Вы находитесь на странице: 1из 815

RILEM Bookseries

Simon Aicher
H.-W. Reinhardt
Harald Garrecht Editors

Materials and Joints

in Timber Structures
Recent Developments of Technology
Materials and Joints in Timber Structures
Volume 9

RILEM, The International Union of Laboratories and Experts in Construction

Materials, Systems and Structures, founded in 1947, is a non-governmental scien-
tific association whose goal is to contribute to progress in the construction sciences,
techniques and industries, essentially by means of the communication it fosters be-
tween research and practice. RILEMs focus is on construction materials and their
use in building and civil engineering structures, covering all phases of the building
process from manufacture to use and recycling of materials. More information on
RILEM and its previous publications can be found on www.RILEM.net.

For further volumes:

Simon Aicher H.-W. Reinhardt
Harald Garrecht

Materials and Joints

in Timber Structures
Recent Developments of Technology

Simon Aicher Harald Garrecht
University of Stuttgart University of Stuttgart
Stuttgart Stuttgart
Germany Germany

H.-W. Reinhardt
University of Stuttgart

ISSN 2211-0844 ISSN 2211-0852 (electronic)

ISBN 978-94-007-7810-8 ISBN 978-94-007-7811-5 (eBook)
DOI 10.1007/978-94-007-7811-5
Springer Dordrecht Heidelberg New York London

Library of Congress Control Number: 2013949137

c RILEM 2014
No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by
any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written
permission from the Publisher, with the exception of any material supplied specifically for the purpose
of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

The use of timber in structures represents one of the most promising approaches
to meeting the urgent needs for sustainability, environmental friendliness and CO2
emission reduction in building technology. Due to the worlds fast growing popu-
lation, the achievement of the mentioned aims is increasingly understood as one of
the primary keys to ensure future life on earth.
Timber construction has experienced considerable progress within recent years.
Advancement can be seen within the both equally important interacting construc-
tion components - materials and joints. With regard to materials, it is undeniable
that cross-laminated timber (cross-lam) has widened the range of possibilities of
timber constructions the most. Its re-invention of plywood, now based on sawn
boards instead of peeled veneers, is the basis for multi-storey buildings of up to 9
storeys and for realistic perspectives of up to 30 storeys. Though itself rigid, cross-
lam in combination with energy dissipative joints is well suited for buildings in
zones with high seismic activity. The recent production of glulam and LVL made
of high strength hardwoods such as beech, oak, chestnut and several tropical hard-
woods, further broadens architectural horizons and meets the demands of forestry
towards sustainable and soil/climate apt tree cultivation. The spectrum of glulam
possibilities has also been expanded substantially due to so-called block gluing,
whereby several glulams are glued to form massive cross-sections which may, for
instance, be ideally used in bridges. An additional renewable, wood-like material,
bamboo, is also considerably extending the possibilities and efficiency of wooden
materials. Processed and densified bamboo strands enable the production of plates
and beams with strength properties similar or even higher than those of steel. Low
density wood fibre boards, representing a highly ecological insulation material, are
now simultaneously used for structural bracing purposes as well. Timber-concrete
composites with different connection technologies are increasingly gaining trac-
tion in buildings, most notably as cost-effective floor elements, due to good proven
performances regarding strength, stiffness, vibration and response to fire. Innova-
tive, glued, lightweight, composite deck-element constructions based on solid wood
and panel materials are also now manufactured industrially with lengths of over
30 meters.
VI Preface

Regarding mechanical joints, self-tapping screws produced at lengths of up to

1.5 m have substantially changed jointing and reinforcement technology. The use
of the screws as primarily tension or compression transferring devices at an angle
to the fibre direction represents a new highly efficient jointing and reinforcement
technology. Contrary to dowel type fasteners which act by embedment and hereby
inherently induce splitting of the wood, self-tapping screws overcome the problem
of tension perpendicular to the grain, hereby increasing load capacity of the joint
and changing the failure mode from brittle to ductile. Further, self-tapping screws
used in combination with specially formed i.a. dovetailed metal plates enable archi-
tecturally esteemed, completely hidden, end-grain connections.
The diversity of structural adhesives (now comprising phenolic resorcinol,
melamine urea, one-component polyurethanes, emulsion polymer isocyanates and
epoxies) with properties tailored to specific timber products and production pro-
cesses has increased considerably within the last years. In the case of joints with
glued-in rods, recently approved and reliable adhesive systems have been introduced
into timber construction technology and now enable new jointing solutions, i.a. for
wide-span grid-like spherical domes. Further, glued-in, perforated steel plates allow
for stiff and strong connections, i.a. in timber-concrete composite structures. Re-
cently, a new glued jointing technology for glulam beams which enables the trans-
fer of forces without any reduction of the full cross-sectional capacity has also been
The present Conference follows the earlier RILEM conference Joints in Timber
Structures held in Stuttgart from the 12th to 14th of September, 2001. The objec-
tive of the Conference is to bring together world leading experts in the mentioned
fields of timber materials, joints and construction to present the latest state of tech-
nology and to identify future research needs to conserve the pace of progress. The
main subjects of the conference are: cross-laminated timber, glued laminated timber,
timber based facades, modified wood, i.a. acetylated and thermally treated wood,
adhesives, self-tapping screws, glued and mechanical joints, seismic and cyclic be-
haviour of joints and structures, durability issues, surface treatment, timber-concrete
compounds and bamboo-based materials.
The financial support from the sponsoring companies is gratefully acknowledged.
Sincere gratitude is owed to the staff of the Division of Timber Structures of the Ma-
terials Testing Institutute (MPA), Otto-Graf Institute, University of Stuttgart, par-
ticularly to Maximilian Henning, Nikolai von Ruckteschell and Zachary Christian
for all of their organizational efforts. We also greatly appreciate the effort and en-
gagement from the editorial staff of Springer in the compilation of the Proceedings,
particularly Ms. Anneke Pot.

Simon Aicher
H.-W. Reinhardt
Harald Garrecht

The Conference was kindly supported by the following enterprises:

VIII Sponsors

Part I: Structures
Horizontal Displacements in Medium-Rise Timber Buildings:
Basic FE Modeling in Serviceability Limit State . . . . . . . . . . . . . . . . . . . . . 3
Ida Nslund, Helena Johnsson
Improving the Moment Resistance of a Concealed Timber Post Base
Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Jrme Humbert, Sang-Joon Lee, Joo-Saeng Park, Moon-Jae Park
The Multifunctional TES-Faade Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Stefan Loebus, Stephan Ott, Stefan Winter
Green-Glued Products for Structural Applications . . . . . . . . . . . . . . . . . . . 45
Erik Serrano, Jan Oscarsson, Magdalena Sterley, Bertil Enquist
Experimental Analysis of a Post-tensioned Timber Connection . . . . . . . . 57
Flavio Wanninger, Andrea Frangi
Risk Based Investigations of Partly Failed or Damaged Timber
Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Gerhard Fink, Jochen Kohler
Naturally Grown Round Wood Ideas for an Engineering Design . . . . . . 77
Matthias Frese, Hans Joachim Bla
Recycling and End-of-Life Scenarios for Timber Structures . . . . . . . . . . . 89
Annette Hafner, Stephan Ott, Stefan Winter
Advancements for the Structural Application of Fibre-Reinforced
Moulded Wooden Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Jrg Wehsener, Tom-Egmont Werner, Jens Hartig, Peer Haller
X Contents

Sole Plate Fixing Details for Modern Methods of Timber

Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Jesus M. Menendez, Kenneth Leitch, Robert Hairstans
Thin-Walled Timber Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Benoit P. Gilbert, Steven B. Hancock, Henri Bailleres
Recommendations for the Design of Complex Indeterminate Timber
Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Andrew Lawrence
Novel Lightweight Timber Composite Element: Web Design in Shear
and Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Simon Aicher, C. Stritzke
New Timber Bridges: Inventive Design by Block-Gluing . . . . . . . . . . . . . . 149
Frank Miebach, Dominik Niewerth

Part II: Mechanical Connections

Steel-to-Timber Joints with Very High Strength Steel Dowels Using
Spruce, Beech and Azob . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Jan-Willem van de Kuilen, Carmen Sandhaas, Hans Joachim Bla
Wood Load-Carrying Capacity of Timber Connections: An Extended
Application for Nails and Screws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Pouyan Zarnani, Pierre Quenneville
Ductility in Timber Structures: Investigations on Over-Strength
Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Frank Brhl, Jrg Schnzlin, Ulrike Kuhlmann
An Experimental Study on Bearing Strength in Compression
for Bolted Joint of Plywood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Akiko Ohtsuka, Sumiya Takahashi, Takumi Ito, Wataru Kambe
Investigations Concerning the Force Distribution along Axially
Loaded Self-tapping Screws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
A. Ringhofer, G. Schickhofer
Experimental Analysis on the Structural Behaviour of Connections
with LVL Made of Beech Wood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Peter Kobel, Ren Steiger, Andrea Frangi
The Embedment Failure of European Beech Compared to Spruce
Wood and Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Steffen Franke, Nolie Magnire
Contents XI

Modelling of Non-metallic Timber Connections at Elevated

Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Daniel Brandon, Martin P. Ansell, Richard Harris, Pete Walker,
Julie Bregulla
Analysis of the Brittle Failure and Design of Connections Loaded
Perpendicular to Grain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Bettina Franke, Pierre Quenneville
Structural Performance and Advantages of DVW Reinforced Moment
Transmitting Timber Joints with Steel Plate Connectors and Tube
Fasteners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Daniel Brandon, Adriaan Leijten
Fully Threaded Self-tapping Screws Subjected to Combined Axial
and Lateral Loading with Different Load to Grain Angles . . . . . . . . . . . . 265
Robert Jockwer, Ren Steiger, Andrea Frangi
Alternative Approach to Avoid Brittle Failure in Dowelled
Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Daniela Wrzesniak, Massimo Fragiacomo, Andr Jorissen

Resistance and Failure Modes of Axially Loaded Groups of Screws . . . . . 289

U. Mahlknecht, R. Brandner, A. Ringhofer, G. Schickhofer
A Method to Determine the Plastic Bending Angle of Dowel-Type
Fasteners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Michael Steilner, Hans Joachim Bla
Low-Damage Design Using a Gravity Rocking Moment Connection . . . . 307
Mamoon Jamil, Pierre Quenneville, Charles Clifton

Part III: Glued Joints and Adhesives

Finger Jointing of Freshly Sawn Norway Spruce Side Boards
A Comparative Study of Fracture Properties of Joints Glued with
Phenol-Resorcinol and One-Component Polyurethane Adhesive . . . . . . . 325
Magdalena Sterley, Erik Serrano, Bertil Enquist, Joanna Hornatowska
Pressure Distribution in Block Glue Lines Analyzed by Theory
of Beams on Elastic Foundation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
Gordian Stapf, Simon Aicher
EPI for Glued Laminated Timber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
Kristin Grstad, Ronny Bredesen
Bonding of Various Wood Species Studies about Their Applicability
in Glued Laminated Timber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Y. Jiang, J. Schaffrath, M. Knorz, Stefan Winter
XII Contents

Fatigue Performance of Adhesive Connections for Wooden Wind

Towers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
Leander Bathon, Oliver Bletz-Mhldorfer, Jens Schmidt,
Friedemann Diehl
Multifunctional Wood-Adhesives for Structural Health Monitoring
Purposes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
Christoph Winkler, Ulrich Schwarz
Assessment of the Glue-Line Quality in Glued Laminated Timber
Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
Bettina Franke, Florian Scharmacher, Andreas Mller
Review of Recent Research Activities on One-Component
PUR-Adhesives for Engineered Wood Products . . . . . . . . . . . . . . . . . . . . . 405
Christian Lehringer, Joseph Gabriel

Part IV: Timber and Concrete/Cement/Polymer

Development of a High-Performance Hybrid System Made
of Composites and Timber (High-Tech Timber Beam ) . . . . . . . . . . . . . . 423
Markus Jahreis, Martin Kstner, Wolfram Hdicke, Karl Rautenstrauch
Experimental Study of the Composite Timber-Concrete SBB
Connection under Monotonic and Reversed-Cyclic Loadings . . . . . . . . . . 433
Manuel Manthey, Quang Huy Nguyen, Hugues Somja, Jrme Duchne,
Mohammed Hjiaj
The Predictive Model for Stiffness of Inclined Screws as Shear
Connection in Timber-Concrete Composite Floor . . . . . . . . . . . . . . . . . . . . 443
F. Moshiri, R. Shrestha, K. Crews

Shear Performance of Wood-Concrete Composite with Different

Anchorage Length of Steel Rebar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
Sang-Joon Lee, Jrme Humbert, Kwang-Mo Kim, Joo-Saeng Park,
Moon-Jae Park
Development of Prefabricated Timber-Concrete Composite Floors . . . . . 463
Petr Kuklk, Pavel Nechanick, Anna Kuklkov
Wood-Based Construction for Multi-story Buildings: Application
of Cement Bonded Wood Composites as Structural Element . . . . . . . . . . 471
Alireza Fadai, Michael Fuchs, Wolfgang Winter
Rehabilitation, Upgrading and Repair of Historic Timber Structures
with Polymer Concrete and FRP-Reinforcement . . . . . . . . . . . . . . . . . . . . 485
Markus Jahreis, Karl Rautenstrauch
Contents XIII

Fatigue Performance of Single Span Wood-Concrete-Composite

Bridges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
Leander Bathon, Oliver Bletz-Mhldorfer
Hybrid Wall-Slabs for Multi-storey Buildings: Made of Timber
with a Directly Applied Mineral Cover Layer . . . . . . . . . . . . . . . . . . . . . . . 499
Christian Dorn, Alexander Stief, Markus Jahreis, Karl Rautenstrauch
An Innovative Prefabricated Timber-Concrete Composite System . . . . . . 507
Roberto Crocetti, Tiziano Sartori, Roberto Tomasi, Jos L.F. Cabo

Part V: Cyclic, Seismic Behaviour

A Component Model for Cyclic Behaviour of Wooden Structures . . . . . . 519
Giovanni Rinaldin, Massimo Fragiacomo
Overview of a Project to Quantify Seismic Performance Factors
for Cross Laminated Timber Structures in the United States . . . . . . . . . . 531
M. Omar Amini, John W. van de Lindt, Shiling Pei, Douglas Rammer,
Phil Line, Marjan Popovski
Force Modification Factors for CLT Structures for NBCC . . . . . . . . . . . . 543
Marjan Popovski, Shiling Pei, John W. van de Lindt, E. Karacabeyli
Experimental Testing of a Portal Frame Connection Using Glued-In
Steel Rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
James Walker and Robert Xiao

Part VI: Hardwood, Modified Wood and Bamboo

Bending Strength and Stiffness of Glulam Beams Made of Thermally
Modified Beech Timber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569
Robert Widmann, Wilfried Beikircher, Jos L.F. Cabo, Ren Steiger
Structural Veneer Based Composite Products from Hardwood
Thinning Part I: Background and Manufacturing . . . . . . . . . . . . . . . . . . 577
Ian D. Underhill, Benoit P. Gilbert, Henri Bailleres, Robbie L. McGavin,
Dale Patterson
Glue Laminated Bamboo (GluBam) for Structural Applications . . . . . . . 589
Y. Xiao, B. Shan, R.Z. Yang, Z. Li, J. Chen
Glulam Composed of Glued Laminated Veneer Lumber Made
of Beech Wood: Superior Performance in Compression Loading . . . . . . . 603
Gerhard Dill-Langer, Simon Aicher
XIV Contents

Structural Performance of Accoya Wood under Service Class 3

Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615
Julian Marcroft, Ferry Bongers, Fernando Perez Perez, John Alexander,
Ian Harrison
Structural Veneer Based Composite Products from Hardwood
Thinning Part II: Testing of Hollow Utility Poles . . . . . . . . . . . . . . . . . . . 629
Benoit P. Gilbert, Ian D. Underhill, Henri Bailleres, Robbie L. McGavin
Glulam from European White Oak: Finger Joint Influence
on Bending Size Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
Simon Aicher, Gordian Stapf
Non-homogeneous Thermal Properties of Bamboo . . . . . . . . . . . . . . . . . . . 657
Puxi Huang, Wen-Shao Chang, Andy Shea, Martin P. Ansell,
Mike Lawrence

Part VII: Cross-Laminated Timber

Tapered Beams Made of Cross Laminated Timber . . . . . . . . . . . . . . . . . . . 667
Marcus Flaig, Hans Joachim Bla
Influence of the Connection Modelling on the Seismic Behaviour
of Crosslam Timber Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677
I. Sustersic, B. Dujic, Massimo Fragiacomo
Behaviour of Cross-Laminated Timber Panels under Cyclic Loads . . . . . 689
Igor Gavric, Massimo Fragiacomo, Marjan Popovski, Ario Ceccotti
CLT Plates under Concentrated Loading Experimental
Identification of Crack Modes and Corresponding Failure
Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 703
Georg Hochreiner, Josef Fssl, Josef Eberhardsteiner, Simon Aicher
Seismic Strengthening of Existing Concrete and Masonry Buildings
with Crosslam Timber Panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713
I. Sustersic, B. Dujic
In-Plane Stiffness of Traditional Timber Floors Strengthened
with CLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725
Jorge M. Branco, Milos Kekeliak, Paulo B. Loureno
Propose Alternative Design Criteria for Dowel Type Joint With CLT . . . 739
Shoichi Nakashima, Akihisa Kitamori, Takuro Mori, Kohei Komatsu
Contents XV

Part VIII: Properties and Testing of Wood

Length Effects on Tensile Strength in Timber Members With
and Without Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
R. Brandner, G. Schickhofer
New Perspectives in Machine Strength Grading: Or How to Identify
a Top Rupture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 761
Julia K. Denzler, Andreas Weidenhiller
Aspects of the Difference between the Local and Global Modulus
of Elasticity of Structural (hardwood) Timber . . . . . . . . . . . . . . . . . . . . . . 773
G.J.P. Ravenshorst, P.A. de Vries, Jan-Willem van de Kuilen

Part IX: Glulam

A Study of Australian Glulam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787
H.R. Milner, Con Y. Adam
Improving Strength of Glulam Laminations of Norway Spruce
Side Boards by Removal of Weak Sections Using Optimized
Finger Jointing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 801
Jan Oscarsson, Anders Olsson, Bertil Enquist
Double Span Continuous Glulam Slabs Strengthened with GFRP . . . . . . 813
Jorge M. Branco, Marco P. Jorge, Jos Sena-Cruz
Simplified Design of Glued Laminated Timber Girders
for the Torsional Moment Caused by Stability Effects . . . . . . . . . . . . . . . . 823
R. Hofmann, Ulrike Kuhlmann

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831

Keyword Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 835

Part I
Horizontal Displacements in Medium-Rise
Timber Buildings: Basic FE Modeling in
Serviceability Limit State

Ida Nslund and Helena Johnsson

Division of Structural and Construction Engineering, Lule University of Technology,

SE-971 87 LULE Sweden

Abstract. Higher and larger timber buildings are built today. The building is ex-
posed to static lateral wind loads that cause displacements that might lead to dis-
comfort and non-function of the building. To determine the size and behavior of
horizontal displacements, two timber systems has been studied, a light frame sys-
tem with shear walls and a post and beam system with diagonal bracing. The sta-
bilizing wall segments have been analyzed with a FE model and subjected to static
lateral wind load and vertical dead load in serviceability limit state. Both plywood
and particle board were used as sheet materials. To reduce flanking transmission
Sylomer is applied in the light frame system. The total lateral displacement va-
ries between 4 to 125 mm in the light frame system and 2 to 5 mm in the post and
beam system. Removing the Sylomer damping material from the light-frame
system would decrease the lateral displacements with 1.5-3 times, which needs to
be further investigated.

Keywords: Stabilizing element, finite element model, lateral displacement, tim-

ber-frame, post and beam, serviceability limit state.

1 Introduction
Higher and larger timber buildings are constructed as progression in technology
moves forward. Serviceability aspects such as displacements, movements, and
deformation become more important when designing higher buildings. When ex-
posed to horizontal action e.g. from wind, a medium-rise building can experience
large lateral displacements due to the low mass density of wood. There are differ-
ent ways to stabilize medium-rise timber buildings against static lateral wind load.
In light-framed systems shear walls are used for stabilization and in beam and post
buildings different principles are used e.g. diagonal bracing, moment resisting
connections, and wall diaphragms. Shear walls take up load in their own plane and
work through diaphragm action in roof and floors. The diagonal bracing can be
wooden and work in both compression and tension or a steel chord which works
only in tension.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 3
DOI: 10.1007/978-94-007-7811-5_1, RILEM 2014
4 I. Nslund and H. Johnsson

To transfer load between structural elements connections are used. How the
connection acts between floors and foundation is important for the whole struc-
ture. Depending on the design the strength and stiffness will vary in the connec-
tion. It must be designed so unexpected slip is avoided. Slip is also generated in
materials used for vibration damping in connections. There is no limit established
for static displacements in structures in Eurocode 5 (2009), merely on single
building elements as beams. Clients are responsible for setting up displacement
limits in tendering. In an early version of Eurocode 5 (2009) a maximum of hori-
zontal displacement was set to H/300 were H are the height, but this has been
removed. The Recommendation in the German design code is H/500 (Kllsner &
Girhammar, 2008).
Depending on the building system and stabilization method, lateral displace-
ments will vary. To qualify as a stabilizing wall segment it needs to be continuous
from the ground to the top of the house. A building consists of several stabilizing
walls that must resist lateral load, but in this study one wall is studied.
Figure 1 and Figure 2 present stabilizing wall segment for a shear wall system
and a diagonal bracing system respectively. Several models to predict the capacity
of a shear wall subjected to static lateral load has been proposed, both analytical
and numerical e.g. Kllsner & Girhammar (2009); Falk & Itani (1989); Kllsner
(1984); Foschi (1977). Basic FE modeling is made for two different cases, a light-
frame system and a beam and post system. The light frame model consists of
studs, sheets, sheathing-to-frame nails, anchorage and Sylomer to reduce flank-
ing transmission. Sheets will be either plywood or particle board. The post and
beam structure is stabilized by wooden diagonals connected by bolts.
The aim is to study the total static horizontal displacement in stabilizing walls
in a light-frame system and a post and beam system in serviceability limit state.
Second order theory is not taken into account in this study. The frequent load
combination was chosen to represent serviceability limit state.

1.1 Case Study Houses

The case study houses are medium-rise apartment buildings with a timber frame.
They are 3-5 storeys, which correspond to a height of 11 to 18 meter. The length
of the houses varies between 24 to 62 meters and the width between 9 to 36
meters. The number of stabilizing walls in each house varies between 1 to 9 walls
with a length of 2.4 to 9.6 meter. No openings and walls shorter than 1.2 meter
counts as a stabilizing segment because of high loads at the end of the walls
(Kllsner & Girhammar, 2008). For the beam and post system, tr8 is used as an
example (Tlustochowicz, 2011). The wind is applied according to Eurocode 1
(2009) and distributed through rigid floor diaphragms to the stabilizing element;
see Figure 1 and Figure 2.
Horizontal Displacements in Medium-Rise Timber Buildings 5

Dead load
Lateral displacement

Wind load

Fig. 1 Shear wall segment, see figure 3 for detail A

Dead load
Lateral displacement

Wind load

Fig. 2 Post and beam wall segment, see figure 3 for detail B

2 Materials and Method

2.1 Light-Frame System

The timber-framed wall consists of timber of strength class C24 (SS-EN 338,
2009) and two different sheathing materials; plywood and particle board (EN
12369-2, 2011; SS-EN 312). In Table 1 material properties are presented. The
cross-section of the timber element was set to 45x120 mm2 and the thicknesses of
6 I. Nslund and H. Johnsson

the sheets were 9 mm and 12 mm respectively for plywood and particle board.
The width and height of one wall segment is 1.2x3.0 m2. To minimize flanking
transmission between floors a damping material, Sylomer is applied under each
stud, between the upper and lower sills. Sylomer is a polyurethane (PUR) elas-
tomer material that transfers compression loads and small shear loads. Different
stiffness of Sylomer is used for different compression loads. Due to production
ease, the same stiffness is often used on at least two floors. Material properties for
Sylomer are assumed according to the product sheet from the supplier and are
presented in Table 2. Anchorage devices are applied between each floor to resist
the tensile up-lift, 15 nails are applied. The stiffness of the anchorage is calculated
according to Eurocode 5 (2009) and is presented in Table 3.

Table 1 Material properties

Timber Glulam Plywood

C24 GL28c P30
Elastic modulus
11000 12600 1050 2000
Shear modulus
Not used Not used 500 960
Poissons ratio Not used Not used 0.05 0. 4
Density [kg/m3] 420 380 Not used Not used

Table 2 Stiffness of Sylomer

SR 850 SR 1200
304 342
2736 3952

Sheets are connected to the frame by nails. The stiffness of the sheathing-to-
timber joints are calculated with data collected from U. A. Girhammar, Bovim, &
Kllsner (2004). In U. A. Girhammar, Bovim, & Kllsner (2004) tension tests on
sheathing-to-timber joints is presented for three different sheet materials; hard-
board, particle board and plywood. The tests were performed both perpendicular
and parallel to the grain of the frame member. Data from the test were analyzed
here-in and the slip modulus calculated according to SS-EN 26 891. In the finite
element model the sheathing-to-timber joints are modeled by two springs in each
corner, x- and y-direction in Figure 3. The fastener force distribution varies along
the framing member and the resulting stiffness of the joints is not the number of
fasteners times the stiffness (Kllsner & Girhammar, 2009). A numerical calibra-
tion was performed to arrive a stiffness value for the corner springs to simulate
nailing along the perimeter of the sheathing. The distance between fasteners was
Horizontal Displacements in Medium-Rise Timber Buildings 7

set at 200 mm with fasteners on the perimeter studs only. The numerical analysis
was performed in Matlab R2012b with properties that are described in the para-
graph finite element model. The stiffness was determined by comparing lateral
displacements for the two models and adjusting the stiffness of the corner springs
until it coincided with the stiffness of the model with springs around the perimeter,
Table 3.

Table 3 Stiffness of connections

Type Plywood Particleboard Anchorage Bolts

2564 3566 Not used Not used
1318 2437 26093 Not used
K [kNm] Not used Not used Not used 106

2.2 Post and Beam System

In the post and beam system glulam GL28c (SS-EN 1194, 1999) was used with a
cross-section of 360x160 mm2. In Table 1 material properties are listed. To
connect the diagonal bracing four bolts were applied. The rotational stiffness was
calculated according to Eurocode 5 (2009) and presented in Table 3. The post and
beam system has no Sylomer and no anchorage devices.

2.3 Load
Wind load was calculated according to Eurocode 1, (2009) with a basic wind
velocity of 24 m/s (Stockholm, Sweden) and a pressure factor of 1.1. The wind is
exponentially distributed over the height of the building. The frequent load com-
bination was used, which represents a reversible state with temporary damage e.g.
temporary deflection of a beam due to a short high loads (Eurocode 0, 2009).
Dead loads are applied on the building: 0.3kN/m2 for the walls and 0.6kN/m2 for
the slabs where the ceiling is included.

2.4 Finite Element Model

For the calculation Matlab R2012b with a finite element toolbox, CALFEM (Ola
Dahlblom et al., 1986) was used. The toolbox is based on (Niels & Saabye Otto-
sen, 1992). The light-frame model consists of three basic elements: frame mem-
ber, a sheathing member and fasteners. The frame members are modeled as beam
elements with three degrees of freedom in each node. A plane stress
8 I. Nslund and H. Johnsson

isotropic element is used for the sheathing with two degrees of freedom in each
node (four node element). Two springs represent fasteners between the frame
member and the sheet which are connected in the translation direction only to the
corner. In Figure 3 a) all connections are displayed. Since the studs are pinned
towards the top and bottom rail an additional degree of freedom (rotation) was
added in those nodes that contain frame members, since the studs and top/bottom
rail will not have the same rotation. The top/bottom rail will behave as a conti-
nuous beam. To model the Sylomer between each floor, single spring pairs were
used in the translation direction. All floors are connected to each other with hold-
downs which only can take tension load and are placed at the end of the wall.
Hold-downs are represented by a spring in the model in the vertical direction. The
bottom rail at the first floor is hindered to move in any translational direction, it is
rigid to the foundation.




Sylomer Pinned
Pinned Column

Sheating-to-frame joints Frame Sheet

a b
Fig. 3 Detail A in the shear-wall system and detail B in the beam-and-post system

The post and beam system consists of columns and beams that are modeled as
beam elements, Figure 3b. Diagonals are also modeled as beam elements and are
connected with bolts to the columns. The bolted connection is represented by rota-
tional springs. Diagonals can take both tension and compression. Columns are pin
connected to the beams (additional degree of freedom) and the beams are modeled
as continuous. The bottommost columns are hindered to move in any translational
direction and are rigid to the foundation
All springs have two nodes where each node has one degree of freedom. For
springs, different types are available see e.g. Judd & Fonseca (2005). In this study
all springs behave linearly due to the serviceability limit state.
The wind load is distributed to each beam element node. Load distribution are
set to 36 m and distributed to 5 walls in the light-frame model and distributed to 2
walls in the post and beam system. In the light-frame model the wind load is di-
vided in two since walls often consist of sheets on both side of the frame. Dead
load is applied to all beam element nodes.
Horizontal Displacements in Medium-Rise Timber Buildings 9

3 Result

For the system with shear walls as stabilizing elements, the number of shear walls
has been varied using two different sheathings. The number of floors in the build-
ing has also been varied. For the post and beam system, the number of floors and

a b

c d

e f
Fig. 4 Building height-top corner horizontal displacement curves for the light frame system,
a) - c) 3-5 storeys plywood, d) - f) 3-5 storeys particle board
10 I. Nslund and H. Johnsson

the number of wall segments was varied. Figure 4 a)-c) present top corner
horizontal displacements (Figure 1) for 3-5 storeys stabilized with shear walls of
plywood. Referring to Figures 4b) and c): using two or three wall segments pro-
duces a non-linear response. This is due to bending of the stabilizing segments and
changing stiffness of Sylomer higher up in the building. In Fig. 4a) the response
is almost linear since the building is low. The results for the particle board in Fig-
ure 4 d)-f) show the same overall behavior, but the response is stiffer overall. In
the light frame system the displacements vary between 6 125 mm for plywood
and 4 102 mm for particle board. If removing the Sylomer from the FE simula-
tion, the horizontal displacements are 1.5-3 times lower i.e. maximum lateral
displacement amounts to 42 mm in the light-frame system
Figure 5 presents the post and beam system where the top corner displacements
(Figure 2) vary little with the number of wall segments and the behavior is linear.
The displacement requirements H/300 and H/500 are fulfilled easily. The post and
beam system shows a displacement between 2-5 mm.

a b

Fig. 5 Building height-top corner horizontal displacement curves for the post and beam
system a)-c) 3-5 storeys
Horizontal Displacements in Medium-Rise Timber Buildings 11

4 Conclusion and Discussion

This study provides an overview of the possible horizontal displacement range for
medium-rise timber building stabilized by either a light frame system or a post and
beam system. To stabilize the light-frame system, 5-8 wall segments of 1.2 m
length is a minimum if considering the frequent load combination. When taking
into account any eccentrical loading, this number may increase. In the beam and
post system, horizontal displacements are much smaller. For several of the wall
configurations in the light-frame system, the horizontal displacement fall short of
the recommendations, (H/300 or H/500) which shows that it is important to check
displacements in higher buildings. The analysis here-in was performed with the
frequent load combination, which represents a load value that is exceeded 1% of
the life-time of the structure. Since the light-frame model considers pinned
connections where the real behavior is semi-rigid, the model presented here is
somewhat weaker than the actual case. Sylomer seems to affect the building
negatively when looking at horizontal displacements, further investigations are
Further research is to; model these semi-rigid connections, study the effect of
Sylomer in detail and also produce a 3D model that considers the surrounding

Acknowledgements. The authors wish to acknowledge the financial support from the cen-
tre for Lean Wood Engineering sponsored by VINNOVA in Sweden and SWECO struc-
tures in Sweden. The authors also wish to thank Bertil Nslund for advice in programming.

EN 12369-2, Wood-based panels characteristic values for structural design part 2. Ply-
wood (2011)
Eurocode 0, SS-EN 1990: Eurocode 0 basis of structural design. SIS, Stockholm (2009)
Eurocode 1, SS-EN 1991-1-4 (2005): Eurocode 1: Actions on structures - part 1-4. General
actions - wind actions. SIS, Stockholm (2009)
Eurocode 5, SS-EN 1995-1-1(2004): Eurocode 5: Design of timber structures - part 1-1.
General - common rules and rules for buildings. SIS, Stockholm (2009)
Falk, R.H., Itani, R.Y.: Finite-element modeling of wood diaphragms. Journal of Structural
Engineering 115(3), 543559 (1989)
Foschi, R.: Analysis of wood diaphragms and trusses - 1 diaphragms. Canadian Journal of
Civil Engineering 4(3), 345352 (1977)
Girhammar, U.A., Bovim, N.I., Kllsner, B.: Characteristics of sheathing-to-timber joints in
wood shear walls. In: 8th World Conference on Timber Engineering, Lathi, Findland
Judd, J.P., Fonseca, F.S.: Analytical model for sheathing-to-framing connections in wood
shear walls and diaphragms. Journal of Structural Engineering-ASCE 131(2), 345352
12 I. Nslund and H. Johnsson

Kllsner, B.: Panels as wind-bracing elements in timber-framed walls, Stockholm, Sweden

Kallsner, B., Girhammar, U.: Analysis of fully anchored light-frame timber shear walls-
elastic model. Materials and Structures 42(3), 301320 (2009)
Kllsner, B., Girhammar, U.A.: Horizontal stabilising of light frame timber structures.
Plastic design of wood-framed shear walls. SP Technical Research. Institute of Sweden,
2008:47 (2008) (in Swedish)
Niels, S.O., Saabye Ottosen, N.: Introduction to the finite element method. Prentice Hall,
New York (1992)
Dahlblom, O., Peterson, A., Petersson, H., Dahlblom, O., Peterson, A., Petersson, H.:
CALFEM - a program for computer: aided learning of the finite element method. Engi-
neering Computations: International Journal for Computer: Aided Engineering and
Software 3(2), 155160 (1986)
SS-EN 1194, Timber structures - glued laminated timber - strength classes and determina-
tion of characteristic values (1999)
SS-EN 26 891, Timber structures joints made with mechanical fasteners general prin-
ciples for the determination of strength and deformations characteristics (1991)
SS-EN 312, Particleboards - specifications (2010)
SS-EN 338, Structural timber strength classes (2009)
Tlustochowicz, G.: Stabilising system for multi-storey beam and post timber buildings.
Doctoral thesis, Lule Univeristy of Technology (2011)
Improving the Moment Resistance
of a Concealed Timber Post Base Joint

Jrme Humbert*, Sang-Joon Lee, Joo-Saeng Park, and Moon-Jae Park

Korea Forest Research Institute (KFRI), 57 Hoegiro, Dongdaemun-gu, Seoul, Korea


1 Introduction
The Korean traditional house, Hanok, is a post-and-beam timber structure with a
large tiled roof designed to keep the house cool in summer (Fig. 1). The solid
wood members are connected together without metallic connectors, using complex
woodwood joints, while the posts are fitted inside cornerstones. Associated with
the stabilizing effect of the heavy roof, this provides a basic lateral resistance to
the house, which historically proved effective against the few earthquakes which
occurred in this area of low seismicity. Nowadays, the Hanok house is still fairly
popular among Koreans thanks to its cultural, aesthetic, and eco-friendly aspects.
However, some concerns include e.g. a limited thermal insulation, the presence of
woodwood connectors with brittle behavior, and a design not suited to high-
rising structures. In a country where over one-third of the population lives in the
metropolitan area around the capital city of Seoul, the lack of space implies
favoring multi-story buildings over single-story ones. Moreover, the recent
seismic events in the region raised concerns on the seismic resistance of buildings.
Additionally, the complexity and diversity of traditional timber joints require
skillful artisans and are not well-suited to industrialized processes. Finally, in an
eco-friendly approach, the structural use of domestic wood species implies using
engineered wood products to overcome limited mechanical properties.
In this context, the Korea Forest Research Institute (KFRI) developed a new
modernized design of multi-story timber house based on the visual identity of
Hanok and aiming at addressing the mentioned issues. A modular hybrid design
was proposed, coupling an outer post-and-beam structure with inner lateral-
resistant elements 66. Using engineered timber members and metallic connectors,
the outer post-and-beam structure provides the necessary resistance for vertical
loads while ensuring a safer design. Indeed the semi-rigid metallic connectors
provide a source of increased ductility and energy dissipation, and prevent the
unsafe brittle fracture of wood. Moreover, the use of glued laminated (glulam)
timber members associated with factory pre-cutting together provide reduced
mechanical and geometrical variabilities, and thus better-quality constructions
with easier assembly than what is obtained with on-site works using naturally
highly-variable solid wood elements.

Corresponding author.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 13
DOI: 10.1007/978-94-007-7811-5_2, RILEM 2014
14 J. Humbert et aal.

Fig. 1 A traditional Korean timber house Hanok composed of an apparent post-and-beaam

structure and a heavy tiled ro
oof. (Source: Wikipedia)

Concerning the lateraal resistance, the post-and-beam structure is known tto

intrinsically display limitted performances. Therefore, additional lateral-resistannt
elements are used. Severral technologies were considered, including the heavilyy-
studied light-frame shearrwalls with sheathed OSB panels, which served as a
reference design 6, and thhe structural insulation panels (SIP), which provides botth
a lateral resistance and an
n increased thermal insulation 6.
One concern regarding g the development of this hybrid timber house design waas
to maintain the visual ideentity of the Hanok. In particular, replacing woodwoood
connections with metalliic connectors in the apparent post-and-beam structurre
called for the use of co oncealed metallic connectors. Therefore new metalllic
connectors were designed d, based on existing commercial models adapted to suuit
the modular hybrid design n developed.
In this study, we investigate the moment resistance of a concealed post basse
joint connector under monotonic
m and reversed cyclic loading. Although thhe
primary use of the post-and-beam structure is to undergo vertical loadings, thhe
enclosing configuration of o the outer glulam members around the inner lateraal-
resistant structure has an effect
e on the overall mechanical behavior of the structurre
under lateral loading 6. Moreover,
M although fairly limited the lateral resistance oof
the outer post-and-beam structure
s alone is not completely negligible as compareed
to those of the inner light--frame shearwall. Finally, assessing the rotational rigiditty
and other moment-related d mechanical properties of the joint is of interest both foor
structural design and num merical modeling.
Results of bending testts were already reported in 6.
In the following, we present
p in a first step experimental tests under reverseed
cyclic loading, and a stu udy of the moment-resisting properties of the post basse
joint. After concluding onn the limited performance of the connector, we propose iin
a second step some mod difications of the connector to improve its ductility annd
energy dissipation. A new w series of experimental tests is then conducted to validaate
Improving the Moment Resistance of a Concealed Timber Post Base Joint 115

these modifications. In paarallel, a refined 3D finite element (FE) model is derived,

and results from numericaal simulations are compared with experimental ones.

Keywords: post and beam

m structure, cyclic loading, metallic connections, ductility..

2 Experimental Tests

2.1 Materials and Methods

A picture of the metallicc connector is presented on Fig. 2 along with its maiin
dimensions. A schematic diagram of a quarter of the lower part of the connector is
presented in Fig. 3. In thee upper part, it is composed of four vertical metal wingss
arranged to form a cross, and connected to the timber member using four metalllic
pins of d = 16 mm diam meter. The pins are grouped in two pairs arrangeed
perpendicularly at 20 mm m and 50 mm from the top of the connector. This ensurees
that a sufficient end distaance is maintained when connecting the timber membeer.
For reference, the Europeean design codes 6 prescribe a minimum timber membeer
end distance of max(7d, 80 mm) = 112 mm. In these tests we comply with thhis
The lower part of the connector
c is composed of two 170 mm 170 mm paralllel
horizontal metal plates, with
w a 30 mm gap between them leaving enough space foor
the bolt nut and washer. TheT upper plate features four 30 mm holes to insert thhe
bolt nuts. The lower platee has four 20 mm holes for the anchor bolts. The platees
are connected together wiith small vertical metallic segments located both directlly
under the wings and alon ng one of the half-borders of the connector, leaving thhe
other half-border open forr the insertion of a wrench. All metallic plates composinng

Fig. 2 Metallic connector

16 J. Humbert et aal.

Fig. 3 Diagram of the lower section of the connector showing the position and length of thhe
welds (thicknesses not to sccale). (1) Full-length weld between the upper face of the uppper
plate and the vertical wings.. (2) Short weld between the lower face of the upper plate annd
the vertical segments.

the connector have a thickkness of 5 mm. The welding of the metallic parts togetheer
is presented in the diagram
m on Fig. 3. In the upper part, the wings are welded botth
together and to the upper horizontal plate of the base along their full length (welld
(1) on Fig. 3). This is cllearly visible on Fig. 2. In the lower part however, thhe
welds between the horizontal plates and the vertical segments do not cover the fuull
length of the pieces (weld (2) on Fig. 3). This difference has a strong impact oon
the mechanical behavior, as explained below in the experimental results section.
Improving the Moment Resistance of a Concealed Timber Post Base Joint 117

The post-and-beam structure of the hybrid timber house design is composed oof
Japanese larch (Larix kaeempferi) glulam timber members with a cross-section oof
180 mm 180 mm, fix xed at their base to the studied metallic connectorrs.
Therefore, in this study we
w perform pseudo-static reversed cyclic moment tests oon
the joint using timber sppecimens with the same wood species and cross-sectioon
dimensions. The specimeens of timber posts are, depending on the tests, eitheer
800 mm or 1000 mm talll, and are horizontally loaded at their top with a 100 kkN
oil jack, effectively appllying a moment load to the joint. The reversed cycllic
loading (Fig. 4) is based on the ISO 16670 protocol 6, and is composed of cyclees
of increasing amplitude, namely
n 1, 5, 10, 20, 40, and 60 mm, followed by a finnal
monotonic ramp up to 80 0 mm which usually leads to the failure of the joint. Thhe
loading speed is 1 mm/ss, which allows keeping the test duration within a 330
minute span while avoidin ng possible dynamic effects.
The internal load cell of the oil jack and an external laser displacement sensoor
provide a measurement of o the rotation and moment in the joint. Because of thhe
experimental setup and the
t effect of rotation, for large displacements the actuual
horizontal displacement as measured by the laser sensor at constant height (aas
shown on Fig. 4) differss slightly from the displacement of the actuator itsellf.
Therefore, in order to compute the rotation angle we use the displacement valuue
measured by the externaal laser sensor rather than the internal measure of thhe
actuator stroke. The momment is derived directly from the load cell and the height at
which the load is applied. In order to monitor a possible differential displacemennt
between the connector an nd the post after damaging started, an additional linear
variable differential traansformer (LVDT) displacement sensor provides a
measurement of the verticcal displacement of the base of the post. The base of thhe
connector is affixed to the metallic testing frame with four bolts of 20 mm m
diameter representing the anchor bolts usually used with post base connectors.
Experimental tests are performed at least two times per configuration (see otheer
configurations below) in order
o to minimize experimental discrepancies.

Fig. 4 Reversed cyclic displaacement loading as recorded by the external laser sensor
18 J. Humbert et aal.

2.2 Experimental Results

The reversed cyclic loadin ng applied to the joint always leads to its failure througgh
the damaging of the meetallic connector. In the initial configuration presenteed
above, the glulam membeer is not damaged, nor are the metallic pins. The failurre
occurs only in the connector base. The connector first begins to plasticize; namelly,
the upper plate begins to o bend. However, this phenomenon is brief, as the loaad
quickly reaches the resistance limit of the welds, leading to their tearing. Morre
specifically, the short welds
w binding the upper horizontal metal plate of thhe
connector with the verticcal elements underneath, labeled weld-(2) on Fig. 3,
experience the earliest faiilure as shown on Fig. 5. Indeed, their length is limited tto
a mere 20~30 mm, and d therefore they constitute the weakest point of thhe
connector under this tractiion-inducing moment loading.

Fig. 5 Failure of short-lengtth welds on the lower face of the upper horizontal plate of thhe
connector base

Fig. 6 presents the evo

olution of the moment M in the connector function of thhe
rotation angle . The joint
j displays a semi-rigid non-linear behavior witth
hysteresis loops typical of timber joints with metal connectors. The momennt
resistance is 11 kN.m in n the positive direction and 8 kN.m in the negativve
direction, and the initiall elastic rotational stiffness is approximately equal tto
460 kN.m/rad. We can no ote an asymmetry in this behavior, which is due in part tto
the loading history: the co
onnector is first loaded on the positive side, and then oon
the negative side. The dammaging of the connector when loaded on the positive sidde
lowers its overall resisttance, leading to a lower maximum moment wheen
Improving the Moment Resistance of a Concealed Timber Post Base Joint 119

subsequently loaded on thet negative side. In this particular test, an experimenttal

error also adds up to the lo
oading to increase this asymmetry.
However, in the preseent case we can clearly see two different parts in thhe
evolution. Initially, the evolution
e displays a high rotational stiffness. Yet, in a
second step after the tearring of the welds the rotational stiffness is much loweer,
and so is the resistance. Despite
D preventing the brittle fracture of wood in favor oof
the plasticizing of the metallic
m connector, this behavior limits the ductility annd
energy dissipation of the joint.

ment M in the joint function of the rotation angle

Fig. 6 Evolution of the mom

Fig. 7 Bilinear model (EEEP curve) for the original configuration, displaying a low ductility
20 J. Humbert et aal.

In order to quantify th he ductility, a possible approach is to use the Equivalennt

Energy Elastic-Plastic (EE EEP) method from 6. The EEEP curve corresponding tto
the test result presented in
n Fig. 6 is shown on Fig. 7. It is obtained using the secannt
at 10% and 40% of the peeak moment Mmax as the initial elastic rotational stiffnesss.
The ultimate moment Mu is computed to equal the energies. The ductility m is theen
defined as the ratio of u over
o v. In this case, we obtain a ductility value m = 1.46,
which is close to 1.0 (elasstic fragile behavior) and indicates a low ductility for thhe

Fig. 8 FE mesh of the specim

Improving the Moment Resistance of a Concealed Timber Post Base Joint 21

Concerning the uplift, the measured values show no significant relative

displacement between the metallic connector and the timber member. This is
consistent with post-test observations which do not show any damage in the pins
or the timber member.
In conclusion, the failure of the joint occurs in the metallic connector as it could
be expected for a use under reversed cyclic loading. However, considering an
application in a structure undergoing a seismic loading, the ductility and energy
dissipation of this joint remain low because of the failure in the welds rather than
in the pins, as usually found in pinned timber connectors.

3 Numerical Study
In order to complement the experimental results, and as a preparation for future
modelings of a complete timber house, we perform a numerical analysis using a
refined Finite Element (FE) model.

3.1 Numerical Model

The FE model of the joint is composed of a mesh of approximately 105 second-
order volume elements (20-node hexahedra and 10-node tetrahedra) with a size
ranging from 50 mm in areas of lesser interest down to 1 mm in areas of
concentrated stresses, i.e. typically around pin holes and weldings (Fig. 8). We
perform static nonlinear simulations using the FE analysis software ANSYS. The
nonlinearities arise only from the presence of contact between the wood and the
metal. The base of the connector is fixed, and a monotonic load is applied on the
upper part of the wood member on an area of 100 mm height (0.018 m2) roughly
equal to the size of the loading metal plates used in the experimental tests, in the X
direction as shown on Fig. 8.
The material parameters of the FE model are summarized in Table 1. The wood
is modeled using a linear elastic orthotropic material with a density w of
520 kg/m3, a longitudinal elasticity modulus Ew,L of 14.2 GPa (along the column
long side, vertically), and a tangential modulus of elasticity Ew,T of 1 GPa. Those
parameters correspond to a larch species as reported in 6. The steel of the
connector and the pins is modeled with a linear elastic isotropic material with an
elasticity modulus Es of 210 GPa and a density s of 7850 kg/m3.

Table 1 Material properties of the FE model

Material Parameter Value

Density w = 520 kg/m3
Wood Elasticity modulus (L) Ew,L = 14.2 GPa
Elasticity modulus (T) Ew,T = 1 GPa
Density s = 7850 g/m3
Elasticity modulus Es = 210 GPa
22 J. Humbert et aal.

3.2 Numerical Sim

mulation Results
The numerical results shoow a concentration of high stresses around the pins and iin
the lower part of the metaallic connector, that is in the regions of mesh refinemennt
(Fig. 9). The maximum sttress in the connector is reached at the base, between thhe
horizontal upper plate and
a the vertical wing above it. Qualitatively, this is
consistent with the observ
vations of the experimental tests since this is precisely thhe
location of the failure.

Fig. 9 Stress in the metallic parts

p of the joint
Improving the Moment Resistance of a Concealed Timber Post Base Joint 23

Quantitatively, because these numerical results are computed using materials

with a linear elastic behavior, there is no strain or stress limitation during the
simulation. Therefore, additional criteria must be verified a posteriori to ensure
physical consistency. The selected criteria are summarized in Table 2. For the wood,
the embedding strength of wood fh is taken equal to 35 MPa for a load parallel to
wood grain 6, and the tensile strength ft is taken equal to 3 MPa for a load
perpendicular to wood grain 6. The ultimate strength of the steel constituting the
connector itself and the pins is taken equal to s,u = 460 MPa. As seen on Fig. 9,
this limit is reached for a load of approximately 10 kN. This value is consistent
with the resistance of the connector observed in the experimental tests.

Table 2 Physical criteria for FE model post-processing

Material Parameter Value

Embedding strength fh = 35 MPa
Tensile strength ft = 3 MPa
Steel Ultimate strength s,u = 460 MPa

4 Improving the Joint

Based on those results, we can draw two main conclusions. First, the failure of the
joint occurs in the metallic connector rather than in the timber member. This is
generally considered safer because the sudden brittle fracture of wood is replaced
by the much slower plasticizing of steel, allowing for a visual hinting about an
imminent mechanical failure. Conversely, despite the failure occurring in the
metallic connector, the experimental tests underlined the relatively low ductility
and energy dissipation of the joint. Therefore, when considering even a moderate
seismic event the energy released during the earthquake cannot be absorbed in the
connectors and need to be dissipated elsewhere, possibly leading to the failure of
structural elements and the collapse of the structure.
Based on these conclusions, we propose improving the joint to enable a greater
ductility and larger energy dissipation. Because in the original configuration the
failure occurs in the weldings in the base of the connector, we consider two
possible modifications. First, for the sake of simplicity, we simply bolt the
connector directly from the upper plate, allowing the metallic wings to be more
directly connected to the ground (Fig. 10). With this setup, we need to cut the end
of the timber member for practical reasons. Alternatively, a second possibility is to
reinforce the welds by extending them to cover the full length of the wings (see
Fig. 5). Finally, a mixed setup is also considered using both modifications.
In the following, we present the experimental results of tests conducted using
those alternative connector configuration, all test parameters being otherwise
equal to those presented earlier for the original joint configuration. We also
present results on a mixed configuration using both approaches at the same time.
24 J. Humbert et aal.

Fig. 10 Plasticizing of the connector

c when using the modified setup with a cut-end timbber
member. The failure still occcurs in the short weldings of the connector, but is delayed bby
the deformation of the upper plate. Full-length weldings at the metal wing base do nnot
experience any damage.

4.1 Cut-End Config

The momentrotation ev volution in the joint for the cut-end configuration is
presented in Fig. 11. As itt was the case for the original configuration, the failure oof
the joint occurs in the weak
w welds. Yet the resistance of the connector (peaak
moment) is lower than the one in the original configuration. The reason is thhe
absence of contact betweeen the lower end of the timber member which was cuut

Fig. 11 Evolution of the moment

m M function of the rotation angle in the cut-ennd
Improving the Moment Resistance of a Concealed Timber Post Base Joint 225

and the upper plate of th he connector. In this configuration, the rotation is onlly
limited by the metal wing gs. However, the failure occurs for a much larger rotation,
over 0.1 rad. This, assocciated with the shape of the hysteresis loops, togetheer
suggest a larger ductility and
a energy dissipation.
The EEEP curve correesponding to the test of Fig. 11 is presented in Fig. 12.
In this case, we obtain an a ultimate moment Mu = 7.5 kN.m lower than the onne
nfiguration. However, the ductility factor is larger, with a
obtained in the initial con
value of m = 2.64 indicating a more ductile behavior as compared to the originnal
configuration (m = 1.46). The
T direct bolting of the upper plate therefore delayed thhe
failure of the welds and allowed
a a greater plasticizing of the metallic connector, aas
shown on Fig. 10, while th he absence of contact lowered the moment resistance.

Fig. 12 EEEP curve (bilinearr model) for the cut-end configuration. Mu = 7.5 kN.m

4.2 Full-Length Welds

In a second step, we consider a configuration with full-length welds only, i.e.
without a cut-end timbeer specimen. The momentrotation evolution in thhis
configuration is presented d in Fig. 13. In this case, the resistance of the joint is
much higher than in th he previous configurations, with a maximum momennt
Mmax = 16.5 kN/m, that is an increase of roughly 50% in resistance. The associateed
EEEP curve is presented in Fig. 14. We obtain a ductility ratio m = 2.36 which is
intermediate between the initial configuration and the cut-end configuration. Thhe
behavior is still ductile, and
a some energy dissipation occurs. The failure occurs iin
the welds again, but wood d fracture is also visible directly above the metal wings.
26 J. Humbert et aal.

ment M function of the rotation angle with full-length welds

Fig. 13 Evolution of the mom

Fig. 14 EEEP curve (bilinear model) for the full-length weld configuration. Mu = 155.9

4.3 Mixed Cut-End

d and Full-Length Welds Configuration
Lastly, we perform exp perimental tests on a mixed configuration using botth
previous approaches at thhe same time: the timber specimen is cut at its lower end,
the upper plate of the con
nnector is directly bolted to the ground, and the welds arre
extended to cover the full length of the metal elements.
Improving the Moment Resistance of a Concealed Timber Post Base Joint 227

Fig. 15 presents the momentrotation

m evolution in the joint. The evolution is
similar to the one obtaained in the cut-end configuration, with a resistancce
Mmax = 10.7 kN/m slightlly higher. Yet the resistance remains much lower thaan
when using full-length welds alone because of the absence of contact. Indeed w we
can observe in the experim
mental tests some fracture directly above the metal winggs,
indicating that the momeent in the area of the pins is larger. The EEEP curve is
presented in Fig. 16 and gives
g a ductility factor m = 2.70, which is slightly higher
than when bolting the low
wer plate to the ground.

Fig. 15 Evolution of the moment M function of the rotation angle with mixeed

Fig. 16 EEEP curve (bilinearr model) for the mixed configuration. Mu = 9.5 kN.m.
28 J. Humbert et al.

Table 3 Comparison of the performance of the joint in the tested configurations

EEEP Ductility
Joint Strength EEEP limit
stiffness factor
configuration Mmax [kN.m] Mu [kN.m]
[kN.m/rad] m [-]
Initial (I) 10.9 9.0 458 1.46
Cut-end (C) 8.8 7.5 214 2.64
Full-length 16.5 15.9 351 2.36
welds (W)
Mixed (C+W) 10.7 9.5 250 2.70

5 Conclusions
We presented in this paper some experimental and numerical results on the
moment resistance of a concealed timber post base joint aimed at replacing in a
modern design introduced lately the woodwood joints used in the traditional
Korean Hanok timber house.
Experimental and numerical results show that the original configuration of the
joint offers a limited moment resistance and a low ductility and energy dissipation.
In an attempt to mitigate those limitations without undergoing major changes in
the connector, three new configurations are proposed and investigated. Motivated
by the wish to prevent the early failure in welds, a first approach consists in
directly bolting the connector's upper plate to lower the stress on the weak welds.
Alternatively, another approach focuses on increasing the strength of these welds
by extending their length to the full width of the metal wings. Finally, a third
configuration investigates the effect of those two approaches combined.
Experimental results on those three alternative configurations show that cutting
the end of the timber member to allow direct bolting of the upper plate also have
the effect of reducing the moment strength of the joint. This is because there is no
more contact at the lower end, and therefore the timber specimen is only
constrained by its contact on the wings. Unfortunately, a gap is generally present
between the connector and the timber member to allow for an easier setup. This
gap enables some free rotation, and favors the splitting of wood, overall producing
a larger ductility factor but lowering the joint resistance.
On the other hand, increasing the strength of the welds by extending their
length has a positive effect on the strength, with a gain of over 50% as compared
to the initial configuration, while also improving the ductility of the joint, albeit in
a more moderate way as the previous configuration.
The mixed approach benefits from both alternatives but unfortunately also
shares their weakness. The resulting strength is about the same as the original
configuration, while the behavior is much more ductile. Yet wood splitting is
increased, and the concealing of the connector somewhat decreased. Moreover
cutting the timber member generates a potential concern regarding the vertical
Improving the Moment Resistance of a Concealed Timber Post Base Joint 29

In conclusion, we believe that reinforcing the welds is the best option among
the presented ones. The resulting joint exhibits a larger moment strength and a
much more ductile behavior than in the original configuration. As a result, we
believe that this connector is fitted for use in earthquake-resistant structures,
provided that the timber frame is associated with suited lateral-resistant structural
elements, and that its tensile resistance is found to be sufficient to sustain the uplift
forces usually produced by a horizontal loading. This last point is currently being
investigated experimentally.

Acknowledgment. The authors thanks M. Whi Lim Son for his help during the experi-
mental tests.

This work is a continuation of the study presented in 6. Therefore, some initial results
already published are reported here for the sake of consistency, along with some corrections
resulting from the analysis of new tests. The second part of this work presents new
configurations and is completely original.

[1] Hwang, K., et al.: Shear performance of post and beam construction by pre-cut pro-
cess. Journal of the Korean Wood Science and Technology (Mokchae
Konghak) 35(6), 112 (2007)
[2] Park, M.-J., et al.: Shear performance of hybrid post and beam wall system with
structural insulation panel infill. In: Proceedings of the 11th WCTE, paper 456 (2010)
[3] Humbert, J., et al.: Cyclic Behavior of Timber Column Concealed Base Joint. Journal of
the Korean Wood Science and Technology (Mokchae Konghak) 41(2), 123133 (2013)
[4] Hwang, K., et al.: Resistance performance of cross-shaped metallic joint for use in
multi-story timber structures. In: Proceedings of the Korean Society of Wood Science
and Technology (KSWST) Annual Meeting, paper D-12, Gwangju, Korea, April 4,
pp. 6465 (2011) (in Korean)
[5] International Organization for Standardization, ISO 16670: Timber Structures
Joints made with mechanical fasteners Quasi-static reversed-cyclic test method,
Geneva, Switzerland (2003)
[6] European Committe for Standardization (CEN), Eurocode 5: Design of timber struc-
tures Part 1-1: General Common rules and rules for buildings, EN 1995-1-
1:2004, Brussels, Belgium (2004)
[7] Architectural Institute of Japan, New estimation method for shear wall performance.
Standard for Structural Design of Timber Structure, 104 p. (2002) (in Japanese)
[8] Yamada, M., et al.: Seismic Performance Evaluation of Japanese Wooden Frames. In:
Proceedings of the 13th WCEE, paper 753, Vancouver, British Columbia, Canada,
August 1-6 (2004)
[9] Kretschmann, D.: Wood Handbook, Chapter 05: Mechanical Properties of Wood.
Technical Report FPL- GTR-190 (2010)
[10] Sawata, K., Yasumura, M.: Determination of embedding strength of wood for dowel-
type fasteners. J. Wood Sci. 48, 138146 (2002)
The Multifunctional TES-Faade Joint

Stefan Loebus, Stephan Ott, and Stefan Winter

TU Mnchen, Chair of Timber Structures and Building Construction, Germany


Short Summary. In comparison to a demolition of existing buildings with severe

technical deficits, usually the retrofitting of buildings is more effective in order to
prepare them for low energy consumption and new necessities as communication
and media connection or HVACinstallation (Heat, Ventilation and
AirConditioning). Prefabricated retrofit solutions are developed throughout
Europe to enable higher levels of industrialization in building envelope
modernization and hence additionally improvements in energy efficiency. Five
years of experience and a reasonable number of demonstrations done with
timberbased element system (TES) faades show tendencies for best-practice
building construction.
This paper focuses on the jointing between single faade elements and the
connection of those elements to the existing building. Being a crucial construction
detail within the TESfaade, the joint area shall meet various requirements and
challenges, from load bearing over hygro-thermal to fire safety functionality. The
results of in-depth construction detailing lay out the requirements and principles of
the TES joint.

Keywords: Energy efficiency, refurbishment, faade construction, timber

construction, prefabrication, fire safety, building envelope.

1 Introduction

Prefabrication a well-known concept in new timber construction still needs the

implementation of proper detailing considering the different circumstances and
assembly methods in refurbishment in order to apply it for retrofit projects. TES
Energy Faade and the follow-up smartTES, two international research consortia,
work on this idea since 2008 [1]. Since then the TES faade elements for energy
efficient retrofit of the building envelope with prefabricated timber based elements
are demonstrated in several projects. A first milestone was the definition of an
integrated work process starting from planning, digital survey, prefabrication and
assembly [2]. A key feature was the development of horizontal and vertical joints
of faade elements for certain assembly routines and structural properties. Further
findings are dealing with fire safety, ecologic properties, thermal and moisture
behaviour in specific climate conditions. This article shows the development of

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 31
DOI: 10.1007/978-94-007-7811-5_3, RILEM 2014
32 S. Loebus, S. Ott, and S. Winter

principles of detailing timber framework in refurbishment; analogy to existing

systematics in timber construction, the requirements, and generic solutions.

2 Retrofit with Prefabricated Faades

Modern, prefabricated timber frame construction is highly developed as a building

system for new construction. There is also no limit for small series or so-called
pilot prefabrication [3][2]. Because of the flexibility in the production planning
and manufacturing, custom timber framed components can be produced without
overhead [4]. This fact makes timber framed faade elements very interesting for
the refurbishment of existing buildings, because individual elements are required
for almost each building and project. Numerous pilot projects and demonstrators
in the context of research projects have shown that the transferability of the
principles of the timber framed building on a conceptual level is possible and
useful [5][6][7]. But at the detail level numerous new developments and
adjustments are necessary to meet the requirements of a reliable and economically
feasible design.
The accurate measurement of the faade is a prerequisite for precise
prefabrication and mounting Geometry data have to be combined with a thorough
survey of the existing faade construction, assessing materials, detailing, load
bearing capacity and if necessary building services und further in depth
information from the existing substance [2].

Fig. 1 Vertical section of Fig. 2 TES horizontal joint Vertical

typical semi-balloon frame
construction [9]

3 Timber Framed Construction in Refurbishment

Systematics of New-Built Timber Framework and Analogy

The three principles of timber frame construction are the platform frame
construction, balloon frame, and the semi or virtual balloon-frame [8].
The Multifunctional TES-Faade Joint 33

The virtual balloon-frame construction is the most suitable type as an analogy

for a TES EnergyFacade element. The elements have storey-high size and are
horizontally oriented; they can be produced in original upright position, put on the
truck and handled on site the same way without changing orientation. All elements
are stacked on top of each other and the dominant joint is the horizontal one. This
has the advantage of a continuous faade layer or building skin, without
interruptions and unbroken functional layers for air tightness, insulation or fire
safety. In analogy to the ceiling support edge beam a so called coupling beam is
introduced as a connection between the existing ceiling and the timber-framed
wall. The coupling beam is adjusted exactly in height and depth. It has the
additional advantage to assist horizontal alignment of TES elements. In contrast to
new building in virtual retrofitting balloon-frame, no vertical forces are transferred
to the coupling beam. The vertical loads are transferred only through the timber
framed wall. The horizontal loads have to be induced into the existing ceiling
plate. Hygrothermal exposure and wood protection of such a detailing solution can
be solved in analogy to new building. The fire safety must be examined in depth
because facades have a certain risk in the spread of fire and fire safety has always
observed in relation with the boundary conditions of the existing building.



Ecology Fire safety

Assembly Building services

Fig. 3 Parameters of a TES faade system

4 Definition of Requirements

The number of parameters affecting the TES element and its details is large. Most
influential are apart from structural issues, planning building physics of timber
construction, correct production of multilayered elements and assembly on
34 S. Loebus, S. Ott, and S. Winter

existing exterior walls. The main parameters about strengths and fixtures are
derived from the load-bearing structure of the elements. As a result, they are at the
same time in conjunction with the materiality, manufacture, structural
arrangement, aka. tectonics, and the assembly process. The other group of
influences originates from building physics. They affect TES elements on the
main inner and outer protection functions and therefore define the requirements
for layering and details. In addition to the production, they are in interaction to the
assembly, economy, ecology and technical building equipment and must be linked
to the static requirements. These interactions take place not only within the TES
system, but and above all, they are in interaction with the existing building. Some
protective functions of the existing exterior wall are partially replaced by the TES,
such as the thermal protection; they are fulfilled primarily by the TES envelope.

5 Limitations of the Analogy

The development of the retrofit construction system starts from the main structural
influences and the appropriate way of tectonics of parts. The focus is on the area
of the horizontal element joint. It is the connection of the existing exterior walls
with the existing ceiling. The horizontal joint is the most relevant detail to solve
from the technical and economic perspective, because it is the most frequent joint
of a virtual balloon-frame. A 1950ies multi-story house with 8 apartments on four
floors has a total length of basement and roof eaves joint line between circa 80 and
130 m. All other horizontal joints between the floors have a total length of
approximately 240 and 390 m. The complete vertical joint length is around 65 to
75 m. It matches a ratio between vertical and horizontal joints from 1:5 to 1:7. The
high deviation in this ratio clearly shows how important detailing of the horizontal
joint has to be done. It influences assembly effort and costs to a large extent. The
technical requirements and the effort in this area of the element have to be
adjusted very carefully to avoid high costs or unforeseeable risk. The developed
solutions have to be in line with the condition of efficiency. A robust solution of
but- jointing two elements will reduce the risk of the construction system for
failures. The detailed analysis of the construction principles on the applicability
and specialties of the construction process with large-format element in
refurbishment results in clear advantages for the balloon-frame principle. The
requirements of the continuity of the insulation layer and the uninterrupted surface
with defined axis of fixing, as alignment, and mounting aid for the faade
elements are established as significant characteristics.
A further invention is the adaption layer between the uneven surface of the
existing exterior wall and the plane back side of the TES element. The gap should
be at least fifty millimetres wide for further filling with blown-in insulation
material, preferably cellulose fibre. The process of blowing thermal insulation
material in the gap has the advantage of adapting the thickness, where it is needed.
Around penetrations of the exterior wall and the elements like windows, the gap
should be stuffed with non-combustible mineral wool.
The Multifunctional TES-Faade Joint 35

Fig. 4 Development of platform frame, (a) balloon-, and semi balloon-frame (b-c) to TES
(d) in analogy to (c)

6 Results in Timber Framework for Refurbishment

Development of the Horizontal Construction Joint
Principles of Timber Framing for Refurbishment Existing
As in classic timber frame construction the solid wood studs are positioned in a
span of 625 mm and butt-jointed perpendicular to sill-beam and wall-plate. On the
element inside an OSB is applied and on the outside a gypsum fiber board. The
cavities are filled with thermal insulation, e.g. cellulose fiber. Usually a timber-
frame element measures about 12,0 x 3,2 m, limited by production facilities and
transportation. [5]

Fig. 5 Mounting a TES-Element

36 S. Loebus, S. Ott, and S. Winter

While the TES-facade is built with a tolerance of +/-2 mm, the existing exterior
wall deviates +/-40 mm in depth. This unevenness is leveled by the coupling
beam, which is anchored into the existing wall and runs around the entire building
horizontally on each floor level. Thus, while mounting, its possible to align the
faade elements to the coupling beam and fix it with self-drilling timber screws.
The assembly of the elements happens horizontally row by row. The design of the
element joint was inspired by the simple joining method of a tongue-and-groove
connection and enhances the common butt-joint known from classic timber-frame
constructions. The tongue consists of the upper elements sill-beam. The groove
consists of wall-plate and the overlapping gypsum fiber board of the lower
element. That way the upper element is fitting precisely into the lower element,
without any larger alignment work necessary and still guaranteeing a force-fitting
connection. [5] The vertical connection is a butt-joint of two members without any
further connection.

Fig. 6 TES horizontal joint Vertical Fig. 7 TES horizontal joint Horizontal

All facade components such as the window generally are integrated in the TES
element, see Figure 7. This has the advantage of better manufacturing control and
simplified construction site operations. The new window can be installed in
different configurations as a replacement for the existing one. It can be integrated
only as a supplementary window in the TES, which enhances the existing window
and improves its thermal and acoustic properties. This solution is an important
option for upgrading existing components of good quality, speeds up the
construction process significantly and reduces disturbances of the residents.
The variety of applicable cladding materials is analog to new timber-frame
buildings; however, retrofitting a larger multi-storey building, the options are
strongly limited by fire safety requirements. In this paper the construction is given
with a rear ventilated wooden faade.

7 Structural Joint

Essentially there are two main loads on the TES-faade that have to be taken in to
account constructing the loadbearing system: Self-weight and wind-load.
The Multifunctional TES-Faade Joint 37

Fig. 8 Force conduction

The self-weight of the given construction is with about 80 kg/m (varying by

thickness, thermal insulation material and cladding) comparably high. This is due
to the extended timber loadbearing structure, the required fire safety boards, and a
dense cellulose fibre thermal insulation. The vertical force caused by self-weight
conducts all the way through the timber frame into a foundation system at the foot
of the faade. (For faade foot construction solutions see [1]) As the vertical force
accumulates over the building height, the horizontal joint between the elements is
built as full form fit contact connector. (Fig. 8 No. 1) The horizontal forces from
the wind-load are induced shortest way into the next ceiling. The connection is
built with anchor bolts and self-tapping screws. By placing coupling beam and
element joint on same height with the ceiling, eccentricities can be minimized and
heights of cross-sections optimized with the minimum edge distances of the bolts
and screws as one leading factor. Hence the sill-beam of the top element and the
wall-plate of the bottom element need to be built in one level. (Fig. 8 No. 2)

8 Hygrothermal Behaviour of TES Elements

The avoidance of thermal bridges is the first principle in detail development as no

high heat-conductive material from the hot to the cold side goes through, so that
conduction of heat is minimised. Thus, the impact of material-related thermal
bridges can be largely reduced by the use of wood or wood-based materials in the
joint area. A flexible variation of the insulation layer thickness can be achieved by
the adaptation of the depth of studs.
Regarding diffusion safety of the TES layer, the back panelling of the element
has to be vapour retardant. Diffusion open (to the outside) cross section of the
element with a low 0,3 sd 4m on the outside of the element and an sd at least
six times higher on the inside of the element is recommended. [10]
38 S. Loebus, S. Ott, and S. Winter

Air and wind tightness of timber framed elements are one of the most important
measures to stop convection of warm and moist air to the cold side of wooden
parts. A proper airtight construction also improves fire safety and sound protection
by overcoming leakages.
The horizontal and vertical TES joints have to be sealed for airtightness reasons
to ensure the moisture performance and avoid heat conductivity through wet wood
or insulation materials in the joint area.

Fig. 9 Hygrothermal detail

Wind tightness is provided by a wind barrier as an outer panelling or membrane

of the closed TES element. The wind barrier has to be sealed at connection by
tapes or clamped joints.
Besides construction material has to be dry, and elements are protected during
construction by closed cavities and heavy weathering to minimize moisture intake.
Diffusion open layering supports the drying capacity of elements. Cavities
between TES and existing walls have to be avoided to reduce the risk of
uncontrolled convection. An appropriate measure is the filling of cavities with soft
and dense insulation material during the assembly process. There is a good
experience with blown-in cellulose fibre that has a sorption quality that can buffer
the moisture from the existing exterior walls and reduces the moisture load on the
wooden parts of TES. The existing wall and the window connection is very often
permeable, therefore the coupling beam is horizontally sealed to inhibit vertical
The elements and element joints are made air-tight with sealing stripes
positioned between the elements and all paneling joints; thereby it is important
that the inner sealing stripe is connected air-tightly with the OSB by sealing tape.
(Fig. 9).
The Multifunctional TES-Faade Joint 39

9 Wood Protection and Durability

The built-in wood and wood-based materials are sufficiently protected through
compliance with the previously mentioned hygrothermal requirements and
moisture management in closed timber frame components. Thus the TES elements
fulfill the requirements of use class 0. [12]

10 Fire Safety

Normally, the existing structure already fulfills the fire safety objectives. However
the condition of the existing building has to be verified by a fire safety survey. [1]
The TES-faade layer has to fulfill the requirements of the relevant building
classes. In building classes 4 and 5, in which most of the potential TES-retrofitting
objects are classified, or buildings 3 storeys, non-loadbearing exterior walls
have to be at least fire-retardant (Fig. 10 No. 1) and difficult combustible cladding
and insulation material of at least class C- or B-s2,d0 is required. Timber or other
combustible faade materials are possible but need special construction and
additional approval. Where combustible material and ventilated gaps are used, the
cladding must have a storey-wise separation either by suitable fire stops or with a
floor wise horizontal separating construction. In this construction a metal sheet
serves as fire stop. It is disposed slightly above the joint area in order not to
obstruct the screwing of the sill-beam to the horizontal coupling beam. Cladding
and inner TES-Element is separated by a gypsum board to inhibit an ignition
mainly of combustible thermal insulation in case of a burning cladding. Closing

Fig. 10 Fire safety detail

40 S. Loebus, S. Ott, and S. Winter

the gypsum board layer at the element joint, requires a minimum of on-site work,
the connection is backed with an additional stripe of gypsum board and
horizontally with another below the wall-plate. Thus the joining of the elements
does not require any additional gluing of the boards. (Fig. 10 No. 2) [1][14][15]
The component requirements for the ceiling have to be taken in account especially
if the existing wall is removed or quite thin. The spread of fire and smoke into the
next storey in detour over the TES-faade has to be prevented. If there does not
exist an exterior wall, by definition, the TES-joint becomes part of the existing
ceiling. (Fig. 10 No. 3)

11 Mounting

An important difference to new building is the accessibility of elements rear side

while mounting, because one side is blocked by the existing wall. The higher the
prefabrication level the less work needs to be done on site, but covered connection
points will be harder to access this way. Especially if a visible cladding has been
prefabricated, necessary joining works must leave the visible side without
Pilot projects have shown that by beveling the plane between upper sill-beam
and lower wall-plate and additionally building a tongue-and-groove connection
with the front gypsum board, the mounting of the elements is simplified a lot. The
tongue-and-groove connection has another advantage of enabling a screwing of
both elements to the coupling beam in one horizontal level and thereby minimizes
the compulsory gap in the cladding.
Before the prefabricated TES-Elements are mounted to the building, its
necessary to set the foot foundation, building up surrounding scaffolding and
remove existing windows, if no elevating platform is used.

Fig. 11 Mounting process

The Multifunctional TES-Faade Joint 41

The mounting process is divided in four steps (Fig. 11):

(1) Fixing the coupling beam to the wall: For a fast mounting process and a
clean plane stop for the elements the coupling beam is positioned by using a laser-
assisted positioning system. The fixation points were taken during site measuring
in the survey phase. With the help of computer-aided design a leveled plane is
calculated onto the uneven surface of the existing building. The gap between
existing wall and coupling beam is filled with swelling mortar.
(2) Placing the lower element, butting the wall-plate against the coupling beam,
and fixing the connection with screws.
(3) Filling the adaption layer by blowing cellulose fiber through an opening in
the coupling beam.
(4) Insertion of the upper element into the lower element, screwing the sill-
beam to the coupling beam, and closing the wind-barrier.

12 Principles of Timber Framed Jointing in Retrofit

For the structural connection of the TES-Faade to an existing building, the

construction of the existing ceiling needs to be examined closely regarding
stability and positioning. The pull-out strength of anchoring into the floor slab
needs to be determined. In some cases there is a thin layer of masonry in front of
the ceiling edge, this has an influence on the anchorage length. The self-weight of
the faade needs be carried by the existing foundation. For the foundation system
of the TES-Faade various solutions are available [1] and should be chosen
according to the existing load-bearing capabilities.
The moisture management has to be dealt with like in new building. The claims
for hygrothermal protection are airtightness of timber framed elements back-
panelling and horizontal air stops in the vertical gap zone at each storey.
Convection and airborne sound might find also horizontal ways to neighbouring
units through leakages around window openings in exterior wall. Faults in
assembly may occur but can be improved from the inside afterwards.
By adding the TES-Faade as wall part not only the fire safety of the wall as
such has to be taken into account, but also the fire and smoke spread through gaps
between new TES-Faade and existing structure.

13 Conclusion

Generally spoken the interface of the existing building envelope and the new
timber-framed wrap around has to be planned very carefully. The building survey
has to examine all existent types of exterior walls and junctions of ceilings with
exterior walls. These are evident for structural jointing and critical routes for air,
moisture, sound and fire propagation. In the early planning process it is
prerequisite for risk mitigation. Timber framed elements for faade retrofit have
no high risks as shown in the proposed robust solutions.
42 S. Loebus, S. Ott, and S. Winter

14 Outlook

The timber-frame-system enables a self-bearing faade. Additionally loadbearing

components like balconies, spatial extensions, or roof-top extensions can be built
and integrated in the same system. Also HVAC-Systems can be included into
TES. Research to potential topology and detailing is ongoing.

Acknowledgements. TES EnergyFacade, run from 2008 to 2009, and still running
SmartTES, from 10/2011 to 09/2013 are funded in the European ERA-Net Woodwisdom-
Net by the German federal ministry for research BMBF, represented by PTJ.

[1] TES ENERGYFACADE: TES EnergyFaadeprefabricated timber based building
system for improving the energy efficiency of the building envelope, Period from
01/2008 to 12/2009, funded by BMBF, represented by PTJ,
[2] Larsen, K.E., Lattke, F., Ott, S., Winter, S.: Surveying and digital workflow in energy
performance retrofit projects using prefabricated elements. Automation in
Construction 20(8), 9991011 (2011), doi:10.1016/j.autcon.2011.04.001
[3] Davidson, C.H.: The challenge of organizational design for manufactured construction.
Construction Innovation. Information, Process, Management 9(1), 4257 (2009),
doi: 10.1108/147141709109315342009
[4] Jensen, P., Olofsson, T., Johnsson, H.: Configuration through the parameterization of
building components. Automation in Construction 23, 18 (2012),
[5] Ott, S., Loebus, S., Winter, S.: Vorgefertigte Holzfassadenelemente in der
energetischen Modernisierung. Bautechnik 90(1), 2633 (2013), doi:10.1002/
[6] Lattke, F., Ott, S., Winter, S.: TES EnergyFaade - Innovative vorgefertigte
Fassadenelemente aus Holz. Holzbau (dnq) 3 (2011)
[7] Lattke, F., Ott, S., Winter, S.: TES EnergyFacade Vorfertigung bei der
energetischen Sanierung. In: Bautechnik Innovative Fassadentechnik, vol. 88(9).
Ernst & Sohn Verlag, Berlin
[8] Kolb, J.: Holzbau mit System Tragkonstruktion und Schichtaufbau der Bauteile.
Birkhuser, Basel (2007)
[9] Hubweber, et al.: Holzrahmenbau. Holzbau Handbuch. In: Holzabsatzfonds (ed.)
Informationsdienst Holz. Reihe 1, T. 1, Folge 1, Bonn (2008) ISSN 0466-2114
[10] DIN 68800-2:2012-02 Wood Preservation. Part 2: Preventive constructional
measures in buildings. Beuth Verlag, Berlin (2012)
[11] DIN 68800-1:2011-10 Wood Preservation. Part 1: General. Beuth Verlag, Berlin
[12] EN 335:2006-10: Durability of wood and wood-based products Definition of use
classes. XXX Verlag, City (2006)
The Multifunctional TES-Faade Joint 43

[13] E2REBUILD: Industrialised energy efficient retrofitting of residential buildings in

cold climates, Period from 01/2011 to 06/2014, funded by the EU within the Seventh
Framework Program (FP7), http://www.e2rebuild.eu
[14] HTO Research Report HTO TP2 (High-Tech-Offensive Bavaria, Subproject 2),
Brandsicherheit im mehrgeschossigem Holzbau, TU Mnchen (2009)
[15] HTO Research Report HTO TP 11 (High-Tech-Offensive Bavaria, Subproject 11),
Mechanismen der Brandweiterleitung bei Gebuden in Holzbauweise, TU Mnchen
Green-Glued Products for Structural

Erik Serrano1,2, Jan Oscarsson2, Magdalena Sterley2, and Bertil Enquist1

Linnaeus University, Department of Building and Energy Technology, Vxj, Sweden
SP Technical Research Institute of Sweden, Bors

Abstract. The results from bending tests on 107 laminated, green-glued, beams
manufactured from Norway spruce side boards are presented. The beams were
made by face gluing 21-25 mm thick boards using a commercial one-component
moisture curing polyurethane adhesive. In addition to the bending test results,
results from shape stability measurements after climatic cycling and bond line
strength and durability test results are also presented. The results from the bending
tests show that, by applying very simple grading rules, it is possible to obtain
beams with high bending strength (with a 5%-percentile characteristic value of
40,1 MPa) and substantial stiffness (mean value of 14360 MPa). Also the shape
stability of the beams and the strength and the durability of the interlaminar bonds
were found to be satisfactory.

Keywords: green gluing, glulam, bond line shear strength, durability, adhesive

1 Introduction

For a typical south-Swedish saw mill, approximately 30% of the produced volume
consists of so-called side boards (here we focus on saw mills producing spruce
timber, Picea abies). Such side boards are of rather narrow dimensions, 18-25 mm
in thickness, and are not of major interest for structural applications. At the same
time, the huge number of narrow-dimension boards to be dealt with increases
considerably the costs for internal logistics at the saw mills.
The side boards are cut from the logs at a relatively large distance from the pith
and it is well known (see e.g. Steffen et al. 1997) that basic mechanical properties,
such as the modulus of elasticity along grain, are positively correlated to the dis-
tance from pith. Thus, the narrow-dimension side boards are of great interest if it
would be possible to use them in structural applications.
With the above in mind, an obvious approach would be to produce large-
dimension products from the side boards, similar to how traditional glued
laminated timber (glulam) production is done. However, in traditional glulam
production, the laminations have to be dried to approximately 12% moisture con-
tent (MC), and, in addition, laminations have to be planed within a short period of

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 45
DOI: 10.1007/978-94-007-7811-5_4, RILEM 2014
46 E. Serrano et al.

time prior to gluing. If kiln-dried thin side boards are used, such planing would
cause a substantial waste. If instead the laminations could be face-glued already in
the wet state, with a minimum of planing (or no planing at all) much would be
gained. This is the basic idea behind the research presented here and thus the ap-
proach has been to investigate the mechanical properties of green-glued laminated
products, produced from freshly sawn spruce side boards.
The research presented herein relates to three research projects running at Lin-
naeus University from 2006 to 2013. Only the main results in terms of mechanical
behaviour (stiffness and strength) from beam bending tests are reported in detail
here. In addition, investigations on the shape stability of beams exposed to varying
climates, on the durability and strength of adhesive bond lines, on the fracture
mechanical behaviour of bond lines and on the performance of green-glued finger
joints have been performed. Only some results from these investigations are
briefly mentioned here, but further details can be found in e.g. Serrano et al.
(2010), Serrano et al. (2011) and Sterley (2012).

2 Materials and Methods

2.1 Material Selection and Manufacturing Methods

2.1.1 Wood Raw Material, Grading and Adhesives

All wood material used is Norway spruce (Picea abies). The material was taken
from ordinary production at the saw mills, and was not graded before delivery to
the production facilities were the laminated beams were manufactured.
In two out of a total of three test series, the laminations were used as is, i.e. in
the saw falling quality delivered to the production facility, although with some
rejects due to vane. In the third series an effort was done to produce cross sections
with high quality outer laminations. In that test series it was decided to use two
grading criteria in order to classify the boards into inner laminations, outer lamina-
tions and rejects. First, a visual grading was performed, in which maximum knot
size was checked. For outer laminations, the maximum wide face knot size was set
to 25 mm. No criterion on maximum knot size was set for inner laminations. Fol-
lowing the visual grading, a hand-held grader (Timber Grader MTG, Brookhuis
Brookhuis Micro-Electronics) was used in order to determine the dynamic
modulus of elasticity (MOE) along the board direction of each lamination. For
outer laminations the MOE-criterion was 12 000 MPa, boards with MOE-values
between 7 000 and 12 000 MPa were used as inner laminations and boards with
MOE less than 7 000 MPa were rejected. Thus, laminations with MOE > 12 000
MPa, but with maximum knot size being > 25 mm were rejected.
In all tests, the laminations were delivered at an MC close to the fibre saturation
point, or considerably above that. The same adhesive was used in all test series a
commercial one-component, moisture curing polyurethane. The adhesive is not
specifically designed for green gluing, but is in fact an approved structural wood
Green-Glued Products for Structural Applications 47

adhesive for load bearing applications according to the European standardisation

and approval system when used in traditional dry gluing.

2.1.2 Beam Lay-Ups and Manufacturing

The original test plans included a total of 128 beams in three main series with
different variations. The variations included the annual ring orientation of the
laminations, grading of the laminations, and various climatic cycling schemes.
The various beam test series are denoted series I-III, with letters a-d denoting the
variations within each series.
Lamination lay-ups that were discussed in the planning phase are shown in
Fig. 1. Note that beam lay-ups B and C rely on the laminations being split to their
final width before gluing the laminations together to form the beam cross-section.
Beam lay-ups A and D can be manufactured by splitting the beam after gluing the
laminations together. Splitting the beam after gluing has the advantage of mini-
mising waste, since only two sides have to be planed and since the waste from the
saw cut in general is less than the waste from planing.
It was decided, based on the results from simple finite element analyses, to in-
vestigate orientations type A, B and D in the various test series. Lay-ups A and D
were chosen since these are the most practical in production and lay-up B was
chosen since the finite element analyses showed that orientation B would lead to
less distortion of the cross-section.
Two different approaches in manufacturing of the beams from the laminations
were thus used. In series I (comprising lay-up A) and in series II (comprising lay-
up D) the beams were manufactured with 120 mm wide laminations and the
beams were split into two halves of approximately 60 mm width after curing.
Thus, each such pair of beam halves comprises two beams with a good match of
the lamination properties.
In test series III annual ring orientations B and D were used, orientation D be-
ing used for reference to test series II. By comparing the distortion between orien-
tations B and D for pairs of beams with matched laminations the variability could
be kept at a minimum. The matching procedure was realised by first splitting
120 mm wide boards into two halves, one such half being used as a type D-
lamination in a type D-beam, the other half being used as a type B-lamination at
the same position in the corresponding type B-beam. Consequently, test series III
included pairs of beams, each pair being one type B-orientation beam and one type
D-orientation beam. Each lamination in a type B-beam was thus matched as close
as possible with the lamination at the corresponding position of the type D-beam.
Before gluing the laminations in series III, the laminations were also graded, as
described below. This meant that not all laminations could be used, since there
was a constraint that both halves should be placed at the same position in corre-
sponding beams. Thus, if one half was graded (based on either stiffness or knot
size) as being an outer lamination and the other as an inner lamination, both halves
had to be rejected. The somewhat complicated procedure used in series III was
used in order to minimise variability, and thereby simplify the evaluation of the
results. In production, clearly this approach cannot be used, for practical reasons.
48 E. Serrano et al.

Fig. 1 Lamination lay-ups used in the various test series

All beams were manufactured with full length laminations, thus no finger joints
were used. The beam gluing was done in semi-industrial scale beam presses. A
new pressing equipment was developed for use in series II and III due to insuffi-
cient performance of the equipment used in series I. The nominal clamping
pressure was around 0,5 MPa in series I and 0,9 1,0 MPa in series II and III. The
press time varied slightly between the various series, but was in general 45 - 60
minutes. The nominal spread rates used in applying the adhesive was 250 g/m2 in
series I and 200 g/m2 in series II and III.
Green-Glued Products for Structural Applications 49

After curing, and in relevant cases after splitting of the beam into two halves,
the beams were kiln dried together with structural timber of similar dimensions
(thickness). Finally, planing to target dimensions was done after kiln drying. All
test specimens manufactured for the three series are specified in Table 1. After
delivery to the laboratory, the beams to be tested without climatic cycling were in
all cases stored under controlled climate (20C/50%RH or 20C/65%RH) for at
least 3 months before testing. The beams to be exposed to climatic cycling were
stored for 3-7 months in the various climates.

Table 1 Beam series and parameter variation in manufacturing and climatic conditioning.
Laminations marked with * were split to half their width (giving two matched lamination
halves) prior to assembly of the beam.

Test Number Lay-up Laminations Grading of Beams Gluing Climates,

series of beams thicknesswidth laminations Widthdepth condi- RH (%).
length (mm3) (mm2) tions
Ia 15 A 211204900 No 50300 Dry 501
Ib 12 A 211204900 No 50300 Green 501
Ic 13 A 211204900 No 50300 Green 50-90-501
IIa 16 A 251205400 No 50300 Green 652
IIb 16 A 251205400 No 50300 Green 65-35-652
IIc 16 D 251205400 No 50300 Green 652
IId 16 D 251205400 No 50300 Green 65-35-652
IIIa 5 B 21,51205400* Yes 50300 Green 652
IIIb 7 B 21,51205400* Yes 50300 Green 65-35-90-652
IIIc 5 D 21,51205400* Yes 50300 Green 652
IIId 7 D 21,51205400* Yes 50300 Green 65- 35-90-652
Beams were to be tested after conditioning at climate 20C/50%RH.
Beams were to be tested after conditioning at climate 20C/65%RH.

2.2 Beam Bending Tests

The bending stiffness and the bending strength of the beams were evaluated. The
tests were all performed in 4-point bending, as is prescribed in EN 408. Due to the
limited length of the beams in series 1, the standard 18 times beam depth for the
total span was not possible to use. Instead, approximately 15,3 times the beam
depth was used in series I and 16,7 times the beam depth was used in series II and
III. The minimum requirement according to the standard is 16 times the beam
depth. The distance between the loads was, however, always 6 times the nominal
beam depth. The test set-up used is depicted schematically in Fig. 2 for the case of
16 times the beam depth.
Two different types of deflection measurements were used, denoted v and w in
the figure. The local deformation, v, includes the deformation due to bending only,
since there is no shear force in the mid span. The deformation denoted w includes
50 E. Serrano et al.

deformation due to shear but also due to local effect at supports. Therefore, the
most relevant stiffness to report is the one calculated using the deformation v. This
stiffness is also known as local modulus of elasticity (local MOE). All tests were
performed in displacement control with the loading speed set such that failure was
reached within 5-10 minutes.
Prior to testing to failure, the stiffness was determined by calculating the de-
formation between two load levels within the elastic domain. All test results in
terms of strength have been evaluated in relation to EN 14358, and EN 14080,
assuming a lognormal distribution of bending strength with unknown standard
deviation. The characteristic values are determined as 5-percentiles at 75% confi-
dence level. The size correction factor according to EN 1194 has been used, and
such corrected strength values are thus representative for a 600 mm deep and 150
mm wide beam. Since the beams in fact were approximately 300 mm in depth, the
actual strength of the beams in testing was approximately 13% higher than the
reported values.

Fig. 2 Test set-up used in beam bending tests

2.3 Shape Stability, Bond Line Shear Strength and Durability

The shape of the beams was measured at delivery to the University and after com-
pleting the different climatic cycles. In those measurements a special rig making it
possible to measure twist, cup, bow, crook and cross-sectional distortion (meaning
deviation from 90 angle at the corners). Here, twist, bow and crook were meas-
ured over a length of 3,72 m while cup was measured over a 270 mm length in the
height direction of the cross section.
For assessment of the adhesive bonds in the beams, shear strength and wood
failure percentage (WFP) were determined according to the European standards
EN 392 and EN 386, respectively. The delamination of the bonds according to
EN 391 was also measured. In a first test series, specimens for shear and for de-
lamination tests were taken from four beams from series II (two with a annual ring
orientation of type A and two of type D) that had been tested in bending (without
climatic cycling). In total 48+48 bonds from the type A beams and 24+24 bonds
from the type D beams were tested for shear strength and delamination, respec-
tively. A second test series was performed by taking additional specimens from
Green-Glued Products for Structural Applications 51

two type A that were manufactured especially for the shear and delamination tests
and thus had not been tested in bending or been exposed to climatic cycling. These
beams had two different spread rates: 150 and 200 g/m2. Here in total 24+24
bonds for each spread rate were tested for shear strength and delamination, respec-
tively. A third test series was realized by cutting specimens from beams that had
been tested in bending after being exposed to climatic cycling. These specimens
were used only for delamination tests.

3 Results

3.1 Beam Bending Tests

The results from all tests performed are summarised in Table 2 and Table 3. The
characteristic strength values of Table 3 correspond to the 5%-percentile estimates
calculated according to the provisions of EN 14358. Size-effect corrected values
are calculated according to EN 1194. Bending MOE values are based on local
MOE, cf. Fig. 2. Note that in manufacturing the test specimens according to the
schedule in Table 1, there was at one occasion a malfunction of the pressing
equipment, and thus three of the planned tests of Series III could not be per-
formed. In addition, cracks running from the end of the beams had developed
during kiln drying in three specimens, such that at testing of these beams failure

Table 2 Bending strength, fm, MOE, density and MC. kh is the size factor according to EN
1194. Density and MC (only measured in one series) refer to beam mass/beam volume at
the time of testing, after conditioning in 20C/50%RH or 20C/65%RH, cf. Table 1.

Series N fm fm kh MOE Density MC

Mean COV Mean Mean COV Mean Mean

[MPa] [%] [MPa] [MPa] [%] [kg/m3] [%]
Ia 15 43,9 9,1 38,8 14210 4.2 495 -
Ib 12 44,8 9,3 39,6 13860 6.2 498 -
Ic 12 42,6 12,6 37,6 13750 5.6 504 -
IIa 16 45,5 10,0 40,2 13680 4.7 495 ~14-16
IIb 16 45,5 14,8 40,1 13800 6.6 492 ~13-14
IIc 16 46,8 12,1 41,3 14030 5.5 507 ~14-16
IId 16 50,4 7,1 44,4 14510 6.8 504 ~13-14
IIIa 2 48,7 - 43,0 13880 - 500 -
IIIb 7 52,5 5,7 46,3 14690 3.8 495 -
IIIc 3 51,8 - 45,7 13830 - 493 -
IIId 7 49,8 5,1 44,0 14380 4.6 493 -
52 E. Serrano et al.

was due to shear along that same pre-existing crack. Such specimens have been
excluded from the evaluation. Thus, out of the planned 128 beams, only 122 are
Table 3 gives the results statistics for various combinations of beam lay-ups (cf.
Table 1). Most of the variations studied seem to have a relatively small influence
on average bending strength, except combination D (graded laminations) which
gives a higher strength, and also a lower COV, resulting in a considerably higher
characteristic bending strength. The MOE variation was in all cases small, both
within the test series and between test series.

Table 3. Test results, combinations of test series. kh is the size factor according to EN 1194

Series N fm fm kh MOE
Mean COV Char Mean Char Mean COV
[MPa] [%] [MPa] [MPa] [MPa] [MPa] [%]
A: Series Ib+Ic+II+III 107 47,0 11,2 36,9 41,5 32,6 14020 5.9
B: Series II+III 83 48,0 10,7 37,9 42,3 33,5 14080 5.9
C: Series II 64 47,1 11,5 36,5 41,5 32,2 14000 6.3
D: Series III 19 51,0 5,2 45,4 45,0 40,1 14360 4.3

3.2 Shape Stability, Bond Line Durability and Strength

An example of results from the shape stability measurements are given in Fig. 3,
showing the twist of the beams (measured over a length of 3,72 m) of Series III
after the various climatic conditioning schemes.

Fig. 3 Measured twist after various climatic conditioning schemes for series III
Green-Glued Products for Structural Applications 53

Results from the bond line durability and shear strength tests are presented in
Table 4. The results fulfil on average the strength versus wood failure percentage
requirements of EN 386. Individual specimens do however, for one of the test
series, not fulfil the requirements: Using only 150 g/m2 adhesive spread rate, two
individual tests (out of 24 tests) failed to comply with the strength requirements of
EN 386 (for structures in service class 3). For the delamination tests on bonds that
had not been subjected to climatic variations a 3,0-5,2 % delamination was
obtained. On average this meets the requirements and the spread rate 200 g/m2
appears to be enough to obtain both sufficient mechanical strength and to obtain
adequate resistance to delamination (note that some of the delamination tests were
done on specimens taken from tested beams). The test series showing highest
delamination values (see Table 4) are tests performed on specimens subjected both
to bending tests and climatic cycling.

Table 4 Results from tests of adhesive bonds. Standard deviations in parenthesis.

Lay-up and specimen type Mean shear Mean wood Delamination

strength failure (%) (%)
A. (200 g/m2, after bending test, no climatic cycling) 9,3 (1,4) 94 (12) 3,4
D. (200 g/m2, after bending test, no climatic cycling) 10,3 (1,4) 91 (13) 5,1
A(150 g/m , no bending test, no climatic cycling) 9,5 (1.9) 86 (13) 5,2
A (200 g/m2 no bending test, no climatic cycling) 9,9 (1.2) 93 (10) 3,3
A (200 g/m2 after bending test, climatic cycling) - - 19 (9)
Min 1 Max 34
D (200 g/m2 after bending test, climatic cycling) - - 17 (7)
Min 3 Max 26
Requirements for average according to EN386 6 Min. 90
8 Min. 72 Max 5
11 Min. 45

4 Discussion and Concluding Remarks

With the approach presented here it was possible to obtain laminated beams with
strength corresponding to GL36h (according to EN 1194) and almost matching the
MOE requirement (14360/14700=97,7%) of that same glulam grade. Using very
simple visual grading rules and MOE-measurements based on axial excitation, it
was possible to obtain a characteristic beam bending strength of 40,1 MPa.
In terms of shape stability (distortion of cross section and shape changes due to
moisture variations) it was found that the influence of lamination lay-up is only of
minor importance. In general, distortions (cup, twist, bow and crook) were very
small. The interlaminar bond line strength, achieved with reasonable amounts of
54 E. Serrano et al.

adhesive spread rate (200 g/m2) and moderate clamping pressure (0,5-1,0 MPa)
fulfil the requirements of the EN standards for glued laminated timber to be used
in service class 2 (on average the strength requirements for service class 3 are also
fulfilled). For some of the delamination tests series performed the requirements of
the EN-standards were also fulfilled, but for other test series this was not the case.
Since the specimens used in some of the delamination tests had been subjected to
climatic changes already before the delamination test it is difficult to assess
whether they would have passed the requirements if the tests had been performed
on specimens cut from virgin beams which is assumed to be the case in the
standard. However, for the only series performed with a normal adhesive spread
rate and virgin material specimens, the requirements were fulfilled.
On-going research (see e.g. Sterley 2012) has shown that finger-joints made in
the green state, show a good potential in terms of creating close-contact and
homogeneous bonds with high strength. It is believed that including such finger-
joints in combination with defect elimination in the laminations even better beam
performance can be expected.
In order for the timber industry to be able to take full advantage of the green
gluing technology in timber engineering applications, there is a need for standardi-
sation and/or approval procedures. Todays EN standards do not include the pos-
sibility of producing e.g. CE-marked green glued laminated beams. A CUAP
procedure for the type of product presented here is under development, with the
aim of having this approved during 2013.

Acknowledgments. This research was possible thanks to the financial support from The
Knowledge foundation, Linnaeus University and CBBT (Centre for building and living
with wood). The authors gratefully acknowledge this support.

EN 386:2003. Glued laminated timber Performance requirements and minimum produc-
tion requirements
EN 391:2003. Glued laminated timber Delamination test of glue lines
EN 392:1995. Glued laminated timber Shear test of glue lines
EN 1194:1999. Glued laminated timber Strength classes and determination of characteris-
tic values
EN 14080:2005. Timber structures Glued laminated timber Requirements
EN 14358:2006. Timber structures Calculation of characteristic 5-percentile values and
acceptance criteria for a sample
Anon: THE SWEDISH FOREST INDUSTRIES Facts and figures 2011. Swedish Forest
Industry Federation (2011), http://www.forestindustries.se/ (accessed
May 14, 2013)
Serrano, E., Blixt, J., Enquist, B., Kllsner, B., Oscarsson, J., Petersson, H., Sterley, M.:
Wet glued laminated beams using side boards of Norway spruce, Linnaeus University,
School of Engineering, Report No 5 (2011) ISBN: 978-91-86491-79-6
Green-Glued Products for Structural Applications 55

Serrano, E., Oscarsson, J., Enquist, B., Sterley, M., Petersson, H., Kllsner, B.: Green-glued
laminated beams High performance and added value. In: Proc. of World Conference
on Timber Engineering, WCTE 2010, Riva del Garda, Italy (2010)
Steffen, A., Johansson, C.-J., Wormuth, E.-W.: Study of the relationship between flatwise
and edgewise moduli of elasticity of sawn timber as a means to improve mechanical
strength grading technology. Holz als Roh und Werkstoff 55, 245253 (1997)
Sterley, M.: Characterisation of green-glued wood adhesive bonds. Dissertation, Linnaeus
University press (2012) ISBN: 978-91-86983-57-4
Experimental Analysis of a Post-tensioned
Timber Connection

Flavio Wanninger and Andrea Frangi

Chair of Structural Engineering, ETH Zurich, Zurich, Switzerland


Abstract. The moment-rotation-behaviour of a post-tensioned beam-column timber

joint has been analysed extensively with a series of static bending tests. The timber
joint was loaded at the end of the beams in order to apply a moment to the connec-
tion. The tests were conducted with various forces in the tendon, from 300 kN up to
700 kN. The bending tests were performed with a controlled load level, so that no
failure perpendicular to the grain in the column occurred.
The maximally allowable vertical load to be applied was estimated using a simple
spring model. A final bending test was conducted in order to study the failure mode
of the post-tensioned timber beam-column joint. The vertical load on the beams was
increased until the tendon-elongation got so high that the test had to be stopped for
safety reasons.
This paper presents the main results of bending tests on the post-tensioned timber
joint. Attention will be given to the structural behaviour and to the influence of the
applied post-tensioning force on the connection stiffness. The experimental results
will be compared to the results obtained using a simplified analytical calculation

Keywords: beam column connection, post-tensioning, LVL, rocking behaviour.

1 Introduction
In the past decades precast concrete frames were developed using tendons to
connect the columns and the beams [1, 2]. These systems showed favourable seis-
mic behaviour, being able to avoid residual deformations after an earthquake (self-
centering systems). Furthermore a model, the monolithic beam analogy, was
developed to describe the connection behaviour [3].
A similar system for timber was introduced in New Zealand at the University
of Canterbury [4, 5]. A timber frame made of laminated veneer lumber was post-
tensioned, resulting in a good structural behaviour. Design proposals were published
[6, 7] and first buildings were constructed [8].
Post-tensioned timber joints are also being studied at the Institute of Structural
Engineering at the ETH in Zurich. A post-tensioned beam-column timber joint has

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 57
DOI: 10.1007/978-94-007-7811-5_5,  c RILEM 2014
58 F. Wanninger and A. Frangi

been developed using glued laminated timber with local strengthening of the joint
using hardwood (fraxinus). No steel elements are required for the moment-resisting
timber joint exept from a single straight tendon, which is placed in the middle of the
beam. The developed post-tensioned beam-column timber joint is characterised by
a high degree of pre-fabrication and easy assemblage on site.
Within the framework of a doctoral thesis the post-tensioned timber frame will
be further developed to also withstand significant horizontal loading. A simplified
analytical spring based calculation model is being developed in order to predict the
structural behaviour of a post-tensioned timber connection.

2 Specimen and Test Setup

The test specimen consists of two beams and a column made of glulam. The glulam
beams are made of spruce GL24h [9] except for the three bottom lamellae, which
are made of ash D40 [9]. The column is a hybrid element as well, made of spruce
and ash. The hardwood is used in the connection area between the column and the
beam (darker areas in figure 1(a)), where high stresses perpendicular to the grain
A straight tendon is inserted in the middle of the beams and attached at each
side of the beams. A 50 mm thick steel plate is used at the end of the beams for the
load transmission from the tendon into the beams. The moment-resistant connection
does not require further steel elements. The shear force between beam and column
is transferred via friction and through the small supports under the beams, which
were produced by cutting a notch into the column.
All the tests were performed at the ETH Zurich on a strong floor. A rigid steel
frame was built for the tests (figure 1(b)). The frame consists of two columns, which
are fastened to the strong floor with high-strength pre-stressed bolts. The columns
are further connected to a strong wall with two horizontal beams. The test specimen
is attached to the columns with stiff steel profiles. The load F on the beams is applied
by two cylinders, which are connected to the beams at a distance of 1.24 m from

1.62 0.36 1.62


(a) (b)

Fig. 1 1(a): Test specimen (dimensions in m), 1(b): Test set up with steel frame
Experimental Analysis of a Post-tensioned Timber Connection 59

the interface column/beams (see figure 1(a)). The cylinders allow applying a load
of max. 300 kN on each side of the specimen. The cylinders are connected to a
hydraulic pump, but can be controlled separately, so that several load cases can be
investigated. It is for example possible to apply the load only on one beam, while
the other one is unloaded. One beam can also be loaded to a certain constant force
while the second beam is loaded to a different force. The weight of one beam with
two cylinders is 480 kg, which has to be accounted for in the analysis.

3 Experimental Analysis
3.1 Structural Behaviour of the Connection
While prestressing the tendon, the beams are being pressed against the column.
This leads to an initial compression at the interface, which has to be measured. A
mean initial value for the compression was estimated using LVDTs, as shown in
figure 2(a), where the measurements show a uniform horizontal displacement of
approximately -0.4 mm at the load level F=0 kN. This deformation is the initial
compression after stressing the tendon (and before the beams are loaded).
If the vertical load is applied on the beams, the interface starts rotating which
leads to decompression and to an opening of a gap. Figure 2(b) qualitatively shows
the rotation of the beams and the recorded horizontal displacements from the LVDTs
during a test for different loads F (figure 2(a)).
From the horizontal displacement measured at the interface, a rotation can be
calculated (slope of the lines in figure 2(a)). All results herein will be presented as a
function of the rotation in the connection. Since this parameter is crucial, inclinome-
ters where placed on top of the beams, which measured the inclination and therefore
the rotation. The values were used to verify the rotations measured with the LVDTs.



Position [mm]


0 kN
200 20kN
100 60 kN
80 kN

1 0.5 0 0.5 1 1.5 2 2.5 3
Displacement [mm]
(a) (b)

Fig. 2 2(a): Horizontal displacement of the beam-column interface. Position = Height of the
beam measured from the bottom. Load levels = load F on the beam, 2(b): Definition of the
60 F. Wanninger and A. Frangi

From the measured displacements at the interface it is also possible to calculate

the height of the compressive zone, the second key variable to describe the structural
behaviour of the connection.
The third value, calculated using the acquired measurements, is the maximal
compressive stress at the beam-column-interface. It is assumed that the stress distri-
bution is elastic, no plastic behaviour has been taken into account during the exper-
imental analysis.
The fourth value needed to describe the structural behaviour of the connection is
the tendon force, which is measured during the tests with a load cell.

3.2 Test without Tendon Elongation

Figure 3 shows the main results for a test conducted with a post-tensioning force of
550 kN. The structural behaviour of the post-tensioned beam-column timber joint is
characterised by 4 different stages:
Before decompression: The beam-column timber joint is subjected to compres-
sion over its total beam height (see Figure 3, x- -diagram), i.e. the compressive
zone x = 600 mm. The stiffness of the beam-column timber joint corresponds
to the slope of the moment-rotation-curve (M- ), which has the highest value
in stage 1. The two connections (left and right) behave identically.
After decompression: The gap starts opening between the column and the beam
and thus the compressive zone gets smaller. The reduction of the compressive
zone leads to a decrease of the stiffness of the joint. The moment of decom-
pression, when the gap opening occurs, is estimated to be 55 kNm. It is also
noticeable that the two connections (left and right interface) behave differently;
the right connection is stiffer than the left one. The beam-column connections
interact with each other, i.e. when the left beam is being pulled down, the right
beam wants to move upwards and vice versa. It may be possible, that the left
beam was pulled down slightly before the right one, making the right connec-
tion stiffer. This effect was very distinctive for small tendon forces and became
smaller as the tendon force increased.
Tendon elongation: When the gap opening reaches the position of the tendon in
the middle of the beam, the tendon starts to be elongated. However, in this test
this phenomenon is hardly visible; only at the end of the test, at a rotation of
5 mrad, a small increase in the tendon force can be noticed. This is confirmed by
the height of the compressive zone at the left interface, which became smaller
than 300 mm, i.e. the gap opening has reached the position of the tendon. This
effect can be seen more clearly in the test described in the next section.
Failure: Failure can not be observed during this test. There are no residual de-
formations as can also be seen in figure 3, where the moment-rotation curve
always goes back to its origin after unloading the specimen.
During the test in figure 3 the specimen remains purely elastic. The non-linear
behaviour is due to the gap opening and not due to plastic behaviour of timber. No
residual deformations could be measured at the end of the test.
Experimental Analysis of a Post-tensioned Timber Connection 61

120 700
100 600
M [kNm]

P [kN]
40 left left
right right
20 model 100 model
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
[] 3 [] 3
x 10 x 10

600 10

500 8

inf [N/mm2]
x [mm]

200 left left
right 2 right
100 model model
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
[] 3 [] 3
x 10 x 10

Fig. 3 Results of the test without tendon elongation. Key: : rotation, M: moment, P: tendon
force, x: height compressive zone, in f : maximal compressive stress at the interface.

3.3 Test with Tendon Elongation

Figure 4 shows the main results of a final test conducted with a post-tensioning
force of 550 kN. The vertical load on the beams was increased as much as possible
in order to gain information about the failure behaviour of the connection.
The connection shows a hysteretic behaviour in the moment-rotation-curve (dif-
ferent loading and unloading paths, see figure 4). Since only minor plastic deforma-
tions occurred during the test (the residual deformation were negligible) nearly no
energy was dissipated from the system. The hysteretic behaviour is mainly due to
loss of the post-tensioning force during the test (leakage in the hydraulic system).
The tendon force was 550 kN at the beginning of the test, and 510 kN at the end.
The tendon force (P) starts to increase at a rotation of 6 mrad. The gap open-
ing reaches the position of the tendon, which is therefore elongated as the load is
The compressive zone is constant at a value of 600 mm up to a rotation of
1.2 mrad, at which point it starts to decrease, indicating the moment of decom-
pression. The height of the compressive zone decreases to 150 mm. This means that
the gap has a total height of 450 mm and is 150 mm under the position of the tendon,
leading to a remarkable elongation of the tendon, which results in a tendon force of
750 kN.
62 F. Wanninger and A. Frangi

200 800

150 600
M [kNm]

P [kN]
100 400

50 test 200 test

model model
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
[] []

600 25
500 model 20
400 inf [N/mm2] 15
x [mm]

5 test
100 model
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
[] []

Fig. 4 Results of the test with tendon elongation. Key: : rotation, M: moment, P: tendon
force, x: height compressive zone, in f : maximal compressive stress at the interface

The maximal compressive stress, calculated based on elastic behaviour, reaches

values over 20 MPa and is therefore not in the elastic range of the material any more.
Some measuring devices were used beyond their measuring range at large ro-
tations. This can be seen for the compressive zone and the maximal compressive
stress, whose values are incorrect after a rotation of 20 mrad. The tendon force and
the rotation were measured correctly at large rotations, since the force was measured
directly with a load cell and the rotation was estimated with the inclinometers.

3.4 Discussion of the Test Results

The rocking behaviour of the post-tensioned timber connection was studied with a
series of static bending tests. The joint is characterised by an initial (higher) stiff-
ness up to the point where a gap opening occurs (decompression point). After the
decompression point, the connection shows a non-linear structural behaviour due to
a decreasing contact zone between the beams and the column.
If the load is increased, the gap opening reaches the position of the tendon and the
tendon starts to be elongated while the tendon force increases. At some point, the
column will start to yield due to the ductile embedment failure of timber subjected
to compression perpendicular to the grain.
Experimental Analysis of a Post-tensioned Timber Connection 63

Thus, one important mechanical property governing the design is the timber em-
bedment strength perpendicular to the grain. However, during the tests only minor
plastic deformation of the timber column subjected to compression perpendicular
to the grain occurred. The residual deformations were small and the damage at the
beam column interface negligible. The test had to be stopped due to the increase
in tendon force. However, it has to be considered, that the tendon was very short
leading to a fast increase in the tendon force.
The applied tendon force governs the structural behaviour of the beam-column
connection. The tests showed that the initial stiffness (stiffness up to the moment of
decompression) increases as the tendon force increases (figure 5 left). However, the
relationship between initial stiffness and tendon force is non-linear and the tendon
force cannot be increased infinitely; an upper limit exists due to the limited strength
of the column.

x 10
5 250

4 200
KI [kNm/rad]

M [kNm]

3 150

2 100

1 tests 50 tendon elongation

curve fitting no elongation
0 0
0 200 400 600 800 0 0.01 0.02 0.03 0.04
P [kN] []

Fig. 5 Initial stiffness as a function of the tendon force (left), moment-rotation behaviour
with and without considering tendon elongation (from analytical model)

After decompression, when the gap has reached the position of the tendon, the
latter begins to elongate which leads to an increase in stiffness as opposed to the
case without tendon elongation (figure 5, right). Based on this observation it would
be more favourable to change the position of the tendon when designing frames with
post-tensioned timber connections for gravity loads, i.e. the tendon should be placed
in the upper part of the beam, so that the elongation of the tendon already begins at
a lower load level.

4 Analytical Model
4.1 Without Tendon Elongation
In order to describe the structural behaviour of the beam-column interface, an an-
alytical model is being developed. The elastic model simulates the column with
64 F. Wanninger and A. Frangi

springs, which are only able to bear compressive forces. This allows a gap to open
as soon as decompression is reached. The beam is modelled as a rigid body, since it
is much stiffer than the column which is loaded perpendicular to the grain.

Fig. 6 Model before decompression (left) and after decompression (right)

The equilibrium of forces and bending moments leads to equations for the mo-
ment, the rotation and the compressive zone. The resulting compressive stresses are
calculated based on a linear elastic theory (see table 1).

Table 1 Equations for calculating the compressive zone, rotation, moment and stresses be-
fore decompression (left column) and after decompression (right column)

before decompression after decompression

h M
x=h (1) x = 3 (2)
2 P0

1 2 1 1 1
= (3) = (4)
c x c x

b h3 h 2 P0
M = c = c IB (5) M = P0 (6)
12 2 9 bc

2 P0
P0 M P M 1 =   (8)
1,2 = = 0 + bh2 (7)
A W bh 3 b h2 PM0

Key: b: width of the beam, h: height of the beam, P0 : initial tendon force, M: bending moment,
: rotation, : stresses, c: spring constant, x: height of compressive zone

The only parameter needed for the model is a spring constant c. It is suggested
to calculate the spring constant by placing a uniform load on the interface. The
spring constant is then defined as the ratio between the compressive stress and the
Experimental Analysis of a Post-tensioned Timber Connection 65

displacement due to the load, which only depends on the geometry and the modulus
of elasticity of timber perpendicular to the grain.

4.2 With Tendon Elongation

The tendon is positioned in the middle of the beam. If the gap opens and reaches the
tendon, the latter is elongated and its force increased. Both values can be calculated
by using geometric expressions and Hookes law (equation 9 and 10).

L p = (d x) (9)

(d x)
P = P0 + A p E p (10)
Key: L p : tendon elongation, L p : length tendon, A p : cross section area tendon, E p : modulus
of elasticity tendon, P0 : initial tendon force, P: tendon force, d: position of the tendon, x:
height compressive zone, : rotation

Since the tendon force has become a function of the rotation and the height of
the compressive zone, an iterative procedure is necessary to obtain a solution for the
case with tendon elongation.

5 Comparison Tests Model and Conclusions

In figures 3 and 4 the prediction made by the spring model is plotted amongst the test
data. The spring stiffness was calculated with a modulus of elasticity perpendicular
to the grain of 860 MPa, as recommended by the EN 338 [9]. This value leads to
accurate estimations for the initial decompression of the beams.
The model fits the test data for the test without tendon elongation well. The initial
stiffness up to the moment of decompression is slightly underestimated. However,
the moment of decompression is predicted well as is the height of the compressive
zone and therefore the moment when tendon elongation starts. The stresses also
seem to be modelled reasonably well.
For the case where tendon elongation occurs, the differences between the model
and the test become noticeable. The model takes the elongation of the tendon into
account. Therefore an increase in the tendon force is visible in the test as well as
in the model. However, the increase according to the model is higher than during
the test. This leads to a model which describes a stiffer structural behaviour of the
beam-column connection than the actual behaviour of the connection, as can be seen
in the moment-rotation diagram.
One possible reason for the difference between model and tests is that the model
does not take the plastic deformations of timber subjected to compression perpen-
dicular to the grain into account. However, embedment failure was observed during
the test (a residual deformation of 2 mm was measured by the column at the in-
terface under the beams, which corresponds to an additional rotation of 1.6 mrad).
66 F. Wanninger and A. Frangi

A second possible reason is that the tendon may have some space to move, since the
hole in the specimen is larger than the tendon itself.
The results of the extensive experimental analysis show the favourable structural
behaviour of the developed post-tensioned beam-column timber joint using glued
laminated timber with local strengthening of the joint with hardwood. Further ex-
perimental and numerical analysis are planned in order to study and optimise the
structural behaviour of the beam-column connection and to establish economic and
reliable design rules. A first building will be built at ETH in autumn 2013 using the
developed post-tensioned timber frame construction and will be monitored during
construction and later use.

Acknowledgements. The authors gratefully acknowledge the financial support by the

Swiss Commission for Technology and Innovation (CTI) and the industrial support of
Haring & Co.AG.

1. Priestley, N.M.J.: Overview of PRESSS Research Program (1991)
2. Pampanin, S., Priestley, N.M.J., Sritharan, S.: Analytical Modelling of the Seismic Be-
haviour of Precast Concrete Frames Designed with Ductile Connections. Journal of Earth-
quake Engineeringl (2001)
3. Palermo, A., Calvi, G.M., Castellani, A.: The Use of Controlled Rocking in the Seismic
Design of Bridges. PhD thesis, Politecnico di Milanol (2004)
4. Palermo, A., Pampanin, S., Buchanan, A.: Experimental Investigations on LVL Seismic
Resistant Wall and Frame Subassemblies (2006)
5. Buchanan, A., Deam, B., Fragiacomo, M., Pampanin, S., Palermo, A.: Multi-Storey Pre-
stressed Timber Buildings in New Zealand. Structural Engineering International (2008)
6. Newcombe, M.P.: Seismic Design of Multistorey Post-Tensioned Timber Buidlings
7. Newcombe, M.P.: Multistorey Timber Buildings Seismic Design Guide (2010)
8. Buchanan, A., Palermo, A., Carradine, D., Pampanin, S.: Post-tensioned timber frame
buildings. The Structural Engineer (2011)
9. CEN. DIN EN 338 - Structural Timber - Strength classes; German version. European
Committee for Standardization (2009)
Risk Based Investigations of Partly Failed or
Damaged Timber Constructions

Gerhard Fink1 and Jochen Kohler2

1 Institute of Structural Engineering, ETH Zurich, Wolfgang-Pauli-Strasse 15,
8093 Zurich, Switzerland
2 Department of Structural Engineering, NTNU Trondheim, Rich. Birkelandsvei 1A,
7491 Trondheim, Norway

Abstract. In this paper a framework for updating the load and resistance proper-
ties of a partly failed timber construction based on different kind of information is
presented. That includes information available, such as the resistance of the failed
member(s) or the results of destructive and non-destructive measurements. In ac-
cordance to the type of information different updating methods are presented. The
application of the framework is exemplary illustrated on the case study.

Keywords: failures of structures, risk analysis, duration of load, damage,


1 Introduction
In the last decade a significant amount of failures in timber constructions has been
reported and analysed [1, 2]. Fortunately the majority of those constructions show
only partly failure. Depending on spatial extension of the failure (e.g. a single com-
ponent, a single connection, a group of components, etc.) the construction has to be
repaired or in the worst case rebuild, to guaranty a safe use.
In particular cases the failure occurs only within one structural member while
the remaining structural members are apparently unaffected but could be damaged
also due to additional loading caused by load redistribution. In such cases the cor-
responding engineer has to decide for different repair alternatives. Possible alter-
natives could be (a) exchange of the failed member, (b) reinforcement of the all
members or (c) complete renovation of the entire construction.
To find the optimal decision it is essential to know the extend of the damage,
the load bearing capacity of the remaining structural elements as well as the load
at the time of failure or better the entire load history of the construction. Here it
has to be considered that the load bearing capacity of the remaining structural ele-
ments is composed of the load bearing capacity at the time of construction and the
deterioration during utilization.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 67
DOI: 10.1007/978-94-007-7811-5_6,  c RILEM 2014
68 G. Fink and J. Kohler

It is well known that a failure within one structural member occur if the effective
stresses through the applied load exceed the strength of the member. Therefore both
a huge realization of the load and a small realization of the resistance is essential.
A huge realization of the load can be a result of an extreme event (e.g. extreme
snow load) or a deviation from the planed conditions (e.g. green roof: excessive
fill) or a combination of both. On the other hand a low realization of the resistance
might have the following two reasons - or their combination: (a) the basic population
of the structural components do not fulfil the requirements which are given in the
corresponding requirements; e.g. the actual bending capacity fm,g,k a is smaller than
the target load bearing capacity fm,g,k . (b) the basic population of the structural
components fulfil the requirements and the failed member is a very low realization
of the population. Depending on the existing reasons different repair alternatives
should be chosen.
In general it is difficult to estimate the actual material properties of existing struc-
tural components. Under the consideration of information available, such as the age
of the building, the history of load (e.g. amount and duration of load) or the amount
of the damage a first estimation of the material properties can be made. However
in the majority of the cases the gained knowledge might be not sufficient for an ap-
propriate estimation of the material properties to make a final decision. Additional
to the information at hand, several destructive or non-destructive measurements can
be performed to enhance the estimation. It is obvious that an increasing amount of
additional measurements lead to an increase of the quality of the estimation.
In the present paper a framework will be presented on which the material prop-
erties can be estimated or rather the estimated material properties can be updated
based on the gained information (load history and/or test results). Following the
application of the framework will be shown in a case study.

2 Updating
A convenient way to combine all kind of information is by using Bayes updating.
The given information (so-called prior information) will be updated with the re-
sults of inspections of measurements. It seems likely that the planed conditions,
such as the target material properties can be used as prior information. For updating
the prior it has to be differentiated between so-called equality and inequality in-
formation. Equality information corresponds to measured variables and inequality
information denotes the information obtained by a measurement that some variable
is greater than or less than some predefined limit. Further it can be differentiated
between direct and indirect information; i.e. direct measurements of the quantity of
interest and the measurement of some indicator of the quantity respectively. Using
that aspects the information can be subdivided into 4 types; c.f. Table 1.
The procedure of the updating process is depending on the type of information,
in particular between equality and inequality information. Both procedures will be
described in the following.
Risk Based Investigations of Partly Failed or Damaged Timber Constructions 69

Table 1 Types of information

Information Description Example

Type A Equality Direct Bending capacity of the failed member

Type B Inequality Direct Minimum bending capacity of the non-failed members
Type C Equality Indirect Stiffness measurement on the non-failed members
Type D Inequality Indirect Status inspection

2.1 Updating Using Equality Information

The inspected parameter (e.g. resistance of the structural members) is represented
by the variable X with the probability distribution function FX (x). The parameters
= (1 , 2 , ..., n )T of the distribution function are not known with accuracy; i.e.
they are product of engineering knowledge, physical understanding or earlier ob-
servations of the quantity. The parameters are in general expressed as random
variables itself specified by the so-called prior density function f (). The uncer-
tain parameters can be updated based on new observations of realizations of the
variable X, x = (x1 , x2 , ..., xn )T . The general scheme for updating of the parameters
= (1 , 2 , ..., n )T is:

f ()L(|x) n
f (|x) = with L(|x) fX| (xi |) (1)
f ()L(|x)d i=1

where f (|x) is the posterior distribution function of the parameters , L( |x) is
the likelihood function, representing the knowledge gained by the observations and
n is the number of observations. Based on f (|x) it is possible to calculate the
predictive distribution according to:

f  (x) = fX (x| ) f (|x)d (2)

Having a normal or log-normal distributed variable with a known standard devia-

tion the approach can be simplified. In the following the analytical solution for the
predictive distribution of the strength R of structural components are described. R
can be assumed log-normal distributed. Thus Z = log(R) is a normal distributed
variable, with the parameters = (z , z )T . The standard deviation z is assumed
to be known and the mean value is a random variable z N(  ,  ). Having direct
or indirect observations the variable Z can be updated as following:
Direct information: Assuming the variable Z will be updated with n observations x.
Then the posterior distribution of the mean is normal distributed z N(  ,  ).
Following also the predictive distribution of the variable Z  is normal distributed
Z  N(  ,  ); c.f. [3, 4].
70 G. Fink and J. Kohler

nx +  n z z2
 =  = n = n = n +n  = 2 + z2 (3)
n n 2

Indirect information: The variable Z can be also updated with n indirect observations
of an indicator . The interrelation between the estimated and the actual strength
might be expressed through log(r) = log( ) + ; where is a normal distributed
error term N(0, ). Thus Eq. (3) can be extended with the expected mean
strength log( ) and its variability / n. The predictive distribution of the variable
Z  is normal distributed Z  N(  ,  ).

 nlog( ) + n  z2 + 2 /n
= = (4)
n n

2.2 Updating Using Inequality Information

The prior information can also be updated having inequality information. Assuming
a member will be proof loaded up to specific stresses level l without failure. Thus
the resistance can of the structural members can be represented by the variable X
following a truncated distribution function. In Faber et al. 2000 [5] the following
approach is proposed; here FR (r) is the distribution function of the resistance:

FR (r) FR (l )

FR (r) = r l (5)
1 FR (l )

3 Collecting Information
The first and most likely step for the estimation of the remaining reliability of the
non-damaged components is the collection of information available. That includes
information about the strength class of the investigated members, the load bearing
capacity of the failed member, the minimum load bearing capacity of the non-failed
members, the design load and a detailed investigation of the applied load. Here
permanent loads (especially for green roofs) and live loads should be investigated.
Updating the Load: Based on the investigations the prior information of the perma-
nent load (based on the design) can be updated (information Type A). An example
therefore could be the density of the fill. There the target density can be updated
with measured densities. Significant differences between the target and the mea-
sured density, indicates that the build in material might be wrong. In that case more
detailed investigations should be performed and if it became more accurate that the
wrong material is built in, the prior information has to be neglected.
Load Bearing Capacity of the Failed Member: Based on the updated load the ac-
tual stresses l at the time of damage can be calculated. It obvious that the ac-
tual stresses are consistent with the load bearing capacity of the failed member
ri = l (information Type A). However, if uniform loading is assumed, and all other
Risk Based Investigations of Partly Failed or Damaged Timber Constructions 71

members have not failed, thus l is the lower limit of their load bearing capacity
r j l , j = {1, 2, ..., n\i} (information Type B). This information can be used to
update the prior information of the strength properties (strength class). At this point
it has to be mentioned that based on the strength properties of a single structural
component it is not possible to identify if the the basic population of the structural
components do or do not fulfil the requirements which are given in the code. The
failed member can be a very low realization of the population. To analyse this further
investigations are essential.
Inspection Methods: In general the results, identified through the procedure de-
scribed in Section 2, might not be sufficient to make a final decision. Thus addi-
tional destructive and non-destructive measurement can be performed to enhance
the estimation. In the following a few selected methods are discussed. Thereby it is
assumed that the failed structural member is a glued laminated timber beam (GLT).
In Table 2 an overview of the selected methods is given.
Bending tests (GLT): The only way to gain information Type A is by testing one (or
more) of the structural components. Therefore the components have to be removed
from the building and tested into a lab. It is obvious that this method is time and
money consuming. However the gained information might be very efficient for fur-
ther decisions. It has to be considered that the adjacent components might be already
partially damaged through the additional load.
Tensile tests (Lamella / FJ): From the destroyed GLT beam the lamellas can be
cut out and tested under tension. There the tensile strength and the tensile stiffness
of the lamellas and the finger joint connections (FJ) are estimated. The specimens
should be taken from low loaded areas of the GLT to exclude lamellas that are
already deteriorated. Assuming that all GLT are produced out of the same source
material and under similar conditions (e.g. same factory) the material properties of
the beams can be estimated. Therefore models from the literature (e.g. [6]) can taken
into account.
Proof loading: The construction could be loaded again up to certain (higher) load
level. If the construction do not fail an additional information Type B is collected
(Otherwise information Type A).
MOE measurements: An indirect way to estimate the bending capacity is by using
the correlation to the bending modulus of elasticity (MOE). Therefore different non-
destructive methods exists; e.g. ultrasonic runtime or Eigenfrequency. Alternative
the MOE can also measured during a proof loading test (elastic deformation).
Visual inspections: The GLT members can investigated in detail with the focus on
the identification of possible indication of deterioration. However it is difficult to
quantify the outcomes. In best they can be used to estimate an upper value of the
load bearing capacity.
In Fig 1 a possible outcome of the updating process is illustrated. Based on the
density distributions the probability of failure and thus the reliability of the mem-
bers as well as the reliability of the structural system can be calculated. Following
72 G. Fink and J. Kohler

Table 2 Compilation of selected DT and NDT inspection methods

Inspection method DT / NDT Information Expected Efficiency

type costs
Bending tests (GLT) DT Type A ***** *****
Tensile tests (Lamella / FJ) DT Type C *** ****
Proof loading NDT (DT) Type B (A) **** **
MOE measurements NDT Type C * *
Visual inspections NDT Type D * *

Fig. 1 Schematic illustra-

tion of a possible outcome S R actual
of the updating process target

decisions concerning the further procedure, such as repair alternatives or more de-
tailed investigations, can be made.

4 Duration of Load Load History

As mentioned before it has to be considered that the load bearing capacity of the
remaining structural elements at the investigated time ti is composed of the load
bearing capacity at the time of construction t0 and the strength deterioration during
utilization. A typical realisation of R(t) and S(t) is illustrated in Fig 2.
In timber engineering strength deterioration during utilization is of particular in-
terest in the case of huge load [7]. According to the literature load of 60% of the
maximal load bearing capacity R(t) leads to strength reduction. Here it has to be
considered that R(t) is a function of the time, thus the probability of reaching the
critical load level is increasing with time. However deterioration during utilization
is a complex topic itself and will not disused in detail within this paper.

5 Case Study
In order to illustrate the developed methodologies a case study is made. Thereby a
risk analysis will be made for a partly damaged timber-hall located in Zurich (Year
of construction: 2000). The main structural elements are n = 100 single supported
beams (b h = 200 1600mm; GL28h) with a length of 20m each. The elements
of the secondary structure are single supported beams, located between the main
structural elements. A troughed sheet together with a green roof filling constitutes
Risk Based Investigations of Partly Failed or Damaged Timber Constructions 73

Fig. 2 Typical realisation of R, S

the load and resistance over
Realisation of R
the time, adapted from [8]
Realisation of S

Table 3 Load and resistance

Load / Resistance Char. value Distribution function

Permanent load gk = 14kN/m Normal distribution g N(14, 0.5)
Snow load gs,k = 8kN/m Gumbel distribution gs G(3, 1.5)
Bending capacity fm,k = 28MPa Log-normal distribution f m L(36.2, 5.43)

Assuming COV=0.15

the exterior shield of the roof structure. In Table 3 the applied load and resistance
are summarized.
The following failure scenario was observed: One GLT member of the structural
system failed in bending and the adjacent members showed no visible damage (Year
of damage: 2012). The load at the time of failure is investigated limitedly after the
damage and is thus assumed to be known: g + q = 20 + 4 = 24kN/m (78% of design
In a first step the actual load has to be compared with the designed load. It is
obvious that the measured permanent load is significantly above the the target value
(which might be a results of the wrong fill material), whereas the snow load has not
reached a critical magnitude. For the following investigation it is assumed that the
fill will be exchanged g N(14, 0.5).
Correspondingly fm can be updated. Concerning the relatively huge load over the
entire live time of the building a constant reduction of bending strength of 20% is
assumed. Given this assumption the probability that 1 of 100 member fails under
the allied stresses m 14MPa is rather small Pf 8.1 105. This indicates that
the assumed bending bending capacity might overestimate the real bending capac-
ity; i.e. the mean bending capacity is significantly lower. Thus in addition to the
variability of fm through constant standard deviation (z COV = 0.15) a standard
deviation of the uncertain mean of  = 0.10 is assumed. Following the information
of the bending capacity can be updated using information available (m = 14MPa)
and information based on additional inspections. In this particular case study the
following inspections are assumed: Bending tests on 3 GLT beams. NDT strength
estimation on 5 GLT beams (measurement uncertainty = 5MPa). Two different
74 G. Fink and J. Kohler

Table 4 Compilation of the estimated material properties and the corresponding relability

Description Information fm fm,k Pf

n fm [MPa] [MPa] [MPa] [-] [-]

1 14
No test 22.9 17.4 3.79 7.60 105
99 14
4 {14, 18, 22, 17}
Case A 96 14 20.2 15.8 3.45 2.85 104
20 = 20
4 {14, 29, 24, 28}
Case B 96 14 25.1 19.4 4.20 1.31 105
20 = 26
structural reliability of a single component

0.14 1
Probability density


0.08 0.6
0.06 0.4 No tests
Case A
0.2 Case B
0 0
0 20 40 60 0 20 40 60
Bending strength [MPa] Bending strength [MPa]

Fig. 3 Estimated bending strength

scenarios of test result are assumed (Case A, B; c.f. Table 4). Based on information
available and the experimental investigations the corresponding strength properties
are estimated using Eq. (3-5). For practical reasons primary the equality informa-
tion is used for updating. Using the updated resistance the reliability of the structural
components expressed through the reliability index can be calculated.
In Tab 4 a compilation of the results is given. The estimated strength properties
of the remaining GLT beams are significantly different for the two assumed inspec-
tion scenarios. In Case A the estimated strength properties are significantly below
the required values fm = 20.2MPa, fm,k = 15.8MPa. This indicated that the fail-
ure occurs as a result of inadequate material properties (wrong strength grade) or a
significant amount of strength deterioration as a result of the high permanent load.
However the results clearly show that the remaining structure does not fulfil the re-
quirements. A reinforcement of the members might be the optimal solution. In Case
B the fm = 25.1MPa, fm,k = 19.4MPa which is slightly above the design values.
Even in this case reinforcement might be a solution. Alternative an exchange of the
Risk Based Investigations of Partly Failed or Damaged Timber Constructions 75

roof structure (to reduce the permanent load) might be adequate. However in both
scenarios the experimental investigation leads to an improvement of the estimation
which could be very useful for the responsible engineer. At this point it has to be
mentioned that the here presented method is sensible to the chosen prior informa-
tion and the real strength deterioration through duration of load. For a practical
application both should be investigated in detail.

6 Conclusions and Outlook

In this study a framework for updating the load and resistance properties of a
partly failed timber construction based on different kind of information is presented.
Thereby it is primary focused on constructions where only a marginal part of the
construction failed (1 of 100 structural members). It is presented how information
available, like the load bearing capacity of the failed member or the minimum load
bearing capacity of the non-failed members can be used to estimate the load bearing
capacity of the remaining structural component. Further it is shown how information
of different types of destructive and non-destructive measurements can implemented
in the estimation. The application of the framework is illustrated on the case study.
The here presented model can be optimized having detailed knowledge about the
load history and its influence on the strength deterioration. This should be taken into
account in detail within a further study.

Acknowledgements. Part of this work was supported by COST Action FP1101 Assess-
ment, Reinforcement and Monitoring of Timber Structures.

1. Fruhwald, E., Serrano, E., Toratti, T., Emilsson, A., Thelandersson, S.: Design of safe
timber structures - How can we learn from structural failures in concrete, steel and timber?
Technical report, Lund University (2007)
2. Bla, H.J., Frese, M.: Schadensanalyse von Hallentragwerken aus Holz. Karlsruher
Berichte zum Ingenieurholzbau 16. KIT Scientific Publishing (2010)
3. Rackwitz, R.: Predictive distribution of strength under control. Materiaux et Construc-
tion 16(4), 259267 (1983)
4. Faber, M.H.: Statistics and Probability Theory: In Pursuit of Engineering Decision Sup-
port, vol. 18. Springer (2012)
5. Faber, M.H., Val, D.V., Stewart, M.G.: Proof load testing for bridge assessment and up-
grading. Engineering Structures 22(12), 16771689 (2000)
6. Bla, H.J., Frese, M., Glos, P., Denzler, J.K., Linenmann, P., Ranta-Maunus, A.: Zu-
verlassigkeit von Fichten-Brettschichtholz mit modifiziertem Aufbau, vol. 11. KIT Sci-
entific Publishing (2008)
7. Srensen, J.D., Svensson, S., Stang, B.D.: Reliability-based calibration of load duration
factors for timber structures. Structural Safety 27(2), 153169 (2005)
8. Melchers, R.E.: Structural reliability analysis and prediction. John Wiley & Son Ltd.
Naturally Grown Round Wood
Ideas for an Engineering Design

Matthias Frese and Hans Joachim Bla

Karlsruhe Institute of Technology (KIT),

Timber Structures and Building Construction,
R.-Baumeister-Platz 1,
76131 Karlsruhe, Germany

Abstract. The hardwood species European oak, European chestnut and black
locust are hardly used for engineering structures in their naturally grown shape,
although such a use is generally possible. From forest thinnings these hardwoods
are available as low-cost material to a certain extent and their durability and re-
markable strength can fulfil the demands of robust timber structures. However, the
irregular shape of naturally grown logs and the subsequent complexity of connec-
tions inhibit engineering applications. Thus, there is a gap between the potential of
an available low-cost material and the possibility of a value-added application. In
order to bridge this gap the present contribution aims first at giving basic ideas to
make naturally grown logs more calculable and second at presenting a modular
screw connection. Hence, compression tests on naturally grown logs, screw with-
drawal tests and tests on screw connections were performed. Based on the experi-
mental results, strength models for the compression capacity and the withdrawal
resistance of screws were developed and a high strength and ductile connection
type was designed. The findings of the work may widen the possibilities in the
engineering design of timber structures made of naturally grown logs.

Keywords: round wood, logs, strength model, oak, chestnut, steel connectors.

1 Introduction

In the past, logs in their naturally grown shape played always a certain role in
heavily loaded structures and branches of the industry where material costs rather
than architectural appearance are the main aspect. Bridges built by the Romans
were founded on timber poles and more or less debarked round wood was used to
support mining adits and tunnels under construction.
A brief literature review about the structural use of naturally grown logs (NGL)
and of machined cylindrical logs (MCL) revealed various activities in the two past
decades. Broader overviews on the topic are given in Ranta-Maunus et al. (1999),
Wolfe (2000) and Stern (2001). Specific ideas how to connect NGL/MCL to each

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 77
DOI: 10.1007/978-94-007-7811-5_7, RILEM 2014
78 M. Frese and H.J. Bla

other are presented in Eckelman and Senft (1995), Lusambo and Wills (2002),
Eckelman (2004), Eckelman et al. (2007), Shim et al. (2009), Gorman et al. (2012)
and Brito and Junior (2012). The works of Wolfe and Moseley (2000), Wolfe and
Murphy (2005) and Green et al. (2008) report, among others, results of bending
and compression tests that were performed with NGL. In addition to the first three
references above, case studies about the use of NGL/MCL in buildings and struc-
tures are described in Burton et al. (1998), Cantrell et al. (2004) and Yeh and Lin
(2007). Round wood projects were erected in Doncaster (Clark R 2000) and Dor-
set (Architectural Association Inc. 2009, cf. Burton et al. 1998) in the UK.
Trinkert (2008/10/12) presents three different round wood tower structures.
The mentioned works show in particular the potential of softwood species in
naturally grown shape for various load carrying structures. On this base, the pre-
sent work aims at adding some new aspects. From thinnings in German forests
small-diameter European oak, European chestnut and black locust are available as
low-cost material to a certain extent. Due to their natural durability and hardwood
strength these species could be used in weathered space frame timber structures
(fig. 1) that need an engineering design. The problems of the present work do
therefore address the following key aspects in relation with green timber: 1. the
compression capacity of NGL and its prediction, 2. the withdrawal capacity of
screws and a corresponding strength model and 3. an idea how to join NGL mem-
bers with regard to a modular connection as well as a high capacity and pro-
nounced ductility. The present work is a continuation with wider coverage of a
prior study by the authors (Bla et al. 2012).

Fig. 1 Towers, pylons and bridges: visions of timber structures with durable hardwoods

2 Material and Methods

For the research project, 32 European chestnut trees (castanea sativa) were har-
vested in autumn 2010 and 23 European oak trees (quercus robur) as well as 16
Naturally Grown Round Wood Ideas for an Engineering Design 79

black locust trees (robinia pseudoacacia) were harvested in early spring 2011. The
stands were in Southern Germany; the one of the chestnuts was Oberkirch (Black
Forest) and the one of the oaks and the black locusts was close to the Rhine not far
away from Freiburg. The breast height diameter including bark was between 20
and 40 cm. The trees were then divided into single logs. Occasionally up to three
logs at least 5 m in originating length came from one tree. In doing so, 54
chestnut, 30 oak and 21 black locust logs were realised, in total 105 pieces. Imme-
diately after dividing the trees, the logs were X-rayed to measure the natural three-
dimensional shape of the surface between the sapwood and the bark. With the data
obtained the corresponding surface shape was numerically described to determine
the geometric properties in a later step. Prior to testing the logs in January/Febru-
ary 2012 they were stored outside without any weather protection. It was intended
to test the timber in green condition. Thus, no special measures regarding drying
of the logs were undertaken. All tests described hereafter were performed as short
term tests. Bark and bast were not removed.

Fig. 2 Preparation of the specimens for the buckling and compression tests; position of the
tree discs for the moisture and density distribution

The 105 logs were tested twice: A longer section of each log was used for a
buckling and a shorter one for a compression test. Fig. 2 shows the corresponding
division of the logs. Section 1 with a constant log length of 4350 mm and a con-
stant buckling length (sk) of 4810 mm was used for determining the compression
capacity (Fc1,max); section 2 was planned as a control for determining the compres-
sive strength (fc2). Compared to the log length, the buckling length was longer due
to mounted hinges at both ends, cf. fig. 3c. Before testing section 1, the section
modulus (W1), the cross sectional area (A1) and the maximum eccentricity (eef,max)
were calculated using the numerical surface data. W1 and A1 are mean values
representative of the middle third of the buckling length; eef,max represents the
80 M. Frese and H.J. Bla

maximum eccentricity in the middle third. These terms are exemplified in fig. 3a.
The section modulus, the cross sectional area and the maximum eccentricity do
only refer to the present wood substance without the bark. The compression mem-
bers were compressed (fig. 3b) far beyond the compression capacity with a
displacement of at least 30 mm. The compressive strength of the shorter section 2
was determined with the compression capacity (Fc2,max) and its mean cross sec-
tional area (A2). Finally, two tree discs were sawed off as stated in fig. 2 of each
log. These specimens were used to measure the moisture content (u) and the wet
density (u). In doing so, in each tree disc two outer cubes (A) were planned to
represent the outer annual rings and one inner cube (B) to represent the inner
annual rings.
Fig. 4a shows the test scheme to determine the screw withdrawal resistance
(Rax). Fully threaded screws made of stainless steel were used. The nominal di-
ameter was 8 mm. The screws were inserted into holes that were predrilled with a
diameter of 5 mm. The tensile capacity of the screws lies between 16.3 and
16.8 kN. The penetration depths (ef) were 40 to 60 mm for black locust, 60 mm
for oak and 60 to 100 mm for chestnut. Different orientations in radial und tangen-
tial direction were used. After each test the wet densities and the corresponding
moisture contents were examined. Fig. 4b exemplifies a typical failure mode
observed at the surface of a chestnut specimen.
The screw connection described in Fig. 5a was newly developed. It is shown in
an early development stage. The connection principle should include the use of
inclined screws, ensuring high stiffness, and steel connectors, easy to mount.
Regarding later realisation in timber structures as space frame structures, the
preparation of the members should be performed by computerised numerically
controlled woodworking machines. The final aim is to produce modular members.
The main features of the connection are: chestnut logs were machined to a cross-
shaped specimen. Two times four steel angles were arranged in the re-entrant
edges. Specific washers were inserted in appropriate holes in the angles to ensure
proper load transmission in the inclined screws ( = 45) connecting the timber
and the two legs of the angles (fig. 5b). The screw type was the same that was
used in the withdrawal tests. It was intended to arrange the screws with tangential
orientation to avoid screw positions in probable radial V-cracks which are typical
for round wood containing the pith. The connection was designed so that the
screws are fully embedded in the durable heartwood. The testing load was applied
at both ends at the midpoint of the four angles. It is, therefore, more or less equally
distributed between the angles. Three specimens identical in construction were
tested in tension up to a displacement of at least 15 mm in one of both connections
(either top or bottom). After the tests the moisture content and the wet density of
the wood that directly surrounds the inserted screws was determined. These
values are later used to link the results of the withdrawal tests to the results of the
connection tests.
Naturally Grown Round Wood Ideas for an Engineering Design 81

a) b) c)
Fig. 3 Geometric log data (a), buckling test (b) and equipment for the hinge joint of the
buckling tests, the bearing shell was mounted at the end of the logs (c)

a) b)
Fig. 4 Withdrawal test with fully threaded screws (a) and typical failure (b)

a) b)
Fig. 5 Tension test (a) and connection detail with inclined screws (b)
82 M. Frese and H.J. Bla

3 Results

Fig. 6a shows load displacement curves of buckling tests. Three curves are plotted
for each species. These curves are representative of log members with the highest,
medium and lowest compression capacity. Independently of the species compres-
sion capacity is reached between 10 and 20 mm vertical displacement. The geo-
metric, mechanical and physical data of the log sections are compiled in table 1.
Since the buckling length was constant for all tested log sections, the oaks and
chestnuts feature the highest cross sectional areas and the lowest mean eccentrici-
ties and do therefore show the highest compression capacities.

Table 1 Statistics of the geometric, mechanical and physical data

N mean std min max

European oak
A1 in cm 30 411 101 220 610
eef,max in mm 30 69.5 30.1 21.3 136
Fc1,max in kN 30 377 228 110 1070
fc2 in N/mm 30 30.3 3.32 24.2 36.4
u (A) in % 120 61.7 9.41 24.9 78.6
u (B) in % 60 63.4 8.60 30.5 78.4
u (A) in kg/m 120 939 58.1 704 1070
u (B) in kg/m 60 962 52.1 789 1050
European chestnut
A1 in cm 54 427 136 207 803
eef,max in mm 54 46.9 24.4 11.5 131
Fc1,max in kN 54 509 256 126 1170
fc2 in N/mm 54 27.3 3.32 20.5 33.9
u (A) in % 216 82.7 15.8 49.5 140
u (B) in % 108 77.2 14.1 52.9 130
u (A) in kg/m 216 847 80.2 622 1022
u (B) in kg/m 108 791 66.2 606 935
Black locust
A1 in cm 18 367 70.2 259 470
eef,max in mm 18 106 38.6 41.7 171
Fc1,max in kN 21 297 118 123 571
fc2 in N/mm 21 38.9 5.10 30.2 50.1
u (A) in % 84 40.5 6.22 21.8 59.1
u (B) in % 42 38.7 5.27 26.3 47.3
u (A) in kg/m 84 851 62.0 745 1080
u (B) in kg/m 42 798 61.6 683 912
Naturally Grown Round Wood Ideas for an Engineering Design 83

The mean compressive strengths that were determined with the control sections
2 are 27, 30 and 39 N/mm for chestnut, oak and black locust, respectively. The
stated mean moisture contents and mean wet densities are calculated on the basis
of the two tree discs containing four outer (location A) and two inner cubes (loca-
tion B). Corresponding moisture and density differences between two discs be-
longing to one log (cf. fig. 2) were low. The comparison between the locations A
and B shows moderate moisture and density differences. Considering all cubes at
the locations A and B, the mean moisture contents (mean wet densities) are ap-
proximately 40 (830), 60 (950) and 80 % (830 kg/m) for black locust, oak and
chestnut. The relation between the compression capacity and the cross sectional
area is shown in fig. 6b. Black locust and oak show lower capacities for the same
cross sectional area compared to chestnut. This is most likely due to the more
pronounced deviation from the ideal cylindrical shape of the first two species.
Regarding black locust, the influence of the compressive strength on the compres-
sive capacity seems to be low.

1.2 1.2
____ Oak
_ _ _B. locust
Fc1,max [MN]

0.8 0.8
Fc1 [MN]

0.4 0.4

B. locust
0.0 0.0
0 5 10 15 20 100 300 500 700 900
Displacement [mm] A1 [cm]
a) b)
Fig. 6 Load displacement curves of the buckling tests (a) and compression capacity and
cross sectional area (b)

Fig. 7a shows the relation between the withdrawal resistance and the penetration
depth. The mean moisture contents (mean wet densities) are 46 (854), 43 (659)
and 30 % (842 kg/m) for the oak, chestnut and black locust specimens. Black
locust does therefore show the highest withdrawal resistance and chestnut the
lowest for a given penetration depth. The performance of oak is similar to that of
black locust. The withdrawal resistance is limited by the maximum tensile capac-
ity of 16.8 kN of the screws used.
Fig. 7b shows the load displacement curves for the three connection tests. For
each test the tension force is plotted over the displacement of the failed connection
side. Each connection showed one side that survived and one side that failed. The
capacities range from 302 to 347 kN related to a displacement of approximately
5 mm. Beyond this displacement the load slip behaviour was ductile. For the ulti-
mate displacement of 20 mm each connection showed a capacity of approximately
70 % of the maximum.
84 M. Frese and H.J. Bla

18.0 400

15.5 300
Rax [kN]

Ft [kN]
13.0 200

10.5 100 Test

Oak 1
Chestnut 2
B. locust 3
8.0 0
30 40 50 60 70 80 90 100 110 0 5 10 15 20
lef [mm] Delta on the failure side [mm]
a) b)
Fig. 7 Withdrawal resistance of screws and penetration depth (a) and load displacement
curves of the connection tests (b)

4 Strength Models and Validation of the Connection Tests

Based on the data of the buckling and withdrawal tests two strength models were
found. For that, a nonlinear and a linear regression analysis, respectively, were
performed. Here, the specified models are a scientific base to reproduce the capac-
ity of compression members and the withdrawal capacity of screws. In both
models, r is the coefficient of determination, e is the error of the prediction and N
denotes both the number of observations used and a normal distribution with mean
value and the standard deviation (in brackets), respectively. The models can be
applied independently of the three examined hardwood species.
Equation (1), a nonlinear model, describes a static equilibrium between the
maximum external compression force and the resulting internal bending and
compression stress. Therefore, the equation features linear bending-compression
interaction. It enables a capacity prediction denoted as Fc1,max,p. The prediction
necessitates the given eccentricity, the section modulus and the cross sectional
area as independent variables. The unknown model parameters, the ideal bending
(fm,i) and ideal compressive strength (fc,i), were estimated by the regression analy-
sis. The specified values are 33.5 N/mm for the ideal bending and 24.3 N/mm for
the ideal compressive strength. It is not a contradiction, that the ideal compressive
strength is lower than the ideal bending strength. From technological point of view
one would expect a higher bending strength in comparison with the compressive
strength particularly for green timber. As the ideal compressive strength compen-
sates for further capacity reducing influences that are not covered by the static
equilibrium, it is less than the mean compressive strength of fc2 = 27.3 N/mm
which is the mean compressive strength of chestnut and the lowest among the
three species, respectively (table 1). The agreement between the experimental and
predicted capacities is shown in fig. 8a. The plot shows that an individual error
Naturally Grown Round Wood Ideas for an Engineering Design 85

distribution is indicated in particular for black locust. Normal distributions of er-

rors that depend on the three species are therefore given in term (2). Their unit is
Newton. As a result, equation (1) provides a means to calculate a characteristic
compression capacity for compression members. For that, individual limits for the
eccentricity and for the minimum cross sectional areas have to be found. For the
three hardwood species different error distributions apply. Calculated capacities
are associated with a maximum buckling length of 4810 mm. Consequently, de-
viations from this have to be adequately considered in practical applications.

e 1
Fc1,max,p = 1/ ef,max + + e
f W
m,i 1 f c,i A1 (1)
N = 102 r = 0.92
e:N ( 1.3 10 ;69 10 3 3
Oak: N = 30 e:N ( 25 103 ;62 103 )
Chestnut: N = 54 e:N ( 4.0 103 ;73 103 ) (2)
B.locust: N = 18 e:N ( 46 103 ;43 103 )

Equation (3) states a linear model. It is suitable to calculate a prediction for the
withdrawal resistance (Rax,p in Newton) of screws with a nominal diameter of 8
mm. The prediction depends on penetration depth (in mm), wet density (in kg/m)
and moisture content (as ratio). The agreement between the experimental and pre-
dicted values is shown in fig. 8b. As opposed to fig. 8a the individual errors are
well balanced around the bisection line. The model and its errors do therefore
apply without major restrictions to the three hardwood species.

ln(Rax,p ) = 6.91 + 0.0305 ef + 0.00178 u 1.33 u 2 0.000118 ef 2 + e

N = 78 r 2 = 0.81 e : N(0;0.0730)

1.2 9.8

Fc1,max [MN]

ln(Rax) [-]


Oak Oak
Chestnut Chestnut
B. locust B. locust
0.0 9.0
0.0 0.4 0.8 1.2 9.0 9.2 9.4 9.6 9.8
Fc1,max,p [MN] ln(Rax,p) [-]
a) b)
Fig. 8 Experiment over prediction: compression capacity (a) and withdrawal resistance (b)
86 M. Frese and H.J. Bla

Estimated values of the withdrawal capacity (Rax,p) were compared to the

experimental capacities (Ft,max) of the three connection tensile tests. For that, equa-
tion (3) was used. Inserting the penetration depth of 80 mm, individual wet densi-
ties and moisture contents (present in the cross-shaped specimens, cf. fig. 5a)
result in individual withdrawal capacities (table 2, column 5). A total of 28 screws
and its inclination of 45, regarding the static equilibrium, were considered to
calculate the estimated tension capacity Ft,p (column 7). The experimental capaci-
ties Ft,max (column 8) amount to 109 % of the calculated ones. The difference can
be explained among others by friction in the wood-steel interface and a screw
angle different from 90. As in general the influence of screw angles between 90
and 45, regarding the fibre deviation, on the withdrawal capacity is low, the fric-
tion coefficient that is effective in the wood-steel interface amounts to not more
than 0.09.

Table 2 Validation of the connection tests

Test ef u u Rax,p ncos Ft,p Ft,max Ft,max/Ft,p

mm kg/m - N - N N -
1 80 651 0.303 15200 280.707 300,000 326,000 1.09
2 80 787 0.480 16100 280.707 318,000 347,000 1.09
3 80 588 0.263 14000 280.707 277,000 302,000 1.09

5 Conclusions

The results obtained confirm that naturally grown logs (NGL) of the durable
hardwood species European oak, European chestnut and black locust are suitable
to be used in engineered structures that are exposed to the weather. The tests on
NGL show high compression capacities that mainly depend on the diameter and
the eccentricity. The tested connection prototype for NGL involving inclined
screws provides a high load transmission and a pronounced ductility. This is a
necessary criterion for NGL to be used e.g. in space frame structures. All the ex-
periments were performed with NGL having moisture contents above the fibre
saturation point. Consequently, the results apply to green timber and regarding
structural use, drying measures are not necessary.
More research should address the grading of NGL and the determination of
suitable limits for deviations from the ideal cylindrical shape. Screw connections
with economic steel connectors have to be enhanced. Open questions concern a
probable debarking and the long term behaviour of the less durable sapwood.
Characteristics of the axial stiffness have to be found and the effect of load
duration on the deflections is still not known.

Acknowledgements. This work was funded by Baden-Wrttembergs Ministry for Rural

Area and Consumer Protection and by the European Union. Parts of the support are means
Naturally Grown Round Wood Ideas for an Engineering Design 87

of the European Regional Development Fund (ERDF) with the aim of Competitiveness and
Employment. The authors would like to thank their project partners University of Freiburg
Institute of Forest Utilization and Work Science, HECO-Schrauben GmbH & Co. KG
Schramberg and Forstliche Versuchs- und Forschungsanstalt Baden-Wrttemberg for their

Ranta-Maunus, A. (ed.): Round small-diameter timber for construction. VTT publications
383, Technical Research Centre of Finland, Espoo (1999)
Wolfe, R.: Research challenges for structural use of small-diameter round timbers. Forest
Products Journal 50(2), 2129 (2000)
Stern, E.G.: Construction with small-diameter roundwood. Forest Products Journal 51(4),
7182 (2001)
Eckelman, C.A., Senft, J.F.: Truss system for developing countries using small diameter
roundwood and dowel nut connection. Forest Products Journal 45(10), 7780 (1995)
Lusambo, E., Wills, B.M.D.: The strength of wire-connected round timber joints.
Biosystems Engineering 82, 339350 (2002)
Eckelman, C.A.: Exploratory study of high-strength, low-cost through-bolt with cross-pipe
and nut connections for square and roundwood timber frame construction. Forest Prod-
ucts Journal 54(12), 2937 (2004)
Eckelman, C.A., Haviarova, E., Erdil, Y.: Exploratory study of small timber trusses con-
structed with through-bolt and cross-pipe heel connectors. Forest Products Jour-
nal 57(3), 3947 (2007)
Shim, K.B., Wolfe, R.W., Begel, M.: Nailed mortised-plate connections for small-diameter
round timber. Wood and Fiber Science 41, 313321 (2009)
Gorman, T.M., Kretschmann, D.E., Begel, M., et al.: Assessing the capacity of three types
of round-wood connections. In: Quenneville, P. (ed.) Proceedings of the 12th World
Conference on Timber Engineering, Auckland, July 15-19. Curran Associates Inc., Red
Hook (2012)
Brito, L.D., Junior, C.C.: Types of connections for structural elements roundwood used in
Brazil. In: Quenneville, P. (ed.) Proceedings of the 12th World Conference on Timber
Engineering, Auckland, July 15-19. Curran Associates, Inc., Red Hook (2012)
Wolfe, R., Moseley, C.: Small-diameter log evaluation for value-added structural applica-
tions. Forest Products Journal 50(10), 4858 (2000)
Wolfe, R., Murphy, J.: Strength of small-diameter round and tapered bending members.
Forest Products Journal 55(3), 5055 (2005)
Green, D.W., Gorman, T.M., Evans, J.W., et al.: Grading and properties of small-diameter
Douglas-fir and ponderosa pine tapered logs. Forest Products Journal 58(11), 3341
Burton, R., Dickson, M., Harris, R.: The use of roundwood thinnings in buildings a case
study. Building Research & Information 26(2), 7693 (1998)
Cantrell, R., Paun, D., LeVan-Green, S.: An empirical analysis of an innovative application
for an underutilized resource: small-diameter roundwood in recreational buildings. For-
est Products Journal 54(9), 2835 (2004)
Mc, Y., Yl, L.: Use of small thinning logs in a round-wood trussed bridge. Forest Products
Journal 57(3), 3438 (2007)
88 M. Frese and H.J. Bla

Clark, R.: The Solar Canopy and sustainable energy promotion. In: Sayigh, A.A.M. (ed.)
Proceedings of the 6th World Renewable Energy Congress, Brighton, July 1-7. Elsevier
Science Ltd., Oxford (2000)
Architectural Association Inc., Hooke Park, Buildings & Projects (2009),
hookebuildings.php (accessed April 24, 2013)
Trinkert, A.: Der Turmbau zu Warstein. Bauen mit Holz 110(12), 1013 (2008)
Trinkert, A.: Ein Blick bis zu den Alpen. Bauen mit Holz 112(4), 1417 (2010)
Trinkert, A.: Mit Blick ber das Wasser. Bauen mit Holz 114(7/8), 2023 (2012)
Bla, H.J., Frese, M., Brchert, F., et al.: Computertomographische Untersuchungen und
Druckversuche an Robinienrundholz. Report No. KIT-SR 7610, KIT Scientific Publish-
ing, Karlsruhe (2012)
Recycling and End-of-Life Scenarios for Timber

Annette Hafner, Stephan Ott, and Stefan Winter

Chair of Timber Structures and Building Construction,

Technische Universitt Mnchen, Germany

Abstract. In consideration of sustainable buildings, closing life cycle loops be-

comes more and more important. Up to now reuse and recycling is taken rarely
into account in building processes. With rising consumption of wood for energetic
use recycling of material becomes more important.
Up to now there are various studies in EU market ([1], [2], [3]), which quantify
the usage of wood in market shares. Explicit calculations on recycling of wooden
material in the building sector have not yet been done. In general the demand for
reclaimed wood products in the building sector will rise due to the fact that the
preferred option has to be the reuse and the recycling of reclaimed wood. The
thermal use of wood is the last option in the cascade of use. On this option
the refinement of reclaimed wood for innovative products as well as the broaden-
ing and enhancement of the cascades of reuse and recycling is strongly needed for
the timber construction industry. Long-term and resource efficient use of wood
from premium quality (like laminated wood, plywood, timber frame construction)
is necessary to ensure sustainable construction with wood. In the process of plan-
ning new wooden construction the dismantling and reuse / recycling of the prod-
ucts has to be considered too.
In this paper outcomes of the woodwisdom-net research project ECO2 wood
in carbon efficient construction as well as calculations on wood consumption of
wide-span timber structures and investigated case studies on a very detailed level
are brought together to show the state of art and theories to improve resource effi-
cient usage of wood. Aim is a realistic estimation of theoretical scenarios for end
of life and their influence on planning processes as well as the influence on life
cycle assessment according to EN 15978. In another approach the total demolition
of an old wooden house in the Alps was evaluated. It is a typical example for a
long-used construction with numerous repair intervals, changes, and additions.
This leads to a wide variety of fractions and often to a contamination of wood
from preservatives. The fractions of the demolished house mainly consist of small
bits and pieces dedicated to different recycling options than wide span structures.
The different waste wood fractions in strength, scale, and size will tolerate certain
processing options with an emerging range of recycling products.
A better management of its renewable resources supports the material supply of
the wood sector to ensure a long-term availability of solid wood products at

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 89
DOI: 10.1007/978-94-007-7811-5_8, RILEM 2014
90 A. Hafner, S. Ott, and S. Winter

reasonable prices. This will allow preservation and also gain market shares now
and in the future.

Keywords: Life cycle, end of life, reuse, recycling, timber structures, ECO2-

1 Introduction
With a growing importance of wood as significant biomass component of the re-
newable energy supply, there might be a shortage in the availability of wood at
reasonable prices in the future [4]. The European Commission therefore proposed
to increase the efficiency in the production and the use of wood [5]. A better man-
agement of its renewable resources helps the wood sector to ensure a long-term
availability of solid wood products. Solid and glued wood products should have
the chance of a second, similar product life instead of recycling without preserva-
tion of solid wood properties. Today's most important challenge is to activate the
potential of the wood sector to provide construction products fulfilling these
For sustainable buildings closing life cycle loops becomes more and more
important. Up to now reuse and recycling of existing buildings are not examined
systematically. If at all reuse and recycling is rarely taken into account for new
building processes. The advantage of wood as a carbon storage can be obtained
through reuse and recycling of the material. Only after all options of recycling are
exploited the additional advantage of energy recovery should be used. This has a
positive effect, which will lead to improved lifecycle assessment (LCA) indicators
for wood based products. This paper shows the state of art and theories to improve
resource efficient usage of wood.
As shown in [6] the consumption of wood for material use will rise slowly in
the future, whereas the consumption of wood for energetic use will rise dramati-
cally. In Germany wood for energetic use already exceeds the material use in
2012, five to six years earlier as actual calculations from Mantau have predicted in
2010. This fact causes a rising share of fresh wood thermally used, and a shortage
and in long run rising cost for raw material supply of the timber construction sec-
tor. The use of good quality, recovered wood not only as energy source reduces
this economic pressure. Old timber structures can provide the necessary dimen-
sions and quantities of wood.
Therefore various research was done estimating the amount of material for re-
cycling in wooden buildings and wide-span timber structures. In addition related
outcomes of finished woodwisdom-net project ECO2 wood in carbon efficient
construction are included in this paper.

2 General Framework on Recycling of Wood

The European Union has set an objective to develop itself as a recycling society,
where waste generation is avoided, and waste generated is utilised as a resource.
Recycling and End-of-Life Scenarios for Timber Structures 91

The latest waste directive from 2008 [7] contains an article for reuse and recycling
of materials. Among other things, it requires that the member countries have to
proceed with necessary actions to recycle materials and products. To fulfil the
normative requirements, the industries and R&D are in charge to develop products
that can be easily recycled. In the wood product sector, the waste hierarchy is
largely underdeveloped, so far. A lot of wood products that could be utilised in
secondary product life cycle are burned for energy recovery or are down-cycled
that they lose material properties of solid wood. For realistic recycling scenarios,
the knowledge about the sources, the quantity and quality of reclaimed wood is
There are various studies in EU market, which quantify the usage of wood in
market shares [1], [2], [3]. Up to now no exact evaluation is confirmed on how
much recycled wood exists in our building stock and how high the potential of
reuse or recycling is from this material source. There is already a small amount
from recovered wood used in the production of particle boards or wood fibre prod-
ucts but it is a small share. Explicit calculations on quantities of recovered wooden
material in building sector have not yet been done.
Beside precise evaluation about volumes of wooden material for reuse purpose,
reuse and recycling faces other obstacles. Wooden material can only be reused if it
is not contaminated with harmful substances. Timber has been treated with poi-
sonous chemicals since the beginning of the 20th century, to reduce the risk of
mould or to impregnate it against insects. This fact decreases the possibility of
reuse and even recycling of reclaimed wood nowadays. For reuse purposes the
contaminated wood has to be identified and its amount has to be quantified. Addi-
tionally the type of chemical treatment has to be analysed for bio-hazardous sub-
stances and human-hazard toxicity of treated timber construction. These quantities
have to be sorted out and treated according to the rules for hazardous wastes [8].
According to German law the term reclaimed or recovered wood (Altholz) means
used wood from production and end user, as far as it is covered by the German
life-cycle Resource Management Act [9].

3 Related Outcomes of ECO2-Project: Wood in

Carbon-Efficient Construction

The European woodwisdom research project ECO2 wood in carbon efficient

construction focused on creating holistic understanding of carbon efficiency in the
full life-cycle of a wooden building and defining technical potential and obstacles
for the use of wood in carbon efficient construction. Aim was a clear minimization
of carbon emissions during production, construction and in the full life-cycle.
Different case studies on life cycle of wooden products and buildings were exam-
ined and energy and carbon balance examined.
92 A. Hafner, S. Ott, and S. Winter

3.1 General Framework on Lifecycle Assessment in Buildings

According to the standard EN 15978: sustainability of construction works As-
sessment of environmental performance of buildings Calculation method the
lifecycle of a building is divided up in different modules from A to D. Fig. 2
shows the division in phases and the input of residues in lifecycle. In that context
all scenarios for recycling and in general end of life of a product are integrated in
module C. It includes deconstruction / demolition (C1), transport to the products
waste processing (C2), waste processing for reuse, recovery or recycling, recovery
and/or disposal (C3), disposal (C4).
To show possible benefits and loads of materials beyond the product system
boundary, an additional module D is introduced. This means that in module D the
recycling potential, the persistence of mineral building products, avoided impacts
of thermal energy recovery, embedded renewable energy or carbon stored in the
product can be shown. Up to now end-of-life scenarios for wooden products
nearly only consist in incineration and energy recovery.

Fig. 1 Residues from wood production in different phases of lifecycle

The treatment of stored carbon in life cycle analysis is a crucial factor which is
not handled equally in LCA calculations up to now. Trees embed carbon in the
material while growing in the forest. Several LCA calculations therefore regard
wood as GHG neutral and dont calculate them at all. This is only the case if
the wood is not leaving the forest system and if these forests are not being har-
vested. Wood from forest is counted with negative greenhouse gas emissions
when it enters module A due to the carbon stored in the product. The biogenic
carbon emissions directly attributed to a wood based product result either from the
use of biomass energy in production phase (module A) or from combustion of the
product after end-of-life (module C). These emissions are equal to the amount of
carbon sequestered in the growing tree which provides the biomass for the wood
or the energy used. Prerequisite for that is always that sustainable forestry is ap-
plied. LCA calculations done according to the standards of EN15804 and
EN15978 do not give instructions for the handling of wood and sequestration of
carbon. But according to these standards the carbon and primary energy have to be
accounted separately in the different modules. This requires that carbon balance is
shown divided up in the modules. Hence wooden materials become a negative
value in module A1 and a positive value in C4. Energy gains and the carbon stored
Recycling and End-of-Life Scenarios for Timber Structures 93

in the product (if it is reused or recycled) have to be shown in module D. Carbon

storage gets prolonged by recycling or reusing wooden material at end-of-life for a
second life-cycle.

3.2 Characteristics of Wood in End-of-Life of LCA

While buildings are seen as a whole in the use phase, for end-of-life it comes
down to the specific construction and the materials they are made of. Building
components can be decomposed into different layers to get a deeper understanding
of their impact at end of life. The layers of the building have different exposures,
durability and therefore a different life span. The disassembly allows the identifi-
cation of required service life of building parts and has to be considered in main-
tenance, inspection, end-of-life scenarios. In modern (timber) buildings different
layers are also common to fulfil a wide variety of technical requirements.
The principles of faade layering are divided into its structural and functional
layers. The primary construction needs to outlast the whole life span, wooden
primary construction has high mass and can store a high amount of carbon over
this period. The other layers e.g. cladding will be replaced many times and there-
fore are relevant in terms of recycling potential and burdens / benefits at EoL
stage. This can be used as design methodology for the improvement of environ-
mental performance. Through the interdependency of use- and EoL-scenarios, the
different layers can separately be optimized more easily and then be designed for
reuse. The jointing between layers and the frequency of renewal are additional
criteria for the end-of-life phase apart from the material impact. [11]
Because of a wide spread of unplanned demolition reuse of construction parts is
problematic. For optimal reuse the available construction parts have to be classi-
fied at least for different end-of-life scenarios. Construction wood and engineered
wood products can be dismantled easier, because they are used in dedicated layers,
are often used in modular assemblies and are light-weight. All these properties
allow professional waste streaming and management. After the planned disman-
tling process of removable parts, the rough demolition of monolithic portions can
take place. Such fractions may contain smaller amounts of wooden bits and pieces
which are reserved for recycling or thermal use.

4 Case Study on an Existing Wooden Building for

Calculation of Quantities

Aim of this study was a realistic assessment of the material fractions in existing
old wooden buildings. Basis for the examinations were an overall inventory analy-
sis and documentation of existing construction with classification in qualities and
quantities, see [12]. Used methods were: research with existing planning material,
survey on site and examination of material samples for hazardous substances.
94 A. Hafner, S. Ott, and S. Winter

The amount of wood contained in a historic cottage in the Alps, dedicated to

demolition, was evaluated. It is an example for construction with a long life cycle
with numerous repair intervals, changes, building extensions, and a conglomerate
of different materials.
Thereby reusing and recycling is not possible without a careful separation of
the materials, which are non-toxic and toxic. Via pyrolysis and use of spectros-
copy harmful substances like polycyclic aromatic hydrocarbon (PAK) and Penta-
chlorphenol (PCP) in coatings were detected. Result of the analysis showed, that
the main part of the timber construction could be reused as further components.
Only parts of the faade cladding and materials included in the construction phase
of the 1960s contained hazardous substances. Most material from the original
cottage was free of harmful substances and was also in good and dry condition.
6.7% of wooden material contained harmful substances, mainly the cladding
which was exposed to painting in the last decades of the 20th century.
Although the cottage is an example from wooden buildings material results
show the following. Analysis of material fractions in volume shows a high per-
centage of massive material in the cottage. The share of wooden material reaches a
share of 38% in volume with very simple wooden constructions. The main mass is
concrete / stone 63%. Complete analysis of construction is shown in Fig. 4.

Fig. 2 Construction material different fractions. Comparison of volume [12]

To compare results to calculations done in chapter 5 a combined value of m

wood per m gross floor area is introduced. It is 5.1 m/m or as a reciprocal value
0.20 m wood per m gross floor area. This value shows the amount of used wood
in comparison to the m of gross floor area.

5 Analysis of Wide-Span Timber Structures

The problems of the reuse of contaminated wood shift the focus on younger build-
ings without chemical treatment of the timber construction. High quality engi-
neered wood products like glulam, solid wood panels or CLT are in centre of
further investigation.
Recycling and End-of-Life Scenarios for Timber Structures 95

Structures with wood in every grade and in every size can be found throughout
entire Germany. TUM has a large database of with detailed information about
wide span timber structures. The presence of comprehensive data such as the
number of wood structures, their type and span, gives precise figures about the
wood volume stored. The collection of data is limited to roof structures made of
wood, as they are relatively easy to evaluate, compared to other wooden buildings.
The available beam dimensions and types of wooden roof structures are well
suited for reuse because they are serialized, modular and similar in dimension and
grade. For evaluating 116 hall structures, as well as other large-scale structures, a
wide range of use type, size and age, is recorded in tabular form. The evaluation is
adapted as a whole as well as on the individual structure type.
The classification of the parameters starts with the desired result, the amount of
used wood. To estimate the amount of used wood, span, beam depth, truss spacing
and truss number as input variables are necessary in addition to the truss form and
design. To describe the use of wood in a structure accurate, it is necessary to ex-
plain it according to its shape-defining parameters. This is the consumption of
wood in structures in relation to the footprint of the structures.


This ratio m/m allows a comparison of trusses with each other. The relation-
ship is suitable for a general assessment of the truss, since it contains all necessary
parameters; with the counter, namely the footprint, which includes span, truss
number and spacing, as well as the denominator, consumption of wood. The out-
put size will give the covered area and standardizes it according to the quantity
wood used. Parameters truss number and distance is secondary. There is no influ-
ence on wood consumption coefficient due to correlation between area and wood
volume, compare Fig. 3. The representation of wood consumption in relation to
the floor space for each type is calculated as an own indicator from the averages of
the demand for wood, compare Fig. 7. The nominal size of this factor increases
the resource savings. The inverse value defines the amount of exploitable recycla-
ble material per square meter floor space.

Table 1 Average results of lattice versus solid cross section trusses

Floor area Span S Amount n Distance e Wood

[m] [m] [-] [m] [m]

Solid-CS 1415 25 9 6 91

Lattice 836 19 12 5 41
96 A. Hafner, S. Ott, and S. Winter

The type of the saddle roof truss is the most used with over 40 observations in
the database. According to Table 1 the solid cross section has advantages in dis-
tance and number of trusses although these numbers do not influence the wood
volume. Fig. 3 shows the results of saddle roof truss with a lattice construction
versus a solid cross section. The lattice truss obviously has the lower wood con-
sumption but that definitely proofs an assumption that can be made easily. The
distribution of spans of lattice truss in Fig. 3 shows a concentration around 15
meters while the solid truss distributes over full bandwidth. A reason might be the
production effort of lattice truss is lower in short span while long spans are more
effective in solid construction due to continuous lamination technology.

Poly. (Voll-QS)
volume wood [m]

300 R = 0.74
Power (FW)
R = 0.67
0 5 10 15 20 25 30 35 40 45
Span [m]

Fig. 3 Saddle roof truss framework (FW) compared to solid cross section (Voll-QS) in
correlation with span

Fig. 4 Comparison of wide-span roof structures shows the covered floor space in relation to
1 m3 of used wood

The evaluation could be completed with the statement, that the most used
trusses have the lowest covered floor space in relation to 1 m of used wood and
therefore have the highest wood consumption. The scattering of values around the
Recycling and End-of-Life Scenarios for Timber Structures 97

average of 20 m/m and the plateau at this point indicates that regardless of truss
type the evaluation shows evenly distributed results without outliers. Results of
Fig. 4 normalized to the span of the three-hinged arch truss offers a variation of
8.4 to 13.6 m/m of wood consumption for the other types of trusses. The types
with the highest deviation are parallel trusses and three-hinged arch trusses.
Higher wood consumption of parallel truss can be explained with the inefficient
structural shape compared to for example saddle roof truss. The three-hinged arch
has the same problem of a continuous cross section although there is an advantage
that the normal force follows the arch shape. The type load bearing structure and
therefore the bending moment also have an influence on consumption of wood.

6 Conclusions

For future planning the reuse and recycling of buildings has to be integrated quite
early in the planning process to be able to reuse the materials in the best way.
Small scale drawings also show how components are built in and how jointing is
done. This information is useful for dismantling.
With existing building documentation fictive mass calculation could be pro-
duced which then can be updated by partial surveys. In future planners contracts
have to be enlarged to include also reuse of material in early planning stages. For
recycling of existing buildings exact calculation of actual mass and volume as well
as building condition is necessary. Therefore drawings (as built) help to calculate
overall quantities. They need to be verified on-site with survey. On-site survey
focuses on quality of materials. Only on-site reliable information on the qualities
of wood and usage of harmful substances can be gained. These analyses are im-
portant to choose material for recycling without introducing critical substances in
the recycling process. For reuse the classification system has to be improved to the
same level as the reclaimed wood classes regarding the contamination. This could
be difficult, because the way of reuse is unclear and needs further development. In
general conclusion can be drawn that the Alpine cottage is an example for a sim-
ple building with a minimized construction which allows a high level of reuse. A
comprehensive evaluation but also a heterogeneous result is achieved with data
from wide span timber structures. In difference to the historic building the amount
of wood per square meter is significantly lower and the reuse options for similar
and modular parts are different. The existing data in literature about characteristic
span of truss types is mostly correct [13]. The average amount of 0.05 m/m of
wood in these structures which equals about 21 kg/m of softwood. Although there
are losses due to the removal of construction joints and areas with dowels or nails,
there is still a reasonable amount of quality engineered wood which can be
reclaimed for a second use phase.
Cascaded use of wood can affect the LCA positively, as carbon storage is
prolonged. The material choice for wooden products will be influenced more posi-
tively, if the used wooden material is not burned immediately, but used as materi-
als. Therefore, there is a need of a recycling proof and elimination concept for a
future use of reclaimed wood of high quality.
98 A. Hafner, S. Ott, and S. Winter

Acknowledgments. This paper was based on research done in woodwisdom project ECO2,
bachelor thesis of Mathias Woznik, Florian Hofbauer and on-site survey with students in
summer 2012.

[1] Lang, A.: Charakterisierung des Altholzaufkommens in Deutschland. Rechtliche
Rahmenbedingungen Mengenpotenzial Materialkennwerte. Mitteilungen der
Bundesforschungsanstalt fr Forst- und Holzwirtschaft, Kommissionsverlag
Buchhandlung Max Wiedebusch Hamburg, vol. 215 (2004)
[2] Weimar, H.: Empirische Erhebungen im Rohstoffmarkt am Beispiel der neuen
Sektoren Altholz und Grofeuerungsanlagen. Sozialwissenschaftliche Schriften zur
forst- und Holzwirtschaft, vol. 9. Internationaler Verlag der Wissenschaften Frankfurt
am Main (2009)
[3] Mantau, U.: Rohstoffknappheit und Holzmarkt. In: Waldeigentum: Dimensionen und
Perspektiven. Springer, Heidelberg (2010)
[4] Holzzentralblatt, Leinfelden-Echterdingen (November 2012)
[5] Green Paper. On Forest Protection and Information in the EU: Preparing forests for
climate change SEC, 163 final (2010)
[6] Mantau, U., et al.: EUwood - Real potential for changes in growth and use of EU for-
ests. Final report. Hamburg/Germany, 160 p. (June 2010),
[7] Directive 2008/98/EC of the European Parliament and of the Council of 19 November
2008 on waste and repealing certain Directives. Official Journal L 312, 00030030 (2008),
OJ:L:2008:312:0003:01:EN:HTML (July 16, 2012)
[8] Verordnung ber Anforderungen an die Verwertung und Beseitigung von Altholz
Altholzverordnung AltholzV (February 24, 2012)
[9] Ordinance on the Management of Waste Wood, Bundesministerium fr Umwelt (Oc-
tober 25, 2012), http://www.bmu.de/files/pdfs/allgemein/
[10] Hafner, A., et al.: Rohstoffverwertung in Nutzungskaskaden. Holzbau die neue
Quadriga, Wolnzach 3, 4145 (2012)
[11] ECO2-book, http://www.eco2wood.com (forthcoming, Summer 2013)
[12] Hafner, A., et al.: Hllentalangerhtte case study for end of life reuse and recycling
methodologies. In: Proceedings of Sustainable Building Conference sb 2013 Munich,
Munich, pp. 14571466 (2013)
[13] Natterer, J., et al.: Holzbau Atlas Zwei, 1st edn. Birkhuser (2001) ISBN 978-
Advancements for the Structural Application
of Fiber-Reinforced Moulded Wooden Tubes

Jrg Wehsener1, Tom-Egmont Werner2, Jens Hartig1, and Peer Haller1

Technische Universitt Dresden, Institute of Steel and Timber Structures,
01062 Dresden, Germany
STM Montage GmbH, Cossener Strae 2, 09328 Lunzenau, Germany

Abstract. Thermo-hygro-mechanical processes can be used to densify and to form

wood. This is applied to produce wooden tubes of structural size by shaping of
boards, which were densified previously transverse to the grain. Profiles with
hollow cross sections, like tubes, have several advantages compared to those with
compact cross sections. Due to the larger moments of inertia, a higher load-
bearing capacity for bending and buckling with a given amount of material can be
reached. Furthermore, the low thickness of the tube walls allows employing small-
sized tree sections, which are currently used only energetically or for fibre produc-
tion. Thus, production of these tubes might increase the added value of the forests.
In this contribution, a brief overview over the load-bearing behaviour of circular
moulded wooden tubes exposed to axial compression, bending and torsion is
given. Moreover, one of the first practical applications of the tubes, the tower of a
small-sized wind power plant is presented.

Keywords: moulded wooden tubes, densified wood, axial compression, poles,

wind power plant.

1 Introduction

Thermo-hygro-mechanical processes can be used to densify and to form wood

(Sandberg et al. 2013). This ability is applied to produce wooden tubes of struc-
tural size by shaping of plates, which were densified previously transverse to the
grain (Haller 2004, Haller et al. 2006).
The process steps are shown schematically in Fig. 1. At first, planks are densi-
fied up to about 50 % transverse to the grain in a heating press at temperatures of
about 120 C to 140 C. This process takes advantage of the softening of the lig-
nin leading to an almost damage-free plastic deformation of the wood cell walls.
After cooling down to ambient temperature and resetting of the lignin (< 80 C),
the densified planks are cut into slats with a thickness according to the desired

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 99
DOI: 10.1007/978-94-007-7811-5_9, RILEM 2014
100 J. Wehsener et al.

wall thickness of the tube. The slats are then glued to boards with the densification
direction of the wood oriented parallel to the plane direction. Subsequently, the
boards are steam-treated with water steam temperatures of about 100 C resulting
in plastic deformability of the wood due to re-softening of the lignin. Thus, a po-
tential recovery strain of about 100 % can be reached allowing for reshaping the
plane board to a tube. Therefore, suitable moulds and compression facilities are
necessary. After cooling down and adhering the open joint, the tube is ready for
application. In a further optional processing step, a reinforcement layer, e. g.
of fibre reinforced polymers, can be applied on the outer surface of the tube for
increasing durability, load-bearing capacity and ductility.

Fig. 1 Processes for producing moulded wooden tubes (Haller et al. 2013)

The tubes have several advantages compared to structural elements with com-
pact cross sections. Due to the larger moments of inertia, a higher load-bearing
capacity for bending and buckling with a given amount of material can be reached.
Furthermore, the low thickness of the tube walls allows employing small-
sized tree sections, which are currently used only energetically or for fibre
production. Thus, production of these tubes might increase the added value of the
Since the first ideas for producing moulded wooden tubes, about ten years ago
(Haller 2004, 2007), a considerable number of investigations was carried out to
develop applicable processes, at first, and to determine the load-bearing behaviour
and ecologic impact, afterwards. The investigations regarding the load-bearing
behaviour are briefly summarised in this contribution. More detailed information
on the ecologic impact of moulded wood compared to other building materials can
be found in (Manthey et al. 2010, Haller et al. 2011).
Moulded wooden tubes have various fields of application, for instance struc-
tural elements like columns and transport elements like industrial water lines
(Putzger et al. 2012, Haller and Nendel 2013). A first application, which was re-
cently taken in service, is the pole of a wind power plant with a height of 9 m. The
respective construction is described in here.
Advancements for the Structural Application 101

2 Performance of Moulded Wooden Tubes

2.1 Load-Bearing Behaviour

The load-bearing behaviour of moulded wooden tubes was investigated for differ-
ent types of loading with and without additional fibre reinforcement on the outer
surface. A quite detailed summary can be found in (Haller et al. 2011, 2013).
Thus, in this contribution only a very condensed overview is given.
Each of the tested specimens had a diameter of about 27 cm and a wall thick-
ness of about 2 cm. The specimens were made of spruce. Tests were performed
regarding axial compression, torsion and bending. Except for the case of axial
compression, additional fibre reinforcement on the outer surface was crucial to
compensate the low tensile strength transverse to the grain and shear strength typi-
cal for wood and to obtain a reasonable load-bearing capacity for building pur-
poses. The fibre reinforcement comprising of glass or carbon fibres and polymeric
resins was applied by a winding technique.

2.1.1 Axial Compression

The most extensive investigations (Heiduschke and Haller 2009, 2010, Haller et
al. 2011, 2013) were performed concerning compression in axial direction, which
is the most important case of loading for column-type applications. The tested
columns had a length of 2.5 m. The mean ultimate load of the unreinforced col-
umns (8 samples) was 564 kN, which corresponds to a ultimate compressive stress
of 42 N/mm and is close to the nominal compressive strength of spruce according
to DIN 68364 (fc=45 N/mm). The material behaviour is linear elastic. The failure
occurs in a brittle manner due to exceeding the tensile strength transverse to the
grain by the circumferential stresses leading to longitudinal splitting of the tube
With an outer fibre reinforcement, the low tensile strength of the wood trans-
verse to the grain can be effectively compensated. Thus, the reinforcement serves
as a confinement. With the confining fibres, increases of more than 50 % of the
ultimate load and more ductile failure modes were observed. Only a relatively low
amount of reinforcement in the magnitude of a few hundred g/m of glass or car-
bon fibres are necessary to achieve the increase in the load-bearing capacity. The
fibre orientation played a less important role concerning the performance as long
as the confining function was ensured.

2.1.2 Torsion

Wood has a relatively low resistance against torsion loading because of its low
shear strength. However, it is known that a fibre reinforcement with +/-45 % ori-
entation of the fibres can effectively increase the respective load-bearing capacity.
102 J. Wehsener et al.

As an example, a tube was reinforced with 4 layers of carbon fibre reinforcement

with inclination angles of 7/-7/45/-45 (Haller et al. 2011, 2013). The amount
of fibres in each layer was 300 N/mm. The compound tube resisted a torsion
moment of about 34 kNm and failed at a principal strain of about 4 , which
corresponds to a maximum twist of 3.5/m, due to shear failure.

2.1.3 Bending

It is known that ring shaped profiles are poorly suited for bending loads. Neverthe-
less, bending almost always occur in structural applications. Consequently, a car-
bon fibre reinforced tube was exposed to a four point bending test (Haller et al.
2011, 2013) to determine the load-bearing behaviour. Two layers of carbon fibres
oriented in +/-45 direction and a fibre amount of 300 N/mm in each layer were
applied on the surface of the tube. The tube consisted of two pieces, which were
connected at midspan by finger jointing. The span of the beam was 4.5 m. The
loading positions had a distance of 1 m.
The ultimate load was reached at 63 kN. While linear elastic behaviour was ob-
served up to a load of 60 kN, gradual degradation occurred while increasing the
load until failure, which resulted from local buckling at the position of load appli-
cation. This led to a delamination of the reinforcement. However, a rupture of the
reinforcement in the tension zone was not observed after exceeding the ultimate

3 Application Pole of a Wind Power Plant

Moulded wooden tubes were applied to construct a small-sized wind power plant,
which is in service since January 2013, cf. Fig. 2. The pole of the wind
power plant, which has a height of 9 m, is constructed as a bundle of three tubes.
Each tube consists of six smaller tubes of 1.5 m length due to the limitations of the
production facility. At the joints of the tubes ring-shaped internal wooden cores
with a thickness of 3 cm and a length of 40 cm are arranged, cf. Fig. 3a,c. More-
over, each two tubes are reinforced with a wounded reinforcement of polyester
resin impregnated glass fibres with a fibre weight per unit area of about
1000 g/m. The wounding was performed with a woven fabric and a fibre orienta-
tion of 0/90 related to the longitudinal axis of the tube. For improving the
surface and the durability, a glass fibre fleece and a weather-resistant topcoat were
These 3 m long tube pieces are connected by steel brackets at heights of 3 m
and 6 m, which also connects the tube bundle, cf. Fig. 3d. The fastening between
the steel and wood parts is performed by pre-drilled commercially available
screws (diameter 5 mm).
Advancements for the Structural Application 103

Fig. 2 Wind power plant with moulded wooden tubes in service

On top, a commercially available Darrieus wind turbine is arranged, which has

a power of 1 kW. It is attached to the tubes with another steel part connecting also
the tubes at this position. For fixation of the cantilever type construction at the
base, a sleeve foundation of steel is used, which is again attached to the tubes by
means of screws, cf. Fig. 3b,c.
Moreover, for reasons of redundancy and reducing lateral deflections at the
top of the tower, steel rods are attached to the steel brackets in longitudinal
direction of the tower to participate in bearing tensile loads in the case of bending,
cf. Fig 2.
104 J. Wehsener et al.

a) b)

c) d)

Fig. 3 Construction details of the pole of the wind power plant a) Connected tubes (length
3m) before application of fibre reinforcement; b) Detail base construction; c) View into the
hollow tube (thickness 2 cm) with stiffening ring (thickness 3 cm); d) Detail steel clamp to
connect the 3m tube pieces

4 Static Verification

4.1 Material Properties

The static verification of the column consisting of three wooden tubes, which are
connected in a pointwise manner by steel clamps is a challenging task. One of the
Advancements for the Structural Application 105

uncertainties arise from the mechanical properties of the densified and subse-
quently recovered wood. It is not possible to rely on standards or design codes
because such does simply not exist for densified and recovered wood, hitherto.
Thus, the mechanical properties have to be estimated based on available ex-
perimental results. It is known that densified wood has increased stiffness and
strength compared to the base material because the void space given be the cell
lumen is reduced (Vorreiter 1949, Kollmann 1955, Haller & Wehsener 2004).
Therefore, stiffness and strength increase towards respective values of the cell
walls, which are the upper limits that can be reached. However, it is also known
that the brittleness increases because of a beginning thermal degradation (Vorre-
iter 1949, Kollmann 1955, Haller & Wehsener 2004).
During recovery of the densification the void space increases again and, conse-
quently, also stiffness and strength of the wood reduce. Because the crimping of
the cell walls and thermal transformations will result in a certain degradation, it
has to be assumed that stiffness and strength values will be reduced compared to
the initial material before densification (Wienhaus 1999). However, the experi-
mental investigations regarding the axial compression of the tubes, mentioned
previously, revealed that the reduction of strength is only low.
For the static verifications, it might be a convenient approach to use material
values provided by Eurocode 5 (DIN EN 1995) and choose a less grading class.

4.2 Static System

Another challenge arises from the construction. It is difficult to estimate the actual
shear stiffness between the tubes provided by the steel pieces attached by screws.
For reasons of simplicity, it is assumed for the static verification that the steel
brackets do not apply shear rigidity but only distribute the loads uniformly to the
three tubes. Thus, an increase of the moment of inertia is not taken into account
due to the shear connection but the tubes act like three parallel cantilevers with
equal loading.
The steel rods, which are attached to bear some of the tensile loads in case of
bending, are also neglected for reason of simplicity and are only considered as a
means for increasing redundancy of the structural system.
Moreover, it is assumed that tensile stresses at the tube joints are carried by the
FRP layers or the steel brackets.

4.3 Static Verifications

Following these assumptions, static verifications were performed. Besides the
dead loads of the wooden tubes including reinforcement layer and stiffening rings
(3.6 kN), the steel fasteners (0.75 kN) and the rotor (0.55 kN), wind loads were
taken into account. The wind loads were estimated with 0.8 kN/m according to
Eurocode 1-4 (DIN EN 1991-1-4). According to the datasheet, a maximum
bending moment of the rotor at the attaching position to the tubes of 1.2 kNm was
106 J. Wehsener et al.

considered. The maximum torsion moment of 90 Nm caused by the rotor accord-

ing to its datasheet is neglected.
Following the assumption of equal load sharing between the tubes, each tube is,
thus, loaded at the fixing position at the base with a normal force of 2.1 kN, a
transverse force of 2.2 kN and a bending moment of 3.8 kNm. These are already
design values considering a partial safety factor of 1.35 for the dead loads and 1.5
for the wind loads according to Eurocode 5.
For estimating the material design properties, it was assumed that the moulded
wood has at least properties according GL 24h according to DIN EN 1194. This is
a conservative approach, because the applied wood was free of knots. The design
compressive/bending and shear strengths are, thus, fc,0,g,d = fm,g,d = 11.1 N/mm and
fv,g,d = 1.2 N/mm, respectively.
The verifications in the ultimate limit state were performed according to Euro-
code 5. The verification against failure of the wood at the base due to combined
bending and compression (Eurocode 5 Section 6.2.4) led to a ratio between al-
lowable and actual stresses of about 0.3, which is considerably smaller than 1 and
a conservative result. For shear failure of the wood at the base (Eurocode 5 Sec-
tion 6.1.7) the respective ratio has a value of about 0.15. The ratio of allowable
and actual stresses for the case of buckling due to bending and compression ac-
cording to (Eurocode 5 Section 6.3.2) is about 0.4.
As already mentioned, each tube consists of 6 tube pieces, which are connected
either by the fibre reinforced composite on the surface or by the steel clamps and
predrilled screws. Both connections have to be verified. It is sufficient to verify
that the appearing tensile stresses are transferred by the connections. The com-
pressive stresses are assumed to be transferred by contact pressure at the tube
cross sections. The screw connections were designed corresponding to the rec-
ommendations in (Eurocode 5 Section 8.7.1). The screw distance in circumfer-
ential direction is 4 cm. At the base, six and at the clamps in the middle of the
tubes three staggered rows of screws with a distance 1.8 cm between each row are
arranged, cf. Figs. 3b,d.
The amount of glass fibres in the composite was determined in an estimating
manner by experience such that the composite will not fail at the joints due to the
mentioned types of loading.
The verification of the deformations in the serviceability limit state is difficult
to perform in an exact manner due to the combination of different materials and
the compliance of the connections. According to (Eurocode 5 Section 7.2), the
tolerable deflections for cantilevers are for the final state (t=) in the range of
L/175 = 4.0 cm up to L/125 = 5.6 cm. Neglecting steel connections, stiffening
cores, fibre reinforcement and compliance of the fixation at the base and assuming
a mean modulus of elasticity in the final state of E0,mean,inf = 7250 N/mm (E0,mean =
11600 N/mm), the calculated deflection at the top of the cantilever is about 7 cm.
This is only about 50 % more than the allowed value. Taking into account the
fibre reinforcement, the shear connections between the tubes and the additional
steel rods will supposedly reduce the occurring deflections to the allowed value.
Advancements for the Structural Application 107

5 Summary and Conclusions

In this contribution, the concept of producing moulded wood profiles was briefly
reviewed and recent advancements regarding the determination of the load-
bearing behaviour as well as one of the first practical applications were addressed.
The completion of the aforementioned first wind power plant is an important step
towards placing moulded wooden tubes on the market of structural elements. An
important challenge for the future is the production of longer moulded wooden
tubes in order to reduce the effort for jointing. Moreover, it would be advanta-
geous to have tubes with larger diameters and wall thicknesses to increase the size
of the wind power plant. For a broader application of moulded wood profiles in
the building industry, a standardisation will be necessary.

Acknowledgements. The authors gratefully acknowledge the financial support from Ger-
man Federal Ministry of Economics and Technology and AIF Projekt GmbH for ZIM re-
search project "Pressholzhohlprofil fr Windkraftanlagen" under grant numbers
KF2132401WZ8 und KF2132601WZ8. Furthermore, the authors also want to thank Mr.
Thomas Hndel for producing the tubes.

DIN 68364:2003-05, Properties of wood species - Density, modulus of elasticity and
strength. Deutsches Institut fr Normung, Berlin (2003)
DIN EN 1194:1999-05, Timber structures Glued laminated timber Strength classes and
determination of characteristic values. Deutsches Institut fr Normung, Berlin (1999)
DIN EN 1995:2010-12, Eurocode 5: Design of timber structures. Deutsches Institut fr
Normung, Berlin (2010)
DIN EN 1991-1-4:2005-07, Eurocode 1: Actions on structures - Part 1-4: General actions,
Wind actions. Deutsches Institut fr Normung, Berlin (2005)
DIN EN 1995:2010-12, Eurocode 5: Design of timber structures. Deutsches Institut fr
Normung, Berlin (2010)
Haller, P.: Vom Baum zum Bau oder die Quadratur des Kreises. Wissenschaftliche
Zeitschrift der Technischen Universitt Dresden 53(1-2), 100104 (2004)
Haller, P.: Concepts for textile reinforcements for timber structures. Mater. Struct. 40(1),
107118 (2007), doi:10.1617/s11527-006-9153-5
Haller, P., Nendel, K.: Holz von der Rolle - Formholzrohre leiten heie Solen unter Druck.
Deutsches Ingenieurblatt 2013(5), 2831 (2013)
Haller, P., Wehsener, J.: Festigkeitsuntersuchungen an Fichtenpressholz (FPH). Eur. J.
Wood Wood Prod. 62(6), 452454 (2004), doi:10.1007/s00107-004-0516-8
Haller, P., et al.: Hochleistungsholztragwerke HHT Entwicklung von hochbelastbaren
Verbundbauweisen im Holzbau mit faserverstrkten Kunststoffen, technischen Textilien
und Formpressholz: Abschlussbericht zum BMBF-Forschungsvorhaben 0330722A-C.
Technische Universitt Dresden, Dresden (2011)
Haller, P., Putzger, R., Wehsener, J., Hartig, J.: Formholzrohre Stand der Forschung und
Anwendungen. Bautechnik 90(1), 3441 (2013), doi:10.1002/bate.201200065
108 J. Wehsener et al.

Haller, P., Wehsener, J., Ziegler, S.: Wood Profile and Method for the Production of the
Same. European Patent 1390181 B1 (2006)
Heiduschke, A., Haller, P.: Zum Tragverhalten gewickelter Formholzrohre unter axialem
Druck. Bauingenieur 84(6), 262269 (2009)
Heiduschke, A., Haller, P.: Fiber-Reinforced Plastic-Confined Wood Profiles Under Axial
Compression. Struct. Eng. Int. 20(3), 246253 (2010), doi:10.2749/101686610792016772
Kollmann, F.: Technologie des Holzes und der Holzwerkstoffe, vol. 2. Springer, Berlin
Manthey, C., Guenther, E., Heiduschke, A., Haller, P., Heistermann, T., Veljkovic, M.,
Hajek, P.: Structural, economic and environmental performance of fibre reinforced
wood profiles vs. solutions made of steel and concrete. In: Braganca, L., et al. (eds.)
Sustainability of Constructions - Integrated Approach to Life-time Structural Engineer-
ing. European Science Foundation - COST Action C25, Timisoara, pp. 275289 (2009)
Putzger, R., Eckardt, R., Eichhorn, S., Nendel, K., Haller, P.: Investigation on fibre rein-
forced moulded wood pipes for the conduction of brine at high temperature and pres-
sure. In: COST Action FP0904 Workshop, Book of Abstracts Current and Future
Trends of Thermo-Hydro-Mechanical Modification of Wood Opportunities for New
Markets, pp. 174176. Universit de Lorraine, Nancy (2012)
Sandberg, D., Haller, P., Navi, P.: Thermo-hydro and thermo-hydro-mechanical wood
processing: an opportunity for future environmentally friendly wood products. Wood
Mater. Sci. Eng. 8(1), 125 (2013), doi:10.1080/17480272.2012.751935
Vorreiter, L.: Holztechnologisches Handbuch, vol. 2. Verlag Georg Fromme, Wien (1949)
Wienhaus, O.: Modifizierung des Holzes durch eine milde Pyrolyse abgeleitet aus den
allgemeinen Prinzipien der Thermolyse des Holzes. Wissenschaftliche Zeitschrift der
Technischen Universitt Dresden 48(2), 1722 (1999)
Sole Plate Fixing Details for Modern Methods
of Timber Construction

Jesus M. Menendez, Kenneth Leitch, and Robert Hairstans

Institute for Sustainable Construction, Edinburgh Napier University, Edinburgh, UK


Abstract. In order to resist lateral loads, modern methods of timber construction

are reliant on the in-plane shear strength of the walls orientated parallel to the
applied action. In closed panel systems, the shear stresses are transferred to the
foundations by the sole plate through the sheathing board, which is usually me-
chanically jointed to the timber frame. Since closed panels are delivered to site as
single units, access to the internal bottom rail is rather restricted and novel effi-
cient solutions to secure the panel to the substrate are required. Sole plate fixing
components for open and closed panel systems were tested in isolation and com-
bination in order to validate a simplistic version of the weakest link theory. As a
result, findings were embedded into a software database with a direct link to a
previously developed sole plate and racking design application. This integrated
process facilitates the structural optimization of the sole plate detail.

Keywords: timber to concrete, shear wall, racking, closed panel system, MMC.

1 Introduction

The housing sector in the United Kingdom (UK), regardless of the current finan-
cial scenario, is experiencing an upward trend in housing prices mainly caused by
the long term imbalance between housing demand and number of homes built [1].
In order to mitigate this trend, the UK government has encouraged the use of
Modern Methods of Construction (MMC) to produce a higher quantity and quality
of houses [2] with a clear strategy for efficiency and elimination of waste [3].
Timber presents ideal properties to be manufactured offsite under lean princi-
ples and with a considerable high level of finishing detail. Furthermore, timber
frame closed panel systems also benefit from excellent carbon footprint, low
thermal conductivity, high strength-to-weight ratio and ease of construction. As a
result, offsite timber frame closed panel systems can be considered as a cost-
effective MMC for low energy buildings [4].
These panels are used in general as load bearing shear walls, roof and floor cas-
settes for low-rise platform construction. Nevertheless, closed panel shear walls,
contrary to open panel construction, must deal with a relatively complex connec-
tion system to the foundations or substrate. This collection of connections is

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 109
DOI: 10.1007/978-94-007-7811-5_10, RILEM 2014
110 J.M. Menendez, K. Leitch, and R. Hairstans

commonly referred as the sole plate base fixing detail (SPFD). On site access to
the SPFD is rather difficult in prefabricated closed panel units. The solution to this
problem is often found in the design of structurally convoluted sole plate geome-
tries and the inclusion of additional shear planes. As a consequence, the SPFD for
closed timber frame wall panels presents a potential decrease of strength and stiff-
ness in comparison with standard open panel construction. Furthermore, additional
design caution must be taken to comply with the minimum end and edge connec-
tion distances, particularly, if slanted fasteners are used.
This study describes the research work carried out by the Centre for Offsite
Construction and Innovative Structures (COCIS) at Edinburgh Napier University
on the optimisation of sole plate base fixing designs for closed panel timber frame
walls. The study comprises an empirical sole plate test procedure, an analytical
design methodology and the development of structural software applications for
SPFD and shear walls.

2 Experimental Procedure

This section describes a series of laboratory tests undertaken to determine the

structural behaviour of the sole plate components in isolation as a methodology to
assess its combined performance when using in service. The findings of the ex-
periment provide a dual purpose in this research study; assessing the analytical
method, further explained in section 3 and; as a validated direct input for design
specification in a software application via a database as detailed in section 4.

2.1 Sole Plate Base Fixing Components

Several materials and fasteners form this connection. Fig. 1 illustrates the different
sole plate components for open and closed panel standard construction types. The
standard connection between the bottom rail and the stud is disregarded as it is
considered part of the timber frame structure [5]. The SPFD for the closed panel
system includes a plywood packer to provide a level surface for the bottom rail to
be secured.
In order to investigate the performance of the connections shown in Fig. 1, the
following sole plate sub-connections were identified and isolated:

a) c) Type: Timber to concrete:

The study of this connection for both open and closed panel systems in-
cludes 7.5x100mm Express nail type fastener. Substrate material is dense
aggregate block of 7N/mm2 compressive strength (DAB07).
b) Type: Timber to timber:
Apart from joining the wall framing members, this connection is also
found at the base of the open panel and it is critical in terms of transfer-
ring the racking forces from the wall to the foundation. Fasteners tested
include 3.0x90mm smooth wire nail.
Sole Plate Fixing Details forr Modern Methods of Timber Construction 1111

wood to timber:
d) Type: Timber to plyw
This non-standarrd double shear connection comprises of 45x70mm tim m-
ber batten to the sole plate packer through 18mm plywood bby
4.4x115mm self--tapping screw.
e) Type: Timber to OSBB to timber:
Again, this non-sstandard double shear connection horizontally secures thhe
interlock timber to the closed panel rail by 4.4x115mm self tappinng

Fig. 1 Detail of the two sole plate details studied

2.2 Test Methodolo

Table 1 summarises the relevant
r European Standards and bespoke methods useed
in this experiment. Note that
t for fastener determination of yield moment, observeed
deformations did not com mply with BS EN 409:1993 recommendations [6]. As a
result, the double plastic hinge deformation model developed by Coste [7], annd
based on the work of Jorrissen and Blass [8], was adopted instead. Furthermorre,
embedment tests resulted d in bending of the fastener. BS EN 383 standard invalli-
dates embedment test resu ults if bending of the fastener occurs [9].
The current standard foor the design of racking walls in the UK, PD6693-1 [100],
allows the SPFD for restrraint provision against overturning moments. Thereforre,
fastener withdrawal tests were undertaken in order to assess its mechanical
112 J.M. Menendez, K. Leitch, and R. Hairstans

Table 1 Description of the tests undertaken in this study

Test Method Equipment Applied load

Lateral load capacity BS EN1380 100kN SCHENK BS EN26891
Tensile strength BS EN ISO898-1 30kN Lloyd R30k 1mm/min
Yield moment BS EN409 30kN Lloyd R30k 1mm/min
Embedment BS EN383 30kN Lloyd R30k BS EN26891
Pull through BS EN1383 100kN SCHENK 2mm/min
Withdrawal BS EN1382 100kN SCHENK 2mm/min

The overall strength performance of the SPFD is determined by the capacity of

the weakest sub-connection as a revised model of the weakest link theory. On the
other hand, the overall stiffness of the SPFD is defined by the accumulative dis-
placement occurred at each shear plane [5].

2.3 Test Results

The results of the isolated connection tests are given in Table 2. Due to the ductile
nature of the connections, the ultimate strength is based upon the measured force
resistance at 15mm of displacement.

Table 2 Strength and stiffness results of the isolated sole plate components

Connection Ultimate strength Slip modulus(1)

fmax (N) Kser (N/mm)
a) c) 4087 1594
b) 3159 1500(2)
d) 5727 1297
e) 11470 1542
Slip modulus taken as linear stiffness between 0.1-0.4Fmax
Slip modulus in accordance with BS EN 26891

Fig. 2 shows the connection types a) b) and type d) being tested with their asso-
ciated load slip curve results.
In addition to the lateral shear resistance for each connection type, the experi-
ment also includes testing of fasteners for tensile strength, yield moment and axial
withdrawal capacity. On the other hand, nominal wire and root diameter were
considered to analytically determine the mechanical properties of nails and screws
respectively. Test results, as mean values, are shown in Table 3.
Sole Plate Fixing Details for Modern Methods of Timber Construction 113

a) b)


Fig. 2 Test execution and displacement results of isolated sole plate components

Table 3 Mechanical properties for sole plate fasteners

Fastener type Tensile Yield moment With- Pull-

strength My (Nmm) drawal through
fu (N/mm2) strength(1) strength(1)
(N/mm2) (N/mm2)
3.0x90mm 894 5126 6.4 50.1
4.4x115mm 1079 21562 18.2 30.5
7.5x100mm 1290 44845 3.58(2) 50.0
Values given refer to the performance of the fastener when installed in C16
Withdrawal strength given when installed in DAB 7N

In order to confirm the performance of the SPFD in isolation with the overall
performance of the complete detail, full sole plate fixing open and closed panel
details were also tested. These tests were performed according to a heavily modi-
fied version of the BS EN 1380 test set up so as to replicate the shear load being
transferred from the wall panel to the substrate [5]. Fig. 3 illustrates the SPFD test
set up and the load slip curves for the open (OP) and closed panel (CP) systems.
114 J.M. Menendez, K. Leitch, and R. Hairstans



Fig. 3 Test execution and displacement results for full open and closed panel sole plate

Furthermore, Table 4 presents the strength and stiffness test results of the full
SPFD. The maximum strength value, fmax, is determined when 15mm displacement
of the bottom rail relative to the substrate occurs. Similarly, stiffness, Kser, is based
upon the displacement of the bottom rail at 40% of its ultimate strength.

Table 4 Strength and stiffness results for both open and closed panel SPFD

Sole plate Ultimate strength Slip modulus

fmax (N) Kser (N/mm)
OP 2841 1364
CP 4036 745

2.4 Sole Plate Overall Performance

Test results for the open and closed panel SPFD contained in Table 2 and Table 4
corroborate that the stiffness of the sole plate connection equals the sum of the
displacements of each individual connection [5].
On the other hand, the overall strength resistance of the SPFD is dictated by the
weakest connection with no provision for load sharing enhancing factors. Table 5
compares the test results of the connection detail in combination with the isolated
components. Hence, the proposed isolation-combination methodology is validated.
Sole Plate Fixing Details for Modern Methods of Timber Construction 115

Table 5 Isolation and combination test results for open and closed panel sole plate details

Connection Isolation (N) Combination Ratio

a) b) c) d) e) (N)
OP 4087 3159 2841 0.90
CP 4087 5727 11470 4036 0.99

This validation allows for the optimisation of the SPFD in MMC by designing
effective shear planes of similar strength. However, the spacing of the fasteners
for each component along the runner also influences the design as the strength
capacity of the sole plate is given in load per meter run (N/mm).

3 Design Tool for Optimisation

In order to achieve the advantages of timber as structural material for MMC, inno-
vative products and processes need to be further exploited by the architectural,
engineering and construction (AEC) sector. Osterrieder and Richter [11] con-
cluded that the timber industry should be provided with viable design aided tools
in order to facilitate the knowledge transfer from research to the final market and
that these tools should enable data sharing instead of data exchange.
COCIS, in collaboration with industrial partners, is developing a platform to
streamline the release of research findings on innovative products and methodolo-
gies to the AEC industry [12]. The software for structural engineering Tedds, from
CSC (UK) Ltd, is used to both develop an application to design timber frame
walls and to populate a database with the SPFD test results for direct use in the

Fig. 4 Tedds racking application optimisation flow chart

116 J.M. Menendez, K. Leitch, and R. Hairstanns

Tedds application. The Tedds

T designing optimisation procedure for the rackinng
application is shown in Fiig. 4.
The development of prrofessional Tedds applications is driven by the sequentiial
completion of a series of pre-established tasks. Firstly, the user interface is deveel-
oped in C++ through the add-on application Tedds Interface Designer. Alongsidde
with the interface, detailed active drawing sketches of the wall panel, design notees
and references are produ uced. Then, according to the relevant code of practicce,
restrictive parameter valuues are defined in the interface so the design falls withiin
the scope and assumption ns of the method. Fig. 5 illustrates a snapshot of the rackk-
ing application interface.

Fig. 5 User interface for Ted

dds Racking App

Once the interface is operative, a flow chart with logical expressions of the caal-
culation methodology is programed
p in C++ through another auxiliary applicationn:
Tedds Calc Designer (Fiig. 6). This process will create the output report of thhe
calculation in a word padd fashion. The user can select a summary or a full outpuut
In the meantime, the results
r achieved in the SPFD tests are transferred into a
Tedds database by the app plication Tedds DataList Designer. Variables are defineed
to describe the geometry, properties and components of the SPFD. The name annd
units of these variables are consistent with the Tedds applications in order to
ensure a correct informatiion data sharing.
Sole Plate Fixing Details for Modern Methods of Timber Construction 117

After the application is completed and tested, developmental files are packed
up into one single executable file and distributed to the Tedds users who will in-
stall the new calculation directly on the standard Tedds program.

Fig. 6 Calculation flow chart for Tedds Racking App

The user begins the calculation by introducing the design input parameters. In-
stantaneously, the outcome of these parameters is displayed in the interface. As a
result, the user can accordingly adjust relevant variables, such as member density,
member thickness or spacing of fasteners, in order to optimise the overall strength
and stiffness of the SPFD and the racking wall panel. Once the user accepts the
final design, the software will report the calculation ready for verification and

4 Discussion and Conclusions

In comparison to on-site open panel construction, sole plate fixing details in tim-
ber MMC systems, such as racking wall panels, are critical as accessibility is
restrained resulting in the introduction of more and complex shear planes in the
118 J.M. Menendez, K. Leitch, and R. Hairstans

A series of laboratory tests for two different SPFD, open and closed panel, were
carried out in order to verify the hypothesis of the isolation combination meth-
odology where the overall sole plate resistance is given by the strength of the
weakest component. The stiffness of the sole plate fixing results in the addition of
the individual displacement of the components which is directly proportional to
the size of the fastener and the density of members. As a result, the research find-
ings can facilitate the optimisation of the sole plate fixing detail by enhancing the
limiting structural parameters of the critical component part.
Furthermore, findings on structural timber research can rapidly be disseminated
by means of structural software applications in Tedds (CSC UK Ltd). Tedds has a
potential outreach of more than 15,000 engineers in the UK. Moreover, the output
of this research is released in the Tedds Racking Design application via direct
database input. The final application is installed in the standard Tedds library of
calculations after being beta tested.

Acknowledgments. The authors would like to appreciate the financial support provided by
the Engineering and Physical Science Research Council (EPSRC). Software licenses and
technical support offered by CSC (UK) Ltd is also acknowledged.

[1] Barker, K.: Review of Housing supply - Delivering stability: securing our future
future housing needs. HMSO, London (2004) ISBN 1-84532-010-7
[2] Egan, J.: Rethinking construction. Department of the Environment, Transport and
the Regions, London (1998)
[3] Office C. Government Construction Strategy. Cabinet Office, London (2011)
[4] Hairstans, R.: Off-Site and Modern Methods of Timber Construction: A Sustainable
Approach. TRADA Technology Limited (2010)
[5] Leitch K. The development of a hybrid racking panel. Edinburgh Napier University,
Edinburgh (2013)
[6] BSI. BS EN 409:1993. Timber structures - Test methods - Determination of the
yield moment of dowel type fasteners - Nails. BSI (1993)
[7] Coste, G.: The assessment and applications of a new connector type for use in struc-
tural timber systems. In: University EN, ed. United Kingdom (2010)
[8] Jorissen, A.J.M., Blass, H.J.: The fastener yield strength in bending. Savonlinna
[9] BSI. BS EN 383:2007. Timber Structures - Test methods - Determination of em-
bedment strength and foundation values for dowel type fasteners. British Standard
Institution, London (2007)
[10] BSI. PD6693-1. Complementary Information to Eurocode 5 Design of timber struc-
tures Part 1 General Common rules and rules for buildings, p. 66. British Standard
Institution, London (2012)
[11] Osterrieder, P., Richter, S., Fischer, M.: A product data model for design and fabri-
cation of timber buildings. In: WCTE 2004 Proceedings, Lahti (2004)
[12] Menendez, J., Hairstans, R., Leitch, K., Turnbull, D.: A Structural Engineering Plat-
form for Timber Modern Methods of Construction. In: International Conference on
Innovation in Architecture, Engineering and Construction, Sao Paulo, August 14-18
Thin-Walled Timber Structures

Benoit P. Gilbert1, Steven B. Hancock1, and Henri Bailleres2

Griffith School of Engineering, Griffith University, Australia
Salisbury Research Centre, Department of Agriculture, Fisheries and Forestry,
Queensland Government, Australia

Abstract. Due to their efficiency, lightweight, ease of erection and low cost, steel
and aluminium thin-walled structures have become very popular in the construc-
tion industry over the past few decades. Applications include roof and wall sys-
tems (purlins and girts), storage racks, and composite concrete and steel slabs. The
effectiveness of these structures lies in the cross-sectional shape of the profiles
which enhances their strength by controlling the three fundamental buckling
modes: local, distortional, and global. However, despite the attractiveness of these
structures, steel and aluminium are greenhouse gas intensive materials and do not
produce sustainable structural products. This paper presents an investigation per-
formed at the Griffith School of Engineering, Griffith University, which shows
manufacturing these types of profiles in timber is possible. Short composite thin-
walled timber Cee-sections (500 mm long) were fabricated by gluing together thin
softwood (Araucaria cunninghamii) veneers (1 mm thick). Two types of Cee-
sections were considered, one with a web stiffener to increase the local buckling
capacity of the profile and one without. The profiles were tested in compression
and the test results are presented and discussed in the paper in terms of structural
behaviour and performance. Further research directions are proposed in order to
provide efficient and lightweight sustainable structural products to the timber

Keywords: thin walled timber C-sections, veneers, axial compression, column

tests, post-buckling behaviour, Araucaria Cunninghamii.

1 Introduction

Cold-formed steel and aluminium thin-walled profiles are commonly used in the
construction industry due to their efficiency, lightweight, ease of erection and low
cost [1]. The worldwide market for these products is significant and estimated at
more than two billions dollars. Typical industrial and civil engineering applica-
tions include roof and wall systems (purlins and girts), steel storage racks and
composite concrete and steel slabs. Typical cold-formed steel profiles are shown
in Figure 1.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 119
DOI: 10.1007/978-94-007-7811-5_11, RILEM 2014
120 B.P. Gilbert, S.B. Hancock, and H. Bailleres

Due to the nature of the manufacturing process, consisting of bending a thin

sheet of metal to a desired cross-sectional shape, open profiles such as Cee- or
Zee-cross-sections are generally used. Three fundamental buckling modes,
referred to as local, distortional and global (see Figure 2), can occur for these pro-
files and usually govern the compressive and bending strength. The section
capacity for local buckling is typical enhanced by adding intermediate stiffeners to
walls having significantly large width-to-thickness ratios (see Figure 1), while the
section capacity for distortional buckling can be enhanced by changing the cross-
sectional shape, mainly by adding lip stiffeners.

Fig. 1 Examples of cold-formed steel profiles, with and without intermediate stiffeners

(a) (b) (c)

Fig. 2 Fundamental buckling modes for thin-walled open cross-sections in compression, (a)
local, (b) distortional and (c) global

However, despite the attractiveness of these types of profiles, steel and alumin-
ium are greenhouse gas intensive materials and do not produce sustainable struc-
tural products [2]. Manufacturing profiles similar to the ones shown in Figure 1 in
timber is possible, but to the authors best knowledge, has not yet been investigat-
ed in the published literature. This paper presents the first step of a study aiming at
developing extra light and structurally sound timber profiles for applications in-
cluding amongst others (i) emergency shelters which can be rapidly assembled
and disassembled, (ii) purlins for major timber buildings, as a substitute to steel
purlins currently used and (iii) wall stubs.
Thin-Walled Timber Structures 121

Short composite thin-walled timber Cee-sections (500 mm long) were fabri-

cated by gluing together thin softwood (Hoop pine - Araucaria cunninghamii)
veneers. Two types of Cee-sections were considered, namely one with a web stiff-
ener to increase the local buckling capacity of the profile and one without, and this
paper introduces the steps involved in the manufacturing process. The sections
were tested in compression and the test results are presented and discussed in the
paper in terms of structural behaviour and performance. Further research direc-
tions are also proposed.

2 Investigated Sections

2.1 General
Three sets of two 210 mm deep 105 mm wide Cee-sections (with 50 mm lip
stiffeners) were manufactured and investigated. The two cross-sections in a set are
of different type. Type A cross-section had no intermediate stiffeners, as shown in
Figure 3 (a), and the cross-section was designed so local buckling of the web
would govern the strength of the profile. On the contrary, a web stiffener was
added to the second type of cross-section (Type B), as shown in Figure 3 (b), so
the local buckling capacity of the cross-section would theoretically be higher than
the maximum compressive strength of the material. The nominal section proper-
ties of the cross-sections are given in Table 1.

(a) (b)

Fig. 3 Cee-sections manufactured and tested, (a) Type A and (b) Type B

Each cross-section was composed of 5 layers of nominal 1 mm thick Hoop pine

rotary sliced veneers. The grain of the three inner layers was orientated along the
member longitudinal axis, while the grain of the two outer layers was perpendicu-
lar to the inner layers and orientated in the member transverse direction (see insert
122 B.P. Gilbert, S.B. Hancock, and H. Bailleres

in Figure 3). In this configuration, the three inner layers would mainly resist the
compressive load while also providing resistance against bending of the walls in
the longitudinal direction, and the two outer layers would mainly provide re-
sistance against bending of the walls in the transverse direction, therefore enhanc-
ing the local buckling capacity of the profile.

Table 1 Nominal section properties

Type Area Imajor axis Iminor axis Warping J

(mm2) (mm4) (mm4) (mm6) (mm4)
A 2375 1.58107 3.92106 4.971010 1.98104
B 2466 1.58107 4.64106 5.031010 2.06104

2.2 Manufacturing Process

Veneers were delivered in sheets of 1.2 m 1.3 m and sheets with a minimum
number of or with no natural defects (knots, resin veins, etc...) were selected. Dif-
ferent veneer sheets were used for each layer constituting a profile. Yet, all cross-
sections in a set were manufactured from the same sheets, glued in the exact same
order, allowing comparison between the two types of cross-sections in a set.
Each cross-section was manufactured as:
Step 1: Sheets were cut to size and veneers were soaked in water for 48 hours.
Step 2: The veneers were laid flat on a bench, with the grain in the appropriate
orientation (see section 2.1), heated using a steamer and bent around a jig
to form the cross-section.
Step 3: Each flat side of the cross-section was clamped to the jig, as seen in Fig-
ure 4 for Type B cross-section.
Step 4: The jig was placed in an oven at 40C for about 12 hours, till the bends
were dried and hold their shapes.
Step 5: The veneers were unclamped and removed from the jig, further left to dry
in the oven for about 4 hours, and stored in an air-conditioned room until
their moisture content reached equilibrium.
Step 6: The veneers were glued around the jig at ambient temperature with Res-
orcinol formaldehyde structural adhesive. Similar to Step 3, each flat side
of the cross-section was clamped to the jig. Rubber sheets were inserted
between the veneers and the clamping plates to uniform the applied pres-
sure. The glue was left to set for a minimum of 48 hours before unclamp-
ing the cross-section.
Step 7: The excess wood in the lip stiffeners were cut to form the final cross-
To simplify the manufacturing process and obtain more reliable products, espe-
cially for the bends, a vacuum press is planned to be used in the future. Figure 5
shows a photo of the final products.
Thin-Walled Timber Structures 123

Fig. 4 Type B Cee-section during Fig. 5 Final Cee-sections

manufacturing Step 3

To determine the mechanical properties of each set of cross-sections, flat panels

were also manufactured using the same sheets as their associated cross-sections
glued in the exact same order and orientation. Before testing, cross-sections and
flat panels were left in an air-conditioned room until their moisture content
reached equilibrium.

3 Material Testing

3.1 MOE Parallel to Grain

The Modulus of Elasticity (MOE) parallel to the grain of the veneers used to
manufacture the cross-sections was estimated using a non-destructive resonance
method [3]. A 300 to 500 mm long 30 mm wide sample of each veneer was
simply supported on rubber bands and impacted with a hammer. The natural fre-
quency of the tested sample was recorded using a microphone and analysed using
the software BING (Beam Identification by Nondestructive Grading) [4]. Table 2
gives the measured MOE parallel to the grain of each veneer used.

Table 2 Modulus of Elasticity for each veneer used

MOE parallel to the grain (MPa)

Set Layer 1 (Inside) Layer 2 Layer 3 Layer 4 Layer 5 (Outside)
1 12969 13279 11955 12471 14648
2 20399 19317 16811 19164 15113
3 17187 16186 16725 18776 14401
124 B.P. Gilbert, S.B. Hancock, and H. Bailleres

3.2 MOR Parallel to the Longitudinal Axis

The flat panels of each set of cross-sections were cut in four and reglued to form
20 mm nominal thick panels. Nominal 100 mm long 46 mm wide samples were
then tested in compression parallel to the longitudinal axis of the cross-sections, in
a 500 kN capacity MTS testing machine, following the method proposed in the
Australian and New-Zealand Standard AS/NZS 4357.2 [5]. The apparent Modulus
of Rupture MORapp is calculated as,
MOR app = (1)
where Pmax is the maximum applied load, and b and d are the measured width and
depth of the sample, respectively. Table 3 gives the average apparent MOR for
each set of cross-sections.

Table 3 Average apparent Modulus of Rupture for each set of cross-section

Set Number of tests Average MORapp (MPa) CoV

1 2 29.1 0.014
2 3 36.3 0.034
3 3 34.8 0.023

4 Stub-Column Tests

4.1 Test Set-Up

The profiles were tested in compression in a 500 kN capacity MTS testing ma-
chine in which the lower platen was fixed, while the upper platen was mounted on
a half sphere bearing which could rotate so as to provide full contact between the
platen and the specimen, as shown in Figure 6. The tests were performed in dis-
placement controlled at a stroke rate of 0.6 mm/min. Load was applied through the
theoretical centroidal axis of the cross-sections.
This test arrangement, i.e. consisting of testing short columns of lengths typi-
cally less than twenty times their least radius of gyration, aims at determining the
effect of local buckling on the column performance and allows the determination
of the column strength (or section capacity for local buckling) [6].
For Type A cross-sections only, three Linear Variable Displacement Transduc-
ers (LVDT) recorded the horizontal displacements of the centreline of the web at
heights of 175 mm (LVDT 2), 250 mm (mid-height LVDT 1) and 325 mm
(LVDT 3), and one LVDT recorded the horizontal displacement of the centreline
of the north flange at mid-height (LVDT 4), as shown in Figure 6.
Thin-Walled Timber Structures 125

(a) (b)

Fig. 6 Stub-column test set-up (shown for Type A cross-section), elevations: (a) flange
view, (b) web view

4.2 Test Results

Table 4 gives the recorded section capacity Ns of each cross-section (defined as
the maximum applied load) and the associated squash load Nsquash, representing the
upper bound capacity of the cross-section, defined as,
N squash = MOR app A (2)

where MORapp is given in Table 3 and A is the nominal cross-sectional area given
in Table 1.

Table 4 Stub-column test results

Type A Type B
Set Ns (kN) Nsquash (kN) Ns/Nsquash Ns (kN) Nsquash (kN) Ns/Nsquash
1 41.3 69.1 0.60 54.7 71.8 0.76
2 47.5 86.2 0.55 63.5 89.5 0.71
3 49.8 82.7 0.60 61.3 85.8 0.71
Average 46.2 79.3 0.58 59.8 82.4 0.73

It can be seen from Table 4 that the addition of the web stiffener in Type B
cross-section resulted in an increased average section capacity of 29% for an in-
creased cross-sectional area of 3.8%. Type A and B cross-sections reached an
average of 58% and 73% of their upper bound capacity Nsquash, respectively.
126 B.P. Gilbert, S.B. Hancock, and H. Bailleres

Figure 7 and Figure 8 plots the recorded displacements of the four LVDTs and
shows the failure mode, respectively, for the Type A cross-section of the third set.
It can be noticed in Figure 7 and Figure 8 a large post-buckling behaviour, similar
to the one encountered in steel structures.

Fig. 7 Test results for set n3, Type A cross-section Fig. 8 Failure mode for set n3,
Type A cross-section

4.3 Comparison with Cold-Formed Steel Structures

Using typical densities of 550 kg/m3 for hoop pine [7] and 7,850 kg/m3 for steel,
the cross-sectional area required for a steel profile to equal the linear weight of the
studied cross-sections is in the range of 166 (Type A cross-section) to 173 mm
(Type B cross-section). This range of cross-sectional area would lead to a maxi-
mum section capacity Ns of 74.7 kN to 77.9 kN for 450 MPa graded steel, i.e.
about 39% and 22% greater than the average observed capacities of cross-sections
Type A and B, respectively. Yet, these steel profiles would likely have slender
wall elements leading to low local buckling capacities and therefore reduced col-
umn strength.
For instance, the C7510 cold-formed steel Cee-section (75 mm deep and 1.0
mm wall thickness) from the Australian manufacturer Lysaght [8] has a nominal
cross-sectional area of 162 mm2 and a stub-column capacity of 46.2 kN [9] for
450 MPa graded steel. This capacity is equal and 29% less than the average ob-
served capacities for cross-sections Type A and B respectively. Moreover, the
bending stiffness EI about the major axis of bending of the C7510 is equal to
2.881010 N.mm2 (E = 200,000 MPa and I = 1.44105 mm4), while the bending
stiffness of Type A cross-section is equal to 1.231011 N.mm2 (E = 13,000 MPa at
Thin-Walled Timber Structures 127

12% moisture content [7] and I = 9.45106 mm4 when only considering the three
inner layers resisting bending), i.e. 4.3 times greater than the C7510 steel profile.
Similarly, Yap and Hancock [10] performed stub-column tests on 128 mm deep
and 0.42 mm wall thickness complex high strength cold-formed steel open pro-
files. The nominal yield stress of the profiles was equal to 550 MPa, the cross-
sectional area to 158 mm2 and the average tested stub-column capacity to 45.7 kN,
i.e. 1% and 31% less than the average observed capacities for cross-sections Type
A and B respectively. The profiles tested in [10] had a bending stiffness about the
strong axis of bending equal to 4.741010 N.mm2 (I = 2.37105 mm4), i.e. about
2.6 times less than the studied timber profiles.

5 Future Studies

By advancing the structural and mechanical knowledge on thin-walled timber

structures, the final aim of this project is to eventually develop design rules, so
engineers can safely use the proposed products. Current researches involve devel-
oping a Finite Element model to reproduce the test results presented in this paper.
Moreover, the ultimate strength and post-buckling behaviour of thin-walled struc-
tures is significantly influenced by geometric imperfections [11] and understand-
ing the influence of the moisture content on these imperfections, and therefore on
the capacity of the cross-section, is important. Additional proposed future research
also includes (i) developing joining details to connect manufactured lengths of
cross-sections, (ii) testing and understanding the structural behaviour of long col-
umns and beams, (iii) extend the Finite Strip Method [12] currently used to deter-
mine the buckling curves of thin-walled isotropic and orthotropic materials to
composite materials, and (iv) extend the Direct Strength Method [13] currently
used to design cold-formed steel profiles to thin-walled timber profiles.

6 Conclusion

This paper presented an experimental investigation on thin-walled timber struc-

tures. 210 mm deep, 105 mm wide and 5 mm wall thickness composite timber
Cee-sections were fabricated by gluing together thin softwood (Hoop pine - Arau-
caria cunninghamii) veneers. Two types of cross-section, with and without inter-
mediate web stiffener, have been investigated. The manufacturing process is
detailed in the paper. Stub-column tests were performed and results show that (i)
manufacturing structurally sound thin-walled timber structures is possible, (ii) the
investigated cross-sections have significant post-buckling behaviour and (ii) are
able to compete with cold-formed steel products with similar or greater weight to
compressive capacity ratio and significantly higher bending stiffness.
128 B.P. Gilbert, S.B. Hancock, and H. Bailleres

[1] Hancock, G.J.: Design of cold-formed steel structures (to AS/NZ 4600:2007), 4th
edn. Australian Steel Institute. North Sydney, Australia (2007)
[2] Yan, H., Shen, Q., Fan, L.C., Wang, Y., Zhang, L.: Greenhouse gas emissions in
building construction: A case study of One Peking in Hong Kong. Building and En-
vironment 45, 949955 (2010)
[3] Brancheriau, L., Bailleres, H.: Natural vibration analysis of clear wooden beams: a
theoretical review. Wood Science and Technology 36, 347365 (2002)
[4] CIRAD, BING (Beam Identification by Nondestructive Grading) software,
http://ur-bois-tropicaux.cirad.fr/produits/bing/usage (ac-
cessed on April 23, 2013)
[5] AS/NZS 4357.2, Structural laminated veneer lumber, Part 2: Determination of struc-
tural properties - Test methods, Standards Australia, Sydney, Australia (2006)
[6] Galambos, T.V.: Guide to stability design criteria for metal structures. Wiley and
Sons, New York (1998)
[7] Kingston, R.S.T.: The mechanical properties of Queensland grown hoop pine,
Proj.TM.9-7, Final report, Commonwealth of Australia - Council for Scientific and
Industrial Research - Division of Forest Products, South Melbourne, Australia
[8] BlueScope Steel Limited, LYSAGHT Zed & Cee Purlins and Girts,
purlins-and-girts (accessed on April 23, 2013)
[9] AS/NZS 4600, Cold-formed steel structures, Standards Australia, Sydney, Australia
[10] Yap, D.C.Y., Hancock, G.J.: Experimental Study of Complex High-Strength Cold-
Formed Cross-Shaped Steel Section. ASCE Journal of Structural Engineering 134,
13221333 (2008)
[11] Schafer, B.W., Li, Z., Moen, C.D.: Computational modeling of cold-formed steel.
Thin-Walled Structures 48, 752762 (2010)
[12] Cheung, Y.K.: Finite Strip Method in structural analysis. Pergamon Press, Inc., New
York (1976)
[13] Schafer, B.W.: Designing cold-formed steel using the direct strength method. In:
LaBoule, R.A., Yu, W.W. (eds.) Proceedings of the 18th International Specialty
Conference on Cold-Formed Steel Structures, Orlando, Florida, pp. 475490 (2006)
Recommendations for the Design of Complex
Indeterminate Timber Structures

Andrew Lawrence

Arup, MA (Cantab), CEng, MICE, MIStructE, London, United Kingdom


Abstract. Complex indeterminate structures with multiple loadpaths are now in-
creasingly common in timber. However, due to connection fit-up and the material
variability, one loadpath might be stiffer and attract forces which are perhaps two
or three times the level predicted by a simple elastic analysis. To achieve overall
ductile behavior is important to ensure that the connections are sufficiently ductile
to protect the brittle timber members from damage. This paper examines whether
the rules in current design codes are adequate for the design of such structures.

Keywords: indeterminate structures, grillage type structures, load paths, brittle

damage, ductile damage.

1 Introduction

Timber has traditionally been used in one-way spanning structures such as arches
and trusses. However, advances in analysis and fabrication techniques over the
past 5-10 years, have enabled the construction of ever more complex two-way
spanning indeterminate timber grillages. Such structures, including the Metz
Pompidou, the Metropol Parasol (see Figure below) and the Canary Wharf Station
Roof, in which the author has been involved, and are now very much part of the
modern architectural aesthetic.
Timber structures have traditionally been designed using a single elastic analy-
sis. However, for complex indeterminate timber structures with multiple loadpaths
this will no longer necessarily be a safe design method. For example, if there are
several close supports, then due to connection fit-up and material variability the
load in some members near the supports could be significantly higher than pre-
dicted by the simple analysis. As the loads attempt to redistribute, damage to the
brittle timber elements could result.
Elastic analysis is typically used for ductile materials such as steel and (under-
reinforced) concrete, which can accommodate a degree of load redistribution.
However, an elastic analysis will only be applicable for a brittle material such as
timber, if the overall structure has been designed to behave in a ductile manner.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 129
DOI: 10.1007/978-94-007-7811-5_12, RILEM 2014
130 A. Lawrence

This paper examines whether the rules in current timber design codes are ade-
quate to ensure overall ductile behaviour. There are two obvious alternatives to
such an approach:
a) Making the structure determinate this would reduce the overall robust-
ness as it would remove the possibility of alternative loadpaths in the
case of an accidental design situation;
b) Using a sensitivity analysis to consider the effect of varying connection
stiffness except for very simple buildings, this will generally lead to
structures which are expensive to design and construct, because of the
need to design for multiple loadpaths.
While these approaches will sometimes be useful, they will not be discussed
further in the current paper. Ensuring ductile behaviour is believed to offer several
- it should lead to a more efficient design (because it avoids the need to
design for multiple alternative loadpaths);
- it will be more robust in the event of accidental damage;
- the large deflections which would occur before failure would provide a
visible warning of any potential problems.

2 Ensuring Ductile Behaviour of the Overall Structure

Timber is an inherently brittle material. This is especially true in shear and in ten-
sion perpendicular to grain (because of the weak bond between the fibres) and also
Recommendations for the Design of Complex Indeterminate Timber Structures 131

in tension parallel to grain (because of the effect of natural defects such as knots).
Therefore the easiest way to ensure overall ductile behaviour of the structure is:
- to design the connections to fail before the (brittle) timber members,
thereby limiting the global load which the members can attract;
- to ensure the connections fail in a ductile manner;
- to ensure that the connections have adequate ductility to allow the loads
within the structure to redistribute, thereby compensating for uncertainty
over the initial loadpath.
This is similar to the Capacity Based Design referred to in seismic codes.
Each of these three requirements will now be discussed in turn.

3 Ensuring Ductile Behaviour of the Connections

There are three common types of timber connection:

- bearing connections (including traditional joints);
- dowel type fasteners;
- bonded rods.
To avoid damage to the connected member, it is important to ensure that the
connection fails in a ductile manner under the various applied loads (axial, bend-
ing, shear and, if applicable, torsion). It is assumed that the member is able to
resist any directly applied external loads; the issue is the extent to which it can
support the overall structure beyond and therefore it is necessary to limit the load
which can be applied to the member by the connection. Each of the three types of
connection will now be examined in detail.

3.1 Bearing Connections

These rely mainly on compression parallel and perpendicular to grain. Such fail-
ures are generally ductile. However, notched ends are susceptible to a brittle split-
ting failure in tension perpendicular to grain. This is recognised in design codes
which limit the shear on notched ends to prevent splitting. However, in complex
structures, where the load on those notched ends might be two to three times the
predicted value, there is still a risk of splitting. The risk of splitting therefore needs
to be prevented, which is easily done by introducing cross-grain reinforcement,
such as screws or cross-veneers.

3.2 Dowel Type Metal Fasteners in Shear

These include nails, screws, dowels and bolts. Formulae for the potential failure
modes were originally developed by Johansen [4] and are now included in modern
132 A. Lawrence

design codes. The formulae rely on either crushing of the wood, yielding of the
fastener or a combination of both (Eurocode 5, Figure 8.2). Several measures can
be taken to encourage ductile behaviour:

- Johansen failure modes which include yielding of the fastener will obvi-
ously be more ductile. The fasteners should therefore be so proportioned
(relative to the thickness of the wooden parts) to ensure that these modes
- Johansen assumes that the full crushing (i.e. embedment) strength of the
timber can be developed and that premature brittle failure does not occur
due to splitting ahead of the fastener, tensile failure of the residual timber
section or plug shear failure. To avoid splitting, codes impose minimum
spacings, end and edge distances. Codes also require additional checks
to be made to ensure that brittle tensile, block or plug shear failure does
not occur;
- Since the rules to prevent splitting are empirical and could be compro-
mised by, for example splits due to drying shrinkage, it is recommended
that screws are incorporated as cross grain reinforcement. For a small ex-
tra cost these will give a more reliable behaviour and also allow more
ductility and distortion before failure. Guidance on their design will
hopefully be included in future revisions of Eurocode 5;
- Where several fasteners in-line are loaded parallel to grain, testing shows
that the load distribution between the fasteners is uneven (partly due to
fit-up) and that a strength reduction factor needs to be applied (the neff
factor in Eurocode 5) to prevent premature splitting which would then
prevent the redistribution of forces within the connection. Testing [9, 10]
shows that cross-grain reinforcement screws can again be used to ensure
more reliable behaviour and to enable the full crushing strength of the
timber to be mobilized;
- Testing [6] also shows that multiple bolted connections can have less
ductility, because the oversized holes may mean that not all the bolts are
in contact with the timber, again leading to an uneven load distribution. It
is therefore recommended that dowelled connections which have a more
reliable behavior are used in preference to bolts;
- Smaller fasteners will obviously be more ductile than large fasteners and
should therefore be used in preference, especially if a large amount of
ductility and rotation is required to redistribute the forces within the
structure. It is interesting to note that Eurocode 8 limits fasteners to only
12mm diameter. However, as well as achieving ductility this will be to
ensure good energy dissipation and therefore is considered to be an ex-
cessive limit for non-seismic areas.
Recommendations for the Design of Complex Indeterminate Timber Structures 133

3.3 Bonded Rods

Limited research to date [11] suggests that to ensure ductility:
- Mild steel rods should be used, rather than high strength steel;
- The rods should have a sufficient unbonded length that is necked down to
allow for elongation without damaging the glue bond. The proportions
can be varied to control the degree of ductility required;
- Cross-grain reinforcement screws should be used to prevent splitting of
multiple bonded rods.

4 Ensuring That the (Brittle) Timber Members Are Stronger

Than the (Ductile) Connections

As noted, to ensure that the overall structure behaves in a ductile manner, it is not
sufficient just to make the connections ductile; it must also be ensured that the
(brittle) timber members are stronger than the (ductile) connections.
Fortunately, connections are generally the weak points of timber structures and
therefore this criterion will usually be satisfied automatically. However, checks
should always be made to ensure that this is the case. In making this check, it
should be remembered that the connection strengths derived from design codes are
conservatively low (to account for the variability of the timber and the approxima-
tions of the Johansen formulae [4]) and therefore the timber members must exceed
the code-derived connection strength multiplied by a suitable overstrength
Limited testing to date [7] on dowel type fasteners suggests that this factor is
about 1.6. If there is any uncertainty about the members being stronger than the
connections, then the connections may need to be tested; reference [7] that sug-
gests testing on at least 10 similar connections to determine the overstrength factor
(based on the 95th percentile of connection strength).

5 Ensuring That the Connections Have Adequate Ductility

In some cases it will be necessary to undertake a non-linear analysis with elastic-
perfectly plastic springs to confirm that the connections have adequate ductility
(for example to ensure that they can accommodate sufficient rotation) to allow
redistribution of the forces within the structure. More research and testing is
needed to help predict how much deformation a connection can accommodate
before failure.

6 Summary
For complex indeterminate timber grillage-type structures with multiple loadpaths
it is recommended that the following conditions are met:
134 A. Lawrence

(1) The connections should be ductile. For connections with dowel type fasteners,
this will be the case if the fasteners are so proportioned to ensure that one of the
Johansen [4] failure modes governs which incorporate yielding of the fasteners.
Dowels should also be used in preference to bolts. To improve the level of ductili-
ty smaller fasteners and reinforcement perpendicular to grain can be used; the
latter will ensure more reliable and predictable behaviour. Checks should also be
made to ensure that the dowels yield (with an adequate factor of safety) before the
onset of brittle plug shear or tensile failure.
(2) Members should be stronger than the connections.
(3) In some cases it will be necessary to undertake a non-linear analysis, incorpo-
rating elastic-perfectly plastic springs to represent the ductility of the connections,
to demonstrate that the connections have adequate ductility, possibly backed up by
testing to help understand the degree of deformation which the connections can

Acknowledgements. The author is grateful to colleagues at Arup for their assistance and
also to the detailed research which has been undertaken at Eindhoven, Karlsruhe, Stuttgart,
UBC and Zurich which has been referred to in this paper. There is still much work to be
undertaken in the field and it is hoped that this paper may help guide the areas for future

[1] EN 1995-1-1:2008, Eurocode 5 Design of timber structures part 1-1: General
common rules and rules for buildings
[2] EN 1998-1:2004, Eurocode 8 - Design of structures for earthquake resistance part
1: General rules, seismic actions and rules for buildings
[3] SIA 265. Swiss code for timber structures
[4] Johansen, K.W.: Theory of Timber Connections. IABSE 9, 249262 (1949)
[5] Mischler: Dowelled timber connections with high efficiency. In: Proceedings of the
First International RILEM Symposium on Timber Engineering (1999)
[6] Leijten, A.J.M.: Requirements for moment connections in statically indeterminate
timber structures. Engineering Structures 33, 30273032 (2011)
[7] Jorissen, A., Fragiacomo, M.: General notes on ductility in timber structures. Engi-
neering Structures 33, 29872997 (2011)
[8] Bruhl, F., Kuhlmann, U., Jorissen, A.: Consideration of plasticity within the design
of timber structures due to connection ductility. Engineering Structures 33, 3007
3017 (2011)
[9] Blass, H.J., Schadle, P.: Ductility aspects of reinforced and non-reinforced timber
joints. Engineering Structures 33, 30183026 (2011)
[10] Lam, F., Wrede, M.S., Yao, C.C., Gu, J.J.: Moment resistance of bolted timber con-
nections with perpendicular to grain reinforcements. In: Proceedings 10th WCTE
Miyazaki, Japan
[11] Tlustochowicz, G., Fragiacomo, M., Johnsson, H.: Provisions for ductile behaviour
of timber-steel connections with multiple glued-in rods. ASCE Journal of Structural
Engineering (September 13, 2012)
Novel Lightweight Timber Composite Element:
Web Design in Shear and Compression

Simon Aicher and C. Stritzke

Materials Testing Institute, University of Stuttgart, Department of Timber Constructions,

Pfaffenwaldring 4b, 70569 Stuttgart, Germany

Abstract. The paper examines the build-up and some aspects of the mechanical
behavior of a novel lightweight timber composite called the Keel-web element.
The element is a double-skinned composite similar to a multiple box beam ele-
ment. The flanges consist of finger jointed lumber chords arranged and glued in
parallel. The name-giving specific characteristic of the element consists of the
multiple S-shaped webs made of plywood or OSB, resembling ship keels, glued in
between the flanges. The element, which can be produced in a fully automatized
process, as straight or cambered, with lengths of up to 35 m, has recently obtained
a German technical building approval. The build-up will first be examined, and
afterwards, the engineering design approach for the shear force and compression
capacity, partly following Eurocode 5, is shown.
In order to analyse in greater detail the stability and nonlinear bending of the
pre-curved webs at end supports, the load deflection behavior of the webs at an
increasing support force is studied by 2nd and 3rd order beam column and plate
theory. Additionally, the bending stresses introduced from the manufacturing pro-
cess of the S-shaped webs which are subject to relaxation, must also be consid-
ered. With increasing loads, the out of plane web displacements are restrained by
the adjacent webs, leading to a reduction of the free web height. Solving the dif-
ferential equation for a fixed-end beam, valid results for small deformations can
be calculated, while a nonlinear finite element simulation is performed in order to
consider large deformations and contact boundaries. It is shown that the stress-
state of a 3D shell model with contact simulation of the webs provides a good
estimate of the experimental load capacities.

Keywords: Lightweight timber element, Keel-web element, glued composite

structure, S-shaped webs, Buckling, Nonlinear bending, Support design, Shear
capacity, Plywood, OSB.

1 Introduction

Wood as a construction material can be considered as lightweight due to its rela-

tively high strength-to-weight ratio. In comparison to other common building

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 135
DOI: 10.1007/978-94-007-7811-5_13, RILEM 2014
136 S. Aicher and C. Stritzke

materials, the tearing length of a softwood member loaded in tension parallel to

the fiber direction is large. Depending on the strength class, defect state and cross
sectional size, the tearing length is between 1.5 to 3 times larger as compared to
usual construction steel. Additionally, the compression strength of wood is within
the same range as that of concrete. However, as wood is increasingly being used
due to its undisputed ecological benefits (for example CO2-storage) as well as its
sustainability, it is approaching the limits of sustainable forestry. The result is a
steady price increase for wood products, which is expected to rise significantly in
the coming years. The higher costs already have had an impact on the competi-
tiveness of massive wooden building products such as cross-laminated timber
(CLT). Hence, a growing future demand consists in the development of material
resource preserving lightweight structures.
Recently, the palette of wooden lightweight structural elements was expanded
with a new innovative glued element type, known as the Keel-web element [1].
Based on extensive trials, testing and analysis, the construction principle has re-
cently obtained a German Technical building approval [2]. The elements are al-
ready being produced on an industrial scale and are increasingly being used in
building structures.
The elements have a very high load capacity and stiffness and are ideally suited
for long-span roof structures as well as ceilings. To get high stiffnesses in all spa-
tial directions, the webs are bent in the manufacturing process to the characteristic
S-shape. This particularity leads to interesting structural mechanical behaviour,
especially of the web structure. One of the major interests concerns the buckling
behaviour and nonlinear bending in the end support areas, which is the focus of
this paper.

2 Build-Up of Keel-Web Elements

The Keel-web element is a double-skinned wooden composite element with thick

skins made of finger jointed softwood lumber chords glued in parallel at the nar-
row edges. The lumber cross-sections are within the size range of (thickness x
width) 40 mm x 70 mm to 90 mm x 175 mm, depending upon the specific build-
up of the element, with a height between 228 mm and 800 mm (see Table 1). Be-
tween two adjacent lumber cords, which shall have a minimum strength class of
C24, the ends of the webs are glued-in (see Figs. 1 and 2). The webs consist of
characteristically S-shaped bent thin wooden based panels, being either three-
layered plywood acc. to EN 636 [3] (suitable for damp areas) with a thickness of
4.8 mm or OSB/3 material acc. to EN 300 [4] with panel thicknesses of 8, 10 or
12 mm. Webs made of plywood are confined to element depths of 228 mm to
380 mm, whereas OSB/3 webs are used for elements with depths of 485 mm to
800 mm.
Novel Lightweight Timber Composite Element 137

Fig. 1 View of Keel-web elements with webs made of plywood (Height: 380 mm)

The bending strength class of the plywood material must be at least class
F60/10 and at most class F70/15. The flexural modulus of elasticity (MOE) class
must be at least class E100/5 and at most class E40/20. Both panel materials shall
be CE-marked on the basis of EN 13986 [5]. The web and flange dimensions can
be designed according to the applied loading, within the size and geometry re-
strictions specified in the German technical Approval Z.9.1-831 [2].
In the layout of the element, special consideration shall be given to the bending
stresses induced in the web during production to achieve the typical S-shape, ob-
tained by a relative horizontal offset of the tension and compression flange chords
by half of the width, /2, of the lumber cross-sections. The maximum web stress
to be considered is then (where: , = width and depth of flange (see Fig. 2)
and , , = MOE in bending of panel material perpendicular to span):

, , =3 , , /(2 ) . (1)

Taking into account a high degree of stress relaxation, the bending stress in the
webs at the connection to the flanges is limited in [2] to

, , , < 0.8 , , for plywood webs and (2a)

, , , / < 1.0 , , for OSB/3 webs. (2b)

In the manufacturing process performed by Kulmer Bau GesmbH & CoKG, Aus-
tria, the webs are bent into the S-shape and bonded with the lumber chords in a
single operation, thus forming elements with standard widths of 1.2 m. At the
present time, the maximum element length is up to 35 m and can optionally be
produced pre-cambered.
Table 1 gives an overview of typical product ranges allowable given the
provisions in [2].
138 S. Aicher and C. Stritzke

Table 1 Cross-sectional size and geometry ranges of Keel-web elements according to [2]

web-material Element height [mm] Dimensions of flanges [mm]

(thickness )
228 70-90 40-50
280 70-112 40-57
(4.8 mm)
380 70-112 40-67
485 70-120 40-67
560 70-140 57-80
OSB/3 615 70-155 57-80
(8, 10, or 12 mm) 730 70-140 67-80
785 90-165 80-90
800 90-175 80-90

3 Simplified Design Rules

3.1 General
For a sufficiently accurate simplified engineering design in the case of static load-
ings, a splitting of the cross-sectional structure into equal I- or half box beams (see
Fig. 2) can be performed [2]. Similar to I- or box beams, the following stresses
must be checked [2; 3]:
- / , / , bending stresses on the top and bottom of the cross sectional
- , / compression / tensile stresses at the centroid of the cross sectional ar-
ea of flanges
- , shear stresses in the glue lines between flanges; and webs, and in
web-web glue lines, respectively
- , shear stresses at the centroid of the cross sectional area of the webs,
considering buckling (see below)

3.2 Shear Force Design

Due to the comparatively slender webs, it is of upmost importance to ensure that
buckling of the web is excluded. The design verification implemented in [2] con-
siders the web stability in an implicit manner through use of a material-specific
rule for the geometry dependent effective shear strength , , for the web. The
design procedure follows, apart from material related aspects, the shear force ca-
pacity design procedure specified in Eurocode 5 [6] for glued I- and box beams
with straight thin webs. The latter is briefly outlined below to demonstrate the
analogy of standard design of the shear capacity design to that of the Keel-web
Novel Lightweight Timber Composite Element 139

Fig. 2 Simplification of the cross-sectional structure of Keel-web elements[1] for design in

[2] a) unit cell of real structure; b) I-beam approximation; c) half box beam (for shear
capacity design)

elements with the S-shaped webs. Eurocode 5 specifies that unless a detailed
buckling analysis is performed, and provided that the web geometry suffices
/ 70 (see Fig. 2c), the subsequent design procedure should be followed:

, , , , (3)


design value of acting shear

, ,
force of each web
design shear force re-
, , = , , , (4)
sistance of each web
0.5 , + ,
, = 1+ effective web cross-section (5)

or in case of equal flange thicknesses = , /

, = + = , effective web cross-section (6)


, , = , , / 35 (7)
, , = , , = , , 35 / 70 . (8)
140 S. Aicher and C. Stritzke

The design strength values and the characteristic values are related as usual via

, , , = , , , / . (9)

where is the strength modification factor for influence of load duration and
moisture content; and is the partial strength (safety) factor. The above equa-
tions are slightly modified from those given in [6] by the introduction of , ,
and in order to enable the calculation of the S-shape in a manner similar to that
for a standard I-shape. The equations show that EC5 does not consider any buck-
ling phenomenon up to a web aspect ratio of / = 35 whereby the design (d)
or characteristic (k) shear force capacity is fully determined by the effective cross-
sectional web area and the characteristic in-plane shear strength value , , of the
web panel material. For larger aspect ratios, the buckling effects caused by shear,
in-plane compression close to the support and from in-plane bending are consid-
ered together as an effective shear strength value , , whereby the reduction
factor depends exclusively on the aspect ratio / .
In the derivation of the design equations for the technical approval for the Keel-
web element, for sake of simplicity the same approach as specified in EC5 for
glued I- and box beams for computing shear force capacity per web on the basis of
an effective shear strength value , , was pursued. Based on a best fit to the
5 %-quantile values of the shear force capacities of full scale element tests with
several different geometry and size configurations (see Fig. 4), the material specif-
ic effective shear strength expressions analogous to Eqs. ( 7) and ( 8 ) were
derived and are given as follows (graphs of , , acc. to Eqs. ( 10 ) to ( 12 ) are
shown in Fig. 4):

, , , = , , , , , , / , = , , / , (10a,b)

In case of plywood ( = 4.8 mm)

=1 / < 30 (11a)
= 0.1124 + 772( / ) 30 / 61 (11b)

and for OSB/3 ( = 8, 10 and 12 mm)

=1 / < 46 (12a)
= 0.0133 + 2144( / ) 46 / 64. (12b)

The characteristic nominal in-plane shear strength values for the plywood and the
OSB/3 material were derived in the tests for [2] as:

, , , = 7.5N/mm and , , / , = 4N/mm . (13)

Novel Lightweight Timber Composite Element 141

Fig. 3 Test setup for shear capacity tests of Keel-web elements

8 f v,0,plywood,k
effective shear strength

approval [2]
EC5 [3]
fv,eff,k [ N/mm ]

f v,0,OSB/3,k

approval [2]
2 EC5 [3]

0 10 30 40 50 60 70
web geometry ratio (hw / bw )

Fig. 4 Effective shear strength dependent on web geometry ratio / and material type

4 Buckling and Bending Due to Support Normal Forces

4.1 General and Eigenvalue Results

It has been previously revealed in [7] that the vertical reaction force at the end
support locations leads to a complex interaction of normal force, bending moment
and shear force loading of the S-shaped web. Super-imposed to the section forces
from the vertical reaction force is the bending moment ( = 1 /12)

[ ] = 6 , , ( 2 )/ (14)

introduced during the production process due to the horizontal web displacement
= /2. The normal force per unit length of the web in span direction
[ ]= , [ ] imposed at the bottom side of the web by the support force
142 S. Aicher and C. Stritzkke

(Fig. 5) decreases in an approximately

a linear manner towards the upper unloadeed
edge of the element simillar to a classic I-beam, see Fig. 6 and [7]. The comprees-
sion stress distribution , ,90 thus depends considerably on the length of the supp-
port and the excess length
h. The pure compression strength verification of the weeb
can be performed in the case of sufficient excess length /4 analogous tto
EC 5 [6] as

, [0] ,
, , = = = (15)

where n = number of webs; V= total shear force per element and

= +2 . (166)

Flanges Web
11000 5500
370 370

370 1000
690 700
690 700
69 200
0.35 0.0422

0.35 0.0155
0.015 0.35

Fig. 5 Geometry notations att end support and material parameters used for stress analysis

a) b)

Fig. 6 Compressions stress distribution in the web of I-beams with straight web at ennd
supports (see Fig. 5) at V = 1 kN (dimensions as for KSE380, see below), calculated wiith
FEM a) with excess length; b)
b without excess length
Novel Lightweight Timber Composite Element 143

As opposed to a straight web, an additional bending moment [ ] results

from the linearly diminishing normal force [ ] along the S-shaped web (see Fig.
7b). The maximum value at the support edge of the varying moment distribution
along depth of the web has been determined in [7] numerically as

[0] = (1.5 10 ) (17)


= ( /(2 )) (18)

is the averaged linearized inclination of the web.

For a straight, two-sided clamped beam-type column subjected to a constantly
distributed tangential compression line load q, an analytical expression for the
critical buckling load

= = 8.2445 , , / (19)

has been derived in [8]. For a two-sided clamped S-shaped beam subjected to a
uniformly distributed tangential axial line load, no analytical solution exists. Finite
element computations performed with S-shaped beams revealed as for the case of
a linearly variable tangential compression load [7] a rather small decrease of the
critical load (1st eigenvalue) as compared to the straight (reference) beam on the
order of approximately 2 6 %. This decrease is within practically relevant rota-
tion angles in the range of 10 to maximally 20. Fig. 8 shows the critical beam
buckling load for an S-shaped web with dimensions tested at element type
KSE380 ( = 294 mm, = 4.8 mm, = 108 mm and , = 1000 MPa);
further i.a. the result according to Eq. (21) is given.

Fig. 7 Section forces of S-shaped web a) [ ] from production; b) [ ] and [ ] from

support force N = V
144 S. Aicher and C. Stritzke

4.2 Higher Order Theory Bending

It is obvious that for the pre-curved web, the eigenvalues of a classical buckling
problem are rather arbitrary numbers. Nevertheless, they may however serve as a
verification of the results of 2nd order theory computations which must converge to
the eigenvalue result [9]. It can be easily seen that the pre-curved web starts to bend
stably under the action of an applied increasing normal force [ ], leading to higher
order deflection and hence to a stress/ strength problem for ultimate capacity.
Some features will be highlighted considering first a beam situation followed
by plate solutions. The governing differential equation is

[ ] + (1 / )( [ ]+ [ ]) = 0 (20)


= , , ; = ; [ ]= (3 2 ) / and = /2 (21)

represents the bending curve or shape function of the initial S-shape of the web at
zero normal force. [ ] is the additional lateral deflection resulting from the
linearly distributed normal force, which when integrated along the total web
height , delivers the vertical reaction force , [0].
Employing the boundary conditions [0, ] = 0 and [0, ] = 0 one ob-
tains the relationship of load vs. mid-web deflection according to 2nd order theory
as shown in Fig. 8 for a specific element configuration (Note: The results pre-
sented in Fig. 8 refer to the geometry and stiffness configurations of element type
KSE380 where = 380 , , / = 43 , = 108 , = 294 m
and = 4.8 ). It is evident from the graph that the large deflections of the
2nd order theory solution render the results to be simply of theoretical interest and
are practically irrelevant.
In a further step, the problem of large deflections can be handled as a 3rd order
theory problem resulting in a coupled partial differential equation system where
deflection and longitudinal displacements are coupled. The beam solution ob-
tained from finite element analysis, is shown in Fig. 8, revealing the typical 3rd
order theory effect of stiffening of the structure at large deflections. It can be seen
that the load deflection curve now reaches the characteristic test load and the cal-
culated shear capacity according to Eqs. (10a) and (12). Nevertheless, the pure
3rd order theory approach is again irrelevant for determining realistic failure loads
(after conversion to stresses) as the horizontal mid-depth displacement of the web
is by far too large as compared to any realistic situation or test result. (Note: The
almost exact agreement between characteristic test load and calculated shear ca-
pacity shown in Fig. 8 results from the above explained derivation of the effective
shear strength).
In a further attempt to obtain the true load capacity, the authors have considered
the experimentally witnessed and mechanically plausible fact that two adjacent
Novel Lightweight Timber Composite Element 145

webs move closer together in the horizontal direction at increasing vertical loads
until making contact along a fictive vertical line/plane. This issue is schematically
illustrated in Fig. 7b. The empirically proven phenomenon was implemented in
the FE-modelling by a contact boundary which prevents any horizontal displace-
ment beyond the vertical line at y = 0, i.e. at the symmetry line of the keel formed
by adjacent webs.

load [kN]

char. test load Vu,k
calculated shear capacity
3,5 3rd order theory
contact boundary
3,0 beam modell 3rd order theory
Pcrit,straight beam modell
3rd order theory 2nd order theory
1,5 contact boundary beam modell
3D modell
1st order theory
0,5 beam modell element type:
0 10 20 30 40 50 60 70 80 90
displacement vN[hw/2] [mm]

Fig. 8 Curves of vertical load vs. horizontal displacement at mid-height of web acc. to
beam and 3D shell analysis for different boundary conditions

It is apparent that the contact between the webs from moving closer together
reduces the free height of the S-shaped web. Fig. 8 reveals for a beam solution and
for a full 3D shell model that the vertical boundary contact leads to a significant
reduction of the horizontal displacements and hence to a stiffening and stress re-
duction of the webs at increasing loads. Fig. 9 shows the changes in the moment
distribution along the web height for increasing loads, hereby revealing the differ-
ences which result from the discussed individual theoretical approaches.
In order to derive the load capacities associated with the nonlinear web
bending, the interacting stresses from normal force and bending moment have to
be considered. Load capacities were computed on the basis of a simple linear
interaction equation
146 S. Aicher and C. Stritzke

, ( ) , ( ) , ( )
(1 ) + + 1 (22)
, , , , , ,

which accounts for the following three contributions: i) the outerfiber bending
stress , [ ] resulting from the initial production process, ii) , [ ] being the
outerfiber bending stress resulting from the bending moment [ ] induced by
support force [ ] and iii) the axial compressive stress , [ ]. In addition, the
stresses , [ ] from the initial bending process are not considered to the full
elastic initial degree as a strong relaxation of the stresses can be assumed, which
has been verified by ongoing experimental investigations. The reduction of the
initial bending stress , [ ] by a relaxation factor in Eq. (22) can be
assumed realistically in the range of about 20 40 % . Fig. 10 shows the evalua-
tion of the stress interaction equation for increasing vertical load further depend-
ing parametrically on the relaxation factor . It can be seen that the strength
normalized stresses deliver a good prediction of the load capacity, especially when
the relaxation factor is assumed in the range of about 0.4.


30000 moment [Nmm]


-30000 3rd order theory considering contact boundary conditions

vertical load = 3000 N
-60000 30000


vertical load = 2000 N
15000 -30000

2nd order theory

3rd order theory

vertical load = 1000 N
-15000 15000
moment [Nmm]

element typ: KSE380 7500

vertical load = 0 N
0 hw/2 hw

Fig. 9 Moment distribution along web height while support force increases
Novel Lightweight Timber Composite Element 147

0 19 horizontal displacement f/hw [%] const. 19

element type:
result of interaction Eq. (24) [-]



char. test load Vu,k

0.5 0.0


-1000 0 1000 2000 3000 4000
zero vertical load [N]

Fig. 10 Stress interaction result vs. vertical load

5 Conclusions

Keel-web elements are innovative, strong and stiff lightweight composites made
of edge-glued solid timber members and thin S-shaped webs of either plywood or
OSB. Contrary to an easy verification of the bending stresses, the design of shear
and compression force capacity at the support as determined by the curved webs
poses a demanding mechanical problem. At present, the shear force verification is
based on an effective shear strength, derived from extensive element tests. It is
further assumed that the compression force design at the supports can be per-
formed as specified in Eurocode 5 for glued I- and box beams where buckling is
not considered explicitly.
The paper revealed that the present design of the new element type smears
several effects of interacting section forces which contribute to the nonlinear
bending of the webs. It is shown that all influences could well be separated and
that it is possible to determine the load capacity by linear interaction of the rele-
vant stresses. Several major issues remain to be solved. It is important to quantify
the relaxation behavior of the high bending stresses induced in the manufacturing
process of the S-shaped webs. Further, simplified equations have to be derived
from the 3D Finite Element solutions to enable a transparent calculation of the
different stress components dependent on curvature and edge aspect ratio of the
web material.

[1] Austrian patent: AT 501521: Trgerartiges/aus Einzelteilen zusammengesetztes
Bauelement sowie Verfahren zur Herstellung des Bauelements. Holder of patent:
Kielsteg GmbH, Graz, Austria
148 S. Aicher and C. Stritzke

[2] DIBt: German Technical Approval Z-9.1-831Kielstegelement, Holder of approval:

Kielsteg GmbH, Graz, Austria (2013)
[3] EN 636: Plywood - Specifications (2012)
[4] EN 300: Oriented Strand Boards (OSB) Definitions, classification and specifications
[5] EN 13986: Wood-based panels for use in construction. Characteristics, evaluation of
conformity and marking (2006)
[6] DIN EN 1995-1-1, Eurocode 5: Design of timber structures Part 1-1: General
Common rules and rules for buildings; German version EN 1995-1-1:2004 + AC:2006
+ A1:2008 (2010)
[7] Stritzke, C., Aicher, S.: Traglastverhalten von innovativen geklebten Holzleichtbau
Elementen. In: Kuhlmann, U., Brhl, F. (eds.) Holzbau Forschung + Praxis, pp. 115
124. University of Stuttgart (2012)
[8] Hauger, W.: Die Knicklasten elastischer Stbe unter gleichmig verteilten und linear
vernderlichen, tangentialen Druckkrften. Ingenieurarchiv 35, 221229 (1966)
[9] Petersen, C.: Stahlbau, 4th edn. Springer (2013)
New Timber Bridges: Inventive Design by

Frank Miebach and Dominik Niewerth

Ingenieurbro Miebach, D-Lohmar, Germany


Summary. The technology of gluing in wood construction has evolved

considerably in recent years. This has been shown especially in timber bridges that
Glulam is now the main building material. A further development is the so-called
block gluing, which provides a good basis for supporting structures. Numerous
bridges, especially in central Europe, appeal by unique design and monolithic and
solid construction. These structures base on two main developments: block
lamination of glulam and the composite of timber and concrete to one structural

Keywords: Block-Gluing, Glulam, Timber-Concrete-Composite.

1 Introduction

1.1 History in Europe after Industrialization

The history of bridge construction has always been connected to the technical
state of the art. It started with applications of natural material as stones and wood.
During the industrialization construction materials changed to steel and afterwards
to concrete and high developed wood constructions got displaced by concrete or
steel structures. Due to a mind change to ecological matters there is a rediscovery
of wooden structures in the last decade. And the characteristics of wood light but
strong, easy to process push this natural material into focus of architectural

1.2 Retarding in Evolution

Timber constructions have their own industrialization, but with some retardation.
In 1906 a German Otto Hetzer patented the gluing of wooden beams. The modern
glue-lamination was born. Glulam is a structural timber product consisting of
number of layers glued together to a certain cross section. Additionally it is
possible to block laminate a number of glulam beams together to achieve bigger
cross sections.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 149
DOI: 10.1007/978-94-007-7811-5_14, RILEM 2014
150 F. Miebach and D. Niewerth

This invention provides the possibility to produce wooden beams with huge
dimension in a constant quality: length of 50m and more, width and heights of 3m
and more. The gluing process even offers the possibility to produce curved
elements or even twisted elements. This allows thinking and constructing in free
Since 40 50 years glulam is used in halls not just for the roof but also for
the posts and reaches a respectable market share by now. Wooden bridges are on
the rise since almost 15 20 years.
There is a diversity of reasons for using of timber in bridges. For example,
prefabricated large-sized parts are easy to handle and assemble because of their
lower weight - a well-known advantage for economical constructions.
The durability of the construction often is discussed seriously because latest
planning often does not consider historic knowledge as constructive wooden
The development of gluing technology including block lamination is a type of
the retarded timber industrialization.

2 Block Laminated Structures

Block lamination generates massive timber cross section to achieve loadbearing

structures for large spans with a thin side view. Block lamination comprises three
main steps: limitless gluing of lamellas through finger jointing, gluing layers of
lamellas to a glulam beam and finally gluing numbers of beams to a massive

A certain form like a curve is created in step 2 lamination of layers. The

curve is always vertical to the longitudinal axis because the layers only are
flexible according to the thickness. Therefore smaller radiuses require thinner
According to the required form a block laminated beam may have a curvature
in top view or in side view. The difference in production is
a) Horizontal block lamination (curvature in side view)
New Timber Bridges: Inventive Design by Block-Gluing 151

or b) vertical block lamination (curvature in top view).

Even twisted forms are possible. Twisted beams are curved around two axes.
The first curvature is generated in step 2 lamination of layers as usual.
Afterward a horizontal band saw divides the curved beams into thin layers. These
thin layers are sufficient flexible to be curved again in a second direction during
block lamination.
152 F. Miebach and D. Niewerth

A part from curved gluing, through block lamination stepped beams are feasible
both in cross section and in longitudinal section. There for glulam beams with
different height or length are block-laminated to one block beam.

3 Timber-Concrete-Composite Bridges

The increasing role of ecology and sustainability of building materials more and
more influences construction industry and leads to a rethink on site of the builders.
The symbiosis between wood and concrete has the best conditions to meet all
resulting claims. Constructions of massive wood beams statically connected to a
concrete slab on the upper side ensuring optimum utilization of the material
specifications of both materials. In this case the wooden cross-section is
considered to take tensile forces and the concrete slab takes the pressure forces.
Special connectors ensure the interaction of timber beams and concrete slab to get
a more effective loading capacity and serviceability.
Advantages in comparison to conventional timber bridges:
- higher load capacity with lower height of construction
- good structural wood protection through cantilevered concrete slab on the top
- optimal load spreading of point loads by the concrete slab
- better cross bracing
- use of proven details in connections to the concrete
New Timber Bridges: Inventive Design by Block-Gluing 153

Advantages in comparison to conventional concrete bridges:

- lower weight of the superstructure and thus more efficient structure
- fast and efficient installation with high degree of prefabrication without
extensive formwork
- cost savings in foundation and the abutment
- improved energy balance and eco-balance, sustainability through CO reduction

3.1 HBV-Shear Connector

The HBV-shear connector by TiComTec (Haibach) is an expanded metal part that
is glued perpendicularly into the wooden structure. The dimension and number of
connectors is determined to the static needs. The concrete slab usually has a
thickness of at least 20 cm and has several functions: road decking, carrier plate
for the dispersal of transverse loads and constructional wood protection.

3.2 Head Bolts

As known from steel-concrete bond structure in this alternative the connecting
parts between wood and concrete are steel bars with welded head bolts. The bars
fit exactly into milled kerfs and are fixed with screws. To achieve an efficient
utilization, the axial distances of the dowel bars correspond to the traffic load and
transverse force caused by the traffic load.
154 F. Miebach and D. Niewerth

3.3 Kerfs and Glued in Reinforcement Bars

The shear connection in this type of timber concrete composite is implemented by
a combination of kerfs and glued-in steel bars. The kerfs are milled in transverse
direction and transfer shear forces (from longitudinal direction) into compressive
forces to the vertical sides of the kerfs. To avoid a lifting of the concrete additional
fasteners or connectors are considered (e.g. glued in reinforcement bars).

4 Conclusions

A changed awareness on ecological matters is the proper basis for timber bridges.
Timber is the one and only material that saves and stores CO2 permanently. And
the technical possibilities are still growing. For example, researches on modern
glues and gluing methods show in near future glues will be more efficient
(temperature-resistant and the level of bonding pressure during the gluing process
will be halved from 0,4 N/mm2 to 0,2 N/mm2). Against this background the
combination of block lamination and timber-concrete-composite is the beginning
of the latest development: The comeback of timber bridges as adequate road

Simon, A.: Analyse zum Trag- und Verformungsverhalten von Straenbrcken in Holz-
Beton-Verbundbauweise. PhD. thesis, Bauhaus - Universitt Weimar (2008)
Gerold, M.: Holzbrcken am Weg. Eigenverlag, Karlsruhe (2001)
Miebach, F.: Holzbrckenbau 3.0 als Weiterentwicklung. Brckenbau (March 2012)
Dietrich, J.R.: Faszination Brcken. Verlag Georg D. W. Callwey, Mnchen (1998)
Part II
Mechanical Connections
Steel-to-Timber Joints with Very High Strength
Steel Dowels Using Spruce, Beech and Azobe

Jan-Willem van de Kuilen1 , Carmen Sandhaas2 , and Hans Joachim Bla3

1 Holzforschung Munchen, Winzererstr. 45, 80797 Munchen
Delft University of Technology, Timber Structures and Wood Technology,
Stevinweg 1, 2628 CN Delft
2 Karlsruhe Institute of Technology, Timber Structures and Building Construction,
R.-Baumeister-Platz 1, 76131 Karlsruhe
Delft University of Technology, Timber Structures and Wood Technology,
Stevinweg 1, 2628 CN Delft
3 Karlsruhe Institute of Technology, Timber Structures and Building Construction,
R.-Baumeister-Platz 1, 76131 Karlsruhe

Abstract. Present-day very high strength steel (vhss) grades show high plastic
deformation capacity whilst reaching tension strengths of up to 1400 MPa. These
properties open new application fields in timber engineering. Replacing mild steel
dowels in timber joints with vhss dowels should lead to higher load carrying capaci-
ties or to leaner joints (thinner dowels and smaller cross sections) without losing the
desired joint failure mode with one or two plastic hinges per shear plane. Extensive
test series on double-shear timber joints with slotted-in steel plates have been car-
ried out using 12 mm and 24 mm dowels. The chosen timber species were spruce,
beech and azobe. One, three and five dowels in a row were tested and the used dowel
steel grades were high strength steel (hss) with a mean tension strength of 590 MPa
and vhss with a mean tension strength of 1390 MPa. The test outcomes have shown
that joints with vhss dowels reach a higher load carrying capacity than the same
joints using hss dowels, but are still able to develop plastic hinges. No correlation
between density, load carrying capacity and stiffness within one wood species could
be found. The effective number of fasteners showed a trend to be lower for the joints
with vhss dowels and at the same time, is dependent on the used wood species as
generally, ductile species such as beech show large deformations and subsequently
high load carrying capacities if one dowel is used.

Keywords: steel-to-timber joints, high strength steel dowels, spruce, beech, azobe,
plastic hinges, failure modes.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 157
DOI: 10.1007/978-94-007-7811-5_15, RILEM 2014
158 J.-W. van de Kuilen, C. Sandhaas, and H.J. Bla

1 Introduction
In timber joints with dowel-type fasteners, the preferred failure mode is combined
failure of yielding of the fastener (one or two plastic hinges per shear plane) and
embedment of the timber. Apart from the geometry, the load carrying capacity is
therefore defined by the embedment strength of the timber and the yield moment of
the dowel-type fastener.
The utilisation of very high strength steels (vhss) with subsequent higher yield
moments is promising in order to optimise joints and get high performance joints.
Thinner dowels and thinner member sections should lead to the same load carrying
capacity as have thicker mild steel dowels with bigger member cross sections - as
long as failure mode 2 or 3 is reached (one or two plastic hinges per shear plane).
Another option is that fewer dowels could be used to obtain the same performance.
As for the plastic deformation capacity of the joints, very high strength steels avail-
able nowadays are able to ensure a good overall ductile behaviour without brittle
dowel failure modes.
This is especially valid and even more advantageous for high density timber. The
embedment strength of species with high densities is higher. Therefore, less member
thickness would be needed when using vhss dowels in comparison to softwood.

2 Review
The database of test results on timber joints with dowel-type fasteners, both dowels
and bolts, is huge [5, 8, 12, 16]. Most of the research was focussed on softwood and
only some research was done with hardwood [5, 16]. In previous research, hss or
vhss dowels were mainly used to ensure brittle failure modes in joints [14] and not
to investigate possible higher load carrying capacities.
Since 2006, first tests were undertaken at TU Delft to investigate the applicability
of hss dowels in double-shear timber-to-timber joints [6]. Azobe and spruce mem-
bers were assembled with steel dowels of grade S690 and diameters of d = 8, 16, 24,
30 mm. Further tests were carried out on double-shear timber joints with slotted-in
steel plates [7, 11, 15]. In [11, 15], tests were carried out with spruce and 8 mm
hss dowels of grade 12.9 (i.e. ultimate strength of 1200 MPa and yield strength of
1080 MPa) with 1, 3 and 5 dowels in a row. In [7], tests undertaken on double-shear
timber joints with slotted-in steel plates with tropical hardwoods and hss and mild
steel dowels. The dowel diameter was 8 mm and 1, 3 and 5 dowels in a row were
In [9, 10] all above-mentioned research was analysed and results and conclusions
were summarised. The main conclusion with regard to the practicability of joints
with vhss fasteners was that the failure modes could be predicted with the Johansen
equations and plastic hinges could develop in the vhss dowels.
Timber Joints with Very High Strength Steel Dowels 159

3 Materials and Methods

Double-shear timber joints with slotted-in thick steel plates and two different dowel
steel grades were tested. Three different wood species were used, two different
dowel diameters of d = 12, 24 mm and 1, 3 and 5 dowels in a row. Per series, 5
specimens were tested which led to a total of 180 tests.
Test Setup. The tests were carried out according to EN 26891 [4]. The loading
scheme was load-controlled up to 70% of the estimated maximum force and after-
wards displacement-controlled up to failure. Four LVDTs, one on front and rear of
each timber member, were used to record the relative slip between timber members
and steel plate.
Steel Properties. Analoguously to [6, 7, 11, 15], the joints were designed with
the Johansen equations using mean material properties. The yield moment of the
dowels was calculated with Equation (1) from mechanics and not with the empirical
Equation (2) from EC5 [3].

My = fy , (1)
My = 0.3 fu d 2.6 , (2)
where d = dowel diameter, fu = ultimate strength, fy = yield strength

The empirical Equation (2) derived by Blass et al. [1] is based on the assumption
that only at a bending angle of 45, the plastic capacity of the dowel is fully de-
veloped which is never the case for timber joints unless the fasteners are very thin.
Furthermore, Equation (2) was derived for steel grades with a ratio between yield
and ultimate strength of about 0.65 which is not the case for hss and vhss steel dow-
els where this ratio is 0.95 (see Table 1). Especially for large diameter fasteners,
Equation (2) is punishing [10].
Tension tests have been carried out on the used steel dowels to establish their
mechanical properties, see Table 1. The used steel dowels could be assigned to the
grades high strength steel with an ultimate strength between 500 and 700 MPa and
very high strength steel with an ultimate strength between 800 and 1100 MPa.
Wood Properties. The embedment strength parallel-to-grain fh,0 on the other hand
was calculated according to [2] as the embedment strength equation in EC5 was

Table 1 Results of tension tests on steel dowels

Steel grade n fy [MPa] fu [MPa] strain at failure [%]

hss 9 563 590 11.3

vhss 12 1311 1389 9.4
160 J.-W. van de Kuilen, C. Sandhaas, and H.J. Bla

Table 2 End and edge distance, spacing, d = dowel diameter

Loaded end a3,t 7d

Spacing a1 5d
Edge distance a4 3d

considered to yield too conservative results. Equation (3) resulted from a regression
analysis of embedment test results on hardwoods:

fh,0 = 0.102 (1 0.01d) mean (3)

where mean is the mean wood density and d is the dowel diameter.

The chosen timber species were spruce (Picea abies) with a mean density of
mean = 445 kg/m3, beech (Fagus sylvatica) with mean = 715 kg/m3, and azobe
(Lophira alata) with mean = 1120 kg/m3 . Spruce and beech were stored at a 20/65
climate chamber whereas azobe was stored in a wetter climate chamber with 20C
and 85% relative humidity. The final mean wood moisture content was 12% for
spruce, 9% for beech and 21% for azobe.
Design. The member thickness was designed with the above-described assump-
tions for yield moment and embedment strength in such a way that the failure modes
of the joints with hss dowels lay exactly between failure mode 2 and 3 with 1 or 2
plastic hinges per shear plane respectively.
The end and edge distances listed in Table 2 were the same for all specimens and
corresponded to the minimum distances specified in EC5. It should be noted that the
fastener spacing a1 is smaller than the end distance a3,t .
All further information on used materials and methods can be taken from [13].

4 Experimental Results
The test results in terms of moisture content, density, maximum load and Kser (ac-
cording to [4]), all at time of test, are given in Table 3. The increase in load carrying
capacity when using vhss steel dowels is ranging between 10% for the beech joints
with three 12 mm dowels and 69% for the azobe joints with one 24 mm dowel. This
increase in load carrying capacity is expected to be even higher if mild steel dowels
would have been compared with vhss dowels. The wood quality of the spruce joints
with three 24 mm vhss dowels was very low which may explain their low load car-
rying capacity due to their premature splitting. (All results and descriptive statistics
can be found in [13].)
Fig. 1 shows exemplarily the load-slip graphs for the beech joints with 12 mm
vhss dowels. Plastic deformation behaviour is observable with mean deformations
at maximum load of 16 mm for the joints with one dowel (COV 14.5%), 7 mm for
the joints with three dowels (COV 34.9%) and 5 mm for the joints with five dowels
Timber Joints with Very High Strength Steel Dowels 161

Fig. 1 Load-slip graphs for

beech joints with 12 mm
300 five dowels
vhss dowels
three dowels
250 one dowel


Load [kN]



0 5 10 15 20
Displacement [mm]

(COV 17.4%). Failure mode 2 with one plastic hinge per shear plane was observed
in all tests shown in Fig. 1 and as expected, the possible maximum deformation
decreases with increasing number of fasteners.
Effective Number of Fasteners. In Fig. 1, the influence of an effective number of
fasteners can be seen as the load carrying capacity of joints with three (170 kN) and
five (300 kN) dowels did not correspond to multiples of the capacity of a joint with
one fastener (70 kN).
In EC5, the effective number of fasteners ne f is calculated according to Equation
ne f = min a1 (4)
n0.9 4 13d

where n is the number of fasteners, a1 is the fastener spacing and d is the dowel

In EC5, the effective number of fasteners is only depending on the fastener diam-
eter and the spacing, other possible influence factors such as slenderness or steel
grade/density are not taken into acount. For the tested specimens with dowel diam-
eters d of 12 mm and 24 mm and a spacing of 7d, the effective numbers of fasteners
according to Equation (4) are resulting to 2.12 for joints with three dowels and 3.35
for joints with five dowels.
If the EC5 predictions are compared with the effective number of fasteners re-
sulting from the tests as given in Table 4, ne f from tests are generally higher. If
comparing ne f for joints with vhss and hss dowels, it can be observed that espe-
cially for the timbers with higher densities, the effective number for the joints with
vhss dowels is lower than ne f for the joints with hss dowels.
Dependency on Density. Figs. 2 and 3 show the dependency of the load carrying
capacity and the stiffness Kser on the density of the timber members where a uniform
162 J.-W. van de Kuilen, C. Sandhaas, and H.J. Bla

Table 3 Joint test results and ratio of Fmax with vhss dowels over Fmax with hss dowels

Number of Steel Moisture content Density Fmax [kN] Kser [kN/mm]

dowels grade [%] [kg/m3 ]
mean COV mean COV mean COV mean COV
Spruce, 12 mm
1 hss 11.9 2.4 456 3.6 32 8.4 21 34.5
vhss 12.1 1.6 457 5.8 44 3.4 20 32.6
3 hss 12.7 1.6 462 7.0 90 9.4 51 20.9
vhss 12.4 2.4 485 2.7 115 6.3 51 8.4
5 hss 12.5 1.9 460 5.6 136 9.3 52 12.3
vhss 12.6 1.5 478 1.9 169 15.7 68 6.5
24 mm
1 hss 12.2 1.9 424 10.8 110 12.2 126 27.5
vhss 12.0 1.8 413 4.2 165 4.2 118 13.8
3 hss 11.9 1.1 418 3.7 324 5.1 127 22.3
vhss 11.8 1.7 410 3.5 297 18.0 122 6.6
5 hss 12.4 1.7 436 4.4 482 9.3 137 6.1
vhss 12.5 2.3 439 4.7 570 9.4 247 44.0

Beech, 12 mm
1 hss 8.6 3.4 720 1.6 59 3.6 36 40.6
vhss 8.7 3.1 722 2.5 70 6.0 49 44.2
3 hss 8.5 0.7 730 2.2 157 4.6 61 28.8
vhss 8.7 1.4 732 4.8 173 4.9 67 8.4
5 hss 8.8 12.6 718 3.1 240 2.5 110 4.2
vhss 8.5 1.0 703 2.3 294 2.5 129 13.3
24 mm
1 hss 8.7 8.0 718 2.2 208 3.6 114 20.4
vhss 8.5 2.0 710 1.0 300 2.4 157 7.7
3 hss 8.4 0.8 711 1.7 510 5.0 202 17.2
vhss 8.3 1.3 724 0.9 659 5.1 205 2.9
5 hss 9.8 2.9 701 1.0 823 5.3 171 7.2
vhss 9.4 4.5 695 2.3 927 3.5 192 6.3

Azobe, 12 mm
1 hss 21.4 1.4 1115 2.3 57 6.5 45 29.5
vhss 22.0 1.4 1115 2.3 72 2.7 53 12.4
3 hss 21.8 0.9 1117 1.4 128 7.6 39 6.7
vhss 22.9 3.1 1131 4.0 183 7.9 64 11.2
5 hss 21.7 0.8 1115 1.6 233 14.1 132 6.0
vhss 17.2 2.4 1130 3.9 281 13.2 158 22.1
24 mm
1 hss 21.5 5.9 1122 3.1 181 10.6 103 15.1
vhss 21.5 3.1 1154 1.9 306 6.4 201 17.5
3 hss 22.6 10.7 1104 2.3 539 7.2 186 33.3
vhss 21.3 6.6 1116 1.9 706 5.5 246 13.8
5 hss 22.5 6.9 1118 2.4 867 11.0 321 16.8
vhss 15.9 15.7 1101 1.3 1128 16.0 374 40.9

F(vhss)/F(hss) Spruce Beech Azobe

12 mm 24 mm 12 mm 24 mm 12 mm 24 mm
one dowel 1.38 1.50 1.18 1.44 1.26 1.69
three dowels 1.29 0.92 1.10 1.29 1.43 1.31
five dowels 1.25 1.18 1.22 1.13 1.21 1.30
Timber Joints with Very High Strength Steel Dowels 163

Table 4 Effective number of fasteners ne f

Spruce Beech Azobe

hss vhss hss vhss hss vhss
diameter 12 24 12 24 12 24 12 24 12 24 12 24
dowels 2.80 2.94 2.62 1.80a 2.65 2.45 2.48 2.20 2.23 2.98 2.53 2.31
five dowels 4.24 4.38 3.84 3.45 4.04 3.96 4.20 3.09 4.06 4.80 3.90 3.69
a Spruce joints with three 24 mm vhss dowels suffered from premature splitting due to very

low wood quality

200 Beech Azob Number

Spruce dowels
300 300 1

Fmax/dowel [kN]
Fmax/dowel [kN]
Fmax/dowel [kN]

150 3
200 200 Steel grade
24mm dowels 24mm dowels 24mm dowels mild
12mm dowels 100 12mm dowels 100 12mm dowels

0 0 0
380 400 420 440 460 480 500 675 700 725 750 775 1075 1100 1125 1150 1175 1200
Density [kg/m3] Density [kg/m3] Density [kg/m3]

Fig. 2 Density versus maximum load per dowel

250 Spruce 250 Beech 250 Azob Number dowel

measured Kser per dowel
measured Kser per dowel
measured Kser per dowel

200 200 200 5

150 150 150 Diameter

100 100 100

50 50 50

0 0 0
380 400 420 440 460 480 500 675 700 725 750 775 1075 1100 1125 1150 1175 1200
Density [kg/m3] Density [kg/m3] Density [kg/m3]

Fig. 3 Density versus Kser per dowel and two shear planes

load and stiffness distribution between the single dowels was assumed. Within one
wood species, no correlation between density and Fmax and Kser could be found.

5 Conclusions
A comprehensive joint test series including the wood species spruce, beech and
azobe has been carried out using two different steel grades, vhss and hss. The ex-
perimental results are summarised below:
In timber joints, hss dowels can be replaced by vhss dowels. Joints with vhss
reached higher loads than the same joints with hss dowels.
164 J.-W. van de Kuilen, C. Sandhaas, and H.J. Bla

No strong correlation could be found between density of the wood and load car-
rying capacity of the joints using one wood species.
The failure modes can be predicted with the Johansen equations. This is valid
also for different steel grades and wood species.
Ductile failure modes with one or two plastic hinges per shear plane were pos-
sible also when using vhss dowels as modern vhss steel grades possess enough
deformation capacity.
Compared to hss dowels, the failure modes of the joints with vhss dowels were
shifted towards failure modes with less plastic hinges. Furthermore, the bigger
the dowel diameters, the more a shift towards modes with no plastic hinges or
one plastic hinge per shear plane could be observed.
The observed effective number ne f for more dowels in a row showed a trend to
be lower for the joints with vhss dowels. The difference between ne f for hss and
vhss dowels was smaller for beech. This may be due to the higher deformation
capability of beech and hence higher ultimate loads in the joints with one dowel.
On wood species level, the stiffness Kser is not strongly related to the density
of the used wood species and only weakly to the dowel diameter. An effective
number ne f could be used for Kser .

Acknowledgements. Thanks to the companies Pollmeier Massivholz GmbH&Co KG for

donating the beech wood, De Groot Vroomshoop for sponsoring the spruce glulam and to
Groot Lemmer for sponsoring and manufacturing the azobe specimens.

[1] Blass, H.J., Bienhaus, A., Kramer, V.: Effective bending capacity of dowel-type fasten-
ers. In: CIB-W18 Meeting 33, Paper 33-7-5, Delft, The Netherlands (2000)
[2] Ehlbeck, J., Werner, H.: Softwood and hardwood embedding strength for dowel-type
fasteners. In: CIB-W18 Meeting 25, Paper 25-7-2, Ahus, Sweden (1992)
[3] EN1995-1-1: Eurocode 5. Design of timber structures - Part 1-1: General - Common
rules and rules for buildings (2004)
[4] EN26891: Timber structures - Joints made with mechanical fasteners - General princi-
ples for the determination of strength and deformation characteristics (ISO 6891) (1991)
[5] Gehri, E., Fontana, M.: Betrachtungen zum Tragverhalten von Passbolzen in Holz-Holz-
Verbindungen. Tech. Rep. Int. Bericht 83-1, Institut fur Baustatik und Stahlbau, ETH
Zurich, Switzerland (1983)
[6] Hieralal, R.: Application of high strength steel in steel pin connections and double-shear
timber joints. Master thesis, Faculty of Civil Engineering, University of Technology
Delft, The Netherlands (2006)
[7] Islamaj, L.: Application of high strength steel dowels in steel-to-timber double-shear
joints with slotted-in steel plate. Master thesis, Faculty of Civil Engineering, University
of Technology Delft, The Netherlands (2009)
[8] Jorissen, A.: Double-shear timber connections with dowel-type fasteners. Dissertation,
University of Technology Delft, The Netherlands (1998)
Timber Joints with Very High Strength Steel Dowels 165

[9] Van de Kuilen, J.W.G.: Verbindungsmittel aus hochfesten Stahlen - Stabdubelver-

bindungen. In: Tagungsband Ingenieurholzbau - Karlsruher Tage: Forschung fur die
Praxis, Karlsruhe, Germany (2009)
[10] Van de Kuilen, J.W.G., De Vries, P.A.: Timber joints with high strength steel dowels.
In: 10th World Conference of Timber Engineering WCTE, Miyazaki, Japan (2008)
[11] Langedijk, A.: Houtverbindingen met hoge sterkte staal. Bachelor thesis, Faculty of
Civil Engineering, University of Technology Delft, The Netherlands (2007)
[12] Mohammad, M., Quenneville, J.H.P.: Behaviour of wood-steel-wood bolted glulam
connections. In: CIB-W18 Meeting 32, Paper 32-7-1, Graz, Austria (1999)
[13] Sandhaas, C.: Mechanical behaviour of timber joints with slotted-in steel plates. Dis-
sertation, University of Technology Delft, The Netherlands (2012)
[14] Schmid, M.: Anwendung der Bruchmechanik auf Verbindungen mit Holz. Dissertation,
University of Karlsruhe, Germany (2002)
[15] Van Groesen, J., Kranenburg, M.: Houtverbindingen met hoge sterkte staal. Bachelor
thesis, Faculty of Civil Engineering, University of Technology Delft, The Netherlands
[16] Werner, H.: Untersuchungen von Holz-Verbindungen mit stiftformigen Verbindungs-
mitteln unter Berucksichtigung streuender Einflussgrossen. Dissertation, University of
Karlsruhe, Germany (1993)
Wood Load-Carrying Capacity of Timber
Connections: An Extended Application for Nails
and Screws

Pouyan Zarnani and Pierre Quenneville

Department of Civil and Environmental Engineering, Faculty of Engineering,

University of Auckland
pzar004@aucklanduni.ac.nz, p.quenneville@auckland.ac.nz

1 Motivation

The wood engineering community has dedicated a significant amount of effort

over the last decades to establish a reliable predictive model for the load-carrying
capacity of timber connections under wood failure mechanisms. Test results from
various sources (Foschi and Longworth 1975; Johnsson 2003; Quenneville and
Mohammad 2000; Stahl et al. 2004; Zarnani and Quenneville 2012a) demonstrate
that for multi-fastener connections, failure of wood can be the dominant mode.
In existing wood strength prediction models for parallel to grain failure in
timber connections using dowel-type fasteners, different methods consider the
minimum, maximum or the summation of the tensile and shear capacities of the
failed wood block planes. This results in disagreements between the experimental
values and the predictions. It is postulated that these methods are not appropriate
since the stiffness in the wood blocks adjacent to the tensile and shear planes
differs and this leads to uneven load distribution amongst the resisting planes
(Johnsson 2004; Zarnani and Quenneville 2012a).
The present study focuses on the nailed connections. A closed-form analytical
method to determine the load-carrying capacity of wood under parallel-to-grain
loading in small dowel-type connections in timber products is thus proposed. The
proposed stiffness-based model has already been verified in brittle and mixed
failure modes of timber rivet connections (Zarnani and Quenneville 2013b).

Keywords: nailed connections, rivet connections, brittle, ductile and mixed failure
modes, elastic spring model, ultimate capacity, design model.

2 Stiffness-Based Predictive Model

The proposed analysis for wood strength is best explained using the analogy of a
linear elastic spring system in which the applied load transfers from the wood

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 167
DOI: 10.1007/978-94-007-7811-5_16, RILEM 2014
168 P. Zarnani and P. Quenneville

member to the failure planes in conformity with the relative stiffness ratio of each
resisting adjacent volume to the individual failure plane (Fig. 1). By predicting
these volumes stiffness, one can derive the portion of the connection load that is
channelled to each resisting plane and from the resistance of each failure planes,
one can determine which failure plane triggers the connection failure.

Fig. 1 Proposed elastic spring model

The difference in the loads channelled to the tensile and shear planes is a
function of the modulus of elasticity and modulus of rigidity, the volume of wood
surrounding each of the failure planes (bottom, end and edge distances-dz, da and
de) and also the connection geometry (Fig. 2). For details regarding the
determination of stiffness of the resisting planes, refer to Zarnani and Quenneville

Fig. 2 Simplified analytical model

By predicting the stiffness of the wood surrounding each of the failure planes
(Kh, Kb and Kl), one can predict the proportion of the total connection load applied
to each plane, Ri = K i K . By further establishing the resistance of each of the
failure planes as a function of a strength criterion, one can verify which of the
Wood Load-Carrying Capacity of Timber Connections 169

failure planes governs the resistance of the entire connection. It should be asserted
that the strength of the shear planes cannot be higher than the tensile capacity of
the adjacent wood volume where the load is channelled to these resisting planes
(Fig. 3). If the attracted load by the resisting shear planes be larger than the tensile
capacity of in-contact wood volume, then the wood block torn out from the
member would be as wide as or as deep as the member corresponding to wood
failure mode (b) and (c) (Fig. 4).

Fig. 3 Loads acting on the wood volume adjacent to the shear resisting planes: (a) bottom
block, (b) lateral blocks

Thus, the wood load carrying capacity of the connection (Eq. 1) is the load
which results in the earlier failure of one of the resisting planes due to being
overloaded and equals to the minimum of Pwh, Pwb and Pwl. It is important to note
that the connection resistance given by Eq. 1 is a summation of the critical plane
failure load plus the load carried by the other planes.

Kb Kl
Pwh = f t ,m Ath (1 + + ) , Mode (a)
Kh Kh

K h Kl f v,m C ab Asb , Mode (a)

Pw = np . min Pwb = (1 + + ). min (1)
Kb Kb f t ,m X l d z , Mode (c)
K h Kb f C A , Mode (a)
Pwl = (1 + + ). min v,m al sl
Kl Kl 2 f t ,m t ef d e , Mode (b)

In Eq. 1, ft,m and fv,m are the wood mean strength in tension and shear along the
grain (MPa). Ath, Asb and Asl are the areas of the head, bottom and lateral resisting
planes with respect to the wood effective thickness, tef, subjected to tension and
shear stresses. Also, Cab and Cal are the ratios of the average to maximum stresses
on the bottom and lateral shear planes respectively (Zarnani and Quenneville
2013b) and Xl is the joint width. For a one-sided joint having only one plate, np=1,
the member thickness value, b, to be used to determine dz = b/2-tef is twice the
thickness of the wood.
170 P. Zarnani and P. Quenneville


Mode (a) Mode (b) Mode (c)

Fig. 4 Different possible failure modes of wood block tear-out

It should be noted that when one plane fails, then the entire connection load
transfers to the remaining planes in accordance with their relative stiffness ratios.
It could be possible that the occurrence of the first failure of one plane does not
correspond with the maximum load of the connection. This is more susceptible in
case of either small edge or bottom distances accompanying large shear resisting
areas which leads to wood failure mode (b) and (c) respectively (Fig. 4).
Therefore, Eq. 1 needs to be checked again for the remaining planes by defining
no value for the terms related to the failed planes.
In the case of fasteners which are inserted into predrilled holes, the area
corresponding to the cutting diameter is to be subtracted from the resisting plane
surfaces. This affects the strength of the tensile and shear resisting planes and not
their stiffnesses.

3 Design Procedure

The design of timber joints using small dowel-type fasteners such as rivets, nails
and screws is governed by either the brittle, mixed or ductile failure mode of the
joint. The occurrence zone of these potential failure modes is illustrated on a
typical load-displacement curve of a timber joint (Fig. 5).
In the brittle zone, the fasteners deflection is in the elastic range, therefore, the
effective wood thickness for the joint corresponds to the elastic deformation of the
fasteners, tef,e. In this failure zone, the wood capacity of the connection, Pw,tefe, is
less than the fastener yielding resistance, Pr,yld. It should be noted that the Pr,yld is
not an ultimate failure but constitute a boundary. As the yield point is reached, the
effective wood thickness reduces if the yield mode is not Mode I. This reduction
in effective wood thickness, tef,y, leads to the generation of a new connection
failure mode. If the wood capacity of the new connection, Pw,tefy, cannot resist the
Wood Load-Carrying Capacity of Timber Connections 171

Fig. 5 Occurrence zone of potential failure modes of timber joints

fastener yielding load, a sudden failure with slight deflection on the fasteners
which is called mixed failure mode occurs. Even if Pw,tefy > Pr,yld, the mixed failure
mode can happen as the deflection of the connection progresses if Pw,tefy is lower
than the connection ultimate ductile strength, Pr,ult. If the wood strength based on
tef,y is greater than Pr,ult, the ductile failure governs and there is no wood rupture.
By following the described mechanism for the potential failure modes, the
connection ultimate capacity, Pc,ult, can be predicted as follows (Zarnani and
Quenneville 2013a);

Pw,tefe if Pw,tefe < Pr,yld (Brittle mode)

Pr,yld if Pw,tefy < Pr,yld Pw,tefe (Mixed mode)
Pc,ult = Pw,tefy if Pr,yld Pw,tefy Pr,ult (Mixed mode) (3)
Pr,ult if Pr,ult < Pw,tefy (Ductile mode)

If a designer wants to rely only on the yield limit state as the connection
maximum capacity, therefore, the above design procedure can be simplified to
Pc,ult = min (Pw,tefe, Pr,yld).

4 Material and Experimental Method

4.1 BNG, MNG and DNG Test Series on LVL

The laboratory tests (Fig. 6) involved specimens with nail configurations designed to
observe the three possible modes of failure and to evaluate the effect of geometry on
connection strength. Specimens were manufactured from 90*200*1200mm and
172 P. Zarnani and P. Quenneville

90*240*1200mm New Zealand Radiata Pine LVL grade 11. The tests series
involved 8 groups with 3 replicates for each configuration. Two sizes of nail were
used; nails with a shank length of 30mm and a diameter of 3.2mm (Lp=22mm) and
nails with a shank length of 33mm and a diameter of 3.8mm, (Lp=25mm). The
geometric parameters of connections (Fig. 7) tested varied from 8 to 10 for number
of rows (nR) and 5 to 10 for number of columns (nC) as shown in Table 1. Sp of 32,
38 and 72mm (Sp 10d) and Sq of 8 and 10mm (Sq 5d) were adopted conforming
with NZS3603 (Standards New Zealand 1993). The holes in the nail plate along the
columns were staggered by half of the column spacing (Sp) to increase the amount of
nails for the failure block size while still conforming to the codes minimum spacing
perpendicular to grain. Edge distance, de varied from 55 to 92mm and end distance
from 50 to 100mm (da >12d). The specimens had nailed plates on both faces of
timber, resulting in a symmetric connection. The nail plates were 8 mm thick of 300
grade steel with predrilled 3.3mm and 3.9mm holes to ensure the head of the nail
was rotationally fixed. Nail holes on the wood were not pre-drilled. In six of the
groups, the characteristics were specified in order to prompt wood failure and
maximize the amount of observations on the brittle and mixed failure mechanism.
All specimens were conditioned to 20C and 65% relative humidity to attain a
target 12% equilibrium moisture condition (EMC). The wood had an average
density of 607 kg/m3 with a coefficient of variation of 2% at the time of the tests
and an average moisture content of 13.2%. The wood mean tensile and shear
strengths (ft,m, fv,m) used in the analysis were determined from the characteristic
values provided by the manufacturer. A coefficient of variation of COV=15% and
COV=10% for the wood tensile and shear strengths respectively was used to back
calculate the average values. The estimated ft,m and fv,m were 39.8 MPa and 7.2
MPa correspondingly.

Fig. 6 Typical specimen Fig. 7 Definition of connection

in testing apparatus geometry variables
Wood Load-Carrying Capacity of Timber Connections 173

4.2 REC, GRP, NOR, SPR and TEN Test Series on Glulam
The REC, GRP, NOR, SPR and TEN test series presented in Table 1 were
conducted by Johnsson (2003). A similar test setup was used as shown in Fig. 7.
The specimens cross sections were 90*225 to 360 mm from glulam of strength
class GL28c (produced from Norway Spruce). In the RECX series, the thickness
of the specimen varied between 66, 78, 90, and 140 mm with 5 replicates in each
group. The GRP series had nails placed in two groups with a gap in between equal
to 75, 150 and 300 mm for GRPS, GRPL and GRPX respectively. All the joints
consisted of one steel plate on one side. The nail used had a diameter of 4.0 mm
and penetration depth of Lp=40 mm. Nails were inserted in pre-drilled holes for all
the test series.
The wood had an average density of 454 kg/m3 and an average moisture
content of 11.2% (Johnsson 2004). The wood mean tensile strength reported in
Johnsson (2004) and the mean shear strength found by Crocetti et al (2010) for the
same timber class of GL28c were used as inputs to the proposed model. The ft,m
and fv,m were 28.4 MPa and 4.9 MPa respectively. For the stiffness properties,
based on data available in the literature, an average ratio of modulus of rigidity to
modulus of elasticity (G/E) is considered equal to 0.045 and 0.069 for LVL and
glulam respectively in order to make the planes stiffness equations independent
of G and E values.

5 Results and Discussion

5.1 Test Observations

The load slip curve for each specimen tested in Auckland was plotted and the
ultimate load and the types of failure were recorded. The peak loads ranged from
180 kN to 324 kN. When brittle or mixed failure occurred, the connection did not
develop the full ductile capacity of the fasteners and the ultimate capacity of the
connection was dictated by plug shear failure initiation. Specimens that failed
brittlely experienced displacements of approximately 1.5mm. Specimens for
which a mixed failure was observed suffered slightly more slip, experiencing
minor plastic deformation of the fasteners before plug shear failure occurred at a
slip of 1.5-2.5mm. Brittle and mixed failure occurred in tests groups with a large
number of tightly spaced nails (Fig. 8). When a brittle plug shear failure occurred,
a block of wood contained within the nail group was pulled away from either one
side or both sides of the specimens. In the case of ductile failures, the connections
were able to reach the full capacity of the nails. Before the ultimate capacity was
reached, the connections experienced large deformations, far beyond yield
(defined by the 5%-offset method) and design serviceability allowances.
The specimens were named based on the observed modes of failure. BNG,
MNG and DNG stand for nail group configurations with brittle, mixed and ductile
modes of failure respectively.
174 P. Zarnani and P. Quenneville

Fig. 8 Specimen and failed block exhibiting the brittle/mixed modes of wood failure

Table 1 Configuration of the nailed joints on LVL and glulam

Spacing of
Nail diameter
No. of rows rows
Test and penetration Member Member End
and columns and columns thickness width distance
groups* (mm)
b (mm) W (mm) da (mm)
nR nC Sp Sq d Lp
BNG1-L 10 10 32 8 3.2 22 90 200 50
MNG2-L 10 10 32 8 3.2 22 90 200 50
MNG3-L 8 10 32 8 3.2 22 90 200 50
MNG4-L 8 10 32 8 3.2 22 90 240 50
MNG5-L 8 10 32 8 3.2 22 90 200 100
DNG6-L 8 6 32 8 3.2 22 90 200 50
MNG7-L 10 8 38 10 3.8 25 90 200 50
DNG8-L 10 5 72 10 3.8 25 90 200 50
RECS-G 9 7 28 7 4.0 40 90 225 66
RECL-G 19 8 28 7 4.0 40 90 265 80
RECX0-G 19 15 28 7 4.0 40 66 265 60
RECX1-G 19 15 28 7 4.0 40 78 265 60
RECX2-G 19 15 28 7 4.0 40 90 265 60
RECX4-G 19 15 28 7 4.0 40 140 265 60
GRPS-G 19 10 28 7 4.0 40 90 265 56
GRPL-G 19 13 28 7 4.0 40 90 265 56
GRPX-G 19 18 28 7 4.0 40 90 265 56
NORS-G 7 5 40 10 4.0 40 90 225 60
NORL-G 13 6 40 10 4.0 40 90 265 60
NORX-G 15 11 40 10 4.0 40 90 280 60
SPR-G 19 8 40 7 4.0 40 90 265 56
TENS-G 13 2 28 7 4.0 40 90 225 60
TENL-G 33 2 28 7 4.0 40 90 360 60
L and G stand for LVL and glulam respectively. In LVL groups, the joints are double-
sided (np=2) and in glulam one-sided.
Wood Load-Carrying Capacity of Timber Connections 175

5.2 Wood Effective Thickness

The thickness of the wood block pulled out of the brittle and mixed failure mode
specimens was measured. The thickness was compared to the predicted effective
thickness, tef (Table 2). For the brittle failure mode, the elastic deformation of the
nail modelled as a beam on an elasto-plastic foundation was assumed to determine
the tef,e (Zarnani and Quenneville 2012c) and in the case of mixed failure, the tef,y
was predicted based on the bearing length corresponding to the governing yielding
mode of the nail (Zarnani and Quenneville 2013b).

Table 2 Experimental results on wood effective thickness compared to predictions

Nail tef,e (mm) tef,y (mm)

diameter, for brittle failure mode for mixed failure mode
length, Lp
d Average Average
(mm) predicted predicted
(mm) observed observed
3.2 22 0.87Lp = 19.1 17.5 16.4 14.6
3.4 25 0.84Lp = 21.0 - 18.7 19.9
4.0 40 0.75Lp = 30.0 27.5* 15.1 -
Average of 25-30 mm, based on the tests conducted by Johnsson (2003).

The average measured block depth was within 2mm of the predicted effective
thickness for both the brittle and mixed failure modes. For the mixed failure
modes, there was a visible deformation of the nails, in the form of a plastic hinge
formed at the plate (Fig. 9). This plastic hinge is consistent with the predicted
ductile yield failure mode IIIm using Johansens yield theory (1949) which is the
foundation for the European Yield Model (EYM) prediction formulas in Eurocode
5 (2004). As shown in Table 2, a good agreement can be found between the
predictions and observations for the effective wood thickness.

Fig. 9 Wood mixed failure showing deformed nails

176 P. Zarnani and P. Quenneville

5.3 Validation of the Proposed Stiffness-Based Model

Results for the current LVL groups tested and tests on glulam reported in
Johnsson (2003) are listed in Table 3 along with the predominant modes of failure
observed. The connection ultimate capacities were calculated using the proposed
stiffness-based model and the algorithm presented to determine the connection
failure mode. As shown in Table 3, for the BNG test group, the estimated wood
strength corresponding to the nail elastic deformation, Pw,tefe, was lower than the
nail yielding resistance, Pr,yld. For this configuration, the connection ultimate
capacity is predicted as Pc,ult=Pw,tefe with a brittle failure mode which is consistent
with the observation. However, in the MNG test groups, the predicted wood
strength for tef,e is higher than the nails yielding strength. The strength of the
connection is thus checked for the possible mixed or ductile modes of failure. In
these test series a mixed mode failure occurred which can be explained by the
wood strength corresponding to tef,y being weaker than the nails ultimate strength
(Pw,tefy < Pr,ult). In the MNG2 test group, since Pw,tefy is greater than Pr,yld, thus the

Table 3 Nailed connection ultimate strength prediction using the proposed analysis
compared to experimental results on LVL and glulam

Wood strength Nail Wood strength Nail
Ultimate strength
corresponding yielding corresponding ultimate Failure mode
Test Pc,ult (kN)
to nail elastic strength to nail strength
groups* Ratio
deformation Pr,yld yielding mode Pr,ult
Pw,tefe (kN) (kN) Pw,tefy (kN)* (kN)* Predicted Observed of pred. Pred. Obser.
(COV%) to
BNG1-L 248 271 N/A N/A 248 298 (9%) 0.83 BRT BRT
MNG2-L 262 244 250 338 250 263 (3%) 0.95 MIX MIX
MNG3-L 230 216 205 N/A 216 231 (4%) 0.94 MIX MIX
MNG4-L 230 216 205 N/A 216 228 (1%) 0.95 MIX MIX
MNG5-L 214 216 N/A N/A 214 263 (7%) 0.81 BRT MIX
DNG6-L 222 130 192 180 180 180 (5%) - DUC DUC
MNG7-L 305 272 293 440 293 324 (5%) 0.90 MIX MIX
DNG8-L 292 170 283 275 275 275 (4%) - DUC DUC
RECS-G 73 136 N/A N/A 73 88 (9%) 0.83 BRT BRT
RECL-G 138 329 N/A N/A 138 162 (5%) 0.85 BRT BRT
RECX0-G 199 634 N/A N/A 199 250 (9%) 0.80 BRT BRT
RECX1-G 198 634 N/A N/A 198 200 (12%) 0.99 BRT BRT
RECX2-G 196 634 N/A N/A 196 256 (15%) 0.77 BRT BRT
RECX4-G 196 634 N/A N/A 196 255 (9%) 0.77 BRT BRT
GRPS-G 168 329 N/A N/A 168 181 (7%) 0.93 BRT BRT
GRPL-G 184 329 N/A N/A 184 217 (5%) 0.85 BRT BRT
GRPX-G 218 329 N/A N/A 218 229 (8%) 0.95 BRT BRT
NORS-G 88 81 77 N/A 81 97 (5%) 0.84 MIX BRT
NORL-G 164 166 N/A N/A 164 178 (6%) 0.92 BRT BRT
NORX-G 229 361 N/A N/A 229 293 (5%) 0.78 BRT BRT
SPR-G 187 329 N/A N/A 187 253 (3%) 0.74 BRT BRT
TENS-G 46 48 N/A N/A 46 59 (4%) 0.78 BRT BRT
TENL-G 123 152 N/A N/A 123 136 (11%) 0.90 BRT BRT
Not applicable (N/A) for all the test groups based on the presented design algorithm.

BRT, MIX and DUC stand for brittle, mixed and ductile failure modes, correspondingly.
Wood Load-Carrying Capacity of Timber Connections 1777

connection ultimate cap pacity is determined as Pc,ult=Pw,tefy. However, foor

the MNG3 and MNG4 test series where Pw,tefy < Pr,yld, Pc,ult is predicted as Pr,yld. IIn
the case of the DNG test groups,
g the wood strength based on tef,y is greater than thhe
connection ultimate ductile strength, Pr,ult, therefore, the ductile failure governns
(Pc,ult=Pr,ult) and there wass no wood rupture.
For the REC, GRP, NO OR, SPR and TEN test series (Johnsson 2003), since thhe
nails were inserted into predrilled holes, the area corresponding to the cuttinng
diameter (0.8d) was subtrracted from the resisting plane surfaces. This affected thhe
strength of the tensile and d shear resisting planes and not their stiffnesses. For thhe
GRP series in which the nails
n were placed in two groups with a gap in between, it
was assumed that the gap g is filled with the nails by the same pattern (thhis
assumption was only app plied for defining the wood capacity). As shown in Tabble
3, the predicted wood strength
s corresponding to the nail elastic deformation,
Pw,tefe, was lower than th he nails yielding resistance, Pr,yld (except for the NO OR
group). Therefore, Pc,ult=P= w,tefe and the observed thickness of the brittle faileed
blocks is comparable to the t bearing length based on the nail elastic deformatioon
(Table 2). The yield and ultimate capacities of the nails with 3.2 and 3.8 mm iin
diameter were derived basedb on the conducted tests and based on the valuees
reported in Johnsson (20 003) for the tests with nails of 4.0mm diameter. Thhe
yielding strength used in the analysis were 1.4, 1.7 and 2.3 kN for the 3.2, 3.8 annd
4.0 mm nails diameter resspectively and for the ultimate resistance, 1.9, 2.8 and 3.2
kN correspondingly. For all the experimental groups with the brittle/mixeed
fashion, the failure modess of the wood block tear-out was predicted to be mode (a)
(Fig. 4) which was in linee with the test observations.

Fig. 10 Comparison between

n predictions and test data in brittle/mixed failure modes
178 P. Zarnani and P. Quenneville

As shown in Table 2 and 3, there is very good conformity between the

predictions and observations for the thickness of the failed block, the governing
failure mode, and the strength of the connection. Fig. 10 shows the strength
predictions of the experimental groups compared to the tests results. One can note
that the proposed analysis results in precise predictions with a correlation
coefficient (r2) of 0.92, a mean absolute error (MAE) of 16.6% and a standard
deviation (STDEV) of 9.4%. It is assumed that this method of calculation could
also be applicable to screwed connections.

6 Conclusions

The closed form predictive model for determining the wood capacity of rivet
connections in glulam and LVL, developed by Zarnani and Quenneville (2013b),
can be applied to nailed connections. The method takes into account the strength
of the failure planes and the stiffness of the adjacent wood which distributes the
member load to these planes. Also, a design algorithm is presented which allows
the designer to calculate the resistances associated with the predictions of the
different possible brittle, ductile and mixed failure modes. Results of nailed joint
tests on LVL and the test data available from Johnsson (2003) on glulam confirm
the validity of this new method and show that it can be used as a design provision
for the wood load-carrying capacity prediction in timber connections. The
proposed method can be extended to other small dowel type fastener such as
screws to predict accurately the connection ultimate capacity and its failure mode.

Acknowledgements. The authors wish to thank the New Zealand Structural Timber
Innovation Company (STIC) for funding this research work. Likewise the authors would
like to express their gratitude to The University of Auckland undergraduate students
Alexander Jessep and Andrew McQueen, who did the experimental work as part of their
final year project.

Crocetti, R., Gustafsson, P., Danielsson, H., Emilsson, A., Ormarsson, S.: Experimental and
numerical investigation on the shear strength of glulam. In: Proc., International Council
for Research and Innovation in Building and Construction, CIB-W18, New Zealand,
paper 43-12-5 (2010)
European Committee for Standardization (CEN). Eurocode 5-Design of timber structures.
EN 1995-1-1:2004, Brussels, Belgium (2004)
Foschi, R.O., Longworth, J.: Analysis and design of griplam nailed connections. J. Struct.
Div. ASCE 101(12), 25372555 (1975)
Johansen, K.W.: Theory of timber connections. Publications of International Association
for Bridge and Structural Engineering 9, 249262 (1949)
Johnsson, H.: Plug shear failure in nailed timber connections: experimental studies. In:
Proc., International Council for Research and Innovation in Building and Construction,
CIB-W18, Colorado, paper 36-7-2 (2003)
Wood Load-Carrying Capacity of Timber Connections 179

Johnsson, H.: Plug shear failure in nailed timber connections. Doctoral dissertation,
Department of Civil and Environmental Engineering, Lulea University of Technology,
Sweden (2004)
Quenneville, P., Mohammad, M.: On the failure modes and strength of steel-wood-steel
bolted timber connections loaded parallel-to-grain. Can. J. Civ. Eng. 27(4), 761773
Stahl, D.C., Wolfe, R.W., Begel, M.: Improved analysis of timber rivet connections. J.
Struct. Eng. ASCE 130(8), 12721279 (2004)
Standards New Zealand. NZS 3603:1993 Timber Structures Standard, Wellington, New
Zealand (1993)
Zarnani, P., Quenneville, P.: A stiffness-based analytical model for wood strength in timber
connections loaded parallel to grain: Riveted joint capacity in brittle and mixed failure
modes. In: Proc., International Council for Research and Innovation in Building and
Construction, CIB-W18, Sweden, paper 45-7-1 (2012a)
Zarnani, P., Quenneville, P.: Reliable yield model for strength prediction of timber rivet
connection under ductile failure. In: Proc., 12th World Conference on Timber Eng.,
Auckland, New Zealand (2012b)
Zarnani, P., Quenneville, P.: Wood effective thickness in brittle and mixed failure modes of
timber rivet connections. In: Proc., 12th World Conference on Timber Eng., Auckland,
New Zealand (2012c)
Zarnani, P., Quenneville, P.: Design procedure to determine the capacity of timber
connections under potential brittle, mixed and ductile failure modes. In: Proc.,
International Council for Research and Innovation in Building and Construction, CIB-
W18, Vancouver (2013a)
Zarnani, P., Quenneville, P.: Wood block tear-out resistance and failure modes of timber
rivet connections - A stiffness-based approach. Journal of Structural Engineering, ASCE
(in press, 2013b)
Ductility in Timber Structures: Investigations
on Over-Strength Factors

Frank Bruhl1 , Jorg Schanzlin2, and Ulrike Kuhlmann1

1 Universitat Stuttgart, Institute of Structural Design, Pfaffenwaldring 7,
70569 Stuttgart, Germany
2 Konstruktionsgruppe Bauen AG, Bahnhofplatz 1, 87435 Kempten, Germany

Abstract. This paper presents a study on the implementation of connection ductil-

ity in timber structures. Regardless for which purpose ductility in timber structures
is needed, it is necessary to avoid a brittle failure of the timber element before the
ductile element is in the stage of yielding. An over-strength factor is introduced
to consider the required distance of the load-bearing resistance of the beam ele-
ment from the introduced bending moment initiated by the load-carrying capacity
of the fasteners. Hence, a Monte-Carlo simulation was conducted, focusing partic-
ularly on the scattering of the material properties to determine a reliability index of
a joint loaded in bending. Since the reliability index is based on the application of
the ductility, a range of over-strength factors is given for different reliability indices.
The Monte-Carlo simulation is based on the mean material properties of the exper-
imental specimens. The experiments are explained and the non-linear behaviour is
displayed not only for connections loaded in tension but also for joints loaded in

Keywords: dowel type connections, reinforcement, over-strength factor, reliability,

Monte-Carlo simulation.

1 Introduction
The question of ductility of structural elements has attracted more and more interest
in recent years. The demand for ductile connections is widespread. The moment-
rotation behaviour of a connection is important in seismic design [1], in the con-
sideration of the redistribution of internal forces [2, 5.1(3)] or if robustness [3] is
demanded. However, no information about the non-linear behavior of fasteners is
given [2]. Therefore it is necessary to gain knowledge about the ductile behaviour
of connections and joints in timber structures.
Several types of connections behave in a ductile manner; in particular dowel type
fasteners generally show a significant ductile behaviour [4] if timber is not at risk
of splitting within the connection [5, 6]. By the introduction of simplified nonlinear
load-slip relations of different types of fasteners, it is possible to integrate the ductile

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 181
DOI: 10.1007/978-94-007-7811-5_17,  c RILEM 2014
182 F. Bruhl, J. Schanzlin, and U. Kuhlmann

behaviour for practical applications as already included in EN 1995-1-1 [2, 5.1(3)].

The inherent material properties of timber only allow the formation of a plastic hinge
within the joint. Therefore it is indispensable to introduce an over-strength factor,
which ensures that the joint is in the state of yielding before brittle failure occurs.

2 Ductile Behaviour of Dowel-Type Connections

2.1 Material
The width of all timber members was 180 mm, with an associated timber grade
of GL24h. A homogenous lamination was chosen to ensure that all dowels were
embedded in the same timber grade. The mean value of density in the connection
experiments was 443.5 kg/m3 and 444.4 kg/m3 for the joint test setup.
Previous investigations on dowel-type connections showed that the tensile strength
is usually higher than the target value of the ordered steel grade [7, 8]. Within these
investigations, the dowels had an ordered steel grade of S235 and a diameter of 12
mm. The tensile strength was tested at the beginning of the test series to ensure that
all of the dowels belonged to the same lot. Again, an enhanced value of 581 N/mm2
was ascertained, which is outside the range of 360-510 N/mm2 given in [9].
All of the experiments conducted were displacement-controlled following EN
26891 [10]. Failure was defined as a rupture of the specimens or as a decrease of the
load to 80% of the maximum load [11].

2.2 Connection Tests

The connection tests were conducted on three different dowel arrangements with a
slotted plate connection. All of the specimens were reinforced with fully
threaded screws and tested under a tension
loading. The examined dowel group was set
at the bottom of the specimen; a steel cover
was attached on the top of the specimens with
fully threaded screws (see Figure 1). There-
fore it was ensured that only one connection
was tested until the defined failure occurred,
regardless of the single-load-carrying capac-
ity. The dowel arrangements varied from a
wide alignment of 24 dowels to a rather
stretched arrangement of 52 dowels. The
first number indicates the number of dowels
parallel to the grain and the second the num-
ber of dowels perpendicular to the grain. Fig-
ure 2 shows the mean test results, based on
four experiments conducted for the different
Fig. 1 Connection test setup dowel arrangements. All of the experiments
Ductility in Timber Structures: Investigations on Over-Strength Factors 183

showed a constant ductile behaviour. Hence the ductility was evaluated based on the
definition Df = uf /uy [12]. A ductility ratio of 9 to 9.5 could be achieved. The con-
nections are therefore classified as highly ductile [8]. The displacement at yielding
(uy ) was determined by the regulations of [11, 13], and uf describes the displace-
ment at failure. If the classification is not only based on the relative value Df , but
also on the absolute value Dfy = uf -uy , a plastic displacement of more than 20 mm
could be achieved [14].
In order to compare the different dowel arrangements, the applied load was nor-
malized to the bearing resistance calculated based on the initial material properties
of one fastener [15]. It is shown that the load-carrying capacity is almost identical
for one fastener within each arrangement. Therefore it can be confirmed that within
reinforced connections the effective number of fasteners is equal to the installed
number [6, 2]. Furthermore, the initial stiffness shows a good accordance regardless
of the fastener arrangement.

2.3 Simplified Mechanical Model

No information on the non-linear load-displacement behaviour is given in the cur-
rent standards [2, 13]. In order to achieve practical applicability, it is necessary to
introduce a simplified method. The introduced approach is based on known meth-
ods associated with the load-bearing resistance and the stiffness known from the
standard [2]. The simplified method describes the load slip behaviour as a tri-linear
approach which follows the procedure of EN 1993-1-8 [16].
Thus the first part of the triangular graph is characterized by the initial stiffness
(Kser ) and the load-carrying capacity (Fv,Rk ) of the fastener based on Equation (9).
The first describing point is found by:


F / Fv, Rk [15]




Mean 24
Mean 33
0,2 Mean 52

0 5 10 15 20 25 30 35
Displacement [mm]

Fig. 2 Mean values of the load-displacement behaviour in the experiments conducted (

184 F. Bruhl, J. Schanzlin, and U. Kuhlmann

F1 2
2 u1 = = 3 1,5 (2)
F1 = F v,Rk (1) K1 k
3 23
The stiffness of the second part is given by one third of the initial stiffness. Hence
the second point is calculated as:
F Fv,Rk
u2 = u1 + = u1 +
3 K1 K1
F2 = F v,Rk (3)
Fv,Rk Fv,Rk 5
= 3
+ = 3
K1 K1 K1
The third part is characterized by an infinitesimal stiffness.
Figure 3 shows the comparison of the gained load-displacement behaviour based
on the simplified model and the mean value obtained in all experiments on speci-
mens with a diameter of 12 mm. Two different proposals are shown: on the one hand
the simplified behaviour based on the initial material properties, and on the other the
material properties according to the standard. Within the standardized determination
a steel grade of S355 (fu,k =510 N/mm2 ) was chosen and a mean value of the density
of g,mean =420 kg/m3 [17] was used to calculate the bearing resistance.
The load-bearing resistance found experimentally shows a good conformity with
the tri-linear approach based on the initial properties. Since the input values of the
standardized determination are lower than the actual properties, the load-bearing re-
sistance underestimates the experimental value. On the other hand, the initial stiff-
ness given by the characteristic density g,k based on [17] confirms the experimental
The displacement at failure (uf ) is determined as the 2% percentile based on all
experiments conducted on specimens with a diameter of 12 mm. Assuming a log-


K3 =0
F / Fv,Rk [15]

F1 K2 = 31 K1

Mean 12mm
0,2 Approach based on [15]
K1 Approach based on experimental data
0 5 10 15 20 25 30
u1 u2 uu
Displacement [mm]

Fig. 3 Comparison of the load-displacement curve of the tri-linear approach with the mean
value of the tension component obtained experimentally ( 12mm)
Ductility in Timber Structures: Investigations on Over-Strength Factors 185



M / Mmax, 52

Mean 24
Mean 33
0,2 Mean 52

0 20 40 60 80 100 120 140 160
Rotation [mrad]
Fig. 4 Mean values of the moment-rotation behaviour observed experimentally ( 12mm)
(normalized to the maximum moment of dowel arrangement, 52)

normal distribution, a 2% percentile value of 25.4 mm is determined with a mean of

32.1 mm and a coefficient of variation of 11%.

2.4 Joint Tests

Besides the pure connection tests, experiments on joints were conducted to prove
the rotational capacity of moment-resistant joints. The experiments were performed
as four-point bending tests with a joint acting in the middle of the test setup. The
bending moment is split into a compression force and a tension force (see Figure
4). The previously described connections are acting in the tension zone, whereas
the compression force is transferred via a defined compression zone (steel block 65
mm180 mm).
Figure 4 shows the mean values determined on the basis of three experiments
on specimens with a dowel diameter of 12 mm. The given rotation is associated
with either side of the joint. A dowel arrangement of 52 dowels consists of the
highest bending moment capacity due to the larger inner lever arm and the largest
number of fasteners. Furthermore it can be seen that the decrease in the plastic
level is more pronounced for a stretched arrangement (52) compared to a rather
wide arrangements (24). The decrease in the moment capacity is associated with
a geometrical effect, which is described in [18].

3 Introduction of an Over-Strength Factor

3.1 General
The previous chapter has shown that the variation of the material properties has a
significant influence on the load-bearing resistance of the dowel-type fastener (see
186 F. Bruhl, J. Schanzlin, and U. Kuhlmann

brittle element ductile element brittle element

Rcs Rjoint Rcs
Rjoint kcs Rcs

Fig. 5 Series of different structural elements [19]

Figure 3). Regardless of the application of ductility in timber structures, it is impor-

tant to ensure that the plastification of the ductile element takes place before a brittle
element fails. The capacity design method was developed by Paulay and Priestley
[19] with regard to the earthquake safety of reinforced concrete structures. The gen-
eral idea is illustrated in Figure 5. An over-strength factor kcs is introduced (comp.
Equation (5)), which ensures that no brittle failure occurs with a certain probability.
M joint
kcs kcs < 1.0 (5)
The current version of the Swiss timber code [13,] already implies an over-
strength factor. Hence the brittle element must consist of a 20% higher bearing re-
sistance as the ductile element.

3.2 Determination of the Over-Strength Factor

The over-strength factor is determined based on a Monte-Carlo simulation. The in-
vestigations are conducted following the geometrical and material properties of the
experiments loaded in bending (comp. Section 2). The reliability index is deter-
mined with a C++ program based on 108 calculations following the equation given
in EN 1990 [20] assuming a normal distribution of the limit state function:
= (6)

where g represents the mean value of the limit state function and g the corre-
sponding standard deviation. The limit state function is given by [20]:

g = RE (7)

R represents the resistance and E the effect on the system.

Within this consideration the load-carrying capacity of the dowel-type fasteners
is set as an effect on the brittle element. The limit state function therefore becomes:

g = Wnet fm XM cs n Fv,Rk e = Mcs XM cs M joint (8)

Ductility in Timber Structures: Investigations on Over-Strength Factors 187

Table 1 Input variables based on [22] and test results [24]

tensile timber bending model

diameter width height
strength density strength uncertainty
fu [MPa] [mm] [kg/m3 ] fm [MPa] [mm] [mm] [-]
log-normal normal log-normal log-normal
distribution normal normal normal
[21] [22] [22] [21]




2x4 mean value 441.7 33.8

3x3 mean value 449.6 33.9


5x2 mean value 440.3 33.4

COV 0.04 [21] 0.001 0.1 [22] 0.15 [22] 0.0025 0.0015 0.1 [21]

Wnet : net section modulus fm : bending stress

XM : model uncertainty cs : variable to determine kcs
n : number of fasteners e : inner lever arm (see also [18])

Rk,1 = fh,i,k ti d [2]

Fv,Rk = min Rk,2 = fh,i,k ti d 4My,k
2 + f dt 2 1 [2] (9)

h,i,k i
R = 2 2 M f
k,3 y,k h,i,k d [15]
Table 1 shows the main input variables of the Monte-Carlo simulations. The material
properties of the lamellae used for the test beams were recorded during the manu-
facturing process. Therefore it was possible to gain knowledge about the density and
the tension strength of the fabricated lamellae within the different specimens. The
bending strength of the beams was determined based on the tension strength of the
boards, following the equation given in [23].
Certain material properties of timber are showing a dependence on each other
[22]. According to [22] a correlation coefficient of 0.6 was introduced to take into
account the dependency of the bending strength on the density within the beam

3.3 Results of the Monte-Carlo Simulation

Figure 6(a) shows the reliability index obtained by the Monte-Carlo simulation for
different input values of cs . The load-carrying capacity of the fasteners has a direct
influence on the reliability index. The decrease of the reliability index for a dowel
arrangement of 52 with a larger inner lever arm and a larger number of fasteners
is higher compared with a dowel arrangement of 24 dowels with a smaller lever
A hidden over-strength factor is inevitably integrated within the design of joints.
This is the result of the differences in the design of the bearing resistance of the
188 F. Bruhl, J. Schanzlin, and U. Kuhlmann

24 8
7 33 (kcs ) = 7.65 7.65 kcs
52 7
6 33
5 52

2 2
1 1
0 0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1,6 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1,6
cs kcs
(a) Factor cs for different reliability indexes (b) Normalized factor kcs for different relia-
of various dowel arrangements bility indexes

Fig. 6 Over-strength factors cs /kcs depending on the the reliability index based on prop-
erties given in Table 1

cross-section (Mcs,design ) and the design of the load-carrying capacity of the joint
(Mjoint,design ).
An over-strength factor between 0.5 and 0.75 was considered in the design of the
experimental joint. Therefore the normalized over-strength factor kcs can be found
M joint,design
kcs = cs (10)
The design values Mjoint,design and Mcs,design are determined based on the mean in-
put values of the Monte-Carlo simulation (comp. Table 1). Figure 6(b) shows the
normalized over-strength factors kcs of the different reliability indexes for the
considered joints [14]. The obtained reliability line can be expressed by:

(kcs ) = 7.65 7.65 kcs (11)

The verification of the conducted reliability line shows a reliability index of zero
for a kcs value of one. A reliability index of zero gives a failure probability (Pf ) of
0.5. This means that the system is in an indifferent condition. This is a reasonable
number since both the effect and the resistance are of the same magnitude.

4 Conclusions and Outlook

Reinforced dowel-type connections show a significant ductile behaviour not only
when loaded in tension but also when loaded in bending. A simplified model is
given to describe the non-linear behaviour of connections based on known parame-
ters. Regardless for which purpose ductility is needed, it is important to ensure that
the ductile behaviour is activated before a brittle member fails. A reliability line is
determined based on a Monte-Carlo simulation, which gives certain over-strength
factors. Thereby an over-strength factor can be chosen.
Ductility in Timber Structures: Investigations on Over-Strength Factors 189

Further investigations of the reliability line need to be conducted to consolidate

the results.

Acknowledgements. This research work (16184N) coordinated by International Associa-

tion for Technical Issues related to Wood (iVTH e.V.), is supported in the program of Indus-
triellen Gemeinschaftsforschung (IGF) and financed by the German Federation of Industrial
Research Association (AiF) [24]. Furthermore we would like to thank the Deutschen Institut
fur Bautechnik (DIBt) for its support [25].

[1] EN 1998-1:2004 + AC:2009: Eurocode 8: Design of struictures for earthquake
resistance- Part1: General rules, seismic actions and rules for buildings. European Com-
mittee for Standardization (CEN), Brussels (2004)
[2] 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), Brussels
[3] EN 1991-1-7: Eurocode 1: Actions on structures - Part 1-7: General actions - Acciden-
tal actions. European Committee for Standardization (CEN), Brussels (2006)
[4] Bruhl, F., Kuhlmann, U., Jorissen, A.: Consideration of plasticity within the design of
timber structures due to connection ductility. Structural Engineer 33, S.3007S.3017
[5] Jorissen, A.: Double shear timber connections with dowel type fasteners. Delft Univer-
sity of Technology, The Netherlands, Dissertation (1998)
[6] Bejtka, I.: Verstarkungen von Bauteilen aus Holz mit Vollgewindeschrauben. Univer-
sitat Karlsruhe, Lehrstuhl fur Ingenieurholzbau und Baukonstruktionen, Dissertation
[7] Schickhofer, G., Augustin, M., Jeitler, G.: Einfuhrung in die Verbindungstechnik
mit Stabdubeln, Schrauben und eingeklebten Stahlstangen. In: 6. Grazer Holzbau-
Fachtagung (2007)
[8] Smith, I., Asiz, A., Snow, M., Chui, I.H.: Possible Canadian/ISO approach to deriv-
ing design values from test data. In: Proceedings of the Meeting No. 39 of Working
Commission W18 - Timber Structures, CIB, Florence, Italy (August 2006)
[9] EN 10025-2: Hot rolled products of structural steels. Technical delivery conditions for
- non-alloy structural steels. European Committee for Standardization (CEN), Bruxelles
[10] EN 26891: Timber structures; Joints made with mechanical fasteners; General princi-
ples for the determination of strength and deformation characteristics. European Com-
mittee for Standardization (CEN), Brussels (1991)
[11] EN 12512: Timber structures - Test methods - Cyclic testing of joints made with me-
chanical fasteners. European Committee for Standardization (CEN), Bruxelles
[12] Munoz, W., Mohammad, M., Salenikovich, A., Quenneville, P.: Need for a harmo-
nized approach for calculations of ductility of timber assemblies. In: Proceedings of the
Meeting No. 41 of Working Commission W18 - Timber Structures, CIB, St. Andrews,
Canada (August 2008)
[13] SIA 265:2012: Holzbau. Schweizerischer Ingenieur- und Architektenverein (2012)
[14] Bruhl, F.: Ductility in timber structures - possibilities and requirements. Institut fur
Konstruktion und Entwurf, Universitat Stuttgart, Dissertation (in preparation)
190 F. Bruhl, J. Schanzlin, and U. Kuhlmann

[15] DIN EN 1995-1-1/NA: National Annex - Design of timber structures - Part 1-1: Gen-
eral rules and rules for buildings. DIN-Deutsches Institut fur Normung e.V. (2010)
[16] EN 1993-1-8: Eurocode 3: Design of steel structures- Part 1-8: Design of joints. Euro-
pean Committee for Standardization (CEN), Brussels (2005)
[17] E DIN EN 14080: Holzbauwerke - Brettschichtholz und Balkenschichtholz - An-
forderungen; deutsche Fassung prEN 14080:2011. DIN-Deutsches Institut fur Normung
e.V. (DIN) (2011)
[18] Bruhl, F., Kuhlmann, U.: Requirements on Ductility in Timber Structures. In: Proceed-
ings of the meeting No. 45 of Working Commission W18 - Timber Structures, CIB,
Vaxjo, Schweden (paper No 45-7-5) (August 2012)
[19] Paulay, T., Priestley, M.J.N.: Seismic design of reinforced concrete and masonary build-
ings. John Wiley & Sons, Inc. (1992)
[20] EN 1990: Eurocode: Basis of structural design. European Committee for Standardiza-
tion (CEN), Brussels (2002)
[21] Kohler, J.: Reliability of timber structures, Swiss Federal Institute of Technology,
Zurich, Dissertation (April 2006)
[22] J OINT C OMMITTEE ON S TRUCTURAL S AFETY (JCSS): Probabilistic Model Code.
Part 3: Resistance Models (3.5 Properties of timber) (2006)
[23] EN 14080: Timber structures - Glued laminated timber and glued solid timber - Re-
quirements. European Committee for Standardization (CEN), Brussels (2011)
[24] Kuhlmann, U., Bruhl, F.: Vorteilhafte Bemessung von Holztragwerken durch duk-
tile, plastische Anschlusse / Institut fur Konstruktion und Entwurf. 2012.
Forschungsvorhaben im Auftrag der iVTH, gefordert duch die AiF, Forschungsbericht
AiF 16184 N (2012)
[25] Kuhlmann, U., Bruhl, F.: Robuste Holztragwerke durch duktile Anschlusse mit
stiftformigen Verbindungsmitteln/Institut fur Konstruktion und Entwurf. 2011.
Forschungsvorhaben im Auftrag des Deutschen Instituts fur Bautechnik DIBt (2011)
An Experimental Study on Bearing Strength
in Compression for Bolted Joint of Plywood

Akiko Ohtsuka1, Sumiya Takahashi1, Takumi Ito1, and Wataru Kambe2

Dept. of Archi., Fac. of Eng., Tokyo Univ. of Sci.,
6-3-1, Niijyuku, Katsushika-ku, Tokyo, Japan
Dept. of Arch. and Env. Design, Kanto Gakuin Univ., Dr. Eng.
1-50-1, Mutsuura-higashi, Kanazawa-ku, Yokohama-city, Kanagawa, Japan

Abstract. In recent years, in the field of wood industry, especially in Japan, it is

required that planed raw materials switch from import materials to domestic
lumber and development of new usage suitable for domestic lumber is strongly
demanded. To promote and increase the demand of domestic cedars, the various
types of products or building members by use of plywood has been suggested in
Japan. We suggested the combined structure which sandwiched a steel material in
plywood. And also, to investigate the structural resistant mechanism and
performance under compression, the loading test of the sandwiched members has
been conducted. Furthermore, we assume that the combined structural system
would apply to sheet lightweight section steel structures, and the examination
about the stiffening effect with the system is done, too. In addition, it is thought
that it is necessary to establish enough clearance between the hole of plywood and
bolt because of the accuracy of finishing of materials and precision of
constructions. On the other hand, it is desirable that the clearance becomes small
as much as possible, because the excessive clearance causes a drop of the structure
unity of the combined material and invites the paths for thermal air environmental
and humidity. Then, the moderate clearance around the joint would be inevitable,
it is necessary to clarify the influence on the stress transfer mechanism and
modification quality by the bearing pressure of the circumference of a bolt joint,
and the above. In this research, the two types of filling up method for the
clearance are studied; the first is a wet construction method by use of filler, and
the other is a dry construction method with steel hardware. Herein, to develop and
propose the above filling up methods, the structural resistant performance is not
only evaluated, the difficulties of production, the costs and the human body effects
are estimated. To select the suited filler materials, the construction examination
and material testing were carried out. Form this selection, an urethane, an epoxy
and a cement were picked up. And the bearing pressure experiment tests with the
parameter of materials, the size of clearance, i.e. the thickness of filler, and the
procuring period were conducted. The next, the steel hardware is designed with
consideration the absorbing the gap resulted from construction and production

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 191
DOI: 10.1007/978-94-007-7811-5_18, RILEM 2014
192 A. Ohtsuka et al.

errors. And the bearing pressure experiment study is conducted. From the
experimental result, the relations of rigidity and strength with the parameters of
fillers are clarified. And also, from the comparison of the results of filler and the
ones of steel hardware, it can be said the steel hardware becomes the one of the
most suitable methods.

Keywords: Plywood, Joint, Resistant mechanism.

1 Introduction
Nowadays there is a movement that is transferred to domestic materials from
import one. It is require that the way to use of domestic timber. Especially the
ways of using domestic cedar are positive suggested, in order to plan promotion of
utilization of domestic Japan cedar materials. [1] We proposed the composite
member which sandwiched the steel component with plywood, as shown in Fig. 1,
and the experimental examination about the structural resistant mechanism and
performance under compression were performed. [2]
This panel system is that a steel component is put with plywood, a bolt is
inserted in the hole prepared beforehand, and it joins by hand-clapping for
attaining simplification and the increase in efficiency of construction,
management, etc.. In this time, if it takes into consideration that processing
accuracy of the bolt hole of each component and the accuracy of execution when it
is erected, it is thought that it is necessary to prepare sufficient clearance to the
design intention value of a bolt position, as shown in Fig. 2. On the other hand, he
clearance becomes small as much as possible, because the excessive clearance
causes a drop of the structure unity of the combined material and invites the paths
for thermal air environmental and humidity. In this research, it aims at clarifying
the following two points about the clearance of the circumference of the joints.-1)
Influence which it has on the structure performance and the dynamic mechanism
of a sandwich panel component.2) Influence which it has on the stress transfer and
modification quality by bearing strength in compression for bolted joint of
plywood. Although we are examining 1) separately, in this time it is an
examination experimental about the influence of the clearance given here to.

Design objective(Dashed line


hole of plywood
steel frame

Fig. 1 Proposed composite member [1] Fig. 2 Clearance around the hole
An Experimental Study on Bearing Strength in Compression 193

In this paper, as the filling method for clearance, two kinds of methods, the wet
construction method by one filler and a dry type construction method with a steel
hardware, are proposed. And the bearing pressure experiment is conducted to
examine about structural resistant mechanism and performance under

2 Outline of Clearance and Filling Method

We propose two methods for filling up with clearance, the wet construction
method by use of filler, and the dry type construction method which uses ones of
steel hardware.

2.1 The Wet Construction Method by Use of Filler

It is desirable that a clearance is filled up with a suitable method, since the
excessive clearance of the circumference of a joint may lead to the problem of
reducing the one of steel materials and plywood in sandwich panel structure. For
example, the crevice resulting from the crack may affect safety, usability,
durability, etc. by corrosion etc. Therefore, various kinds of fillers are developed
and put in practical use. Then, these fillers are investigated, and when filled up
with the circumference of the bolt joints of a sandwich panel, the optimal filler is
selected from various viewpoints. The basis of selection of fillers and its process
are as being shown below.

1). Set Up Basis of Selection from a Viewpoint of Safety, Construction, and

Environment, and Select Eight Sorts from 100 Sorts of Materials.
2). Carry Out Construction Examination about Material Selected by 1, and Select
Five Sorts from
3). Carry Out Material Testing about Material Selected by 2), Material
characteristics are evaluated.

2.1.1 The Basis of Selection of Material

The basis of selection is uniquely set up and selected from a viewpoint of safety,
construction, and environment to 100 sorts of materials. The details of a basis of
selection are as follows.
Safety The standard of non-formaldehyde is cleared.
Construction Processes, such as becoming hot at the time of construction, are
not required.
Environment It can construct without using a special tool.
We select also as that of the basis of selection, and decide to carry out a
construction examination to four sorts of epoxy, urethane, emulsion, silicone, and
eight sorts of materials of cement.
194 A. Ohtsuka et al.

2.1.2 The Method and Result of a Construction Examination

Using t = 24 mm of thickness of plywood, a bolt is set to M16-F10T. 17, 19, 23
and 26mm hole are opened in the center of plywood, and a 16-mm bolt is inserted.
It is not filled up about 1 mm of clearance between a point hole and a bolt, but the
clearance 3, 5 and 10 mm are filled up with the filler using the fillers selected
2.1.1. The specimens are cured in the temperature-controlled room (temperature of
20 , 60% of humidity). It is considered as curing period 1, 3, 7, and 28 days, and
prepared one body each. A result is shown in Table. 1.

Table 1 Construction test result

working The days to protect fillers

filler Construction Appearance
life 1 3 7
Epoxy A
Epoxy B
Epoxy C
Epoxy D
: good : Defect : Mean value
2.1.3 Method and Result of a Construction Examination
From the experimental result of 2.1.2, we experiment monotonous compression
loading about two sorts of epoxy, urethane, emulsion, and cement In the
temperature-controlled room (temperature of 20 , 60% of humidity), it is cured
in each filler. The curing period was made into 1, 3, 7, 14, and 28 days, and is
prepared every one body each. About the specimen, it should be based on JIS-
K7181 except cement, and it should be based on JIS-A1132 about cement.
The Young's modulus in a material-testing result is shown in Table 2. The
following four points can be said as a feature.
1). Young's Modulus of Epoxy A, Urethane, and Emulsion is near 0.
2). Although Urethane changes greatly, after unloading it is highly rich in
3). Epoxy B shows stable Young's Modulus after the days 7th.
4). As for cement-fulled, the increase in Young's modulus is checked in
connection with days in care-of-health between 3 and 14 days.
Except epoxy, cement, the Young's modulus which have distinguished Young's
modulus were hardly able to be checked based on the above experimental result.
But the characteristic of each material, it was filled up with the filler using the
urethane which is rich in elasticity and returns to the form near a basis after
modification. As a result, its decided to do the next examination.
An Experimental Study on Bearing Strength in Compression 195

Table 2 Young's modulus in the material 16 15 15 16
.5 6 .5 17 .5 6 .5
-testing result of a filler [N/mm2]

The days to protect fillers

Filler Steel pipe
1 3 7 28 6 t= 6mm
= 60 mm
Steel plate (SS400)
1. Epoxy A t = 6 mm

2. Urethane 7.75 7.59

3. Emulsion Hole of the bolt
= 17 mm
4 Epoxy B 109.1 103.2

5. Cement 1622 902 4464 4483



Fig. 3 The steel hardware

Table 3 The material-testing result of hardware

Young's Yield Tensile Surrender

Cross-sectional Fracture
modulus strength strength distortion
shape growth [%]
[N/mm2] [N/mm2] [N/mm2] []
Monotonous 2.07105 288 443 38 2112
Circular steel pipe 2.17105 334 512 21 2479

2.2 Wet Construction Method by Use of Steel Hardware

By the dry type construction method using steel hardware, it is expected that the
ease of carrying out, the adjustment of the pressure area under compression are
improved by comparison with the fillers. Then, the method of filling up clearance
with this research with hardware as shown in Fig. 4 is devised. This hardware is
preparing several kinds of positions of the point hole of an inside steel plate, and
has intention of absorbing an error. The steel plate of the outside and an inner side
is set to SS400(from Japanese Industrial Standards :JIS grade), and the steel pipe
is set to STKM13A. The mechanical properties are shown in Table 3.

3 Specification of a Specimen

The bearing pressure experiment with an assumption on the mechanical condition

around the bolt joint by reference of BS EN3833 to is conducted.
196 A. Ohtsuka et al.

3.1 The Wet Construction Method by Use of Filler

Using t= 24 mm of thick of plywood, a bolt is set to M16-F10T. 17, 19, 23 and
26mm hole are opened in the center of plywood shown in Fig5, and 16-mm bolt is
inserted. It is not filled up about 1 mm of clearance between a point hole and a
bolt, but the clearance 3,5 and 10 mm are filled up with the filler. The specimens
are cured in the temperature-controlled room (temperature of 20 , 60% of
humidity). It is considered as curing period 1, 3, 7, and 28 days, and prepared one
body each.

3.2 Dry Construction Method by Use of Steel Hardware

As previous mentioned in 3.1, thick material plywood (t= 24 mm) is used, and a
bolt is set to M16-F10T(from JIS grade. The Scale and layout of test specimens
are shown in Fig4 and 5. The specimen has as show Fig5, the 63mm-hole opened
for steel hardware.

200 24
100 200 24

224 100

112 112
42 Steel Hardware
08 3
1 08


Fig. 4 Specimen figue(filling) unit:mm Fig. 5 Specimen figure(Steel Hardware) unit:mm

Load Elastic line



Rigidity 2

Rigidity 1

Fig. 6 The outline of a bearing pressure experimental result

An Experimental Study on Bearing Strength in Compression 197

4 Results of Bearing Pressure Examination

4.1 Load--Relations
The outline of the load displacement relation to the experimental result of this
paper is shown in Fig. 6.

4.2 The Wet Construction Method by Use of Filler

In an experimental result, load displacement relations are shown in Fig.7and it is
presented by a type of filler and curing days. While it is not concerned with curing
days in the case of a urethane-filled and rigidity hardly changed, in the case of the
epoxy -filled or the cement -filled, when care-of-health days progressed, it is
observed that rigidity becomes large. From the Fig.7, the difference of rigidity is
confined whether the test specimen has a filled or not. About the urethane-filled,
in case of urethane, the rigidity is almost same concerning with curing days. On
the other hands, in case of epoxy-filled and cement-filled, the rigidity becomes
large as the case of curing days.
From the results of urethane-filled, the rigidity shows very small, however, the
it is almost same as curing days is progressed of 3-day-curing.
From the result of epoxy-filled after 7-day-curing, it is revealed that the rigidity
becomes large compared with plywood. Between 7 to 28 curing-day, the rigidity is
shown as almost same.
Cement filled it is observed that the rigidity of cement-filled after 3, 7 and 28
curing-days shows almost same with its of plywood only.

16 16 16
14 14 14
12 12 12



10 10 10
8 8 8
6 6 6
4 4 4
2 2 2
0 0 0
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
Diplacement[mm] Diplacement[mm] Diplacement[mm]

(a)Urethane -filled (b)epoxy -filled (c)cement -filled

explanatory notesa legend :1day :3days :7days :28days no filling
Fig. 7 The load-displacement relation

4.3 In Case of Hardware

The result of load-displacement relation in case of steel-hardware-used is shown
in Fig 8. And the results issue of plywood and cement-filled of 28-curing-day.are
198 A. Ohtsuka et al.

compared in Fig.8, too. From the comparison of Fig.8, proof stress and the rigidity
in case of steel hardware-used are presented about 2-times compared with other

Steel Hardware (for one body)

15 No Filling(17mm)
5 Cement System 28days(19mm)
-5 0 5 10 15 20 25

Fig. 8 The load-displacement relation of using steel hardware

5 Analytical Examination about Rigidity

From the experimental result of the preceding chapter, it is observed that the
difference of strength, rigidity and behavior are presented as s parameter with the
kind of filler, curing days, etc. That is, it is supposed that the material-
characteristics value of the filler influence to bearing pressure performance. Then,
as shown in Fig. 10, a model is assumed as a series -filled of a filler and plywood.
And an analytical examination is performed about rigidity herein. Rigid body
assumption of the bolt shall be carried out, and it shall be given by the following
formula as a series system of a filler and plywood. That is, the bearing pressure
rigidity K shall be given by rigid K2 of rigid K1 and the filler of plywood by the
following formula.
K= (1)
1 K1 + 1 K 2

The rigidity K1 of plywood shall be set to rigid K1 when sinking into plywood
based on "woody structural design standard" 4 in Japanese, shall be devoted fiber
rectangular cross direction k90 with the fibrous direction k0 of the following
formula, and shall be given by rigid average value K1=Ac/(k0+k90)2.

A fibrous direction caves in and it is rigidity: k E1

0 = (2-1)
31.6 + 10.9 d

The fiber rectangular cross direction k

k90 = 0
3.4 (2-2)
caves in and it is rigidity:

Next, the rigidity K2 of a filler is obtained from the following formula.

An Experimental Study on Bearing Strength in Compression 199


h 2

h 2

( t
P = 2 E2 t R0 + cos2 + G2 t R0 + sin2 d = E2 + G2 R0 +
h h h

t h
K2 = (E 2 + G 2 ) R0 + (3)
h 2

Analytical results by use of Formula (1)-(3) are shown in Fig.9. And also test
results are compared in Fig.9. also.


Days Days Days

(a)Urethane system (b)epoxy -filled (c)cement -filled

Introductory notes: experimental resultSolid line( ), analysis resultDotted line (---)
The size of the hole:19,21,26

Fig. 9 Comparison of the experimental result and analysis result

6 Comparison of Experiment and Analysis

1) In cases of epoxy-filled and a cement-filled, the rigidity becomes large with the
high Youngs modules of the material.
Moreover, in the case which uses the filler with high Youngs modules, rigidity
and proof stress become large, with large clearance.
Of course, the large clearance will expand the contact area between the filler
and plywood. So then the rigidity becomes large if filler with high Youngs
modules is used.
Moreover, when hardware is used, it is guessed that it is easy to secure proof
stress and rigidity also with a repeated load.
2) Since rigidity and proof stress were high as compared with the case which
uses a filler and transfer of the strength to plywood was also large when hardware
was used, it turned out that it excels in the bearing pressure performance.
Moreover, when hardware is used, it is guessed that it is easy to secure proof
stress and rigidity also with a repeated load.
3) From an analysis result, it is thought that the bearing pressure mechanism of
a joint can be expressed with the series -filled of a bolt, a filler, and plywood.
Moreover, a simple analysis model can estimate bearing pressure rigidity in
general in the case which uses the filler.
200 A. Ohtsuka et al.

7 Conclusion

In this research, the method to be filled up using a filler and steel-hardware is

proposed against the clearance around between the plywood and steel members.
The bearing pressure experiment is conducted as parameter with a kind of filler,
curing days, and so on. From the experiment results, the strength, rigidity and
behavior are investigated. Di the influence of the material characteristic of filler
and curing-day are compared and.
From the experiment, the structural characteristic and construction in a wet
construction method and a dry type construction method have been grasped.
Moreover, a analytical model for rigidity is proposed. And from the
comparison, it shows good agreement with test results.

1. Ito, T., Kambe, W., Kondo, S., Takahasi, S.: An experimental study on compression
resistant mechanism of sandwiched panel of structural playwood and steel member.
Journal of Structural and Construction Engineering, Journal of Architecture and
Building Science, No. 18-40, 941946 (2011) (in Japanese)
2. BS EN383: Determination of embedment strength and foundation values for dowel type
fasteners (2007)
3. Architectural Inst. of Japan: Quality of wood structure design standard, commentary -
permissible stress degree, permission proof stress design law, p. 233 (December 2006)
Investigations Concerning the Force
Distribution along Axially Loaded Self-tapping

A. Ringhofer and G. Schickhofer

Institute of Timber Engineering and Wood Technology,

Graz University of Technology, Austria

Abstract. Self-tapping screws, as simple fasteners with a high load carrying po-
tential if stressed axially, are frequently applied in timber engineering as tensile
joints in wide span GLT truss systems or as reinforcements against stresses
perpendicular to grain. In fact, force distribution along axially loaded screws has a
very important influence on the joint behaviour. Some models based on Volk-
ersens theory combined with fundamentals of linear elastic fracture mechanics
already exist for glued-in rods or lag screws.
This paper provides a measuring technique estimating the force distribution
based on the determined elongation of the threaded part over the inserted length
by several strain gauges while the composite timber-screw is stressed axially.
Therefore, 16 withdrawal push-pull tests were carried out in solid timber vary-
ing the slenderness , given as the ratio lef/d, from 5 to 20 and with angles of screw
axis to grain direction of 0, 45 and 90. The results are used to verify existing
models of comparable configurations which are further adapted to self-tapping
Beside the fundamental knowledge of the withdrawal behaviour, also structural
analysis of tensile joints with self-tapping screws can be improved considering the
location of the stress centre and its impact on eccentricities due to the non-linear
distributions, and in regard to recommendations concerning load introduction
perpendicular to grain.

Keywords: Axial loading, Self-tapping screws, Push-pull, Withdrawal capacity,

Force distribution, Strain gauge measurement.

1 Introduction

Modern self-tapping screws are usually produced with tensile strengths of more
than 1,000 N/mm which enables assembling up to lengths of 1,500 mm without
pre-drilling. Steel hardening, necessary to reach these high strength values, led
to an optimisation of axial bearing resistance and consequently to inclined

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 201
DOI: 10.1007/978-94-007-7811-5_19, RILEM 2014
202 A. Ringhofer and G. Schickhofer

positioning. Concerning those connections, three failure modes of the single screw
are worth to be mentioned: (i) withdrawal failure, (ii) steel failure in tension and
(iii) pull-through failure of the screw head. While scenarios (ii) and (iii) are easy
to handle or rarely decisive in the design process, the withdrawal behaviour, influ-
enced by several material, environmental and constructive conditions, may be
regarded as the essential research topic concerning the axial bearing resistance of
self-tapping screws. Partially influenced by several scientific activities in the past,
shown e. g. in [2,9,16], European standardisation modified the design process of
self-tapping screws over the last decade which currently results in the determina-
tion of the characteristic withdrawal capacity Fax,,Rk of a screwed connection, in
[N], according to EN 1995-1-1:2009 [7], see

nef f ax , k d lef kd
Fax , , Rk = lef k ,
0.5 0.1
with f ax , k = 0.52 d
1.2 cos ( ) + sin ( )
2 2

where fax,k is the characteristic value of the withdrawal strength, in [N/mm],

which depends on the geometrical parameters d as the thread diameter of the
screw and lef as the effective penetration length, both in [mm] on the one side and
on the material parameter k as the characteristic density of the timber product, in
[kg/m] on the other side. Further parameters in this model are nef as the effective
number of fasteners and kd as reduction factor for screw diameters smaller than 8
mm. With regard to these parameters which base on empirically determined
regression models, it is worth mentioning that none of them provide explicit
information about the mechanical behaviour of the axially stressed composite
timber-screw. This model thus enables a simplified application for practical use,
but is not able to represent the real behaviour of the connection analytically which
requires adequate physical parameters. Motivated by that, several investigations in
this topic were carried out in the last decades:
On the one hand, two tests series [1,4] where force distributions of axially
loaded rods were measured by several strain gauges (SG) positioned along the
inserted lengths in timber specimen with numerous varying geometrical parame-
ters are worth to be mentioned. Both showed qualitatively similar and nonlinear
distributions of elongations and consequently lead to nonlinear distributions of
axial force in the rods. They provide an empirical background for further numeri-
cal and analytical investigations.
On the other hand, there are many research activities to be noted, whose scope
was to describe the withdrawal behaviour analytically, mainly of axially loaded
glued-in rods (in steel, glass fibre or carbon fibre), but also of lag or self-tapping
screws (parallel to grain direction), see e. g. [5,11,12,18,19]. All of these men-
tioned contributions principally base on the theory of Volkersen [20] which has
been adjusted considering different bond behaviour (linear-elastic, nonlinear-
elastic, ideal-plastic, etc.) or in combination with maximum or mean stress failure
criterions basing on linear elastic fracture mechanics (LEFM). In contrast to
Eq (1), those models contain mechanical parameters such as fracture energy Gf,
Investigations Concerning the Force Distribution 203

maximum shear stress f as well as elastic modulus E to describe the withdrawal

Considering the fact that force distribution of self-tapping screws has not been
determined extensively by measuring their elongations if loaded axially, the main
scope of the project presented in this paper was to fill this lack of information.
Therefore, the following sections illustrate those experimental investigations
which contained strain gauge measurements in the frame of withdrawal tests on
diameter 12 mm screws under the variation of the parameters penetration length,
lef and angles of screw axis to grain, . Consequently, test results are used to verify
the applicability of the mentioned analytical models for self-tapping screws under
the consideration of other test series which mainly focused on the determination of
geometrical parameters for screwed connections (see [10,15,17]). Finally, recom-
mendations, not only for practical use but also for further investigations in this
research field, are given.

2 Experimental Investigations

2.1 Test Configuration and Procedure

The experimental program for determining the force distribution of axially loaded
self-tapping screws in the frame of withdrawal tests was carried out in three main
steps. The first test series, containing a variation of lef (60, 120, 180 and 240 mm,
two tests for each step) while = 90 has been kept constant, was executed. Every
single test was done as follows:
1. preparation of solid timber (ST) test specimen of Norway spruce (Picea abies);
2. cutting test specimen in the middle (perp. to grain direction) followed by press-
ing together;
3. pre-drilling and inserting of the screw centrally in the saw kerf;
4. removing of screw and application of strain gauges (SG);
5. non-destructive tension test of the screw for calibration of the strain gauges;
6. embedding the screw in the screwed hole and bonding of timber pieces;
7. withdrawal test of the screw (pulsating, cyclic load with nonlinearly increasing
load steps and constant loading rate)
In the next step, tests were carried out with = 0 and = 45 where, based on
first results, lef was only varied from 180 mm to 240 mm. For determination of the
elastic modulus, Ec,0 and the compression strength, fc,0 (both parallel to grain
direction), useful parameters for verification with analytical models, small sam-
ples (40 x 40 x 80 mm) were cut out of each timber specimen and exposed to
compression tests. Additionally, density 12 (for 12 % moisture content) was de-
termined. It should be noted that the height of these small samples deviates from
the regulations for the determination of Ec,0 and fc,0 given in EN 408:2010 [6].
204 A. Ringhofer and G. Schickhofer

copper wires
F [N]

strain gauges (SG)

adhesive films
t [sec]
2 SG

20 SG saw kerf
60 240 mm

10 SG 24 46 220 250 mm

6 SG

130 160 mm

6 SG

2 SG

Fig. 1 Withdrawal test configuration using strain gauges to determine force distribution

2.2 Test Results and Discussion

In Table 1 mean values of the basic parameters 12, Ec,0, fc,0 as well as the with-
drawal strength fax,corr (density correction according to [3], required for compari-
son of rather inhomogeneous values of 12,mean in dependence of ) reached for the
tests are shown. With regard to the given values, a significant difference of fax,corr
can be observed from = 45 to 0 while results of = 45 and 90 are more or
less equal. This corresponds to conclusions from several investigations done in the
past, see e. g. [16]. Furthermore, similar relations of Ec,0, fc,0 and also fax (for
= 45 and 90) in dependence of 12 can be noted.

Table 1 Basic parameters and results of withdrawal tests

N lef No. SG fax,mean 12,mean fax,corr,mean E0,mean fc,0,mean

[] [-] [mm] [-] [N/mm] [kg/m] [N/mm] [N/mm] [N/mm]
90 8 60240 2446 14.5 401 13.0 9303 35.2
45 4 180240 4046 16.2 448 13.3 12637 42.1
0 4 180240 4046 9.98 403 8.92 9503 36.1

A compilation of the determined force distributions along the screw axis, as the
main objective of these investigations, is shown in Fig. 2 (axes are scaled in [%]).
Therefore, the measured electric values from the strain gauges during the with-
drawal tests had been calibrated for two different load levels considering (i) 30 %
of Fmax as loading rate situated in the linear elastic area of the force-deformation
relationships and (ii) 90 % of Fmax which already is located in the nonlinear part
close to the total failure load. Inconsistent values measured by the strain gauges at
100 % Fmax level made assessment at this point impossible. The data points deter-
mined for each test at these load levels have a certain degree of variation which
Investigations Concerning the Force Distribution 205

depends on different parameters like eccentric load transfer or measuring inaccu-

racies of the strain gauges. Therefore, polynomial trend lines (TL, 2nd order for 1st
load level and 3rd order for 2nd load level data points) were used to describe the
measured force distributions shown in Fig. 2. With regard to these distributions,
some principle facts have to be discussed more in detail:
Firstly, measurement errors occurred during three tests ( = 90, lef = 60 mm
(both) and 240 mm (one)) which had therefore to be rejected for further investiga-
tions. Secondly, all test results given in Fig. 2 indicate clearly a slight nonlinear
force distribution along the penetration length lef depending on the load levels
described before. All determined trend lines at 30 % Fmax have a similar and con-
vex shape with small deviations while those at 90 % Fmax show higher variation in
their course, especially for = 90. Thirdly, it is remarkable that apart from this
higher dispersion for = 90 the influence of different angles from screw axis to
grain direction and also penetration lengths lef on the qualitative course of the
lines appears to be negligible.

100% 100% 100%

= 90 = 45 = 0

75% 75% 75%

% Fmax

% Fmax

% Fmax
50% 50% 50%

25% 25% 25%

0% 0%
0% 25% 50% 75% 100% 0% 25% 50% 75% 100%
0% 25% 50% 75% 100%
% lef % lef % lef
120_30_01 120_90_01
180_30_01 180_90_01 180_30_01 180_90_01
120_30_02 120_90_02
180_30_02 180_90_02 180_30_02 180_90_02
180_30_01 180_90_01
240_30_01 240_90_01 240_30_01 240_90_01
180_30_02 180_90_02
240_30_02 240_90_02 240_30_02 240_90_02 240_30_02 240_90_02

Fig. 2 Distributions of Fax in dependence of : polynomial regression graphs of measured


3 Verification with Models Adapted for Self-tapping Screws

Within this section the former discussed force distributions along the screw axis
are compared with the solution of Volkersens differential equation considering
linear-elastic bond behaviour declared in Eq. 2 (according to [18], some parame-
ters are renamed).

sinh( x ) G 1 1
N ax ( x ) = Fax , with =
dc + , (2)
sinh( lef ) t ( EA ) st ( EA )w
206 A. Ringhofer and G. Schickhofer

where Nax(x) is the force distribution along the screw axis, in [N], G the shear
modulus of the shear layer, in [N/mm], t the thickness of the shear layer and dc
the core diameter of the screw, both in [mm]. Consequently, only distributions
determined for forces situated in the linear-elastic part (at 30 % of Fmax) of the
force-deformation relationship are considered in this section. All parameters nec-
essary to describe the behaviour analytically are given in Table 2. In addition,
Fig. 3 shows the assumed specifications of the geometrical parameters which dif-
ferentiate in dependence of .

Table 2 Parameters for the analytical determination of force distribution

lef E00 E90 | E45 a GLR | GLT aw,1 bw,1 aw,2 bw,2 ash bsh t
[] [mm] [N/mm] [N/mm] [N/mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm]
90 120 10118 440 389 42.0 30.0 123 67.5 36 18 22.0
120 9367 407 360 42.0 30.0 123 67.5 36 18 22.0
180 8847 385 340 42.0 30.0 123 67.5 36 18 22.0
180 9245 420 356 42.0 30.0 123 67.5 36 18 22.0
240 8820 383 339 42.0 30.0 123 67.5 36 18 22.0
45 180 12767 1505 491 42.0 30.0 97.5 77.0 36 18 22.0
180 13190 1554 507 42.0 30.0 97.5 77.0 36 18 22.0
240 12144 1431 467 42.0 30.0 97.5 77.0 36 18 22.0
240 12447 1467 479 42.0 30.0 97.5 77.0 36 18 22.0
0 180 9366 - 446 30.0 30.0 118 80.0 18 18 14.5
180 10495 - 500 30.0 30.0 118 80.0 18 18 14.5
240 8791 - 419 30.0 30.0 110 67.5 18 18 14.5
240 9359 - 446 30.0 30.0 110 67.5 18 18 14.5
Screw (steel) parameters: Est = 230,000 N/mm; d = 12 mm; dc = 7 mm
( )
determined according to [13,14]: E45 = E00 1/ 23 sin3 (45) + 1/ 23 cos3 (45)

On the one hand, material parameters such as the elastic and shear modulus
dont have a large impact on the results if varied in a realistic bandwidth and thus
are determined empirically using E90 = E00/23, GLR = E00/26 and GLT = E00/21
(according to [13]).
On the other hand, geometrical parameters, especially the size of the effective
area of the timber, Aw influences the results in a significant way. Due to this fact,
two different approaches lead to the values of aw and bw given in Table 2. The first
approach (index 1, Aw determined as ellipse) assumes aw and bw to be minimum
distances necessary to reach withdrawal failure of an axially loaded screwed con-
nection. In this case, aw and bw are equal to a1/2 and a2/2 (a1 and a2 as minimum
distances according to [7]) which were determined in several investigations on
self-tapping screws and glued-in rods, e. g. [10,15,17], see
Investigations Concerning the Force Distribution 207

= 90 | 45 : aw = a1 / 2 = 3.5d ; bw = a2 / 2 = 2.5 3.5d

= 0 : aw = bw = 2.5d

The second approach (index 2, Aw determined as rectangle) assumes aw and bw

to be maximum values reachable and therefore equal to the timber specimen di-
mensions used for the withdrawal tests, see Table 2.
In addition to that, ash and bsh, necessary for the determination of the shear layer
thickness t (for = 90 and 45 an equivalent circular cross-section has been used)
result from observations made when withdrawal failure occurred. Both values are
assumed (according to [8]) to be three times the distance from the screw axis to
the fracture zone which was observed to be 0.5d x 0.5d for = 0 and 1.0d x 0.5d
for = 90 and 45, see

= 90 | 45 : ash = 3d ; bsh = 1.5d = 0 : ash = bsh = 1.5d (4)

= 45|90 = 0

Aw Aw


Ash Ash
ash aw,1




t Ast t Ast
bsh bw,1 bsh bw,1
bw,2 bw,2

b b

Fig. 3 Assumed specifications of the geometrical parameters used in the model

Finally, a comparison of the determined force distributions with those predicted

by the model approach according to Eq 2 is shown in Fig. 4 to Fig. 6 in depend-
ence of lef and . The two different possibilities of Aw are referenced by M1 and
M2. With regard to the given results, three main aspects should be discussed
more in detail.
First of all and already mentioned before, the interpretation of the effective area
of the timber, Aw has a high impact on the quality of the prediction. All predicted
force distributions (except those with = 45) with Aw determined by the second
approach (maximum condition) correspond very well to the experimental results
while those where the first approach (minimum condition) was applied show a
distinguished convexity over all, significantly underestimating the measured be-
haviour. In addition to that, the angle of the screw axis to grain direction, , influ-
ences the deviation between experiment and prediction in a major way. Compared
to the other angles, the qualitative shapes of the prediction lines for = 45 devi-
ate more from those of the experimental distributions which may is caused by the
assumed values of the model parameters for this configuration, especially E45, aw
208 A. Ringhofer and G. Schickhofer

and bw. Furthermore, it can be noted that the difference between test results and
prediction slightly increases with increasing lef. It is worth mentioning that a good
correlation is given up to lef = 240 mm (slenderness = lef/d = 20), which is al-
ready located close to the upper limit where withdrawal failure is barely decisive
for connection design.
8.00 12.00 16.00
= 90 = 90 = 90
6.00 9.00 12.00
Fax [kN]

Fax [kN]

Fax [kN]
4.00 6.00 8.00

2.00 3.00 4.00

0.00 0.00 0.00

0 30 60 90 120 0 45 90 135 180 0 60 120 180 240
lef [mm] lef [mm] lef [mm]
M1 01 M1 01 M1 01 M1 02 M1 02
M2 01 M2 02 M2 01 M2 02 M2 02
TL 01 TL 02 TL 01 TL 02 TL 02

Fig. 4 Comparison of force distributions: test results vs. predictions, = 90

12.00 16.00
= 45 = 45
M1 01
9.00 12.00 M2 01
Fax [kN]

Fax [kN]

M1 02
6.00 8.00 M2 02
TL 01
TL 02
3.00 4.00

0.00 0.00
0 45 90 135 180 0 60 120 180 240
lef [mm] lef [mm]

Fig. 5 Comparison of force distributions: test results vs. predictions, = 45

8.00 12.00
= 0 = 0
M1 01
6.00 9.00 M2 01
Fax [kN]

Fax [kN]

M1 02

4.00 6.00 M2 02
TL 01
TL 02
2.00 3.00

0.00 0.00
0 45 90 135 180 0 60 120 180 240
lef [mm] lef [mm]

Fig. 6 Comparison of force distributions: test results vs. predictions, = 0

Investigations Concerning the Force Distribution 209

4 Conclusion

A measurement technique which enables the determination of elongation and

force distributions along the inserted threaded part of an axially loaded self-
tapping screw has been developed and presented in this paper. In addition, test
results have been used to verify the suitability of Volkersens theory in depend-
ence of essential parameters such as penetration length, lef and angle of screw axis
to grain direction, . With regard to that a good predictive capability of this ana-
lytical model can be reached for = 0 and 90 under certain conditions which
especially concern the geometrical parameters Aw and t. Consequently it can be
stated in principle that this model approach is also applicable for the design of
axially loaded self-tapping screws.
Investigations concentrating on the prediction of the withdrawal behaviour in
the nonlinear part of the force-deformation relationship as well as the considera-
tion of timber products inherent orthogonal characteristics by this model approach
are seen as the next steps to be carried out in this field.
The fact, that the stress centre is located in between the first third of the
threaded penetrated part of an axially loaded self-tapping screw should not only be
considered for the declaration of minimum distances of screwed connections but
also for practical cases of load introduction perp. to grain influencing the position-
ing of the threaded part in the timber product.

Acknowledgement. The authors want to thank all persons involved in the research project
which led to this contribution, especially Gernot Pirnbacher, Thomas Krpfl, Uwe Flp
and Bernd Heissenberger.

[1] Bernasconi, A.: Tragverhalten von eingeleimten Gewindestangen. Presentationat 2.
GraHSE 2008 Verbindungstechnik im Ingenieurholzbau. Graz University of Tech-
nology, Graz (January 30, 2008)
[2] Bla, H.J., et al.: Tragfhigkeit von Verbindungen mit selbstbohrenden
Holzschrauben mit Vollgewinde. In: Band 4 der Reihe Karlsruher Berichte zum
Ingenieurholzbau. Universittsverlag Karlsruhe (2006)
[3] CUAP 06.03/08 Common Understanding of Assessment Procedure for European
Technical Approval according to Article 9.2 of the Construction Products Directive
Self tapping screws for use in timber constructions (version December 2010)
[4] Ehlbeck, J., Siebert, W.: Praktikable Einleimmethoden und Wirkungsweise von
Gewindestangen unter Axialbelastung bei bertragung von groen Krften und bei
Aufnahme von Querzugkrften in Biegetrgern. Teil 1: Einleimmethoden,
Meverfahren, Haftspannungsverlauf. Versuchsanstalt fr Stahl, Holz und Steine,
Universitt Karlsruhe, Karlsruhe (1987)
[5] Ellingsb, P., Malo, K.A.: Withdrawal capacity of long self-tapping screws parallel to
grain direction. In: World Conference on Timber Engineering (WCTE), Auckland,
New Zealand (July 2012)
210 A. Ringhofer and G. Schickhofer

[6] EN 408: 2010-07, Timber structures Structural timber and glued laminated timber
Determination of some physical and mechanical properties
[7] EN 1995-1-1: 2009-07, Eurocode 5: Design of timber structures Part 1-1: General
Common rules and rules for buildings
[8] Foschi, R., Longworth, J.: Analysis and design of griplam nailed connections. J.
Struct. Div. ASCE 101(12), 25372555 (1975)
[9] Frese, M., Bla, H.J.: Models for the calculation of the withdrawal capacity of self-
tapping screws. In: 42nd CIB-W18 Meeting, Duebendorf Switzerland, paper 42-7-3
[10] Gehri, E.: Eingeklebte Anker Anforderungen und Umsetzungen. Paper Presented at
Internationales Holzbau-Forum (IHF 2009), Garmisch-Partenkirchen (December
[11] Gustafsson, P.J., et al.: A strength design equation for glued-in rods. In: International
Symposium on Joints in Timber Structures, Stuttgart. RILEM proceedings, vol. 22
[12] Jensen, J.L., et al.: A simple unified model for withdrawal of lag screws and gluedin
rods. Eur. J. Wood Prod. (2010), doi:10.1007/s00107-010-0478-y
[13] Keylwerth, R.: Die anisotrope Elastizitt des Holzes und der Lagenhlzer. VDI-
Forschungsheft 430. Ausgabe B, B and 17 (1951)
[14] Kollmann, F.: Technologie des Holzes und der Holzwerkstoffe, vol. 2. Band 1.
Springer, Gttingen (1951)
[15] Mahlknecht, U.: Widerstand und Bruchverhalten axial beanspruchter
Schraubengruppen in Vollholz und Brettschichtholz. Report. Graz University of
Technology (2013)
[16] Pirnbacher, G., Brandner, R., Schickhofer, G.: Base parameters of self-tapping
screws. In: 42nd CIB-W18 Meeting, Duebendorf Switzerland, paper 42-7-1 (2009)
[17] Plieschounig, S.: Ausziehtragverhalten axial beanspruchter Schraubengruppen. Mas-
ter Thesis, Graz University of Technology (2010)
[18] Prtner, C.: Untersuchungen zum Verbund zwischen eingeklebten stiftfrmigen
faserverstrkten Kunststoffen und Holz. Dissertation, University of Kassel (2005)
[19] Serrano, E., Gustaffson, P.J.: Fracture mechanics in timber engineering Strength
analyses of components and joints. Materials and Structures 40, 8796 (2006)
[20] Volkersen, O.: Die Schubkraftverteilung in Leim-, Niet- und Bolzenverbindugnen.
Energie und Technik, Teil 1-3 (1953)
Experimental Analysis on the Structural
Behaviour of Connections with LVL Made
of Beech Wood

Peter Kobel1, Ren Steiger2, and Andrea Frangi1

ETH Zurich, Institute of Structural Engineering IBK, Zurich, Switzerland
EMPA, Swiss Federal Laboratories for Materials Science and Technology,
Structural Engineering Research Laboratory, Dubendorf, Switzerland

Abstract. Despite its higher strength and stiffness properties as compared to most
softwood species, beech wood is today almost entirely used for energetic purposes
or non-structural applications. Benefitting from elevated mechanical properties
and a reliable high degree of homogeneity, Laminated Veneer Lumber (LVL)
made of beech has a great potential for applications in high performance structural
elements, for instance in large span truss structures. As the performance of timber
truss structures predominantly depends on the efficiency of the connections, an
experimental analysis of dowel-type connections in beech LVL was performed. A
series of embedment tests was carried out according to EN 383:2007, the parame-
ters being the dowel diameter and the end distance of the connectors. The tests
showed a very ductile behaviour of the material, high values of embedment
strength and a low scatter in the results (CoV < 5%). These findings were
confirmed in subsequent tensile tests on full dowel connections. The tested con-
nections consisted of four dowels and two slotted-in steel plates, the examined
parameters were the dowel diameter and the spacing. Provided that an adequate
spacing is guaranteed, the full connections also showed a very ductile behaviour.
The common problem of premature splitting failure did not occur due to the fa-
vourable effect of the cross-layers in the LVL. It was further found, that the
adequate spacing has to be determined with regard to shear plug failure. The
experimental analysis has confirmed the potential for efficient dowel-type connec-
tions with LVL made of beech.

Keywords: laminated veneer lumber, beech LVL, dowel type connections,

embedment strength, ductility.

1 Introduction

Timber is highly complex due to its anisotropic nature and natural inhomogeneity.
The mechanical and physical properties are influenced by duration of load,
moisture and temperature. As a consequence, the reliability of structural timber

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 211
DOI: 10.1007/978-94-007-7811-5_20, RILEM 2014
212 P. Kobel, R. Steiger, and A. Frangi

elements is often inadequate and the full potential of timber in the building and
construction sector has not been exploited yet.
The higher strength and stiffness properties of beech wood as compared to most
softwood species are well known. In Switzerland and other European countries
beech is available in large quantities. However, beech wood is today almost en-
tirely used for energetic purposes or non-structural applications (e.g. in the wood
furniture industry). In Switzerland, for example, almost 60% of the harvested
hardwood is used directly for energetic purposes without adding value to it by
considering other applications. A current research project at ETH Zurich and
Empa aims at developing sustainable innovative and reliable timber structures
using Laminated Veneer Lumber (LVL) made of beech wood. Due to its industri-
alised production, reliable and high strength and stiffness properties, improved
dimensional and form stability, structural elements made of beech LVL have a
great potential to be strong and reliable as steel and sustainable as wood.
With its large ratio between weight and strength timber in general is very suit-
able for large span structures. For large spans, usually truss structures are applied,
as they allow an efficient utilisation of the material due to predominantly axial
forces in the members. However, due to the brittle behaviour of timber, the con-
nections in timber truss structures generally become expansive and complex.
Therefore, the connections are often governing the design of the whole structure
and thus can lead to overdesigned timber members. This constitutes technical as
well as economical limits for conventional truss structures.
In order to improve the performance of timber truss structures, the presented
project focuses on developing more efficient connections by applying LVL made
of beech. In a first step, dowel-type connections were investigated, as this is a very
common type of connection for truss structures.
To provide a basis for the design of dowel-type connections using LVL made
of beech a series of embedment tests was carried out. Based on the results from
these tests the load-carrying capacity of full connections can be calculated accord-
ing to the Johansen theory, taking into account additional design criteria given in
current design codes (e.g. Eurocode 5). Finally, the validity of the current design
criteria for dowel-type connections in LVL made of beech was verified in a pre-
liminary tensile test series with full connections.

2 LVL Made of Beech

As a result of its layered structure, LVL is a very homogeneous material with a

very low scatter in its properties. Compared to softwood, the use of beech veneers
significantly improves the strength values. Furthermore, specific properties can be
enhanced and adjusted by applying cross-layers. Most significantly, the tensile
strength perpendicular to the grain can be increased by a multiple (Table 1). This
is particularly favourable in dowel-type connections, where splitting along the
grain is a major problem.
Experimental Analysis on the Structural Behaviour of Connections 213

The LVL material used in this project was manufactured from 2.5 mm thick ro-
tary-peeled beech veneers. In the manufacturing process the veneers were
compacted to a thickness of about 2.3 mm. The cross-sectional layout (Figure 1)
corresponds in principle to that of Kerto Q (established LVL product made of
softwood, usually Norway spruce), with a proportion of cross-layers of around
23%. The average density of all 38 specimens was 762 kg/m3, with a coefficient of
variation of CoV = 1.1%.

Table 1 Influence of cross-layers on the tensile

strength perpendicular to the grain [1]

Beech LVL
Tensile strength
No cross- Cross-layers
perp. to grain
layers (20-25%)
Fig. 1 Cross-sectional layout of the
ft,90,k [N/mm2] 1.6 24.5
used beech LVL (II-III--III-II)

3 Dowel-Type Connections

Connections play an important role with regard to stiffness and load-carrying ca-
pacity of timber structures and thus have been subject of several research studies.
However, so far the research mainly focused on the development of optimised
connection systems and design rules for softwood, rather than hardwood.
Connections with dowel-type fasteners are very common in timber structures.
Their structural behaviour is very complex as it depends on several parameters
like the geometry (fasteners spacing, edge and end distances) and the mechanical
properties of wood and steel. Usually, dowel-type connections are designed
according to Johansens yield theory [2]. This theory is based on a simplified
approach, approximating the behaviour of both timber and steel as ideally rigid-
plastic. By applying equilibrium conditions, the load-carrying capacities of differ-
ent ductile failure modes can be estimated, depending on the geometry and the
material properties. For steel-to-timber connections with slotted-in steel plates the
following three basic failure modes are relevant [3]:

(1) Fv , Rk = f h , k t d
4 M y , Rk
(2) Fv , Rk = f h , k t d 2 + 1
fh ,k t d

(3) Fv , Rk = (1.15 ) 2 M y , Rk f h d

Fig. 2 Basic failure modes acc. to Johansen for dowel-type connections with slotted-in steel
plates. (1) Embedment failure, (2) combined failure, (3) bending failure of the dowel [3].
214 P. Kobel, R. Steiger, and A. Frangi

The smallest value from equations (1) - (3) determines the failure mode and the
capacity per dowel and shear plane. For connections with several shear planes the
capacity per dowel can be derived by determining the failure mode for each shear
plane and then adding up the corresponding capacities.
Johansens yield theory provides the basis for the design of dowel-type connec-
tions. However, in current design codes additional design criteria apply, in order
to cover further failure modes, group effects as well as beneficial effects due to
friction (friction between members, rope effect) which are not considered in
Johansens theory.
Eurocode 5 (EC 5) [4] prescribes a minimum spacing according to Table 2 to
prevent premature brittle failures due to splitting along the grain and plug shear
failure of the timber. To take into account the negative effect of several fasteners
in a row, a reduction factor nef (effective number of fasteners) has to be taken into


Fv ,ef , Rk = nef Fv , Rk with nef = min a1 (4)
n 4

Depending on the failure mode EC 5 also allows for beneficial effects related to
friction [3]. In failure mode (3) the deformation of the dowel implicates a lateral
contact force between the timber and the steel plate, which in turn leads to friction
between these elements. For this case EC 5 suggests a friction factor of 15%, as
included in equation (3).
Generally, EC 5 also allows for a rope effect. However, as for dowels the with-
drawal resistance is solely dependent on friction between dowel and timber, this
effect is neglected in the design of connections with dowels.
As the mentioned design rules have been defined mainly for softwood solid tim-
ber and glulam, the relevant parameters still have to be determined for beech LVL.

Table 2 Spacing according to EC 5 for dowel-type


Spacing and Minimum spacing or

edge/end distance edge/end distance
a1 0 360 (3 + 2|cos|)d
a2 0 360 3d
a3,t -90 90 max(7d; 80mm)
a3,t 90 150 max(a3,t |sin|)d; 3d)
150 210 3d
210 270 max(a3,t |sin|)d; 3d)
max((2 + 2sin)d;
a4,t 0 180
Fig. 3 Spacing according to EC 5
a4,c 180 360 3d
for dowel-type connections
Experimental Analysis on the Structural Behaviour of Connections 215

4 Embedment Tests

4.1 Specimens and Test Setup

To determine the embedment strength for LVL made of beech a series of tensile
embedment tests was carried out. A total of 30 specimens (Table 3) were tested
according to the test standard EN 383:2007 [5], each specimen being equipped
with two dowels (Figure 4). The tests were carried out with four different dowel
diameters. The geometrical properties of the specimens B, C, E and F correspond
to the requirements specified in EN 383. In all specimens of groups D and G, the
prescribed end distance l3 was reduced from 7d to half the value (3.5d) to investi-
gate the effect of the end distance on the embedment strength.
High quality steel dowels with a tensile strength of 960 - 1,100 N/mm2 were
used, in order to prevent a failure of the dowel. The holes for the dowels were pre-
drilled with an equal nominal diameter as the respective dowel.
The tests were run on a displacement controlled servo-hydraulic universal test-
ing machine. The loading procedure included one reloading cycle between 10%
and 40% of the estimated maximum load. During testing the applied load was
recorded, as well as the relative displacement between the dowel and the timber.

Table 3 Specimens for embedment tests

Dowel End No. of

diameter distance specimens
L B t
d [mm] l3 [-] n [-]
[mm] [mm] [mm]
B 20 7d 5 880 120 49
C 16 7d 4 704 96 49
D 16 3.5d 5 592 96 49
E 12 7d 5 528 72 21 Fig. 4 Tensile embed-
F 8 7d 5 352 48 21 ment test acc. to EN 383:
G 8 3.5d 5 296 48 21 a1 = 3d; l3 = 7d; l4 = 40d

4.2 Results
According to EN 383 the embedment strength is derived from either the maximum
load or the load at a displacement of 5 mm, whichever occurs first:

{F=5 mm ; Fmax }
fh = (5)
d t
For specimens with a full end distance according to EN 383 the load at a dis-
placement of 5 mm always became decisive. Figure 5 shows the load-
displacement behaviour of all conducted embedment tests.
216 P. Kobel, R. Steiger, and A. Frangi

120 100

Embedment strength fh [N/mm2]

B Test results
100 80
Load F [kN]

C 60
D 40
40 Spruce
E 20
0 0
0 5 10 20 30 40 50 0 6 8 12 16 20 30
Relative displacement [mm] Dowel diameter d [mm]

Fig. 5 Load-displacement behaviour of all Fig. 6 Comparison of test results. Acc. to

conducted embedment tests (for notation of EC 5 design rules: fh,m=0.082(1-0.01)m with
groups see Table 3). Limit displacement m(spruce)=470kg/m3 , m(beech)=760kg/m3.
acc. to EN383: =5mm. : l3=7d; x: l3=3.5d.

All configurations with the full recommended end distance of 7d (i.e. B, C, E,

F) showed a very pronounced plateau after an initial linear elastic phase, and the
maximum embedment stresses fh,max occurred at very large displacements and
were significantly higher than the values fh,=5mm derived according to EN 383.
The tests were continued until failure occurred on one side of the specimen.
Hence, fh,max was only reached for one out of two dowels. In Table 4 both these
values are listed for each dowel diameter together with the coefficients of varia-
tion (CoV). For the values derived according to EN 383 the average CoV was
below 5%.

Table 4 Embedment test results

fh,=5mm (EN 383) fh,max

Configuration n fh,=5mm CoV n fh,max CoV (Fmax)
[-] [N/mm2] [%] [-] [N/mm2] [%] [mm]
B (d=20mm) 10 75 5.1 4 104 2.2 39
C (d=16mm) 8 77 3.6 4 100 5.0 21
E (d=12mm) 10 84 3.4 4 101 4.7 29
F (d=8mm) 10 87 6.7 5 101 4.8 16

Table 5 shows the influence of a reduced end distance. In Figure 5 it can be

seen that the initial slopes of the curves are very similar for specimens with half
and full end distances. However, for specimens of groups D and G with only half
the suggested end distance, failure in the form of a shear plug occurred at a much
lower displacement.
Experimental Analysis on the Structural Behaviour of Connections 217

Table 5 Influence of the end distance l3 on the embedment strength

d = 16 mm d = 8 mm
fh,=5mm (EN 383) fh,=5mm (EN 383)
Configu- Configu-
l3 n fh,=5mm CoV l3 n fh,=5mm CoV
ration ration
[-] [-] [N/mm2] [%] [-] [-] [N/mm2] [%]
C 7d 8 77 3.6 F 7d 10 87 6.7
D 3.5d 7 72 3.9 G 3.5d 5 81 2.6
Relative loss: -6.5% Relative loss: -7.3%

In Figure 6 the results are compared with the expected values according to the
design rules in EC 5 for solid timber and glulam made of Norway spruce and
beech. The obtained values for beech LVL are significantly higher, not only
compared to solid spruce timber, but also compared to solid beech timber,
which confirms the beneficial effect of the cross-layers. The influence of a re-
duced end distance resulted in a loss in embedment strength of around 7% (for
d = 8 / 16 mm).

5 Full Connection Tensile Tests

5.1 Specimens and Test Setup

To investigate the behaviour of full dowel-type connections with LVL made of
beech a series of tensile tests was carried out. A total of eight specimens were
tested, each specimen being equipped with two connections, and each connection
consisting of four dowels and two slotted-in steel plates (Figure 7). All steel parts
were of the quality S355. The examined parameters were the dowel diameter and
the spacing (Table 6). The dowel diameter was varied in order to study the behav-
iour for rigid dowels (d = 20 mm) and slender dowels (d = 8 mm). Spacing
according to EC 5 was applied, as well as a reduced spacing with only half the
distances, to investigate the influence of the spacing.

Table 6 Configurations of the tested connec-

tions. Specimen dimensions: L=1,275mm,
B=240mm, t=100mm.

Configu- Spacing
d [mm] a3 a1 a4
20F 20 7d 5d 3d
20H 20 3.5d 2.5d 3d
8F 8 7d 5d 3d
8H 8 3.5d 2.5d 3d Fig. 7 Schematic of the tested connections
218 P. Kobel, R. Steiger, and A. Frangi

The tensile tests were carried out according to EN 1380:2009 [6] and
EN 26891:1991 [7] on a displacement controlled servo-hydraulic universal testing
machine. The loading procedure included one reloading cycle between 10% and
40% of the estimated maximum load. The tests were continued until failure oc-
curred on one side of the specimen. During testing the applied load was recorded,
as well as the relative displacement in the connection between the steel plates and
the timber.

5.2 Results
Figure 8 shows the load-displacement behaviour of all conducted tests. The ob-
tained maximum loads and displacements including the observed failure modes
are listed in Table 7.

Load F [kN]


100 8H

0 5 10 15 20 25
Relative displacement [mm]

Fig. 8 Load-displacement behaviour of the Fig. 9 Failure modes for d=20mm. Left: full
tested connections (notation as in Table 6) spacing; right: half spacing

The tests have shown a very ductile behaviour of connections with a spacing of
the dowels according to EC 5. In configuration 8F the ductility resulted mainly
from the bending deformations of the dowels. In configuration 20F, however, the
dowels remained rigid during the whole test, which means in this case the ductility
was provided by the ductile embedment failure in the LVL.

Table 7 Test results

Configuration n [-] Fmax [kN] [mm] Failure mode

20F 2 523 13.4 Embedment failure
20H 2 346 3.4 Shear plug failure
8F 2 203 15* Dowel failure
8H 2 137 5.4 Dowel and shear plug failure
*Limit displacement acc. to EN 26 891.
Experimental Analysis on the Structural Behaviour of Connections 219

Configurations with only half the spacing failed prematurely due to shear plug
failure. Consequently, lower bearing capacities and a significantly reduced ductil-
ity were observed. Configuration 20H showed a quite brittle behaviour, whereas in
configuration 8H the bending failure of the dowels maintained a certain degree of
ductility. No splitting failures were observed in any of the specimens.

6 Conclusions

The presented experimental investigation was carried out to evaluate the potential
of dowel-type connections with LVL made of beech. The tests have shown that
the favourable material properties of beech LVL are well reflected in the perform-
ance of the connections:
The mean embedment strength values obtained in beech LVL are significantly
higher compared to corresponding values for solid Norway spruce and beech tim-
ber. This confirms the advantage of beech compared to Norway spruce (as the
embedment strength is a function of the timber density) and the beneficial effect
of the cross-layers.
Given an adequate spacing, a very ductile behaviour was observed in the em-
bedment tests, along with a very low scatter (CoV < 5%).
These findings were reflected in tensile tests with full dowel connections. The
results have shown that a notable ductility can be provided by the beech LVL
material, as even connections with rigid dowels (d = 20 mm) showed a ductile
overall behaviour. Furthermore, the problem of premature splitting failure could
be eliminated by the cross-layers. The variation of the dowel spacing yielded that
the adequate spacing has to be determined with regard to shear plug failure.
The observed ductile behaviour leads to the assumption that the negative group
effect should be small or even negligible (nef 1). However, further experimental
and numerical analysis has to be conducted in order to validate this hypothesis.
The experimental investigation has shown the potential for efficient connections
in LVL made of beech. This, along with the high strength and stiffness values, con-
firms the suitability of beech LVL for high performance timber structures, such as
large span trusses.

1. Van de Kuilen, J.-W., Knorz, M.: Prfbericht Nr. 10511: Ergebnisse der
Zulassungsversuche fr eine, allgemeine bauaufsichtliche Zulassung (abZ) von
Furnierschichtholz aus Buche. Technische Universitt Mnchen (TUM), Germany
2. Johansen, K.W.: Theory of Timber Connections, pp. 5669. IABSE, Publication No. 9,
Bern, Switzerland (1949)
3. Porteous, J., Kermani, A.: Structural Timber Design to Eurocode 5. Wiley-Blackwell
220 P. Kobel, R. Steiger, and A. Frangi

4. DIN EN 1995-1-1:2010-12. Eurocode 5: Design of timber structures - Part 1-1: General

- Common rules and rules for buildings. DIN German Institute for Standardization
5. DIN EN 383:2007. Timber Structures - Test methods - Determination of embedment
strength and foundation values for dowel type fasteners. DIN German Institute for
6. DIN EN 1380:2009. Timber structures - Test methods - Load bearing nails, screws,
dowels and bolts. DIN German Institute for Standardization
7. EN 26 891:1991. Timber structures; joints made with mechanical fasten-ers; general
principles for the determination of strength and deformation characteristics (ISO
6891:1983). CEN European Committee for Standardization
The Embedment Failure of European Beech
Compared to Spruce Wood and Standards

Steffen Franke and Nolie Magnire

Bern University of Applied Sciences, Architecture, Wood and Civil Engineering,


Abstract. The embedment behaviour of European hardwoods in dowel-type

connections is investigated through the analysis of tests performed on beech wood.
Experimental results are evaluated using the 5% offset method. They are com-
pared with estimations provided by the Eurocode 5 and the SIA 265 and with
similar tests performed on spruce. Influences of the load-to-grain angle and the
fastener diameter d are investigated. Beech has a stronger hardening behaviour
than spruce, and is very influenced by the fastener diameter. The comparison with
the standards shows a general overestimation (at least 20%) of the experimental
results, especially for loading perpendicular to the grain. The model provided by
Eurocode 5 to consider the load-to-grain angle is better than the SIA 265 but is
optimized for softwoods only. Thus a formula optimized for both species is

Keywords: hardwood, beech solid wood, Fagus sylvatica, embedment strength,

load-to-grain angle, dowel diameter.

1 Introduction

This paper focuses on the behaviour of dowel-type connections which are widely
used in the building industry. It is thus crucial to estimate accurately the stiffness
and ultimate strength of such connections. The European Yield Model (EYM)
proposed by Johansen in 1949 to design dowel connections is today commonly
accepted. In this model, the overall behaviour of the connection has two compo-
nents: the fastener yield capacity and the embedment strength which directly de-
pends on the wood species. Indeed the embedment strength is defined in the norm
ISO/DIS 10984-2:2009 as the average compressive stress at maximum load in a
piece of timber or wood-based product under the action of a stiff linear fastener.
To achieve smaller connections for higher loadings, European hardwoods are
increasingly used in timber construction. But the embedment strength design for-
mula provided by the EN 1995-1-1:2005 (EC 5) has been obtained experimentally
through many tests performed on softwoods and tropical hardwood species mainly

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 221
DOI: 10.1007/978-94-007-7811-5_21, RILEM 2014
222 S. Franke and N. Magnire

(Leijten et al. 2004). Whereas only 75 tests were conducted for European hard-
woods species (Hbner 2008). This results in a large uncertainty regarding the
accuracy of European hardwoods embedment strength values provided by the
standards. Consequently our current standards must be reviewed to ensure
the same reliability both for hardwoods and softwoods designs. Thus this paper
aims at evaluating the embedment behaviour of the most common European
hardwood specie in Switzerland: beech (Fagus Sylvatica).
To achieve this goal, 160 embedment tests were conducted on beech speci-
mens. Both, the EC 5 and the Swiss standard SIA 265:2012 (SIA 265) describe the
embedment strength as a function of:
The wood species (softwood or hardwood)
The wood density [kg/m ]

The fastener diameter d [mm]
The load-to-grain angle []
Four dowel diameters d (6, 12, 20 and 30 mm) and four load-to-grain angles
(0, 25, 55 and 90) were tested. Similar tests with the same set of parameters were
carried out on common softwood specie: spruce (Picea Abies). 10 specimens were
tested for each combination of parameters to enables a statistical analysis of the
data obtained. The results of both tests series were compared to each other and to
the estimation obtained using the SIA 265 and the EC 5 formulas.

2 Material and Methods

2.1 Testing Standards

Embedment tests can be carried out either using the half-hole design (described in
the ASTM D 5764-97a:1997 or in the ISO/DIS 10984-2) or the full-hole design
(described in ASTM, ISO/DIS or EN 383:1993). The half-hole design enables to
obtain a result free from any influence of the fasteners bending and is suitable for
hardwoods such as beech. It is thus the design which was chosen for this study
(see Fig. 1). According to the ISO/DIS 10984-2, the tests were carried out with a


1 - Steel, 2 - Fastener, 3 - Specimen

Fig. 1 Half hole test set up in ASTM and experiment
The Embedment Failure of European Beech 223

constant speed of 1 mm/min to reach the maximum load in 300 s 120 s. The
tests were stopped when either failure or 8 mm displacement was reached. A pre-
loading cycle between 0.4 Fmax and 0.1 Fmax was performed for each test.

2.2 Beech and Spruce Wood Specimens

Beech and spruce specimens were cut from boards of a constant thickness of
40 mm. Their width and height were obtained from a linear interpolation of the
values provided by the ISO/DIS 10984-2 (see Table 1). The specimens were con-
ditioned at 20 C and 65% of relative humidity until mass consistency was
reached with a mean moisture content of 13.3%. The density distribution of each
series was narrow (see Fig. 2) with, in average, 734 kg/m3 (COV 1.8%) for beech
and 447 kg/m3 (COV 4.4%) for spruce.

Table 1 Sizes of beech and spruce specimens

Width/Height/Thickness [mm]
Load-to-grain angle d = 6 mm d = 12 mm d = 20 mm d = 30 mm
= 0 36/70/40 70/72/40 120/100/40 180/150/40
= 25 45/70/40 85/70/40 143/80/40 180/85/40
= 55 50/70/40 100/70/40 170/85/40 242/75/40
= 90 60/70/40 120/70/40 200/70/40 300/70/40

Load F [kN]

800 Beech
750 F5 mm
Density [kg/m3]

F5 %
600 d=6 mm d=12 mm d=20 mm d=30 mm
d=6 mm d=12 mm d=20 mm d=30 mm
450 K
400 1
Spruce 0.1 F5 mm
0 40 80 120 160 u0 0.05d-offset 5 mm
Sample number [-] Displacement u [mm]

Fig. 2 Density of beech (filled) and spruce Fig. 3 Evaluation methods

2.3 Evaluation of the Results

The evaluation of embedment tests is a point of discussion among experts as various
methods have been used in the past (see for instance in Leijten et al. (2004)). This
led to incompatibility in experimental results. In this study, the curves were evalu-
ated following the 5% offset method to obtain the yield load Fyield. The maximum
load was either the ultimate load or the load at 5 mm displacement, according to
EN 383 and ISO/DIS 10984-2 respectively. Furthermore, the stiffness K was
224 S. Franke and N. Magnire

determined. In the analysis, the yield load was used to calculate the embedment
strength of each test following this formula:
fh = (1)
d t
Where t = thickness of the wood specimen [mm]
The load-displacement curves were described and evaluated as well. A
summary of the information extracted from the experimental curves is provided in
Fig. 3.

2.4 Design Standards

The test results were confronted with the values obtained from the EC 5 and the
SIA 265. Both standards predict the embedment strength at an angle based on
the values parallel and perpendicular to the grain, fh,0 and fh,90 respectively. In the
EC 5, the interpolation between fh,0 and fh,90 is performed using a trigonometric
model: the Hankinson formula:

f h,0
f h, = (2)
k90 sin + cos 2


f h,0 = 0.082 (1 0.01d ) k (3)

1.35 + 0.015d for softwoods

k90 = (4)
0.90 + 0.015d for hardwoods

In the SIA 265 however, a linear interpolation between fh,0 and fh,90 is provided:

f h, = ( 9000
+ 0.15) d 0.3 for softwoods (5)

f h, = ( 9000
+ 0.19 ) d 0.3 for hardwoods (6)

For both standards, is the wood density [kg/m3], d is the fastener diameter [mm]
and is the load-to-grain angle [].

3 Results

3.1 Embedment Behaviour

For both series, the load-displacement curves show a strong relation of the em-
bedment behaviour with the angle . Fig. 4 provides the mean curves of the
The Embedment Failure of European Beech 225

embedment tests performed on beech with a dowel of 20 mm in diameter. They

are representative for the other series as well. For load parallel to the grain ( = 0),
the samples present a quasi-ideal elastic/plastic behaviour with a high slip
modulus followed by a constant load carrying value until failure in splitting along
the plane of a growth ring. When increases a reduction of the slip modulus and
the embedment strength just after yield appears. This corresponds to a hardening
of the wood due to the crushing of fibres under the dowel. The hardening behav-
iour is already present for 25 and 55 angle to the grain and clearly appears at 90
angle. Such behaviour agrees with results found in the literature both for soft-
woods (for instance Franke and Quenneville (2011) or Sawata and Yasumura
(2002)) and hardwoods (Hbner 2008).

Beech, d = 20 mm - mean curves

Load [kN]

5 = 25 = 55
= 90 = 0
0 1 2 3 4 5 6 7 8
Displacement u [mm]

Fig. 4 Typical load-displacement curves for various angles

3.2 Embedment Strength Results

Beech experimental results are given in Table 2. The stiffness K significantly de-
creases when increases (60% to 85%). The yield strength values fh,,5% obtained
were compared with the maximal strength fh,,max. Large differences between fh,,5%
and fh,,max would mean the evaluation method chosen to calculate the yield load
has a strong influence on the final result. Table 2 shows that, in average, fh,max is
26% higher than fh,,5% for beech whereas it was only 17% higher for spruce.
Therefore beech embedment tests are even more sensitive to evaluation method
than spruce ones. An agreement of the experts to one evaluation method is thus a
necessity for further research about hardwood embedment strength.

3.3 Comparison with the Standards Estimation

The characteristics of each specimen (, d and ) were used to calculate its em-
bedment strength according to the EC 5 and the SIA 265, fh,,EC5 and fh,,SIA respec-
tively. They were then compared with the experimental strength found (see Fig. 5
and Table 2). Generally, the embedment strength is overestimated. For both spruce
and beech, only series with = 0 provided the experimental values higher than the
expected ones. The most accurate estimation is for the beech series using the EC 5
226 S. Franke and N. Magnire

formulas. For this group the mean overestimation is only of 18% whereas it ranges
between 50% and 60% for the three other groups. When looking at the evolution
with the angle , the EC 5 provides slightly better results than the SIA 265, espe-
cially for beech. It is important to note that fh,90 is overestimated by least 20% for
both standards and all diameters. fh,90 is then used as a basis to calculate
embedment strength at intermediate angles. Therefore there is a need to review the
calculation of fh,90 to obtain more accurate values. Otherwise the inaccuracy of
fh,90 estimation impacts the estimation of any embedment strength calculated for

Table 2 Embedment strength for beech

Yield strength Max. strength
f h, ,max f h, , EC 5 f h , , SIA
Group Stiffness fh,,5% [MPa] fh,,max [MPa]
HE-d K N/mm] Mean COV Mean COV f h, ,5% f h, ,5% f h, ,5%
HE0-6 46271 70.4 5% 76.9 4% 1.09 0.81 1.16
HE25-6 22703 42.3 3% 62.6 1% 1.48 1.30 1.81
HE55-6 15698 40.0 5% 68.2 6% 1.71 1.45 1.94
HE90-6 16724 45.1 8% 71.0 12% 1.57 1.22 1.56
HE0-12 65751 59.0 5% 62.5 3% 1.06 0.92 1.15
HE25-12 39812 37.1 4% 46.9 3% 1.27 1.37 1.68
HE55-12 32713 33.7 6% 47.7 4% 1.42 1.49 1.84
HE90-12 22107 36.6 1% 55.3 5% 1.51 1.28 1.54
HE0-20 76953 48.5 6% 49.1 5% 1.01 0.96 1.13
HE25-20 56420 41.3 3% 46.4 2% 1.12 1.13 1.34
HE55-20 35286 32.7 7% 39.6 5% 1.21 1.34 1.68
HE90-20 32285 34.0 7% 45.0 6% 1.32 1.22 1.55
HE0-30 116358 50.0 11% 51.0 9% 1.02 0.86 1.02
HE25-30 72105 36.5 4% 39.1 4% 1.07 1.11 1.36
HE55-30 38863 27.3 4% 31.9 3% 1.17 1.31 1.80
HE90-30 30572 25.8 4% 29.6 4% 1.15 1.18 1.71
Average 5% 5% 1.26 1.18 1.52

Beech Spruce
fh,,5% - Standard value [MPa ]

fh,,5% - Standard value [MPa ]

75 35

65 30

55 25
EC 5 SIA 265 EC 5 SIA 265
45 20
= 0 = 0
= 25 = 25
35 = 55 15 = 55
= 90 = 90
25 35 45 55 65 75 85
fh,,5% - Experimental value [MPa ]
fh,,5% - Experimental value [MPa ]

Fig. 5 Standard vs experimental embedment strength for the EC 5 (filled) and the SIA 265
The Embedment Failure of European Beech 227

3.4 Influence of the Dowel Diameter

The average value of embedment strength for each group of dowel diameters is
shown in Fig. 6. The whiskers represent the coefficient of variation of the group.
Two different trends appear: beech series present a clear negative correlation be-
tween strength and fastener diameter, spruce series however, do not show an obvi-
ous trend with relatively constant strength.
Recent studies on softwoods such as Franke and Quenneville (2011) or Sawata
and Yasumura (2002) found fh vs d relatively constant. Whereas Hbner, on his
tests carried out on ash (2008), found a negative correlation between fh and d.
These studies agree with the trend of both series. However, for beech, a decrease
of the trend with increasing angles is observed which contradicts the results of
Hbner (2008). The models used to consider the evolution of fh, with d are
common to softwoods and hardwoods in both the SIA 265 and the EC 5. Both
standards result in a decrease of fh, with d which corresponds better to hardwoods
behaviour than softwoods behaviour.

Beech Spruce
75 50
Embedment strenght fh,,5% [MPa ]

Embedment strenght fh,,5% [MPa ]

= 0 = 25 = 0 = 25
65 = 55 = 90 = 55 = 90

25 10

15 0
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Dowel diameter d [mm] Dowel diameter d [mm]

Fig. 6 Evolution of the embedment strength with the dowel diameter

3.5 Influence of the Load-to-Grain Angle

The embedment strength decreases when increases (see Fig. 7). The reduction
between fh,0 and fh,90 ranges from 30% to 65% with a stronger tendency for spruce
series. The experimental curves in Fig. 7 can be split in two parts. In the left part,
(up to 25 for beech and 55 for spruce) most of the reduction occurs. In the
right part, the strength reduction is less significant with even a slight increase for
Interpolations of fh,0 and fh,90 experimental values were performed to obtain
values for fh,25 and fh,55. The interpolation models used were linear, as in the
SIA 265 and the Hankinson formula, as in the EC 5. Both models provide curves
less convex than the experimental ones and thus overestimate fh at intermediate
228 S. Franke and N. Magnire

Beech Exp. Hankinson Spruce Exp. Hankinson

Embedment strenght fh,,5% [MPa ]

Embedment strenght fh,,5% [MPa ]

d = 6 mm d = 6 mm
70 d = 12 mm d = 12 mm
d = 20 mm 30 d = 20 mm
60 d = 30 mm d = 30 mm
40 15

30 10

20 5
0 20 40 60 80 0 20 40 60 80
Load to grain angle [] Load to grain angle []

Fig. 7 Embedment strength fh,5% vs (full lines) and comparison with the adapted
Hankinson model (dotted lines)

EC 5 model is the most accurate. However with only a correlation R2 of 0.72

for beech compared to 0.92 for spruce, it is optimised for softwoods. An optimiza-
tion of the trigonometric coefficients of the formula was performed. Formula (7)
was obtained which describes accurately the behaviour of beech (R2 =0.95) and
spruce (R2 = 0.94). We can thus hypothesize that it is possible to use a common
model for hardwoods and softwoods to describe the reduction of the embedment
strength with . This model could be based on the approach provided by Hankin-
sons formula. However still further research is necessary to optimize the trigo-
nometric coefficients to more hardwoods and softwoods species.

fh, = (7)
k90 sin1.2 + cos1.5

4 Conclusion

Tests of the embedment strength with various load-to-grain angles and dowel
diameters were performed on beech specimens. The results of these tests com-
pared to the ones of identical tests carried out on spruce specimens show that both
species present different embedment behaviours. These differences can be as-
sumed to be representative of the ones between softwoods and European hard-
woods in general. The values obtain from the Swiss SIA 265:2012 and European
EN 1995-1-1:2004 show that a review of these standards is necessary to obtain
values also representative for European hardwoods. But they also show that the
embedment strength values are currently overestimated even for softwoods when
the load-to-grain angle is different from 0. A better estimation of the embedment
strength perpendicular to the grain is needed. A reduction of at least 20% of cur-
rent estimated values would be necessary. This study also shows that a model
adapted from the one of the Eurocode 5 describes accurately the influence of the
load-to-grain angle on the embedment strength for both beech and spruce wood.
The Embedment Failure of European Beech 229

ASTM International, ASTM D 5764-97a Standard Test Method for Evaluating Dowel-
Bearing Strength of Wood and Wood-Base Products (1997)
European Committee for Standardization, EN 383 Test Methods - Determination of Em-
bedment Strength and Foundation Values for Dowel Type Fasteners (1993)
European Committee for Standardization, EN 1995-1-1 Eurocode 5 Design of Timber
Structures, Part 1-1: General Common Rules for Buildings (2004)
Franke, S., Quenneville, P.: Bolted and Dowelled Connections in Radiata Pine and
Laminted Veneer Lumber Using the European Yield Model. Australian Journal of Struc-
tural Engineering 12(1), 13 (2011)
Hbner, U.: Embedding Strength of European Hardwoods. In: CIB-W18, Paper 4175, St.
Andrews, Canada (2008)
International Organization for Standardization, DRAFT ISO/DIS 10984-2 Timber Struc-
tures - Dowel-type Fasteners Part2: Determination of Embedding Strength and Foun-
dation Values (2009)
Johansen, K.: Theory of Timber Connections. International Association for Bridge and
Structural Engineering Publications 9, 249262 (1949)
Leijten, A., Khler, J., Jorissen, A.: Review of Probability Data for Timber Connections
with Dowletype Fasteners. In: CIB-W18, Paper 37713, Edinburgh, UK (2004)
Sawata, K., Yasumura, M.: Determination of Embedding Strength of Wood for Dowel-type
Fasteners. Journal of Wood Science 48(2) (2002)
Swiss Society of Engineers and Architects, SIA 265 Timber Structures, Zrich, Swiss
Modelling of Non-metallic Timber Connections
at Elevated Temperatures

Daniel Brandon1, Martin P. Ansell2, Richard Harris1, Pete Walker1,

and Julie Bregulla1,3
BRE CICM, Dept. Civil Eng. & Architecture, University of Bath, UK
BRE CICM, Dept. Mechanical Eng., University of Bath, UK
BRE, Watford, UK

Abstract. Models estimating the slip modulus and the load capacity, including
temperature dependent effects, of non-metallic timber connections are presented.
Previous studies, including the work of Thomson, have shown that Glass Fibre
Reinforced Polymer (GFRP) rods are suitable connectors for timber and that
Densified Veneer Wood (DVW) functions effectively as a flitch plate material.
Thomsons model predicting the slip modulus of the connection with GFRP rods
and DVW plates is revised in this paper. The revised model is used to predict the
slip modulus and the failure load at room temperature and elevated temperatures.
The latter is achieved by predicting local temperatures in the connection and tak-
ing corresponding reduced material properties into account.

Keywords: non-metallic timber connections, pultruded GFRP dowels, densified

veneer wood, embedment strength, connection slip.

1 Introduction

Previous studies have shown that Glass Fibre Reinforced Polymer (GFRP) rods
are suitable shear connectors for timber (Pederson 2002; Drake, 2003; Thomson,
2010) and that Densified Veneer Wood (DVW) functions effectively as a flitch
plate material (Thomson, 2010). A section of the connection type considered in
the study of Thomson is shown in Fig.1. Thomson et al. (2010) state that the
advantages of this connection type, compared to metallic connections, are higher
corrosion resistance and increased fire safety. The latter is, however, not yet con-
firmed. Therefore, a model able to predict the behaviour of the connection at ele-
vated temperatures, as proposed in this paper, is required. This model is part of an
on-going study of the behaviour of the connection in fire where the model will be
validated experimentally.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 231
DOI: 10.1007/978-94-007-7811-5_22, RILEM 2014
232 D. Brandon et al.



Distance (mm)



F/2 10

axis of symmetry -0.01 0.01 0.03 0.05
Deflection (mm)

Bernoulli beam theory

Structural node Timoshenko beam

Thermal node Higher order beam
(Thomson, 2010)

Fig. 1 Section of a failed connection (left) model (middle) and deflection at 1kN load

Only a small number of analyses of timber connections in fire have been pub-
lished in the past. A part of these consist of 3-dimensional (3D) structural finite
element (FE) analyses and a 3-dimensional heat transfer model (Racher et al.,
2010; Peng et al., 2011; Audebert et al., 2012; Laplanche and Racher, 2004). The
FE analyses are not easy to get accurate because cracking of the timber locally
around the dowel should be taken into account. 3D FE analyses that do not model
the cracking or local failure will predict behaviour which is too stiff and must be
calibrated to experimental results to perform accurately. A good solution for 3D
modelling of the embedment of a dowel in timber at room temperature was devel-
oped by Xu et al. (2012), by implementing a modified Hill yield criterion. How-
ever, this method requires a high level of expertise in finite element modelling
under the condition of at constant temperature.
Konig and Fontana (2001) concluded from experimental research that the num-
ber of dowels has an insignificant influence on the time to failure. This indicates
that instead of modelling the full connection, it is adequate to model just a single
dowel to predict the behaviour of the connection. Moss et al. (2010) successfully
implemented modified Johansen yield equations (Johansen, 1949) to predict fire
resistance. The Johansen equations are, however, based on a yield moment of the
dowel, which is adequate for steel dowels. GFRP dowels in shear connections,
however, fail in shear, making the Johansen equations unsuitable for predicting
the fire resistance of the non-metallic connection (Thomson et al., 2010).
Modelling of Non-metallic Timber Connections at Elevated Temperatures 233

This research presents a model based on the structural FE model of Thomson

(2010). He used beams on an elastic foundation to model the deformation of the
dowel in the connection where the beam represents the dowel and the foundation
represents the timber. To gain more accuracy the element mesh is refined and the
beams on elastic foundation are replaced with shear deformable beams on elastic
springs. A comparable structural model with a beams on springs was recently
introduced in a fire model by Cachim and Franssen (2009). However, their design
is made for timber connections with steel dowels. The beam theory used for the
beam elements is not suitable for predicting the performance of connections with
GFRP rods. To accurately predict the behaviour of these connections shear de-
formable beam theories should be used, since the shear modulus and shear
strength of the dowel have a significant influence on the connection behaviour.
These properties are determined experimentally at elevated temperatures for
GFRP and the results are presented in this paper. Cachim and Franssen used the
stiffness reduction factors provided by Eurocode 5, to take into account the re-
duced embedment stiffness at elevated temperatures. The embedment stiffness is
however highly dependent on local cracking of the timber around the dowel and is
therefore not only dependent on the stiffness of the timber. Timber embedment
tests at elevated temperatures are presented in this paper to qualify this assump-
tion. The FE models are made using Matlab v7.11, but different computational
software can be used to make the models.

2 Modelling at Room Temperature

Thomson (2010) proposed an FE model to determine the slip modulus using
beams on elastic foundation, but it does not correspond well to his test results. In
this paper beams on multiple springs are used to predict deflections and to con-
struct the basis for a temperature-dependent model. The proposed FE model
shown in Fig 1 (middle) contains beam elements, which represent the GFRP
dowel, and springs which represent the stiffness of the timber and the DVW (thick
lines). Only one symmetrical half of the connection is modelled to reduce the cal-
culation time. The other half is replaced by a boundary condition that restrains the
Different beam elements are considered. The classic or Bernoulli beam theory
assumes that plane cross sections of a beam remain plane and normal to the mid-
line (Fig.2a). The theory is based on pure bending of a beam and therefor does not
take shear deformation into account, in contrast with shear deformable beams. The
Timoshenko beam is a first order shear deformable beam and allows rotation of
the cross section around the midline (Fig.2b). This rotation corresponds to a shear
stress, but since the cross section in Timoshenko beams still remains plane, the
shear stress is constant over the depth of the beam. However, the shear stress has
to be zero at the top and the bottom of the beam and is highest at the midline of
the beam. Therefore, in reality the cross section curves (Fig.2c). The Timoshenko
theory can be corrected with a shear correction factor. A value of 0.88 for a
circular section is assumed in this study.
234 D. Brandon et al.

dw/dx dw/dx

a) Bernoulli b) Timoshenko c) Higher order

beam theory beam theory beam theory

Fig. 2 Beam theories

Higher order beam theories take the warped section into account. The beam
element published by Eisenberger (2003) is used to simulate the dowel. The mate-
rial properties of the timber are determined with standard embedment tests accord-
ing to ASTM D5764 (2002) and the material properties of GFRP are determined
with torsion tests that are discussed below. The results of the two dimensional FE
analysis parallel to the grain can be seen in Fig.1. The results correspond to a load
F of 1 kN and are compared with experimental results of Thomson et al. (2010).
The point representing the experimental results is an average gained from 95
tested dowels and the deformation of the DVW plate is excluded from it. A
slightly more accurate model can be obtained by adding spring elements near the
shear surface. However, to maintain stability of the heat transfer analysis
discussed in the next section the element mesh is chosen as shown in Fig.1.
From the results, it is concluded that predictions with the Bernoulli theory are
not adequate for describing the connection behaviour. This indicates that the shear
deformation plays an important role in the connection behaviour. The higher order
beam theory resembles the experimental results only slightly better than the Ti-
moshenko beam theory. Because of its simplicity, the Timoshenko beam theory is
chosen for further simulations.
The shear strength of the dowel is determined using an expression for the
maximum shear stress in a solid circular section given by Gere (2004):
max = ,
where V is the shear force on the dowel and A is the cross sectional area.

3 Modelling at Elevated Temperatures

Timoshenkos beam element adopted in the FE model of Fig.1 is used as a basis

for the temperature dependent model. To include the temperature effects, a heat
Modelling of Non-metallic Timber Connections at Elevated Temperatures 235

transfer model is introduced, which estimates the local temperatures in the connec-
tion. An increase in local temperature will result in reduced material properties.
The stiffness and strength of each of the spring and beam elements of the FE
model are adjusted to the local temperature estimated by the heat transfer model.
The test setup to determine the reduced shear stiffness and shear strength of the
individual beam elements is shown in Fig.3a. The setup to determine the reduced
stiffness of the individual spring elements is shown in Fig. 3b.

Fig. 3 Material tests at elevated temperatures

3.1 Heat Transfer Model

The heat transfer model presented in this paper is 1-dimensional in the direction of
the GFRP fibres. In this model a full timber section is assumed i.e. no GFRP heat
transfer is modelled. The timber has a higher conductivity and a lower heat capac-
ity per unit volume than GFRP, indicating that the assumption is conservative for
deep beams. However, mass transfer as well as heat transfer is involved during the
heating of timber. For example, moisture will evaporate which requires energy
and slows the heating process down. The validity of assuming a 1D heat transfer
model consisting of only timber will be validated in the on-going research by
comparisons with 2D and 3D models and by experimental validation. The heat
transfer model consists of nine nodes in the length of the dowel. These nodes are
shown in Fig.1 as circles.
The explicit finite difference solution of Fouriers heat equation to predict the
temperatures in the nodes presented by Gilbert (2005) is used:

t Tmp1 Tmp Tmp+1 Tmp

Tmp +1 = + + T4

Cm Rm 1,m Rm , m +1
236 D. Brandon et al.

Where Tmp is the temperature at time p at point m. R is the thermal resistance, Cm

is the heat capacity and t is the time step. The temperature of a node in each
following time step is calculated from the temperature difference with the adjacent
nodes. 1D explicit finite difference solutions require a stability criterion, which is
in this case (Incropera, 2013):
Fo 1

Fo is the Fourier number. This means that to maintain stability the time steps
have to be small in the ratio of the node distances. If node distances are too small
there is a requirement for small time steps and long calculation times.
The heat flux due to convection and radiation is calculated as follows
(Buchanan, 2002):

q" = hc (T f Ts ) + eff (T f4 Ts4 ) ,

where hc is the convection coefficient, is the configuration factor, is the Stefan
Boltzmann constant, eff is the effective emissivity, Tf is the fire temperature and
Ts is the surface temperature. The convection coefficient is taken as 25 W/m2K
(Buchanan, 2002). The configuration factor is taken as 1.0 to simulate zero
distance between the heat source and the surface. The effective emissivity eff can
be calculated with the furnace emissivity f and the surface emissivity s (Peng et
al., 2011):
eff =
f + s fs
f and s are both taken as 0.9 (Peng et al., 2011).
Thermal properties of timber required for the heat transfer model are specific
heat, density and conductivity. The mass transfer however should not be ignored.
Different authors in the past have included mass transfer in the heat transfer prop-
erties of timber (Knig, 2006; Frangi, 2001; Fredlund, 1993). The authors treat
complex phenomena like pyrolysis and evaporation and fissures in the charcoal
differently, but all confirmed their set of properties with test results. Therefore, the
practical difference between the works is small.
The properties presented by Frangi (2001) are implemented in the heat transfer
model of the present study. The energy required for evaporation of moisture is
included in the heat capacity curve. Similarly the energy required for pyrolysis is
included in the function. At higher temperatures over 400C the cracking of the
char layer is included by implementing an additional conductivity.

3.2 Material Properties at Elevated Temperatures

Tests to determine the shear properties of GFRP and the embedment properties of
timber at elevated temperatures were performed in an Instron 3111 convection
Modelling of Non-metallic Timber Connections at Elevated Temperatures 237

oven that fits in the testing machine. The test setups are shown in Fig.3. In both
tests thermocouples were used to measure the temperature. One thermocouple was
placed in a drilled hole in the steel fixture, one thermocouple was placed in a
drilled hole inside an unloaded sample to determine the time required to heat the
sample to the right temperature and another thermocouple measured the tempera-
ture of the surface of the specimen. The test was initiated when all thermocouples
showed an error of less than 3C.
In the torsion test the upper clamp was fixed in all degrees of freedom. The
lower clamp was free to move upwards and downwards and was fixed in the other
degrees of freedom. Four samples were tested at temperatures of 20C, 60C,
100C and 140C and two samples were tested at 180C. The GFRP used in these
tests consisted of S-glass fibres with a polyester resin matrix without fire resistant
The embedment test was performed according to ASTM D5764 because of the
ease of loading. Santos et al. (2010) showed, with a large number of tests, that this
method is equivalent to the method described in EN383 (2004). The specimens
had a small thickness of 20 mm to improve the heat transfer. The test was per-
formed at temperatures of 20C, 60C, 100C, 140C, 160C and 200C. Lami-
nated veneer lumber Kerto-S was used for these embedment tests.

4 Results and Discussion

Fig.4 shows test results for the embedment stiffness (left) and the embedment
strength (right) at elevated temperatures. Here the values are normalized to the
average at 20C. At 20C the average embedment stiffnesses are 463 N/mm2 and
190 N/mm2 for the parallel and perpendicular to grain direction respectively. The
average embedment strengths for these directions are 361 N/mm2 and 322 N/mm2
respectively. Fig.4 also shows the stiffness and the strength of timber in compres-
sion at elevated temperatures according to Eurocode 5 (BS EN 1995-1-2:2004,
2006) for advanced fire calculations. It can be seen that the test results do not re-
semble the code. The compressive stiffness and compressive strength of timber are
at room temperature not linearly related to the embedment stiffness and embed-
ment strength. There are no grounds for assuming that these properties would
decrease in similar ratios when temperature increases. A notable stiffness increase
after 100C in parallel and perpendicular to grain direction, can be caused by the
evaporation of moist at that temperature. However, the scatter of results is large
and the number of tests is not sufficient to confirm this stiffness increase.
Timber starts charring at approximately 300C at which the timber has no
significant strength left. For the present model, linear interpolation between the
average (AVG.) test results per temperature is used, as shown in Fig.4. After
200C, a linear decrease is assumed which reaches zero strength or stiffness at
238 D. Brandon et al.

1.2 1.2
Normalized embedment

Normalized embedment
1 1

0.8 0.8


0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 50 100 150 200 0 50 100 150 200
Temperature (C) Temperature (C)

Parallel to grain Perpendicular to grain

AVG. parallel to grain AVG. perpendicular to grain
BS EN 1995-1-2:2004

Fig. 4 Results of embedment tests at elevated temperatures

Shear strength
Normalized value

Shear modulus

Modelled shear
0.2 strength

20 60 100 140 180 Modelled shear
Temperature (C)

Fig. 5 Results of torsion tests at elevated temperatures

The results of the torsion tests are shown in Fig.5. The average shear strength
and shear stiffness are 40.3MPa and 2.59 GPa respectively. It can be seen that
both shear strength and shear modulus decrease rapidly between 20 and 60C. It
can also be noted that the deviation between results becomes very small at higher
temperatures. Similar relative reduction is assumed for the flexural modulus and
strength. This will be studied in future experiments.
Fig.6 shows the predicted deflections of the symmetrical half of the dowel dur-
ing a standard fire. The load is equal to 40% of the maximum load and the side
member thickness is 48 mm. It can be seen that the deformation accelerates during
a fire. According to the model, shear failure in the dowel will occur after 26
Modelling of Non-metallic Timber Connections at Elevated Temperatures 239

minutes. The shear failure occurs at the shear plane where the temperature reaches
77 C at the moment of failure.
The use of pultruded rods with a polyester resin results in a rapid decrease of
connection properties. The fire resistance can be enhanced by using pultruded rods
that contain a resin with a higher glass transition temperature, like vinyl-ester,
epoxy or phenolic resins.

Deflection (mm)

0 min
0 10 20 30 40 50 60
10 min
20 min
Location (mm)

Fig. 6 Dowel deflection of the symmetrical half during a standard fire

5 Conclusions
A model to describe the connection slip and predict the dowel failure at elevated
temperatures of non-metallic timber connections with pultruded glass fibre rein-
forced polymer dowels and a densified veneer wood flitch plate was presented.
This model is part of an on-going study of the fire performance of these non-
metallic connections. The model consists of a 2-dimensional structural model and
a 1-dimensional heat transfer model. Mechanical material properties at elevated
temperatures required for the model were measured experimentally. In further
studies at the University of Bath the model will be validated with fire tests and the
1-dimensional heat transfer model will be compared with a 2 and 3 dimensional
heat transfer model. The most important findings of this study can be summarized
as follows:
The embedment stiffness and the strength reductions resulted from tests, do not
coincide well with the stiffness and strength reductions of timber in compression
at elevated temperatures according to BS EN 1995-1-2:2004 (2006).
The GFRP dowel of this study has a thermosetting polyester matrix without
fire resistant additives. This material degrades rapidly at elevated temperatures,
resulting in a short time to failure. GFRP rods which perform better in fires will be
considered in future research.
240 D. Brandon et al.

ASTM D5764. Standard test method for evaluating dowel-bearing strength of wood and
wood-based products (2002)
Audebert, M., Dhima, D., Taazount, M., Bouchar, A.: Behavior of dowelled and bolted
steel-to-timber connections exposed to fire. Engineering Structures 39, 116125 (2012)
Buchanan, A.H.: Structural design for fire safety. Wiley, Chichester (2002)
BS EN 1995-1-2:2004. UK National Annex to Eurocode 5: Design of timber structures
part 1-2: General Structural fire design, British Standards (2006)
Cachim, P.B., Franssen, J.M.: Numerical modelling of timber connections under fire load-
ing using a component model. Fire Safety Journal 44, 840853 (2009)
Drake, R.D.: The advancement of structural connection techniques for timber buildings.
PhD Thesis, University of Bath (2003)
Eisenberger, M.: An exact high order beam element. Computers & Structures 81, 147152
EN383, Timber structures. Test methods: Determination of embedding strength and foun-
dation values for dowel type fasteners. European standard, Brussels (2004)
Frangi, A.: Brandverhalten von Holz-Beton-Verbunddecken. Eidgenssische Technische
Hochschule, Zrich (2001)
Fredlund, B.: Modelling of heat and mass transfer in wood structures during fire. Fire Safe-
ty Journal 20, 3969 (1993)
Gere, J.M.: Mechanics of materials with infotrac. Thomson/Brooks/Cole (2004)
Gilbert, B.: Thermal mass and the effects of dynamic heat flow. Master Thesis, University
of East London, School of Computing and Technology (2005)
Incropera, F.P.: Foundations of heat transfer. John Wiley, Singapore (2013)
Johansen, K.W.: Theory of timber connections. International Association for Bridge and
Structural Engineering 9, 249262 (1949)
Knig, J.: Effective thermal actions and thermal properties of timber members in natural
fires. Fire and Materials 30, 5163 (2006)
Knig, J., Fontana, M.: The performance of timber connections in fire. In: Proceedings of
the RILEM Symposium Joints in Timber Structures, vol. 22, pp. 639648 (2001)
Laplanche, K., Dhima, D., Racher, P.: Predicting the behaviour of dowelled connections in
fire: Fire tests results and heat transfer modelling. In: 8th World Conference on Timber
Engineering, Lahti, Finland (2004)
Moss, P., Buchanan, A., Fragiacomo, M., Austruy, C.: Experimental testing and analytical
prediction of the behaviour of timber bolted connections subjected to fire. Fire Technol-
ogy 46, 129148 (2010)
Pedersen, M.: Dowel type timber connections. PhD Thesis, Danmarks Teknise Universitet
Peng, L., Hadjisophocleus, G., Mehaffey, J., Mohammad, M.: Predicting the fire resistance
of woodsteelwood timber connections. Fire Technology 47, 11011119 (2011)
Racher, P., Laplanche, K., Dhima, D., Bouchar, A.: Thermo-mechanical analysis of the fire
performance of dowelled timber connection. Engineering Structures 32, 11481157
Modelling of Non-metallic Timber Connections at Elevated Temperatures 241

Santos, C.L., De Jesus, A.M.P., Morais, J.J.L., Lousada, J.L.P.C.: A comparison between
the EN 383 and ASTMD5764 test methods for dowel-bearing strength assessment of
wood: Experimental and numerical investigations. Strain 46, 159174 (2010)
Thomson, A.: The structural performance of non-metallic timber connections. PhD Thesis,
University of Bath (2010)
Thomson, A., Harris, R., Ansell, M., Walker, P.: Experimental performance of non-metallic
mechanically fastened timber connections. The Structural Engineer 88, 2532 (2010)
Xu, B.-H., Bouchar, A., Taazount, M., Racher, P.: Numerical simulation of embedding
strength of glued laminated timber for dowel-type fasteners. Journal of Wood Science,
17 (2012)
Analysis of the Brittle Failure and Design
of Connections Loaded Perpendicular to Grain

Bettina Franke1,2 and Pierre Quenneville2

Bern University of Applied Sciences, Switzerland
The University of Auckland, New Zealand

Abstract. The brittle failure behaviour and the design of double shear connections
with mechanical fasteners loaded perpendicular to grain are analysed using
experimental and numerical test series and the fracture mechanics methods. The
results observed define the important geometry parameters which influence the
load carrying capacity. Based on the failure criteria determined for double shear
connections, a new design approach are proposed and compared to current interna-
tional design standards. The correlation between the new design proposal and the
experimental test results done or published confirms the methods used and the
failure criteria determined.

Keywords: mechanical fasteners, dowel type connection, double shear connec-

tions, design proposal, fracture mechanics, European spruce, Canadian spruce.

1 Introduction

Timber constructions almost include joints with mechanical fasteners. The con-
nections can be loaded under tension, compression and under angle to the grain
orientation of wood. Depending on the size, layout as well as the loading situation,
the connection describes brittle or ductile failure behaviour. The ductile failure
depends on the embedment strength of the wood and the bending capacity of the
dowel. For the prediction of the ductile failure the established European Yield
Model is used in international standards. The brittle failure of connections occurs
due to tension or shear stress in the wood member. Based on the anisotropy of
wood the tension stress perpendicular to grain leads to very brittle failure behav-
iour and it is the most critical case for the failure of connections.
For the prediction of the splitting failure of double shear connections loaded
perpendicular to grain, different design equations are available in various interna-
tional standards. The design standards are based on one hand on a strength crite-
rion which was introduced by Ehlbeck, Grlacher and Werner (1989) and on the

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 243
DOI: 10.1007/978-94-007-7811-5_23, RILEM 2014
244 B. Franke and P. Quenneville

other hand on fracture mechanic, introduced by v. d. Put and Leijten (1990).The

brittle failure occurs due to the splitting of the wood under tension perpendicular
to grain and/or shear stress. Current research results and publications show that
there are disagreements between the experimental results and the international
design standards, e.g. Franke and Quenneville(2010, 2011), Schoenmakers(2010),
Jensen (2005).

l Fig. 1 Double shear dowel type connection

lr loaded perpendicular to grain

2 Experimental and Numerical Test Program

Experimental and numerical test series are used for the analyses of the failure
behaviour. The experimental test series with multiple dowel-type connections and
different loaded edge distances were conducted in laminated veneer lumber of
Radiata Pine (LVL). The test specimens were in large scale format to provide
comparable results to practical construction details. The detailed description of the
test setup is summarized in Franke et al. (2012).
In addition various numerical test series with single or multiple dowel-type
connections with different loaded edge distances, positions, different number of
dowels per row or column as well as different spacing within the connection are
simulated. The numerical model used is defined as a 2-dimensional model based
on the purpose to especially investigate the splitting of the wood and is presented
in Franke and Quenneville (2010). The numerical model simulates the splitting
failure of wood under tension perpendicular to grain and shear as well as the duc-
tile failure of wood due to compression. However, since the focus is to investigate
the brittle failure mechanism, the bending of the dowel is not included in the
numerical model and the influence is neglected. This is acceptable if the dowel
slenderness ratio is small and ductile behaviour due to dowel bending is mini-
mized. The capabilities of the numerical model were proofed on comprehensive
experimental test series done in Canadian spruce glulam and also with the experi-
mental test series done in LVL. The complete experimental and numerical test
programme is summarized in Table 1. In each test series the dowel diameter is
20 mm.
Analysis of the Brittle Failure and Design of Connections Loaded Perpendicular 245

Table 1 Experimental and numerical test program

Connection Layout
Test Dowel Loaded edge Connection Connection
Material setup Sizes b / h / l [mm] mxn distance he/h width ar height ac
LVL exp. 63/400/1600 3x1 0.2/ 0.4/ 0.6 8d -
LVL exp. 63/600/2400 3x1 0.13/ 0.4/ 0.6 8d -
LVL exp. 63/400/1600 3x2 0.4/ 0.6 8d 4d
LVL exp. 63/400/2400 3x2 0.4/ 0.6 8d 4d
LVL exp. 63/400/2400 3x2 0.4/ 0.6 8d 4d
LVL exp. 63/400/2400 3x2 0.4 8d 4d
LVL exp. 63/600/2400 5x2 0.4 16d 4d
LVL num. 63/(200, 400, 600)/4h 2x2 0.2/ 0.4/ 0.6/ 0.8 3d/ 6d/ -
GL num. 80/(190, 300, 400, 600, 1x1 0.2/ 0.3/ 0.4/ 0.5/ -
800)/610 0.6/ 0.7/ 0.8
GL num. 80/190/4h+ar 2x1 0.2/ 0.3/ 0.4/ 0.5/ 3d/ 4d/ 6d/ 8d/ 10d/
0.6/ 0.7/ 0.8 15d/ 20d/ 25d
GL num. 80/(300, 400, 600, 2x1 0.3/ 0.5/ 0.7 3d/ 6d/ 8d/ 10d/
800)/(4h+ar) 20d
GL num. 80/190/610 1-6x1 0.4/ 0.6 (n - 1) 3d -
GL num. 80/304/1320 1-4x1-3 0.44/ 0.7 16.8d (m-1) 3d
GL num. 80/304/1320 2-3x2 0.44/ 0.7 16.8d 3d, 4.5d,

3 Failure Behaviour of Connections Loaded Perp. to Grain

3.1 Failure Behaviour Observed during the Experimental Tests

The load capacity of double shear dowel-type connections transversely loaded
depends mainly on the loaded edge distance but also on the connection layout, as
shown in Fig. 2 and Fig. 3. Increasing the loaded edge distance increases not only
the splitting load capacity but also allows a load reserve between crack initiation
and separation up to the complete failure. The crack initiation describes the load
when first cracks occur and the crack separation load the load level at complete
separationof the inner part of the connection. The analysis of the distribution of
the load capacity over the number of dowels shows that each row of the connec-
tion carries the same load portion and that the dowels at the edge of the connection
with the largest loaded edge distance carries the greatest load portion and also
fully trigger the failure, Franke and Quenneville (2010), Franke et al. (2012).
246 B. Franke and P. Quenneville

Fig. 2 Crack initiation, crack separation and splitting failure load for b/h = 63/400 mm and
63/600 mm with m x n = 3 x 1

Fig. 3 Load distribution for the dowels of the top row for a connection with two rows

3.2 Failure Behaviour Analysed by Fracture Mechanics

For the characterization of the brittle failure behaviour of double shear connec-
tions investigated, the energy balance method together with crack resistance
curves as one of fracture mechanic methods were used. The crack process is clas-
sified in three different modes. The fracture mode I describes the symmetric sepa-
ration under normal tension stress, mode II contains the in plane shear stress and
mode III the out of plane shear stress. Double shear dowel-type connections
loaded perpendicular to grain fail, considering the stress situation, in interaction of
fracture modes I and II.
The crack resistance curves evaluated for double shear connections show a
nonlinear material behaviour. The crack grows stable as long as the crack resis-
tance increases more than the crack extension force under a constant load during
crack propagation. If the crack resistance exceeds the critical value, the crack will
grow in an unstable manner and the system fails. The crack resistance curves
Analysis of the Brittle Failure and Design of Connections Loaded Perpendicular 247

respectively the critical fracture energies determined were split into the fracture
mode I and mode II using the method of Ishikawa et al. (1979). The critical frac-
ture energies determined are not comparable with the fracture energies known for
different materials for the pure fracture modes I or II. Because they are caused by
a stress situation related to one connection layout and do not relate to test setups
for the investigation of a single fracture mode. Therefore, in this paper, the frac-
ture energies determined for dowel-type connections are called specific critical
I , II
fracture energies, spec ,c .

Fig. 4 shows as example the numerical solution for a 3 by 2 dowel type connec-
tion and Fig. 5 the corresponding crack resistance curve for the dowel at the edge
with the largest loaded edge distance. In the case shown, the failure load is 60 kN.
Analysing the critical fracture energies for the complete connection shows that the
dowels at the edge with the largest loaded edge distance trigger the unstable fail-
ure, Franke and Quenneville (2011). The cracks between the outer dowels (inner
part of the connection) become also unstable and the wood between these dowels
is then completely separated. For the splitting failure design proposal the outline
of the connection layout, described by the loaded edge distance he and the connec-
tion width ar, will be used as observed during the experimental test series.

Fig. 4 Numerical solution - transverse Fig. 5 Crack resistance curve for mode I with
stress for 3 x 2 dowel type connection energy lines for the outside side of the dowel
marked in Fig. 4

3.3 Failure Criteria

For each dowel of the connection investigated, the crack resistance curves as well
as the critical specific fracture energies were determined, as described before. The
distribution of the specific critical fracture energies of the dowels at the edge with
the largest loaded edge distance shows a dependency on the loaded edge distance
he, the connection width ar and the depth of the member h, as shown in Fig. 6 and
Fig. 7. In general, the normalized fracture energies for solid wood respectively
glulam shows a different behaviour than LVL. Due to the multi-layered cross
section of the engineered wood product LVL, it is more homogeneous and the
brittle failure behaviour occurs different to solid wood.
248 B. Franke and P. Quenneville

The distribution of the specific critical fracture energies in relation to the mem-
ber depth shows an influence on the fracture mode I but not on mode II, as shown
as example for single dowel-type connections in Fig. 7. The test programme con-
siders member depths from 150 mm up to 600 mm. The increase of the member
depth results in decreasing of the normalized specific critical fracture energy for
fracture mode I. Whereas the values respectively the corresponding curves are all
close together for the fracture mode II.

LVL-test series
2.5 GL-test series
Gspec b/F90 10





0.0 0.0
100 0.2
200 0.4
300 0.6
400 0.8
ar [mm] 500 he/h
600 1.0

Fig. 6 Comparison of specific critical frac- Fig. 7 Dependency of the specific critical
ture energies for mode I forGlulam and LVL fracture energies on the member depth for
test series modes I and II for Glulam test series

The splitting failure of dowel-type connections can be summarized using com-

mon failure criteria. The failure criteria describes the interaction between the frac-
ture modes I and II. The distribution of the numerical test series investigated are
compared with the linear-, quadratic- and Wu-failure criterion (1967), as shown in
Fig. 8. The assessment of the distribution of the specific critical fracture energies
for the test programme shows that the quadratic failure criterion mostly encloses

1.2 Glulam test series

LVL test series
1.0 Linear criterion 3.0
Quadratic criterion 2.5

Criterion from Wu (1967)

Gspec b/F90 10

Kspecific /Kc

0.6 1.5

0.0 0.0
200 0.4
0.0 300 0 .6
400 0.8
0.0 0.2 0.4 0.6 0.8 1.0 1.2 ar [mm] 500 he / h ]
Kspecific /Kc [-] 600 .0

Fig. 8 Failure criteria for dowel-type connec- Fig. 9 3-dim. curve of the normalized
tions loaded perpendicular to grain for fracture energies for mode I compared to
Glulam- and LVL test series the individual test results for solid wood/
Analysis of the Brittle Failure and Design of Connections Loaded Perpendicular 249

all failure cases for Glulam and LVL test series. The quadratic failure criterion
will be used for the prediction of the splitting failure behaviour of dowel-type
connections. The criteria found takes into account the important parameters of the
connection layout of single and multiple dowel type connections as well as the
ratio between the fracture modes I and II.

4 Design Approach

The load capacities of double shear connections loaded perpendicular to grain

depend on the connection layout, the number of fasteners and rows as well as the
loaded edge distance and the cross section of the beam. An influence on the posi-
tion of the connection along the span of the beam could not be observed and is
therefore not considered in the design approach, Franke and Quenneville (2011).
Depending on the connection layout and its position over the member depth,
the connection fails either ductile such as bending of the dowel or the embedment
failure of the wood or in splitting of the wood. Therefore the design for double
shear connections loaded perpendicular to grain has to be used in combination
with the European yield model for the prediction of the ductile failure behaviour
as given in Eq. (1).

Fductile ( EYM )
Fconnection = min (1)
Fsplitting = F90

For the design proposal for the splitting failure, the quadratic failure criterion de-
termined will be used to consider the interaction between the transverse tension
and shear failure. Substituting the individual critical specific fracture energies with
I , II
the distribution of the normalized fracture energies norm , the splitting load F90
for dowel-type connections in wood becomes:

b 103
F90 = kr (2)
c cII

WhereF90 in [N] is the load capacity depending on the splitting failure of the
wood. cI and cII [Nmm/mm] are the critical material fracture energies for the
fracture mode I or II and b [mm] the width of the member. The normalized frac-
I , II
ture energies norm enclose all individual critical specific fracture energies of the
connection considered. Therefore the critical specific fracture energy of each con-
nection layout investigated was normalized with the specimen width b and the
splitting load F90, see Eq. (3).
250 B. Franke and P. Quenneville

I , II
spec b
I , II
norm = (3)

The distribution of all values for solid wood respectively glulam test series were
expressed with a 3-dimensional group of curves, which depends on the loaded
edge distance he/h, the connection width ar [mm] and the member depth h, as
shown in Eq. (4) and Eq. (5) and for LVL as in Eq. (5) and Eq. (6). Fig. 9 shows
as example the 3-dimensional curve for the member depth h = 190 mm compared
to the individual values of the test series. The empirically determined Eq. (4), Eq.
(5) and Eq. (6) are based on more than 200 different connection layouts investi-
gated in solid wood, glulam or LVL.

norm =e
( h ( 20010h h
ar )) (4)

norm = 0.05 + 0.12 + 1 103 ar (5)

norm =e
( 0.81.6 h h
3 ar
110 ) (6)
The analysis of the results of the test programme shows, that the load capacity
increases with increasing the number of rows. For an increasing number of rows,
the load capacity increases and later on become constant as well as the stress situa-
tion, as shown in Fig. 10. This behaviour could be summarized using the quadratic
interaction of the areas of the tension stress perpendicular to grain and the shear
stress besides the dowels at the corner of the top row, see Franke and Quenneville
(2011). These results lead to the following factor kr to respect the effect of the
number of rows:

1 for n =1
kr = (7)
0.1 + ( arctan ( n ) )
for n >1

The results of test series, which enclose different numbers of rows with different
spacing between the rows, show that the load capacity does not increase when
spacing increases, Franke and Quenneville (2011). Therefore a dependency on the
spacing between the rows is not included in the factor kr.
For wider connections with more than two columns, e.g. nail plate connections
as shown in Fig. 11, the splitting load has to be determined as for the whole con-
nection with the complete connection width ar but also as for single connections
with the individual connection widths ar,i. The minimum of the load capacities of
Eq. (8) is the governing splitting load capacity of the connection.

F1 ( ar ,1 ) + F2 ( ar ,2 )
F90 = min (8)
F3 ( ar )
Analysis of the Brittle Failure and Design of Connections Loaded Perpendicular 251

Numerical test results
1.8 he/h = 0.44
he/h = 0.7
Factor kr [-]

1.0 Interaction of stress
kr Design proposal
0 1 2 3 4 5 6
Number of rows n [-]

Fig. 10 Factor kr, depending on the numerical Fig. 11 Design for nail plates or double
load capacities and stress situations deter- dowel-type connections

5 Discussion and Comparison to Current Design Standards

5.1 Comparison to Experimental Test Results

The design proposal is compared with experimental test series in solid wood, glu-
lam and LVL. Fig. 12 includes the correlation with the experimental test series in
Canadian spruce glulam done by Reshke (1999), Kasim (2002) and Lehoux
(2004). The test series covers single and multiple dowel-type connections.
Furthermore the design proposal are also compared with the experimental test
series from: Ballerini (2004, 2003, 1999) who investigated mainly different depths
of the member and different loaded edge distances; Mhler and Lautenschlger
(1989) who observed different numbers of rows, connection widths and loaded
edge distances; Ehlbeck and Grlacher (1989) who did tests with nailed steel-to-
wood connections; and Schoenmaker (2010) where test series encloses various
double-shear connections in European spruce. All are summarized in Fig. 12. The
comparisons are always related to Eq. (1), because for all cases of the experimen-
tal test results published the differentiation between the ultimate or splitting load
capacities is not given. The material values used for Canadian spruce glulam and
European spruce are cI = 0.225 Nmm/mm2 and cII = 0.650 Nmm/mm2 , as referenced in
Vasic (2000) respectively Larsen and Gustafsson (1990). Fig. 12 always shows a
close correlation in the comparison of the experimental test series and the design
Fig. 13 shows the comparison of the splitting load capacity and the design load
F90 for the experimental and numerical test series in LVL. In this case the splitting
load is known and the direct correlation to the new design approach can be shown.
The material values used for LVL are cI =1.0 Nmm/mm2 and cII = 6.0 Nmm/mm2 , as
referenced in Franke and Quenneville (2012), Ardalany et al. (2012).
252 B. Franke and P. Quenneville

Fig. 12 Comparison of design proposal and Fig. 13 Comparison of design proposal and
experimental test series published for Euro- experimental as well as numerical test se-
pean spruce or Canadian spruce ries for LVL

6 Conclusion

A new design proposal is presented for double shear connections in solid wood,
glulam and also LVL which allows the design of the splitting failure of the wood
due to connections loaded perpendicular to grain. The design approach is based on
fracture mechanic methods including the important parameters which influence
the load capacity of the connection. The comparison of the design results with
comprehensive experimental test series done in Canadian and European spruce
confirms the procedure of the design proposal. The good agreement is based on
over 200 different experimental test configurations and 600 numerical test results.

Acknowledgments. The research work was generously supported by the Structural Timber
Innovation Company (STIC) from New Zealand.

Ballerini, M.: A new prediction formula for the splitting strength of beams loaded by dow-
el-type connections. In: CIB-W18, paper 37-7-5, United Kingdom (2004)
Ballerini, M.: A new set of experimental tests on beams loaded perpendicular to grain by
dowel-type joints. In: CIB-W18, paper 32-7-2, Austria (1999)
Ballerini, M.: Beams transversally loaded by dowel-type joints: Influence on splitting
strength of beam thickness and dowel size. In: CIB-W18, paper 36-7-8, USA (2003)
Canadian Standards Association, O86-09 Engineering design in wood, Canada (2009)
Deutsches Institut fr Normung e. V., DIN 1052:2008 Entwurf, Berechnung und
Bemessung von Holzbauwerken Allgemeine Bemessungsregeln und
Bemessungsregeln fr den Hochbau, Berlin, Germany (2008)
Ehlbeck, J., Grlacher, R., Werner, H.: Determination of perpendicular to grain tensile
stresses in joints with dowel-type fasteners. In: CIB-W18, paper 22-7-2, Germany
European Committee for Standardization (CEN), EN 1995-1-1:2004 Design of Timber
structures, Brussels, Belgium (2004)
Analysis of the Brittle Failure and Design of Connections Loaded Perpendicular 253

Franke, S., Franke, B., Quenneville, P.: Analysis of the failure behaviour of multiple dowel-
type connections loaded perpendicular to grain in LVL. In: Proceedings of 12th World
Conference on Timber Engineering, New Zealand (2012)
Franke, B., Quenneville, P.: Investigation of the splitting behaviour of Radiata pine and
LVL under tension and shear. In: Proceedings of 12th World Conference on Timber En-
gineering, New Zealand (2012)
Franke, B., Quenneville, P.: Numerical modeling of the failure behavior of dowel connec-
tions in wood 137, 186, 155226 (2011)
Franke, B., Quenneville, P.: Analyses of the failure behaviour of transversely loaded dowel
type connections in wood. In: Proceedings of 11th World Conference on Timber Engi-
neering, Italy (2010)
Franke, B., Quenneville, P.: Failure behaviour and resistance of dowel-type connections
loaded perpendicular to grain. In: CIB-W18, paper 43-7-6, New Zealand (2010)
Ishikawa, H., Kitagawa, H., Okamura, H.: J Integral of a mixed mode crack and its applica-
tion. In: ICM 3, Cambridge, England, vol. 3 (1979)
Jensen, L.J.: Splitting strength of beams loaded perpendicular to grain by dowel joints.
Journal of Wood Science 51, 480485 (2005)
Kasim, M.H.: Bolted timber connections loaded perpendicular-to-grain - Effect of row
spacing on resistance. Royal Military College of Canada, Kingston, Canada (2002)
Larsen, H.J., Gustafsson, P.J.: The fracture energy of wood in tension perpendicular to
grain results from a joint testing project. In: CIB-W18, paper 23-19-2, Portugal (1990)
Lehoux, M.C.G.: Bolted timber connections loaded perpendicular to grain. A proposed
design approach. Royal Military College of Canada, Kingston (2004)
Ardalany, M., Deam, B., Fragiacomo, M.: Experimental results of fracture energy and
fracture toughness of Radiata Pine laminated veneer lumber (LVL) in mode I (opening).
Materials and Structures 45, 11891205 (2012)
Reshke, R.: Bolted timber connections loaded perpendicular-to-grain - Influence of joint
configuration parameters on strength. Royal Military College of Canada, Kingston
Schoenmakers, J.C.M.: Fracture and failure mechanisms in timber loaded perpendicular to
the grain by mechanical connections. Doctoral thesis, University of Technology Eind-
hoven, Netherlands (2010)
Van der Put, T.A.C.M., Leijten, A.J.M.: Evaluation of perpendicular to grain failure
of beams caused by concentrated loads of joints. In: CIB-W18, paper 33-7-7, The
Netherlands (2000)
Vasic, S.: Applications of fracture mechanics to wood. Doctoral dissertation, University of
New-Brunswick, Canada (2000)
Wu, E.M.: Application of fracture mechanics to anisotropic plates. Journal of Applied
Mechanics, Series E 34(4), 967974 (1967)
Structural Performance and Advantages
of DVW Reinforced Moment Transmitting
Timber Joints with Steel Plate Connectors
and Tube Fasteners

Daniel Brandon1,* and Adriaan Leijten2

Dept. of Civil Eng. and Architecture, University of Bath, UK
Structural Design Unit, Eindhoven University of Technology, The Netherlands

Abstract. This paper presents a study to the moment-rotation aspects of two 3-

member DVW reinforced timber connections with an inter-connecting steel plate
used as middle member. Previous studies showed that reinforcing dowel-type
timber connections with densified veneer wood (DVW) and using expanded tube
fasteners results in connections with superior structural properties compared to all
conventional connections. In this connection type, the DVW prevents premature
timber splitting. The tube fasteners aid a high initial stiffness, a high ductility and
a high reliability. A drawback of the connection, already in a 3-member connec-
tion, is the total thickness. By using only two side members and a much thinner,
steel middle member, the thickness is strongly reduced. The steel middle member
is used as a connecting interface in a flitch plate connection. This generally results
in a 50% reduction of the rotational stiffness. However, it is shown by an analyti-
cal and numerical study, that the rotational stiffness of two closely spaced, flitch
plate DVW connections acting in series remains unchanged if certain conditions
are fulfilled. Two full connection tests are performed to confirm the analytical and
numerical results. Additionally, the paper presents a comparison to a conventional
connection, which confirms the structural quality of the reinforced connection.

Keywords: densified veneer wood (DVD), DVD reinforced connections, ex-

panded tube fasteners, flitch plate connection, moment capacity, moment-rotation

1 Introduction
Moment connections in timber structures in general lead to limitations of struc-
tural strength. One conventional connection of this type is a moment connection

Corresponding author.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 255
DOI: 10.1007/978-94-007-7811-5_24, RILEM 2014
256 D. Brandon and A. Leijten


Fig. 1 The single connection (a), the flitch plate connection before assembly (b) and during
assembly (c)

with shear dowel-type fasteners. These dowel-type connections usually do not

exceed 20% of the strength of the connected beams. Also the rotational stiffness is
significantly lower than the bending stiffness of the beams.
A connection type developed in the 1990s showed a high rotational stiffness
and moment capacity (Leijten, 1998). This connection is shown in Fig.1a. Ex-
panded steel tubes are used as fasteners and a densified veneer wood (DVW)
plate is glued on each timber member to reinforce the timber. DVW is cross wise
layered plywood compressed at high temperatures and prevents timber splitting in
the connection. This material is significantly stronger and stiffer than timber and
shows almost isotropic material properties.
The low rotational stiffness in traditional moment connections can be blamed
on the low embedment stiffness of timber and the hole clearance necessary to
assemble the connection easily. The DVW reinforcement proposed by Leijten
(1998) has an embedment stiffness that is significantly higher than the embedment
stiffness of timber. Additionally, Leijten used hollow steel tubes instead of con-
ventional solid dowels. These easy fitting tubes are expanded with a hydraulic
jack after positioning them in the hole, Fig.1c. Not only does the hole clearance
disappear as a result of this process, but the connection becomes pre-stressed and
the initial stiffness increases significantly.
The total thickness of the DVW reinforced connection is a downside of the
connection type. This becomes even more significant in 5-member or multi-
member connections of this type. To reduce the thickness of the connection, two
connections are placed in series and have a much thinner steel plate as a middle
member, as can be seen in Fig.1b and Fig.1c. The two connections in series are in
this paper further denoted as one flitch plate connection.
Usually the rotation of a single connection s (Fig.2a) is half the rotation of a
flitch plate connection fp (Fig.2b) for any bending moment. The study presented
in this paper, however, shows in an analytical, numerical and experimental way
that the flitch plate connection can have a similar stiffness and a higher strength
than the single connection. In the flitch plate connection the connected timber and
DVW members that are in the same plane can suppress each others rotation. In
Structural Performance and Advantages of DVW Reinforced Moment 257

Fig.2c it can be seen that the rotation will cause a contact between these members
if the timber members are closely spaced. The contact force that comes to exist
becomes larger when the rotation increases. The moment of first contact between
these members is called gap closure in this paper. The significance of the rotation
suppressing effect is dependent on the size of the initial gap between the timber
members. The study presented in this paper predicts the relationship between rota-
tion fp and bending moment M analytically and numerically. It also confirms the
predictions with experiments.

a b c
s fp = 2s fp /= 2s


Fig. 2 The rotation difference of the single (a), the flitch plate connection without (b) and
with (c) rotation suppressing effect

2 Analytical Approach

The analytical approach aims to describe the bending moment-rotation curve for
the flitch plate connection. The rotation suppressing effect is studied for different
initial gap sizes between the timber members. The nonlinear load-slip behaviour
of a DVW reinforced connection with a single tube is previously studied by Lei-
jten (1998) and his results are used for this analysis.
As a basis for the analytical approach, assumptions are made. Firstly, it is as-
sumed that half of the flitch plate connection can be modelled as a single connec-
tion replacing the other half by a symmetry plane, as shown in Fig.3. Therefore,
the analysis is restricted to splice connections. Secondly, in the analytical model it
is assumed that the distance between the tube fasteners stays unchanged. There-
fore, the tubes rotate around one single rotation centre when the connection is
loaded. A translation can hereby be seen as a rotation around an infinitely far
point. Thirdly, it is assumed that applying the steel plate as a middle member does
not influence the load-slip behaviour of the tube connection. Lastly, it is assumed
that tube failure is governing the strength of the connection. Other failure mecha-
nisms are studied in the numerical and experimental work.
Fig.3 schematically shows a symmetrical half of the connection. If a bending
moment rests on the connection, the rotation centre R is initially located in the
gravity centre of the tubes. This will remain unchanged until the timber and DVW
members of both symmetrical halves come into contact. At that moment additional
contact force Fc will come to exist. To maintain equilibrium the rotation centre has
to move over the path of rotation centre given in the figure. An iterative calcula-
tion procedure is used to determine the location of the rotation centre on this path.
258 D. Brandon and A. Leijten

The height of the compressed contact surface hc is calculated by using the equilib-
rium of forces and geometrical requirements. If the results of both methods are
similar, the location of the rotation centre is determined.

Fig. 3 Forces in the symmetrical half of the connection before (left) and after (right) gap

3 Numerical Approach

The numerical approach aims to predict the moment rotation behaviour of the
DVW reinforced flitch plate connection including different failure mechanisms.
The numerical study consists of three-dimensional finite element analyses per-
formed using ABAQUS v6.8. This part of the research is only presented briefly in
this paper and will be more thoroughly discussed in a following paper by the au-
thors. The aims of the numerical approach were to study the failure mechanisms
that can occur and to predict the moment-rotation relationship.
Special attention is put on failure of the bond line between DVW and timber in
this study. The fractural mechanical properties of DVW-timber bond lines with a
polyurethane (PU) adhesive and with an epoxy (EP) adhesive that are used, are
presented by Brandon (2010). The bond line stress distribution around one tube is
studied using a local model (Fig.4a). Bond line stress distribution in other areas is
studied in a global model (Fig.4b). Cohesive elements, that are able to describe a
reduction in strength with an increasing slip, are used to describe the bond line
behaviour. The steel and DVW are modelled as isotropic plastic materials and the
timber is modelled as an orthotropic elastic material. Failures or other significant
phenomena are estimated per material as follows: the strain of DVW exceeds
0.07; the stress in the timber perpendicular to grain is smaller than -3.6 MPa or
larger than 1.6 MPa; the stress parallel to grain is smaller than -42 MPa or larger
Structural Performance and Advantages of DVW Reinforced Moment 259

than 42 MPa; or the tube displacement exceeds 9mm. The modelled thickness of
the timber and the DVW are 60 mm and 15 mm respectively. The modelled tube
diameter (d) is 18 mm and the edge and end distances are 3.5d. In the global
model, the beam depth is 300 mm and the DVW plate is 300 mm square.

Applied bending moment

Applied loads DVW glued on timber

Steel plate

DVW glued
on timber
Steel tube
Non linear
load slip relationship
Steel plate
a) Local model b) Global model

Fig. 4 Geometry of finite element models

4 Experimental Approach

The experimental study aimed to verify the analytical and numerical predictions.
A number of two four-point bending tests (Fig.5) were performed. The depth of
the timber members was 300 mm. The thickness of each member was 60 mm and
was chosen so that the connection was the weakest link in the structure. The initial
length of the tubes before expansion was 16% longer than the connection thick-
ness and the expanded tube diameter was 18 mm. The distances between the

Fig. 5 Full scale connection tests of splice connection

260 D. Brandon and A. Leijten

supports and the load points and between the two load points were 1600 mm. Two
hydraulic jacks applied the load simultaneously and both loads were measured
with a force gauge. These loads were kept similar, resulting in practically zero
shear forces and pure bending moments at the location of the connection. The
rotation and translations were measured between the steel plate and the timber

5 Results and Discussion

Fig.6 is obtained from the analytical study. It shows the bending moment-rotation
relationship of a single connection (grey curve) and flitch plate connections with
different initial gap sizes, tg (Fig.3) (black curves). The kink in the non-linear parts
of the black curves is caused by gap closure when DVWs compressive stresses
pick up. Current production practice in the Netherlands indicate that gaps are usu-
ally smaller than 4mm and, as such, moment-rotation curves are expected to run
between the two black curves on the furthest left. So although the influence of the
initial gap on the bending moment versus rotation is substantial, accurate assembly
resulting in a small initial gap, will make it a non-issue. Additionally it can be
seen that the flitch plate connection has the same rotational stiffness as a single
connection and an increased moment capacity, if the initial gap is small.

45 single connection
Bending moment (kNm)

35 flitch plate connection;
30 0 mm gap
25 flitch plate connection;
20 4.3 mm gap
flitch plate connection;
10 8.6 mm gap
flitch plate connection;
13 mm gap
0 0.05 0.1 0.15
Rotation (rad) flitch plate connection;
30.2 mm gap

Fig. 6 Influence of the initial gap size tg

Fig.7 shows the experimental, numerical and analytical results. Different num-
bered points on the numerically gained curve indicate significant points during
loading. At point 3 the tensile stress of the timber perpendicular to grain exceeds
1.6 MPa, resulting in cracking of the timber near the contact surface. However,
because of the glued-on DVW reinforcement, the size of the crack will be limited
and the structural effect will be insignificant. At point 4 the compression in the
timber will exceed 42MPa, resulting in yielding of the timber parallel to grain.
Structural Performance and Advantages of DVW Reinforced Moment 261

This will have a significant influence on the structural performance, since the
timber bears approximately 60% of the contact force. The dotted curve represent-
ing the numerical results with yield correction in Fig.7 is obtained by reducing the
bending moment after point 3 with the bending moment that is caused by the addi-
tional contact force in the timber after point 3. This contact force and its moment
arm are determined from the numerical results. The curve stops when the tube
deformation is 9 mm and tube failure occurs. It is concluded that bond lines with
PU and EP adhesives do not fail prior to the tubes in the flitch plate connection
with splice-type arrangement.
In Fig.7 it can be seen that the numerical study predicts the stiffness of the con-
nection well, but overestimates the strength of the connection. This is very likely
due to the yielding of the timber which is not taken into account in the model. The
analytical study gives a good prediction of the entire moment-rotation behaviour.

Bending moment (kNm)

50 3 Numerical results
40 Numerical results with
1 yield correction
Experimental results
10 Analytical results

0 Traditional connection
0 0.05 0.1 (Leijten, 1987)
Rotation (rad)

Fig. 7 Analytical, numerical and experimental results

6 Comparison with Conventional Moment Connections

This section aims to show the advantages of the DVW reinforced connection in
comparison with conventional connections. Results of Leijten (1987) who per-
formed tests of a conventional three member moment connection are used for the
comparison. The dimensions of the tested specimens are shown in Fig.8 and the
connection contained 10 steel dowels of 8 mm. The beam depth of the tested con-
nections was similar to the beam depth of the present experiments. The ends of the
beams were pulled inwards as shown by the arrows. Results of these tests are
shown in Fig.7. From this figure it can be concluded that the DVW reinforced
connection has a significantly higher moment capacity and the rotational stiffness
than the traditional connection. An even stiffer and stronger connection can be
obtained by using more tubes in one connection.
262 D. Brandon and A. Leijten





0 0
15 30

Fig. 8 Traditional connection test dimensions (Leijten, 1987)

7 Conclusion

An analytical and numerical study to the relationship between bending moment

and rotation of a densified veneer wood (DVW) reinforced flitch plate connec-
tion with expanded tube fasteners is performed. Also, experiments are performed
to verify the analytical and numerical models.
The studied flitch plate connection consists of two single connections with a
steel plate as the middle member acting in series (Fig.1a and Fig.1c). The flitch
plate connection, however, can have a higher rotational stiffness and moment
capacity than two single connections in series. This is caused by the rotation sup-
pressing effect that comes to exist when the DVW members and the timber mem-
bers come into contact due to rotations.
The most important findings can be summarised as follows:
the moment capacity of the flitch plate connection (Fig.1c) exceeds the moment
capacity of the single connection (Fig.1a), under the condition that the timber
and DVW members are closely spaced;
under the same condition, the stiffness of the flitch plate connection exceeds
the stiffness of two single connections;
the analytical approach accurately predicts the bending moment-rotation rela-
tionship of the connection;
the numerical approach accurately predicts the stiffness, but overestimates the
moment capacity, which is likely due to the ignorance of timber yielding in the
a comparison with a more conventional dowel connection showed the structural
quality of the DVW reinforced connection.
Structural Performance and Advantages of DVW Reinforced Moment 263

Brandon, D.: Mechanical properties of bond line between timber and densified veneer
wood, STSM report of COST E55. Short term scientific mission code: STSM-E55-
06127, Brussels, Belgium (2010)
Leijten, A.: Standard short duration tests on locally reinforced moment joints, 8 mm dowels
and bolts. Technische Universiteit Delft (1987)
Leijten, A.J.M.: Densified Veneer Wood Reinforced Timber Joints with Expanded Tube
Fasteners. PhD Thesis, Technische Universiteit Delft (1998)
Fully Threaded Self-tapping Screws Subjected
to Combined Axial and Lateral Loading with
Different Load to Grain Angles

Robert Jockwer1, Rene Steiger1 , and Andrea Frangi2

1 Empa, Swiss Federal Laboratories for Materials Science and Technology,
Structural Engineering Research Laboratory, Ueberlandstrasse 129,
CH-8600 Dubendorf, Switzerland
2 ETH Zurich, Institute of Structural Engineering IBK, Wolfgang-Pauli-Strasse 15,
CH-8093 Zurich, Switzerland

Abstract. In order to benefit from the advantages of fully threaded self-tapping

screws as reinforcing elements, it is essential to have detailed knowledge about the
strength and stiffness of screws with different shank to grain angles subjected to
combined axial and lateral loading. In this paper a design model is proposed for
the calculation of the load-carrying capacity and stiffness of screws with different
shank to grain angles subjected to loads perpendicular to the grain by accounting for
the effective embedment of the screw in the timber. The proposed model is based
on commonly used material properties and other established design models and fits
well the influence of the angle between shank and grain direction on the joints
capacity and stiffness. However, detailed knowledge about the specific input param-
eters in the design approaches is necessary in order to achieve a reliable prediction
of the load-carrying capacity and stiffness of the individual screw.

Keywords: self tapping screws, fully threaded screws, inclined screws, embedment
length, design approach, load carrying capacity.

1 Introduction
The use of self-tapping screws increased in recent years due to the development in
materialization and shaping of the screws. In contrast to other types of fasteners
screws provide the advantage of load transfer into the wood not only perpendicular
to the shank of the screw but also in axial direction and thus of making optimal
use of the high pull-out capacity of the screw. For reinforcement purposes screws
can be used to reinforce the timber in shear and perpendicular to the grain. They
are used to strengthen timber structural elements loaded in compression and tension
perpendicular to the grain and in shear.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 265
DOI: 10.1007/978-94-007-7811-5_25,  c RILEM 2014
266 R. Jockwer, R. Steiger, and A. Frangi

At notches combined shear stresses and tensile stresses perpendicular to the grain
occur. Screws are wanted to reinforce notches against both these acting stress com-
ponents. Due to the brittle failure behavior of timber, the reinforcement should be
as stiff as possible. However, screws exhibit their highest stiffness in axial direction,
whereas stiffness in lateral direction is lower. Hence, the load carrying behavior (in
terms of strength and esp. stiffness) of screws set in different angles to the grain and
loaded in shear and tension is of focal interest.
In order to benefit from the advantages of fully threaded self-tapping screws as
reinforcing elements at notches it is essential to develop detailed approaches for the
estimation of strength and stiffness of screws with different shank to grain angles
subjected to combined axial and lateral loading.

2 Current Design Approaches for Inclined Screws

Beside geometric parameters (outer thread diameter d or d1 , inner thread diameter
dcore , effective penetration depth le f , edge distances and spacing) the load carrying
capacity of screws is influenced by the yield moment My , the embedment strength
of the timber fh and by the withdrawal strength fax of the thread.
The design equations for screws loaded in lateral direction in Eurocode 5 (EC5)
[1] are identical to the ones for other dowel type fasteners. They are based on inves-
tigations by Johansen [2]. The design approach for screws loaded in axial direction
is based on studies by Bejtka [3]. In the design approaches only pure lateral and
pure tensile loading of the screws are considered.
For combined lateral and tensile loads (Fv and Fax , respectively) in EC5 it is
referred to the interaction formula for the calculation of the load-carrying capacity
of threaded nails (Equation 1). In this interaction formula the load to grain angle is
not taken into account. Information about how to account for the stiffness of screws
subjected to combined lateral and axial loading is not given in EC5.
 2  2
Fax,Ed Fv,Ed
+ 1.0 (1)
Fax,Rd Fv,Rd

Bejtka and Bla [4] propose a design approach for shear connections with in-
clined screws for different shank to grain angles . The resistance of a connection
with inclined screws results from the portions of the axial and the lateral resistance
of the screw (Rax and R , respectively) in direction of the interface of the members
and from the friction force between the members. Corresponding to failure mode 3
according to the Johansen model with plastic hinges in the screw, the resistance R
of a shear connection with one single inclined screw loaded in axial tension can be
calculated by Equation 2.

R = Rax ( sin + cos ) + R (sin cos ) (2)

Tomasi et al. [5] performed tests on shear connections with screws with dif-
ferent shank to grain angles . Connections with screws acting in tension, shear,
Fully Threaded Self-tapping Screws Subjected to Combined Axial 267

compression and combined tension and compression were studied. The approach
according to Equation 3 for calculating the stiffness Kser of the connection is pro-
posed, taking into account angle and both the stiffness in axial and lateral direction
(K and K , respectively).

Kser = K sin (sin cos ) + K cos (cos + sin ) (3)

In all these approaches the force acting on the connection runs in the parallel
to the grain direction of the member. This corresponds to a force to grain angle
of = 0 . However, when self-tapping screws are beeing used as reinforcement
for notched beams, combined parallel and perpendicular to the grain loading of
the cracked region occurs. This issue is not covered by the above equations but
will be further analysed in the following and an according design approach will be

3 Analytical Approach
In equivalence to the model proposed by Bejtka and Bla [4] in Figure 1b a design
model for the calculation of the effective embedment of the screw was developed
as illustrated in Figure 1a. When an inclined screw is loaded perpendicular to the
grain in tension, the embedment stresses are acting in direction of the timber surface.
At the very surface of the timber the layer resisting the embedment loads has zero
thickness. As a result zero embedment stresses can be carried by this layer. With
increasing loads compression and splitting failure will occur. The full embedment
strength fh can be achieved at a distance x1 from the surface, at which the timber
layer perpendicular to the screw shank is thick enough to transmit the embedment
stresses along two shear planes at both sides of the screw. Depending on the screw
shank to grain angle rolling shear fv,roll is the dominating strength parameter. The
resulting length x1 along which only reduced embedment stresses can be transmitted
is given in Equation 4 and illustrated in Figure 1a.

f h de f
x1 = (4)
2 tan fv,roll

Based on Johansens model [2] the load-carrying capacity of an inclined screw

loaded perpendicular to the grain can be calculated using the reduced embedment
length. As a simplification, zero embedment strength is assumed within length x1
along the screw. The resulting lateral load-carrying capacity of the screw is given
in Equation 5. The withdrawal capacity of the screw can be calculated using the ef-
fective length reduced by x1 . The resulting load-carrying capacity on the joint with
an inclined screw loaded perpendicular to the grain can be determined according to
Equation 2. Due to the opening of the joint, no friction between the members will
occur ( = 0).
268 R. Jockwer, R. Steiger, and A. Frangi



x1 Fax Fax

(a) (b)
Fig. 1 Forces and stresses in joints with inclined screws loaded (a) perpendicular and (b)
parallel to the grain

R = f h x1 d e f + 2My + fh x21 de f fh de f (5)

The stiffness of the joint is markedly influenced by the embedment of the screw. As
a simplification, again zero embedment strength is assumed within length x1 along
the screw. As a further simplification, clamping of the screw beyond length x1 is
assumed for the stiffness in lateral direction K as given in Equation 6. The axial
stiffness is calculated using the effective length reduced by x1 .

3 Esteel dcore
K = (6)
64 x31

According to [4] the capacity of the joint consists of the portions of the capacity of
the screw in lateral and in axial direction. The stiffness of a joint loaded parallel to
the grain was proposed by Tomasi et al. [5] to be calculated as a sum of the portions
resulting from the stiffness in lateral and in axial direction. In contrast the stiffnesses
in the joint when loaded perpendicular to the grain are acting in series. The effective
stiffness Kser is equal to the sum of the inverse values of the stiffnesses in lateral and
axial directions (K and K respectively) (Equation 7):

1 1 1
= + (7)
Kser K K

4 Experimental Verification of the Proposed Model

Tests have been carried out in order to determine the capacity and the stiffness of
joints with screws with different shank to grain angles and loaded under different
load to grain angles. Within these test series, the parameters shank to grain angle
, load to grain angle and density have been varied. In two different test se-
tups three test series with = 90 , 60 and 45 were tested using two groups of
Fully Threaded Self-tapping Screws Subjected to Combined Axial 269

110 110




(a) (b)
Fig. 2 Test-setups for (a) pulling and (b) shearing tests of joints with inclined screws

samples with high and low timber density. The test setup for the application of a
tension force perpendicular to the grain ( = 90 ) is illustrated in Figure 2a and
named pulling test. The second test setup for a force applied parallel to the grain
( = 0 ) is similar to a block shear test and given in Figure 2b. The tests were run
displacement controlled with constant speed of load. In order to reach ultimate load
within approximately 5 min a speed between 1.0 and 3.0 mm/min was chosen.
Norway spruce glulam of high and low density was produced out of two sam-
ples (each of around 100 boards) which have been selected from a bulk of over
500 boards with mean = 419 kg/m3 (CoV = 12.4%), k = 344 kg/m3 . The
bulk of boards had a density corresponding to solid timber of strength class C24
according to EN 338 [6] (mean = 420 kg/m3, k = 350 kg/m3). According to
prEN 14080 [7] the laminations of glulam of strength class GL24h correspond
to solid timber of strength class C24. The moisture content of the glulam speci-
mens when tested was approximately MC = 10%. The density of the glulam was
mean,high = 464 kg/m3(CoV = 2.5%) for the specimens with high density and
mean,low = 360 kg/m3(CoV = 1.8%) for the specimen with low density.
Fully threaded self-tapping screws of type SFS WR-T-13xL400 as specified in [8]
have been used in the tests. The screws have an outer thread diameter d1 = 13 mm,
an inner core diameter of dcore = 8.5 mm and a length of l = 400 mm. In the national
technical approval published by the Deutsches Institut fur Bautechnik [8] a yield
moment of the screws of My,k = 80.0 Nm is given. Furthermore the characteristic
withdrawal parameter perpendicular to the grain is specified as fax,k = 80 106 k2 .
Two parallel screws were used per joint. The distance between the screws perpendic-
ular to the grain was 60 mm or 4.5d and the edge distance perpendicular to the grain
was 40 mm or 3d. The effective length of the threaded part of the screw in the side
members of the block shear tests and the pulling tests was le f = 110 mm, 127 mm
and 155 mm or 8.5d, 10d and 12d for screw axis to grain angles of = 90 , 60
and 45 , respectively.
270 R. Jockwer, R. Steiger, and A. Frangi

40 40
= 90 = 90
35 = 60 35 = 60
= 45 = 45
30 30

Load per screw [kN]

Load per screw [kN]

25 25

20 20

15 15

10 10

5 5

0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Opening displacement [mm] Shearing displacement [mm]

(a) (b)
Fig. 3 Joint behaviour in (a) pulling and (b) shearing tests. Individual results in grey, nor-
malised values for = 420 kg/m3 in black with solid line for = 90 , dashed-dotted line
for = 60 and dashed line for = 45 .

In the gap between the side members and the central member of the shear tests a
teflon foil of 1.5 mm thickness was placed in order to avoid frictional forces between
the members. With this foil inserted, the members were screwed together tightly.
The members of the pulling tests were screwed together tightly without such a teflon

5 Results and Discussion

The load-displacement diagrams are shown in Figure 3. Shearing and opening dis-
placements had been measured in the timber at the location of the screws. In order to
compare the series with high and low density the results are normalized with regard
to their density to a reference density of 420 kg/m3. The effective stiffness Kser is
determined in the range of 0.1 Fult and 0.4 Fult by linear regression following EN
12512 [9]. In EN 12512 a limit deformation of 30 mm is specified. However, due
to the intended application of the screws as reinforcement, a limit deformation of
15 mm was chosen. Thus, the ultimate load Fult was taken as the minimum of the
highest load and the load at 15 mm deformation.
The ultimate load Fult and the stiffness Kser normalised to a density =420 kg/m3
are given in Table 1. The results of the two test series in shearing and pulling with
= 90 serve as a reference in the common design approaches as specified in EC5
[1] and in Bejtka and Bla [4]. The required material strength parameters used in
these approaches can be found in literature: fh,k , fax,k and K in EC5, fh,mean and
fax,mean in Bejtka [3], My,k and K in [8]. For the calculation of the load-carrying ca-
pacity and the stiffness according to Equation 5 and Equation 7 the material property
values published in [3] and [8] were used.
Fully Threaded Self-tapping Screws Subjected to Combined Axial 271

Table 1 Test results compared to the proposed design model and to values from literature

Load-carrying capacity Stiffness

Tests EC5 [4] Eq. 5 Tests [5] Eq. 7
Fult Rint,k Rmean Rmean Kser Kser Kser
[ ] [kN] (CoV) [kN] [kN] [kN] [kN/mm] (CoV) [kN/mm] [kN/mm]
Pulling 90 21.2 (3%) 16.2 18.7 18.7 28.3 (25%) 17.9
60 21.6 (4%) 12.3 16.3 8.9 (15%) 6.5
45 17.7 (8%) 11.8 15.3 3.0 (17%) 3.4
Shearing 90 10.5 (5%) 6.8 6.2 1.9 (22%) 2.5
60 20.5 (4%) 8.6 15.0 5.9 (8%) 6.3
45 27.4 (10%) 11.8 20.3 13.6 (16%) 10.2

All design approaches give lower load-carrying capacities for = 90 compared

to the ultimate loads reached in the tests. The ratio of the mean values from tests
and the characteristic values of the capacities determined according to EC5 is in the
range of 1.3 to 1.5. Due to the low variation of the test results, the characteristic
value of the test results is expected to be higher than the characteristic value accord-
ing to EC5. The stiffness measured in the pulling tests for = 90 is larger than the
one calculated according to the national technical approval of the screws [8]. The
stiffness in the shearing tests for = 90 is lower than the one calculated according
to EC5.
The influence of the angle is modelled in a appropriate way by the proposed
approach of reducing the embedment length (Equation 5 and Equation 6). The de-
viation of the predicted load-carrying capacities and stiffnesses from the test re-
sults can be eliminated by using more precise input parameters valid for the specific
screws used in the tests instead of the general input parameters proposed by stan-
dards and literature: A better fit of the stiffness Kser in pulling is achieved when a
larger stiffness in withdrawal direction K = 40 d le f is used instead of K = 25 d le f
as proposed in [8]. The mean withdrawal strength as suggested by Bejtka ([3]) for
a variety of self-tapping screws is fax,mean = 13.1 N/mm2 wherereas for the screws
used in the tests a larger value of fax,mean = 14.8 N/mm2 was found.
The length x1 according to Equation 4 is calculated using the material property
values given in prEN 14080 and given by Bejtka [3]. A maximum length of approx-
imately x1 = 30 mm occurs in the pulling tests for = 45 . This length corresponds
well to the observations made in the tests.

6 Conclusions
Joints with fully threaded self-tapping screws show highest load-carrying capacity
and stiffness when the screws are loaded in axial direction. The more the screw is
loaded in lateral direction the lower are the capacities and the stiffnesses. By choos-
ing an appropriate angle between screw shank and grain direction, the behaviour of
the joints can be optimised. Existing design approaches do not cover inclined screws
272 R. Jockwer, R. Steiger, and A. Frangi

loaded perpendicular to the grain. The model proposed in this paper gives a good
estimate of the structural behaviour of inclined screws loaded perpendicular to the
grain. By accounting for a reduced embedment length the strong decrease in stiff-
ness can be adequatly modelled. The reduction of the embedment length strongly
depends on the properties of the timber and the screw. The values for these prop-
erties given in literature differ strongly which leads to big variations in estimated
load-carrying capacity and stiffness. For a reliable estimate of capacity and stiffness
a precise knowledge of the actual properties is necessary. Precise property values
for the specific screw used in the joint should be used instead of general property
values for dowel-type fasteners as can be found in EC5.

1. European Committee for Standardization CEN, EN 1995-1-1: Eurocode 5: Design of
timber structures - Part 1-1: General - Common rules and rules for buildings + AC (2006)
+ A1 (2008). CEN, Bruxelles, Belgium (2004)
2. Johansen, K.W.: Theory of Timber Connections. IABSE Publications 1949(9), 249262
3. Bejtka, I.: Verstarkung von Bauteilen aus Holz mit Vollgewindeschrauben. Ph.D.Thesis,
Fakultat fur Bauingenieur-, Geo- und Umweltwissenschaften, Karlsruhe University,
Karlsruhe, Germany (2005)
4. Bejtka, I., Bla, H.J.: Self-tapping screws as reinforcement in beam supports, Paper 39-
7-2. In: Proc. of CIB-W18 Meeting, Kyoto, Japan, vol. 39 (2002)
5. Tomasi, R., Crosatti, A., Piazza, M.: Theoretical and experimental analysis of timber-to-
timber joints connected with inclined screws. Construction and Building Materials 24,
15601571 (2010)
6. European Committee for Standardization CEN, EN 338: Structural timber - Strength
classes. CEN, Bruxelles, Belgium (2009)
7. European Committee for Standardization CEN, prEN 14080: Timber structures - Glued
laminated timber and glued solid timber - Requirements. CEN, Bruxelles, Belgium
8. Deutsches Institut fur Bautechnik DIBt, Z-9.1-472, Allgemeine bauaufsichtliche Zu-
lassung, SFS Befestiger WT-S-6,5; WT-T-6,5; WT-T-8,2; WR-T-9.0 und WR-T-13 als
Holzverbindungsmittel. DIBt, Berlin, Germany (2011)
9. European Committee for Standardization CEN, EN 12512: Timber structures - Test
methods - Cyclic testing of joints made with mechanical fasteners + A1 (2005). CEN,
Bruxelles, Belgium (2001)
10. European Committee for Standardization CEN, EN 26891: Timber structures - Joints
made with mechanical fasteners - General principles for the determination of strength
and deformation characteristics (ISO 6891:1983). CEN, Bruxelles, Belgium (1991)
Alternative Approach to Avoid Brittle Failure
in Dowelled Connections

Daniela Wrzesniak1, Massimo Fragiacomo2, and Andr Jorissen3

DIA Department of Engineering and Architecture, University of Trieste,
Piazzale Europa 1, 34127 Trieste
DADU Department of Architecture, Design and Urban Planning,
University of Sassari, Piazza Duomo 6, 07401 Alghero
Eindhoven University of Technology, P.O. Box 513, 5600 MB, Eindhoven, and
SHR Timber Research, Wageningen, The Netherlands

Abstract. Ductile behavior of timber connections with metal fasteners is essential

to achieve a robust structure. Moreover, a ductile behavior of the fastener and the
timber prior to failure is necessary to fulfill the boundary conditions for applying
the Johansen theory (1949). If these boundary conditions are not fulfilled, the
capacity of a connection is overestimated and brittle failure may occur. However,
using sufficient spacing, end and edge distances reduces tensile stresses
perpendicular to the grain, the main stresses initiating brittle failure. If in addition
to that the influence of the relation between fastener diameter and timber volume
is considered, brittle failure mechanisms can be avoided.
This paper discusses quasi static tests carried out on dowelled connections with
different spacing and loaded end distances chosen in accordance to the above
criterion. The results were compared with the capacity calculation for dowelled
connections of Eurocode 5, chapter 8.2, the design approach against block shear
failure of Eurocode 5, Annex A, and the design proposal for the avoidance of
block shear failure proposed by Hanhijrvi and Kevarinmki (2008). The
experimental failure load achieved was in all cases higher than predicted values by
all three design approaches.

Keywords: dowelled connections, brittle failure, block-shear failure, splitting.

1 Introduction

Johansens theory, which was presented in 1949 and later extended by Meyer in
1957, is a widely used and acknowledged design method for timber connections
with metal fasteners. This theory is fully incorporated in Eurocode 5 (EC5, EN
1995-1-1:2004) and some other design codes.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 273
DOI: 10.1007/978-94-007-7811-5_26, RILEM 2014
274 D. Wrzesniak, M. Fragiacomo, and A. Jorissen

It is well known that in Johansens theory a full-plastic behavior of the timber

and the fasteners is assumed. Both the dowel in bending and the timber in
embedding are assumed to attain the plastic state prior to failure.
For both materials this looks a logical assumption since also timber in
compression responses Eplastically. To satisfy these assumptions, Johansen noted
that basic design rules such as sufficient spacing between fasteners and sufficient
edge and end distances have to be taken into account.
For dowelled connections he suggested a minimum spacing between fasteners
of a1=10d, a loaded end distance of a3=7d and an edge distance of a4=3d (d =
dowel diameter) to ensure timber plasticization and the connection failure
according to his theory (Figure 1).

Fig. 1 Required spacing and distances between fasteners for dowelled connections
(Johansen, 1949)

Later studies refined and extended these design rules concerning the spacing,
edge and end distances, the slenderness ratio and the effective number of dowels
in a row (Jorissen, 1998). The slenderness ratio is defined by
t with t = timber
thickness and d = dowel diameter. An optimized slenderness ratio was suggested
by Mischler (1998):
1.4 (1)

with Mu,95 being the upper 5th percentile of the bending strength of the dowel and
fh,05 being the lower 5th percentile of the embedding strength of timber.
As indicated in Eq.(1) and pointed out by Mischler the strength of the steel,
spacing and distances should be related to the properties of the timber.
However, brittle failure in timber structures such as splitting along the row of
fasteners and tearing out of timber parts has been observed ever since. Great effort
was made by numerous scientists to describe especially the plug-shear and block-
shear failure mechanisms, which have been found inadequate in Annex A of
Eurocode 5 (Zarnani and Quenneville (2012), Johnsson (2003), Hanhijrvi et al.
(2006)). An alternative design approach was presented by Hanhijrvi and
Kevarinmki (2008), who suggested that the connection should be divided in inner
and outer parts, and their capacity should be calculated independently as indicated
in Figure 2.
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 275

Inner Part

Outer Parts

Fig. 2 Example of a timber specimen with dowels indicating the division of inner and outer
parts after Hanhijrvi and Kevarinmki (2008)

Due to its complexity, the approach results in difficult application in common

An extensive literature research has been conducted thereafter by the writers,
realizing that researchers specifically dealing with block and plug shear failure
mechanisms came to similar conclusions for the avoidance of this failure
Kangas and Kevarinmki (1997 and 1998) established a model which optimizes
the distances and spacing between nails in timber joints. They also came to
the conclusion that in order to avoid block shear failure, the spacing between
nails and the joint area have to be increased to achieve the full capacity of the
Johnsson (2003) suggested that in order to avoid plug shear failure the joint
width should be increased and the number of fasteners in a row should be
Kairi (2004) came to the conclusion that slenderness ratio and number of
dowels should be optimized such that fastener yielding is attained before timber
failure occurs.
Although the brittle failure mechanisms dealt with in the above literature were
mainly related to nailed connections, the authors agreed upon the application of
basic design principles to all fastener types such as fastener spacing, diameter and
timber volume to avoid brittle failure mechanisms like plug shear or block shear
Based on the literature research it was decided to conduct tests on dowelled
connections aiming to show that considering basic design rules, brittle failure
mechanisms like those described by Hanhijrvi and Kevarinmki (2008) and in
Annex A of Eurocode 5 can be prevented.
276 D. Wrzesniak, M. Fragiacomo, and A. Jorisseen

2 Experimental Tests

35 steel-timber-steel dow wel connection specimens were tested to failure at thhe

laboratory of the Neue Holzbau
H AG in Lungern, Switzerland. The test set up annd
load apparatus is presenteed in Figure 3.


Steel plates
fixed to both sides
of the specimen
S235, t=5mm

Timber Board
ft,mean = 23N/mm^2

Steel Block

Front View
Specimen Load

Fig. 3 Test Set-up and Load application

2.1 Geometrical Prroperties

The specimens were consstructed with either one or two rows of 12 mm diameter
dowels. The number of dowels
d in a row as well as edge and end distances varieed
(Figure 4). A maximum of 4 dowels per row were used. The fastener spacing a1
was 7.5d which is less th
han suggested by Johansen but higher than the minimum m
values required accordingg to Eurocode 5 (5d for forces parallel to the grain) annd
Swiss standard SIA 265 (7d). The distance between dowels perpendicular to thhe
grain was chosen to be 5dd (minimum values according to Eurocode 5 and SIA arre
both 3d).
An additional test wass carried out with 4 dowels, a spacing a1 of 15d and a
loaded end distance a3 off 12d. The aim was to investigate the influence of endd-
distance and spacing on the load carrying capacity and the embedment strengtth.
The timber board thickneess is 40 mm. Detailed specimen properties are presenteed
in Table 1.
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 277

a3 90 90 Varies 90 90 a3

a4 Varies

a4= 30, 40, 60, 75mm

a3= 30, 40, 50, 60, 70, 90, 120mm


a4= 30, 40, 60, 75mm

a3= 30, 40, 50, 60, 70, 90, 120mm


a4= 30, 40, 60, 75mm

a3= 30, 40, 50, 60, 70, 90, 120mm


a4= 30, 40, 60, 75mm

a3= 30, 40, 50, 60, 70, 90, 120mm
12 Dowels
Timber Boards
of ft,mean=23N/mm^2

Fig. 4 Detailed layout of test specimens

2.2 Material and Loading Protocol

Ten boards of European spruce with similar grain pattern, density and moisture
content were visually selected out of 100 boards. The Youngs Modulus was
determined by measuring the sound wave velocity, which ranged from 5850 to
6150 m/s. The timber boards had a characteristic tensile strength ft,0,k = 20 N/mm2
according to their strength grade.
Knots and defects such as compression wood were cut off to achieve
representative results. The density of the boards was measured at a moisture
content of w 10% varied between 420 and 500 kg/m3 and 12 mm diameter
dowels of high strength steel (ETG 100, fu = 1000 N/mm2, fy = 900 N/mm2),
smooth surface, and steel plates (S235) with a thickness of 5 mm were used for the
tests. The holes in the steel plates were oversized by 0.5 mm, whereas the holes in
the boards were drilled with diameter of 12 mm. The tensile force was applied by
a hand pump (ENERPAC). The pump speed was applied to reach failure after
approximately 5 minutes.
278 D. Wrzesniak, M. Fragiacomo, and A. Jorissen

3 Test Results

3.1 Observations and Analysis of Failure Loads

The slenderness ratio of the dowel was 3.33 in all cases. With a steel side plate
thickness of 5 mm and a dowel diameter of d = 12 mm the steel plates can be
regarded as relatively thin:

8 4 4 900
< = 6,2
, 3 , 3 30

with fh,k being the embedment strength according to Eq. (2). The fasteners can be
regarded stocky (Jorissen, 1998). Stocky dowels of high strength class were
chosen to avoid any ductile behavior of the fastener in order to verify the approach
presented by Hanhijrvi and Kevarinmki (2008/9).
The observed failure mechanisms were all brittle, such as row splitting, cracks
along the row of the fasteners, tensile failure of the cross section or side parts of
the timber. However, for large spacing (up to a1= 12d) brittle failure was observed
after large deformations, indicating that the brittle failure mode is not governing.

Fig. 5 Photos showing different failure mechanisms, from top: 1st and 2nd =Splitting and
Cracks along the row, 3rd =Row Shear combined with tensile failure, 4th =Tensile failure
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 279

Table 1 Summary of Specimen Properties and Test Results

Test Results
One Row a4=30, 4Dowels
No. of
Name of Dowels Number Failure Failure
Specimen Distances in row of rows Density Load Mode
a4 a3
mm mm kg/m^3 kN
30/30-1 30 30 468.0 35.2 SC
30/50-1 30 50 464.4 65.7 SC
4 1
30/70-1 30 70 458.6 64.5 T
30/90-1 30 90 447.5 63.9 Td
One Row a4=40mm, 4Dowels
40/30-1 40 30 430.9 20.3 I/C
40/50-1 40 50 452.3 55.6 SCd
40/60-1 40 60 478.5 67.3 SCd
4 1
40/70-1 40 70 494.8 63.7 SC
40/90-1 40 90 447.7 66.1 SCd
40/120-1 40 120 503.9 69.1 T
One Row a4=60mm, 4Dowels
60/40-1 60 40 524.3 38.3 SC
60/50-1 60 50 472.2 44.3 SCd
4 1
60/70-1 60 70 466.1 61.7 SCd
60/90-1 60 90 451.7 72.8 Scd
One Row a4=75mm, 4Dowels
75/60-2 75 60 479.8 - I/C
4 1
75/120-2 75 120 470.9 157.3 T
Two Rows a4=40mm, 4Dowels
40/60-2 40 60 4 2 500.8 122.5 SC
Two Rows a4=60mm, 4Dowels
60/60-2 60 60 481.0 62.5 RST
4 2
60/120-2 60 120 472.0 114.6 RST
One Row a4=30, 3Dowels
30/40-1.1 30 40 454.1 33.5 SCd
30/50-1.1 30 50 497.2 48.5 SC
30/60-1.1 30 60 469.0 53.2 T
3 1
30/70-1.1 30 70 484.0 49.1 SCd
30/90-1.1 30 90 447.3 39.1 SCd
30/90-1.1b 30 90 456.6 53.4 T
One Row a4=40mm, 3Dowels
40/60-1.1 40 60 440.0 55.3 Sd
3 1
40/90-1.1 40 90 459.0 56.1 Sd
One Row a4=60mm, 3Dowels
60/60-1.1 60 60 451.0 44.9 S
3 1
60/90-1.1 60 90 436.0 53.1 S
Two Rows a4=40mm, 3Dowels
40/60-2.1 40 60 469.0 87.0 RST
40/90-2.1 40 90 3 2 464.0 106.3 RST
40/120-2 40 120 489.7 104.2 Td
Two Rows a4=45mm, 3Dowels
45/60-2.1 45 60 456.0 76.9 RST
3 2
45/120-2.1 45 120 453.0 103.3 RST

Table 1 contains the geometrical and mechanical properties of the specimens,

and the related failure load and failure mechanisms. The following abbreviations
280 D. Wrzesniak, M. Fragiacomo, and A. Jorisseen

are used: SC=splitting and d/or cracks along the row, T=tensile failure of the timbeer
cross-section, RST=row w shear combined with tensile failure, I/C=teest
interrupted/test cancelled d. Furthermore, an index d (from primarily ductile) is
added to the failure modees which first showed relatively large timber deformatioon
before brittle failure (e.g. SCd, Td and RSTd).
The different failure modes
m are displayed above (Figure 5).
The experimental failu ure load of each specimen per dowel was then compareed
with the embedding streength which was calculated using the semi empirical
formula of Eq. (2), Euroco ode 5.

, = 0.082 (1 0.01 ) (22)

The calculated embeddin ng strength was then turned into a load per fastener bby
multiplication with timbeer board thickness t and dowel diameter d and compareed
with the load per fasteneer achieved during testing. The results are presented iin
Figure 6 and Figure 7.
Note that the embeddiing strength of each specimen was calculated using thhe
density of each specimen obtained during testing.
It was realized that thee embedding strength of the timber was reached when thhe
loaded end distance a3 waas greater than 60 mm (5d) (Figure 6) or the a3/a4 rattio
was greater 1.5 (Figure 7)).
This indicates that althhough stocky dowels were used, if distances and spacinng
are kept sufficiently larrge, the maximum timber capacity in regards to thhe
embedding strength can sttill be reached.

Fig. 6 Comparison between n experimental failure load per dowel and embedding strenggth
(based on EC5) vs. end distaance a3
Alternative Approach to Avo
oid Brittle Failure in Dowelled Connections 2881

Fig. 7 Comparison between n experimental failure load per dowel and embedding strenggth
(based on EC5) vs. a3 / a4 rattio

Furthermore, the failure load was compared with the design methods for thhe
avoidance of brittle failurre mechanisms due to block shear described in Eurocodde
5, Annex A and by the forrmulas presented in the research report VTTS-07046-009
(Hanhijrvi and Kevarinm mki 2009).
The formulas of Euroccode 5, Annex A for the avoidance of block shear failurre

1.5 , , ,
, = (33)
0.7 , ,

with , and , be
eing the net areas of the cross-section perpendicular tto
the grain and parallel to o the grain and , , and , being the characteristtic
tensile and shear strengthh of the timber respectively. The results are presented iin
Figure 8a to 8d. Altho ough the failure mechanism due to the unfavorabble
relationship between dow wel diameter and board thickness were brittle, the desiggn
approaches discussed ab bove underestimated the load-carrying capacity of thhe
connection by far. The vaalues obtained during testing were on average three timees
higher than predicted usin ng the Finnish approach and twice as high as the valuees
calculated based on the Annex
A A of EC5.
The test results in Figgure 8.d show higher results for the connection with aan
edge distance a4 of 40mm m. A reason for this is assumed to be a more evenly loaad
induction into the group of
o fasteners due to a better relationship between a3 and a4,
that is a3 and a4 being moore equal to one another. In addition, the test results werre
compared with Johansens equations of Eurocode 5 concerning double shear steeel-
to-timber connections witth thin steel plates as presented in Eq. (4).
282 D. Wrzesniak, M. Fragiacomo, and A. Jorissen

Fig. 8.a

Fig. 8.b

Fig. 8.c
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 283

Fig. 8.d

Fig. 8.a to 8.d Comparison between observed and predicted failure loads using EC5, EC5
Annex A, and Hanhijrvi and Kevarinmki (2008/9) approach versus end distance a3. Fi
indicating the results based on Hanhijrvi and Kevarinmki, a4 is the edge distance a

0.5 , ,
, = (4)
1.15 , , , +

The prediction of the load carrying capacity is the minimum of the values
obtained by the formulas presented in Eq. (4) and being the characteristic
withdrawal capacity of the fastener.
The approximate analytical equation of EC5 was used to calculate the yield
, = , 0.3 . (5)

with d signifying the dowel diameter of 12mm and , being its characteristic
tensile strength of 1000 / .
The failure load based on the embedding strength was always governing. This
is due to a low slenderness ratio and the great strength of the dowels. The results
are presented in Figure 8.a to 8.d and it was found that the failure load is predicted
well by the Eurocode 5 equations, which are always conservative.

3.2 Analysis of the Effective Number of Fasteners

The effective number of dowels was calculated for each specimen by dividing the
failure load per dowel by the load per dowel derived from the embedding strength
as described in Chapter 3.1.
This number was then compared with the effective number of dowels
calculated using the Eurocode 5 formula, Eq. (6):
284 D. Wrzesniak, M. Fragiacomo, and A. Jorisseen

= .

where n is the actual number of fasteners. To obtain a representative value thhe

edge distance for the Eurrocode value was set to 80mm as recommended in Tabble
8.5, Eurocode5. In the serries with one row of 4 dowels and one row of 3 dowells,
the effective number off fasteners was greater than the effective number oof
fasteners calculated accorrding to the Eurocode 5 and equal to the actual number oof
fasteners when the conditiion a35d is satisfied (Figure 9).

Fig. 9 Effective number of fasteners

f neff versus end distance a3 for specimens with one roow
of 4 dowels

In the series with two rows

r of 3 and two rows of 4 dowels the effective numbeer
of fasteners was always greater
g than the effective number of fasteners calculateed
according to Eurocode 5 when
w a3 was greater than 5d (Figure 10).

Fig. 10 Effective number of fasteners neff versus end distance a3 for specimens with tw
rows of 3 dowels
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 285

Therefore, the effective number of dowels in a connection can be increased by

simply increasing the end and edge distances as well as the spacing between

3.3 Analysis of Embedding Capacity

In the following, the influence of fastener spacing, end and edge distances on the
embedding strength was investigated. During testing it was observed that
specimen 75/120-2 which was constructed with two rows of four dowels and an
end and edge distance of 120 mm (10 d) and 75 mm (6.5d) respectively, failed at a
load which was about 2.4 times higher than predicted based on Eurocode 5, for
which the embedment strength was calculated according to equation (2) using the
actual density.
An additional test with four bolts (two rows of two dowels), an end distance of
12d (144 mm) and a spacing a1 of 15d (180 mm) was carried out. At a spacing of
a1 of 15d, no interaction between dowels is expected.

Fig. 11 Plastic deformation in timber

As a result, the failure load reached in the test was Fmean=22 kN per dowel and
the deformation in the timber was around 5 mm (Figure 11) prior to failure. Based
on EC5, Eq. (2), using a characteristic timber density of = 0.7 = 479
0.7 = 368.5 / , the estimated embedding strength is 26.59 N/mm2.
Converting the experimentally obtained failure load per dowel into the embedding
strength by dividing the load by the timber board thickness and the dowel
diameter, the value obtained is 32.08 N/mm2.
The difference between the two values is considerable.
The derivation of the embedding strength is defined in EN 383 (Chapter 6,
Table 1) and provision is made for the use of an end distance of 7d. This would
indicate that current values for the embedding strength are only valid for an end
distance of 7d. Increasing end distances and bolt spacing leads to a higher load
carrying capacity and, consequently, to higher embedment stresses.
Therefore, the embedment strength seems to be not only dependent on the
density and the diameter of the fastener but also on the design of the connection.
However, further analyses of tests already carried out and probably additional tests
are necessary to confirm this assumption.
286 D. Wrzesniak, M. Fragiacomo, and A. Jorissen

4 Conclusions
The aim of this study was to show that focusing on simple design principles
concerning the end and edge distances and spacing is sufficient to achieve a
ductile failure mechanism, hence fulfilling the requirements of the Johansen
theory and therefore predicting the capacity of timber connections with dowel
fasteners accurately. Since the slenderness ratio was kept constant ( = 3,33) the
expected positive effect of a large slenderness ratio on the connection ductility is
not studied.
The test results presented in this paper provided evidence that by applying the
above principles, splitting is delayed and ductile failure mechanisms can be
attained. It was demonstrated that:

1. The embedding strength of the timber based on EC5 semi-empirical formulas

was reached.
2. The effective number of fasteners was equal to the actual number of fasteners.
3. The load reached in the tests was always higher than predicted using the
Finnish proposal for block shear failure and according to Eurocode 5, Annex
4. The load carrying capacity increased in respect to the value calculated using
the EC5 formulas, indicating that the embedding strength may not only be
dependent on the timber density, the fastener yield capacity and the diameter
of the fastener but also on the layout of the joint, i.e. fastener spacing end and
edge distances.

The design method of Eurocode 5 - Chapter 8.2-, based on Johansens equations

predicted the load carrying capacity accurately. This proves that applying the
conditions on spacing and end and edge distances already set by Johansen, this
theory is a reliable design tool for the capacity calculation of connections with
dowel fasteners. However, it should be noticed that the values for spacing, end-
and edge distances must be considerably larger than the minimum values
according to Eurocode 5.
Furthermore, it is suggested that the strength class of timber and the timber
thickness should be correlated with the steel grade and the fastener diameter. To
transfer large tensile forces, alternative options like glued-in rods or screws
inserted parallel to the grain like the SFS system should be considered.
A pilot test program was conducted which will have to be extended by
additional tests to draw conclusions which are not only qualitatively but also
quantitatively representative.

Acknowledgements. The authors would like to express their gratitude to Neue Holzbau
AG Lungern, Switzerland who provided the laboratory facilities and the material which
made the tests possible. Special thanks goes to Prof. E. Gehri for the supervision of the
tests, the background knowledge provided, and the assistance in writing this paper. His kind
support and technical advice are very much appreciated.
Alternative Approach to Avoid Brittle Failure in Dowelled Connections 287

Eurocode 5 - Design of timber structures - Part 1-1, General-Common rules and rules for
buildings. EN 1995-1-:2004 (E) (2004)
EN 383 Timber structures - Test methods - Determination of embedment strength and
foundation values for dowel type fasteners (2007)
Hanhijrvi, A., Kevarinmki, A.: Timber failure mechanisms in high-capacity dowelled
connections of timber to steel - Experimental results and design, VTT Publications 677,
Espoo, Technical Research Centre of Finland (2008) ISBN 978-951-38-7090-4
Hanhijrvi, A., Kevarinmki, A., Koski, R.Y.: Block Shear failure at dowelled steel-to
timber connections. In: CIB 39-7-4, Working Commission W18 Timber Structures
Meeting, Florence, Italy (2006)
Johansen, K.W.: Theory of timber connections. International Association of Bridge and
Structural Engineering 9, 249262 (1949)
Johnsson, H.: Plug Shear Failure in nailed timber connections. In: CIB 36-7-2, Working
Commission W18 Timber Structures Meeting, Colorado, United States (2003)
Johnsson, H., Parida, G.: Prediction model for the load-carrying capacity of nailed timber
joints subjected to plug shear. Materials and Structures 46(13) (2013) ISSN: 1359-5997
Jorissen, A.J.M.: Double shear timber connections with dowel type fasteners. Doctoral
Thesis, University of Delft, Delft (1998)
Jorissen, A.J.M., Fragiacomo, M.: General notes on ductility in timber structures.
Engineering Structures 33(11), 29872997 (2011)
Kairi M.: Block Shear failure tests with dowel-type connection in diagonal LVL structure.
In: CIB 37-7-7, Working Commission W18 Timber Structures Meeting, Edinburgh,
Scotland (2004)
Kangas, J.: Design on timber capacity in nailed steel-to-timber joints. In: CIB 31-7-4,
Working Commission W18 Timber Structures Meeting, Savonlinna, Finland (1998)
Kangas, J., Kevarinmki, A.: Modelling of Block tearing failure in nailed steel to timber
joints. In: CIB 30-7-2, Working Commission W18 Timber Structures Meeting,
Vancouver (1997)
Kevarinmki, A.: Design method for timber failure capacity of dowelled and bolted
glulamconnections, Research report No VTTS0704609, Finnish Glulam Association,
Helsinki, Finland (2008)
Meyer, A.: Die Tragfhigkeit von Nagelverbindungen bei statischer Belastung. Holz als
Rohund Werkstoff, 15Jg. Heft 2, pp. 96109 (1957)
Mischler, A.: Bedeutung der Duktilitt fr das Tragverhalten von Stahl-Holz-
Bolzenverbindungen. Dissertation Nr. 12561 ETH Zuerich (1998)
NTC - Norme Tecniche per le Costruzioni. Decreto del Ministero delle Infrastrutture e dei
Trasporti D.M. (2008)
Quenneville, P., Bickerdike, M.: Effective in-row capacity of multiple-fastener connections.
In: CIB 39-7-1, Working Commission W18 Timber Structures Meeting, Florence
Schoenmakers, J.C.M., Jorissen, A.J.M.: Failure mechanisms of dowel-type fastener
connections perpendicular to grain. Engineering Structures 33(11), 30543063 (2011)
Resistance and Failure Modes of Axially Loaded
Groups of Screws

U. Mahlknecht1,*, R. Brandner1,2, A. Ringhofer2, and G. Schickhofer2

Competence Centre holz.bau forschungs gmbh | Inffeldgasse 24, 8010 Graz, Austria
Graz University of Technology, Institute of Timber Engineering and Wood Technology
Inffeldgasse 24, 8010 Graz, Austria

Abstract. Screwed connections provide high resistance in strength and stiffness.

Arranged to a group the screws interact and influence each other in dependency of
their in-between spacings. A test setup was found to investigate (i) the influence of
the spacings in-between the screws, and (ii) the anchoring depths on the failure
modes and resistances of groups of screws. We conducted tests on axially loaded
and under a stress-fiber angle of 90 placed groups of screws in solid timber (ST)
and glued laminated timber (GLT) of Norway spruce. Steel fracture, withdrawal
failure and also block shear failure mode, till now for self-tapping screws not
considered by design codes, were observed. Additionally and based on a simple
mechanic load shearing consideration model for the block shear failure mode was
developed for the investigated axially loaded groups of screws. Verification with
test results confirms congruent but conservative results.

Keywords: groups of screws, failure modes, block shear failure, minimum

spacings, load shearing.

1 Introduction

Latest screw technologies offer multifaceted possibilities to connect timber mem-

bers in load bearing systems. Screw joints allow to submit high loads together
with high stiffness. Joint design depends on the magnitude and direction of the
internal forces, and on the dimension and material of the jointed members. Timber
to timber and recently steel-timber connections with outer steel plates are common
for load transfer by a few dozen or even hundreds of screws. Joints with more than
200 screws have been already used to transfer loads of more than 2 MN. To secure
sufficient reliability of these large joints the need for investigations on groups of
screws focusing on their interaction is obvious. In doing so some failure modes
can be clearly distinguished.

Corresponding author.

S. Aicher et al. (eds.), Materials and Joints in Timber Structures, RILEM Bookseries 9, 289
DOI: 10.1007/978-94-007-7811-5_27, RILEM 2014
290 U. Mahlknecht et al.

1.1 Failure Modes of Groups of Screws, Screwed Joints

Self-tapping screws are optimized for axially loading. The resistance of a screwed
joint may either be limited by the jointed timber members, e.g. by net cross sec-
tion failure, by tension stress perpendicular to the grain or interactions with shear
stresses caused by the local load insertion, by withdrawal, or by the screw itself.
Thus steel fracture, withdrawal and head pull-through failure modes can be distin-
guished. These are well known and explicitly considered by limit state verifica-
tions provided e.g. in EN 1995-1-1 (EC 5). This standard gives also regulations for
groups of screws involving minimum spacings and an effective number of screws
given as nef = n0.9. Of course by testing the occurrence of quasi brittle failure
modes (i) splitting of each row of fasteners along the grain, and (ii) block shear
failure of the whole group was observed, although the regulations concerning the
minimum spacings of EC 5 were met (Plieschounig (2010) and Mahlknecht
(2011)). The interaction of the fasteners in the group to an amount not considered
so far is found to be responsible for these failure modes.
Usually the minimum spacings in-between the fasteners and to the edges of
jointed members are given as multiple of the screw diameter d. According to EC 5
the minimum spacing in-between two screws in grain and perpendicular to grain
are given as a1 = 7 d and a2 = 5 d, respectively, and that in grain to the end grain
and transverse to the edge as a1,CG = 10 d and a2,CG = 4 d, respectively.

1.2 State-of-the-Art Concerning Groups of Fasteners

Kevarinmki (2002), Bejtka and Bla (2002), Bla et al. (2006) and Krenn and
Schickhofer (2009) studied tension joints with inclined screws of combined axial
and lateral loading. In particular Bejtka and Bla (2002) and Bla et al. (2006)
studied tension joints where two timber members got connected by inclined
screws. Firstly by one row of crossed screw pairs (one in compression and the
other in tension), and secondly a row of parallel arranged tension screws
d = 8 mm). Withdrawal or pull-through failures at spacings in-between of
200 mm, without reduction in resistance its compared with one screw pair or
single screw, respectively, was observed (nef = n). Kevarinmki (2002) proposed
for parallel arranged tension screws a1 = 8 d based on similar findings. For high
loaded tensile joints with lateral outer steel plates Krenn and Schickhofer (2009)
proposed to overlap the tips of the screws within the jointed members to avoid
failures in tension perpendicular to grain.
Gehri (2001) investigated the behavior of groups of 4 and 9 threaded rods at a
stress fiber angle of 0 and observed the same failure mode and resistance as in
single rods so far the amount of surrounding timber is comparable.
Splitting failure of groups of fasteners has been reported only on joints with
other fasteners than screws. Ehlbeck et al. (1989) and Schoenmakers (2010) exam-
ined groups of dowel type fasteners loaded perpendicular to the grain, which pro-
voked failure of the timber members in tension perpendicular to the grain.
Resistance and Failure Modes of Axially Loaded Groups of Screws 291

Stahl et al. (2004), Hanhijrvi et al. (2006) and Zarnani and Quenneville (2012)
analysed the block shear behavior of rivet fasteners transmitting a load parallel to
the grain. Hanhijrvi et al. (2006) reviewed the block shear verification methods
given in EC 5 (Annex A) testing groups of dowel fasteners and observed that just
block shear failure and no plug shear of each dowel row occurred. Thus an inter-
action between the dowel rows must be present. Stahl et al. (2004) proposed an
evaluation model based on the capacity of wood failure planes around the group,
in which the plane with the highest resistance governs the joint resistance. Also
the model of Zarnani and Quenneville (2012) is based on the capacity of fracture
planes by considering load shearing between the planes according to their stiff-
ness. This model is further discussed in chapter 3.
In contrast to EC 5 some European Technical Approvals (ETAs) for screws al-
low lower minimum spacings of a1 = 5 d or a2 = 2.5 d given that a1 a2 25 d.
These minimum spacings base on a test procedure of CUAP 06.03/08, which base
on screwing-in tests developed by Bla and Uibel (2009). Of course with this
procedure the splitting along the grain can be examined by screwing-in and not by
loading, which in fact also provokes splitting.
In fact quasi brittle failure modes can be avoided so far adequate minimum
spacings and anchoring depths are provided. Of course Plieschounig (2010) and
Mahlknecht (2011) showed that the spacings given in EC 5 and ETAs are too
small to exclude them. In addition the parameter anchoring depth is until now not
included in any regulation. As it is generally required to design joints tightly ar-
ranged, simply extending the spacings may not be meaningful for all circum-
stance. Thus a verification method is requested to determine the resistance of
groups of screws including block shear.

2 Tests of Groups of Axially Loaded Screws

We investigated the behavior of groups of axially in tension loaded screws at a

stress-fiber angle of 90. To avoid possible homogenisation effects on the with-
drawal behavior of screws in cases where more than one timber lamination is pen-
etrated (e.g. GLT), the first test series were executed in solid timber (ST). As the
dimensions of ST are limited smaller anchoring depths were investigated. In GLT
higher cross sections can be tested. The anchoring depths were defined equivalent
to that of single screws showing (i) clear steel fracture, as well as (ii) clear with-
drawal failure. For classification of failure modes the spacings between the screws
along and perpendicular to the grain were varied.
To account for the tip-effect of screws the proposal of Pirnbacher et al. (2009)
to subtract 1.17 d from the threaded length in the timber to get the effective length
lef was applied. Also the effect of embedding of the thread on the withdrawal ca-
pacity fax reported in Pirnbacher et al. (2009) was considered. Pilot tests in ST
showed an optimal resistance at an embedding depth temb = 4 d.
Limited by the testing device investigations were done with screws of
d = 6 mm, although in timber engineering diameters of 8 to 12 mm are more
292 U. Mahlknecht et al.

common. Groups of m n are defined by m screws along and n screws perpendicu-

lar to the grain.

2.1 Test Procedure and Setup

The screws were inserted through an almost rigid steel plate providing a matrix of
holes for variation of in-between spacings, see Fig. 1. Every screw was pre-
tensioned with a torque wrench and by a twisting moment of 6 Nm. After some
cycles of pre-tensioning a nearly homogeneous load distribution within all screws
and an equal embedding of all threads could be assumed. During loading the spec-
imen in the push-pull loading setups was pressed on a second rigid but fixed steel
plate with a rectangular opening, large enough to allow local deformations of the
timber surface surrounding the edge screws. The length and width of the opening
was adapted for each variation in spacings and group size. The distances between
the supporting edge and the boundaries of the group were chosen equal to the
spacings in timber a1 and a2, see Fig. 1.
Reference tests were done with single screws of the same type, with the same
anchoring depth and in the same timber soureces. Tests on groups in ST were
way-controlled according to EN 1382 using a constant velocity of 4.5 mm/min.
The maximum load Fmax was reached after 180 60 sec. Tests stopped after a
decrease of 30 % from Fmax or a deformation 10 mm. Also the tests with GLT
were way-controlled, but accomplished according to EN 26891, with 1 mm/min
and stopped at F 0.8 Fmax | t.

Fig. 1 Test setup: upper loading plate with hole matrix and the fixed supporting steel plate
with a rectangular opening

2.2 Test Material

ST and GLT of Norway spruce (Picea abies) was used. The cross section of ST
(C24 according to EN 338) was approximately 163 x 252 mm and that of GLT
(GL24h according to EN 1194) 240 x 180 mm. To provide sufficient supporting
area against compression failures perpendicular to the grain during the test, the
specimens were laterally strengthened by edge bonded boards or smaller GLTs. In
particular the ST specimens were strengthened in depth to provide sufficient stiff-
ness against bending deflection. The serial spacings to the end grain and to the
Resistance and Failure Modes of Axially Loaded Groups of Screws 293

edge of the glued specimen were > 18 d = 108 mm. For the tests in ST self-
tapping partially threaded screws 6x300/75 from Schmid Schrauben Hainfeld
GmbH were used and for tests in GLT with lef = 17.8 d and 28.3 d a special screw
with a longer shank and/or a longer threaded length were used.

Table 1 Test setup: group of screws with a stress-fiber angle of 90, d = 6 mm

mean of Fmax (*or F1st if available) [kN]

failure # W / # S
specimen length

12,mean [kg/m]

lef / d (temb / d)
source, height,

and (# failure mode)

mean density

single screw
width [mm]

a1 = 10.0 d

a1 = 12.5 d
a1 = 5.0 d

a1 = 7.5 d

a2 / d

ST 400- 406 5x5 50 /0 3.5 122* (10B) 141* (9B) 157* (10B) 151*
163 600 11.2 (9B)
256 (4)

GLT 350 453 3x4 0/41 2.5 170/182 (8B/2S) 164/183 (4B/5S) 182/184 (1B/4S) -
240 28.3 3.5 184 (5S) 184 (2S) 184 (2S) -
180 5.0 184 (5S) 185 (2S) 184 (5S) -
GLT 350 453 3x5 17/3 3.5 120* (8B) 141* (8B) 138*/165(7B/1W) -
240 17.8 5.0 124/151 134/162/182 135/156 -
180 (1B/4W) (1B/1W-2M/1S) (3B/2W-1M)
7.0 165(3W-3M) 166(4W-2M) 167(5W-1M) -
Wwithdrawal, Bblock shear, Mmixed failure mode, Ssteel fracture

2.3 Analysis of Failure Modes

After testing each specimen was cut along and perpendicular to the grain deter-
mining cracks and their position. With the developed test-setup only block shear,
withdrawal, screw steel fracture and interactions of these failure modes were
Steel Failure Mode (S): Testing GLT of lef = 28.3 d and a2 3.5 d all screws
failed by steel fracture, see Fig. 2 a).
Withdrawal Failure Mode (W): In specimen this was only observed at
a2 = 7 d, see Fig. 2 b) and in some of the specimen with a2 = 5 d.
Block Shear Failure Mode (B): In both, ST and GLT specimen, cracks in
depth at the position of the screw tips limited by the boundary screw rows along
the grain were located. Crack planes appeared also in-between both outer screw
rows along the grain at the same depth, see Fig. 2 c). In some GLT specimens a
crack caused by tension perpendicular to grain at the plane of screw tips and an-
other one between the first and second lamella was recognised. The order of crack
formation is difficult to clarify. Of course a shear failure in timber was not
observed visually.
294 U. Mahlknecht et al.

Mixed Failure Modes (M): Three of the 20 tested single screws in GLT at
lef = 18.3 d failed by steel fracture. Even in the group tests steel failure occurred,
but of course, in none of the tests of the whole group. These tests are further allo-
cated to the mixed failure mode. Local stiffness differences caused by knots and
other growth characteristics are seen as possible reasons.

Fig. 2 Specimens with a) steel b) withdrawal and c) block shear failure mode

2.4 Discussion
The load-deformation curves of the tests done in ST and some of them in GLT
with lef = 18.7 d showed a partial failure followed by load redistribution and partly
by a further load increase, see Fig. 3 a). The partial failure caused a reduction of
the initial stiffness from K1 to K2. A partial failure was classified as a first failure
F1st in cases of F2 / F1 0.96 and K2 / K1 0.9. In some tests an increase in load
was observed although the cross section was severely damaged and of the residual
dimension lef (n 1) a2.
A test series was accomplished by Plieschounig (2010) with the same screw,
lef = 11.3 d, groups of 4, 9, 16 and 25 screws and a1 = a2 = 5 d. Block shear failure
was observed in groups of 9 screws. A second test series done by Plieschounig
(2010) compared the resistance of two partially threaded screws with that of a
single screw. At a1 7 d and a2 3 d no influence was observed. Hereupon we

Fig. 3 a) Typical load-deformation curve b) boxplot F1st, Fmax vs. a1 of groups with
m n = 25
Resistance and Failure Modes of Axially Loaded Groups of Screws 295

did tests of groups in ST with 25 partially threaded screws and an anchoring depth
and embedment length of lef = 11.3 d and temb = 4 d, respectively, whereby
a2 = 3.5 d was kept constant and a1 varied with 5 d, 7.5 d, 10 d and 12.5 d. Test
values of F1st in Fig. 3 b) are shown as grey filled boxes, that for Fmax without
filling. An increase in resistance from a1 = 5 d to 10 d is found, whereby all
specimens failed in block shear.
Fig. 4 shows the results for groups of m n = 12 fully threaded screws, tested
with lef = 28.3 d and lef = 17.8 d, in depending of the area per screw A12. The grey
symbols right of the te