set for the
design and construction of timber buildings, so that sufficient overall quality of the
building can be ensured and that the building is built as it has been designed.
Attention towards execution guidelines are also found in recent publications from
Australia on fire during construction [England, 2014] and moisture controls [MacKen-
zie, 2012] and on tall timber buildings in general in Canada [Karacabeyli and Lum,
2014]. Such initiatives when followed will set a level on the quality of timber con-
struction.
Currently Eurocode 5 has some pages related to detailing and execution of timber
structures. These are given in chapter 10 and this consists of 3 pages. In the code
committee CEN/TC250/SC5 there is a plan to produce a European full execution
standard for timber structures, which may possibly be a separate new part of the Eu-
rocode 5.
In a Nordic project named ‘NEXT Timber’, Nordic views on this document are dis-
cussed and this discussion will also be brought up at least in the Eurocode CEN group.
This paper summarizes the fields of interest and anticipated results as a basis for fu-
ture execution standards. This paper is also a continuation of an INTER note [Nore &
al. 2014] on execution standardization. Here the backgrounds and ideas are brought a
step forward.

423
INTER / 48 - 102 - 04

2 Scope of an execution standard

Until now, the scope of an execution standard for timber construction has not been
fully discussed or agreed at the European level. Additionally, the different structural
materials have their execution standards structured in varying ways and there would
also be a need on harmonizing the structure, contents as well as the scope of these
different execution standards. This harmonization should take place at the
CEN/TC250 level.
The scope of an execution standard is on the quality assurance of the end-product
(which is the building) quality through execution rules, which are achieved by a func-
tional cooperation among the project partners, sufficient coverage and quality of the
design and on the documentation to be produced for the construction process. The
standard should describe at least the following means for quality assurance in detail:
 The project description – A description of the project including all relevant infor-
mation
 The moisture control plan – A plan of how moisture safety is achieved during the
works
 The assembly plan – A plan how the members and elements are erected and the
bracing required
 Tolerances of assembly – Description of tolerance classes and tolerance values for
the construction works, manner of measuring and the checking scheme.

It is proposed here that the execution standard shall cover all the building site work
and related quality control measures from the end of the design phase all the way to
the stage of the finalized building. The recently published Finnish execution standard
[SFS 5978:2014] does actually spread somewhat wider than this as quality aspects of
design are also stated, as well as the ensuring of good communication between de-
sign and building site work. This initial point may vary as design work and on-site
build work are often carried out simultaneously. The execution standard should cover
the aspects related to the planning of the execution works. This should not be mixed
with the design work of the structures.
The execution standard shall be based on producing the building with the intended
level of quality given the design documentation as the starting point.
Two different targets for execution may be given:
1. To apply such a level of execution that the design rules given in Eurocodes ap-
ply. This could be related to assembly tolerances and moisture/weather pro-
tection for instance.
2. To apply execution rules which are known to produce sufficient quality level,
work safety and which is economically sound based on earlier experience.

424
 INTER / 48 - 102 - 04

3 Contents of an execution standard

At present, chapter 10 of Eurocode 5 states some requirements for the execution.
These are requirements on the following topics:
 Connectors and there holes, washers, predrilling etc.,
 Transportation, assembly and erection of elements,
 Detailing of diaphragm connections,
 Some rules for erecting nailplate trusses

The above list is in essence needed. However this is only a small part of the execution
provisions that would be required in a standard to ensure the intended quality level.
At least the following items need to be covered as well.
 The quality assurance on the building site. Definition of the parties involved,
their tasks and responsibilities, quality control measures and inspection
schemes. Execution classes: relevant execution classes (as in SFS5978 defined
EXEC 1, 2 and 3) could be defined based on the consequence classification (EN
1990 annex B) and on the execution difficulty level assumed. This should be di-
rectly linked to the quality assurance scheme to be applied (the level of check-
ing).
 The assembly plan. The assembly of prefabricated elements is a very central
part of the execution of a building. Usually an assembly plan is required. This
plan defines the responsibilities of the parties involved and solutions for Struc-
tural stability during each phase of the erection,
 Fire safety during the site work,
 The moisture control plan. Control of moisture during the building site work,
weather protection methods applied, material storage conditions and the re-
lated inspection scheme.
 Definition of assembly tolerances, how these are measured and the inspection
scheme

4 Requirements coming from the design (Eurocode

5) to the execution standard
The design documentation defines a set of requirements to the execution of the
building. It is self-evident that the design as such has to be followed in the building
phase. In addition, there are a set of requirements which are embedded in the design

425
INTER / 48 - 102 - 04

methods and therefore are conditions to be followed in the execution for the design
to apply. Such characteristics are for instance:
4.1 Tolerances,
As the assembly of building parts and elements can never be done with exact preci-
sion, there has to be an assembly tolerance for which the design equations still apply.
If wider tolerances should be applied, this should be allowed if it is considered in the
design and alterations to the design equations or additional loads may apply. Consid-
ering the tendency towards an increased degree of prefabrication, defined tolerances
are necessary to be agreed on and followed. Well-defined and agreed tolerances will
simplify the communication between de different parties.
Tolerances are needed for:
1. Material sections, which are given in product standards,
2. Prefabricated elements, which should be given in a product standard (when fi-
nalised),
3. Connection and connector placings, which should be given either in the execu-
tion standard or in the design provisions,
4. Assembly of elements/parts on the building site, which should be given in the
execution standard

Tolerance classes for the assembly shall be defined and these should be dependent
on the execution class/consequence class of the building. The material tolerances are
given in the relevant product standards (where available) and the assembly toleranc-
es should be drafted together with the building contractors, code writers and design-
ers in cooperation. No international discussion on this has yet followed.
4.1.1 Tolerance for compressed members
For compressed members the stability criterion in Eurocode 5 limits the straightness
of members to L/300 for solid timber and L/500 for LVL and glulam. These are also
given in the respective EN product standards. This is a basis also for the max assembly
tolerance for wall straightness of ± 1,5 ‰ in table 1, which leaves some
bow/crookedness for the member itself. If this is not fulfilled additional eccentricity
loads should be considered.
Table 1. An example of assembly tolerances for timber wall structures (from SFS 5978)

Maximum deviation allowed

Dimension and location Tolerance class Tolerance class Tolerance class
3 2 1

Wall structures
Side location from base line ± 3 mm ± 5 mm ± 10 mm

426
 INTER / 48 - 102 - 04

Spacing of wall studs ± 3 mm ± 5 mm ± 10 mm

Size of windows and doors ± 3 mm ± 5 mm ± 10 mm
Location of windows and doors ± 3 mm ± 5 mm ± 10 mm
Distance between two adjacent walls ± 3 mm ± 5 mm ± 10 mm
Straightness of walls1) ± 1,5 ‰ ± 1,5 ‰ ± 1,5 ‰
Deviation of the wall frame from vertical line
- height maximum 3 m ± 5 mm ± 5 mm ± 5 mm
- height over 3 m ± 8 mm ± 8 mm ± 8 mm

4.1.2 Tolerances for connections

For connections there are many tolerance measures to be considered. A set of toler-
ances as given in SFS 5978 is presented here as an example in table 2. These toler-
ances may be on the placement of the connection itself, the measure and placement
of the holes if predrilled, spacing’s and edge distances etc. These tolerances should
be considered in the design equations themselves. However, the present tolerance
values are driven more by the practice: what is economically achievable with the ap-
plied production technologies. An interesting tolerance measure is also the support
length where compression perpendicular to grain prevails from the reaction force of
the horizontal member. In the example below for this measure a -10 mm tolerance is
allowed and no plus tolerance is needed. This value is solely based on empirical
judgement.

Table 2. Tolerances for connections in timber structures. The allowed deviation is given from the
nominal location if not mentioned otherwise in the structural design. d is the diameter of the
connector.

description / base
Connection type tolerance value Allowed deviation or gap
Nailed connections
1)
wood-wood location of connector spacing a1,a2 + max(10%; d)
Screwed connection
wood-wood end distance a3 -0 / +10 mm
side distance a4 -d / +10 mm
head flat with sur-
penetration face -0 / +3 mm
Nailing plate connec- holes in the nailing
tion location of hole plate + 3 mm
(also screwed) location of nailing plate in both directions + 5 mm

427
INTER / 48 - 102 - 04

Metal framing plates

and hangers location of connector general + 5 mm
from contact sur-
face + 2 mm 9)
edge or end distance of connector -d
gap to wood surface full contact skewed gap max. 3 mm
simultaneous drilling
2)
Bolt connections location of bolt + 5 mm 5)
location of hole separate drilling + 1,5 mm 3)
full contact be-
tightening tween members skewed gap max. 3 mm
simultaneous drill-
4)
Dowel connections location of connector ing + 3 mm
location of hole 4) separate drilling + 1 mm 6)
length of dowel in wood member t - max(2 mm; 0,05t) 7)
grooved or bat-
gap of contact tened members < min(3 mm; 0,25tt) 8)
Incline screw connec-
tions angle of screw + 5°
position at wood sur-
Glued-in rods location of rod face + 5 mm
Glued-screw connec-
tions skewness of holes drilling length La + La/50
gap contact required skewed gap max. 3 mm
contact length contact length - 10%
Contact connections in grain direction + 10%
length of notch
perp. to grain direc-
depth of notch + 5 mm
tion
1)
In the direction of grain, the nails have to be at least d out of line from each other

if a1 < 14d.
2)
Drilling once through all members or using one drilled member as a template.
3)
When wood members have 1 mm oversize holes and metal parts have 1,5..2 mm oversized holes.
4)
In both surfaces of all wood members.
5)
In the grain direction, the row may be out of line max. 5 mm from each other.
6)
When in wood-metal connections have 1 mm oversized holes in metal parts.

428
 INTER / 48 - 102 - 04

7)
t is the design smooth length of the connector in a wood member.
8)
Gap between wood surface and metal plate, where tt is the metal plate thickness.
9)
For example the distance of a supporting L-plate from the wood-wood contact surface (i.e. preliminary
assembly to truss).

4.2 Moisture control

The design is carried out for a defined service class. This boundary condition is to be
followed during the execution. This may require special weather protection means
for the execution. In some cases it may be allowed to carry out the execution in a
higher moisture condition and then dry out before the execution is finalised or before
the structures are insulated and closed. The designer should instruct the protection
means for the works in such cases also.
Moisture control during the execution is of vital importance when building with
wood. A moisture control plan, can ensure sound construction with minimal undesir-
able moisture influence. At first, a level of weather protection in the building process
must be defined. Further, all stages must be followed from fabrication, transport, de-
livery, storage, assembly and use.
A moisture control plan should cover the whole chain of production and building pro-
cess of the timber structure. It shall be ensured that the contractor follows the mois-
ture control plan.
The content of the moisture control plan is proposed as the following:
1. Basic information of the building project (address and other coordinates of the
building site, the person responsible for the construction on site, the main author
of the moisture control plan),
2. List of wood materials and products to be used in the construction site,
3. The target moisture content of wood and wooden elements at different stages of
the production,
4. The target moisture content of wood and wooden elements when delivered to
the building site, during assembly and when the building is finalized,
5. Inspections on site and the person responsible for them,
6. Possible sources of moisture in the building site (for instance, rain, snow, ground
water etc.),
7. The protection level (PL0-PL3, given in section 4.2.4) chosen for the building phase
and an estimate on the necessary protection duration,
8. The protection of wood on the building site:
- storage method and protection of storage

429
INTER / 48 - 102 - 04

- protection during assembly (as determined by the protection level)

- drying methods applied for wood that has gained moisture (for some reason),
9. Controlled drying of structures to the service conditions of the building
- analysis and prevention of risks caused by moisture, rain among others
- sensitivity of the project to unfavourable weather and other exceptions
- determination of moisture contents of wood, drying times and appropriate dry-
ing conditions
- organizing of drying conditions
- effects of onsite schedules (contingency plans),
10.Moisture measurement plan (measurement method, timetable, documentation
and person responsible)

4.2.1 Design moisture content

In heated internal conditions, timber structures are design to service class 1, in which
case the mean moisture content of wood shall be below 12 % (RH 65 %, 20oC).
In sheltered unheated conditions, timber structures are designed to service class 2, in
which case the target moisture content of wood shall not exceed 20 % (RH 85 %,
20oC).
The internal air humidity of heated buildings may be very low during winter times.
The moisture content of wood may decrease to even less than 5 % during the winter
and during the fall it may increase to about 12 %. In sheltered unheated conditions
the moisture content of wood may vary during the year between 12 – 18 %.

4.2.2 Moisture content of wood during delivery to building site

The moisture content of wood products at delivery varies greatly depending on the
product. Unless otherwise agreed, the following delivery moisture contents may be
assumed:
- sawn timber is delivered usually air-dried at 15 – 25 %. Due to risk of mould
growth this should be lower than 20%.
- glue laminated timber is usually delivered at a moisture content of 10 - 12 %
- plywood and laminated veneer lumber are delivered at a moisture content of 8 –
10 %
- if delivered from storehouse the moisture content of glued laminated timber,
plywood and laminated veneer lumber may be 20 % at most

Suitability of the moisture content measurement instrument for the actual product
shall be checked.

430
 INTER / 48 - 102 - 04

4.2.3 Moisture during construction

Normally during the storage and assembly phases, the wood moisture content in-
creases, except for sawn timber which is delivered usually at a high moisture content
and usually this continues to dry on the building site. On the building site, during
storage and assembly, the material moisture increase is controlled with weather pro-
tection.
The drying of wood should be carried out sufficient slowly so that drying cracks would
not develop. Controlled drying is especially important with massive cross sections.
Drying conditions shall be such that the difference between measured moisture con-
tent and equilibrium moisture content of the drying conditions is at most 6 %. If the
need of the drying of wooden members is more than that, the drying shall be ar-
ranged in phases. The structure is not allowed to be sealed (covered, coated) before
the desired moisture content is achieved.

4.2.4 Weather protection levels

In the moisture control plan and in the assembly plan, the level of weather protection
is decided. During the selection of the weather protection level, exceptional weather
conditions and additional moisture caused by the construction activities shall be con-
sidered.
The weather protection levels and the respective moisture contents expected are:
0. Protection level PL0, no protection:
- moisture content depends on the climate and may not be assigned
- recommended only in winter climates (freezing temperatures) and in short du-
rations
- ground contact is not allowed
1. Protection level PL1, plastic or tarpaulin covering:
- moisture content below 20 %
- sufficient ventilation shall be ensured
2. Protection level PL2, sheltered:
- moisture content below 20 %
- more reliable than PL1
3. Protection level PL3, internal conditions or a tent with heating:
- moisture content below 15%

431
INTER / 48 - 102 - 04

4.3 Assembly plan

The assembly of prefabricated elements is a very central part of the execution. Usual-
ly an assembly plan is required (often by the building law). This plan should define the
responsibilities of the parties involved and the solutions for:
 Structural stability during each phase of the erection,
 Fire safety during the site work,
 Control of moisture during the site work, protection methods, element storage
and related inspections
The assembly plan is a building project specific and requirements set in the moisture
control plan have to be taken into account. Working methods for the safe assembly
of timber structures are presented in the assembly plan. The proposed content of as-
sembly plan is shown in table 3. This example is given for a high consequence class 3
(execution class 3) where the requirements are more rigorous. For other classes
some rows may be omitted and there are fewer requirements.
The assembly plan has to be compatible with the project specification document de-
scribed earlier. The assembly plan may differ from the project specification if there is
a sound reason for it, for example changes in site conditions, and the safety of work is
not decreased. Such changes shall be accepted in appropriate stakeholders and doc-
umented.

Table 3. Proposed content of the assembly plan for execution class 3

1. Site information Building project, address

Supplier of elements
Building developer
General contractor
Designer
General structural designer
Element designer
Site organization:
- general foreman of the site
- site manager of assembly performing contractor
- general timber structures foreman

2. Building elements Element types:

- quantities, outer dimensions and weights,
- required lifting equipment, lifting points and practises
Crane information: crane type, lifting capacity, reach, maximum load of
support on ground

432
 INTER / 48 - 102 - 04

3. Preliminary reports - the building site fulfils the requirements of the security plans
with the following - the stability of the cranes used and their maintenance
requirements - the control of entry to the building site
- ground and foundation conditions and safety on site
- any requirements on dimensions and weights of building parts handled
on site
- environment and weather conditions to be considered (required protec-
tion)
- information on neighbouring buildings effect on the assembly work and
vice versa

4. Reception of build- - transportation and storage on site, storage method and location,
ing elements strength of stock stands
- working schedules on pre-assembly activities
- initial inspection and unloading of the elements
- transportation methods of the elements, machinery needed and path-
ways to be clear

5. Protection of ele- - Weather protection level applied

ments - coatings applied during the execution

6. Lifting, assembling - assembly order per building or storey or building section

and assembly order - support of elements during assembly
- protection of reaction and attachment points of elements
- the stage when the additional supports are removed
- actions caused by final stability and fastening of elements
- detailed assembly order and moisture control relating to it
- introduction of site personnel

7. Accuracies, toler- - tolerance class applied for the assembly

ances, schemes - basic location point for tolerance measurements on site

8. Connection methods Description of connection techniques and used connectors and work
and details on as- method
sembly

9. Work safety - falling protection

- temporary scaffolding, decking or fencing and other security measures,
personal protective equipment

10. Assessment of fire Fire risks on the building site and preparations for these
risks

433
INTER / 48 - 102 - 04

11. Inspection plan Inspections, testing and repair activities are done on the performed work
when compared to the project description structural plan requirements
so that the timber structures have the strength and stability and are du-
rable:
- inspection of materials and products
- inspection on storage
- inspection of components, connections and structures
- inspections and reviews during the process of assembly
- inspection and measurement on moisture content and on building con-
ditions
- inspection on temporary protections

12. Approval of the as- Drafter of the assembly plan

sembly plan Contractor performing the assembly
General timber structures foreman
Main structural designer
Element designer
Main contractor on the site

13. Handling of fault Action procedure, security measures and communication

incidents

5 Conclusion
In this paper some thoughts, ideas and proposals have been presented for the con-
tents of a future EN standard for the execution of timber structures. This work will be
developed in the group CEN/TC250/SC5.
It will be important to identify all the requirements that the design procedures set on
the execution phase of the building. In the present paper such identified cases are at
least a) the assembly tolerances and b) the moisture content limits for wood. There
could be also other overlapping conditions behind the design equations that set pro-
visions for the execution. These should be identified.
The overall target of an execution standard is to set a defined level of quality on the
building site works. This quality level may be different for different execution or con-
sequence classes. Such a standard would underline good building practices and it
should be simple for clients to use or demand and straightforward for the material
producers, building companies and authorities to apply.

434
 INTER / 48 - 102 - 04

6 References
EN 1995 1-1 Eurocode 5. Design of timber structures - Part 1-1: General -Common
rules and rules for buildings. Comité Européen de Normalisation (CEN), Brussels,
Belgium. 2004.
England P. 2014 Fire Precautions during Construction of Large Buildings. Timber-
framed Construction Technical Design Guide. Forest and Wood Products, Australia.
Karacabeyli E and Lum C. 2014. Technical Guide for the Design and Construction of
Tall Wood Buildings in Canada. Special Publications SP-55E. FPInnovations, Canada.
MacKenzie C. 2012 Impact and Assessment of Moisture-affected Timber-framed Con-
struction Technical Design Guide. Forest and Wood Products, Australia.
Nore, K. & al. 2014 Execution of Timber Structures. Note in INTER meeting 2014, Bath
UK.
SFS 5978:2014 Execution of timber structures. Rules for load-bearing structures of
buildings (In Finnish, an English translation is available).

435
INTER / 48 - 102 - 04

Discussion

The paper was presented by T Toratti

U Kuhlmann commented and explained in steel there is an independent executive

standard EN1090 which depends on executive class. The executive class and conse-
quence classes are critical. U Kuhlmann recommended similar approach to be consid-
ered for timber.
S Winter commented that we need to learn from other material and people doing ex-
ecution on site always fear too much regulation even though this is needed. He stated
that national specification document not national standard would be interesting for
industry. S Winter and T Toratti discussed responsibility of different parties in a build-
ing project as this is a grey area.
R Jockwer and T Toratti discussed issues related to tolerances in relationship to work-
ability and practicality. Design equations have to consider these tolerances including
length of members etc.
M Follesa stated in Italy foundation tolerances might not be compatible with timber
tolerance. There are discussions that tolerance is an issue for all materials. In steel,
basic tolerances are available with the consideration of additional tolerance. If basic
tolerance is not followed, redesign might be needed to check if this is ok.

436
437
438
4 INTER Notes, Šibenik, Croatia 2015
Note 1 Start time of charring of timber members protected with gypsum
plasterboards - A Just, K Kraudok
Note 2 Protective effect of insulation materials on charring of timber
elements - M Tiso, A Just
Note 3 Investigation of thread pattern for self-tapping screws as
reinforcement for embedment strength - Cong Zhang, Wen-Shao
Chang, R Harris

439
440
 INTER / Note 1

Start time of charring of timber members

protected with gypsum plasterboards
Alar Just, Tallinn University of Technology
Kairit Kraudok, Tallinn University of Technology

Keywords: Gypsum plasterboards –Start time of charring – Fire resistance

1 Introduction
Gypsum boards are widely used in timber frame assemblies because of their fire pro-
tective properties. The start time of charring and the failure time of the board are
important properties for the fire safety design of timber frame constructions.
EN 1995-1-2 gives simplified design rules for calculation of pull-out failure. However,
the failure time of gypsum plasterboard type F should be determined on the basis of
tests.
As the thermo-mechanical properties of gypsum plasterboard type F are not part of
the classification given in the European product standard for gypsum plasterboards
EN 520, failure times of different makes may vary considerably
For the start time of charring there are rules in EN 1995-1-2 [1] and following the
component additive method in „Fire safety in timber buildings. Technical guideline for
Europe“ [2].
The aim of this study is to verify the calculation rules for start time of charring on EN
1995-1-2 and the component additive method in [2] by the extensive analyse of the
database of full scale fire tests. The study is performed in co-operation with Tallinn
University of Technology and SP.

2 Design methods
There are basically two different design methods regarding start time of charring.
EN 1995-1-2 provides following equation for start time of charring behind gypsum
plasterboards Type A, F and H.
𝑡𝑐ℎ = 2,8 ∗ ℎ𝑝 − 14 (1)

441
INTER / Note 1

The component additive method for design of separating function was developed in
ETH by Schleifer [4] based on extensive experimental results and finite-element
thermal analysis. According to this method the start time of charring can be calculat-
ed as sum of protection times of the cladding layers:

𝑡𝑐ℎ = ∑ 𝑡𝑝𝑟𝑜𝑡,𝑖 (2)

Protection time is defined as the time when temperature rise behind the considered
layer is reaching 250K. Protection times take into consideration the influence of the
backing and preceding layers.

3 Analysis of the database

A database with data from full-scale fire test reports with assemblies including clad-
dings made of gypsum plasterboards in accordance with EN 520 and gypsum fibre
boards in accordance with EN 15283-2 was collected at SP Technical Research Insti-
tute of Sweden. The database consists of results of 388 full-scale tests from different
institutes all over the world, although mainly from Europe.
The first analysis of the results of the full-scale fire tests was carried out by Just et. al.
[3] and the results of the analysis were published in the „Fire Safety in Timber Build-
ings. Technical guideline for Europe“ [1]. The analysis is based on minimum values of
the protection times
Present analysis is performed by finding 5% fractile values and is based on increased
number of test data.

4 Discussions
Start time of charring is depending on thickness of gypsum plasterboards. One layer
cladding consisting of Type A and Type F boards are analysed together. See Figure 1.
Start times of charring behind two layers of the gypsum plasterboards, Type F are
presented. See Figure 2.

442
 INTER / Note 1

80 90

70 80
EN 1995-1-2
60 ETH [4] 70
tch=Σtprot
50 60 5%
fractile

tch [min]
40
tch [min]

EN 1995-1-2 50 [5]
30 40
20 30 ETH [4]
ETH [4] tch=Σtprot
10 5% fractile [5] 20 Walls
tch=Σtprot
Floors
0 10
5 10 15 20 25 30 20 25 30 35
hp [mm] hp,tot [mm]
Figure 1. Full scale tests with recorded start Figure 2. Full scale tests with recorded start time
time of charring behind one layer cladding of of charring behind two layers cladding of gypsum
gypsum plasterboards. 5% fractile values plasterboards, Type F. 5% fractile values
compared to EN 1995-1-2 method and compared to EN 1995-1-2 method and
component additive method [4]. component additive method [4].

This study shows clearly that the generic values for start time of charring are overes-
timated in EN 1995-1-2 for the most cases. 5% fractile values of start times of char-
ring from the database of full scale fire tests show good correlation with the compo-
nent additive method.
Based on this analysis the component additive method [4] is recommended for the
revision of EN 1995-1-2 for determining design start time of charring behind gypsum
plasterboards.

5 References
[1]. Eurocode 5 (2004) Design of timber structures – Part 1-2: General – Structural
fire design. CEN (EN 1995-1-2).
[2]. Fire safety in timber buildings (2010) Technical Guideline for Europe. SP Tech-
nical research Institute of Sweden, Wood Technology. SP Report 2010:19. Stock-
holm, Sweden.
[3]. Just, A, Schmid, J and König, J (2010). Gypsum Plasterboards Used as Fire
Protection - Analysis of a database,” SP Report: 2010:29, Stockholm, Sweden.
[4]. Schleifer, V (2009) Zum Verhalten von raumabschliessenden mehrschichtigen
Holzbauteilen im Brandfall. Zürich, Switzerland.
[5]. Kraudok, K. (2015) Protective effect of gypsum plasterboards for the fire design
of timber structures. Master thesis. TUT. Tallinn, Estonia.

443
444
 INTER / Note 2

Protective effect of insulation materials

on charring of timber elements
Mattia Tiso, Tallinn University of Technology
Alar Just, Tallinn University of Technology

Keywords: Insulation materials –Timber assemblies – Fire resistance

1 Introduction
The insulation materials may give a contribution to the fire resistance of timber frame
assemblies concerning the load-bearing (R criteria) and the separating function (EI
criteria).
Current Eurocode 5 Part 1-2 [1] provides a model for fire design of the load bearing
function of timber frame assemblies with cavities completely filled with stone wool.
The extension of this model for the glass wool for post protection phase [2] is pub-
lished in the European technical guideline Fire Safety in Timber Buildings [4]. For the
verification of the insulation criteria a component additive model is available in Euro-
code 5 Part 1-2 that is further improved and published in the European technical
guideline Fire Safety in Timber Buildings [4].
Eurocode 5 Part 1-2 distinguishes clearly between stone wool and glass wool because
of the different behaviour in fire of those mineral wools. There is similar behaviour of
glass wool and stone wool as long as the wool is protected. According to the Europe-
an product standard of mineral wools [2] those two materials are both mineral wool
and there is no requirement to producer to declare the type (stone or glass). Beside
mineral wool there are other insulation materials that are not included in the design
models according to Eurocode 5 Part 1-2.
The aim of this study is to provide the classification methodology of different insula-
tion materials in terms of fire design of timber structures and improvement of the
fire design model for timber frame assemblies.
The study is performed by Tallinn University of Technology and SP with co-operation
with the international reference group. Since now the investigation of different insu-
lation materials was performed to study the contribution of insulations to the fire re-
sistance of timber frame assemblies.

445
INTER / Note 2

2 Test method
Ten specimens of timber frame assemblies have been tested in horizontal position in
a cubic meter furnace following the standard fire curve according to ISO 834. The
specimens consisted of timber beam with cross-section 45 x 145 mm and the insula-
tion materials applied. The fire side of the tested assembly was protected by 15 mm
thick gypsum plasterboard; Type F. On the unexposed side the particle board with
thickness of 19 mm was used. Thermocouples have been embedded on the timber
beam and insulation material in different depths to follow the charring scenario. In
order to obtain comparable results for the post-protection phase among the insula-
tion materials, the fall off of the gypsum plasterboard was imposed after 45 minutes
from the test start. This was done by releasing the special fastening system. The ex-
pected duration of the tests was 60 minutes; in some tests the specimen was re-
moved earlier due to start of charring in particle board.
At the end of the tests the instrumented beam has been cleaned from the char layer
and the minimum residual cross section has been obtained.

3 Discussions and future works

Ratios between minimum residual moments of inertia and initial moment of inertia
for the tests have been compared with the models in Eurocode 5 Part 1-2 and Fire
Safety in Timber Buildings (Figure 1).
The first comparison indicates that the insulation materials could be divided accord-
ing to their ability to stay only in the protected charring phase and also post-
protection charring phase.
The protection provided by most of the materials is on the safe side respect to the
two existing models analysed. Anyhow the validity of the models for other materials
than stone wool or glass wool is not declared in [1] and [4].
Hence the improved model for timber frame assemblies should be open for wide
range of the insulation materials.
Charring of the timber member has to be considered as two dimensional as general.
Different charring rates have to be regarded for different sides of timber members
and different phases of protection.
There is necessity to develop a qualification methodology that considers the contri-
bution of any insulation materials to the fire protection of timber member. The quali-
fication should not be simply related on the typology of material but on its perfor-
mance.
For comparison reasons in this first step the experiments have been performed with
the same fire protection boards on the exposed side of the specimens and the same

446
 INTER / Note 2

cross sections of timber members. These factors could influence the charring behav-
iour in the protection phase and the load-bearing resistance.
1.0 Design model EC5
Ratio inertia moment In/I0

Design model FSITB

INS1
0.8
INS2

Failure fire protection t = 45 min

INS3
0.6
INS4

INS5
0.4
INS6

INS7

0.2 G type F, 15 mm INS8

45 mm x 145 mm INS9

0.0 INS10
0 10 20 30 40 50 60
Time [min.]

Figure 1. Minimum residual moments of inertia divided initial moment of inertia for the ten
tests compared with the models for stone wool and glass wool

The study will continue with the investigations in order to consider the other cross-
sections and protections. Comprehensive study with more fire tests and thermal sim-
ulations is planned following by analysis and proposals for the classification method-
ology of the insulation materials as well as for the improved design model for timber
frame assemblies.

4 References
[1]. Eurocode 5 (2004) Design of timber structures – Part 1-2: General – Structural
fire design. CEN (EN 1995-1-2).
[2]. EN 13162:2012, “Thermal insulation products for building – Factory made min-
eral wool (MW) products – Specification” European Standard. European Commit-
tee for Standardization, Brussels.
[3]. Just, A, Schmid, J and König, J (2010) Post protection behaviour of wooden wall
and floor structures completely filled with glass wool, Structures in Fire (SIF).
[4]. Fire safety in timber buildings (2010) Technical Guideline for Europe. SP Tech-
nical research Institute of Sweden, Wood Technology. SP Report 2010:19. Stock-
holm, Sweden.

447
448
 INTER / Note 3

Investigation of thread configuration for

self-tapping screws as reinforcement for
embedment strength

Cong Zhang, University of Bath

Wen-Shao Chang, University of Bath
Richard Harris, University of Bath

Keywords: self-tapping screw, thread pattern, embedment strength

1 Introduction
Dowel-type connections are the most common type of connections in timber struc-
tural design, especially in large-scale structures. Recent studies, Blass and Schmid
(2001), Bejtka and Blass (2005) & Blass and Schadle (2011), have shown improvement
on load-carrying capacity of dowel type connections using self-tapping screws (STS).
Similar to other reinforcement such as fibre-reinforced polymers and glued-on ply-
wood panels, the effectiveness of screw reinforcement is implemented and influ-
enced by the bond between wood and screw. Promoting this bond will enhance the
performance of reinforcement. The aim of this study is to investigate the influence of
thread configuration on self-tapping screws to reinforce the embedment strength.
Following this work, the project will progress to optimisa-
tion of self-tapping screws and their use in reinforcement
of dowel-type connection.
20mm
dowel
2 Methodology
Following the guidance by BS EN 383-2007, a total of 150 Distance
to dowel
embedment tests were conducted as shown in Figure 1.
The embedment strength of timber specimens with vary-
ing thread location and length (see Figure 2) was deter- Screw re-
mined. The purpose was to find the most crucial part of inforce-
the thread over the whole length of the screw. ment
Figure 1 Embedment
test set-up

449
INTER / Note 3

Distance
Description to dowel
(d=20mm)

N, 0% thread 1d

S, 100% thread 1d

BS, 33% thread on the pin end 1d

DS, 33% thread on the head end 1d

ES, 100% thread 0.5d

FS, lthread/leff=0.15 on the pin end 1d

GS, lthread/leff=0.35 on the pin end 1d

TTS, 66% thread, two thread segments 1d

Figure 2 Different thread arrangements used in the embedment tests.

3 Results & discussion

Statistical analysis shows that the embedment strength between specimens rein-
forced by a screw with complete thread (S) and screw with 33% thread on the pin
end (BS) does not differ significantly. This implies the thread on the pin side is crucial
in providing the required bonding to prevent splitting failure due to tensile stress per-
pendicular to the grain.
It is hypothesised that a relationship exists between the embedment strength and
the length of thread on the most effective pin end, shown in Figure 3. Embedment
tests with group FS (lthread/leff=0.15) and GS (lthread/leff=0.35) were carried out and the
results of embedment strength vs lthread/leff were plotted in Figure 4. The hypothesis is
based on the assumption that a crack will appear in the middle axis of the specimen.

Figure 3 Indication of length of thread (lthread) on the effective length (leff)

450
 INTER / Note 3

Figure 4 lthread/leff vs strength enhancement (150 specimens)

4 Conclusion
This study found that the thread location and length can influence the effectiveness
of reinforcement. A relationship between embedment strength and thread length
over the most effective pin end on a screw is also discovered.
The work continues in tests on screw reinforced connections.

5 References
Bejtka, I. & Blass, H.J., 2005. Self-tapping screws as reinforcements in connections
with dowel-type fasteners. In: CIB-W18 2005.
Blass, H.J. & Schadle, P., 2011. Ductility aspects of reinforced and non-reinforced
timber joints. Eng Struct, 33(11), pp. 3018-3026.
Blass, H.J. & Schmid, M., 2001. Self-tapping screws as reinforcement perpendicular to
the grain in timber connections. Joints in Timber Structures. Stuttgart, Germany.
BSI, 2007. BS EN 383:2007. Timber structures. Test methods. Determination of
embedment strength and foundation values for dowel type fasteners.

451
452
5 Peer review of papers for the INTER
Proceedings
Experts involved:

Members of the INTER group are a community of experts in the field of timber
engineering.

Procedure of peer review

• Submission of manuscripts: all members of the INTER group attending the

meeting receive the manuscripts of the papers at least four weeks before the
meeting. Everyone is invited to read and review the manuscripts especially in
their respective fields of competence and interest.

• Presentation of the paper during the meeting by the author

• Comments and recommendations of the experts, discussion of the paper

• Comments, discussion and recommendations of the experts are documented

in the minutes of the meeting and are printed on the front page of each paper.

• Final acceptance of the paper for the proceedings with

• no changes
• minor changes
• major changes
• or reject

• Revised papers are to be sent to the editor of the proceedings and the
chairman of the INTER group

• Editor and chairman check, whether the requested changes have been carried
out.

453
454
6 Meetings and list of all CIB W18 and
INTER Papers
CIB Meetings:

1 Princes Risborough, England; March 1973

2 Copenhagen, Denmark; October 1973
3 Delft, Netherlands; June 1974
4 Paris, France; February 1975
5 Karlsruhe, Federal Republic of Germany; October 1975
6 Aalborg, Denmark; June 1976
7 Stockholm, Sweden; February/March 1977
8 Brussels, Belgium; October 1977
9 Perth, Scotland; June 1978
10 Vancouver, Canada; August 1978
11 Vienna, Austria; March 1979
12 Bordeaux, France; October 1979
13 Otaniemi, Finland; June 1980
14 Warsaw, Poland; May 1981
15 Karlsruhe, Federal Republic of Germany; June 1982
16 Lillehammer, Norway; May/June 1983
17 Rapperswil, Switzerland; May 1984
18 Beit Oren, Israel; June 1985
19 Florence, Italy; September 1986
20 Dublin, Ireland; September 1987
21 Parksville, Canada; September 1988
22 Berlin, German Democratic Republic; September 1989
23 Lisbon, Portugal; September 1990
24 Oxford, United Kingdom; September 1991
25 Åhus, Sweden; August 1992
26 Athens, USA; August 1993
27 Sydney, Australia; July 1994
28 Copenhagen, Denmark; April 1995
29 Bordeaux, France; August 1996
30 Vancouver, Canada; August 1997
31 Savonlinna, Finland; August 1998
32 Graz, Austria; August 1999

455
33 Delft, The Netherlands; August 2000
34 Venice, Italy; August 2001
35 Kyoto, Japan; September 2002
36 Colorado, USA; August 2003
37 Edinburgh, Scotland; August 2004
38 Karlsruhe, Germany; August 2005
39 Florence, Italy; August 2006
40 Bled, Slovenia; August 2007
41 St. Andrews, Canada; August 2008
42 Dübendorf, Switzerland; August 2009
43 Nelson, New Zealand; August 2010
44 Alghero, Italy; August 2011
45 Växjö, Sweden; August 2012
46 Vancouver, Canada; August 2013

INTER Meetings:

47 Bath, United Kingdom; August 2014

48 Šibenik, Croatia; August 2015

The titles of the CIB W 18 and INTER papers (starting from 2014) are included in the
complete list of CIB/INTER papers: http://holz.vaka.kit.edu/392.php

456