PHD Thesis Jesper Harrild S Rensen Final Reduceret PDF

Downloaded from orbit.dtu.
dk on: Sep 19, 2019
Design and Modeling of Structural Joints in Precast Concrete Structures
Sørensen, Jesper Harrild
Publication date:
2018
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
Citation (APA):
Sørensen, J. H. (2018). Design and Modeling of Structural Joints in Precast Concrete Structures. Technical
University of Denmark, Department of Civil Engineering. B Y G D T U. Rapport, No. R-384
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
 You may not further distribute the material or use it for any profit-making activity or commercial gain
 You may freely distribute the URL identifying the publication in the public portal
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Jesper Harrild Sørensen
In precast reinforced concrete structures, the joints and connections between the
precast components are essential for the structural performance. This thesis pre-
sents a new design for keyed connections typically used between shear wall pa-
nels. An extensive experimental program is presented in combination with sound
design methods developed on the basis of the theory of plasticity. The theoreti-
cal models include first order upper and lower bound modeling of keyed connec-
tions as well as second order modeling of the load-displacement relationship for
casting joints exposed to a simple displacement field. Design and Modeling of Structural
Joints in Precast Concrete Structures

PhD Thesis
Department of Civil Engineering

2018
DTU Civil Engineering R-384
DTU Civil Engineering

Technical University of Denmark
Brovej, Bygning 118

2800 Kongens Lyngby
www.byg.dtu.dk
ISBN 9788778774798
ISSN 1601-2917
Design and Modeling of
Structural Joints in Precast
Concrete Structures
PhD Thesis

July 2017
Supervisors:
Professor Linh Cao Hoang
Associate Professor John Forbes Olesen
Associate Professor Gregor Fischer
Assesment Committee:
Professor Rune Brincker, DTU Civil Engineering, Denmark
Professor Dr.-Ing, Josef Hegger, Institute of Structural Concrete, RWTH
Aachen University, Germany
Senior Scientist, Miguel Fernandez Ruiz, École Polytechnique Fédérale de
Lausanne, Switzerland
Design and Modeling of Structural Joints in Precast Concrete

Structures
Copyright c 2017 by Jesper Harrild Sørensen
Printed by -
ISBN: 9788778774798
Preface
This thesis is submitted as a partial fulfillment of the requirements for ob-
taining the Danish PhD degree. The study has been undertaken at the
Department of Civil Engineering, Technical University of Denmark, under
supervision of Professor, PhD Linh Cao Hoang and co-supervised by As-
sociate Professor, PhD John Forbes Olesen and Associate Professor, PhD
Gregor Fischer.
During the study, a new design of a structural connection for precast
concrete shear walls has been investigated and the results are promising.
The thesis gives an overview of the different load carrying mechanisms that
have been studied during the period of the project. A collection of papers
is appended.

Kgs. Lyngby
July 2017
Preface to Published Version

The thesis was defended at a public defense on Monday the 22nd of Jan-
uary 2018. Subsequently the PhD degree was awarded from the Technical
University of Denmark.
Compared to the version submitted for assessment, a small number of
minor changes have been implemented. The status of Paper I has changed
from accepted to published and Paper III has changed from submitted to
published.

Kgs. Lyngby
January 2018
iii
Acknowledgments
First and foremost, I would like to express my sincere appreciation for the
great support and the always enthusiastic and helpful guidance from my
supervisors, Professor Linh Cao Hoang, Associate Professor John Forbes
Olesen, and Associate Professor Gregor Fischer, all from the Department of
Civil Engineering at the Technical University of Denmark (DTU). Further-
more, I wish to express my gratitude to Professor Aurelio Muttoni, École
Polytechnique Fédérale de Lausanne, EPFL, for a very interesting research
stay with his group in the structural concrete laboratory, IBETON, during
the summer of 2016.
I also wish to acknowledge a number of people who supported the project.
The experimental work has been supported by the Danish Precast Concrete
Association, the COWI foundation, and CRH Concrete A/S. Furthermore
participation in conferences and the realization of the external research stay
have been financially supported by the Otto Mønsted Foundation, Professor
H. I. Hannovers scholarship, and Erik Hegentofts scholarship. The contri-
butions are much valued.
Furthermore, a number of students have assisted in the experimental
campaigns performed within the scope of this thesis. M.Sc. Baldur Halldórs-
son is acknowledged for his contributions to the triaxial tests and for the
support with the single shear key tests. B.Sc. Frederik Jensen and B.Sc.
Frederik Vind are acknowledged for support with triaxial tests. For the tests
on the tensile capacity of the loop connections, M.Sc. Jacob Svejgaard and
Lasse Øvrelid are acknowledged. For tests on the shear capacity of the keyed
shear connections, B.Eng. Rune Pedersen, B.Eng. Mads Herløv, M.Sc. Ja-
cob Svejgaard, M.Sc. Lasse Øvrelid, M.Sc. Jakob Olsen and M.Sc. Rasmus
Hou are acknowledged for their contributions. Finally, for support with the
tests on double-sided dowel joints, M.Sc. Nikolai Bach and M.Sc. Martin
Hansen are acknowledged. Furthermore, the technical staff of the structural
laboratory at the Technical University of Denmark is greatly acknowledged.
Without the support of the above mentioned persons, the planning and ex-
ecution of the experimental investigations would not have been possible. I
sincerely appreciate the contributions.
In addition, colleagues from the Department of Civil Engineering, DTU,
and colleagues from IBETON, are thanked for many interesting conversa-
tions on various topics and for the always encouraging approach to the life
as a PhD student.
v
Abstract
Precast reinforced concrete components are widely used for construction

of buildings in many industrialized countries. The benefits of the precast
method, as compared to the cast in-situ method, lie primarily in the easier
quality and production control of the structural components and in the on-
site construction speed. The challenges appear in the on-site assembly phase,
where structural integrity has to be ensured by in-situ cast connections in
narrow zones. These connections are essential for the overall structural be-
havior and for this reason, strong and ductile connections that at the same
time comply with the construction sequences for the particular structure,
are important for a well-performing solution.
Current best practice for design of shear connections has been developed
over decades and has primarily been aimed at solutions that are easy to im-
plement on the construction site. The related calculation methods are mostly
based on experience and empirical formulas. The strength and ductility of
the current connection design are not necessarily adequate for structures,
where large loads have to be transferred. The potential for improvement of
the structural connections is therefore significant, as better solutions may
enhance the overall structural behavior and lead to more economic designs.
This study concerns an investigation of in-plane connections between
precast shear walls. A new design with ’2-on-2’ loop connections is suggested.
The significance of the new design is the orientation of the U-bar loops
and the use of a double T-headed rebar in the overlapping area of the U-
bars. The investigation covers several independent research topics, which
in combination provides a broad knowledge of the behavior of keyed shear
connections.
As the first topic, the structural behavior of mortar is investigated. This
is relevant as mortar with small aggregates is typically used to grout the
narrow connections between the precast components. The study comprises
triaxial tests and push-off tests aiming to investigate the behavior of mortar
during failure. Next, the tensile capacity of the new loop connection design
is tested and analyzed by use of upper bound plasticity models. This study is
relevant as the ability of the U-bar loops to transfer tension is a prerequisite
for the shear connection to transfer shear loads. The established models,
supported by tests, can be used to design the loop connection in such a way,
that the tensile capacity is governed by yielding of the U-bars and not by
a brittle failure of the grout. This is important in order to obtain a ductile
vii
response when the connection is loaded in shear.
The main focus of the thesis is test and modeling of keyed shear connec-
tions. An extensive experimental program is presented. The particular lay-
out of the test specimens allows for a direct comparison of the conventional
shear connection design with the new design concept. The performance of
the two designs is evaluated and it is found that the new design is superior
in terms of strength and ductility. Upper and lower bound plasticity models
are developed for strength prediction and satisfactory agreements are ob-
tained when comparing the models with the test results. A theoretical exact
solution is not possible to obtain, as the models are based on assumptions
that are not fully identical. However, the establishment of both types of
models provides a range of expected results and thus valuable information
for practical applications.
Finally, second-order plastic modeling is used to establish the load-displa-
cement relationship for a casting joint loaded in shear and transversely rein-
forced with rebars. Despite the simplicity of the model, rather satisfactory
agreement with tests is found. The model may be used to predict the avail-
able plastic energy and has potential for practical assessment of structural
robustness.
viii
Resumé
Præfabrikerede betonelementer benyttes i stor udstrækning til nybyggeri i

mange industrialiserede lande. Fordelene ved at bygge med præfabrikerede
elementer, sammenlignet med traditionel pladsstøbt beton, findes primært
i en lettere kvalitets- og produktionskontrol samt i opførelseshastigheden.
Udfordringerne findes i udførselsfasen, hvor strukturel sammenhæng skal
sikres gennem pladsstøbte samlinger i smalle åbninger. Disse samlinger er
essentielle for den overordnede strukturelle opførsel og af samme årsag, er
samlinger, der er stærke og duktile og som på samme tid passer ind i ud-
førelsesprocesserne på pladsen, vigtige for velfungerende løsninger.
Den nuværende best practice for udførslen af forskydningssamlinger er
blevet udviklet i løbet af årtier og løsningerne har primært været orienteret
mod udførselsfasen på byggepladsen. De tilhørende beregningsmetoder er
i de fleste tilfælde baseret på erfaring og empiri. Styrken og duktiliteten
af de nuværende samlinger er ikke nødvendigvis tilstrækkelig i tilfælde af,
at der skal overføres større laster. Potentialet for forbedringer er dermed
betydeligt, eftersom bedre løsninger kan styrke den overordnede strukturelle
opførsel og i sidste ende føre til mere økonomiske designs.
Dette studie omfatter en undersøgelse af plane samlinger mellem præfab-
rikerede forskydningsvægge. Et nyt design med ’2-på-2’ U-bøjle samlinger
er foreslået. Kendetegnet for det nye design er, at U-bøjlerne er orienteret
anderledes end i det konventionelle design og at der er brugt dobbelt T-
hovedet armering i U-bøjlernes overlappende areal. Undersøgelsen dækker
flere uafhængige forskningstemaer, som i kombination giver en bred viden
om fortandede forskydningssamlingers opførsel.
Som det første undersøges mørtels strukturelle egenskaber. Det er rele-
vant at undersøge, da mørtel med små tilslag typisk benyttes til at udstøbe
de smalle samlinger mellem de præfabrikerede betonelementer. Studiet om-
fatter triaxiale tests og push-off tests, som har til formål at undersøge mør-
tel under brud. Som det næste, testes trækkapaciteten af den nye U-bøjle
samling og bæreevnen analyseres ved brug af plastiske øvreværdi metoder.
Undersøgelsen er relevant, idet det er en forudsætning for forskydningssam-
lingens evne til at overføre forskydning, at U-bøjlerne er i stand til at overføre
træk. De udviklede modeller er understøttet af tests og kan bruges til at
designe U-bøjle samlingerne således, at trækkapaciteten er bestemt af flyd-
ning i U-bøjlerne og ikke af et sprødt brud i udstøbningsmaterialet. At opnå
flydning i U-bøjlerne er vigtigt for at opnå et duktilt respons, når samlingen
ix
belastes i forskydning.
Afhandlingens hovedfokus er test og modellering af fortandede forskyd-
ningssamlinger. Et omfattende eksperimentelt program er præsenteret. Den
pågældende udformning af forsøgselementerne tillader en direkte sammen-
ligning mellem det konventionelle design og det nye design koncept. Testre-
sponset for de to design sammenlignes, og det findes, at det nye design
er bedre med hensyn til styrke og duktilitet. Plastiske øvre- og nedreværdi
modeller udvikles for estimering af bæreevnen og tilfredsstillende overensstem-
melse mellem model og testresultater findes. Det er ikke muligt at udvikle
en teoretisk eksakt løsning, da modellerne er baseret på antagelser, der ikke
er helt identiske. Ikke desto mindre fører etableringen af begge typer af
modeller til en afgrænsning af forventede resultater, og dermed til værdifuld
information for praktisk anvendelse.
Til slut er andenordens modellering benyttet til at etablere last-flytnings-
relationen for en støbt samling armeret på tværs med armeringsstænger og
belastet i forskydning. På trods af modellens simpelhed, findes overraskende
tilfredsstillende overensstemmelse mellem tests og modelrespons. Modellen
kan benyttes til at estimere den tilgængelige plastiske energi og den har
potentiale for praktisk anvendelse i forhold til vurdering af strukturel robus-
thed.
x
Contents
Notation xv
1 Introduction 1
1.1 Precast Concrete Structures . . . . . . . . . . . . . . . . . . 1
1.2 Motivation of Study . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Objectives of Thesis . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Organization of the Thesis . . . . . . . . . . . . . . . . . . . 6
1.5 Contributions to the Field of Research . . . . . . . . . . . . 12
2 Theory of Plasticity 13
2.1 Yield Condition . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Normality Condition . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Extremum Principles . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Failure Criterion for Cementitious Materials . . . . . . . . . 16
3 Tests of Mortar in Triaxial Compression 23

3.1 Mortar versus Concrete . . . . . . . . . . . . . . . . . . . . . 23
3.2 Triaxial Strength of Concrete . . . . . . . . . . . . . . . . . 25
3.3 Experimental Investigation . . . . . . . . . . . . . . . . . . . 29
3.4 Interpretation of Test Results . . . . . . . . . . . . . . . . . 33
3.5 Thin Section of Tested Specimen . . . . . . . . . . . . . . . 37
3.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . 41
4 Failure of a Mortar Shear Key 43

4.1 Experimental Investigation . . . . . . . . . . . . . . . . . . . 44
4.2 Analysis by Digital Image Correlation . . . . . . . . . . . . . 46
5 Tensile Capacity of Loop Connections 55

5.1 Prediction of Capacity . . . . . . . . . . . . . . . . . . . . . 56
5.2 Parametric Study . . . . . . . . . . . . . . . . . . . . . . . . 58
6 Tests of Shear Connections 67

6.1 Experimental Campaign . . . . . . . . . . . . . . . . . . . . 67
6.2 Test Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
xi
7 Upper Bound Solutions for Keyed Shear Connections 89
7.1 Prediction of Capacity . . . . . . . . . . . . . . . . . . . . . 89
7.2 Comparison of Tests with Theory . . . . . . . . . . . . . . . 96
7.3 Evaluation of Model Parameters . . . . . . . . . . . . . . . . 101
7.4 Practical Application . . . . . . . . . . . . . . . . . . . . . . 108
8 Lower Bound Solutions for Keyed Shear Connections 111

8.1 Development of Lower Bound Solutions . . . . . . . . . . . . 111
8.2 Comparison to Finite Element Limit Analysis . . . . . . . . 119
8.3 Comparison of Tests with Theory . . . . . . . . . . . . . . . 121
8.4 Evaluation of Model Parameters . . . . . . . . . . . . . . . . 126
9 Comparison of Upper and Lower Bound Solutions 131

9.1 Expected Range of Results . . . . . . . . . . . . . . . . . . . 131
9.2 Evaluation of the Models . . . . . . . . . . . . . . . . . . . . 136
9.3 Practical Relevance . . . . . . . . . . . . . . . . . . . . . . . 137
10 Dowel and Catenary Action in Rebars Crossing a Shear

Joint 141
10.1 Second-Order Plastic Modeling . . . . . . . . . . . . . . . . 143
10.2 Comparison of Model with Test Results . . . . . . . . . . . . 155
10.3 Considerations for Model Extensions . . . . . . . . . . . . . 160
11 Conclusions 165
Bibliography 169
A Results of Triaxial Tests 177
B Summary of Push-off Tests on Shear Connections 183
C Upper Bound Calculations for Shear Connections 191
D Lower Bound Calculations for Shear Connections 197
xii
Appended Papers 203
Paper I
”Tensile capacity of loop connections grouted with concrete or
mortar”,
Jesper H. Sørensen, Linh C. Hoang, John F. Olsen, Gregor
Fischer.
Published in: Magazine of Concrete Research (2017), 69(17):892-
904 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Paper II
”Test and analysis of a new ductile shear connection design
for RC shear walls”,
Fischer.
Published in: Structural Concrete (2017), 18(1):189-204 . . 221
Paper III
”Test and lower bound modeling of keyed shear connections in
RC shear walls”,
Jesper H. Sørensen, Morten A. Herfelt, Linh C. Hoang, Au-
relio Muttoni.
Published in: Engineering Structures (2018), 155:115-126 . 239
Paper IV
”Testing and modeling dowel and catenary action in rebars
crossing shear joints in RC”,
Fischer.
Published in: Engineering Structures (2017), 145:234-245 . 253
xiii
Additional Work (not appended)
1. Construction-friendly Ductile Shear Joints for Precast Concrete Pan-
els
Jesper H. Sørensen, Linh C. Hoang, Gregor Fischer, John F. Olesen.
Published in: Fernando, D., Teng, J. G., and Torero, J., editors, Pro-
ceedings of the Second International Conference of Performance-based
and Life-cycle in Structural Engineering, pages 640-649, Queensland
University, Brisbane, Australia, 2015
2. Catenary Action in Rebars Crossing a Casting Joint Loaded in Shear

Jesper H. Sørensen, Linh C. Hoang, John F. Olesen, Gregor Fischer.
Published in: Maekawa, K., Kasuga, A., and Yamazaki, J., editors,
Proceedings of the 11th fib International PhD Symposium in Civil En-
gineering, pages 735-742, University of Tokyo, Tokyo, Japan, 2016
3. Tensile Capacity of U-bar Loop Connections with Precast Fiber Rein-

forced Dowels
Jesper H. Sørensen, Linh C. Hoang, John F. Olesen, Gregor Fischer.
Published in: Beuhausen, H., editor, Proceedings of the fib Symposium
2016 - Performance-based approaches for concrete structures, Univer-
sity of Cape Town, Cape Town, South Africa, 2016
xiv
Notation
Abbreviations
DIC Digital Image Correlation
FELA Finite Element Limit Analysis
LVDT Linear Variable Displacement Transducer
w/c Ratio between water and cement in a mixture
List of Symbols
a Distance between innermost U-bars

a Geometric length (Chapter 8)
Ac Circular area of U-bar overlap
Ad Area of diagonal yield line (Chapter 7)
Ai Area of inclined yield line in a shear key (Chapter 7)
Ak Area of one shear key
A l , At Area of tensile failure plane (Chapter 5)
A l , At Stress resultants (Chapter 8)
As Reinforcement area
AsL Reinforcement area of lacer bar (Chapter 5)
AsL Reinforcement area of locking bar (Chapters 6-9)
b Width of connection
c Cohesion
c Strength enhancement factor for concrete (Chapters 8-10)
Cl , Ct Stress resultants (Chapter 8)
d Diameter of rebar
dk Depth of shear key
dmax Maximum grain size
D Amount of energy dissipated (Chapter 2)
D Internal bend diameter of U-bar (Chapters 5-9)
e Optimization parameter (Chapter 8)
e1 , e2 Geometric length (Chapter 8)
f Yield condition
fc Compression strength of concrete
fc,0 Reference strength of concrete
fcc Enhanced average strength of concrete
xv
ft Tensile strength of concrete
fu Ultimate strength of reinforcement
fy Yield strength of reinforcement
fuL Ultimate strength of locking bar
fyL Yield strength of lacer bar (Chapter 5)
fyL Yield strength of locking bar (Chapters 6-9)
hk Height of shear key
H Overlapping length of U-bars
k Material parameter
K Factor influencing ν
l Longitudinal direction
l Material parameter
l1 , l2 Distances describing the position of plastic hinges
L Total length of shear connection
Lk Length of shear key
m Material parameter
M Moment in rebar
Mp Plastic moment capacity
n Number of shear keys
N Tensile load on loop connection (Chapter 5)
N Tension force in rebar (Chapter 10)
N0 Cracking load of loop connection (Chapter 5)
N0 Initial tension force in rebar (Chapter 10)
Np Plastic tension capacity
P Shear load
Pcal Calculated shear capacity (Chapters 7-9)
Pf Shear resistance from friction (Chapter 10)
PFP Recorded first peak load
Pl , P t Stress resultants (Chapter 8)
Ps Shear resistance of a smooth interface (Chapter 10)
Ptotal Total shear resistance (Chapter 10)
PU Recorded ultimate load
qi Generalized strains
Qi Generalized stresses
s, s1 Distance between outermost U-bars
t Transverse direction
t Thickness of precast concrete element
t Time (Chapter 10)
u Shear displacement parallel to interface (Chapter 10)
umax Shear displacement capacity (Chapter 10)
u Vector containing the relative displacements in a yield line
|u| Length of displacement vector
xvi
ua , ub Displacement vector of segment IIa and IIb (Chapter 5)
ul Displacement in the l-direction
ut Displacement in the t-direction
V Volume of body (Chapter 2)
V Shear force (Chapters 4+5)
V Shear force in rebar (Chapter 10)
Vpeak Maximum recorded shear force
W Plastic work
WE External work
WI Internal work
Greek letters
α Angle of displacement vector
β Inclination of yield line to the l-direction
δ Increment of displacement or strain
∆ Increment of plastic deformation (Chapter 2)
∆ Elongation in plastic hinge (Chapter 10)
∆˙ Rate of elongation in plastic hinge (Chapter 10)
∆max Maximum elongation in plastic hinge (Chapter 10)
ε̄ Strain vector
γ Slope of inclined yield line in a shear key
λ Indeterminate positive factor
µ Coefficient of friction
ν Effectiveness factor, concrete in compression
νt Effectiveness factor, concrete in tension
φ Diameter
φL Diameter of lacer bar (Chapter 5)
φL Diameter of locking bar (Chapters 6-9)
Φ Degree of transverse reinforcement
ΦL Reinforcement degree, lacer bar (Chapter 5)
ΦL Reinforcement degree, locking bar (Chapters 6-9)
ϕ Internal angle of friction
σ Normal stress
σ1 , σ2 , σ3 Principal stresses
σA , σB Compressive stress in Struts A and B
σc Concrete stress
σs Reinforcement stress
τ Shear stress
θ Rotation in plastic hinge (Chapter 10)
θ̇ Rate of rotation in plastic hinge (Chapter 10)
θA , θB Inclination of Struts A and B (Chapter 8)
θk Inclination of key corner (Chapter 8)
xvii
1 Introduction
1.1 Precast Concrete Structures

Precast concrete structures represent a branch of structural concrete, where
the structural components are cast at controlled conditions in a factory and
later assembled on-site. The precast method is often used for buildings
or bridges, however, the method is applicable for any kind of structure and
possibly in combination with traditional in-situ cast solutions. The approach
is considered as an economic, environmental friendly, durable, and well-
performing method with a vast potential.
A precast solution is, however, not to be considered as an attempt to
replace an in-situ cast structural design with precast components assembled
on-site to achieve an overall similar structural behavior as the corresponding
in-situ cast solution. Constructing with precast concrete elements is to be
considered as the employment of a system, where the individual elements and
solutions influence the structural performance and hence the overall design.
Elaborated descriptions for precast solutions can be found in various design
guidelines, see e.g. fib Bulletin 74 (fib, 2014).
After the Second World War, the precast construction method gained
popularity as the demand for fast completion of economical buildings in-
creased. Many industrialized countries adopted this construction method
and completed a large number of buildings. Today construction with pre-
cast concrete is (when applicable) the preferred method in Denmark. One
of the reasons is related to a change in the Danish Building Regulation in
1958, requiring residential dwellings to be designed according to a modu-
lar concept. The choice of construction method was voluntary, however,
a market for precast components fulfilling these modular requirements was
created (Kjærbye and Mork, 1998). Investments in the industry combined
with the apparent benefits of the precast technology eventually dominated
the building industry and the tradition of constructing with precast elements
is continued today.
Another reason for the implementation and retention of the precast
method in Denmark is believed to be found in the preferred design methods
used for concrete structures. The Danish school of concrete plasticity counts
a number of excellent engineers contributing to the development of design
methods for concrete structures. The most prominent contributions to be
mentioned count Ingerslev (1921, 1923), Johansen (1931, 1943), and M. P.
Department of Civil Engineering - Technical University of Denmark 1

1.1 Precast Concrete Structures Introduction
Nielsen (summarized in (Nielsen and Hoang, 2011)). Continuous develop-

ment in this field enables the Danish engineers to explore and challenge the
traditional designs of precast concrete structures. A brilliant example of
such an achievement is the design and construction of the Bella Sky Hotel in
Copenhagen, Denmark, where precast technology was used in combination
with in-situ cast solutions to construct a leaning landmark building (Dahl,
2014; Jørgensen, 2015). Another example is the Tietgen Dormitory, a resi-
dence hall in Copenhagen, Denmark. Both buildings can be seen in Figure
1.1.
(a) Bella Sky Hotel, photo: Dahl (2014) (b) Tietgen Dormitory
Figure 1.1: Examples of landmark buildings in Denmark, built using precast

concrete technology
The prefabrication process promotes quality control, precision, and pro-

duction speed as key features. Nevertheless, the overall performance of pre-
cast structures is highly influenced by the performance of the joints between
the precast elements. The capacity of the precast components are easily
assessed by e.g. the theory of plasticity, however, the major challenge for
the overall performance is the capacity of the structural joints. Clearly,
construction-friendly and ductile connections are essential for well perform-
ing and economical structures. An overview of the current best practice,
including design guidelines, for structural connections in precast concrete
buildings is given in the fib Bulletin 43 (fib, 2008). These solutions are based
on decades of construction with precast components, however, the research
and development of sound structural solutions lag behind the development
in e.g. numerical tools and general improvements in the building industry.
This is exemplified in the code provisions (including the European Norm for
2 Department of Civil Engineering - Technical University of Denmark

Introduction 1.2 Motivation of Study
concrete structures (CEN, 2004)), where empirical relations dominate the

design of connections.
1.2 Motivation of Study

For low rise buildings, the strength requirements for the structural connec-
tions are often met with the available design solutions. However, for heavier
loaded elements, such as in high rises or other non-ordinary structures, there
is a potential for development of connections with a higher capacity, which
at the same time are very ductile. This study is motivated by the need from
the industry to have solutions that ease the processes on the construction
site and meanwhile improve the structural performance of the connections.
The improved performance must be reflected in the design methods. In this
relation, it should be noted that changes in design and calculation meth-
ods do not alone improve the structural solutions. On-site inspections often
reveal examples of insufficient execution of the on-site grouting. This com-
promises the structural performance and hence underlines that the need for
construction-friendly designs is accompanied by requirements to the work-
manship on the construction site.
This study is focused on in-plane connections between shear wall panels
and limited to the case of indented (keyed) shear connections. These con-
nections are important, as they provide overall resistance against horizontal
loads to shear walls consisting of precast panels. A sound design of these
connections may provide a strength which is similar to the shear wall panel
itself, however, the connections may preferably be designed with a lower
strength to ensure a well-defined failure mode of the structure. In this case,
a ductile connection is essential for the overall structural safety.
1.2.1 Development of Design

The conventional design of keyed shear connections between precast ele-
ments is constituted by U-bar loops protruding from adjacent elements and
overlapping in a horizontal plane, see Figure 1.2(a). A challenge for this
design arises in case a shear panel has to be lowered immediately next to
another as the U-bars are overlapping in the horizontal plane. This prob-
lem appears e.g. when continuous vertical reinforcement (post-tensioning)
in corrugated tubes are needed to resist overturning moments. A common
way to overcome this challenge is to pre-bend the U-bars and place them
hidden in the shear keys, when the panels are fabricated. This procedure
also protects the reinforcement loops during transportation. On the con-
struction side the U-bars are subsequently straightened to form the overlap,
and a vertical locking bar is installed in the full length of the connection, see

1.2 Motivation of Study Introduction
Horizontal section Vertical section

Mortar
Precast Concrete (grout)
Element
U-bar
Locking bar Locking bar
(a) Conventional design
Pre-bend U-bars
Horizontal section
Element
lowered
Vertical
section
Element
pre-installed
Vertical
locking bar
(b) Assembly phase of conventional design
Figure 1.2: Conventional design of shear wall connection, illustrations from

Sørensen et al. (2017b)
Figure 1.2(b) for procedure during lowering and appearance before grouting.
This procedure imposes a manual workload, which sometimes has to be per-
formed within a very narrow opening, as the appearance of the connection
on the surface of the panel wall is preferably minimized. Often, the assembly
phase is performed inadequately. Poor quality of the grouting process may
lead to a lack of the required structural continuity or to a limited ductility.
The overall performance of the connection is hence not as assumed in the
design.
Another challenge for the designer is the requirement to the reinforcement
used as U-bars. It has to be strong enough to provide the required strength,
on the other hand, the strength cannot exceed the ability of the worker to
straighten the U-bars and the reinforcement has to possess a ductility that
allows for bending and straightening without breaking.

Introduction 1.2 Motivation of Study

Mortar
(grout)
Precast Concrete
Element
U-bar
Lacer bar Locking bar
(a) New design
Horizontal section
Element
lowered
Vertical
section
Element
pre-installed
Vertical
locking bar
(b) Assembly phase of new design
Figure 1.3: New design of shear wall connection, illustrations from Sørensen
et al. (2017b)
These challenges can be overcome by a change of design. Figure 1.3(a)

shows a new design for U-bar loop connections between precast shear wall
panels. The significance of the new design is the orientation of the U-bar
loops. Contrary to the conventional design, the U-bars are placed in the
same plane as the panel, with a mutual spacing. The spacing enables verti-
cal lowering of the panels without problems of rebar clashing or the need for
post-processing of bend U-bars, see Figure 1.3(b). To ensure the required
structural continuity, the U-bar loop connections are reinforced transversely
with double T-headed bars (lacer bars). In addition to the lacer reinforce-
ment, the design allows for installation of a vertical locking bar similar to
the conventional solution. Throughout the thesis the short double T-headed
rebar placed in the overlapping U-bar area will consistently be denoted lacer
reinforcement. The lacer reinforcement can be considered as transverse re-

1.3 Objectives of Thesis Introduction
inforcement or a type of locking bar, however, to distinguish between the

transverse reinforcement of the shear connection in shape of U-bars, the
short T-headed lacer reinforcement placed in the overlapping area of the U-
bars, and the vertical locking bar placed along the entire joint, the notation
(as introduced in Figure 1.3(a)) is consistently adopted. The design concept
can be expanded to include other types of connections, e.g. corner joints or
T-joints, where the loading may consist of a combination of actions. Such
an investigation is beyond the scope of this thesis.
1.3 Objectives of Thesis

The overall objectives of this thesis are to develop a new design for structural
joints in precast concrete, perform experiments to validate the design, and
develop suitable design models for strength prediction. The goal is to obtain
connection solutions that:
• are easy to complete on the construction site

• provide the necessary structural continuity
• ensure a ductile behavior in the ultimate limit state
The objectives are all improvements of current best practice, as the risk of
errors while handling the important structural connections is reduced and
at the same time, an overall more robust structure is obtained.
1.4 Organization of the Thesis

The thesis has been organized to give a thorough description of the structural
performance of the proposed loop connection design for keyed shear wall
connections. The content of the thesis can be seen in Table 1.1. Some of
the main findings have been published in journals, these are appended and
denoted as described in Table 1.1. All details of the development of the
models included in the papers will not be repeated, but a general overview
of the models and the obtained results will be given. Where additional
information is required the reader will be referred to the relevant paper.
Instead, the models will be subjected to a parametric study or compared
to additional test results generated within the study but not yet published.
Section 1.4.2 contains an overview of the conducted experimental programs
and their appearance in relevant papers.
Chapter 2 contains an introduction to the theory of plasticity, including
the application of the theory to reinforced concrete. This chapter serves as
a general basis for the theoretical interpretation adopted in the thesis.

Introduction 1.4 Organization of the Thesis
Table 1.1: Content of the thesis divided by chapter
No. Content Contained in
2 Theory of Plasticity
3 Test of Mortar in Triaxial Compression -
4 Failure of a Mortar Shear Key -
5 Tensile Capacity of Loop Connections Paper I
6 Tests of Shear Connections Paper II+III
7 Upper Bound Solutions for Shear Connections Paper II
8 Lower Bound Solutions for Shear Connections Paper III
9 Comparison of Upper and Lower Bound Models -
10 Dowel and Catenary Action in Rebars Paper IV
Chapter 3 presents experimental results of triaxial tests on mortar and

concrete. The tests establish estimates of the failure envelopes for the mate-
rials, which provide information on the internal angle of friction, ϕ, for the
materials.
Chapter 4 contains a study of the failure process of a mortar shear key.
The chapter presents experimental results and observations from push-off
tests on specimens with a single shear key. The failure process of the key is
studied by use of digital image correlation (DIC).
Chapter 5 deals with the tensile capacity of the new loop connection
design. The study focuses on establishing the amount of lacer reinforcement
required to ensure tensile yielding of the U-bars. Once yielding of the U-
bars is ensured, the optimal shear capacity can be utilized. The theoretical
solutions related to the tensile capacity of the loop connection can be found
in the appended Paper I (Sørensen et al., 2017a).
Chapter 6 presents experimental tests on shear connections and shows
representative results in terms of load-displacement relationships and general
observations. The tests comprise both the conventional design and the new
design. The results form the basis for a comparison between test results and
the theoretical models developed in Chapters 7 and 8.
Chapter 7 introduces upper bound solutions for keyed shear wall con-
nections. The solutions have been developed and presented in the appended
Paper II (Sørensen et al., 2017b). In the chapter, a comparison to the test
results presented in Chapter 6 is performed.
Chapter 8 contains lower bound solutions for the shear capacity of the
keyed connections. The solutions have been developed in the appended

1.4 Organization of the Thesis Introduction
Paper III (Sørensen et al., 2018) and the theoretical results are compared
with the experimental results presented in Chapter 6.
Chapter 9 evaluates the developed upper and lower bound models. The
solutions supplement each other, and together they can be used to bound the
theoretical exact solution from below and above. The comparisons provide
information for practical application and inputs to the calibration of model
parameters.
Chapter 10 introduces second-order plastic modeling for the assessment
of available plastic energy in a shear joint. A model is presented that con-
siders the combination of dowel and catenary action in rebars crossing a
casting joint loaded in shear. Paper IV (Sørensen et al., 2017c) contains the
development of the model, and in the chapter it is compared to test results of
a shear connection without shear keys. In addition, a discussion on possible
model extensions is given.
Chapter 11 summarizes the major findings of the experimental results
and draws conclusions on the theoretical models.
As the overall objective of the thesis is focused on the shear connection
design, the chapters are organized towards the assessment of the shear con-
nections. Table 1.2 shows the main bodies of the content as they appear in
the thesis, guiding the reader from the basic mechanisms over test results to
the modeling aspects.
Table 1.2: Main bodies of content
Chapter 1 Introduction
Chapter 2 Theoretical basis
Chapter 3 Chapter 4 Chapter 5 Material properties +

Triax Key shearing Tension basic mechanisms
Chapter 6 Shear connection tests
Chapter 7 Chapter 8 Chapter 9

1st order modeling
Upper bound Lower bound Comparison
Chapter 10 2nd order modeling
Chapter 11 Conclusions

1.4.1 Reading Guide

In this section a guide is given to the reader on how the information contained
in the appended papers should be related to the chapters of the thesis.
Chapter 1 and 2 serve introductory purposes. Additional information
cannot be found in the papers. The results presented in Chapter 3 and 4
have not been published yet and additional information cannot be found in
the appended papers.
Chapter 5 gives a brief introduction to the results of Paper I. Details on
the experimental program and the development of the models can be found
in Paper I. The chapter contains an additional parametric study on some
of the essential parameters included in the models. The reader may benefit
from reading Paper I before the chapter.
Chapter 6 presents an overview of the experimental work conducted on
shear connections. For specific information on the test procedure the reader
is referred to the experimental sections of Paper II and Paper III. The aim
of the chapter is, besides a presentation of the experimental program, to
distinguish between tests already published and tests not yet published.
Details on the specimens are contained in Appendix B.
Chapter 7 briefly summarizes the findings of Paper II. The reader may
benefit from reading the part of Paper II that describes the model deve-
lopment before reading the chapter. In the chapter the solutions are briefly
introduced and thereafter compared to the results of all the tests presented
in Chapter 6.
Chapter 8 has the same structure as Chapter 7. Paper III is summarized
and compared to the available test results. The reader may benefit from
reading the theoretical part of Paper III before the chapter. Chapter 9
elaborates some of the discussions given in Paper III.
Chapter 10 repeats the theoretical derivation given in Paper IV. For
details on the experimental tests and for additional comparisons of the model
with test results the reader is referred to Paper IV. The reader may read the
paper before the chapter. In the chapter, the model is linked to the shear
connections and a discussion is given on possible model extensions that can
be considered in future works.

1.4 Organization of the Thesis Introduction
1.4.2 Overview of Experimental Work

The theoretical investigation of the new design for loop connections has
been supported by extensive experimental programs. The programs have
been realized with financial support from the collaborators mentioned in
the acknowledgement and with support from a number of student projects
designed and supervised by the author and the supervisors. The involvement
of students in the experimental work made it possible to perform numerous
tests and it also allowed the author to focus on the design of the experimental
programs and interpretation of the obtained results along with the execution.
Table 1.3 contains an overview of the tests included in the thesis, the relation
to the student projects, and, if published, the appearance in the appended
papers. All tests were performed at the Department of Civil Engineering at
the Technical University of Denmark.
Table 1.3: Experiments performed in relation to the PhD study
Type of tests Method No. of tests Project Status
Triaxial tests Oil chamber 40 a Unpublished

Oil chamber 43 b Unpublished
Tension tests Tension 9 c, d Paper I

Tension 23 - Paper I
Key shearing Push-off 20 a Unpublished
Shear connections Push-off 22 e Paper II

Push-off 8 c Paper II
Push-off 10 d Paper II
Push-off 12 f Paper III
Push-off 12 g Unpublished
Dowel action Push-off 42 h Paper IV

The test results of the following student projects have been included in the
thesis (remaining tests were performed by the author):
Master theses:
a Experimental Study of Strength and Fracture of Mortar under Com-

plex Stress States, Baldur Pór Halldórsson, DTU Department of Civil
Engineering, July 2016.
c Test and Analysis of Keyed Shear Joints between Precast Concrete
Walls - Influence of Indent Area on the Load Bearing Capacity, Ja-
cob Svejgaard, DTU Department of Civil Engineering, July 2015
d Test and Analysis of Keyed Shear Joints between Precast Concrete
Walls - Influence of Key Depth on Failure Mode, Lasse Herstad Øvre-
lid, DTU Department of Civil Engineering, July 2015.
f Test and Analysis of Keyed Joints Reinforced with Vertical U-bar Loops
- Influence of Key Depth on the Shear Behavior, Rasmus Malte Hou,
DTU Department of Civil Engineering, July 2016.
g Test and Analysis of Keyed Joints Reinforced with Vertical U-bar Loops
- Influence of Key Length and Grouting Material on the Shear Behav-
ior, Jakob Schmidt Olsen, DTU Department of Civil Engineering, July
2016.
h Dowel- and Catenary Action in Rebars Embedded in Concrete and
Crossing an Interface Loaded in Shear, Nikolai Schjøtt Bach and Mar-
tin Hansen, DTU Department of Civil Engineering, February 2016.
Bachelor theses:
b Failure Criteria for Concrete and Mortar Subjected to Triaxial Stress

States, Frederik Jensen and Frederik Vind, DTU Department of Civil
Engineering, June 2017.
e Forskydningsbæreevnen af montagevenlige elementsamlinger (English:
Shear Capacity of Construction-friendly Joints for Precast Concrete
Structures), Rune Pedersen and Mads Herløv, DTU Department of
Civil Engineering, January 2015.

1.5 Contributions to the Field of Research Introduction
1.5 Contributions to the Field of Research

The theoretical models developed in this study apply to structural connec-
tions for precast concrete shear walls. However, the theoretical framework is
applicable for a number of other structural solutions, if modified accordingly.
The novel contributions to the field of research can be summarized as:
? Experimental verification of the internal angle of friction for mortar

(ϕ = 30◦ ).
? Development and experimental verification of upper bound plasticity

models for the tensile capacity of loop connections (based on a ’2-on-2’
loop configuration).
? Experimental demonstration of ductile test responses for a new keyed

shear connection design. The significance of the new design is that
the U-bar loops are placed in the same plane as the shear panel and
that a double T-headed rebar is used as lacer reinforcement in the
overlapping U-bar loop area.
? Development of upper bound solutions for keyed shear connections,

accounting for a failure of the shear key by local key corner crushing
or complete key cut off.
? Development of analytical lower bound solutions considering struts

with different inclinations and stress transfer by friction in the indented
interfaces.
? Development of a second-order plasticity model for estimation of the

in-elastic load-displacement relationship for a casting joint crossed by
a rebar and loaded in shear. The model establishes a unique link
between the imposed shear displacements and the sectional forces in
the rebar by use of the kinematical conditions for the mechanism and
the normality condition of plastic theory.

2 Theory of Plasticity
A material experiencing zero strains at stresses below the yield limit and
arbitrary large plastic strains for stresses at the yield limit is defined as a
perfectly rigid-plastic material. Such a material does not exist in reality,
however, it is a convenient assumption for material behavior, when dealing
with determination of the load carrying capacity of structures. The theory
related to such an idealized material behavior is applicable, when the plastic
strain capacity of the material is much larger than the elastic strain domain.
The assessment of structural capacity based on this theory is called limit
analysis and it is adopted in the modeling throughout the thesis. In the
following a brief description of the theory will be given as a theoretical basis
for the work presented in the remaining chapters. For further description,
see e.g. Nielsen and Hoang (2011).
The general theory of plasticity was first developed by Gvozdev in Russia
and before it was known to the western world in the late 1950s (Gvozdev,
1960), a similar theory was developed by the Prager school at Brown Uni-
versity, see i.a. Drucker et al. (1952). Important aspects of the theory
include the concepts of yield condition, normality condition, and extremum
principles. These matters define the theory of plasticity and despite the
rough assumptions for the material behavior, the theory is elegant and its
simplicity provides a powerful tool for structural analysis.
2.1 Yield Condition

The yield condition describes the combinations of generalized stresses, Qi ,
which cause yielding. The yield condition is assumed to be convex and
enclosing the point of no stresses, i.e. Qi = 0. Following the notation of
Nielsen and Hoang (2011), the yield condition can be defined as:
f (Q1 , Q2 , . . . , Qn ) = 0 (2.1)
According to the perfectly plastic material behavior, combinations of

stresses fulfilling f < 0 will cause no strain in the material (and thereby no
failure) and combinations fulfilling f = 0 may cause arbitrary large strains.
Combinations of stresses resulting in f > 0 are not possible. The corre-
sponding generalized strain to a generalized stress, Qi , is denoted qi . In
a three dimensional continuum, the generalized stresses consist of the nine

2.2 Normality Condition Theory of Plasticity
stress components and the generalized strains are the corresponding nine
strain components. The generalized stresses and strains can also be chosen
as sectional forces and corresponding sectional deformation quantities. An
example of the latter is given in Paper IV, also treated in Chapter 10.
2.2 Normality Condition

The product of the generalized stresses and the corresponding generalized
strains defines the work per characteristic dimension. For an increment in
the plastic deformation, ∆qi , the amount of energy dissipated over a certain
volume, D, then becomes:
Z Z
D= (Q1 ∆q1 + ...) dV = W dV (2.2)
V V
where W denotes the plastic work per unit volume.

At yield, von Mises hypothesis of maximum work (Mises, 1928) states
that the stresses corresponding to a given strain field attain such values that
the plastic work, W , becomes as large as possible. When the yield condition
is convex and differentiable (without plane surfaces or apexes), the condition
is fulfilled when the strain vector satisfies the following:
∂f
qi = λ , i = 1, 2, ..., n (2.3)
∂Qi
where λ is an indeterminate positive factor. Hence, the magnitude of the
strains is unknown, only the strain rate (and thereby the direction) is known.
Fulfilling the condition, the strain vector becomes an outward directed nor-
mal to the yield surface. Equation (2.3) is usually referred to as the nor-
mality condition of plastic theory, however, it is also named ’the associated
flow rule’ or ’von Mises flow rule’.
In case the yield condition contains a straight part, the strain vector will
be normal to this line/plane all along the straight parts and hence no unique
relation between the stresses and the strains can be determined. However,
W is uniquely determined from q̄ = (q1 , q2 , ..., qn ). If the yield condition
contains an apex, the strain vector, q̄, may attain any angle between the
normals to the two adjoining parts of the yield condition, see Figure 2.1 for
a graphical representation of a two dimensional yield condition that contains
both a plane part, apexes, and smooth differentiable parts.
2.3 Extremum Principles

The capacity of a structure (or the limit load, where unlimited deformations
are possible), can be determined by use of some very useful theorems of

Theory of Plasticity 2.3 Extremum Principles
Q2 , q2
q̄
q̄
Q1 , q1
q̄
Figure 2.1: Normality condition for a two dimensional yield condition
the theory of plasticity. The theorems were formulated independently by

Gvozdev (1960) and Drucker et al. (1952). Assuming that the yield condition
is convex and that the normality condition is fulfilled, some useful concepts
are introduced (quoted from Nielsen et al. (1978)):
1. A statically admissible stress distribution is a distribution of stresses

which satisfies the equilibrium equations and the statical boundary
conditions.
2. A safe stress distribution is a distribution of stresses corresponding to
a point within or at the yield surface, f ≤ 0.
3. A failure mechanism is a displacement field, compatible with the geo-
metrical boundary conditions.
4. The load-carrying capacity or yield load of a structure of rigid-plastic
materials is the lowest load by which deformations are possible.
Based on the first two concepts, the lower bound theorem can be established:
When a safe and statically admissible stress distribution can be

found for a given load, the load is less than or equal to the yield
load of the structure.
From the two remaining concepts, the upper bound theorem can be estab-
lished:
The load required to form a postulated failure mechanism will be

larger than or equal to the yield load of the structure.

2.4 Failure Criterion for Cementitious Materials Theory of Plasticity
From the extremum principles the capacity of a structure can be assessed

either by lower bound solutions, e.g. analysis of stress fields, or by upper
bound solutions, e.g. established by use of the work equation, where the
rate of work performed by the external loads equals the rate of internal
dissipation. If the loads found from the two theorems are equal, the exact
yield load of the structure is found. This is called the uniqueness theorem.
It was noted by Jensen (1976) that the work equation is not identical to the
principle of virtual work, as the stresses belonging to a chosen deformation
are not required to fulfill the equilibrium equations. However, in order to
fulfill the uniqueness theorem, the equilibrium conditions must be fulfilled.
2.3.1 Capacity beyond Yield Load

Despite the fact that the yield load is defined as the load where deforma-
tion occurs, corresponding to a stress state on the yield surface and thereby
failure of the structure, a material or a structure may possess capacity be-
yond the yield load. This applies to e.g. work hardening materials where
additional capacity can be obtained with deformations beyond yield (see
e.g. Drucker (1950) for stress-strain relationships for work hardening ma-
terials) or structures with boundary conditions that activate constrains as
deformation occurs. Membrane action in slabs is an example of the latter
phenomenon. Analysis of problems where change of structural geometry is
important requires second-order modeling and is not covered by the sim-
ple first-order theory. Paper IV contains an example of such a second-order
analysis where the change of structural geometry is taken into account in the
modeling of a load-displacement relationship for a rebar crossing a casting
joint exposed to shear displacements.
2.4 Failure Criterion for Cementitious

Materials
Structural steel is an example of a material that is close to fulfill the as-
sumption of perfectly plastic behavior whereas structural concrete is a quasi-
brittle material more than a plastic material (Karihaloo, 1995). However,
when adopting appropriate reduction factors (see Section 2.4.1), structural
concrete can be treated by the theory of plasticity. A precise and complete
yield criterion for concrete and similar cementitious materials has not yet
been formulated. Instead, Coulombs frictional hypothesis (Coulomb, 1776)
has generally been considered as an acceptable failure criterion for concrete.
The hypothesis assumes a sliding failure in the material along a failure sur-
face. The condition is described in terms of shear stress, τ , and normal

Theory of Plasticity 2.4 Failure Criterion for Cementitious Materials
fc ft
c
ϕ
σ3 σ2 σ1 ϕ
c σ
Separation failure
Sliding failure τ
Figure 2.2: Modified Coulomb frictional failure criterion
stress, σ:
|τ | = c − µσ (2.4)
where c is the cohesion and µ is the coefficient of friction. The internal
angle of friction, ϕ, relates to the coefficient of friction by: tan ϕ = µ.
Introduction of a limitation on the tensile stress leads to the definition of a
modified Coulomb material. The related failure mode is a separation failure,
described by the tensile strength of the concrete:
σ = ft (2.5)
The modified failure criterion can be seen in Figure 2.2, where Mohr’s
circle is illustrated as well. The failure of a modified Coulomb material can
hence be assessed by the principal stresses, σ3 ≤ σ2 ≤ σ1 , tension taken as
positive. It should be noted that the magnitude of the intermediate principal
stress does not influence the failure of the material.
In terms of principal stresses, the failure criterion for a modified Coulomb
material can be formulated in the following way:
σ1 − σ 3 σ1 + σ3
= c cos ϕ − sin ϕ (2.6)
2 2
σ1 = f t (2.7)
Introducing the parameter k defined as:
1 + sin ϕ
k= (2.8)
1 − sin ϕ

the criterion for sliding failure can then be reduced to:

√
kσ1 − σ3 = 2c k (2.9)
As both c and ϕ are regarded as material constants, it is seen from

Equation (2.9) that the smallest principal stress can be decreased when the
largest principal stress is decreased as well. Such a situation is achieved when
a material is subjected to e.g. triaxial compression (this will be discussed in
Chapter 3). When considering a uniaxial stress state, i.e. σ1 = σ2 = 0 and
σ3 = −fc , the following relation is found:
√
fc = 2c k (2.10)
2.4.1 Structural Concrete

Unlike reinforcement steel, where the material behavior with good accuracy
can be simplified as rigid-plastic, concrete experiences softening after reach-
ing the uniaxial strength. Due to this behavior, an appropriate reduced
strength has to be introduced in order to apply plastic theory to structural
concrete (and other cementitious materials). Furthermore, structural con-
crete as it appears in structures is rarely loaded in a similar way as standard
tests used for determination of the uniaxial strength. An effective plastic
strength of concrete in compression for plastic analysis can be defined by
introduction of an effectiveness factor, ν:
fc,ef = νfc (2.11)
In this context it is important to note that the effectiveness factor is in-

troduced to account for softening effects, size effects, and loading conditions.
Different values of the factor may therefore apply for different problems. Re-
garding the tensile strength of concrete, a similar reduction factor has to be
introduced:
ft,ef = νt ft (2.12)
where ft is the uniaxial tensile strength of concrete. The concept of effective
strengths can be illustrated as shown in Figure 2.3. Throughout the thesis,
the compressive strength of the cementitious materials considered has been
estimated from uniaxial test on φ100x200 mm cylinders, when these have
been available, and the tensile strength, when needed, has been calculated
by the following relation:
ft = 0.26fc2/3 (2.13)
In this context is it relevant to note, that the tensile strength of concrete
normally is neglected in plastic analysis of reinforced concrete structures.

−fc
−νfc
ε
νt ft
ft
σ
Figure 2.3: Real and idealized stress-strain relationship for structural con-
crete, including indication of effective strengths for plastic analysis
This is primarily due to the fact, that concrete loaded in tension usually
cracks before the reinforcement is activated and when the reinforcement is
yielding the tensile strength of the concrete has vanished.
2.4.2 Dissipation Formulas for Concrete

In upper bound analysis, the capacity is assessed by use of the work equa-
tion for a given failure mechanism. This requires calculation of the internal
dissipation along yield lines or yield planes. Expressions for the dissipation
along lines of discontinuity in concrete were developed by Jensen (1976) and
can generally be expressed as:
1 sin α − sin ϕ
WA = νfc (1 − sin α) |u| + ft |u|, ϕ ≤ α ≤ π − ϕ (2.14)
2 1 − sin ϕ
where WA denotes the dissipation per unit area of a failure plane, |u| ex-
presses the relative displacement between the two parts separated by the
yield line, and α describes the angle between the displacement vector, u,
and the yield line. Introducing the following parameters:
νt ft sin ϕ
l =1−2 (2.15)
νfc 1 − sin ϕ
νt f t 1
m=1−2 (2.16)
νfc 1 − sin ϕ

σ2
Plane strain B A
C ft
σ1
D
B fc
Plane stress
Figure 2.4: Failure criteria for plane strain and plane stress conditions
the dissipation can be expressed by the more simple expression:

1
WA = νfc (l − m sin α) |u|, ϕ≤α≤π−ϕ (2.17)
2
It should be noted that when the tensile strength of concrete is disre-
garded, the parameters l and m reduce to unity and Equation (2.17) sim-
plifies slightly. The restriction on the angle of displacement relates to the
normality condition (Equation (2.3)). The restriction is thereby dictated by
the yield condition adopted. Figure 2.4 shows the failure criteria for a modi-
fied Coulomb material in plane strain and plane stress conditions, expressed
in principal stresses. A sliding failure corresponds to the regime between B
and D.
For plain strain conditions, the normality condition dictates that the
angle of displacement for all stress states has to be larger than or equal to
ϕ. In the apexes, i.e. σi = fc − kft , i = 1, 2 (Point B), the angle may attain
values between ϕ and π/2, and for smaller principal stresses the direction of
the strain vector has to be normal to the yield condition. The only exception
is the stress state σ1 = σ2 = ft (Point A). For plane stress conditions, the
same restriction does not apply as α may attain other values in the apexes
of σi = fc , i = 1, 2 (Point C). See Nielsen and Hoang (2011) for an in-depth
description. The dissipation formulas are summarized in Table 2.1 for plane
strain and plane stress, respectively. These will be used in the following
chapters, when upper bound models are developed.

Table 2.1: Dissipation formulas for plane strain and plane stress conditions
for a modified Coulomb material
Plane strain Plane stress
ft 6= 0
1
0≤α<ϕ Not possible WA = νfc (1 − sin α) |u|
2
1 1
ϕ≤α≤π−ϕ WA = νfc (l − m sin α) |u| WA = νfc (l − m sin α) |u|
2 2
1
π−ϕ<α≤π Not possible WA = νfc (1 − sin α) |u|
2
ft = 0
1
0≤α<ϕ Not possible WA = νfc (1 − sin α) |u|
2
1 1
ϕ≤α≤π−ϕ WA = νfc (1 − sin α) |u| WA = νfc (1 − sin α) |u|
2 2
1
π−ϕ<α≤π Not possible WA = νfc (1 − sin α) |u|
2

3 Tests of Mortar in Triaxial
Compression
This chapter presents experimental results of triaxial tests on mortar and
concrete with the aim of establishing the failure envelopes for the materials.
Connections in buildings made of precast concrete elements are often grouted
with mortar (opposed to e.g. bridge structures where concrete grout is
typically used). This is because mortar contains smaller aggregates, which
makes it easier to fill the narrow construction zone. The main motivation for
performing the tests was to obtain an estimate of the material parameters
of normal strength mortar to be applied in plastic modeling. Low strength
and high strength mortars are not considered.
3.1 Mortar versus Concrete

Mortar and Concrete are both cementitious materials that consist of the
same ingredients, namely cement, water, and aggregates. However, the prop-
erties of the mixtures may deviate despite similarities in e.g. compressive
strengths. Figure 3.1 shows the constituents of a concrete mixture. It can
Sand < 4 mm
Aggregate > 4 mm
Cement paste
Figure 3.1: Constituents of a concrete, picture from St John et al. (1998)

3.1 Mortar versus Concrete Mortar in Triaxial Compression
be seen that the concrete consists of a combination of smaller aggregates

(smaller than 4 mm, denoted sand), larger aggregates, and a cement paste
that glues the constituents together. The composition of the aggregates
may vary depending on application. In this study, the distinction between
mortar and concrete is defined by the maximum grain size of the mixture.
Mixtures with a maximum grain size of 4 mm or lower are considered as
mortars whereas mixtures with larger maximum grain sizes are considered
as concretes, see Table 3.1. Inclusion of admixtures may change the prop-
erties of the mixtures, however, the categories remain based on the same
definition. Mixtures of pure cement paste are not considered in this study.
Table 3.1: Ingredients in cement paste, mortar, and concrete mixtures
Ingredient Mixture
Cement
Paste
Mortar
Concrete
Water
Sand 0-4 mm
Aggregates > 4 mm
3.1.1 Influence of Maximum Grain Size

The most important factor influencing the compressive strength of a concrete
mixture is the water to cement ratio (normally measured by weight). A
decrease in ratio increases the strength. This relation was described early
in the 20th century e.g. by Abrams (1919) as a ’water-cement ratio law’.
For a given cement content, the grain size distribution affects the amount
of water required to obtain a workable mixture. Smaller aggregates require
more water to make the mixture workable and hence, a strength increase is
obtained by the use of larger aggregates.
The change in water consumption may be significant when adding larger
aggregates to the mixture, thus design of strong and workable concrete by use
of the water-cement ratio law includes the use of larger aggregates. However,
the increase is not infinite. Walker and Bloem (1960) showed that the use
of aggregates with a maximum size larger than 3/4 inches (19 mm) does not
lead to an increased compression strength. Nevertheless, when comparing
mortar to concrete, the larger maximum grain size of the concrete is expected
to lead to a larger strength of the hardened mixture (assuming a proper
proportioning of the aggregates).

Mortar in Triaxial Compression 3.2 Triaxial Strength of Concrete
3.1.2 Influence of Aggregate Content

As shown in Table 3.1, a concrete consists of two main constituents, aggre-
gates and cement paste. For normal strength concretes, the aggregates are
considered to be stronger than the cement paste. After initial failure of the
cement paste, the aggregates will remain intact and possess the ability to
transfer shear stresses by aggregate interlocking. If the grains are small the
ability to transfer stresses will be reduced compared to a mixture with larger
aggregates. However, it is not only the maximum grain size that influences
this ability. The aggregate content of the mixture (often measured as a
volume percentage) also has an influence. Tests reported in the literature
show that a lower aggregate content leads to a lower strength increase, when
tested in triaxial compression.
Zhang (1997b) developed a micromechanical model that takes the com-
position of the concrete mixture into account. The model assumes that the
aggregates displace the path of the failure through the paste, and hence addi-
tion of aggregates displaces the path further, resulting in a higher strength of
the confined concrete. Fewer aggregates (that are considered strong enough
to resist sliding failure) may result in a decreased strength, as failure occurs
in the cement paste. The model also considers a ratio of effective aggregates
in order to take into account the number of aggregates that are weaker than
the cement paste. Naturally, the effective aggregate ratio is decreased as the
strength of the cement paste is increased. Examples of cracked aggregates
are given in Section 3.5.
3.2 Triaxial Strength of Concrete

The strength of concrete in triaxial compression is higher than the uniaxial
strength. Throughout the last century several researchers have performed
tests confirming this relation. Among many experimental investigations, the
tests of Richart et al. (1928), Balmer (1952), Hobbs (1971), Dahl (1992b),
Imran and Pantazopoulou (1996), and Malecot et al. (2010) can be men-
tioned. From the experimental results, the combination of principal stresses
that lead to failure can be used to establish the failure envelope of the mate-
rial tested. Figure 3.2 shows e.g. how Mohr’s circle for three sets of principal
stresses (a, b, c) relate to the modified Coulomb failure criterion in a τ − σ
coordinate system.
The Coulomb criterion is characterized by a linear relation for increasing
confining pressures. However, the actual failure envelope of a concrete is
non-linear and the shape depends on a number of factors. The shape can
be approximated by a modification to the Coulomb criterion. Dahl (1992b)
suggested the following two-stage failure criterion (only compressive stresses

3.2 Triaxial Strength of Concrete Mortar in Triaxial Compression
fc ft
σ3,c σ3,b σ3,a ϕ

σ1,c σ1,b σ1,a ϕ σ
Figure 3.2: Combinations of principal stresses fulfilling the Coulomb failure

criterion
considered):

4σ − f
1 c
σ3 = max (3.1)
3σ1 − 1.5fc
The criterion can be seen in Figure 3.3. This relation captures the test
results of normal and high strength concrete better, however, it still consti-
−σ3
fc
8 σ3 = 4σ1 − fc
7 4
3
6
5 σ3 = 3σ1 − 1.5fc
4
3
2
1
−σ1
0.5 1.0 1.5 2.0 fc
Figure 3.3: Modified two-stage failure criterion, in terms of principal

stresses, suggested by Dahl (1992b)

Mortar in Triaxial Compression 3.2 Triaxial Strength of Concrete
tutes a simple approximation as the actual failure envelope is smooth. The

simple appearance makes it a suitable choice compared to more complex
models, see Section 3.2.1. Similar bilinear suggestions were made by Nielsen
(1998) for high strength concrete. The criteria are expressed in principal
stresses, which is preferred compared to a criterion expressed in terms of
stress invariants or octahedral stresses (Smee, 1967). Smee describes that
the stress invariants are insensitive to changes in the major principal stress
and hence, the criteria may not distinguish adequately between safe and
unsafe domains.
For mortar, the failure envelope contains the same characteristics as the
failure envelope of a concrete. However, the difference in maximum grain
size and aggregate content of the mixture influence the shape. Smee (1967)
established that an increase in maximum grain size increases the strength
of the mixture, when loaded in triaxial compression. Figure 3.4 shows fail-
ure envelopes as a function of both maximum grain size and the uniaxial
compression strength of the mixture. The investigation included a mix-
ture with sand as the only aggregates. In the study, sand was classified as
passing No. 4 sieve, corresponding to a maximum grain size of 4.75 mm.
For a comparison to the current study, the mixture qualifies as a mortar.
When comparing the estimated failure envelopes of concrete and mortar it
is seen that the strength increase is less for a mortar and that the difference
magnifies with increasing normal stress (σ). Besides an influence of the max-
imum grain size, Smee also found that the uniaxial compression strength of
Figure 3.4: Failure envelopes established by Smee (1967)

3.2 Triaxial Strength of Concrete Mortar in Triaxial Compression
the mixture influences the triaxial strength, where an increase in uniaxial

strength leads to a decrease in confined strength.
Despite the general knowledge of the failure envelope of mortar, studies
of the value of the internal angle of friction for mortar seem scarce in the
literature. For normal strength concrete, a value of k = 4, corresponding
to ϕ = 37◦ , is commonly adopted, see e.g. Richart et al. (1928), Johansen
(1958), Chen (2007), and Nielsen and Hoang (2011). However, a similar
commonly acknowledged value for mortar has to the best knowledge of the
author not been established. Nielsen (1998) indicated that a value of ϕ = 30◦
can be adopted for high strength paste and mortar. Naturally, the value
depends on a number of factors, some of them are addressed in the above.
In order to establish the shape of the failure envelope for well-defined mortar
mixtures, a number of triaxial tests were performed.
3.2.1 Alternative Failure Criteria

A number of attempts have been made to describe the non-linear failure
envelope of concrete subjected to triaxial compression. Chen (2007) gave
an overview of the most common failure criteria proposed, classified by the
number of parameters included in the models. Examples of one-parameter
models are the Rankine maximum tensile stress criterion or the Tresca max-
imum shear stress criterion. The Coulomb criterion is an example of a
two-parameter model. Drucker and Prager (1952) introduced a smooth ap-
proximation to the modified Coulomb failure surface as another example
of a two-parameter model. This modification is useful e.g. for numerical
applications.
Examples of three-parameter models are the Bresler-Pisner criterion and
the Willam-Warnke criterion (Willam and Warnke, 1975). The Willam-
Warnke criterion can be extended to a five parameter model. A more in-
teresting and generally acknowledged failure criterion is the four-parameter
model developed by Ottosen (1977). This model has been adopted in the fib
Model Code (fib, 2013) for assessment of concrete subjected to multiaxially
stress states. In addition to the mentioned models a number of empirical
relations have been established based on curve fitting with experimental
results.

Mortar in Triaxial Compression 3.3 Experimental Investigation
3.3 Experimental Investigation

The experimental program was conducted in two parts, a and b. The primary
aim of the tests was to investigate the influence of the maximum grain size
on the internal angle of friction. As a consequence, the volume content of
aggregates in the mixtures was also varied.
The tests were performed on φ100x200 mm cylinders in a triaxial cell,
where oil pressure was applied as confining pressure. This entails that two
of the principal stresses were equal during the test, i.e. σ1 = σ2 . Hilsdorf
et al. (1973) made a review of experimental procedures for multiaxial testing
of concrete and concluded that this method is sufficient to obtain reliable
results.
Figure 3.5: Test setup used by Dahl, illustration from (Dahl, 1992c)
The testing equipment used in the campaign was identical to the equip-
ment used by Dahl (1992b), Hansen (1995), and Nielsen (1998). The cross
section of the triaxial cell is depicted in Figure 3.5 and the reader is referred
to Dahl (1992c) for a detailed description of the test setup and the test
procedure. It is emphasized that the load applied was first a hydrostatic
pressure until a predefined confinement level (the design of the cell ensured

3.3 Experimental Investigation Mortar in Triaxial Compression
a hydrostatic stress state when applying oil pressure in the chamber) before
imposing additional axial compressive load until failure of the specimen.
Failure was defined as the point where axial deformation continued without
a further increase of the load. For each mixture, only one test was performed
at the same confinement pressure. Nevertheless, the expected smooth fail-
ure envelope was captured well, which is explained by the controlled testing
conditions, where failure was observed gradually and not instantly.
3.3.1 Mixture Proportions

The experimental program included seven mixtures with varying maximum
grain sizes and volume percentages of aggregates. The mixture proportions
can be seen in Table 3.2, where the mixtures are identified by a C denoting
a cementitious material, a number referring to the maximum grain size,
dmax , of the mixture, and an adjoining letter referring to the two parts of
the campaign. Ordinary Portland cement was used in all mixtures. The
densities used when calculating the volume percentages (vol%) were 3100
kg/m3 for the cement, 2610 kg/m3 for the aggregates, and 1000 kg/m3 for
the water. All specimens within a mixture were cast from the same batch.
Table 3.2: Mixtures used in the triaxial test program, values given in kg/m3
Project a Project b
Mixture C2a C4a C8a C16a C4b C8b C16b
Cement 451.9 470.6 404.5 283.1 313.6 317.5 243.2

Water 317.1 289.0 255.5 198.6 220.0 222.8 170.7
Aggr. 0-2 1401.8 0 0 0 0 0 0
Aggr. 0-4 0 1459.6 1256.1 878.0 1765.2 1378.6 927.5
Aggr. 4-8 0 0 346.4 242.3 0 380.6 256.0
Aggr. 8-16 0 0 0 733.0 0 0 774.7
Plasticizer 0 0 0 0 2.50 0.60 0.60
Unit weight 2170.8 2219.1 2262.5 2334.9 2301.3 2300.2 2372.7

w/c ratio 0.70 0.61 0.63 0.70 0.70 0.70 0.70
vol% aggr. 0.55 0.56 0.61 0.71 0.68 0.67 0.75
For a mortar, a smaller volume percentage of aggregates is normally

required to obtain a workable mixture because the smaller aggregates require
more water. A method to obtain workable mixtures with a large volume

Mortar in Triaxial Compression 3.3 Experimental Investigation
content of aggregates is addition of plasticizers. This has been adopted in

the b mixtures to obtain mortar mixtures with a volume content of aggregates
that are comparable to the concrete mixtures.
3.3.2 Specimen Behavior During Test

The compression failure of a concrete is associated with volume expansion.
In the case of triaxially compressed cylinders, the extent of volume expansion
at failure depends on the confining pressure applied. For increasing confine-
ment pressures, the deformation capacity of the material is increased. For a
triaxial test performed in a triaxial cell, the cylinder is naturally compressed
in the axial direction with corresponding expansion in the lateral direction.
The resulting shape of the specimen is a barrel shape, see Figure 3.6, which
shows the specimens of Mixtures C4a and C8a after removal from the triax-
ial cell. The specimens are ordered from left to right by increasing confining
pressure applied during the triaxial test.
(a) Mixture C4a
(b) Mixture C8a
Figure 3.6: Appearance of specimens after triaxial testing
The tendency of increased deformation capacity with increased confining

pressure is clearly observed. In the high confinement region, a test can be
continued after reaching the peak strength of the cylinder until the defor-
mation capacity of the testing apparatus is reached. Figure 3.7 shows an
example of such a specimen that was tested with a confinement pressure of
2.15fc . The initial height of 200 mm was reduced to around 137 mm after
testing. It is interesting to note that the cylinder was not broken into pieces
despite the extensive damage of the matrix, visible on the surface. It is also
observed, that the permanent deformations on the surface seem to originate
from an hour-glass shaped failure pattern.

3.3 Experimental Investigation Mortar in Triaxial Compression
Figure 3.7: Example of highly deformed specimen from Mixture C2a

Mortar in Triaxial Compression 3.4 Interpretation of Test Results
3.4 Interpretation of Test Results

The results of the triaxial tests can be compared in a normalized σ1 − σ3
principal stress coordinate system. The primary aim is to estimate the
internal angle of friction for the tested materials in the low confinement
region and secondly, to assess the strength in the high confinement region.
Figure 3.8 contains the results from all the conducted tests (the test
results are also reported in Appendix A). The overall tendency observed is
that for small confinement pressures (below approximately 0.5 − 0.6 fc ) the
failure envelopes can be described fairly well by a linear relation. However,
for larger confinement stresses the rate of strength increase is declining. The
overall shape of the failure envelope can be described as smooth and convex.
In the figure, the Coulomb criterion with k = 4 is also indicated. The
failure criterion seems to agree well with the tests of the concrete specimens
of Mixture C16a, whereas a slight overestimation is seen for the remaining
mixtures at low confinement pressures and a larger deviation is observed
for higher confinement pressures. This corresponds well with tests from the
literature, see e.g. Dahl (1992b).
12 12
C16a (71%) k=4 C16b (75%) k=4
C8a (61%) C8b (67%)
10 C4a (56%) 10 C4b (68%)
C2a (55%) Coulomb k = 4
8 Coulomb k = 4 8
−σ3 /fc
−σ3 /fc
6 6
4 4
2 2
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
−σ1 /fc −σ1 /fc
(a) Mixtures a (b) Mixtures b
Figure 3.8: Test results compared to the Coulomb failure criterion with k = 4
(corresponding to ϕ = 37◦ )
3.4.1 Low Confinement Pressures

Figure 3.9 contains the results for confinement pressures up to σ1 = σ2 = −fc
with inclusion of the Coulomb failure criterion for k = 4 and k = 3. A value
of k = 3 corresponds to ϕ = 30◦ . It can be seen that for low confinement
pressures, the internal angle of friction for mortar should be approximated
with a value closer to 30◦ rather than 37◦ . This corresponds well with the

3.4 Interpretation of Test Results Mortar in Triaxial Compression
5 5
C16a (71%) C16b (75%)
C8a (61%) C8b (67%)
4 C4a (56%) 4 C4b (68%)
C2a (55%) k=4 Coulomb k = 4 k=4
Coulomb k = 4 Coulomb k = 3
Coulomb k = 3
−σ3 /fc
−σ3 /fc
3 3
k=3 k=3
2 2
1 1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Figure 3.9: Test results from all mixtures in the low confinement region
test results of Nielsen (1998), who suggested the same value for high strength
mortar.
The results also show that the mixtures with a maximum grain size of
dmax = 8 mm have a smaller strength increase than the mixtures with 16
mm grains. However, a larger strength increase is observed for the mixtures
with maximum 8 mm grains than for the mixtures with maximum 4 mm
grains. This leads to the conclusion that an increase in maximum grain
size positively influences the triaxial strength. Yet, the conclusion is partly
based on a simultaneous increase in aggregate volume content, which also
influences the properties.
3.4.2 High Confinement Pressures

Figure 3.10 contains the test results for confining pressures larger than 0.5
fc . The Coulomb failure criteria with k = 4 and k = 1 are also included. A
value of k = 1 corresponds to an internal angle of friction of 0◦ . It can be seen
that the inclination of the failure envelopes for the mixtures with smaller
dmax tend towards a linear relation described by k = 1. For larger dmax ,
the transition between the initial inclination and the decreased inclination
for increased confinement pressure appears at relatively higher confinement
pressures and the decrease in inclination is smaller. In this relation, the
actual expression for the estimated failure criterion (intersection with the
ordinate) is less important. It is more relevant to note, that the failure
criterion of all test series with reasonable accuracy can be describe by a bi-
linear relation as suggested by Dahl (Equation (3.1)), adopting appropriate
k values for the two regimes. For the mortars tested the relationship should

Mortar in Triaxial Compression 3.4 Interpretation of Test Results
12 12
C16a (71%) C16b (75%)
k=4 C8b (67%) k=4
C8a (61%)
10 C4a (56%) 10 C4b (67%)
C2a (55%) Coulomb k = 4
Coulomb k = 4 Coulomb k = 1
8 8
Coulomb k = 1
−σ3 /fc
−σ3 /fc
6 6
4 4
k=1 k=1
2 2
0 0
0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3
Figure 3.10: Test results from all mixtures in the higher confinement region
describe an internal angle of friction of 30◦ for low confinement pressures and
conservatively ϕ = 0◦ can be suggested for the higher confinement region.
The coefficients to adopt would then increase with increasing maximum grain
size and aggregate content of the mixtures.
From a micromechanical point of view, the change in inclination of the
failure envelope can be interpreted as a transition to failure only in the
cement paste without any significant additional strength gained from the
aggregates. This is in accordance with the findings of Palaniswamy and
Shah (1974) who observed that the fracture of concrete at low confinement
pressures was governed by bond failure (between aggregates and cement
paste) whereas the failure at high confining pressures was governed by paste
strength.
3.4.3 Evaluation of Aggregate Content

When comparing the results of C8b and C4b it can be concluded that an in-
crease in maximum grain size has a positive influence on the triaxial strength.
However, the difference is not great and when relating to the results of the a
mixtures it seems that the aggregate content has a greater influence on the
strength increase.
In this relation, it should be noted that the results of Mixture C2a lie
above the results of C4a. This is believed to be due to the fact that the
specific surface area of the aggregates were greater for the C2a mixture
compared to C4a (smaller aggregates with similar volume percentage of ag-
gregate content implies a larger total surface area). The larger surface area
may have an impact similar to an increased volume content. It is also worth

3.4 Interpretation of Test Results Mortar in Triaxial Compression
mentioning that the C16b mixture deviates more from the Coulomb crite-
rion with k = 4 at higher confining pressures than the C16a mixture. As the
compressive strengths and maximum grain sizes were similar, the difference
must be found in the aggregate content. It seems that an increase in volume
percentage above a certain limit does not have a positive influence on the
triaxial strength at high confining pressures. Alternatively, the difference
can be explained by the quality of the aggregates used.
7 7
k=3 k=3
6 6
k=4 k=4
5 5
−σ3 /fc
−σ3 /fc
4 4
3 3
C4a (56%)
2 2
C4b (68%)
Dahl (52%) C8a (61%)
1 Smee (53%) 1 C8b (67 %)
Palaniswamy Dahl (61%)
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
(a) dmax = 4 mm (b) dmax = 8 mm
Figure 3.11: Comparison of test results with identical maximum grain sizes
In Figure 3.11 comparisons of the mixtures with maximum 4 mm and

8 mm grains are shown. In the figure the results of Dahl (1992a), Smee
(1967), and Palaniswamy and Shah (1974) are also included (with estimated
volume percentages) in order to widen the basis for comparison. For the 4
mm mixtures, it is found that the behaviors at low confinement pressures
are comparable (agreeing well with the Coulomb criterion using k = 3). At
the higher confinement pressures (after the transition on the failure enve-
lope), an increase in aggregate content typically affects the strength increase
positively. For the 8 mm mixtures the same tendency is not unambiguously
observed (comparing e.g. the results of Dahl (61 %) to C8a (61 %)). Some
of the explanations may be found in the fact that the mixture tested by
Dahl had a higher concrete strength (fc = 69.61 MPa) than Mixture C8a
(fc = 41.2 MPa). For low confinement pressures, the test results are com-
parable and the inclination of the failure envelope is greater than for the
mixtures with a maximum grain size of 4 mm.
In general, the observations emphasize that the proportions of the mix-
tures influence the test response more than the difference in the maximum
aggregate size. Nevertheless, it has been shown, that increases in the aggre-

Mortar in Triaxial Compression 3.5 Thin Section of Tested Specimen
gate content and the maximum grain size generally have a positive influence
on the triaxial strength.
3.5 Thin Section of Tested Specimen

To gain knowledge of the failure process of a cylinder tested in triaxial
compression, a thin section analysis was conducted for some of the specimens
of the C4a mixture. The analysis consisted of small 32x47 mm samples
impregnated with epoxy. The impregnation reveals any cracking or failure
of the matrix when inspected in a microscope.
Figure 3.12 shows an example from an untested specimen. It can be seen
that the mixture appears homogeneous without lumped aggregates. The
figure also shows that no significant cracking from creep or shrinkage had
developed at the time of analysis. The age of the specimens was not less
than the specimens tested in triaxial compression and hence it constitutes a
sound basis for evaluation of load induced cracks in the matrix.
Figure 3.12: Thin section of untested specimen with a maximum grain size
of 4 mm
According to Bongers and Rutten (1998), cracks in concrete subjected to

multiaxial loading can be described by different phenomenon’s, including:
1. crushing at the interface between aggregate and paste (bond failure)

2. cracking between and around the aggregates (in the cement paste)
3. failure through the aggregates

3.5 Thin Section of Tested Specimen Mortar in Triaxial Compression
The first phenomenon initiates before reaching the peak strength as the
interfacial zones between the aggregates are bond zones with a higher poros-
ity than the cement paste and the aggregates. Failure of this zone (by crush-
ing) explains the initial inelastic compaction of the cylinder. An increase of
the load induces cracking either in the cement paste (phenomenon 2) or
through the aggregates (phenomenon 3) depending on the local strengths
of these. By a combination of the phenomenons, shear cracks form in and
around the aggregates. The development of shear cracks changes the volu-
metric behavior from compaction to dilatation (volume expansion).
In the following, the main observations from thin section analysis of a
specimen tested with a confinement pressure of 1.62fc will be presented. The
confined strength of the concrete corresponded to 4.06 fc . Figure 3.13 shows
an example of the crack pattern at the midsection of the deformed cylinder.
Figure 3.13(a) shows where the thin section analysis was performed on the
longitudinal section of the cylinder (cylinder cut in half) and also the specific
area where the crack pattern was observed. All three phenomenon’s were
present at the midsection, see Figure 3.13(b). Crushing of the interfacial
zone was less pronounced compared to cracks through the cement paste
and through the aggregates. A high number of failed aggregates was also
observed at the midsection for specimens tested with smaller confinement
pressures. However, tests conducted with a higher confinement pressure
generally led to a larger extent of cracked aggregates.
(a) (b)
Figure 3.13: Crack pattern at the midsection of a cylinder tested to failure

with a confinement pressure of 1.62fc

Mortar in Triaxial Compression 3.5 Thin Section of Tested Specimen
(a) (b) Cracks in the side section
(c) (d) Cracks in the top section
Figure 3.14: Apperance of cracks at (a)+(b) the side of the cylinder and
(c)+(d) at the top section of the specimen

3.5 Thin Section of Tested Specimen Mortar in Triaxial Compression
At the sides of the cylinder, see Figure 3.14(b), fewer cracks through the
aggregates were observed. The cracks in the cement paste had a vertical
orientation or a slightly inclined orientation. The cracks appear primarily
around the aggregates and cracks only developed through the weak grains.
Generally, fewer cracks were observed near the side of the specimen compared
to the midsection and they were more difficult to identify.
At the top section the cracks consisted primarily of horizontal cracks
both in the interfacial bond zone but also in the paste, see Figure 3.14(d).
Combining the observed crack patterns with the shape of a tested speci-
men (see e.g. Figure 3.7), the overall crack pattern in the cylinder can be
established, see Figure 3.15. The shape can be described as an hour-glass
shape, where the top and bottom parts of the cylinder may be considered as
cones. Along the cones, cracks develop with an inclined orientation towards
the midsection of the specimen. In the remaining part of the cross section,
additional cracks develop to accommodate the overall volume expansion of
the specimen. This includes horizontal cracking of the top and bottom part
of the specimen.
Figure 3.15: Interpretation of crack pattern in a cross-section of the cylinder

after triaxial testing

Tests of Mortar in Triaxial Compression 3.6 Concluding Remarks
3.6 Concluding Remarks

This chapter presented results of triaxial tests on mortar and concrete. The
mixtures had maximum grain sizes of dmax = 2, 4, 8, and 16 mm. When
comparing the test results to the literature, it is evident that mortar pos-
sesses a smaller strength when confined, compared to concrete mixtures.
The decrease in triaxial strength is a result of the smaller maximum grain
size but also of the aggregate content of the mixture. The shape of the fail-
ure envelope is hence a result of the mixture proportions. The main findings
of the experimental tests are summarized as follows:
? For low confinement pressures (up to about 0.5 − 0.6 fc ) the shape of
the failure envelope can be considered as linear.
For concrete mixtures, a value of k = 4 (ϕ = 37◦ ) may with good

accuracy be adopted to describe the failure envelope.
For mortar mixtures, the value of k to be adopted should be taken
lower than k = 4. For assessment of the limit load a value of k = 3
(ϕ = 30◦ ) can be used with reasonable accuracy.
? For larger confinement pressures the shape of the failure envelope is

dependent on the maximum grain size of the mixture as well as the
volume percentage of aggregates.
For concrete mixtures, the slope of the failure envelope is less

than k = 4 however greater than k = 1.
For mortar mixtures, the slope of the failure envelope can be
estimated with an inclination of k = 1 (also for larger aggregate
contents).
? When combining the findings of the low and high confinement regimes,
the convex failure envelope of both mortar and concrete, can with
reasonable accuracy be described by a simple bi-linear extension of
the Coulomb failure criterion (when reasonable k factors are adopted
reflecting the proportions of the mixture).
? The failure mode of the specimens tested in the high confinement

regime can be characterized by an hour-glass shape.

4 Failure of a Mortar Shear
Key
In this chapter, an experimental study of the failure of a mortar shear key
will be presented. The investigation of the failure process was conducted
on small specimens containing a single shear key loaded in pure shear in
a push-off setup, see Figure 4.1. The main motivations for conducting the
tests were to identify the failure of a mortar shear key and to gain knowledge
of the test response before and after failure of the shear key. A non-contact
measurement technique, in shape of digital image correlation (DIC), was
used to analyze the relative displacements on the surface of the specimens.
Only a few investigations have been made on the behavior of indented in-
terfaces containing a single shear key. The investigations have mainly been
focused on applications for precast segmental bridges, which are typically
cast with concrete. As examples, the work of Zhou et al. (2005) and Issa
and Abdalla (2007) can be mentioned. These works include numerical mod-
eling and comparisons to standards. Kaneko et al. (1993a,b) developed a
mechanical model for the analysis of concrete shear key joints based on test
results of Bakhoum (1991). Common for the investigations was the focus on
Figure 4.1: Push-off test setup used for tests of an indented interface con-
taining a single shear key

4.1 Experimental Investigation Failure of a Mortar Shear Key
Table 4.1: Summary of specimen data and recorded peak loads
Series ID Key length fc,C fc,M VPeak
a 72.78 kN
b 75.12 kN
T40 40 mm 42.1 MPa 35.3 MPa
c 75.68 kN
d 80.11 kN
a 97.22 kN
T50 b 50 mm 42.5 MPa 34.4 MPa 97.92 kN
c 98.60 kN
a 109.80 kN
T60 b 60 mm 39.9 MPa 36.3 MPa 112.03 kN
c 122.46 kN
a 123.90 kN
b 124.84 kN
T70 70 mm 38.8 MPa 36.7 MPa
c 127.52 kN
d 132.90 kN
a 136.54 kN
T80 b 80 mm 35.4 MPa 33.8 MPa 138.71 kN
c 138.99 kN
dry joints or joints with epoxy applied. None of these are traditionally used
for joints in buildings. Connections cast in-place with mortar are usually
used, and the current investigation provides experimental evidence of the
failure of such a key, when the anchorage conditions for the reinforcement
are sufficient to develop yielding.
4.1 Experimental Investigation

The experimental program consisted of 17 push-off tests on specimens with
a 200x200 mm cross section. The specimens were made of two blocks each
reinforced with stirrups. The blocks had a length of 200 mm resulting in
an overall length of the specimen of 400 mm, see Figure 4.1. The concrete
block was cast against smooth formwork and the indented interface was not

Failure of a Mortar Shear Key 4.1 Experimental Investigation
treated before casting the mortar against it. The concrete mixture had a
maximum grain size of dmax = 16 mm, whereas the mortar was a commercial
mortar with dmax = 2 mm. The depth of the key, dk , was for all specimens
equal to 10 mm. The only geometric parameter varied was the key length,
Lk . The in-plane dimension of the key was equal to the specimen’s width
(200 mm) to allow for detection of key failure on the surface of the specimen.
Five key lengths were tested, see Table 4.1 for details on geometry, material
strengths (fc,C is the compression strength of the concrete and fc,M is the
compression strength of the mortar), and the recorded peak loads, VPeak .
The interface was reinforced with four rebars with a diameter of d = 6
mm placed perpendicular to the interface. The reinforcement had a yield
strength of fy = 607 MPa and an ultimate strength of fu = 705 MPa.
4.1.1 Test Setup

The test setup is illustrated in Figure 4.1. The setup is a modification of
the widely used L-shaped elements, e.g. used by Mattock and Hawkins
(1972) and Walraven and Reinhardt (1981) for monolithic joints (cracked
and uncracked), by Zhou et al. (2005) and Issa and Abdalla (2007) for the
single keyed joints, and e.g. by Cholewicki (1971) for joints with multiple
keys. The arrangement of steel plates ensured that the thrust line of the load
coincided with the plane of the interface. In addition, rollers were introduced
to allow for unrestrained movement during failure of the shear key.
During the tests, the load was applied in deformation control with a
constant rate of 1.00 mm/min piston movement. The relative shear dis-
placements were measured with linear variable displacement transducers
(LVDT’s) between the two blocks. The relative displacements in the in-
terface on one surface (where the key was visible) were in addition recorded
with images for DIC analysis.
4.1.2 Test Results

Figure 4.2 contains examples of the tested load-displacement relationships,
where the displacements are relative between the two blocks and measured
with LVDT’s in the load direction. The test results can be described by a
linear phase until the peak load was reached. The peak load was followed
by a drop in load and a subsequent stabilization and gradual increase of the
load until rupture of the reinforcement took place at relatively large shear
displacements.
The peak load was for all specimens related to a failure by complete key
cut off, see Figure 4.3 for examples of failure planes observed after testing.

4.2 Analysis by Digital Image Correlation Failure of a Mortar Shear Key
150
Lk = 40 mm
Lk = 50 mm
Lk = 60 mm
Lk = 70 mm
Lk = 80 mm
100
V [kN]
50
0
0 5 10 15
Shear displacement [mm]
Figure 4.2: Load-displacement relationships obtained from push-off tests on

single key specimens
It can be seen, both from Table 4.1 and Figure 4.2, that a larger key area
increased the first peak load.
The complete key cut off means that the residual capacity after peak
load was governed primarily by the capacity of the reinforcement. The test
results in Figure 4.2 demonstrate that the residual load levels varied less
than the peak loads, when the key length was varied.
4.2 Analysis by Digital Image Correlation

During the tests, pictures were taken manually with a digital camera (24.5
megapixel resolution) with a frequency of up to two pictures per second
around the peak load. In the following, the results of the analysis will be
used to show examples of how the failure by complete key cut off developed
and furthermore give an example of the entire displacement field measured
over the crack that governed the failure of the shear key.
4.2.1 Detection of Key Failure

Figures 4.4 and 4.5 contain examples of crack development observed in the
shear key around the peak load for a key length of 40 mm and 70 mm,
respectively. The results are presented in terms of major principal strain
(superimposed on the original image to visualize the actual crack when vis-
ible to the unaided eye). The recordings reveal displacement primarily at
the interface and cracking of the mortar at the position of the key. The pre-

Failure of a Mortar Shear Key 4.2 Analysis by Digital Image Correlation
(a) Lk = 40 mm
(b) Lk = 70 mm
Figure 4.3: Observed failure planes after testing, representative for all spec-
imens in the experimental campaign
(a) V = 78.48 kN (b) V = 79.15 kN (c) V = 81.17 kN (d) V = 62.93 kN
Figure 4.4: Development of key failure around peak load, Specimen T40d,
detected by use of DIC (major principal strain displayed)

(a) V = 124.53 kN (b) V = 126.01 kN (c) V = 127.00 kN (d) V = 51.84 kN
Figure 4.5: Detection of key failure around peak load, Specimen T70c, de-
tected by DIC (major principal strain displayed)
sented stages are carefully selected to show how the fracture developed in the
mortar. Before reaching the peak load, some displacements were recorded
in all parts of the interface except at the lower key corner, see Figures 4.4(a)
and 4.5(a). This shows that the lower part of the keyed interface was in
compression. The distributions of major principal strain show that before
failure (by shearing) initiated, a small amount of displacement (including
dilatation) had taken place in the interface and a diagonal crack had formed
originating from the lower key corner. These observations are in accordance
with the crack patterns observed by Zhou et al. (2005) and Bakhoum (1991)
for dry single key joints.
From the detailed DIC analysis, it is also found that the failure by shear-
ing of the key developed more or less instantly. It was generally observed,
that the failure did not evolve from the key corner. The strain measurements
indicate that the failure developed within the key and then progressed to the
edges of the key, eventually resulting in complete shearing. In this relation
it should be mentioned that the existing diagonal crack closed partly upon
failure and for Specimen T40d, the strains in the indentation was reduced
(compare part (d) to part (c) in Figure 4.4). For Specimen T70c the strains
in the remaining key area were not relieved as much as in Specimen T40d,
however, the diagonal crack closed. It is also observed that the failure plane
for Specimen T70c was not completely straight. This means, that further
shear displacement had to overcome the roughness of the failure plane. As
will be shown in the next section, this entails transverse displacements.

4.2.2 Analysis of Displacement Field

In the following a detailed example of the displacements measured over the
crack that led to complete key cut off will be given. The displacements can be
divided into several characteristic phases, each described by a displacement
vector and an angle of the displacement vector to vertical. In Figure 4.6 the
failure of a shear key by complete key cut off is schematically illustrated.
Here u1 describes the relative displacements over the crack in Phase 1 and α1
is the direction of the displacement vector. In the succeeding phases similar
parameters can be determined.
In the following, the relative displacements over the governing crack of
specimen T40d will be presented. Three measurements were performed over
the length of the key, at the top, at the center, and at the bottom of the
key, see Figure 4.7. As indicated by the measurement lengths, the results
are solely related to displacements in the crack, i.e. not influenced by any
slip in the interface or diagonal cracking developed in the specimen.
In Figure 4.8 the shear displacement, measured parallel to the interface
at the center line, versus the shear load, can be seen. Referring to Figure
4.4, where no strains were detected at a load close to the peak load, no
significant displacements can be measured before failure of the shear key
l
Element Element
in motion at rest
u1 Concrete
Mortar α1
u2 α Reinforcement
2
Figure 4.6: Idealized relative displacements measured over the governing

crack in a shear key

Top line
Center line
Bottom line
Figure 4.7: Position of measurement points for DIC analysis of relative

displacements at key failure, Specimen T40d
(the peak load is marked with a green circle). After failure of the shear
key the shear displacements over the crack increases until termination of the
measurements.
In Figure 4.9 the transverse displacements measured at the center line
is shown. Here, it can be seen, that some transverse displacement can be
measured before reaching the peak load. This corresponds to the crack ini-
tiation shown in Figure 4.4. After the peak load, a drop in load is seen, and
from this point the load and the transverse displacements increase until the
end of the test. In the figure, some displacement regimes (from A to H) are
indicated. These are carefully chosen as characteristic points in the displace-
ment field. Regime A contains the displacement regime before reaching the
peak load, whereas Regime B contains the drop. In these regimes, the rela-
tive shear displacements are small compared to the total shear displacements
and therefore they are difficult to identify in Figure 4.8.
In Figure 4.10 the relationships between the shear displacements and
the transverse displacements for the three lines in the different regimes are
shown (note the differences in units on the axes). For illustrative purposes,
a shear key with a crack corresponding to complete key cut off is also shown.
It should be mentioned that the displacements are relative (the shear dis-
placements on the ordinate only apply to the top line). In general it can be
seen, that the direction of the displacements changes from an initial relative
large inclination to a smaller inclination and in the end of Regime H, the
displacements are almost parallel to the interface.

100
90
80
70
60
Force [kN]
50
40
30 CD E F G H
20
10
0
0 5 10 15
Figure 4.8: Relationship between the recorded load and shear displacements
measured by DIC over the crack in the mortar key, Specimen T40d
100
90
80
70
60
Force [kN]
50
40
30 AB C D E F G H
20
10
0
0 0.5 1 1.5
Transverse displacement [mm]
Figure 4.9: Relationship between the recorded load and transverse displace-
ments measured by DIC over the crack in the mortar key, Specimen T40d

0
A
CB
D
−5
E Shear key
F
G
−10
H
−15
−2 −1.5 −1 −0.5 0
Figure 4.10: Relative displacements recorded by DIC over the crack in the
mortar key, Specimen T40d
Table 4.2 contains the calculated inclinations (to vertical) of the relative
displacements in the shear key. It is interesting to note that the angles in the
pre-peak regime, Regime A, are in a order of magnitude close to ϕ (which
with reasonable accuracy can be expected to be within 30 − 37◦ ). The drop
in load occurs when the failure of the shear key develops fully (Regime B). In
this regime the angles of the displacements are slightly increased. For further
shear displacements beyond Regime B, the angles decreases and approaches
displacements parallel to the interface.
The general crack pattern and the displacement field observed in a shear
key before and after failure are representative for what can be anticipated
for the behavior of a larger connection with multiple keys. The behavior
can be expected, when tensile yielding of the transverse reinforcement takes
place.

Table 4.2: Average angles of displacement over a crack in a mortar shear

key, measured in the different displacement regimes (measured by DIC)
Regime Bottom line Middle line Top line
αA 42.3◦ 39.4◦ 34.2◦

αB 50.9◦ 45.0◦ 41.7◦
αC 17.6◦ 21.5◦ 24.0◦
αD 9.3◦ 9.4◦ 9.4◦
αE 6.4◦ 6.3◦ 6.2◦
αF 5.8◦ 5.8◦ 5.8◦
αG 4.6◦ 4.5◦ 4.4◦
αH 1.9◦ 1.8◦ 1.6◦

4.3 Concluding Remarks Failure of a Mortar Shear Key

In this chapter, the failure of a mortar shear key was investigated. Experi-
mental results of push-off tests were presented and analyzed by use of digital
image correlation. Examples of the development of failure in the shear key
around the peak load were presented. In addition, the displacement field for
one specimen was presented and the directions of the relative displacements
were estimated in the entire load-displacement regime. The results estab-
lished the general behavior of a keyed joint loaded in shear and indicated the
behavior that can be anticipated in a keyed shear connection with multiple
keys. The findings of the chapter are summarized as follows:
? The peak load of the single key interface was associated with failure
of the shear key and a larger key area led to a larger peak load.
? The residual load carrying capacity after failure of the shear key was
governed primarily by the capacity of the reinforcement.
? The failure of the shear key was detected by use of digital image corre-
lation. The failure by complete shearing of the key developed quickly,
without a significant increase of load, once the crack initiated.
? The failure of a mortar key loaded in shear was associated with trans-
verse displacements.

5 Tensile Capacity of Loop
Connections
This chapter contains an evaluation of the tensile capacity of the new loop
connection design. The chapter is based on Paper I where upper bound
plastic models for the tensile capacity of ’2-on-2’ loop connections were de-
veloped. The models give information on the amount of lacer reinforcement
that should be provided in a loop connection in order to obtain yielding in the
U-bars and thereby avoid a brittle premature failure of the grout. Through
a parametric study, the influence of some parameters on the capacity of the
connection will be highlighted.
When considering a keyed shear connection between precast elements,
the ability of the loop connections to transfer tension is a prerequisite for it
to be able to transfer shear forces. Figure 5.1 shows how a shear load, V ,
can be transferred by uniaxial strut action (compression) between the inden-
tations. The strut action may consist of single struts over one or more shear
keys or a combination of struts with different inclinations, see e.g. Chap-
ter 8. The diagonal compression struts must be accompanied by tension in
the transverse reinforcement (U-bar loops) in order to maintain equilibrium.
This can be seen in Figure 5.1(b) where the vertical force resultant of the
uniaxial stress field between a pair of keys is balanced by the applied shear
load (red arrows), while the horizontal force component must be balanced
by tension in the reinforcement (blue arrows). The largest capacity is ob-
tained when the reinforcement is stressed to yielding, however, this requires
a sufficient strength of the uniaxial compression strut to avoid failure of the
grout. When the reinforcement is stress to yielding, it is possible to obtain
ductile test responses, see e.g. Figure 4.2, where an increase in load was seen
after failure of the shear key.
In the conventional design of shear connections, see Figure 1.2, the longi-
tudinal locking bar provides a passive confining effect on the mortar placed
within the overlapping area of the U-bar loops. Therefore, when the over-
lapping U-bars are placed closely together (i.e. with contact), the strength
of the confined mortar is sufficient to develop yielding in the U-bar loops.
If the overlapping loops are placed with a mutual distance, the capacity of
the compressed mortar might be less than the yield capacity of the U-bars.
This is not considered in design recommendations for shear joints, see e.g.
(CEN, 2004; fib, 2008). Recently Herfelt et al. (2016) developed a numerical

5.1 Prediction of Capacity Tensile Capacity of Loop Connections
Unit for tension test Transverse reinforcement
Precast
element V V V σc V
(a) (b)
Figure 5.1: (a) Loading of shear wall connection and (b) activation of rein-
forcement, when strut action is developed between the shear keys
method based on finite element limit analysis which accounts for a spacing
between the U-bars, however, in practice such tools are rarely available. For
the new shear connection design, an essential feature is that the U-bars are
placed with a mutual distance. This has to be considered when assessing
the capacity. In this relation, it should be noted that contrary to the con-
ventional design, the inclusion of a vertical locking bar in the new design
does not improve the ability of the loop connections to transfer tension.
5.1 Prediction of Capacity

In order to assess the tensile capacity related to grout failure in the new loop
connection design, tensile tests were performed on specimens representing
one unit of a U-bar loop connection, see Figure 5.1(a). The tested con-
nection is classified as a symmetric ’2-on-2’ connection. Besides a variation
of transverse reinforcement (amount of lacer reinforcement), two different
grouting materials were tested - a concrete with maximum aggregate size of
dmax = 16 mm and a mortar with dmax = 2 mm. Details on the experimental
program can be found in Paper I.
In addition to experimental tests, Paper I also contains upper bound
models for assessment of the ultimate capacity of the connection related to
grout failure. The theoretical calculations are based on two models - one
where the tensile strength of the grout is disregarded (to be used when lacer
reinforcement is provided) and one where the tensile strength is included (to
be used when no lacer reinforcement is provided). The model disregarding

Tensile Capacity of Loop Connections 5.1 Prediction of Capacity
the tensile strength of the grout is inspired by the work of Jørgensen and
Hoang (2013) who tested symmetric ’3-on-2’ connections. The geometry is
adjusted to the proposed ’2-on-2’ connection design. The model including
the tensile strength of the grout is inspired by the work of Jørgensen and
Hoang (2015a). A main difference in the development of the models (besides
the change from ’3-on-2’ connections to ’2-on-2’ connections) is that the
works of Jørgensen and Hoang only considered a concrete grout, whereas
this study also considers the use of mortar as grouting material. As shown
in Chapter 3 the material properties of concrete and mortar are different.
The models have to account for this.
Figure 5.2: Failure mechanism for ’2-on-2’ loop connection loaded in tension,
illustrations from Paper I
The assumed failure mechanism for a ’2-on-2’ connection including lacer

reinforcement can be seen in Figure 5.2, where s is the spacing of the U-bars
from each element, a is the distance between the innermost U-bars, H is the
overlapping length of the U-bars, and b is the width of the connection. The
inclination of the yield lines between the tips of the U-bars with vertical is
described by β and α is the angle between the displacement vector and the
yield line. The optimal value of α can be calculated by Equation (5.1), see
Paper I for detailed derivation.
 
Φ
 1−2 L 

α = β + arcsin  ν 
, α≥ϕ (5.1)
s 2
s

 
1+
H
where ϕ is the internal angle of friction. The criterion of α ≥ ϕ ensures that
the normality condition is fulfilled. The capacity related to grout failure

5.2 Parametric Study Tensile Capacity of Loop Connections
(including lacer reinforcement) can be assessed by Equation (5.2), where

the optimal value of α according to Equation (5.1) has been inserted:
 v !
u 2
 u s 4ΦL ΦL s
1− −


 t +
H ν ν H (5.2a)







for α ≥ ϕ and ϕ ≥ β







 2 ! !
s 1 2ΦL s


− tan ϕ + tan ϕ −


 1+
N H cos ϕ ν H


= s (5.2b)
νAc fc 

 1 + tan ϕ



 H
for α < ϕ and ϕ > β






 s
2
s s



1+ −


(5.2c)




 H H

for α < β and ϕ ≤ β

In Equation 5.2, Ac is the circular overlapping area of the U-bars, fc is the

compressive strength of the grout, ν is the effectiveness factor for concrete
in compression, and ΦL is the reinforcement degree of lacer reinforcement.
Comparison of test results with the theoretical models can be seen in
Figure 5.3. The test results are captured well by the models when the
proportions of the grout mixtures are accounted for in the choice of internal
angle of friction and effectiveness factor. Detailed discussion of ϕ and ν are
given below. The capacity related to yielding of the U-bars is included as an
upper limit for the tensile capacity of the loop connection (to explain the test
results above the yield capacity of the U-bars in Figure 5.3(a), the ultimate
strength limit, fu , related to hardening of the reinforcement is included with
a dashed line). It should be noted that Figure 10(a) in Paper I contrary to
Figure 5.3(a) is plotted with a minor mistake. The plateau for small degrees
of ΦL is plotted with a slightly smaller value of N/(νAc fc ) than calculated
by the model. The theoretical representations and equations are correct.
5.2 Parametric Study

From Equation (5.2) it can be seen that besides the amount of transverse
reinforcement (i.e. the double T-headed lacer bar), a number of factors in-
fluence the capacity related to failure of the grout. The following parameters
will be varied in order to highlight and assess their influence:
• Spacing between the U-bars (s)

• Internal angle of fiction (ϕ)
• Effectiveness factor (ν)

Tensile Capacity of Loop Connections 5.2 Parametric Study
1.5 1.5
4As fu
↓ νAc fc
1 1 ↑ 4As fy
N/(νAc fc )
N/(νAc fc )
νAc fc
↑ 4As fy
νAc fc
0.5 0.5
Concrete(C), K=0.88, ϕ = 37◦ Mortar(T), K=0.75, ϕ = 34◦

0 0
0 0.5 1 1.5 0 0.5 1 1.5
ΦL ΦL
(a) Concrete, dmax = 16 mm (b) Mortar, dmax = 4 mm
1.5
↑ 4As fy
1 νAc fc
N/(νAc fc )
0.5
Mortar(M), K=0.75, ϕ = 32◦

0
0 0.5 1 1.5
ΦL
(c) Mortar, dmax = 2 mm
Figure 5.3: Comparison between test results and models, reproduced from
Paper I
Another parameter which also influences the capacity to a great extent

is the overlapping length of the U-bars, H. In the model, H is taken as
the outer mandrel diameter, which results in a circular overlapping area.
If the overlapping area is to be increased, the width of the connection (i.e.
b in Figure 5.2) will also increase, which is not always a desirable change
of design. Nevertheless, it is obvious that an increase in overlapping area
also increases the capacity related to grout failure. This was e.g. shown by
Leonhardt et al. (1973). From a practical point of view, a more interesting
parameter to adjust is the spacing between the U-bars, s, as this indicates
the tolerances for vertical lowering of a shear panel (confer Section 1.2.1).
In this relation, it should be noted that the spacing s refers to the spacing
of the outermost placed U-bars. The spacing between the innermost U-bars
(denoted a in Figure 5.2(a)) can be varied as well. The values given in Table

Table 5.1: Basic values of parameters adopted in the parametric study of the
tensile capacity related to grout failure
Geometry Grout U-bars Lacer Bar
s 30 mm fc 30 MPa φ 8 mm φL 12 mm
D 60 mm ϕ 37◦ fy 550 MPa fyL 560 MPa
a 42 mm ν 0.6
b 100 mm
5.1 will be used as basic parameters when performing the parametric study.
5.2.1 Spacing between the U-bars

The ratio between the spacing of the outermost U-bars and the overlapping
length of the U-bars, s/H, (see Figure 5.2(a)) is influencing the capacity of
the connection to a high extent. The ratio influences both the calculation
of α, Equation (5.1), and the capacity determined by Equation (5.2). For
the practical operations on the construction site, the tolerances for vertical
lowering can be expressed in terms of s (refer to Figure 1.3). The greatest
capacity is achieved when the U-bars are placed closely together (i.e. s ≈ 0).
However, this is practically impossible to achieve with the new design (no
post-processing of the U-bars on the construction site is intended). Hence,
some spacing has to be provided while the connection still has to be designed
to develop yielding of the U-bars.
Figure 5.4 contains calculations, where the spacing of the U-bars is var-
ied. The capacities calculated by Equations (5.2a)-(5.2c) are included to
illustrate which solution that is governing for the design parameters used,
see Table 5.1. It can be seen that for a small spacing, the tensile capacity
of the loop connection is governed by yielding of the U-bars. However, for a
spacing larger than approximately s = 10 mm, the capacity is governed by
failure of the grout. The capacity corresponds to a situation where the angle
α is restricted by the normality condition (i.e. α as calculated by Equation
(5.1) is smaller than ϕ and hence α = ϕ is adopted in the calculation). For
increased spacing (i.e. increased inclination, β, of the yield lines) the ca-
pacity of the connection decreases. For a spacing larger than approximate
60 mm, the inclination of the yield line, β, becomes larger than ϕ and as
an inwards relative displacement is not possible, α must be restricted by
α = β ≥ ϕ in this regime (Equation (5.2c)), see Paper I for details.
The calculated values of α and β for increasing s can be seen in Figure
5.4(b), where also the value of ϕ is included. From the calculated values, it

1
2
Capacity
Eq. 5.2b 0.8

1.5
↓ U-bars yielding ϕ
α, β, ϕ [rad]
N/(νAc fc )
0.6
1
Eq. 5.2c
β = tan s/H
0.4
α, Eq. 5.1
0.5
Eq. 5.2a 0.2
0 0
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
s [m] s [m]
(a) (b)
Figure 5.4: Tensile capacity calculated with a variation of the spacing, s,

between the outermost U-bars
should be noted that for the internal bend diameter of the U-bar loop and
the degree of lacer reinforcement considered, the value of α as calculated by
Equation (5.1) is not applicable for any spacing of the U-bars. Hence, the
capacity is governed by the restriction of the normality condition. In this
relation, it should be mentioned that for the situation where α > β > ϕ
(which is not explicitly covered by Equation (5.2)), the capacity can be
calculated by Equation (5.2a) as well.
For practical applications, a larger amount of lacer reinforcement, i.e.
larger lacer bar diameter, is recommended in order to allow for a larger
spacing between the U-bars. A decrease in the yield capacity of the U-bars
would also allow for a larger spacing.
5.2.2 Internal Angle of Friction

In Paper I different values of the internal angle of friction, ϕ, were adopted
for the grout materials used in the experimental program. Furthermore,
different values for the effectiveness factor, ν, were also adopted, depending
on the grout material. Hence, the direct influence of a change in ϕ was not
clearly illustrated. In the following, the value of the internal angle of friction
will be varied. The results are shown in Figure 5.5.
It can be seen, that for smaller values of ϕ, a larger amount of lacer
reinforcement is required to obtain yielding of the U-bars. This is highly
relevant when using a mortar as grouting material. Referring to the findings
of Chapter 3, the proportion of the mixtures must be considered because fac-
tors like the maximum grain size and volume content of aggregates influence

1.6
1.4 ↓ U-bars yielding
1.2
N/(νAc fc ) 1
0.8
0.6
0.4 30◦◦
32◦
0.2
34◦
36◦
38
0
0 0.5 1 1.5 2
ΦL [-]
Figure 5.5: Influence of ϕ on the tensile capacity of loop connections accord-

ing to the upper bound model
the properties of the mixtures. Therefore, in order to design a connection

layout that is not governed by failure of the grout, an accurate estimate of
ϕ must be adopted in the modeling. Referring to Chapter 3, a choice of
ϕ = 30◦ is a reasonable estimate, when detailed information of the mor-
tar is not available. Comparing to the experimental results, the choices of
ϕ = 34◦ for the mixture with dmax = 4 mm and ϕ = 32◦ for the mixture
with dmax = 2 mm provided good agreement, see Figure 5.3. A choice of
ϕ = 30◦ would be slightly conservative for the considered mixtures. The re-
sults of the tensile tests confirm the findings of the triaxial tests, as a value
of ϕ = 30◦ is to be considered as a lower limit. If detailed knowledge of the
triaxial behavior of the mixture is available, the choice of ϕ can be adjusted
accordingly.
5.2.3 Effectiveness Factor

When using rigid-plastic theory to structural concrete, a proper choice of
effectiveness factor, ν, must be adopted to account for the quasi-brittle be-
havior of concrete. The influence of the choice of effectiveness factor is shown
in Figure 5.6. As expected, it is found that a decrease in effectiveness factor
increases the amount of lacer reinforcement required to obtain yielding of
the U-bars.
The failure of a mortar material is more brittle compared to failure in
concrete. This effect was seen in the tests reported in Paper I and originates
partly from the ability of cracks/yield lines to transfer shear via aggregate

150
↓U-bars yielding
100
N [kN]
50
ν = 0.4
ν = 0.6
ν = 0.8
ν = 1.0
0
0 0.5 1 1.5
Φ L [-]
Figure 5.6: Influence of the effectiveness factor for the grout in compression,
ν, on the tensile strength on the loop connection
interlocking. The effect is reflected in the choice of effectiveness factor, where

a smaller value of ν accounts for a more brittle failure.
In Paper I, the effectiveness factor was estimated based on a formula
originating from beam shear tests (Zhang, 1997a). Jørgensen and Hoang
(2013) adopted the formula for the tensile capacity of loop connections. By
introduction of the factor K to account for differences in brittleness between
concrete and mortar, the effectiveness factor was calculated by Equation
(5.3):
!
K 1
ν=√ 1+ √ , (fc in MPa and H in m) (5.3)
fc H
In Paper I, K was taken as 0.88 for concrete and 0.75 for mortar (based
on the suggestion of Jørgensen and Hoang (2015b)), which corresponded to
an effectiveness factor of ν = 0.63 for concrete and ν = 0.55 − 0.57 for the
mortar materials. Despite the fact, that the basic formula was developed
for beams in shear, the formula seems to apply well for the tension tests as
well. The effectiveness factor accounts for brittleness effects and size effects.
The brittleness effects are accounted for by the term involving fc , which
entails that higher values of fc decreases ν. This relation is well known. By
decreasing the factor K, the higher brittleness of e.g. a mixture with fewer
and smaller aggregates is accounted for.
The size effect is also related to softening/brittleness effects. This has to
be accounted for when using a plasticity approach as the average strength
over the length of the yield line decreases with increasing length. Since the
overlapping area was kept constant, the size effect (expressed in terms of the

characteristic length H) cannot be evaluated in the current testprogram.

However, it can be mentioned that the adopted size effect factor 1 + √1H
makes the theoretical model fit the test results well and that the results of
similar tension tests on symmetric loop connections performed by Jørgensen
and Hoang (2013) showed reasonable correlation with calculations, when the
characteristic length was varied.
5.2.4 Connections without Lacer Reinforcement

For the connections without lacer reinforcement, the tensile capacity as-
sessed by an upper bound solution depends entirely on the tensile strength
of the grout. This constitutes a challenge for practical applications, simply
because the tensile strength of grouts is relatively low and the tensile be-
havior is extremely brittle. When using a mortar as grout material all these
effects become more pronounced compared to a concrete grout, due to the
smaller grain sizes. Nevertheless, the capacity does not vanish entirely if
lacer reinforcement is omitted.
The capacity of a connection without lacer reinforcement was in Paper
I estimated by Equation (5.4), assuming a slightly simplified failure mecha-
nism.
N0 l − m sin α At ν t f t Al ν t f t
= +4 tan (α − β) + , α≥ϕ
νAc fc cos β cos (α − β) νAc fc νAc fc
(5.4)
The parameters l and m are given in Equations (2.15) and (2.16) and
are related to the tensile strength of the grout. The optimal angle of dis-
placement is given by Equation (5.5).
 
!
 m 4A ν f 1
 
t t t 
α = β + arcsin  − s , α≥ϕ (5.5)
 l lAc νfc s 2
 

+1
H
From the equations it is observed that the effective tensile strength, νt ft ,
influences the capacity to a high extent. In Paper I, the effectiveness factor
for concrete in tension, νt , was estimated by a formula adopted for concrete.
This formula was not reduced when applied to mortar and a general value of
νt = 0.65 was calculated for the geometry tested. This νt -formula combined
with the estimation of the tensile strength based on the compressive strength,
Equation (2.13), might overestimate the effective strength of the mortar. On
the other hand, the conservative determination of the areas included in the
failure mechanism counteracts the possible overestimate of the strength and
in summary, the capacity is estimated well, see Figure 5.3(c).

0.8
N/(νAc fc ) 0.6
α=ϕ
α>ϕ
0.4
0.2
N
N0
0
0 1 2 3 4 5 6
ft [MPa]
Figure 5.7: Tensile capacity of loop connections versus the tensile strength,
ft , of the grout
Figure 5.7 shows the influence of the tensile strength of the grout on the
tensile capacity of the loop connections, using the basic parameters given in
Table 5.1. For comparison, the strength estimated by Equation (5.2) is also
included. As expected, an increase in tensile strength increases the tensile
capacity assessed by the model without lacer reinforcement. It can be seen
that the mechanism assuming yielding of the lacer reinforcement estimates
a higher capacity. For small reinforcement degrees (under-reinforced with
respect to lacer reinforcement) the actual capacity is governed by a combina-
tion of tensile yielding of the lacer bar and a contribution from the residual
tensile strength, as the crack may not be complete stress free at the time
of reinforcement yielding. This combination is, however, rather complex
to model. Therefore, for simplicity, the tensile capacity of under-reinforced
connections has to be estimated by ignoring the contribution from the lacer
reinforcement and include the effective tensile strength of the grout. The
purpose of such a solution is to avoid under-reinforced connections, where
failure is brittle. In practice, however, the connection should always be de-
signed for U-bar yielding, as the most ductile response is obtained by this
and not by failure of the grout.

5.3 Concluding Remarks Tensile Capacity of Loop Connections

The chapter discussed the tensile capacity of the proposed ’2-on-2’ loop
connection design. The development of the models and presentation of test
results are contained in Paper I. The models predict the tensile capacity
related to grout failure, and based on this, the amount of lacer reinforcement
that should be provided in order to achieve yielding of the U-bar loops, can
be estimated. The chapter outlines the influence of some theoretical aspects
when applying upper bound modeling. The discussion was presented as
a parametric study and the following remarks should be considered when
applying the design:
? The shear capacity of a keyed shear joint is dependent on the tensile

capacity of the transverse reinforcement (i.e. the U-bar loops).
? The spacing of the outermost U-bars in the symmetric ’2-on-2’ con-

nection influences the capacity to a high extent. A larger spacing can
be obtained without a decrease in capacity, when the amount of lacer
reinforcement is increased.
? The capacity can be assessed by upper bound models with reasonable

accuracy, when adopting appropriate material parameters and effec-
tiveness factors for the grout material used in the connection.
The internal angle of friction influences the capacity significantly

and a lower value should be adopted for a mortar grout compared
to a concrete grout.
The effectiveness factor should account for the brittleness of the
grout, i.e. a smaller value should be adopted for a mortar grout
compared to a concrete grout.
? For under-reinforced connections, the capacity can be assessed by in-

clusion of the tensile strength of the grout, however, for practical ap-
plications the connections should always be designed to obtain yielding
of the U-bars (ductile response).

6 Tests of Shear Connections
In this chapter, experimental tests of shear connections will be presented.
The purpose of this chapter is partly to describe details and results of the
experimental campaign and partly to make it easier for the reader to navigate
between test series already published (Paper II and III) and test series not
yet published. The campaign provided a basis of experimental results that
can be compared with the theoretical models described in the following
chapters. The test series included specimens with the conventional shear
connection design, but the experimental campaign primarily focused on the
new design. Table 6.1 contains an overview of the test series included and
the main parameters varied in the different series. The overall focus was on
the geometry of the shear keys, with the aim of identifying the important
factors for a strong and ductile connection. In this chapter, the shear load
applied in the tests will be denoted P , following the notation adopted in
Paper II.
6.1 Experimental Campaign

The test specimens were designed as two L-shaped reinforced concrete ele-
ments connected by grouted U-bar loop connections. The specimens con-
tained three (n = 3) shear keys and four (n + 1) loop connections. Figure
6.1 illustrates the general layout of the test specimens, including definition
Table 6.1: Overview of the experimental series
Series Design Layout U-bars Variation Grout
1 Conventional ’1-on-1’ 8 mm none mortar

2 New design ’1-on-2’ 8 mm key depth mortar
5 New design ’2-on-2’ 8 mm key length mortar
7 New design ’2-on-2’ 10 mm key length mortar/concrete

6.1 Experimental Campaign Tests of Shear Connections
P P
Precast Precast
element Locking bar, φL element
Locking bar
L k Ak s
hk L
t
dk
A-A Lk B-B
Grout Grout
Precast Precast b
element element
P P
(a) Conventional design (b) New design
A-A B-B D B-B D
s1 s1
D s1 a
s1
(c) Conventional (d) ’1-on-2’ connection (e) ’2-on-2’ connection
Figure 6.1: Layout of test specimens used in the experimental campaign and
indication of the load application points

Tests of Shear Connections 6.1 Experimental Campaign
of geometrical parameters related to the shear keys and the loop connection
design. A particular feature of the tested connections was that the U-bar
loop connections, as seen in Figure 6.1, were placed outside the indented
zones.
The test campaign comprised 64 specimens. In Table 6.1 the campaign
is divided into 7 series, identifying the design and the three different layouts
of loop connections that were tested. In Series 1, the conventional design,
see Figures 6.1(a) and 6.1(c), was tested. The series included 6 identical
specimens denoted R1-R6. In Series 2, the new design with a ’1-on-2’ loop
configuration, see Figures 6.1(b) and 6.1(d), with 8 mm U-bars was tested.
The series contained 4 specimens, S1-S4, with a smooth interface without
shear keys and 8 specimens with shear keys, P1-P8. Tests on this configura-
tion functioned as pilot tests that were used to benchmark the new design
against the conventional design (Specimens P1-P4 had the same amount
of reinforcement in the weakest side of the connection and an identical ge-
ometry of the shear keys as Specimens R1-R6). In Series 3, the ’1-on-2’
configuration was tested with 10 mm U-bars, P9-P12. In Series 4-7 tests
were conducted for the new design with a ’2-on-2’ loop configuration, see
Figure 6.1(e). Series 4 tested 6 mm U-bars and a varying key depth. The
specimens are denoted D followed by a number indicating the key depth.
The results have been presented in Paper III. In Series 5 and 6 specimens
with 8 mm U-bars were tested. These specimens are denoted with roman
numbers and contained in Paper II. Specimens R1-R3 and P1-P4 are also
part of Paper II. Finally, in Series 7 specimens with 10 mm U-bars were
tested. In this series the key length was varied and both mortar and con-
crete was used as grouts. For this reason the specimens are denoted with M
and C followed by a number representing the key length.
Table 6.2: Geometrical parameters used in the experimental campaign, all

dimensions in [mm], including references to Appendix B
Layout b s L hk Lk dk φL Appendix
1 Reference 80 300 1280 85 160 16 12 B.1

2 ’1-on-2’ 80 300 1280 85 160 0-30 - B.2
3 ’1-on-2’ 80 300 1280 85 160 16-30 - B.2
4 ’2-on-2’ 80 300 1280 200 140 10-20 12 B.3
5 ’2-on-2’ 100 300 1280 100 120-180 28 12 B.4
6 ’2-on-2’ 100 300 1280 200 140 10-28 12 B.4
7 ’2-on-2’ 120 320 1340 200 120-180 10 12 B.5

6.2 Test Results Tests of Shear Connections
In the new design, the diameter of the lacer bar was carefully designed
(according to the model presented in Chapter 5) to ensure that the tensile
capacity of the overlapping loops was governed by tensile yielding and not
crushing of the mortar. In the ’2-on-2’ connections, anchorage plates were
mounted in each end of the locking bar in order to eliminate any boundary
effect regarding the anchorage conditions for the locking bar. Table 6.2 con-
tains an overview of the geometrical parameters of the test series. Appendix
B contains all details on the specimens including the recorded first peak
load, PFP , and the ultimate load, PU , recorded at large shear displacements.
6.1.1 Test Arrangement

The tests were performed with the specimens placed horizontally in a steel
frame build on a strong floor. In Appendix B a schematic overview and
a brief description of the test arrangement can be found. The precast ele-
ments rested, on each side of the connection, on smooth plates in order to
minimize friction and allow for unrestricted out of plane deformation of the
joint grout. In Figure 6.1, the load application points are indicated. The
thrust line ensured pure shear at the midsection of the connection. The pre-
cast L-shaped elements were designed with an overstrength to ensure failure
in the connection. A similar test arrangement has been used by e.g. Ha-
lasz and Tantow (1966), Cholewicki (1971), Pommeret (1972), and Fauchart
and Cortini (1972). In this relation, it should be noted that the loading
conditions do not ensure pure shear at the position of the keyed interface.
However, as the connection has a width, b, the condition of pure shear must
refer to the midsection of the connection.
6.2 Test Results

A number of investigations have previously been conducted on the behav-
ior of conventional keyed shear connections. The main findings in these
investigations were summarized by Hansen et al. (1976). The results of the
reference tests performed in relation to the current study, comply well with
the descriptions by Hansen et al. (1976). Nevertheless, when comparing to
tests from the literature, a difference in the capacity is observed. It appears
that the reference specimens of the current study possessed a higher capac-
ity compared to previous test specimens, where the loops were placed inside
the indented areas. This can be explained by the fact that the grout in-
between a pair of shear keys is utilized both to ensure the activation of the
transverse reinforcement and to carry the diagonal strut, when the loops are
placed within the keys (see also Figure 5.1 for illustration of strut action be-
tween the keys). This naturally decreases the capacity and emphasizes that

Tests of Shear Connections 6.2 Test Results
the position of the U-bar loops outside the indented zones ensures a bet-
ter structural integrity, than when the loops are placed inside the indented
areas.
For small shear displacements, the test results for the new design dis-
played great similarities to the results for the conventional design. For larger
shear displacements, the test responses of the new design proved much more
ductile than the responses of the reference specimens. In the following, ex-
amples of typical load-displacement relationships will be given. This includes
a comparison of the conventional layout with the new layout, as well as a
presentation of additional results, which has not been covered by Papers II
and III.
(a) Complete key cut off, Specimen D20B
(b) Local key corner crushing, Specimen D12A
Figure 6.2: Examples of the two experimentally observed failure modes of the
shear keys
The failure mode of the shear keys could be identified either from ex-
aminations of the specimens after testing or by use of DIC. Contrary to
the results of the single key specimens, where only failure by complete key
cut off was observed, two distinctive failure modes were observed for the
shear connections, see Figure 6.2. For the larger key depths, a failure by
complete key cut off was identified as the dominating failure mode, while
the specimens with smaller key depths could suffer from local key corner

(a) Complete key cut off and diagonal cracking, Specimen I2
(b) Local key corner crushing and diagonal cracking, Specimen M180A
Figure 6.3: Examples of diagonal cracking developed in combination with (a)

complete key cut off and (b) local key corner crushing
crushing. In Figure 6.2(a), the interface between the grout and the precast
element has been indicated with a red line to visualize the geometry of the
key. For the failure by local key corner crushing, the contour of the interface
can easily be identified, see Figure 6.2(b). The two failure modes have been
reported in the literature for the conventional loop connection layout, and
as both failure modes were observed for the new design, the observation is
not unique for any of the designs. As will be shown in the following, the
failure mode in the shear keys influenced the load-displacement relationship
of the connection after the failure of the keys had occurred.
The local failure in the shear keys was a part of the global failure mode
of the shear connections. The global failure mode identified for many of the
specimens included a diagonal crack on the surface. This was observed for a
failure by complete key cut off as well as for the local key corner failure, see
Figure 6.3. The extent of diagonal cracking varied for the specimens. Some
experienced one diagonal crack while others experienced several diagonal
cracks. The tests were continued until rupture of the transverse reinforce-
ment and for this reason, the crack pattern that could be assessed after the
test, was not necessarily representative for the global failure mode that ini-
tiated the failure of the shear keys. In Paper II, digital image correlation
(DIC) was used to identify the global failure mechanisms at the instant of
key failure (estimated based on the crack pattern on the surface). In this
relation, it can be mentioned that a failure mode by sliding along the in-

clined interface of the shear key, as described e.g. by Eriksson (1978), was
not observed in the tests.
6.2.1 New Design versus Conventional Design

In Paper II, a comparison between the conventional and the new design was
given. Figure 6.4 repeats the comparison of the tested load-displacement re-
lationships. It should be noted that all depicted response curves correspond
to failure by local crushing of the key corners. The main differences between
the load-displacement relationships of the two designs are:
• The new design displayed a stiffer behavior before reaching the first
peak load.
• The first peak loads (corresponding to failure of the shear keys) were
comparable for identical reinforcement degrees and key geometries.
• The responses after the first peak load were significantly different:
– The new design displayed a ductile response and reached an ul-

timate capacity (corresponding to rupture of the reinforcement
loops) at relatively large shear displacements.
– The conventional design did not display a pronounced ductile
behavior after the first peak load and the shape of the load-
displacement curves seemed random.
400
P2↓ P1
P3 ↓
350
300 P4
250
R3
P [kN]
200
R2
150
100
R1
50
0
0 5 10 15 20 25
Measured longitudinal displacement [mm]
Figure 6.4: Comparison of tested load-displacement relationships for the con-

ventional design (Specimens R1-R3) and the new design (’1-on-2’ design,
P1-P4), graphs from Paper II

(a) U-bars pushed away from each other (Specimen R1)
(b) U-bars pushed towards each other (Specimen R6)
(c) New design ’1-on-2’ design (Specimen P12)
Figure 6.5: Deformation of U-bar loops observed after testing

The gain in ductility of the new design compared to the conventional

design is ascribed to the difference in orientation of the transverse U-bar
loops and the presence of the double T-headed lacer bar. The double T-
headed lacer bar functions as passive confinement for the grout in the area,
where the U-bars overlap. When comparing the deformations of the U-bars
in the two designs after testing, different deformation patterns were observed.
In addition, a dependency of the load direction on the U-bar deformation
was observed for the conventional design. This was not the case for the new
design as the orientation of the U-bars ensures that the response do not
depend on the load direction.
Figure 6.5(a) shows an example of how the U-bars in the conventional
design had deformed, when the U-bars were pushed away from each other,
when exposed to shear displacement (the grout was removed after test). It
was found that the extent of U-bar deformation was little despite of the
large shear displacements imposed during the test. In return, the grout was
extensively cracked at the end of the test.
In Figure 6.5(b), an example is given, where the U-bars of the conven-
tional design were pushed towards each other during the test. It can be
seen that the load direction ensured a self-locking effect between the U-bars,
when the shear displacement was increased. For this reason, the extent of
U-bar deformation was greater (U-bars clashing during shear displacement).
It was also observed that the U-bars deformed out of the plane that they
were placed in. In addition, extensive cracking was observed in the grout at
the end of the test, including a diagonal crack running through the center of
the area, where the U-bars overlapped. Loading in this direction provided
a greater residual capacity after failure of the shear keys than if the U-bars
were pushed away from each other. In the latter case, the softening response
after the peak load was very pronounced. Figure 6.4 contains examples of
both cases. The response curve of Specimen R1 represents a situation, where
the U-bars were pushed away from each other, while Specimens R2 and R3
represent the opposite situation. The difference is clearly seen, both in terms
of the recorded first peak loads and in the ability to carry load after the first
peak capacity was reached. In a real structure, the direction of the shear
load in a connection depends e.g. on the wind load direction. This makes
it impossible to ensure that the better of the two types of response in the
conventional design is always governing.
For the new design, the U-bars deformed differently, see Figure 6.5(c).
The deformations in the U-bars were concentrated at the interface, and the
remaining part of the rebars in overlap remained more or less intact. This
was observed for all diameters of the U-bars tested, both for the ’1-on-2’ and
the ’2-on-2’ configuration. In addition to the different U-bar deformations,
it can also be mentioned that fewer cracks were observed on the surface

(a) Conventional layout, displacement away from the loops
(b) Conventional layout, displacement towards the loops
(c) New layout, independent on load direction
Figure 6.6: Idealized deformations of U-bar loops observed after testing

of the connection for the new design compared to the conventional design.
The idealized deformation patterns of the U-bars in the two designs are
summarized in Figure 6.6. It should also be mentioned that some U-bars of
specimens with the new design had deformed as shown in Figure 6.7. It is not
possible to distinguish whether the deformations appeared as a consequence
of very large displacements or if the U-bars deformed as shown during the
entire test (from when plastic deformations were initiated). Nevertheless,
the response curves show that the new design, with an orientation of the
U-bars in the same plane as the precast element and with the double T-
headed lacer bar used as transverse reinforcement, displayed a significant
ductility compared to the conventional design. In this relation, the ability
of the confined concrete core in the U-bar overlaps to remain intact during
displacement is a deciding factor for a ductile response. It was observed that
the response curve of the new design had the ability to increase the load
again after the first peak capacity, and eventually reach a load level similar
to or higher than the first peak load, PFP . The second local maximum was
recorded as PU , see Appendix B. In the ultimate state, the failure of the
specimens with the new design was governed by rupture of the rebars, and
not failure of the grout. This was not always possible to achieve with the
conventional design.
Figure 6.7: U-bar deformation observed with the new design, Specimen P2
The positive influence on ductility, when changing the orientation of the

U-bars, was most clearly observed in tests on the new ’1-on-2’ design, where
shear keys were omitted, see Figure 6.8. The interfaces of these specimens
were cast against smooth formwork and greased before casting the grout.
The test response did not display a distinct first peak load due to the lack of
shear keys. However, the response is still very ductile, displaying an increase
in load until rupture of the U-bars. The transitions in the load-displacement
relationships at small displacements (from an initially stiff response to an
almost perfect plastic/hardening response) are consequences of the reinforce-

ment yielding. Examinations of the transverse reinforcement after testing

showed plastic deformation of the U-bars similar to the example shown in
Figure 6.5(c).
400
Specimen S1
Specimen S2
350 Specimen S3
Specimen S4
300
250
P [kN]
200
150
100
50
0
0 5 10 15 20 25
Figure 6.8: Tested load-displacement relationships for shear connections with

smooth and greased interfaces without shear keys (8 mm U-bars ’1-on-2’
configuration)
6.2.2 New Design - ’2-on-2’ Connections

In the following, test results of the proposed ’2-on-2’ connection layout will
be presented. Tests of the conventional design with comparable transverse
reinforcement degrees have not been carried out. Figure 6.9 contains exam-
ples of response curves for specimens with 6 mm U-bars (also presented in
Paper III). The test results in the figure are divided in two graphs, relating
to the observed failure modes of the shear keys, confer Figure 6.2.
For the specimens suffering complete key cut off, a drop in load was
observed after the peak load, PFP . Subsequently, a stable residual load level
was maintained until rupture of the U-bars at large shear displacements. For
these specimens, the ultimate load, PU , was smaller than PFP , see Figure
6.9(a). For the specimens suffering local key corner crushing, the first peak
capacity was clearly dependent on the depth of the shear keys. For increasing
key depth, an increase in peak load was seen until a limiting depth, where
the specimens (for the particular transverse reinforcement degree) suffered
complete key cut off. The drop after the peak capacity was less pronounced
compared to failure by complete key cut off and subsequently, the residual
load stabilized at a higher level (larger values of PU , see Appendix B.3) than
seen for the specimens suffering complete key cut off.

600
PF P
500
PU
400
P [kN] 300
200
100
D16 A
Failure mode D18 B
D20 B
0
0 5 10 15
(a) Complete key cut off
600
500
400
P [kN]
300
200
100 D10 B
D12 A
Failure mode D14 B
D16 B
0
0 5 10 15
(b) Key corner crushing
Figure 6.9: Tested load-displacement curves for new design with 6 mm U-

bars (’2-on-2’ configuration), graphs from Paper III
In Chapter 4, the failure of the single shear key was associated with dis-
placements in the transverse direction of the joint. This was also observed
for both types of local key failure in the larger shear connections. In Figure
6.10, the relationships between the measured transverse and longitudinal
displacements are shown for the same specimens as the ones referred to in
Figure 6.9. In order to identify jumps in the displacements, the results are
presented as points. For both failure modes, a jump in displacement was
observed around a longitudinal displacement of 1.5 mm. This corresponded
to failure of the shear keys and was also observed in the tests of single key
specimens presented in Chapter 4. The results confirm that the failure by

complete key cut off was more brittle than the local key corner crushing. It
was found that the amount of transverse displacements was small prior to
the failure of the key, and for all specimens, the amount of transverse dis-
placement increased with increasing shear displacement. The relationships
for the complete key cut off appear somewhat linear, whereas the devel-
opment for the failure by local key corner crushing appears slightly more
random. Nevertheless, it can be concluded that the displacement fields for
all specimens included transverse displacements, and that the displacement
fields were influenced by the failure of the shear keys.
2 2
D16 A D10 B
D18 B D12 A

D20 B D14 B
D16 B
1.5 1.5
1 1
0.5 0.5
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Longitudinal displacement [mm] Longitudinal displacement [mm]
(a) Complete key cut off (b) Key corner crushing
Figure 6.10: Relationships between longitudinal and transverse displace-

ments, specimens with 6 mm U-bars (’2-on-2’ configuration)
In Figure 6.11, test results for specimens with 8 mm U-bars are presented
(the results are also presented in Paper II). The test series with 8 mm U-bars
had specimens suffering complete key cut off as well as specimens failing by
local key corner crushing. In the figure, examples of response curves for spec-
imens suffering complete key cut off can be seen (Specimens II2 and VIII2),
where an initial peak load was observed followed by a drop and a subsequent
increase in load until rupture of the reinforcement. Specimen V2 suffered
from local key corner crushing. The load-displacement relationship of this
specimen confirms that the development after initial peak is dependent on
the failure mode of the shear key. Unlike the results shown in Figure 6.9,
the tests with 8 mm U-bars constitute examples, where the residual capacity
can be greater or similar to the load corresponding to failure of the shear
keys. However, a drop in load after key failure was still observed and the

600
Local key corner crushing
↓
500
400
↑
P [kN] ↑
300 Complete key cut off
200
100
II 2
V2
VIII 2
0
0 5 10 15 20 25
Figure 6.11: Tested load-displacement relationships for specimens with 8 mm

U-bars (’2-on-2’ configuration), graph from Paper II
magnitude of the sudden load decrease depended on the failure mode of the
shear keys.
Figure 6.12 shows tested load-displacement relationships for connections
reinforced with 10 mm U-bars. Both concrete and mortar were used as
grout in the test series. All tests showed failure by local key corner crushing
and the load-displacement relationships proved very ductile with no drop
in load. The test results showed no dependency on the key length (similar
800
700
600
Local key corner crushing
500
P [kN]
400
300
200
C120A
100 M120B
Failure mode
C180A
M180B
0
0 10 20 30 40 50 60
Figure 6.12: Tested load-displacement relationships for specimens with 10

mm U-bars (’2-on-2’ configuration)

results for the three key lengths tested). In addition, the test results did not
show dependency on the grout material. The mortar and concrete grouts
used in the test specimens provided more or less the same capacity for the
connections and a similar behavior during the tests.
In general, it can be concluded that the failure by local key corner crush-
ing facilitates a more ductile load-displacement relationship than the failure
by complete key cut off. In addition, it can be concluded that the amount of
transverse reinforcement governs the residual load level after failure of the
shear keys.
6.2.3 Additional Observations from Digital Image

Correlation
In the following, experimental observations obtained from analysis with digi-
tal image correlation (DIC) will be presented. The analyses were performed
using the software Aramis developed by GOM. Examples of DIC results
were also given in the appended Papers II and III. The following examples
are in line with the observations described above and they supplement the
observations presented in the papers.
The comparison of the conventional design with the new design showed
significant differences in the post peak load-displacement response. DIC
(a) Prior to peak load
(b) Just after peak load
Figure 6.13: DIC analysis (major principal strain) of reference Specimen R4

(U-bars pushed towards each other), showing (a) diagonal cracks prior to
peak load and (b) cracks in the post peak region

One U-bar per loop
Two U-bars per loop

Figure 6.14: Major principal strain recorded on the surface of Specimen P4
at the first peak load, showing no diagonal cracks (figure from Paper II)
results clearly show that the cracking behaviors of the connections were sig-
nificantly different as well. Figure 6.13(a) indicates that diagonal cracks
developed for the conventional design even before reaching the peak load
of the connection. Naturally, the cracking process decreased the stiffness
of the connection as seen in the ascending part of the response curve (the
comparison to the stiffer response curve of the new design is given in Figure
6.4). In the post peak region, when the shear keys had failed, the diagonal
cracks could open widely (partly as a result of the U-bar orientation), see
Figure 6.13(b). The grout therefore gradually lost the ability to maintain
or increase the load level even though the U-bars were pushed towards each
other (the U-bars deformed as shown in Figure 6.5(b)). In the ’1-on-2’ de-
sign, a different crack pattern was observed. Prior to the first peak load,
diagonal cracks had not necessarily developed, see Figure 6.14. With in-
creased shear displacement the diagonal cracks developed over one pair of
shear keys, and not over several keys.
In Paper II, a crack pattern was shown for the ’2-on-2’ design. Less
diagonal cracks were observed in the grout prior to failure of the shear keys
and the observed cracks had smaller crack openings. Figure 6.15 shows
the crack pattern in a late stage of the particular test. Contrary to the
conventional design, where the diagonal cracks were distributed over the
length of the connection, the diagonal cracks in the new design were typically
located around a single pair of shear keys. This was maintained through the
Figure 6.15: Cracks at a late stage of the test, Specimen III2, ’2-on-2’ loop
configuration

(a) M120A, Lk =120 mm
(b) C180A, Lk =180 mm
Figure 6.16: Cracking behavior of specimens with 10 mm U-bars recorded

with DIC
test, despite the large amount of shear displacement that was imposed to
the connection at that stage (approximately 15 mm). At the remaining
shear keys, the displacements concentrated in the interface, where the shear
keys had failed and the grout maintained the ability to carry additional load
until rupture of the reinforcement. In this relation, it is beneficial that the
transverse reinforcement loops were placed outside the keyed areas, i.e. in
the less cracked zones.
The same crack pattern was observed for the specimens with 10 mm
U-bars. Figure 6.16 shows snapshots of the crack development around the
first peak load for two specimens. As can be seen, diagonal cracks between
opposite keys were most dominant, while the zones where the U-bar loops
overlapped were much less cracked (at least what could be observed on the
surface). This can partly explain the very ductile and nearly perfect plastic
behavior seen on the response curves. It should be noted that it was not
only because of the position of the U-bar loops, but also the orientation of
the loops, as well as the presence of the T-headed lacer bars that ensured a
ductile behavior for the new design.
In Paper II and III, it was shown that the failure of a shear key can be
detected by the use of DIC. In Figure 6.17, an additional example of com-
plete key shearing at the first peak load is given (Specimen D20B). It can be
seen that the crack developed where there was no previously measured strain
(detected on the surface). However, the failure developed clearly as complete

(a) Just before peak load
(b) Just after peak load
Figure 6.17: Identification of key failure, Specimen D20B
key cut off. The existing strains/openings at the interfaces and in the diag-
onal crack closed at the failure of the shear key. This is in accordance with
the observations presented in Chapter 4. The DIC measurements confirm
the jump in longitudinal displacement caused by the brittle failure of the
shear key. In this relation, it can be mentioned that the crack appearance
on the surface of the connection indicated that the contour of the failure
plane was not completely parallel to the line of loading. This confirms that
additional longitudinal displacement in the post peak regime had to be ac-
companied by displacement in the transverse direction. This is again similar
to the results of the single key tests.
In Figure 6.18(c), a crack pattern including local key corner crushing and
diagonal cracks is shown. Figure 6.18(a) and (b) show the development of
a diagonal crack as the load-displacement curve enters the regime of plastic
deformations (after failure of the shear keys). The small key depth (10 mm)
favors a failure by local key corner crushing. In this relation, it should be
mentioned that the failure by local crushing of the key corner cannot be
detected as easily as the failure by complete key cut off. For this purpose,
the failure of the key corner is too localized in a too narrow zone.

The local failures of the keys in combination with the diagonal cracks
form the global mechanism. As discussed above, a failure of the key cor-
ners provides the most ductile load-displacement response. The specimens
of the series with ’2-on-2’ connections with 10 mm U-bars are excellent ex-
amples of an ideal combination of key geometry (primarily in terms of key
depth) and reinforcement strength that ensures an almost perfect plastic
load-displacement relationship.
(a) Left key, before key failure
(b) Left key, after key failure
(c) Crack pattern on surface including all keys
Figure 6.18: Identification of key failure, Specimen C150A

Tests of Shear Connections 6.3 Concluding Remarks

In this chapter, representative experimental results of push-off tests on shear
connections, with three pairs of shear keys and four sets of reinforcement
loops, were presented. The results included typical load-displacement rela-
tionships for the different test series, both for the conventional and for the
new design. The recorded first peak loads, PFP , and the ultimate load, PU ,
for all the tests can be found in Appendix B. DIC analyses were performed
to identify the failure modes of the shear keys and the cracking behavior
of the grout. The experimental results serve as a basis for comparison to
the theoretical models that will be developed in the following chapters. The
main findings of the experimental tests are summarized as follows:
? The first peak load of the keyed shear connections was governed by
failure of the shear keys.
? Two different failure modes were observed in the shear keys - complete
key cut off (larger key depths) and local key corner crushing (smaller
key depths).
? Failure by local key corner crushing resulted in the most ductile be-
havior for the connection.
? Diagonal cracking appeared between pairs of shear keys (i.e. diagonal

crack spanning over one key) for both the conventional and the new
design.
? The orientation of the U-bars and the T-headed lacer bars in the new
design for keyed shear connections ensured a ductile test response for
both types of failure in the shear keys.
Specimens with the new layout, but without shear keys, also expe-
rienced a significant increase in capacity after initiation of plastic
deformation.
The tests on the new design were all terminated by rupture of
the reinforcement. This could not always be obtained for the
conventional layout.
? The load direction proved important for the ductility of the conven-
tional design, whereas the response of the new design, due to the re-
inforcement layout, is independent of the load direction.

7 Upper Bound Solutions for
Keyed Shear Connections
In this chapter, upper bound solutions for keyed shear connections will be
presented and compared to the results of the experimental campaign pre-
sented in Chapter 6. The solutions have been developed and presented in
the appended Paper II and they constitute so-called first-order rigid-plastic
solutions. The calculations can be compared to the recorded first peak loads
of the experimental tests. The solutions have been developed with inspira-
tion from the observed crack patterns, i.e. the local failures of the shear keys
and diagonal cracking are considered in the failure mechanisms. The model
furnishes a simple tool to optimize the geometry of a keyed connection in
order to predict the most ductile failure mode. The solutions are applicable
for a variety of keyed connections, and they are not limited to the new or
the conventional design. Nevertheless, the aim is primarily to compare with
tests of the new design.
The upper bound method has been applied for strength prediction of
keyed shear connections by a number of researchers. Jensen (1975) was the
first to establish upper bound solutions and later the same basic assump-
tions was used by e.g. Chakrabarti et al. (1988), Abdul-Wahab and Sarsam
(1991), and Christoffersen (1997). Common for all studies found in the liter-
ature is that they only consider a local failure mechanism corresponding to
complete key cut off. However, several experimental investigations, including
those by Hansen et al. (1976), Eriksson (1978), and Nimityongskul and Liu
(1980) also reported failure by local crushing of the key corners, see Figure
7.1 for distinction between the two idealized failure modes. Since a theo-
retical treatment of the local key corner failure is lacking in the literature,
the present work has been focused on this failure mode in the experimental
campaign, in order to have a better basis for model development. In the
following, the solutions, as they appear in Paper II, will also be compared
to the additional test results not included in the paper. Detailed derivation
of the solutions will not be repeated here.
7.1 Prediction of Capacity

When modeling the capacity of keyed shear connections, a number of as-
sumptions has to be made in order to simplify the analysis. The following

7.1 Prediction of Capacity Upper Bound Solutions
Lk Lk
dk dk
Complete key cut off Local key corner crushing
Figure 7.1: The two local failure modes of the shear keys considered in the
development of the model, see also Paper II
must be fulfilled as a prerequisite for plastic modeling of the structural con-

nections:
• The design of the transverse loop connections allows for tensile yielding
of the U-bars (not failure of the grout)
• The precast elements possess an over-strength compared to the con-
nections
The main theoretical assumptions include:
• Rigid-perfectly plastic material behavior

• The grout is considered as a modified Coulomb material (tensile strength
neglected)
• The problem is considered as a plane strain problem
The following notations are introduced for the average shear stress, τ ,
the degree of transverse reinforcement, Φ, and the degree of longitudinal
reinforcement, ΦL (relating to the locking bar):
Pcal
τ= (7.1)
nAk
n + 1 A s fy
Φ= (7.2)
n Ak f c
AsL fyL
ΦL = (7.3)
nAk fc
where Pcal is the calculated shear capacity, n is the number of shear keys
in a connection, Ak is the cross sectional area of a shear key (Lk · hk ), fc is
the compressive strength of the grout, As is the reinforcement area per loop

Upper Bound Solutions 7.1 Prediction of Capacity
connection, fy is the yield strength of the U-bars, AsL is the cross sectional
area of the locking bar, and fyL is the yield strength of the locking bar.
7.1.1 Failure Mechanisms

Based on the experimental results, four basic failure mechanisms were devel-
oped and presented in Paper II. The mechanisms were developed based on
two considerations. Firstly, the two failure modes of the shear keys should
be reflected (see Figure 7.1). Secondly, the presence of diagonal cracking be-
tween opposite placed shear keys, see e.g. Figure 6.3, should be addressed.
By use of the work equation, the closed formed solutions were developed and
optimized.
Two of the basic mechanisms were based on complete key cut off, see
Figure 7.2, and the remaining two basic mechanisms were based on key
corner crushing, see Figure 7.3. The mechanisms can be categorized as single
body mechanisms, where one rigid body displaces relatively to another rigid
body. Multi-body mechanisms, including relative displacements of several
rigid bodies, can also be considered. Examples of such mechanisms were
given e.g. by Jørgensen et al. (2017), who studied keyed shear connections
reinforced with high strength wire rope loops. For the connection design
developed and tested in this study, calculations showed that the multi-body
mechanisms were not critical. For this reason, they will be not considered
further.
7.1.2 Complete Key Cut Off

The mechanisms with failure by complete key cut off are presented in Figure
7.2. These will be described in the following. Mechanism A involves com-
plete shearing of all n shear keys in one side of the connection, see Figure
7.2(a). The mechanism is similar to the model of Jensen (1975). Adopting
the notation for the shear stress, τ , and the reinforcement degree, Φ, the
capacity and the optimal angle of displacement, α, are calculated by:
Mechanism A
τA 1 − sin αA Φ
= + tan αA (7.4)
νfc 2 cos αA ν
!
2Φ
αA = arcsin 1 − , αA ≥ ϕ (7.5)
ν
As diagonal cracking over one pair of shear keys was observed for many of
the test specimens, Mechanism B with a diagonal yield line was introduced.
The diagonal yield line is assumed to span over one pair of shear keys and
the dissipation in this yield line is therefore influenced by the length of the

P P
t t
l l
Element Element Element Element

at rest in motion at rest in motion
ut ut
α α
ul u ul u
β Lk
b
P P
(a) Mechanism A (b) Mechanism B (and D)
Figure 7.2: Basic failure mechanisms with complete key shearing
keys, Lk , and the width of the joint, b. The relation is expressed by the
angle tan β = b/Lk , see Figure 7.2(b). The capacity is found by:
Mechanism B (with dissipation in diagonal yield line)

τB n − 1 1 − sin αB Ad 1 − sin (β + αB ) Φ ΦL
= + + tan αB + (7.6)
νfc n 2 cos αB nAk 2 cos αB ν ν
t Φ
 
n − 1 + − 2n 
hk ν
αB = arcsin  , αB ≥ ϕ (7.7)

Ad
n−1+
 
Ak
where Ad is the area of the diagonal yield line, hk is the height of the shear
key (in the direction of the panel thickness), and t is the thickness of the
panel. This solution appears to be more complicated than Mechanism A.
However, it is still a closed form solution, which is easy to adopt for practical
applications. Relating to the experimental observations, where a diagonal
crack in many cases was observed before the peak load, it is argued that

the dissipation is reduced in such a diagonal crack due to the existing crack
opening at the time where failure of the shear keys take place, see e.g. Fig-
ure 6.13(a). In addition, the inclination of the crack entails that the relative
displacement here has a direction, which is almost perpendicular to the yield
line (meaning almost no dissipation as the tensile strength of the grout is
neglected). Taking these factors into account, the dissipation is deemed to
be relatively small in the diagonal yield line. To simplify the calculation,
this contribution is simply disregarded, when introducing Mechanism D.
However, the contribution from the locking bar is still included. The mech-
anism appears identical to Mechanisms B, see Figure 7.2, and the capacity
is calculated by:
Mechanism D (without dissipation in diagonal yield line)
τD n − 1 1 − sin αD Φ ΦL
= + tan αD + (7.8)
νfc n 2 cos αD ν ν
!
2nΦ
αD = arcsin 1 − , αD ≥ ϕ (7.9)
(n − 1) ν
7.1.3 Local Key Corner Crushing

For local crushing in the key corners, mechanisms with and without diag-
onal yield lines were likewise developed, see Figure 7.3. The upper bound
solutions for these mechanisms have two optimization parameters, namely
the angle of the displacement vector, α, and the inclination of the yield line
to vertical, γ. The dissipation and thus the capacity are highly dependent
on the combination of the two angles. If the angle γ is small, the failure will
be similar to a complete key cut off mechanism. For increasing values of γ,
the dissipation in the yield line will decrease; on the other hand, the rein-
forcement will dissipate more energy. To counteract this, α must decrease
and, at some point, the restriction of the normality condition (α ≥ ϕ) is
reached. As γ has to attain a certain value to deviate from complete key cut
off, the restriction of the normality condition is reached for smaller degrees
of transverse reinforcement as compared to Mechanisms A and B. In practi-
cal cases, the amount of transverse reinforcement provided will be sufficient
to restrict α to be ϕ. For this reason, the assumption of α = ϕ is adopted,
when developing the model. This reduces the problem to only contain γ as
an optimization parameter.
Mechanism C is developed by assuming local key corner crushing of all
n shear keys in one side, see Figure 7.3(a):

P P
t t
l l
Element Element Element Element

at rest in motion at rest in motion
ut ut
γα γα
ul u ul u
γα u
P P
(a) Mechanism C (b) Mechanism E
Figure 7.3: Basic failure mechanisms with key corner crushing
Mechanism C
τC dk 1 − sin ϕ Φ
= + tan (γC + ϕ) (7.10)
νfc 2Lk sin γC cos (γC + ϕ) ν
 
 s cos ϕ 
γC = arctan 
 Φ 2Lk cos ϕ 
(7.11)
 sin ϕ+ 1+

ν dk 1 − sin ϕ

where Lk is the length of the shear key and dk is the depth of the key (see
e.g. Figure 7.1). The mechanism including a diagonal yield line is denoted
Mechanism E, see Figure 7.3(b). Inspired by the simplification introduced
in Mechanism D, the dissipation in the diagonal yield line is also omitted
here. The contribution from the locking bar is included and the capacity
can be calculated by:

Mechanism E
τE n − 1 dk 1 − sin ϕ Φ ΦL
= + tan (γE + ϕ) + (7.12)
νfc n 2Lk sin γE cos (γE + ϕ) ν ν
 
 cos ϕ 
s
γE = arctan  n Φ 2Lk cos ϕ (7.13)
 

 sin ϕ + 1 + 
(n − 1) ν dk 1 − sin ϕ
It should here be noted that if the dissipation in the diagonal yield line
is taken into account, it will no longer be possible to obtain a closed form
solution for the optimal angle γE . The problem would in this case require an
iterative procedure to solve. Results based on solutions with or without dis-
sipation in the diagonal yield line can be seen in Figure 7.4. The calculations
were performed on the basis of the geometry and the material properties of
the test specimens with 6 mm U-bars, see Appendix B.3. The comparison
shows that it is justified to neglect the dissipation in the diagonal yield line,
when considering Mechanism E. This is due to the fact that the direction of
the relative displacements is almost perpendicular to the diagonal yield line.
0.6
0.5
0.4
[-]
0.3
νfc
τ
0.2 Mechanism C
Mechanism E, without
dissipation in diagonal
0.1 Mechanism E, with
dissipation in diagonal
0
0 10 20 30 40
dk [mm]
Figure 7.4: Comparison of results based on Mechanism C and E
For comparison, Mechanism C is also included in the figure. For this

mechanism, it can be seen that in the limiting case of zero key depth, the
mechanism still predicts a capacity. This is due to the assumption of a fixed
angle of the displacement vector, i.e. α = ϕ. The same capacity would in
fact be found by using a shear-friction approach with a friction coefficient
of µ = tan ϕ and by neglecting the cohesion. In the limiting case of no key
depth, Mechanism E results in a greater capacity than Mechanism C due to

7.2 Comparison of Tests with Theory Upper Bound Solutions
the contribution from the locking bar. The transition between Mechanisms
C and E is of course dependent on the geometry and the number of shear
keys.
7.2 Comparison of Tests with Theory

In the following, results based on the presented upper bound solutions will
be compared to the results of the tests presented in Chapter 6. In the
comparisons, the observed failure modes of the specimens are indicated by
a red symbol for local key corner crushing and a black symbol for complete
key cut off. The calculated transition between the two failure modes are
indicated by a dashed vertical line and the regimes are categorized with
colored text. In this way, the observed failure mode can be compared with
the prediction. Appendix C contains tabulated values of the tested first
peak load, PFP , and the calculated capacity, Pcal , for each specimen. The
appendix also contains information on the observed and the predicted failure
of the shear keys (observed and predicted failure mechanism in brackets).
The calculated capacities depend highly on the values adopted for the two
important parameters, namely the internal angle of friction, ϕ, and the
effectiveness factor, ν. The values adopted in Paper II are therefore shortly
repeated below.

As the failure mechanisms are treated as plane strain problems, the choice
of the internal angle of friction, ϕ, is important for the estimation of the
capacity. In the experimental program, both concrete and mortar grouts
were used to grout the connections. For normal strength concretes, a value
of ϕ = 37◦ is generally accepted (Richart et al., 1928; Johansen, 1958; Chen,
2007; Nielsen and Hoang, 2011). According to the results of the triaxial tests
presented in Chapter 3, the internal angle of friction should be reduced when
the maximum grain size of the mixture is smaller than 16 mm. For mortar
(dmax ≤ 4 mm), a value of ϕ = 30◦ can be adopted with reasonable accuracy,
when no additional information from triaxial tests is available. These two
values for the grouts used in the experimental program are adopted in the
upper bound calculations presented in the following.
7.2.2 Effectiveness Factor

The effectiveness factor, adopted in Paper II originates from calibration with
beam shear tests (Zhang, 1997a), and has been modified for use in structural
connections (Jørgensen, 2014; Jørgensen et al., 2017). The expression is sim-

Upper Bound Solutions 7.2 Comparison of Tests with Theory
ilar to the one used for estimation of the tensile capacity of loop connections
(see e.g. Chapter 5):
!
K 1
ν=√ 1+ √ 6> 1.0, (fc in MPa, Lk in m) (7.14)
fc Lk
As for the assessment of the tensile capacity of the loop connections,

the value of K is chosen as 0.88 for a concrete grout and 0.75 for a mortar
grout. To account for size effects, the characteristic length should relate to
a length measured in the direction of the shear load, hence, the length of
the shear key, Lk , is adopted here (while H was used for estimation of the
tensile capacity in Chapter 5).
Equation (7.14) will be used throughout the following comparison with
the available test results. To study their impact, the parameters included in
the formula (fc and Lk ) will in addition be evaluated.
0.8
Key corner Key cut off
crushing
0.6
[-]
0.4
νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 0.1 0.2 0.3 0.4
Φ [-]
Figure 7.5: Comparison of results from tests on the reference specimens with
the upper bound solutions
7.2.3 Reference Tests

In Figure 7.5, the developed upper bound solutions are compared to the test
results for the reference specimens. The calculated capacities are also tab-
ulated in Appendix C.1. The Mechanisms E and D are highly relevant for
the conventional design as a considerable amount of diagonal cracks was ob-
served before reaching the peak load (see e.g. Figure 6.13(a)). Nevertheless,
the comparison in Figure 7.5 shows that Mechanism C without a diagonal
yield line is governing the theoretical capacity. The prediction of the key
failure mode is in accordance with the test observations. In general, a slight

underestimate of the capacity is found. Nevertheless, the results indicate

that the developed solutions can be applied to the conventional design.
0.8
crushing
0.6
[-]
0.4
νfc
τ
0.2 Mechanism A
Mechanism B
Mechanism C
Test Results
0
0 10 20 30 40
dk [mm]
(a) 8 mm U-bars
0.8
crushing
0.6
[-]
0.4
νfc
τ
0.2 Mechanism A
Mechanism B
Mechanism C
Test Results
0
0 10 20 30 40
dk [mm]
(b) 10 mm U-bars
Figure 7.6: Comparison of results from tests on the ’1-on-2’ connection de-
sign with the upper bound solutions
7.2.4 New Design ’1-on-2’ Configuration

Figure 7.6 shows the comparison between the upper bound solutions and
the results of the ’1-on-2’ test series. The test results are only compared to
Mechanisms A, B, and C. The reason is that diagonal cracking was observed
in the tests only at the peak load level (see e.g. Figure 6.14). It appears
that Mechanism A and Mechanism B lead to practically the same results.

Upper Bound Solutions 7.2 Comparison of Tests with Theory
From Figure 7.6, it can be seen that the upper bound solutions slightly
underestimate the results of the specimens with 8 mm U-bars. The test re-
sults for specimens with 10 mm U-bars are predicted with much better agree-
ment. The predicted failure modes correspond well with the experimental
observations. The results of the comparison are also given in Appendix C.2.

Figure 7.7 shows the comparison of the test results for specimens with 6 mm
U-bars with the upper bound model. Generally, it is seen that the solutions
underestimate the capacity of the connections. However, the tendencies of
the test results are captured rather well. This implies that it is possible to
obtain better agreement by choosing a different value for the effectiveness
factor. The comparisons are also given in Appendix C.3.
0.8
crushing
0.6
[-]
0.4
νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 10 20 30 40
dk [mm]
Figure 7.7: Comparison of results from tests on the new ’2-on-2’ connection
design (6 mm U-bars) with the upper bound solutions
Figure 7.8 contains the results of specimens with 8 mm U-bars compared

to the upper bound solutions. The results are identical to the results pre-
sented in Paper II. Appendix C.4 contains the calculated capacities. It can
be seen that the theory fits the test results well and that the predicted failure
modes of the shear keys are predicted relatively well.
The comparison of test results for specimens with 10 mm U-bars with
the results of the upper bound solutions is shown in Figure 7.9. It can be
seen that the capacity of the specimens grouted with mortar is predicted well
(Figure 7.9(b)). However, the capacity of the specimens grouted with con-
crete is overestimated (Figure 7.9(a)). As seen in Appendix B.5, the concrete

grout and the mortar grout had practically the same compression strength.
Furthermore, comparable specimens (identical geometry) had similar first
peak loads regardless of the grout material. Hence, the overestimation for
specimens with concrete grout must primarily be due to the different K-
values adopted in the effectiveness factor, Equation (7.14). This issue will
be discussed further in the Section 7.3.4.
0.8
Key corner
crushing
0.6
[-]
0.4 Key cut off

νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 0.1 0.2 0.3 0.4
Φ [-]
(a)
0.8
crushing
0.6
[-]
0.4
νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 10 20 30 40
dk [mm]
(b)
design (8 mm U-bars) with the upper bound solutions
It can also be seen that the failure mode of the shear keys is not predicted
correctly for all the specimens. The tests showed local key corner crushing
(due to the small key depths) in all specimens and for the smaller key lengths

Upper Bound Solutions 7.3 Evaluation of Model Parameters
the theory predicts failure by complete key cut off. This can be explained
by the model parameters adopted for the local mechanisms, including a size
effect (in terms of the key length, Lk ) incorporated in the effectiveness factor.
This will be further discussed below.
0.8
Key corner
crushing
0.6
[-]
0.4 Key cut off

νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 0.1 0.2 0.3 0.4
Φ [-]
(a) Concrete grout
0.8
Key corner
crushing
0.6
[-]
0.4 Key cut off

νfc
τ
Mechanism A
0.2
Mechanism B
Mechanism D
Mechanism C
Mechanism E
Test Results
0
0 0.1 0.2 0.3 0.4
Φ [-]
(b) Mortar grout
Figure 7.9: Comparison of results from tests on the new ’2-on-2’ loop con-
nection design (10 mm U-bars) with the upper bound solutions
7.3 Evaluation of Model Parameters

From the comparison between the test results and the theoretical upper
bound solutions, some tendencies can be observed. When considering e.g.
the ’2-on-2’ configuration, it can be seen that the upper bound solutions fit

7.3 Evaluation of Model Parameters Upper Bound Solutions
the test results for specimens with 8 mm U-bars very well. On the other
hand, the results for specimens with 6 mm U-bars are underestimated, while
the results for specimens with 10 mm U-bars are overestimated. Especially,
the test results related to concrete grout are overestimated. In the following,
some of the model parameters will be evaluated. The agreement between the
tests and the calculations are presented as a test-to-calculation ratio based
on the recorded first peak capacity, PFP , and the calculated capacity for each
specimen, Pcal . Naturally, the results are dependent on the model parameters
and the choice of effectiveness factor. In the following, the influence of some
of these parameters will be outlined. The results are contained in Appendix
C, where the observed and the calculated local failure modes of the shear
keys (corner crushing or key cut off) also are included.
7.3.1 Reinforcement Degree

In Figure 7.10, all the test-to-calculation ratios are shown versus the rein-
forcement degree. The results are divided into two groups corresponding to
the shear key failure modes observed in the tests.
Overall, it is seen that the correlation between the test results and the
calculated results is good. The mean value of the PFP /Pcal -ratios for all the
tests is 1.02 and the standard deviation amounts to 0.12. When isolating
the test series, it can be seen that the test-to-calculation ratios of the series
are grouped. This is indicated in the small standard deviation for the series
(maximum 0.07 and generally 0.02-0.04, see Appendix C). This proves that
the tendencies of the experimental results are captured well with the upper
bound solutions. However, the magnitude of Pcal can be predicted better for
some test series.
7.3.2 Connection Width

From the comparison presented in Figure 7.10(a), it is tempting to conclude
that the dimension of the U-bars influences the PFP /Pcal -ratio. A smaller
rebar size seems to provide theoretical results that are lower than the test
results (the capacity of specimens with 6 mm U-bars is underestimated,
whereas the capacity of specimens with 10 U-bars is overestimated). This
tendency can, however, also be related to the width of the connection, since
smaller rebar diameters require smaller mandrel diameters and thus smaller
b. Plots of the PFP /Pcal -ratios versus the width, b, are shown in Figure 7.11.
The width, b, is included as a parameter only in Mechanism B. However, as
this mechanism rarely is governing, the parameter cannot be considered as
fully accounted for in the upper bound solutions. The plots in Figure 7.11
clearly show that the PFP /Pcal -ratio decreases as b increases.

1.2
0.8
PF P /Pcal
0.6
0.4 Reference
1-on-2: 8 mm
2-on-2: 6 mm
0.2 2-on-2: 8 mm
2-on-2: 10 mm M
2-on-2: 10 mm C
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Φ [-]
(a) Key corner crushing
1.2
0.8
PF P /Pcal
0.6
0.4 1-on-2: 8 mm
1-on-2: 10 mm
0.2
2-on-2: 6 mm
2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0
0 0.1 0.2 0.3 0.4
Φ [-]
(b) Complete key cut off
Figure 7.10: Ratio between tested and calculated peak loads versus the degree
of transverse reinforcement
Despite the overall decent agreement between the test and the calculated
results, it seems that an even better agreement can be obtained by taking the
connection width into account when establishing the upper bound solutions.
A possibility is to incorporate an empirical relation into the formula for the
effectiveness factor, as b otherwise only will be included in the solutions, if
dissipation in diagonal yield lines is considered. However, the dissipation in
the diagonal yield line is relatively small, confer e.g. Figure 7.4. A greater
connection width entails a larger inclination of the diagonal yield line to
vertical and thus smaller dissipation of energy - or a governing solution that

favors a mechanism without diagonal yield lines, i.e. no dependency of b.
1.2
0.8
PF P /Pcal
0.6
0.4 Reference
1-on-2: 8 mm
2-on-2: 6 mm
0.2 2-on-2: 8 mm
2-on-2: 10 mm M
2-on-2: 10 mm C
0
60 80 100 120 140
b [mm]
(a) Key corner crushing
1.2
0.8
PF P /Pcal
0.6
0.4 1-on-2: 8 mm
1-on-2: 10 mm
0.2
2-on-2: 6 mm
2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0
60 80 100 120 140
b [mm]
(b) Complete key cut off
Figure 7.11: Ratio between tested and calculated peak loads versus the con-
nection width

In the experimental program, two different grouts were used: concrete and
mortar. As discussed above, different values for the internal angle of fric-
tion have been used in the calculations for the two materials. The value
adopted for mortar resulted in good predictions for the specimens grouted
with mortar, whereas the capacity for specimens grouted with concrete were

overestimated by the choice of ϕ. It is in this context worth mentioning that

all specimens grouted with concrete had small key depths. This relatively
small key depth (10 mm) is comparable to the maximum grain size of 11.2
mm in the grout mixture, where 93 % of the aggregates were smaller than
8 mm. Hence, it is practically impossible that the key corners were filled
with the larger aggregates. This means that the observed shear key failure
(local key corner crushing) most likely took place in a part of the matrix
dominated by smaller aggregate particles. As shown in Chapter 3, the value
of the internal angle of friction should be reduced even for mixtures with a
maximum grain size of 8 mm. In the present case, where the failure of the
key corner is highly localized in an area with mostly small aggregates, it can
be argued that the local effective strength here will be similar to the case of
a connection grouted with mortar. Hence, the choice of the internal angle of
friction should reflect this. In the same way, the value of K can be argued
to be reduced as well.
When adopting the parameters for a mortar grout, namely ϕ = 30◦
and K = 0.75, for upper bound analysis of the connections grouted with
concrete, better agreement between the test results and the upper bound
calculations is obtained. Tests of connections with larger key depths (where
failure would take place a complete key cut off) grouted with concrete most
likely possess a higher capacity than similar connections grouted with mor-
tar. Such an investigation was not performed in the current study. Based on
the discussion above, the recommendation is to use ϕ = 30◦ and K = 0.75
in upper bound analysis of keyed shear connections with shallow key depths
grouted with concrete.
7.3.4 Parameters in the Effectiveness Factor

The comparison above indirectly evaluated the effectiveness factor. A more
detailed evaluation of the effectiveness factor can be performed by assessing
the two parameters currently included in the formula for ν to be used in
upper bound solutions. In Equation (7.14), the compressive strength of the
grout, fc , and the length of the shear keys, Lk , are included. The evaluation
of these two parameters should be divided into the two shear key failure
modes. Both parameters are directly incorporated in the equations for the
solutions with complete key cut off, however, from Equation (7.10) and
(7.12), it is found that Pcal in Mechanisms C and E are independent of Lk .
Considering first the case with failure by complete key cut off, the evalu-
ations of fc and Lk are shown in Figure 7.12. In general, a good correlation
is found. When considering the individual test series independently, it also
appears that the influence of fc and Lk is well accounted for by Equation
(7.14).

1.2
0.8
PF P /Pcal
0.6
0.4 1-on-2: 8 mm
1-on-2: 10 mm
0.2
2-on-2: 6 mm
2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0
25 30 35 40 45 50
fc [MPa]
(a) Compressive strength
1.2
0.8
PF P /Pcal
0.6
0.4 1-on-2: 8 mm
1-on-2: 10 mm
0.2
2-on-2: 6 mm
2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0
100 120 140 160 180 200
Lk [mm]
(b) Key length
Figure 7.12: Evaluation of the parameters incorporated in the effectiveness

factor for specimens suffering complete key cut off
For the test specimens experiencing a failure by key corner crushing, it

is debatable to adopt Equation (7.14) as the effectiveness factor. This is
so because the factor K, as discussed above, may overestimate the effect of
concrete versus mortar, and because size effects are accounted for by taking
Lk as a characteristic length. The test-to-calculation ratios versus fc and Lk
are shown in Figure 7.13.

1.2
0.8
PF P /Pcal
0.6
0.4 Reference
1-on-2: 8 mm
2-on-2: 6 mm
0.2 2-on-2: 8 mm
2-on-2: 10 mm M
2-on-2: 10 mm C
0
25 30 35 40 45 50
fc [MPa]
(a) Compressive strength
1.2
0.8
PF P /Pcal
0.6
0.4 Reference
1-on-2: 8 mm
2-on-2: 6 mm
0.2 2-on-2: 8 mm
2-on-2: 10 mm M
2-on-2: 10 mm C
0
100 120 140 160 180 200
Lk [mm]
(b) Key length
Figure 7.13: Evaluation of the parameters incorporated in the effectiveness

factor for specimens suffering local key corner crushing
For key corner crushing, the consideration of Lk as a size effect parameter

is inconsistent with the actual failure, which is inclined and local. This is
e.g. seen in the test results of the ’2-on-2’ specimens with 10 mm U-bars,
which were practically unaffected by an increase in the key length from 120
mm to 180 mm. A characteristic length, if included at all, should rather
reflect the length of the failure plane. This is not easy to interpret for the
local failure as the direction of the failure plane does not coincide with the
direction of loading. The discussion above shows that it may be relevant to
establish a modified ν-expression for the case of local key corner crushing.

7.4 Practical Application Upper Bound Solutions
Nevertheless, the current test results do not seem extensively affected by the
use of Equation (7.14) for prediction of PFP . Only a 7 % change in ν is seen,
when using Lk = 120 mm compared to Lk = 180 mm. The influence of Lk
and fc can in this case not be independently evaluated, as the size effect in
the end also is related to the brittleness.
7.4 Practical Application

The comparison between the test results and the developed upper bound
solutions showed an overall acceptable agreement, both in terms of predicted
first peak capacity and in terms of predicted failure mode. The mechanisms
including diagonal cracking proved that the upper bound solutions can be
adjusted to account for boundary effects in test specimens with a limited
number of shear keys. In the comparison of the calculations with the test
results, the adjusted solutions proved valuable and provided a better fit
for some specimens. However, for shear connections in practice, where the
number of shear keys can be large, e.g. n ≥ 10, the inclusion of the boundary
effect (diagonal yield line at the end of the connection) will most likely not
lead to results that deviate much from the two basic mechanisms without
diagonal yield lines (i.e. Mechanisms A and C). Alternatively, (n − 1) shear
keys can be included in the calculations to account for any boundary effect
instead of a mechanism including a diagonal yield line.
The conclusion must be that for practical applications, where connections
with a larger number of shear keys are to be designed, Mechanism A with
complete key cut off and Mechanism C with local key corner crushing, should
form the basis for the assessment of the structural capacity. In this relation,
the key depth is an important factor for achieving a ductile response after
failure of the shear keys. Smaller key depths favor a failure by local key
corner crushing, which is the most ductile compared to complete key cut off.
Regarding the effect of the grout material, the test results showed that
the use of a concrete grout for smaller key depths (less than the maximum
grain size in the mixture) did not provide a greater capacity than when
using a mortar. As the larger grains of a concrete mix cannot fill in the
shallow keys, where a highly localized failure may take place, the crushing
of the concrete in the corner corresponds to a failure in a mortar and the
same material properties as for a mortar should be assumed (i.e. ϕ = 30◦
as determined in Chapter 3 and K = 0.75).
The comparison between the tests and the calculations showed that the
adopted effectiveness factor, Equation (7.14), provides reasonable results. It
has been pointed out that the influence of the connection width is not com-
pletely accounted for in the solutions. A larger number of tests is required
to establish a relation for the dependency of the connection width. In fact,

Upper Bound Solutions 7.4 Practical Application
only three values of b have been tested in the experimental campaign. For
practical applications, this means that the design should not deviate sig-
nificantly from a tested connection geometry, if upper bound solutions are
used for strength prediction. If the geometry deviates significantly from the
experimental database, a more detailed analysis, than what is provided by
the unsafe nature of the upper bound method, should be adopted. This can
be done e.g. by use of stress fields, which will be a safe solution, where the
specific geometry of the keyed connection can be taken into account in a
more detailed manner (see Chapter 8).

7.5 Concluding Remarks Upper Bound Solutions

In this chapter, the upper bound solutions developed in Paper II have been
summarized and compared to the results of the tests presented in Chapter
6. Some of the test series have also been treated in Paper II. The compar-
ison between the test results and the calculations showed an overall good
agreement, both in terms of the estimated capacity and for the predicted
failure mode. However, it was also discussed that a better agreement may
be obtained, if the influence of the connection width, b, is accounted for,
i.e. included in the effectiveness factor. The findings of the chapter are
summarized as follows:
? The developed upper bound solutions account for the two observed
failure modes of the shear keys - complete key cut off and local key
corner crushing.
? Four different failure mechanisms were presented based on the two

main failure modes of the shear keys. The mechanisms were established
with or without inclusion of a diagonal yield line.
? The solutions can be used both for the conventional design and the
new design (’1-on-2’ and ’2-on-2’ loop connections were considered).
? The effectiveness factor should include correction terms for both ma-
terial brittleness and size effects.
For a failure by complete key cut off, the inclusion of fc and Lk

in Equation (7.14) provide reasonable results.
For a failure by local key corner crushing, the failure plane is
small and the use of Lk to reflect any size effect is debatable.
? The adopted ν-formula is only applicable in upper bound solutions

(which in the case of keyed shear connections are not necessarily the
exact solutions). However, it should be mentioned that the influence
of b is not accounted for in the ν-formula or in all the mechanisms.
Hence the solutions should only be used for practical applications in
cases supported by experimental data.
? For practical applications, Mechanisms A and C provide sufficient in-

formation to estimate the capacity and distinguish between the two
local failure modes of the shear keys.

8 Lower Bound Solutions for
Keyed Shear Connections
In this chapter, the lower bound solutions developed in Paper III will be
presented and compared to the results presented in Chapter 6. The main
purpose of the lower bound solutions is to supplement the upper bound
solutions and thereby provide a more complete theoretical basis for practical
design. The set of solutions can be used to bound the theoretically exact
solution from below and above. In this relation, it should be mentioned that
it remains an open question whether the effectiveness factor in the upper
bound solutions, which is obtained from calibration with tests, in addition
to brittleness effects also accounts for the fact that the model may not be
the exact solution. Establishment of lower bound solutions, based on an
independent choice of ν may provide the necessary information to evaluate
the validity of the models. This discussion will be given in Chapter 9.
Development of optimal lower bound solutions for narrow connections
between precast elements is complicated as the actual stress field is com-
plex. Analytical attempts have been made previously, see e.g. Christof-
fersen (1997) and Nielsen and Hoang (2011), who assumed a simple stress
field consisting of uniaxial struts spanning over one or more pairs of shear
keys. The single-strut solution is shown in Figure 8.1, where it is assumed
that the uniaxial stress field in the strut is transferred to the precast ele-
ment through a nodal zone subjected to plane hydrostatic pressure (hatched
in the figure). The capacity is dependent on the geometry of the connection,
the compressive strength of the grout, and the strength of the transverse
reinforcement (disregarding any external transverse loading). The single-
strut solution often provides conservative capacities, when compared to test
results.
8.1 Development of Lower Bound Solutions

The lower bound solutions developed in Paper III were inspired by the nu-
merical finite element limit analysis method developed by Herfelt et al.
(2016). The numerical approach ensures that the optimal solution for a
given geometry and reinforcement degree is found. This entails that struts
with different inclinations can be combined and meet in nodal zones at the
key corners. As a further improvement to the single-strut model by Christof-

Shear in Joints 665
8.1 Development of Lower Bound Solutions Lower Bound Solutions
(a) (b)
fc
H H
fc
h=H a Δ
a=a
h =2H + Δ a
a=a
FIGURE 8.32Figure 8.1: Stress fields for the single-strut solution, showing (a) a strut over
Strut action between
one sheartwo
and (b) keys.
shear keys, illustrations from Nielsen and Hoang (2011)
fersen (1997), the ability of the concrete-to-grout interface to transfer shear

stresses by friction is utilized in the numerical model. Inclusion of friction
leads to a non-hydrostatic biaxial stress state in the nodal zones. Based on
the two mentioned improvements (as compared to the single-strut solution),
more complex stress fields can be obtained in the joint grout.L
N keys
Figure 8.2 contains examples of optimized stress distributions (obtained
by finite element limit analysis) in a keyed connection subjected to shear.
The analyses were based on the material properties and the geometry of the
test specimens with the ’2-on-2’ loop configuration with 6 mm U-bars (see
e.g. Appendix B.3). In the numerical model, the reinforcement is assumed
continuous across the njoint, =1 which nensures
=2 that the
n = 3full yield force can be
utilized. The results show S = N that an increase
S = N–1 in keyS =depth
N–2 changes the stress
N: number of keys
field and increases the capacity of the connection. For large key depths, the
S: number of struts
optimized stress field appears similar to the single-strut solution, spanning
over two pairs of shear keys. However, the capacity is increased as friction
FIGURE 8.33
Notation for
is different
includedstrut paths.
in the indented parts of the interface. More detailed examples of
modeling with finite element limit analysis (FELA) can be found in Herfelt
h H (8.72)
(2017). tan A min
a
where h is the active strut depth for the case considered. The corresponding value of
yoyomax is found to be
y o max H cos 2 A min (8.73)
Lower Bound Solutions 8.1 Development of Lower Bound Solutions
(a) dk = 10 mm, PFP = 340.5kN
(b) dk = 14 mm, PFP = 404.1kN
(c) dk = 20 mm, PFP = 486.1kN
Figure 8.2: Examples of optimized distributions of the maximum compressive

stress [MPa] (tension positive), illustrations from Paper III
8.1.1 Inclusion of Friction in Interface

For the friction between the grout and the precast concrete, a Coulomb
friction criterion is assumed with a friction coefficient and a neglectable co-
hesion. The main argument for neglecting the cohesion is that smooth form-
work was used to cast the reinforced concrete elements in the experimental
campaign (smooth formwork is commonly used also in practice). Thus, any
small cohesive resistance in the interface may have (partly or completely)
vanished at the ultimate limit state.
Friction is only considered active in the indented areas of the interface.
The main argument for this is that the normal stress required to activate
friction stems from tension in the transverse reinforcement, which represents
a form of passive confinement, in contrast to active confinement from an
external normal force. Since the load is anticipated to be carried mainly by
strut action between the shear keys, only the indented areas can transfer
friction, as these are the only areas which experience compression.
For interfaces cast against smooth formwork, Eriksson (1978) modeled
the sliding along the inclined interface of a key corner with a coefficient of
friction of µ = 0.7. Cholewicki (1971) also considered sliding in the inclined
part of the key with µ = 0.7. In Paper III, µ was taken as 0.75 for the
smooth and untreated interfaces (relevant for the specimens with ’2-on-2’

loop configurations). This value was also adopted in Nielsen and Hoang
(2011) and Herfelt et al. (2016). In Paper IV, a discussion is given for
the value of µ to be adopted for a greased interface. For consistency, the
same value of µ = 0.3 is adopted for the comparison between test results of
specimens with greased interfaces (reference specimens and specimens with
’1-on-2’ loop configurations) and the lower bound solutions.
σA,2
Lk Lk
σA,1
σB
s s
σA,1 σA,2
l l
σB
t t
σA,1 σA,2
b b
(a) Solution 1 (b) Solution 2
Figure 8.3: The two stress fields considered, illustrations from Paper III
8.1.2 Development of Stress Fields

In Paper III, two analytical lower bound solutions were developed. These
are illustrated in Figure 8.3 for a connection with n = 3 shear keys. Solution
1 considers a distribution of struts with identical inclination, spanning over
one pair of shear keys. This solution appears similar to the single-strut
solution, however, the inclusion of friction in the interface allows for a smaller
inclination of the struts as the stress state in the key corners does not have
to be plane hydrostatic. The difference between the current solutions and
the single-strut solutions by Christoffersen (1997) is particularly seen in

the limiting case of zero key depth (dk = 0), where the single-strut solution
predicts no capacity, while the current solutions predict a capacity stemming
solely from friction.
A large number of combinations of uniaxial struts can be considered
when developing lower bound solutions. The most simple solution is based
on a combination of struts, which span over one or two shear keys, see
Figure 8.3(b). This combination is denoted Solution 2. The stress states in
the struts are uniaxial and the nodal zones in the key corners are stressed in
biaxial compression (hatched area). At the outermost keys of the connection,
the stress transfer is asymmetric in a pair of shear keys, as no strut is assumed
to develop towards the end of the connection. In this relation, it should be
mentioned that any potential locking bar is disregarded in the modeling.
Figure 8.4 shows the stress fields in the key corners for the two solu-
tions. In Solution 1, the nodal zones are identical at both ends of the strut
(Triangle I) due to symmetry. In Solution 2, three triangular zones can be
identified. In these zones, the stress is transferred from the struts to the pre-
cast elements (see Paper III for details). A main feature of the stress field in
Solution 2 is that the grout is stressed in the full length of the shear keys (see
the right shear key in Figure 8.4(b)). The stress transfer in the boundary
key (left shear key in Figure 8.4(b)) is identical to the stress transfer in the
key corners of Solution 1.
θk θk
e1 e1
I II
a e e
III
Lk Lk
e2
σA,1 θA,1 θA,2
e σA,2
a
I I
θB σB
dk dk dk b dk
(a) Solution 1 (b) Solution 2
Figure 8.4: Geometry of key corners and identification of the nodal zones I,
II, and III in the key corners of the two solutions, illustrations from Paper
III

For both solutions, the length e is an optimization parameter and through

this value, the optimal admissible stress field in the joint can be established.
The optimal value of e ensures that the stress field fulfills the failure criteria
of the grout and of the interface. The stresses in the key corners are modeled
by the concept of homogeneous stress fields, see e.g. Nielsen (1969), Muttoni
et al. (1997), and Pistoljevic and Nielsen (2009).
8.1.3 Failure Criteria

The stress field in the grout is limited by the strength of the grout. Failure
criteria for uniaxial stress fields, for biaxial stress fields, and for stress trans-
fer in the interface are considered. For the uniaxial compression (i.e. struts
between the shear keys), the stress cannot exceed the effective strength:
σA,i ≤ νfc , ν ≤ 1, i = 1, 2 (8.1)

σB ≤ νfc , ν≤1 (8.2)
where ν is the effectiveness factor and i refers to Solution 1 or 2 (the stress σB
is not present in Solution 1). It is in this context important to mention that
the ν-factor to be adopted does not have to be the one used in Chapter 7 for
the upper bound solutions, since the upper bounds, as already mentioned,
are not necessarily exact solutions. In fact, to keep the lower bound results
independent of the upper bound results, the adopted ν-factor should not
be affected by any previous calibration with upper bound solutions. In
the following, the value of ν will be taken as unity, when comparing the
analytical lower bound model with numerical calculations. For comparison
with test results, the value will be calculated by the following formula, which
is proposed in the fib Model Code (fib, 2013) and primarily aimed at stress
field modeling of shear related problems:
!1/3
fc,0
ν= 6> 1.0, (fc in MPa) (8.3)
fc
where fc,0 = 30 MPa. Originally, the formula was suggested by Muttoni
(1990) using fc,0 = 20 MPa and later it was modified to the current form.
The formula is considered as a qualified estimate of the effect of material
brittleness. As can be seen, the formula differs slightly from the inverse
square root dependency of fc , which appears in the ν-formula determined
by calibration of upper bound solutions with test results. Further discussion
on the ν-factor to be used in lower bound modeling can be found in Paper
III.
In the nodal zones, the grout is stressed in biaxial compression. As shown
by e.g. Kupfer et al. (1969), the strength of concrete loaded in biaxial com-
pression is greater than when loaded in uniaxial compression. The following

criteria for the stress state in the nodal zones (in terms of principal stresses,
compression positive) are adopted:
σ2 ≤ cfc , c≥1 (8.4)

σ1 ≥ 0 (8.5)
where c can be understood as a strength enhancement factor. The actual

value of c is dependent on the stress state. For simplicity, a constant value
of c can be adopted. The same approach is suggested in the fib Model Code,
where an increase in strength of 10 % is allowed in regions, where significant
biaxial compression exists. In the nodal zones ν can be taken as unity, also
in practice.
The last criterion relates to the frictional resistance in the interface. As
already mentioned, a Coulomb criterion is adopted for the interface:
|τnt | ≤ µσn (8.6)
where τnt is the shear stress and σn is the normal stress acting on a interface.
The criterion applies to the vertical and the inclined parts of the keyed
interface, see e.g. Figure 8.5(a) for distribution of stresses on the boundaries
of Triangle I. In Figure 8.5(b), the stress distributions are expressed in
terms of stress resultants.
σn,P
τP Pl,1
e1 Pt,1
e1
I
I
Cl,1
σn,A σn,C a
At,1 a
Ct,1
l Al,1
t τA τC
dk dk
(a) (b)
Figure 8.5: Homogeneous stress field in the nodal zone of Solution 1, (a)
Stresses along boundaries of Triangle I and (b) resultants of stresses on
boundaries, illustrations from Paper III

The criteria for the interfaces can be described in terms of stress resul-
tants:
|Cl,i | ≤ µCt,i , i = 1, 2 (8.7)

P sin θ − P cos θ
l,i k t,i k
≤ µ, i = 1, 2 (8.8)
Pt,i sin θk + Pl,i cos θk
where Equation (8.7) is valid for the vertical part of the shear key and
Equation (8.8) applies to the inclined part of the shear key. The angle θk
is the inclination of the key corner to horizontal, see Figure 8.4. It can
be mentioned that when choosing traditional values of θk (in this relation
meaning values less than 30◦ ) and normal values of µ (≈ 0.7), the friction
criterion of the inclined part of the key will not be critical. The inclination
of the key corner does not affect the results, as long as it is relatively small
(θk < 30◦ ). However, for practical applications, an inclination larger than
0◦ should be used to ensure adequate grouting of the key corners. A value
of θk = 30◦ is adopted e.g. in Eurocode 2 (CEN, 2004) as a maximum limit
to prevent sliding at the key corner.
In relation to force transfer by friction, it should be noted that when
the optimized value of e entails an inclination of Strut A that fulfills the
condition:
tan θA,i ≤ µ, i = 1, 2 (8.9)
the stress from Strut A, σA , can be transferred by friction in the keyed
interface of Triangle I and the stress state in Triangle I can be considered
as uniaxial (Equation (8.1) then applies for the stress state in the nodal
zone).
In addition to the failure criteria for the grout, the tensile stresses carried
by the U-bars cannot exceed the yield limit:
σs ≤ f y (8.10)
As the transverse loop connections should be designed to enable tensile

yielding of the reinforcement (to avoid a brittle failure of the connection
grout, see Chapter 5), this requirement (Equation (8.10)) ensures that the
transverse capacity of the connection is not exceeded. If the full yield force
cannot be activated, fy should in Equation (8.10) be replaced by the rein-
forcement stress that corresponds to the tensile capacity of the connection
related to grout failure. By use of force equilibrium, the requirement in
Equation (8.10) can be reformulated as a constrain on the stress in the
struts, see Paper III for details.

Lower Bound Solutions 8.2 Comparison to Finite Element Limit Analysis
8.1.4 Optimization of Analytical Solutions

As described in Paper III, two independent parameters are available for
optimization of the solution. As the statical optimization parameter, the re-
sultant Cl,i (which represents how much vertical force that can be transferred
in the vertical interface, see e.g. Figure 8.5(b)) is chosen. An increased stress
transfer leads to increased shear capacity and for this reason, the longitudi-
nal stress transfer is chosen as large as possible. As the stress is transferred
by friction, this means:
Cl,i = µCt,i , i = 1, 2 (8.11)
The only parameter left for optimization of the stress field is hereafter
e. By varying the value of e, the stress state in the connection can be
established and checked against the failure criteria.
8.2 Comparison to Finite Element Limit

Analysis
The developed analytical lower bound solutions can be benchmarked against
the numerical counterpart. Completely identical solutions are not expected
as the numerical optimization operates with a much larger number of ad-
missible stress fields.
Figure 8.6 shows a comparison between the numerical and analytical
solutions. The calculations are performed using a geometry and a reinforce-
ment degree, which are identical to the tests on the ’2-on-2’ shear connection
design reinforced with 6 mm U-bars, see e.g. Appendix B.3. In the calcu-
lations, the coefficient of friction was taken as µ = 0.75. The effectiveness
factor and the strength enhancement factor for the mortar were taken as
unity, i.e. ν = c = 1. The comparison shows an almost perfect correlation
between the numerical and analytical limit analysis. A small deviation is
seen around the transition from Solution 1 to Solution 2, however, the over-
all agreement is excellent. For comparison, the results of the single-strut
solution (Christoffersen, 1997) are also included. It can be seen that the
developed lower bound solutions provide larger load carrying capacity than
the single-strut model.
It should be noted that a benefit of the analytical solution compared to
the current numerical solution is that the different zones in the joint grout
can be assigned different strengths, depending on the stress state. In the
numerical optimization, the same failure criteria must be assumed for all
zones of the grout. This is the reason for choosing ν = 1 and c = 1 when
comparing the models. In the following comparisons, the nominal shear

8.2 Comparison to Finite Element Limit Analysis Lower Bound Solutions
stress, τ , has been determined as in Equation 7.1. For consistency, P is

adopted for the shear load in the lower bound calculations as it was in the
upper bound solutions. The notation deviates from Paper III, where V was
used for the shear force, however, the symbols express the same load. For
the degree of transverse reinforcement, Φ is calculated as in Equation (7.2).
0.25
Solution 2
0.2
↓
0.15 ↑
τ /fc [-]
Single-strut
0.1
↑Solution 1
0.05
Analytical Lower Bound
Single-strut Solution
Numerical Results (FELA)
0
0 5 10 15 20 25 30 35 40
dk [mm]
Figure 8.6: Comparison between numerical and analytical lower bound mod-
eling, illustration from Paper III

Lower Bound Solutions 8.3 Comparison of Tests with Theory
8.3 Comparison of Tests with Theory

In the following, the analytical lower bound solutions will be compared to
the test results presented in Chapter 6. In the comparison, c = 1.15 and
ν = 1 are adopted for the local strength in the key corners (similar to how
compression nodes in strut-and-tie models are treated) and the effectiveness
factor for the compression struts is calculated by Equation (8.3). The coef-
ficients of friction are chosen as discussed in Section 8.1.1. In the figures, an
experimentally observed key failure by local key corner crushing is indicated
with a red symbol while failure by complete cut off is indicated with a black
symbol. The calculated capacities are tabulated in Appendix D along with
the tested first peak capacities. In the appendix, information about which
stress component that is governing is also included.
8.3.1 Reference Tests

Figure 8.7 shows the comparison between the two analytical solutions and
the reference tests (conventional design). The theoretical results are gener-
ated by a variation of fc . The lower bound model predicts capacities that
are lower than the test results. From the calculations (see also Table D.1),
it is found that Solution 1 is governing and that the maximum compressive
stress in Triangle I is the limiting stress component. The underestimation
of the capacity (an average of 30 % underestimation) would be smaller, if a
larger value of µ was used for the greased interface.
0.4
0.3
τ /fc [-]
0.2
0.1 Solution 1
Solution 2
Single-strut
Test Results
0
0.05 0.1 0.15 0.2 0.25
Φ [-]
Figure 8.7: Comparison of results from tests on the reference specimens with
the lower bound solutions (µ = 0.3)

8.3 Comparison of Tests with Theory Lower Bound Solutions

In Figure 8.8, the lower bound solutions are compared to the test results of
the new design with a ’1-on-2’ loop configuration. The plots show capacity
versus key depth. It can be seen that Solution 1 is governing for smaller
key depths, whereas Solution 2 becomes governing for larger key depths.
This is as expected. The model captures the test results reasonable well.
The calculations (see Table D.2) show that different stress components are
governing for the theoretical prediction of the capacity.
0.4
0.3
τ /fc [-]
0.2
0.1 Solution 1
Solution 2
Single-strut
Test Results
0
0 10 20 30 40
dk [mm]
(a) 8 mm U-bars
0.4
0.3
τ /fc [-]
0.2
0.1 Solution 1
Solution 2
Single-strut
Test Results
0
0 10 20 30 40
dk [mm]
(b) 10 mm U-bars
Figure 8.8: Comparison of results from tests on the ’1-on-2’ connection de-
sign with the lower bound solutions (µ = 0.3)


The comparison of lower bound calculations with test results for the new
design with 6 mm U-bars are included in Paper III and repeated in Figure
8.9. Generally, a good fit between tests and theory is found. It is seen that
the capacity is predicted by Solution 2 for all tests. It should be mentioned
that for a key depth of approximately 16 mm, the lower bound solutions
predict a change in the governing stress component. In the tests, both
failure modes (corner crushing and complete key cut off) were observed for
this key depth. However, the change in failure mode cannot be directly
interpreted from the change in the governing stress component.
0.4
Solution 1
Solution 2
Single-strut
0.3 Test Results
τ /fc [-]
0.2
0.1
0
0 10 20 30 40
dk [mm]
design (6 mm U-bars) with the lower bound solutions (µ = 0.75)
In Figure 8.10, the test results of the new design reinforced with 8 mm
U-bars are compared to the lower bound model. For the series with varying
key length (Specimens I-IV, see Table B.4), the results are generated by
a variation of the key length. The remaining specimens had different key
depths (Specimens V-IX, see Table B.5), which is why the capacity for these
tests has been plotted as a function of dk . In general, the lower bound model
agrees reasonable well with the test results. For the series with varying key
length, the correlation is rather good and only a noticeable deviation is seen
for the smallest key length (120 mm). The reason for the apparent decrease
in capacity with increasing key length is due to the fact that the area of the
shear key is used in the definition of the shear stress, τ . A greater key length
results in a steeper inclination of the struts and thereby a higher capacity. In
Appendix D.4 increasing values of Pcal is reported for increasing key lengths.

8.3 Comparison of Tests with Theory Lower Bound Solutions
For the series with varying key depths, it is noted that some of the test
results with larger key depths (Figure 8.10(b)) fall below the lower bound
prediction. This is not expected from the extremum theorems of plastic
theory. However, this may be due to the adopted effectiveness factor. For
a compression strength around 30 MPa (the strength of the grout was 30.6
MPa for the test specimens with large key depths), Equation (8.3) predicts
practically no reduction of the compressive strength due to brittleness. In
reality, the average plastic strength of the mortar should most likely be
reduced, even for fc lower than 30 MPa.
0.4
0.3
τ /fc [-]
0.2
0.1 Solution 1
Solution 2
Single-strut
Test Results
0
100 120 140 160 180 200
Lk [mm]
(a) 8 mm U-bars, varying key length
0.4
0.3
τ /fc [-]
0.2
0.1 Solution 1
Solution 2
Single-strut
Test Results
0
0 10 20 30 40
dk [mm]
(b) 8 mm U-bars, varying key depth

For the ’2-on-2’ test series with 10 mm U-bars, both concrete and mortar
were used as grouts. As the same relations for the effectiveness factor, ν, and
the strength enhancement factor, c, are adopted for the two materials, the
calculations predict the same capacity for the two grouts. The comparison
can be seen in Figure 8.11. It is found that for all specimens, the capacity
is governed by Solution 1, i.e. struts spanning over one pair of shear keys.
This is due to the small key depth used in the tests. Furthermore, it is
seen that the calculations predict the test results better for the longer key
lengths. Again, this is a combination of the small key depth and the length
of the keys (inclination of the struts).
As a concluding remark on the comparison of the lower bound calcula-
tions with test results, it is noted that the developed lower bound model in
all cases provides results that fit the test results better than the single-strut
solution. Higher capacities can be obtained by inclusion of the locking bar
in the model. This extension to the model has not been incorporated in the
current study.
0.4
Solution 1
Solution 2
Single-strut
0.3 Tests Mortar
Tests Concrete
τ /fc [-]
0.2
0.1
0
100 120 140 160 180 200
Lk [mm]

8.4 Evaluation of Model Parameters Lower Bound Solutions
8.4 Evaluation of Model Parameters

Based on the comparisons between calculations and test results, the per-
formance of the lower bound solutions can be evaluated. Figure 8.12(a)
shows the test-to-calculation ratios versus the reinforcement degree. It can
be seen that most of the capacities are predicted below the tested capacity,
which is to be expected from a lower bound model. The mean value of the
PFP /Pcal -ratio for all the tests is 1.17 and the standard deviation amounts
to 0.18.
1.6
1.4
1.2
PF P /Pcal
1
0.8 Reference
1-on-2: 8 mm
0.6 1-on-2: 10 mm
2-on-2: 6 mm
0.4 2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0.2 2-on-2: 10 mm M
2-on-2: 10 mm C
0
0 0.1 0.2 0.3 0.4
Φ [-]
(a)
1.6
1.4
1.2
PF P /Pcal
1
0.8 Reference
1-on-2: 8 mm
0.6 1-on-2: 10 mm
2-on-2: 6 mm
0.4 2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0.2 2-on-2: 10 mm M
2-on-2: 10 mm C
0
25 30 35 40 45 50
fc [MPa]
(b)
Figure 8.12: Ratio between tested and calculated peak loads versus (a) the
reinforcement degree, Φ, and (b) the compressive strength, fc

Lower Bound Solutions 8.4 Evaluation of Model Parameters
In Figure 8.12(b), the correlation can be seen in a PFP /Pcal versus fc

plot. This comparison is relevant as the effectiveness factor for each test is
calculated based on the tested compression strength. The values of ν varied
between 0.86 and 0.99. From a lower bound point of view, the solutions
compare reasonably well with the test results. It appears that the adopted
values of ν and c do not lead to overestimation of the capacity. It should be
noted that the predicted capacity is not always limited by the same failure
criterion and therefore, it is not possible to evaluate the parameters directly
when comparing to all the test results. Nevertheless, the combination of the
two parameters dictates which of stress components that is governing, thus
the overall comparison constitutes a basis for evaluation of the combination
of the two parameters.
8.4.1 Key Depth and Width of Connection

In Figure 8.13, the influence of two essential geometric parameters is evalu-
ated. As mentioned earlier, the test series were designed with a particular
interest in the key depth. It is therefore of interest to study how the lower
bound model agrees with tests as dk varies. This is shown in Figure 8.13(a).
A general tendency which is valid for all test series cannot be observed. For
some series, it seems that the theory overestimates the capacity more for
smaller key depths than for the larger key depths. This can probably be ex-
plained partly by the fact that the simple stress field assumed in the model
consists of narrow struts, when the key depth is small. Where the stress is
transferred to the elements, the real stress state locally in the key corner
might be described better by a triaxial stress state than the biaxial stress
state assumed in the solutions. As the key depth increases, the width of the
struts increases as well (see e.g. Figure 8.2). In this case, the stress con-
centrations may lead to a smaller strength increase than for the smaller key
depths. A more accurate failure criterion for the stress state might capture
the results better for all key depths, however, it is not possible to capture
effects of a triaxial stress state by the simple relation using a constant c
factor.
As shown in Chapter 7, the connection width, b, affected the agreement
between tests and the upper bound solutions, most probably because b is
not properly accounted for in the model. In the lower bound solutions on
the other hand, b plays an important role for the stress field. The influence
of b on the calculated capacity should therefore be reflected better in the
lower bound solutions. This seems in fact to be the case, when comparing
the plot in Figure 8.13(b) with its counterpart in Figure 7.11. The strong
declining tendency in Figure 7.11 is not observed in Figure 8.13(b), which
suggests that the mean value of the PFP /Pcal -ratios, for fixed values of b is

8.4 Evaluation of Model Parameters Lower Bound Solutions
less affected by b, when using the lower bound solutions.
1.6
1.4
1.2
PF P /Pcal
1
0.8 Reference
1-on-2: 8 mm
0.6 1-on-2: 10 mm
2-on-2: 6 mm
0.4 2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0.2 2-on-2: 10 mm M
2-on-2: 10 mm C
0
0 10 20 30 40
dk [mm]
(a)
1.6
1.4
1.2
PF P /Pcal
1
0.8 Reference
1-on-2: 8 mm
0.6 1-on-2: 10 mm
2-on-2: 6 mm
0.4 2-on-2: 8 mm (Lk )
2-on-2: 8 mm (dk )
0.2 2-on-2: 10 mm M
2-on-2: 10 mm C
0
60 80 100 120 140
b [mm]
(b)
Figure 8.13: Ratio between tested and calculated peak loads versus (a) key
depth, dk , and (b) connection width, b

Lower Bound Solutions 8.5 Concluding Remarks

In this chapter, the lower bound model developed in Paper III was compared
to the test results presented in Chapter 6. One of the test series has also been
treated in Paper III. The developed model leads to two solutions: Solution
1 with struts spanning over one pair of shear keys and Solution 2 with a
combination of struts with different inclinations. The local stress states in
the key corners were assessed by the concept of homogeneous stress fields.
A number of other solutions can be developed, however, these two solutions
captured the tendencies of the experimental results reasonable well. The
main results and findings of the chapter are summarized as follows:
? Inclusion of interface friction increases the calculated shear capacity

of the keyed connections compared to the previously developed single-
strut solution.
? Introduction of local strength parameters allows for assessment of dif-

ferent stress states in the grout:
For the uniaxial compression struts, an effective strength of νfc

was adopted. The ν-factor of the model primarily accounts for
material brittleness.
For the nodal zones stressed in biaxial compression, an enhance-
ment factor, c, was introduced to allow for an increase of the
maximum compressive principal stress.
? The influence of the connection width on the calculated capacity is

better accounted for by the lower bound solutions than its upper bound
counterparts.

9 Comparison of Upper and
Lower Bound Solutions
In the previous chapters, upper and lower bound solutions for the ultimate
capacity of shear connections have been presented and compared to the test
results, presented in Chapter 6. One of the main purposes for developing
both types of models is to narrow the range of expected test results. Identi-
cal upper and lower bound solutions are not possible to obtain as the basic
assumptions adopted for development of the solutions are not identical. For
the upper bound model, plane strain conditions are usually assumed for anal-
ysis of the failure mechanism in the narrow connection zone. For the lower
bound solutions, on the other hand, plane stress conditions are assumed.
Nevertheless, the comparison of the two methods provides information on
the expected capacity of a keyed shear connection.
In the following, the upper as well as the lower bound model will be
compared to the test results. In the lower bound calculations, the strength
parameters adopted in Chapter 8 will be used. This includes the coefficients
of friction (µ = 0.3 or 0.75) as well as the strength enhancement factor
c = 1.15. For the upper bound solutions, the same effectiveness factor as
used in the lower bound calculations will be adopted (Equation (8.3)). This
value of ν accounts only for material brittleness and hence, the influence of
the model assumptions (plane strain versus plane stress) can be visualized
while at the same time, any effect related to calibration with upper bound
solutions is avoided.
9.1 Expected Range of Results

Figure 9.1 shows the calculated upper and lower bound results compared
to the test results for the reference specimens (conventional design). The
expected range of results is shaded in grey and bounded by the upper bound
solutions from above and by the lower bound solutions from below. It is
found that the test results lie well within the expected range. The difference
between the upper and lower bound increases with increased reinforcement
degree. Only the result of Mechanism C is included in the graph as this
mechanism was governing in the upper bound calculations. The failure mode
observed in tests also reflected this mechanism (local key corner crushing,
indicated with red symbols).

9.1 Expected Range of Results Upper and Lower Bounds
0.4
C↓
0.3
τ /fc [-] 0.2
0.1 Upper Bound

Lower Bound 1
Lower Bound 2
Test Results
0
0.05 0.1 0.15 0.2 0.25
Φ [-]
Figure 9.1: Expected range of results predicted by the upper and lower bound
solutions, compared with test results of reference specimens (conventional
design)
The plots of the expected range of results for the tests with a ’1-on-2’
loop configuration are shown in Figure 9.2. In this case, the governing upper
bound solutions correspond to Mechanism B (key cut off, indicated with
black color) and Mechanism C (local key corner crushing, indicated with
red color). It is found that the test results for both dimensions of U-bars
fall within the predictions of the upper and lower bound solutions. It can
be mentioned that a greater range is predicted for calculations with 10 mm
U-bars (Figure 9.2(b)). The failure modes predicted by the upper bound
solutions for specimens reinforced with 8 mm U-bars reflect the observed
failure modes, when using the chosen ν-factor. The transition between the
two failure modes (according to the upper bound solutions) is indicated
in the figures. For the specimens reinforced with 10 mm U-bars, correct
prediction of the key failure mode is only obtained for two tests.
Figure 9.3 shows the expected range of results together with test results
for the ’2-on-2’ configuration with 6 mm U-bars. The results of Mechanism
D represent the capacity related to complete key cut off and both Mechanism
C (solid line) and Mechanism E (dashed line) are included for the failure by
key corner crushing. The models provide a very narrow range of expected
results, within which the test results are actually found.

Upper and Lower Bounds 9.1 Expected Range of Results
0.4
crushing
0.3
B↓
τ /fc [-] 0.2

C↓
0.1 Upper Bound

Lower Bound 1
Lower Bound 2
Test Results
0
0 10 20 30 40
dk [mm]
(a) 8 mm U-bars
0.4
Key corner
crushing Key cut off
0.3 B↑
τ /fc [-]
C↓
0.2
0.1 Upper Bound

Lower Bound 1
Lower Bound 2
Test Results
0
0 10 20 30 40
dk [mm]
(b) 10 mm U-bars
solutions, compared with test results of the ’1-on-2’ loop configuration
It should be noted that for very small key depths, the lower bound model
(Solution 1) provides capacities that are greater than the upper bound so-
lution (Mechanism C with local crushing of the key corners). This is incon-
sistent with the extremum theorems of plasticity theory. To obtain results
consistent with the extremum theorems, the upper bound calculations should
be modified to include the friction criterion for the interface and at the same
time fulfill the normality condition at the interface. This would change the
lower limit for α from ϕ = 30◦ (adopted for a mortar) to ϕ = arctan µ = 37◦
(µ = 0.75). There will be no extra dissipation of energy in the upper bound

9.1 Expected Range of Results Upper and Lower Bounds
solution as the dissipation along the interface is zero, when the normality
condition is imposed on a friction criterion without cohesion. In this case,
the upper and lower bound solutions will be identical for dk = 0. This is not
relevant for practical applications, however, consistent with the extremum
theorems.
0.4
crushing
0.3
τ /fc [-]
D↓
0.2 C↓
↓E
0.1 Upper Bound
Lower Bound 1
Lower Bound 2
Test Results
0
0 10 20 30 40
dk [mm]
solutions, compared with the test results of the ’2-on-2’ loop configuration
with 6 mm U-bars
Figure 9.4 contains the expected range of results for the tests on the ’2-
on-2’ loop configuration with 8 mm U-bars. For the tests with varying key
length, both the upper and the lower bound model captures the observed
tendency well, i.e. increasing τ /fc for decreasing Lk . The reason for the ap-
parent decrease in τ is explained in Chapter 8. For the series with varying
key depth, the models provide a reasonable range of expected results. How-
ever, the test results for specimens with larger key depths lie below the lower
bound prediction. A smaller value of ν would provide a better correlation
for both the upper and lower bound model. This was also discussed for the
lower bound model in Chapter 8.
Figure 9.5 shows the expected range of results for the tests on the ’2-on-2’
loop configuration with 10 mm U-bars. Only the mechanisms corresponding
to local crushing of the key corners are included, as this was the observed
failure mode in the tests. The internal angle of friction adopted in the upper
bound calculations is the one suggested for mortar, i.e. ϕ = 30◦ . It can be
seen that the test results lie well within the range of model predictions. It
is found that the difference between the upper and the lower bound models
increases with decreasing key length. The upper bound model predicts the

Upper and Lower Bounds 9.1 Expected Range of Results
experimental tendency better than the lower bound model. On the other
hand, the lower bound model provides a better fit when the key length is
large. This is due to the inclinations of the struts, which are steeper for
larger key lengths and thereby a larger capacity can be predicted.
0.5
0.4 ↓D
τ /fc [-]
0.3
0.2
Upper Bound
0.1 Lower Bound 1
Lower Bound 2
Test Results
0
100 120 140 160 180 200
Lk [mm]
(a) 8 mm U-bars, varying key length
0.5
crushing
0.4
D↓
τ /fc [-]
0.3
C↓
0.2 ↓E
Upper Bound
0.1 Lower Bound 1
Lower Bound 2
Test Results
0
0 10 20 30 40
dk [mm]
(b) 8 mm U-bars, varying key depth
solutions, compared with test results of the ’2-on-2’ loop configuration with
8 mm U-bars

9.2 Evaluation of the Models Upper and Lower Bounds
0.4
Upper Bound
↓C Lower Bound 1
E Lower Bound 2
0.3 ↓ Tests Mortar
Tests Concrete
τ /fc [-] 0.2
0.1
0
100 120 140 160 180 200
Lk [mm]
solutions, compared with test results of the ’2-on-2’ loop configuration with
10 mm U-bars
9.2 Evaluation of the Models

The models have been used to determine the range of expected results
bounded from above by the upper bound solutions and from below by the
lower bound solutions. The calculations were based on the best estimate of
model parameters (including the chosen value of ν) and showed that neither
the upper bound solutions, nor the lower bound solutions, can be regarded
as the exact solution for the problem considered. Nevertheless, for the rele-
vant geometrical parameters, the upper and lower bound models display the
same tendencies, i.e. increase or decrease in capacity by a variation of the
investigated parameter. This is for example seen in the calculations, where
the key depth varies, see e.g. Figure 9.3. At some characteristic key depth,
the upper bound model predicts a transition from local key corner crushing
to complete key cut off (no further increase in capacity from an increase in
key depth). The lower bound model also predicts a change in capacity with
increasing key depth (the governing stress component changes). However,
the failure mode related to the governing stress state cannot be determined
from the lower bound model. Nevertheless, the calculations based on the
two approaches do not contradict each other.
In this relation, it should be noted that the upper bound model consider-
ing local key corner crushing and the detailed lower bound model including
friction and accounting for the local stress states (with different strengths)
constitute models that provide better predictions than the previously devel-

Upper and Lower Bounds 9.3 Practical Relevance
oped models. In the previous models only complete key cut off and single-
strut action were considered. Benchmarked against the previous models,
the developed models constitute an improvement, not only in relation to
the prediction of the capacity, but also regarding estimation of the failure
mode of the shear keys. Relating to the test results presented in Chapter
6, an accurate prediction of the failure mode is beneficial when a ductile
load-displacement relationship is required.
9.3 Practical Relevance

The practical relevance of the upper and lower bound models can be dis-
cussed based on the conducted comparison between model calculations and
test results. The primary merit of the upper bound model lies in the simplic-
ity (closed-form equations). However, the effectiveness factor to be adopted,
i.e. Equation (7.14), must be calibrated with tests in order to compensate
for the fact that the solutions are not exact and thereby unsafe. The lower
bound solutions, on the other hand, provide safe results, but are more com-
plicated to develop. The most accurate stress condition in a shear connection
is most likely somewhere in between the assumptions adopted for the two
approaches (i.e. plane strain versus plane stress). It might be too optimistic
to expect that a solution closer to the exact one can be developed and at
the same time appears as user-friendly as the closed-form upper bound so-
lutions. Therefore the upper bound solutions will still have preference from
a practical point of view, even though they have to be used in conjunction
with an effectiveness factor that not only accounts for the material behavior,
but also functions as an adjustment parameter to compensate for the un-
safe nature of upper bound solutions. Awareness of this is important when
applying the upper bound solutions to practical cases, which deviate signif-
icantly from the experimentally tested configurations. This also means that
the lower bound model should be applied in situations, where the design of
keyed connections deviates significantly from tested designs.
In relation to the practical application of the upper bound solutions,
realistic material properties should also be used in conjunction with the
proper effectiveness factor. The results of the triaxial tests presented in
Chapter 3 proved that mortar does not possess the same material properties
as concrete. This is especially evident, when comparing the values of ϕ
indicated in the triaxial tests. Ideally, detailed information on the internal
angle of friction for the grout should be provided in the design process. As
this requires extensive testing, the suggested parameter, namely ϕ = 30◦ for
mortar, should be adopted. As discussed in Chapter 7, a value of ϕ = 30◦
can also be used in the upper bound solutions, when concrete is used to
grout shallow indentations.

9.3 Practical Relevance Upper and Lower Bounds
As a final remark, it should also be noted that the adopted enhancement

factor in the biaxially compressed key corners has been estimated based on
tests on concrete. For the comparison to test results, the value of c = 1.15
does not seem to be unrealistic. However, additional biaxial testing on
mortar would provide further information about this material property to
be adopted in lower bound modeling. This has not been carried out in
relation to this study.

Upper and Lower Bounds 9.4 Concluding Remarks

In this chapter, the upper bound model, developed in Paper II and presented
in Chapter 7, have been used together with the lower bound model, devel-
oped in Paper III and presented in Chapter 8, to establish an expected range
of results. The expected range was compared to the test results presented in
Chapter 6. The calculations were performed by use of an effectiveness fac-
tor, which accounts only for material brittleness and without any previous
calibration with the upper bound solutions. The comparisons showed that
the first peak loads obtained from tests generally were within the predicted
range. The models have been evaluated based on the comparisons and the
practical relevance of the solutions has been discussed. The findings of the
chapter are summarized as follows:
? The upper and lower bound models are not based on the same assump-
tions and hence, no exact solution can be obtained from the developed
models.
? In most cases, the upper and lower bound models provide a reasonable
narrow range of expected results. The upper and lower bound models
show the same tendencies.
? An exact analytical model would be very difficult to develop as the

conditions in the grout cannot be described as plane stress nor plane
strain.
? The effectiveness factor to be adopted in the upper bound solutions

must account for the material brittleness as well as compensating for
the unsafe nature of the solutions.
? For practical applications, the upper bound solutions (closed-form

equations) are easy to use. However, care must be taken when the
solutions are used in situations that deviate significantly from tested
configurations.
? The lower bound solutions should be applied, when test results are not
available.

10 Dowel and Catenary
Action in Rebars Crossing a
Shear Joint
In this chapter, the second-order plasticity model developed in Paper IV for

modeling of the load-displacement response of a two-sided dowel joint will
be presented. In the preceding chapters, first-order plastic solutions were
developed to model the load-carrying capacity of keyed shear connections.
When comparing with tests, the first-order calculations (upper and lower
bound) refer to the first peak load, which in Chapter 6 was shown to be the
capacity related to failure of the shear keys. The new design for keyed shear
connections displayed a significant ductile response with a pronounced defor-
mation capacity. Examples of load-displacement relationships are given in
Figure 10.1(a), where a higher ultimate load, PU , was reached at larger dis-
placements, than the first peak load, PFP , recorded at small displacements.
The relationships show that the amount of dissipated energy (i.e. area be-
low the response curve) is significantly greater than that of the conventional
design. If this ductile response is to be utilized in an assessment of the over-
all robustness for the structural system, the load-displacement relationship
beyond the first peak capacity should be modeled accurately. In order to
do so within the framework of the theory of rigid-plasticity, second-order
modeling, where change of structural geometry is taken into account, must
be adopted. Examples of large deformation modeling have been demon-
strated by e.g. Calladine (1968), Bræstrup (1980), and Belenkiy (2007). If
such a model is to be developed for keyed shear connections, inspiration can
be found in observations retrieved during and after tests, see e.g. Figure
10.1(b), where the permanent deformations of a U-bar observed after a test
(Specimen P12, ’1-on-2’ design with 10 mm U-bars) can be seen.
In Paper IV, a second-order plasticity model was developed, where the
case of a casting joint crossed by rebar(s) and exposed to pure shear was
considered. The load carrying mechanism consists in this case of a combi-
nation of dowel and catenary action in the rebars. The combination of the
two actions can be related to the imposed shear displacements through the
kinematical conditions of the dowel and the normality condition of plastic
theory. The comparisons between the model and the test results presented in
Paper IV showed good agreements and reasonable estimates of the available

Dowel and Catenary Action
PU
500
PF P
400
P [kN]
300
200
100
Specimen P10, dk =16 mm
Specimen P11, dk =30 mm
0
0 5 10 15 20 25 30
Measured longitudinal deformation [mm]
(a) Load-displacement relationship (b) U-bar deformation
Figure 10.1: Examples of (a) load-displacement relationships where the ulti-

mate load, PU , is larger than the first peak load PFP and (b) U-bar deforma-
tion after test (Specimen P12)
plastic energy can be obtained from the calculations. However, the model,
in its present form, does not provide the means to perform a complete anal-
ysis of the load-displacement relationship for the keyed shear connections
tested. This is partly because the failure of shear keys is not included in
the model and partly because the displacement field assumed is too simple
compared to the experimental observations on the keyed shear connections.
Chapter 4 showed an example of the displacement field for the failure of a
mortar shear key by complete key cut off, see Figure 4.10. It was found
that the failure process of the key includes transverse displacement over the
crack and furthermore that the angle of the relative displacements varied
with increasing shear displacement. Nevertheless, the second-order model
developed in Paper IV assesses a basic problem, where the capacity is in-
creased due to change of structural geometry. The model can be considered
as an introduction to a simple approach, which has potential for practical
assessment of system robustness, e.g. in structural connections of precast
concrete components.
In this chapter, the general considerations and methods adopted for the
second-order modeling will be presented. As the final results are not as self-
explaining as the solutions obtained from first-order modeling, the formulas
from Paper IV will be repeated in the following. However, only a few compar-
isons between the model and test results from the paper will be replicated.
Instead, the model will be compared to the test results of a shear connection

Dowel and Catenary Action 10.1 Second-Order Plastic Modeling
with a ’1-on-2’ loop configuration without indentations (Specimens S4, see

Appendix B.2 for details on the specimen). Based on the comparison, a
discussion will be given on possible extensions of the model that should be
incorporated to account for more complex displacement fields.
10.1 Second-Order Plastic Modeling

The developed second-order plasticity model describes the problem of a rebar
crossing a casting joint loaded in shear. The first phase of the load transfer
mechanism will be dominated by dowel action in the rebar. This was first
treated by Friberg (1938), who modeled the dowel as a linear elastic beam
transversely supported on elastic springs. The obvious shortcoming of this
approach is the linear elastic assumption for material behavior. Due to
the non-linear material behavior, a model based on the theory of plasticity
seems more appropriate to apply for dowel action in the ultimate limit state.
This was done by Rasmussen (1963), who established a first-order plastic
solution for one-sided dowel action (a rebar with part of its length embedded
in a concrete block and loaded transversely at the free end). Rasmussen
assumed yielding of the rebar simultaneously with crushing of the concrete
underneath the dowel.
A joint is called a two-sided dowel joint, when the rebar has both ends
fully embedded in concrete, as shown schematically in Figure 10.2. The
first phase of the load-displacement relationship is, also in this case, gov-
Casting joint/Interface
P (u) Material 2
Idealized plastic deformations

Material 1
d
Rebar, diameter d
fcc,1 ∆
u2 + θ
u u1
θ d
fcc,2
Plastic hinge θ
l1 l2 =
d
P (u) ∆
Figure 10.2: Joint between concretes cast at different times, crossed by a

rebar and subjected to shear displacements, figure from Paper IV

10.1 Second-Order Plastic Modeling Dowel and Catenary Action
erned by development of pure dowel action (plastic hinges and crushing of

the concrete in both blocks, see e.g. Nielsen and Hoang (2011)). However,
with increasing shear displacements, the bending in the rebar is accompa-
nied by tension. Eventually, when further shear displacements are imposed,
the extent of tension increases and in the final stage, the dowel action is
completely replaced by catenary action in the rebar. The development of
pure catenary action requires relatively large shear displacements and it is
evident that the change of structural geometry must be taken into account,
when the capacity in this phase is to be modeled.
In the model, the rates of plastic deformations can be determined by
establishing the kinematic relationship for the assumed mechanism. Then,
by imposing the normality condition of plastic theory on the yield condition
for the rebar and by applying the work equation for increments of displace-
ment, the necessary equations to determine the load, P , as a function of
displacement, u, can be derived. For in-depth details, the reader is referred
to Paper IV.
10.1.1 Assumptions for Second-Order Modeling

As in the first-order plastic models, a number of assumptions have to be
made for the material behavior. The concrete as well as the reinforcing
steel are treated as rigid-plastic materials. However, contrary to the first-
order models, the finite deformation capacity of the materials is taken into
account. Due to the rigid-plastic assumption, the displacements are the sole
results of accumulated plastic deformations. The relative displacement, u,
between the two blocks of concrete is assumed to be parallel to the interface.
The model will be established for the general case, where the casting joint
constitutes the interface between two concretes cast at different times. The
materials on each side of the interface are allowed to have different strengths
and properties when loaded in triaxial compression (e.g. mortar cast against
concrete). The problem is modeled as a displacement-controlled problem
with a monotonically increasing shear displacement. A simple relation for
the time dependent displacement is assumed (a displacement velocity equal
to unity):
u(t) = t (10.1)
Only the case with rebars placed perpendicular to the interface is con-
sidered.
10.1.2 Formation of Mechanism

To establish the starting point for the model, a mechanism must be assumed.
The mechanism is shown in Figure 10.2. When exposed to shear displace-

ments, two plastic hinges must develop in the rebar to accommodate the
displacement. The plastic hinges are developed at the distances l1 and l2 ,
respectively, from the interface. The magnitudes of l1 and l2 depend on the
moment capacity of the dowel and on the local concrete strength on each
side of the interface. In order to comply with compatibility requirements, the
plastic hinges must undergo plastic elongation as well as rotation, θ, when
the shear displacement is increased. Since rigid-plastic material behavior
has been assumed, it is convenient to consider the elongation as a plastic
extension, ∆, concentrated in the plastic hinges. The plastic deformations
are calculated by:
u u1 u2
tan θ = = = (10.2)
l1 + l2 l1 l2
s 2
l1 + l2 l1 + l2 u

∆=− + 1+ (10.3)
2 2 l1 + l2
where u = u1 + u2 , see e.g. Figure 10.2. The plastic deformations can be
regarded as the general strains that correspond to the bending moment, M ,
and normal force, N , acting in the plastic hinges.
The positions of the plastic hinges can be estimated based on equilibrium
considerations (similar to the approach of Rasmussen (1963)). The assumed
displacement field for the rebar implies that local crushing of the concrete
must take place. The actual stress distribution in the concrete is complex
and it is not possible, in this context, to determine the details of the entire
stress distribution. As a simplification, an enhanced average strength, fcc ,
is assumed for the triaxial stress condition of the type developed in the
concrete at the contact zone between the rebar and the concrete:
fcc,i = ci fc,i , ci ≥ 1 (10.4)
where c expresses the enhancement factor for the concrete strength and fc
is the uniaxial compression strength. Index i equals 1 or 2 and refers to the
material on the two sides of the interface. The stress distributions locally in
the concrete at the rebar are assumed to be uniform, see Figure 10.3. In the
undeformed state, i.e. the starting point of the model, the distances l1 and
l2 can be calculated from vertical force equilibrium and moment equilibrium.
The results are given in Equation (10.5) and (10.6):
s v
2 d u fy
u
l1 = s t (10.5)
3 fcc,1 fcc,1
1+
fcc,2
s v
2 d u fy
u
l2 = s t (10.6)
3 fcc,2 fcc,2
1+
fcc,1

Interface
fcc,1
Mp Mp
fcc,2
l1 l2
Mp
-
M
+
Mp
V - -
Figure 10.3: Position of plastic hinges and distribution of contact pressure

at initiation of pure dowel action, illustration from Paper IV
where d is the diameter of the rebar and fy is the yield strength of the rebar.
It should be noted that to fulfill moment equilibrium, the moment in the
rebar at the position of the interface is not zero, when the strengths of the
two materials are different, see moment diagram in Figure 10.3. Hence, to
maintain equilibrium, when fcc,1 6= fcc,2 and when the external action cor-
responds to pure shear, tension must develop in the rebar, which eventually
leads to a distribution of so-called clamping stresses in the interface. The
tension force in combination with the clamping stresses will then be able to
outbalance the (small) bending moment in the rebar at the interface cross
section. Therefore, the assumption of uniformly distributed contact pressure
is not able to fulfill all equilibrium requirements at u = 0 (i.e. initiation of
pure dowel action), when fcc,1 6= fcc,2 . This is, however, acceptable since
the assumption of uniformly distributed concrete stresses primarily was mo-
tivated by the aim of obtaining a simple estimate of the position of the
plastic hinges. In reality, the rebar will probably experience a combination
of tension and bending, already in the elastic range. When having the same
material on both sides of the interface, i.e. fcc,1 = fcc,2 , Equations (10.5)
and (10.6) are identical with the expression obtained by Rasmussen (1963).
At this point, the mechanism has been established and the positions of
the plastic hinges have been estimated. It should here be mentioned, that
the positions of the plastic hinges are assumed to be fixed as the joint is

exposed to shear displacements. The rates of the plastic deformations in the

plastic hinges can be calculated by:
dθ dθ du l1 + l2
θ̇ = = = (10.7)
dt du dt (l1 + l2 )2 + u2
˙ = d∆ = d∆ du = q
∆
u
(10.8)
dt du dt 2 (l1 + l2 )2 + u2
Finally, by use of Equations (10.7) and (10.8), the following condensed

expression for the kinematical condition of the rebar can be established:
˙
s 2
∆ u u

= 1+ (10.9)
θ̇ 2 l1 + l2
10.1.3 Yield Condition of Rebar

As rigid-plastic material behavior is assumed, plastic deformations in the
rebar are only possible, when the entire cross section is stressed to yielding.
In case of pure tension, the plastic tensile capacity is Np . In case of pure
bending, the plastic moment capacity is Mp . The plastic sectional capacities
for a circular cross section are given by:
π 2
Np = d fy (10.10)
4
1
Mp = d 3 f y (10.11)
6
For combinations of bending and axial tension, plastic deformation may
initiate when the yield condition for the rebar is fulfilled, i.e. f (M, N ) = 0.
In previous investigations of dowel action, a number of researchers have
recognized that the presence of axial tension reduces the bending capacity
of the dowel. These includes e.g. Dulácska (1972), Millard and Johnson
(1984), Engström (1990), and Randl and Wicke (2000). Most often, the
simple MN-interaction diagram for a rectangular cross section has been used
to account for the combination of moment and bending:
!2
M N
f (M, N ) = + −1=0 (10.12)
Mp Np
This yield condition is easy to implement in analytical calculations, how-

ever, it is not the correct yield condition for a circular rebar. The MN-
interaction diagram for a circular cross section can be established based on
the stress distribution shown in Figure 10.4.

y0 fy
A1
y1 M
v
d y2 N
fy
A2
Figure 10.4: Full plastic stress distribution in a circular cross section when
subjected to the sectional forces M and N , illustration from Paper IV
The MN-interaction diagram can be calculated by requiring static equiv-

alence between the normal stresses and the sectional forces. The result can
be formulated as follows:
 v 
 !1  !1 u !2
N 2 M 3 M 3u M 3
u

f (M, N ) = +  arcsin  − t1 − −1=0
Np π  Mp Mp Mp 
(10.13)
Figure 10.5 shows the yield conditions of a rectangular and a circular

cross section. It can be seen that the two yield conditions do not deviate
significantly. This means that the yield condition for a rectangular cross
section (when adopting the plastic capacities of a circular cross section)
will be a good approximation for a circular cross section and will only be
a slightly conservative choice. As the shapes of the yield conditions are
roughly similar, the strain vector, ε̄, will have nearly the same direction,
when imposing the normality condition.
In the following, the yield condition of a circular cross section will be
adopted, as this constitutes a more precise description of the combina-
tions of sectional forces that leads to yielding in the rebar. In addition,
it can be mentioned that calculations show a more smooth transition from
a combination of tension and bending to pure catenary action on the load-
displacement curve, when Equation (10.13) is used compared to the use of
Equation (10.12).
From the yield condition of the rebar, the rates of plastic deformations

ε̄
0.8
M/Mp
0.6
0.4
0.2
Rectangular
Circular
0
0 0.2 0.4 0.6 0.8 1
N/Np
Figure 10.5: Comparison of yield conditions for a rectangular and a circular

cross section, including indication of a strain vector fulfilling the normality
condition
can be calculated by imposing the normality condition of plastic theory:

∂f 8 1
θ̇ = λ =λ 3 v (10.14)
∂M πd fy uu !2/3
t1 − M
Mp
˙ = λ ∂f = λ 4
∆ (10.15)
∂N πd2 fy
where λ is a positive constant proportional to the displacement velocity.
˙ θ̇, is established,
When the ratio between the rates of plastic deformation, ∆/
λ cancels out. The condensed expression is given by:
v
!2/3
˙
u
∆ du M
= t1 − (10.16)
θ̇ 2 Mp
10.1.4 Sectional Forces in Plastic Hinges

Based on the rates of plastic deformations established above, a link can
be made between the shear displacement, u, and the sectional forces of
the rebar. It is found that the ratio of plastic deformation rates obtained
from the kinematic relations, Equation (10.9), provides information related

to the material strengths and the shear displacements. From the rates of
plastic deformations given by the constitutive relations, Equation (10.16), a
relation to the sectional forces of the rebar is obtained. Combining these two
expressions, Equations (10.9) and (10.16), the sectional forces in the rebar
as a function of the shear displacement can be established. The bending
moment in the rebar can be calculated by:
2 2 !!3/2
M (u) u u

= 1− 1+ ≮0 (10.17)
Mp d l1 + l2
It can be seen that for zero shear displacement, u = 0, the bending mo-
ment, M (u), equals the plastic bending moment, Mp , and the load transfer
mechanism corresponds to pure dowel action (the first-order solution). Us-
ing the yield condition for the rebar, Equation (10.13), the tension force in
the rebar, N (u), can be obtained:
 v 
 !1  !1 u !2
N (u) 2 M (u) 3
M (u) t
3 u
M (u) 3
u

=1−  arcsin  − 1− ≯1
N p π M p Mp Mp 
(10.18)
According to Equation (10.18), tension develops in the rebar as soon

as u > 0. At a certain shear displacement, the normal force reaches the
tensile capacity, Np , and the regime of pure catenary action is reached. The
displacement at the transition to pure catenary action can be calculated by
equating the right hand side of Equation (10.17) with zero.
10.1.5 Displacement Capacity

In plastic theory, the materials are normally assumed to have unlimited
deformation capacity. In reality, the capacity is limited and for a second-
order model where the load keeps on increasing with increasing deformations,
the range of the model should be restricted by accounting for the capacity
of the materials. The deformation capacity can be evaluated for concrete
and for the reinforcement.
Starting with the concrete, which is the least rigid-plastic material of
the two, the material experiences softening after reaching the peak strength.
The local pressure in the concrete reduces, when the concrete experiences
too large compressive strains. The displacement, u, may become so large
that the concrete near the joint interface spalls off which leaves a zone to
be stress free. With increased shear displacement, this zone will expand

from the position of the interface and into the concrete blocks. Due to
the assumed material properties, redistribution of the contact pressure will
take place as u increases. In addition to this, the redistribution of stresses
must take place in such a way that equilibrium can be maintained, when
catenary action starts to develop in the rebar. In the modeling, the effect of
softening (and spalling of concrete) will indirectly be taken into account by
introducing effective lengths, l1,ef and l2,ef , over which the contact pressures,
fcc,1 and fcc,2 , are assumed to act uniformly, see Figure 10.6.
N (u) M (u)
fcc,1
u
θ
fcc,2
N (u)
M (u)
l1,ef l2,ef
l1 l2
Figure 10.6: Distribution of contact pressures at combined dowel action and

catenary action, illustration from Paper IV
This may be interpreted as an assumption of rigid-plastic behavior with

finite deformation capacity, although there is actually no real information
about the deformation capacity of the concrete, when it is subjected to con-
tact pressures by the dowel. The following relationship between u and the
effective lengths, l1,ef and l2,ef , can be established from equilibrium consider-
ations in the deformed state:
 
v
 u
M (u) 
u1 −
 u 
 2 
li,ef = li 1 − t
 u u fcc,1 fcc,2 ≮0 (10.19)
Mp + d 
2 fcc,1 + fcc,2
 
 
It appears that the effective lengths are identical to the initial lengths,
l1 and l2 , when u = 0. Furthermore, when u has reached a value that
makes M (u) = 0 the effective lengths reduce to zero. It can be interpreted
as a situation, where the rebar will no longer experience contact pressures
between the two hinges, but instead acts as a tie.
Regarding the reinforcing steel, the deformation capacity is also limited,
even though the assumption of rigid-plastic material behavior fits much bet-
ter for the reinforcing steel than for concrete. When exposed to tension, the

elongation of the rebar is uniformly distributed over the length. At a certain

point, concentrated local strains develop, also referred to as necking. Shortly
after, failure by rupture will initiate. The ability to carry tension will remain
until rupture of the rebar and the local deformations in the necking zone will
be relatively large.
The deformation capacity of the rebar can be implemented in the model
as a maximum elongation, ∆max , that can be allowed in the plastic hinges.
To facilitate this, the elongation capacity in the necking region of a rebar
must be studied. In Figure 10.7(a), four examples of experimentally ob-
tained stress-elongation relationships are presented. Figure 10.7(b) shows
the strains in the rebar and the position of the necking region. The mea-
surements were possible by use of DIC. This detailed information cannot
be obtained from the standard methods for measuring the nominal maxi-
mum strain in rebars. In the case of dowel and catenary action, the plastic
hinges can be interpreted as the necking zone of a rebar exposed to tension
as plastic strains accumulate in the hinges.
800
fu
700
600 fy 6
2d

?
500
σs [MPa]
400
300
200
100
0
0 0.5 1 1.5 2 2.5 3 3.5 4
∆ [mm] ∆max
(a) Stress-elongation relation (b) Aramis results at neck-
ing (major principal strain)
Figure 10.7: Local stress-elongation relationship for rebars and indication

of the measurement length spanning the zone where necking occurs, figures
from Paper IV

10.1.6 Load Displacement Relationship

Based on the obtained results, it is now possible to determine the load-
displacement response of a perfectly smooth two-sided dowel joint by use
of the work equation. For this purpose, a stationary situation with dis-
placement, u, and load, P (u), is used as the starting point from which an
increment of displacement, δu, is considered. The external work, WE , is
given by:
WE = P (u) δu (10.20)
The internal work, WI , has contributions from the energy dissipated
in the plastic hinges during deformation as well as the energy absorbed,
when the concrete crushes at the rebar. As the strengths and properties
of the two materials on each side of the interface may be different, the
contributions from the two sides are included separately. The internal work
can be calculated by:
! !
1 l1,ef 1 l2,ef
WI = fcc,1 l1,ef d δu1 + fcc,2 l2,ef d δu2
2 l1 2 l2 (10.21)
+2N (u) δ∆ + 2M (u) δθ
The incremental displacement and deformation quantities, δu1 , δu2 , δ∆,
and δθ, can be expressed in terms of δu through the following relationships
(where δt = δu):
δθ = θ̇δu (10.22)
˙
δ∆ = ∆δu (10.23)
l1
δu1 = δu (10.24)
l1 + l2
l2
δu2 = δu (10.25)
l1 + l2
By setting up the work equation, an expression for the shear load, P , as
a function of the shear displacement, u, can be derived. The result is given
in Equation (10.26):

2 2
 1 f d l1,ef + 1 f d l2,ef + 2N (u) ∆
˙ + 2M (u) θ̇ (10.26a)


cc,1 cc,2
Ps (u) = 2 l1 + l2 2 l1 + l2
˙

 2N ∆

(10.26b)
p
The two parts of the equation reflect the two regimes of the load transfer
mechanism. Equation (10.26a) is valid for N (u) < Np , i.e. a combination
of dowel and catenary action and Equation (10.26b) is valid for N (u) = Np ,
where the capacity stems from catenary action (i.e. pure tension in the
rebar). For u = 0, the capacity stems solely from dowel action.

10.1.7 Inclusion of Friction

Equation (10.26) is established for the case of a perfectly smooth interface,
where there is no resistance from friction. In reality, a casting joint is never
frictionless. From Equation (10.18), it was found that tension develops in
the rebar, when u > 0. The tension force causes compressive normal stresses
to develop in the interface (so-called clamping stresses). From simple shear-
friction considerations, the resistance stemming from friction can be esti-
mated by Equation (10.27), assuming a coefficient of friction, µ, which is
independent of the shear displacement:
Pf (u) = µN (u) cos θ (10.27)
The value of µ must reflect the properties of the interface. In the liter-
ature, several suggestions have been made for the coefficient of friction to
be applied in different situations. Birkeland and Birkeland (1966) suggested
µ = 1.7 for monolithic concrete and µ = 0.8 − 1.0 for ordinary construction
joints. These values reflect calculation models where all the shear capacity
stems from friction without consideration of dowel action (bending in the
dowel). Other suggestions have been made for a combination of actions.
Engström (1990) introduced a model, where friction is combined with dowel
action using a coefficient of friction of µ = 0.3−0.6 for a concrete-to-concrete
interface. Randl and Wicke (2000) also considered a combination of tension
and bending in the rebar with interface friction using µ = 0.5 for smooth
interfaces. However, the mentioned works did not provide a link between
the shear displacement and the sectional forces in the rebar. When this link
is established, the coefficient of friction should reflect the properties of the
interface alone. This is the same property that was sought in Paper III,
where µ = 0.75 was used for the smooth (and untreated) joints. For tests
with greased interfaces, µ = 0.3 was adopted in Paper IV.
10.1.8 Load-displacement Response

The total load-displacement response, including the combination of dowel
and catenary action as well as friction in the interface can be obtained by
addition of Equations (10.26) and (10.27):
Ptotal (u) = Ps (u) + Pf (u) (10.28)
An example of model results is given in Figure 10.8, see Paper IV for

details. The results clearly show the two displacement regimes (transition in-
dicated by a plus sign), where the first phase is governed by a combination of
bending and tension in the rebar and the late phase is governed by catenary

Dowel and Catenary Action 10.2 Comparison of Model with Test Results
action in the rebar. The inclusion of friction raises the load-displacement

response as the tension force in the rebar increases. It is also found that
the contribution from friction does not influence the first-order solution at
u = 0. In the following, a few comparisons with the experimental results
presented in Paper IV will be repeated and in addition, a comparison to the
results of a shear connection test on a specimen without shear keys will be
given.
25
Ptotal
20 ↓
Ps
↓
Force [kN]
15
10
Pf (µ = 0.3)
↓
5
0
0 5 10 15
Displacement [mm]
Figure 10.8: Load-displacement relation for a rebar crossing a casting joint

loaded in shear, illustration from Paper IV
10.2 Comparison of Model with Test

Results
When the model is compared to test results, only a few model parameters
have to be determined, namely the coefficient of friction and the enhance-
ment factors for the concrete materials. For the smooth and greased inter-
face, µ = 0.3 is adopted. Regarding the enhancement factors, ci , Rasmussen
(1963) calibrated his model with test results and found values in the range
of 3.7 − 5.4 for concrete. In case of a two-sided dowel joint, the c-factors
can also be calibrated with tests, however, it requires tests, where the same
material is used on both sides of the interface. The calibration with test re-
sults should be performed at the onset of plastic deformations, e.g. when the
first-order solution is reached, as the model response at u = 0 is independent
of the contribution from friction, see e.g. Figure 10.8.
In Paper IV, reasonable agreement between tests and the model was

10.2 Comparison of Model with Test Results Dowel and Catenary Action
found when a value of c = 5 was used for concrete (also suggested by

Vintzeleou and Tassios (1986)) and c = 4 was used for mortar. It is ar-
gued to use a smaller enhancement factor in the case of mortar, as mortar
normally has a relatively more brittle compression failure compared to con-
crete. An estimate of the local enhancement factor cannot be obtained from
the triaxial tests presented in Chapter 3. However, it was evident from the
triaxial tests that the increase in strength for mortar was less than that for
concrete, when loaded in triaxial compression. Based on this observation it
is justified to adopt a smaller value of c for mortar than for concrete.
40
Elastic Regime
35
←
Plastic Regime
30
→
Force [kN]
25
20
15
10
0
0 5 10 15 20 25
(a) 1 rebar
160
Elastic Regime
140 ←
Plastic Regime
120 →
Force [kN]
100
80
60
40
20
0
0 5 10 15 20 25
(b) 4 rebars
Figure 10.9: Comparison of second-order model with test results (a) with one
rebar and (b) with four rebars, graphs from Paper IV

Figure 10.9 shows the comparison between the model and results from
tests, where concrete was used on one side of the interface and mortar was
cast on the other side (concrete-to-mortar interface). For details on material
properties and test setup, see Paper IV. It is seen that reasonable agreement
is obtained in the entire displacement regime and that the load level at the
onset of plastic deformations is predicted fairly well. In the figure, calcula-
tions of an upper and lower limit for the expected test results are included.
The lower limit is calculated by use of the yield strength of the reinforcement,
fy , as this strength governs the onset of plastic deformations. However, as
can be seen in Figure 10.7(a) the reinforcement hardens already at relatively
small strains/elongations. For this reason the calculations based on the ul-
timate strength of the reinforcement, fu , is also included as an upper limit.
Such reasoning was also made by Mirzaei (2010) and Fernández Ruiz et al.
(2013) for the post-punching behavior of flat slabs. In the comparisons, an
estimate of the elastic displacements has also been included (represented as
a horizontal shift of the plastic load-displacement curve). The calculations
were performed based on the analogy of a beam on an elastic foundation
with a constant stiffness, see Paper IV for details.
From the comparison, it is also seen that the estimate for the maximum
displacement, i.e. umax , which depends on the experimentally determined
∆max , corresponds fairly well with the maximum displacement obtained in
tests. In the calculations ∆max = 3 mm was used. Paper IV contains addi-
tional comparisons between the model and test results. The tests comprised
specimens with 1, 2, and 4 rebars crossing the interface. The tests showed
that the position of the rebars provided results that were proportional to
each other, i.e. there were no interaction between the crushed zones of the
(a) Local crushing of mortar/concrete (b) Rupture of rebar
Figure 10.10: (a) Local crushing at rebars at the casting joint and (b) de-
flected shape of ruptured rebar, illustrations from Paper IV

10.2 Comparison of Model with Test Results Dowel and Catenary Action
concrete. This can be seen in Figure 10.10(a), where the concrete clearly was
crushed and, to some extent, had spalled off at the interface. Moreover, Fig-
ure 10.10(b) shows an extracted rebar after a test. From these observations,
it appears that the model assumptions of concrete crushing and formation
of plastic hinges in the rebar are fair assumptions.
As a general remark, it should be noted that a higher compressive strength
of the materials leads to a higher load at the transition to the non-linear be-
havior, i.e. at full dowel action. However, a higher compressive strength
also leads to a decrease in displacement capacity. Therefore, a lower com-
pressive strength of the concrete/mortar is beneficial for catenary action,
because this requires plastic deformation over a longer length of the rebar
and at the same time makes it easier for the rebar to cut its way through
the concrete/mortar.
10.2.1 Comparison to Test Results for Shear

Connections without Indentations
In Chapter 6, results of tests on shear connections without shear keys were
presented, see Figure 6.8. The specimens were reinforced transversely by
loop connections with the new design in a ’1-on-2’ configuration. The de-
veloped second-order model can be compared to these test results, as the
displacements primarily consisted of sliding along the interfaces without a
noticeable dilatation. As the reinforcement degree was smaller in one side of
the connection, the displacements were primarily concentrated in this inter-
face. However, it was observed that the displacements were not limited to
the interface with the smallest amount of reinforcement. Figure 10.11 shows
results of a DIC analysis of Specimen S4. It can be seen that strains are
also localized in the interface with the largest amount of reinforcement. To
establish test results that are comparable with the displacements assumed
in the model, the displacements (measured in the longitudinal direction of
the shear connection) over the interface with the smallest amount of re-
Two U-bars per loop
One U-bar per loop

Figure 10.11: Major principal strain during test and representation of shear
displacement in the weak side of the connection, Specimen S4

inforcement have been isolated (see right hand side of Figure 10.11). It
appears that the difference to the response curve, where the displacements
are measured over the entire width of the connection (see Figure 6.8), is not
great. At the time of rupture of the first rebar (u ≈ 15 mm), the difference
amounts to approximately 3.5 mm. When comparing the DIC measured
displacements to the second-order model, the same parameters as adopted
in Paper IV can be used. The interface/joint had a precast concrete element
on one side (c1 = 5) and an in-situ cast mortar on the other side (c2 = 4).
In addition, the interface was greased before casting the connection grout
(µ = 0.3). The connection grout was tested to a compressive strength of
fc,2 = 43.0 MPa. Unfortunately, the compressive strength of the precast ele-
ment was not tested in the experimental campaign. Therefore, the strength
has been estimated, and a value of fc,1 = 50 MPa was used in the calcula-
tions. The comparison between the extracted load-displacement relationship
of the weakest side of the shear connection, where 8 rebars crossed the cast-
ing joint, and the model, can be seen in Figure 10.12(a).
The model underestimates the load carrying capacity at the onset of
plastic deformations (i.e. the first-order solution). However, the develop-
ment after transition to pure catenary action is captured surprisingly well,
including the estimate of the displacement capacity, umax , (∆max = 3 mm
was adopted in the calculations). In Figure 10.12(b), the DIC measured
transverse displacements across the interface are presented. For correlation
between the measured displacements, some characteristic points have been
indicated in the figures. On an overall level, it can be concluded that the as-
sumption of pure shear displacement is a relatively good model assumption
in this particular case (only a maximum of 0.3 mm transverse displacement
compared to longitudinal displacements of 10 mm at the same instance).
However, it is found that as the in-elastic displacement regime is entered,
the shear displacement is accompanied by an increase in transverse displace-
ment. The transverse displacements would be far more pronounced in the
case of keyed shear connections, when shear displacements are imposed after
failure of the shear keys.
Although the model does not agree with the test results in all aspects,
the model furnishes a simple tool for robustness assessment. In this relation,
the underestimation seen at small shear displacements in Figure 10.12 is not
decisive in a robustness consideration, as the main concern is the amount of
dissipated energy, i.e. the area under the curve. In this respect, the model
seems to be able to provide a reasonable estimate of the available plastic
energy.

10.3 Considerations for Model Extensions Dowel and Catenary Action
300
Elastic Regime Model
← Specimen S4
250
Plastic
→ Regime
200
P [kN]
150
100
50
0
0 5 10 15 20 25
(a)
300
250
200
P [kN]
150
100
50
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Measured transverse displacement [mm]
(b)
Figure 10.12: (a) Comparison of second-order model with test results ob-
tained from DIC and (b) DIC measured transverse displacements, Specimen
S4
10.3 Considerations for Model Extensions

As mentioned above, the model cannot directly be used to predict the load-
displacement relationship of keyed shear connections. This is because the
model assumes displacements parallel to the interface, whereas the pres-
ence of shear keys in the interface entails that large shear displacements
will be associated with transverse displacements. Examples of transverse
displacements were given in Figures 4.10 and 6.10, which showed that the

Dowel and Catenary Action 10.3 Considerations for Model Extensions
magnitude of transverse displacements cannot be neglected in the modeling.

It would be possible in future works to extend the model to account for
dilatation. In this relation, it should be noted that the failure mode of the
shear keys (local key corner crushing versus complete key cut off) influences
the load-displacement relationship. This is e.g. seen in Figure 10.1(a), where
the failure of Specimen P10 was characterized by local key corner crushing,
whereas Specimen P11 suffered complete key cut off.
The starting point (first-order solution) of the current model assumes
pure dowel action (pure bending in rebar). However, in the first-order so-
lution for the capacity of a keyed connection, the reinforcement is assumed
to be stressed to tensile yielding. The transition from the failure of the
shear keys to new equilibrium conditions after failure of the shear keys must
be considered. The tests showed that the failure of the shear keys may
lead to a brittle drop in load (see e.g. Figure 6.9) before equilibrium is
reestablished and an increase in load can be facilitated. The problem is here
to establish the level of tension in the rebar, N0 , at the new equilibrium
state. In addition, the yield condition has to be established for the sectional
forces in the rebar under these new conditions. The simplest way to extend
the current model to cover this case is to replace the MN-diagram with a
∆M∆N-diagram describing the possible combinations of tension and bend-
ing in a plastic hinge, when N0 is subtracted. With this reduced interaction
diagram and the assumption of the associated flow rule, the sectional forces
in the dowel can be established as a function of the shear displacement in a
similar way as shown above.
Once the starting point is fixed, the nature of the displacement field
must be considered, i.e. dilatation must be accounted for in the model.
Such a mechanism is depicted in Figure 10.13, where the displacement, v,
can be described by the components u and t related by the angle ω. From
this mechanism, the kinematic relations of the problem can be established
as in the current model. The angle of the displacement will be dependent
on the roughness of the failure surface and most likely, the direction of
the displacements will change with increased shear displacement. When
the mechanism includes dilatation, the rate of the plastic elongation, ∆, ˙
in the hinges will be larger than in the current model and as a result, the
displacement capacity, umax , of the joint will be smaller. However, the load
carrying capacity will be greater as tension is introduced earlier and the
transition to catenary action, which constitutes the regime where the highest
capacity is calculated, takes place at a smaller shear displacement.

10.3 Considerations for Model Extensions Dowel and Catenary Action
P (u) Material 2
u
Idealized plastic deformations
Material 1 t ω
v u d
Rebar, diameter d
fcc,1 ∆
u2 + θ
u u1
θ d
fcc,2
Plastic hinge θ
l1 l2 =
d
P (u) ∆
Figure 10.13: Mechanism including dilatation in a casting joint

Dowel and Catenary Action 10.4 Concluding Remarks

In this chapter, the second-order plasticity model developed in Paper IV
for assessment of dowel and catenary action in rebars crossing a shear joint
has been described. The model provides an estimate of the in-elastic load-
displacement relationship for smooth joints exposed to large shear displace-
ments. The model has, in addition to the comparisons with test results
presented in Paper IV, been compared to DIC results of a ’1-on-2’ shear
connection with a smooth and greased interface without shear keys. The
comparison was performed using the same model parameters as established
in Paper IV and a satisfactory agreement was found. The findings of the
chapter are summarized as follows:
? The kinematic relations for the mechanism and the normality condition
of plastic theory were combined to establish a unique link between the
shear displacements and the sectional forces in the rebar.
? The yield condition of a circular cross section was used for the rebar.
This constitutes an improvement compared to the use of the yield
condition of a rectangular cross section.
? The work equation was used to establish an expression for calculation

of the load-displacement relationship for the case of a rebar crossing
a casting joint (Equation (10.28)). The materials on each side of the
interface were allowed to have different strengths.
? The calculation of tension in the rebar provided the necessary infor-

mation to include friction in the interface in a consistent manner.
? As an estimate of the displacement capacity of the joint, the local

deformation capacity of a rebar was determined from uniaxial tension
tests and incorporated in the model as a limit for elongation in the
plastic hinges of the rebar.
? Despite the simplicity of the model and the relatively small number
of required input parameters, the developed second-order plasticity
model has the potential for assessment of structural robustness.

11 Conclusions
This thesis presented a study of structural joints in precast concrete struc-
tures. The main part of the study focused on keyed shear connections be-
tween shear wall panels. The study comprised a large experimental pro-
gram as well as development of calculation models based on limit analy-
sis. The theoretical work consisted of first-order models for assessment of
structural capacity and a second-order plastic model for prediction of the
load-displacement relationship of a shear joint.
The overall objectives of the thesis were to develop a solution for struc-
tural connections that improves the on-site construction phase, provides
structural continuity, and ensures a ductile behavior. The requirements
were met by a modification of the conventional design for keyed shear con-
nections. The traditional orientation of the U-bar loops was changed and
the transverse reinforcement (lacer bars) took the form of double T-headed
rebars. The results of tests on the new configuration were very promising,
and in terms of ductility the new design was superior to the conventional
design. In the following, the main conclusions for the different topics of the
study will be outlined.
Main Research Topics

The thesis treated a number of aspects concerning material properties and
basic mechanisms that led to a detailed investigation of the keyed shear
connections. The main topics and their findings can be summarized as:
? The material properties of mortar were studied, as mortar is

the primary grout material for connections between precast concrete
elements in buildings. From triaxial tests on both concrete and mortar
it was found that a smaller maximum grain size of a mixture led to
a decrease in triaxial strength. The experimental results showed that
for a mortar (dmax ≤ 4 mm) the internal angle of friction can with
reasonable accuracy be taken as ϕ = 30◦ (corresponding to k = 3).
? The failure of a mortar shear key was studied in push-off tests

and the results were analyzed by use of digital image correlation. The
results showed that the recorded peak load corresponded to failure
of the shear key. Analysis of the displacement field, measured over

Conclusions
the governing crack, showed that the shearing of a mortar key was
associated with dilatation.
? The tensile capacity of loop connections was experimentally stud-

ied. Upper bound plasticity models were developed to estimate the
amount of lacer reinforcement required to obtain yielding of the U-bars
(avoiding the more brittle failure of the grout), when the connection
is loaded in tension. The models can be used for concrete as well as
mortar and comparison to test results showed reasonable agreement.
The comparison revealed the importance of using the correct material
parameters for the grout.
? Tests on shear connections were performed. The experimental

campaign comprised tests on the conventional design and the new
design with varying U-bar diameters and key geometries. The tests
showed that the first peak load on the load-displacement curve was
associated with failure of the shear keys, either as complete key cut
off or as key corner crushing. The new design displayed a ductile re-
sponse, where the ultimate load recorded at large shear displacements
could be larger than the first peak load.
? Upper bound solutions were developed for prediction of the first

peak load recorded in the tests. The developed model accounts for
the local failure of the shear keys and the formation of diagonal cracks
between opposite placed shear keys. The local failures of the keys were
either complete key cut off or crushing of the key corner. Overall good
agreement between tests and calculations were obtained, when using
an effectiveness factor which in the literature was calibrated with tests
of similar problems.
? Lower bound solutions were developed as the counterpart to the

upper bound solutions. The essentials of the developed model were
the possibility to combine struts with different inclinations and the
inclusion of friction in the indented interfaces. The solutions constitute
improvements to the previous developed single-strut model and better
agreement with test results was obtained.
? The set of the upper and lower bound solutions provided a

range of expected results. It was not possible to obtain the theoretical
exact solution, as the assumptions for the models were not completely
identical. The comparison of the models emphasized that in prac-
tice, the effectiveness factors to be used in the two models should not
necessarily be identical. In the upper bound solutions, the ν-factor
should account for brittleness effects as well as the model assumptions

Conclusions
in order to address the unsafe nature of upper bound solutions. For

practical situations, the upper bound solutions are easier to adopt,
however, the solutions should only be used, when the ν-factor can be
calibrated with test results. Where experimental results are not avail-
able, the lower bound solutions can safely be used for prediction of the
structural capacity of the keyed connections.
? Second-order modeling of dowel and catenary action in rebars

was performed to establish the load-displacement relationship for a
rebar crossing a casting joint loaded in pure shear. The kinematical
conditions of the two-sided dowel joint and the normality condition of
plastic theory were used to establish a unique link between the shear
displacement and the sectional forces in the rebar.
The model furnishes a simple tool to estimate the amount of plas-
tic energy available in the joint. Comparison with a test on a shear
connection without shear keys provided promising results. However,
the model considers only shear displacements parallel to the interface
and for this reason, it is not directly applicable for analysis of the load-
displacement relationship of keyed shear connections. A few extensions
should be incorporated in the current model to make it applicable for
keyed shear connections.
In conclusion, the test results on the new design for keyed shear connec-
tions were very promising. Regarding ductility, the new design is superior to
the conventional design. The developed theoretical models showed an over-
all satisfying agreement with the test results and it can be concluded that
the models furnish a sound theoretical basis for assessment of the ultimate
behavior of keyed shear connections. Due to the large load carrying capac-
ity and the ductile behavior that can be obtained by the new design, the
connection type has potential for use in high rises and other heavily loaded
structures build by the precast method.

Bibliography
Abdul-Wahab, H. M. S. and Sarsam, S. Y. H. (1991). Prediction of Ultimate
Shear Strength of Vertical Joints in Large Panel Structures. ACI Structural
Journal, 88(2):204–213.
Abrams, D. A. (1919). Design of Concrete Mixtures. Bulletin No. 1, Struc-

tural Materials Research Laboratory, Lewis Institute, Chicago.
Bakhoum, M. M. (1991). Shear Behavior and Design of Joints in Precast

Concrete Segmental Bridges. PhD thesis, Department of Civil Engineer-
ing, Massachusetts Institute of Technology.
Balmer, G. (1952). A General Analytic Solution for Mohr’s Envelope. Pro-

ceedings ASTM, 52:1260–1271.
Belenkiy, L. M. (2007). Upper-Bound Solutions for Rigid-Plastic Beams and

Plates of Large Deflections by Variation Principles. Journal of Engineering
Mechanics, 133(1):98–105.
Birkeland, P. W. and Birkeland, H. W. (1966). Connections in Precast

Concrete Construction. Journal of the American Concrete Institute,
63(3):345–368.
Bongers, J. P. W. and Rutten, H. S. (1998). Concrete in multiaxial com-

pression - a multilevel analysis. Heron, 43(3):159–180.
Bræstrup, M. W. (1980). Dome Effect in RC Slabs: Rigid-Plastic Analysis.

Journal of the Structural Division - ASCE, 106(6):1237–1253.
Calladine, C. R. (1968). Simple Ideas in the Large-Deflection Plastic Theory

of Plates and Slabs. In Heyman, J. and Leckie, F. A., editors, International
Conference on the Applications of Plastic Theory in Engineering Design,
pages 93–127, Cambridge, UK.
CEN (2004). EN 1992-1-1 Eurocode 2: Design of Concrete Structures -

Part 1-1: General Rules and Rules for Buildings. European Committee
for Standardization, Brussels, 3rd edition.
Chakrabarti, S. C., Nayak, G. C., and Paul, D. K. (1988). Shear Characteris-

tics of Cast-in-Place Vertical Joints in Story-High Precast Wall Assembly.
ACI Structural Journal, 85(1):30–45.

Bibliography
Chen, W.-F. (2007). Plasticity in Reinforced Concrete. J. Ross Publishing,

Fort Lauderdale.
Cholewicki, A. (1971). Loadbearing Capacity and Deformability of Vertical

Joints in Structural Walls of Large Panel Buildings. Building Science,
6(4):163–184.
Christoffersen, J. (1997). Ultimate Capacity of Joints in Precast Large Panel

Concrete Buildings, Series R No 25. PhD thesis, Department of Structural
Engineering and Materials, Technical University of Denmark, Lyngby.
Coulomb, C. A. (1776). Essai sur une application des règles de maximis &
minimis à quelques problèmes de statique, relatifs à l’Architecture (En-
glish: Note on an application of the rules of maximum and minimum to
some statical problems, relevant to architecture, Heyman (1997)). Mé-
moires de Mathèmatique & de Physique, présentés à l’Academie Royale
des Sciences, 7:343–382.
Dahl, K. K. B. (1992a). A Constitutive Model for Normal and High Strength

Concrete, Series R No 287. PhD thesis, Department of Structural Engi-
neering, Technical University of Denmark, Lyngby.
Dahl, K. K. B. (1992b). A Failure Criterion for normal and High Strength

Concrete, Series R No 286. PhD thesis, Department of Structural Engi-
neering, Technical University of Denmark, Lyngby.
Dahl, K. K. B. (1992c). The Calibration and Use of a Triaxial Cell, Series

R No 285. PhD thesis, Department of Structural Engineering, Technical
University of Denmark, Lyngby.
Dahl, K. K. B. (2014). Bella Sky Hotel - taking precast concrete to the limit.
Structural Concrete, 15(4):441–447.
Drucker, D. C. (1950). Some Implications of Work Hardening and Ideal

Plasticity. Quarterly of Applied Mathematics, 7(4):411–418.
Drucker, D. C. and Prager, W. (1952). Soil Mechanics and Plastic Analysis

or Limit Design. Quarterly of Applied Mathematics, 10(2):157–165.
Drucker, D. C., Prager, W., and Greenberg, H. J. (1952). Extended Limit

Design Theorems for Continuous Media. Quarterly of Applied Mathemat-
ics, 9(4):381–389.
Dulácska, H. (1972). Dowel Action of Reinforcement Crossing Cracks in

Concrete. ACI Journal, 69(12):754–757.

Bibliography
Engström, B. (1990). Combined Effects of Dowel Action and Friction in

Bolted Connections. Nordic Concrete Research, 9:14–33.
Eriksson, A. (1978). Structural behaviour of vertical joints in large panel

buildings. PhD thesis, Division of Structural Design, Chalmers University
of Technology, Göteborg.
Fauchart, J. and Cortini, P. (1972). Étude expérimentale de joints hori-

zonraux entre panneaux préfabriqués pour murs de batiments. Annales de
L’institut Technique du Batiment et Des Travaux Publics, Paris.
Fernández Ruiz, M., Mirzaei, Y., and Muttoni, A. (2013). Post-Punching

Behavior of Flat Slabs. ACI Structural Journal, 110(5):801–812.
fib (2008). Bulletin 43: Structural Connections for Precast Concrete Build-
ings. fédération international du béton, Lausanne, Switzerland.
fib (2013). fib Model Code for Concrete Structures 2010. Wilhelm Ernst &
Sohn, Berlin, Germany.
fib (2014). Bulletin 74: Planning and Design Handbook on Precast Building
Structures. fédération international du béton, Lausanne, Switzerland.
Friberg, B. F. (1938). Design of Dowels in Transverse Joints of Concrete

Pavements. American Society of Civil Engineers, 64(9):1809–1828.
Gvozdev, A. (1960). The determination of the value of the collapse load for
statically indeterminate systems undergoing plastic deformation. Interna-
tional Journal of Mechanical Sciences, 1(4):322–335.
Halasz, R. and Tantow, G. (1966). Schubfestigkeit der Vertikalfugen im

Grosstafelbau. In Berichte aus der Bauforschung, Heft 39. Verlag Wilhelm
Ernst & Sohn. Berlin.
Hansen, K., Kavyrchine, M., Melhorn, G., Olesen, S. Ø., Pume, D., and
Schwing, H. (1976). SBI-report 97: Keyed shear joints. Technical report,
Danish Building Research Institute, Copenhagen.
Hansen, T. C. (1995). Triaxial Tests with Concrete and Cement Paste, Series
R No 319. PhD thesis, Department of Structural Engineering, Technical
University of Denmark, Lyngby.
Herfelt, M. A. (2017). Numerical limit analysis of precast concrete structures

- A framework for efficient design and analysis. PhD thesis, Department
of Civil Engineering, Technical University of Denmark, Lyngby.

Bibliography
Herfelt, M. A., Poulsen, P. N., Hoang, L. C., and Jensen, J. F. (2016). Nu-
merical limit analysis of keyed shear joints in concrete structures. Struc-
tural Concrete, 17(3):481–490.
Hilsdorf, H. K., Lorman, W. R., and Monfore, G. (1973). Triaxial Testing

of Nonreinforced Concrete Specimens. Journal of Testing and Evaluation,
1(4):330–335.
Hobbs, D. (1971). Strength of Concrete under Combined Stress. Cement

and Concrete Research, 1(1):41–56.
Imran, I. and Pantazopoulou, S. J. (1996). Experimental Study of Plain

Concrete under Triaxial Stress. ACI Materials Journal, 93(6):589–601.
Ingerslev, A. (1921). Om en elementær beregningsmetode af krydsarmerede

plader (English: On an Elementary Method of Calculation for Two-way
Slabs). Ingeniøren, 30(69):507–515.
Ingerslev, A. (1923). The Strength of Rectangular Slabs. The Structural

Engineer, 1(1):3–14.
Issa, M. A. and Abdalla, H. A. (2007). Structural Behavior of Single Key

Joints in Precast Concrete Segmental Bridges. Journal of Bridge Engi-
neering, 12(3):315–324.
Jensen, B. C. (1975). On the Ultimate Load of Vertical, Keyed Shear Joints

in Large Panel Buildings. Technical report, Institute of Building Design,
Report No 108, Technical University of Denmark, Lyngby.
Jensen, B. C. (1976). Nogle plasticitetsteoretiske beregninger af beton og

jernbeton (English: Some Applications of Plastic Analysis to Plain and
Reinforced Concrete). Report no 111, Institute of Building Design, Tech-
nical University of Denmark, Lyngby.
Johansen, K. W. (1931). Beregning af krydsarmerede jernbetonpladers brud-

moment (English: Calculation of the Rupture Moment of Two-way Slabs).
Proceedings of the Danish Society for Structural Science and Engineering,
3(1):1–18.
Johansen, K. W. (1943). Brudlinieteorier (English: Yield Line Theory,

Cement and Concrete Association, London, 1962). Jul. Gjellerups Forlag,
Copenhagen.
Johansen, K. W. (1958). Brudbetingelser for sten og beton (English: Failure

Criteria of Rock and Concrete). Proceedings of the Danish Society for
Structural Science and Engineering, 29(2):25–44.

Bibliography
Jørgensen, H. B. (2014). Strength of Loop Connections between Precast Con-

crete Elements. PhD thesis, Department of Technnology and Innovation,
University of Southern Denmark, Odense.
Jørgensen, H. B. and Hoang, L. (2015a). Strength of Loop Connections

between Precast Bridge Decks Loaded in Combined Tension and Bending.
Structural Engineering International, 25(1):71–80.
Jørgensen, H. B. and Hoang, L. C. (2013). Tests and limit analysis of

loop connections between precast concrete elements loaded in tension.
Engineering Structures, 52:558–569.
Jørgensen, H. B. and Hoang, L. C. (2015b). Load Carrying Capacity of

Keyed Joints Reinforced with High Strength Wire Rope Loops. In Con-
crete - Innovation and Design, fib Symposium, Copenhagen.
Jørgensen, H. B., Hoang, L. C., and Hagsten, L. G. (2017). Strength of pre-

cast concrete shear joints reinforced with high-strength wire ropes. Pro-
ceedings of the Institution of Civil Engineers. Structures and Buildings,
170(3):1–12.
Jørgensen, K. F. (2015). Bella Sky Hotel - exploring the potential in precast

concrete design. Structural Concrete, 16(4):449–457.
Kaneko, Y., Connor, J. J., Triantafillou, T. C., and Leung, C. K. (1993a).

Fracture Mechanics Approach for Failure of Concrete Shear Keys. I: The-
ory. Journal of Engineering Mechanics, 119(4):681–700.
Kaneko, Y., Connor, J. J., Triantafillou, T. C., and Leung, C. K. (1993b).

Fracture Mechanics Approach for Failure of Concrete Shear Keys. II: Ver-
ification. Journal of Engineering Mechanics, 119(4):701–719.
Karihaloo, B. L. (1995). Fracture Mechanics and Structural Concrete. Long-

man Scientific & Technical, Harlow.
Kjærbye, P. and Mork, P. (1998). Precast Buildings - Design and Implemen-

tation of the Precast Method. CIDB, Singapore.
Kupfer, H., Hilsdorf, H. K., and Rusch, H. (1969). Behavior of Concrete

Under Biaxial Stresses. ACI Journal, 66(8):656–666.
Leonhardt, F., Walther, R., and Dieterle, H. (1973). Versuche zur Ermit-
tlung der Tragfähigkeit von Zugschlaufenstößen. Deutscher Ausschuss für
Stahlbeton, Bulletin No. 226, Berlin, Germany.

Bibliography
Malecot, Y., Daudeville, L., Dupray, F., Poinard, C., and Buzaud, E. (2010).
Strength and damage of concrete under high triaxial loading. European
Journal of Environmental and Civil Engineering, 14(6):777–803.
Mattock, A. H. and Hawkins, N. M. (1972). Shear transfer in reinforced

concrete - recent research. PCI Journal, 17(2):55–75.
Millard, S. G. and Johnson, R. P. (1984). Shear Transfer Across Cracks

in Reinforced Concrete due to Aggregate Interlock and to Dowel Action.
Magazine of Concrete Research, 36(126):9–21.
Mirzaei, Y. (2010). Post-Punching Behavior of Reinforced Concrete Slabs.

PhD thesis, Á la Faculté Environnement Naturel, Architectural et Con-
struit, École Polytechnique Fédérale de Lausanne, Lausanne.
Mises, R. v. (1928). Mechanik der plastischen Formänderung von Kristallen.

Zeitschrift für Angewandte Mathematik und Mechanik, 8(3):161–185.
Muttoni, A. (1990). Die Anwendbarkeit der Plastizitätstheorie in der Bemes-

sung von Stahlbeton. PhD thesis, Bericht Nr. 176, Institut für Baustatik
und Konstruktion, ETH Zürich.
Muttoni, A., Joseph Schwartz, and Bruno Thürlimann (1997). Design of

Concrete Structures with Stress Fields. Birkhäuser Verlag, Basel, Switzer-
land.
Nielsen, C. V. (1998). Triaxial Behavior of High-Strength Concrete and

Mortar. ACI Materials Journal, 95(2):144–151.
Nielsen, M. P. (1969). Om jernbetonskivers styrke (English: On the strength

of reinforced concrete discs). Polyteknisk Forlag, Lyngby.
Nielsen, M. P., Bræstrup, M. W., Jensen, B. C., and Bach, F. (1978). Con-
crete Plasticity, Beam Shear - Shear in Joints - Punching Shear. Special
Publication, Danish Society for Structural Science and Engineering, Lyn-
gby.
Nielsen, M. P. and Hoang, L. C. (2011). Limit Analysis and Concrete Plas-

ticity. CRC Press, Taylor & Francis Group, Boca Raton, 3rd edition.
Nimityongskul, P. and Liu, H. Y. (1980). Vertical Shear Strength of Joints

in Prefabricated Loadbearing Walls. Housing Science, 4(2):137–157.
Ottosen, N. S. (1977). A Failure Criterion for Concrete. Journal of the

Engineering Mechanics Division, No. EM4:527–535.

Bibliography
Palaniswamy, R. and Shah, S. P. (1974). Fracture and Stress-Strain Rela-

tionship of Concrete under Triaxial Compression. Journal of the Structural
Division - ASCE, 100(No. ST5):901–916.
Pistoljevic, N. and Nielsen, M. P. (2009). HSTO-metoden. Plasticitetsteori

for armerede betonskiver. Anvendelse af homogene spændingstilstande i
trekantformede områder. Proceedings of the Danish Society for Structural
Science and Engineering, 80(2):19–54.
Pommeret, M. (1972). Le comportement sous charges ou déformations

répétées alternées des joints verticaux entre panneaux préfabriqués. Tech-
nical report, Centre expérimental de recherches et d’études du batiment
et des travaux publics, Saint Remy les Chevreuxe.
Randl, N. and Wicke, M. (2000). Schubübertragung zwischen Alt- und

Neubeton. Beton- und Stahlbetonbau, 95(Heft 8):461–473.
Rasmussen, B. H. (1963). Betonindstøbte tværbelastede boltes og dornes

bæreevne (English: Resistance of Embedded Bolts and Dowels Loaded
in Shear). Proceedings of the Danish Society for Structural Science and
Engineering, 34(2):39–55.
Richart, F., Brandtzaeg, A., and Brown, R. L. (1928). A Study of the Failure
of Concrete under Combined Compressive Stresses. University of Illinois
Bulletin No. 185, 26(12):1–104.
Smee, D. J. (1967). The effect of Aggregate Size and Concrete Strength on

the Failure of Concrete under Triaxial Compression. Civil Engineering
Transactions, The Institution of Engineers, Australia, 9(2):339–344.
Sørensen, J. H., Herfelt, M. A., Hoang, L. C., and Muttoni, A. (2018). Test
and lower bound modeling of keyed shear connections in RC shear walls.
Engineering Structures, 155:115–126.
Sørensen, J. H., Hoang, L. C., Olesen, J. F., and Fischer, G. (2017a). Tensile
capacity of loop connections grouted with concrete or mortar. Magazine
of Concrete Research, 69(17):892–904.
Sørensen, J. H., Hoang, L. C., Olesen, J. F., and Fischer, G. (2017b). Test
and analysis of a new ductile shear connection design for RC shear walls.
Structural Concrete, 18(1):189–204.
Sørensen, J. H., Hoang, L. C., Olesen, J. F., and Fischer, G. (2017c). Testing
and modeling dowel and catenary action in rebars crossing shear joints in
RC. Engineering Structures, 145:234–245.

Bibliography
St John, D. A., Poole, A. W., and Sims, I. (1998). Concrete Petrography -

A handbook of investigative techniques. Arnold, London.
Vintzeleou, E. N. and Tassios, T. P. (1986). Mathematical Models for Dowel

Action under Monotonic and Cyclic Conditions. Magazine of Concrete
Research, 38(134):13–22.
Walker, S. and Bloem, C. L. (1960). Effects of Aggregate Size on Properties

of Concrete. Journal of the American Concrete Institute, 57(9):283–298.
Walraven, J. C. and Reinhardt, H. W. (1981). Theory and Experiments

on the Mechanical Behaviour of Cracks in Plain and Reinforced Concrete
Subjected to Shear Loading. Heron, 26(1A):1–68.
Willam, K. J. and Warnke, E. P. (1975). Constitutive model for the triaxial

behaviour of concrete. Proceedings, International Association for Bridge
and Structural Engineering, Seminar on Concrete Structures Subjected to
Triaxial Stresses, Bergamo, Italy, 19(1):1–30.
Zhang, J.-P. (1997a). Diagonal cracking and shear strength of reinforced

concrete beams. Magazine of Concrete Research, 49(178):55–65.
Zhang, J.-P. (1997b). Strengh of Cracked Concrete: Part 2 - Micromechani-

cal Modelling of Shear Failure in Cement Paste and in Concrete, Series R
No 17. PhD thesis, Department of Structural Engineering and Materials,
Technical University of Denmark, Lyngby.
Zhou, X., Mickleborough, N., and Li, Z. (2005). Shear Strength of Joints in
Precast Concrete Segmental Bridges. ACI Structural Journal, 102(1):3–11.

Appendix A Results of Triaxial Tests
Appendix A
Results of Triaxial Tests
A.1 Series C16a
Table A.1: Mixture C16a, dmax = 16 mm, fc = 30.8 MPa

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
14.28 90.97 0.46 2.95
24.23 123.28 0.79 4.00
33.78 162.60 1.10 5.28
43.63 192.32 1.42 6.24
53.38 219.53 1.73 7.13
63.42 256.61 2.06 8.33
72.48 280.86 2.35 9.12
82.53 302.13 2.68 9.81
92.47 325.87 3.00 10.58

Results of Triaxial Tests Appendix A
A.2 Series C8a

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
9.45 67.15 0.23 1.63
14.48 88.38 0.35 2.15
19.20 103.80 0.47 2.52
24.32 121.32 0.59 2.94
34.07 153.47 0.83 3.73
42.25 175.15 1.03 4.25
53.57 211.77 1.30 5.14
63.03 224.13 1.53 5.44
72.68 240.68 1.76 5.84
82.62 246.62 2.01 5.99
A.3 Series C4a

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
9.65 66.36 0.22 1.50
14.57 86.42 0.33 1.95
19.40 104.57 0.44 2.37
24.23 117.05 0.55 2.65
34.17 137.48 0.77 3.11
43.23 153.63 0.98 3.48
53.97 164.77 1.22 3.73
72.78 180.68 1.65 4.09
82.33 187.93 1.86 4.25

A.4 Series C2a

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
7.04 50.77 0.19 1.35
9.60 61.80 0.25 1.64
11.87 69.12 0.31 1.83
14.43 82.08 0.38 2.18
19.30 95.31 0.51 2.53
24.13 112.85 0.64 2.99
34.09 142.15 0.90 3.77
43.45 155.36 1.15 4.12
53.28 164.46 1.41 4.36
62.68 173.90 1.66 4.61
72.78 184.58 1.93 4.90
82.33 189.83 2.18 5.04

A.5 Series C16b
Table A.5: Mixture C16b, dmax = 16 mm, fc = 29.8 MPa

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
5.11 53.06 0.17 1.78
8.05 64.92 0.27 2.18
11.12 79.19 0.37 2.66
15.08 91.72 0.51 3.08
20.14 105.79 0.68 3.55
25.61 116.51 0.86 3.91
30.32 130.36 1.02 4.38
34.98 146.49 1.17 4.92
40.34 160.07 1.35 5.37
45.02 170.23 1.51 5.71
50.10 184.96 1.68 6.07
55.37 201.87 1.86 6.48
60.18 212.36 2.02 6.81
70.05 225.88 2.35 7.51
80.24 257.80 2.69 8.11

A.6 Series C8b

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
5.13 42.07 0.19 1.60
7.97 55.29 0.30 2.10
10.27 62.35 0.39 2.37
11.00 63.34 0.42 2.41
15.08 75.70 0.57 2.88
20.08 87.50 0.76 3.33
25.12 99.25 0.95 3.77
30.41 110.67 1.16 4.21
35.16 118.02 1.34 4.49
40.12 134.21 1.52 5.10
45.16 139.37 1.72 5.30
50.11 150.01 1.90 5.70
56.51 159.13 2.15 6.05
60.51 169.61 2.30 6.45
70.15 180.93 2.67 6.88

A.7 Series C4b

σ1 σ3
σ1 [MPa] σ3 [MPa] [-] [-]
fc fc
5.18 40.97 0.19 1.50
8.01 52.12 0.29 1.91
10.95 60.95 0.40 2.23
15.03 68.06 0.55 2.49
20.12 85.55 0.74 3.13
25.05 92.78 0.92 3.39
30.11 106.43 1.10 3.89
35.18 112.11 1.29 4.10
39.99 118.03 1.46 4.31
45.23 124.51 1.65 4.55
50.42 134.54 1.84 4.92
54.95 142.07 2.01 5.19
60.10 148.00 2.20 5.41

Appendix B Summary of Push-off Tests on Shear Connections
Appendix B
Summary of Push-off Tests on

Shear Connections
Figure B.1 shows the layout of the test specimens, including definition of
geometric parameters and the three different loop configurations tested.
P
A-A (see Figure 6.1)
Precast D
Locking bar, φL element
(b) Conventional
Lk Ak s B-B D
hk L s1
t s1
dk
(c) ’1-on-2’ connection

Lk B-B
Grout B-B D
Precast b
element s1
a
s1
P
(a) Test setup (loop layout of new design) (d) ’2-on-2’ connection
Figure B.1: General layout of the push-off specimens

Summary of Push-off Tests on Shear Connections Appendix B
In Figure B.2(a) the test arrangement is schematically presented. The

piston applied load to one of the precast elements, and the other precast
element was supported by a steel cylinder (pinned support). The setup
consisted of a steel frame build with U-profiles on a strong floor.
In Figure B.2(b) a picture of a specimen placed in the test arrangement is
shown. In the picture the position of the displacement pins can be seen. Four
LVDT’s were used to measure the longitudinal and transverse displacements
in each end of the connection. In addition, the position of the digital camera
used for collection of data for the digital image correlation analysis can be
seen.
(a) Schematic view of test setup
(b) Specimen placed in test setup before test
Figure B.2: Test setup with (a) schematic view of the test arrangement and
(b) picture of a specimen placed in the test setup before testing

B.1 Reference Specimens

The specimens had a thickness of t = 150 mm and contained three shear
keys and four U-bars loop connections. The U-bars in the loop connections
were placed closely together with the conventional layout, see Figure B.1(b).
The U-bars had an internal bend diameter of D = 60 mm and the loop
connections were placed outside the indentations with a mutual distance of
s = 300 mm. Specimens R1, R2, and R3 were included in Paper II. The
strengths of the reinforcement were:
• 8 mm U-bars: fy = 509 MPa and fu = 650 MPa
• 12 mm locking bar: fyL = 596 MPa and fuL = 697 MPa
The connections had a width of b = 80 mm, a total length of L = 1280 mm, a

key length of Lk = 160 mm, and a key height of hk = 85 mm. The interfaces
were greased before casting the joint grout. The tests were conducted in
relation to student project e.
Table B.1: Tests on reference specimens, conventional design (’1-on-1’ con-

figuration)
ID φ [mm] fc [MPa] dk [mm] Φ [-] ΦL [-] PFP [kN] PU [kN]
R1 8 34.6 16 0.145 0.048 282.43 -

R2 8 35.7 16 0.140 0.046 303.80 -
R3 8 35.7 16 0.140 0.046 337.42 -
R4 8 44.9 16 0.117 0.037 331.75 -
R5 8 44.9 16 0.117 0.037 300.36 -
R6 8 44.9 16 0.117 0.037 310.02 -
B.2 Pilot Specimens (’1-on-2’ configuration)

The specimens had a thickness of t = 200 mm and contained three shear
keys and four U-bar loop connections. The U-bars in the loop connections
were arranged as a ’1-on-2’ configuration using the new design, see Figure
B.1(c). The U-bars had an internal bend diameter of D = 60 mm and the
distance between the U-bars was s1 = 35 mm. In the longitudinal direction,
the loop connections were placed with a mutual distance of s = 300 mm.
Tests were performed using both 8 and 10 mm U-bars. Specimens P1-P4
were included in Paper II.

The strengths of the reinforcement were:
• 12 mm double T-headed lacer bar: fy,Lacer = 552 MPa
The connections had a width of b = 80 mm, a total length of L = 1280 mm, a

key length of Lk = 160 mm, and a key height of hk = 85 mm. The interfaces
were greased before casting the joint grout. The tests were conducted in
relation to student project e.
Table B.2: Tests on the new design, ’1-on-2’ configuration
S1 8 38.1 - - - 134.24 221.72

S2 8 38.1 - - - 140.88 234.15
S3 8 43.0 - - - 129.96 243.29
S4 8 43.0 - - - 127.43 212.06
P1 8 38.1 16 0.132 - 344.24 357.45
P2 8 38.1 16 0.132 - 347.04 368.12
P3 8 42.7 16 0.117 - 342.49 339.97
P4 8 42.7 16 0.117 - 331.42 324.49
P5 8 38.5 30 0.130 - 341.15 324.71
P6 8 38.5 30 0.130 - 358.52 315.93
P7 8 47.7 30 0.105 - 378.09 320.97
P8 8 47.7 30 0.105 - 370.63 300.09
P9 10 42.7 16 0.212 - 398.78 489.66

P10 10 42.7 16 0.212 - 421.68 473.54
P11 10 47.7 30 0.190 - 432.20 454.42
P12 10 46.5 30 0.194 - 439.96 471.37

B.3 New Design - 6 mm U-bars

The specimens had a thickness of t = 200 mm and contained three shear keys
and four U-bar loop connections. The U-bars in the loop connections had a
diameter of 6 mm and were arranged as a ’2-on-2’ configuration using the
new design, see Figure B.1(d). The U-bars had an internal bend diameter
of D = 45 mm, the distance between the outermost U-bars was s1 = 30
mm and the distance between the innermost U-bars was a = 40 mm. In
the longitudinal direction, the loop connections were placed with a mutual
distance of s = 300 mm. Details on the test specimens can also be found in
Paper III. The strengths of the reinforcement were:
The connections had a width of b = 80 mm, a total length of L = 1280 mm,

a key length of Lk = 120 mm, and a key height of hk = 200 mm (equal to
the thickness of the panel). The interfaces were not treated before casting
the joint grout. The tests were conducted in relation to student project f .
Table B.3: Tests on the new design, ’2-on-2’ configuration with 6 mm U-bars
D10A 6 44.6 10 0.073 0.084 448.56 417.38

D10B 6 44.6 10 0.073 0.084 448.62 417.24
D12A 6 44.6 12 0.073 0.084 471.74 437.98
D12B 6 44.6 12 0.073 0.084 496.36 452.90
D14A 6 44.6 14 0.073 0.084 510.91 449.09
D14B 6 44.6 14 0.073 0.084 519.16 462.70
D16A 6 44.6 16 0.073 0.084 543.30 421.02
D16B 6 44.6 16 0.073 0.084 541.57 452.24
D18A 6 42.0 18 0.077 0.090 540.73 380.72
D18B 6 42.0 18 0.077 0.090 537.50 412.51
D20A 6 42.0 20 0.077 0.090 526.62 431.17
D20B 6 42.0 20 0.077 0.090 517.03 398.51


distance of s = 300 mm. Details on the test specimens can also be found in
Paper II. The strength parameters of the reinforcement were:
The connections had a width of b = 100 mm and a total length of L = 1280

mm. The interfaces were not treated before casting the joint grout.
Table B.4 contains a test series where the key length was varied. The
key height was hk = 100 mm and the key depth was dk = 28 mm. The tests
were conducted in relation to student project c.
and with a varying key length
ID φ [mm] fc [MPa] Lk [mm] Φ [-] ΦL [-] PFP [kN] PU [kN]
I1 8 31.2 120 0.349 0.059 379.02 441.21

I2 8 34.2 120 0.318 0.054 416.59 472.92
II1 8 31.2 140 0.299 0.050 366.40 463.79
II2 8 34.2 140 0.273 0.046 414.46 462.48
III1 8 31.2 160 0.262 0.044 393.04 494.70
III2 8 34.2 160 0.239 0.040 473.52 514.87
IV1 8 31.2 180 0.232 0.039 439.44 470.89
IV2 8 34.2 180 0.212 0.036 478.17 515.31

Table B.4 contains a test series where the key depth was varied. The key
length was Lk = 140 mm and the key height was hk = 200 mm (equal to
the thickness of the panel). The tests were conducted in relation to student
project d.
and with a varying key depth
V1 8 31.2 10 0.149 0.025 475.24 488.97

V2 8 34.2 10 0.136 0.023 492.86 535.61
VI1 8 30.6 16 0.152 0.026 527.09 502.55
VI2 8 30.6 16 0.152 0.026 523.82 550.98
VII1 8 30.6 20 0.152 0.026 549.17 451.58
VII2 8 30.6 20 0.152 0.026 524.46 527.85
VIII1 8 30.6 25 0.152 0.026 507.05 528.67
VIII2 8 30.6 25 0.152 0.026 516.97 545.33
IX1 8 30.6 28 0.152 0.026 526.53 534.19
IX2 8 30.6 28 0.152 0.026 527.59 527.07


distance of s = 320 mm. The strength parameters of the reinforcement were:

The connections had a width of b = 120 mm, a total length of L = 1340

mm, a key depth of dk = 10 mm, and a key height of hk = 200 mm. Half the
specimens were cast with a mortar (denoted M) and the remaining were cast
with concrete (denoted C). The interfaces were not treated before casting
the joint grout. The tests were conducted in relation to student project g.
Table B.6: Tests on the new design, ’2-on-2’ configuration, 10 mm U-bars
ID φ [mm] fc [MPa] Lk [mm] Φ [-] ΦL [-] PFP [kN] PU [kN]
M120A 10 42.0 120 0.205 0.022 654.51 745.23

M120B 10 42.0 120 0.205 0.022 646.88 745.98
M150A 10 43.3 150 0.159 0.017 646.43 757.01
M150B 10 44.6 150 0.155 0.017 669.62 752.28
M180A 10 42.0 180 0.137 0.015 607.25 720.64
M180B 10 42.0 180 0.137 0.015 639.55 748.52
C120A 10 41.8 120 0.206 0.023 605.52 689.11

C120B 10 41.8 120 0.206 0.023 676.15 734.61
C150A 10 41.8 150 0.165 0.018 637.77 736.94
C150B 10 41.8 150 0.165 0.018 658.19 733.13
C180A 10 41.8 180 0.138 0.015 689.76 749.10
C180B 10 41.8 180 0.138 0.015 642.81 686.82

Appendix C Upper Bound Calculations for Shear Connections
Appendix C
Upper Bound Calculations for

Shear Connections
C.1 Reference Specimens
Table C.1: Test and upper bound results, reference specimens designed with
the conventional design
PFP Key failure mode

ID PFP [kN] Pcal [kN] ν
Pcal Observed/Predicted
R1 282.43 286.20 0.99 0.45 Corner (E)/Corner (C)

Mean 1.06
Standard deviation 0.07

Upper Bound Calculations for Shear Connections Appendix C
C.2 Pilot Specimens (’1-on-2’ configuration)

It should be noted that the test results were only compared to the calcula-
tions of Mechanisms A, B, and C.
Table C.2: Test and upper bound results, new design with ’1-on-2’ loop con-
figuration

P1 344.24 291.12 1.18 0.43 Corner (C)/Corner (C)

P5 341.15 301.06 1.13 0.42 Cut off (B)/Cut off (B)
Mean 1.16

Mean 1.03

C.3 New Design with ’2-on-2’ Configuration

- 6 mm U-bars
figuration and 6 mm U-bars

D10A 448.56 393.34 1.14 0.44 Corner (E)/Corner (C)

D10B 448.62 393.34 1.14 0.44 Corner (E)/Corner (C)
D12A 471.74 425.08 1.11 0.44 Corner (E)/Corner (C)
D12B 496.36 425.08 1.17 0.44 Corner (E)/Corner (C)
D14A 510.91 450.16 1.13 0.44 Corner (E)/Corner (E)
D14B 519.16 450.16 1.15 0.44 Corner (E)/Corner (E)
D16A 543.30 471.83 1.15 0.44 Cut off (B or D)/Corner (E)
D16B 541.57 471.83 1.15 0.44 Corner (E)/Corner (E)
D18A 540.73 464.55 1.16 0.45 Cut off (B or D)/Cut off (D)
D18B 537.50 464.55 1.16 0.45 Cut off (B or D)/Cut off (D)
D20A 526.62 464.55 1.13 0.45 Cut off (B or D)/Cut off (D)
D20B 517.03 464.55 1.11 0.45 Cut off (B or D)/Cut off (D)
Mean 1.14

Upper Bound Calculations for Shear Connections Appendix C

- 8 mm U-bars

I1 379.02 395.34 0.96 0.52 Cut off (B or D)/Cut off (A)

I2 416.59 403.29 1.03 0.50 Cut off (B or D)/Cut off (A)
II1 366.40 412.67 0.89 0.49 Cut off (B or D)/Cut off (A)
II2 414.46 421.43 0.98 0.47 Cut off (B or D)/Cut off (A)
III1 393.04 427.62 0.92 0.47 Cut off (B or D)/Cut off (D)
III2 473.52 433.99 1.09 0.45 Cut off (B or D)/Cut off (D)
IV1 439.44 438.33 1.00 0.45 Cut off (B or D)/Cut off (D)
IV2 478.17 455.20 1.07 0.43 Cut off (B or D)/Cut off (D)
Mean 0.99
V1 475.24 500.73 0.95 0.49 Corner (E)/Corner (C)

V2 492.86 508.21 0.97 0.47 Corner (E)/Corner (C)
VI1 527.09 538.21 0.98 0.50 Corner (E)/Cut off (D)
VI2 523.82 538.50 0.97 0.50 Cut off (B or D)/Cut off (D)
VII1 549.17 538.50 1.02 0.50 Cut off (B or D)/Cut off (D)
VII2 524.46 538.50 0.97 0.50 Cut off (B or D)/Cut off (D)
VIII1 507.05 538.50 0.94 0.50 Cut off (B or D)/Cut off (D)
VIII2 516.97 538.50 0.96 0.50 Cut off (B or D)/Cut off (D)
IX1 526.53 538.50 0.98 0.50 Cut off (B or D)/Cut off (D)
IX2 527.59 538.50 0.98 0.50 Cut off (B or D)/Cut off (D)
Mean 0.97


- 10 mm U-bars

M120A 654.51 687.92 0.95 0.45 Corner (E)/Cut off (D)

M120B 646.88 687.92 0.94 0.45 Corner (E)/Cut off (D)
M150A 646.43 713.93 0.91 0.41 Corner (E)/Corner (E)
M150B 669.62 716.30 0.93 0.40 Corner (E)/Corner (E)
M180A 607.25 701.43 0.87 0.39 Corner (E)/Corner (E)
M180B 639.55 701.43 0.91 0.39 Corner (E)/Corner (E)
Mean 0.92
C120A 605.52 800.14 0.76 0.53 Corner (E)/Cut off (D)

C120B 676.15 800.14 0.85 0.53 Corner (E)/Cut off (D)
C150A 637.77 840.36 0.76 0.49 Corner (E)/Cut off (D)
C150B 658.19 840.36 0.78 0.49 Corner (E)/Cut off (D)
C180A 689.76 840.68 0.82 0.46 Corner (E)/Corner (E)
C180B 642.81 840.68 0.76 0.46 Corner (E)/Corner (E)
Mean 0.79

Appendix D Lower Bound Calculations for Shear Connections
Appendix D
Lower Bound Calculations for

Shear Connections
D.1 Reference Specimens
Table D.1: Test and lower bound results, reference specimens designed with
the conventional design
PFP
ID PFP [kN] Pcal,1 [kN] Pcal,2 [kN] Stress component
Pcal
R1 282.43 216.88 189.76 1.30 σ2,I
R2 303.80 221.76 193.70 1.37 σ2,I
R3 337.42 221.76 193.70 1.52 σ2,I
R4 331.75 261.08 226.20 1.27 σ2,I
R5 300.36 261.08 226.20 1.15 σ2,I
R6 310.02 261.08 226.20 1.19 σ2,I
Mean 1.30

Lower Bound Calculations for Shear Connections Appendix D
D.2 Pilot Specimens (’1-on-2’ configuration)
Table D.2: Test and lower bound results, new design with ’1-on-2’ loop con-
figuration
PFP
Pcal
P1 344.24 232.31 202.25 1.48 σ2,I
P2 347.04 232.31 202.25 1.49 σ2,I
P3 342.49 252.01 218.49 1.36 σ2,I
P4 331.42 252.01 218.49 1.31 σ2,I
P5 341.15 258.75 321.93 1.06 σ2,II
P6 358.52 258.75 321.93 1.11 σ2,II
P7 378.09 271.01 375.18 1.01 σB
P8 370.63 271.01 375.18 0.99 σB
Mean 1.23
P9 398.78 301.65 270.53 1.32 σA,1

P10 421.68 301.65 270.53 1.40 σA,1
P11 432.20 391.06 404.99 1.07 σ2,II
P12 439.96 388.01 393.93 1.12 σ2,II
Mean 1.23

D.3 New Design with ’2-on-2’

Configuration - 6 mm U-bars
PFP
Pcal
D10A 448.56 279.90 372.35 1.20 σ2,II
D10B 448.62 279.90 372.35 1.20 σ2,II
D12A 471.74 279.90 413.55 1.14 σ2,II
D12B 496.36 279.90 413.55 1.20 σ2,II
D14A 510.91 279.90 453.22 1.13 σ2,II
D14B 519.16 279.90 453.22 1.15 σ2,II
D16A 543.30 279.90 491.21 1.11 σ2,II
D16B 541.57 279.90 491.21 1.10 σ2,II
D18A 540.73 277.69 487.25 1.11 σB
D18B 537.50 277.69 487.25 1.10 σB
D20A 526.62 277.69 496.94 1.06 σB
D20B 517.03 277.69 496.94 1.04 σB
Mean 1.13

Lower Bound Calculations for Shear Connections Appendix D

PFP
Pcal
I1 379.02 241.05 329.60 1.15 σB
I2 416.59 250.12 345.90 1.20 σB
II1 366.40 290.77 371.53 0.99 σB
II2 414.46 300.32 385.73 1.07 σB
III1 393.04 337.40 400.94 0.98 σB
III2 473.52 347.52 419.16 1.13 σB
IV1 439.44 381.45 446.18 0.98 σB
IV2 478.17 392.20 468.64 1.02 σB
Mean 1.07
V1 475.24 385.06 341.04 1.23 σA,1

V2 492.86 392.01 366.08 1.26 σA,1
VI1 527.09 383.57 481.51 1.09 σ2,III
VI2 523.82 383.57 481.51 1.09 σ2,III
VII1 549.17 383.57 539.51 1.02 σ2,II
VII2 524.46 383.57 539.51 0.97 σ2,II
VIII1 507.05 383.57 584.82 0.87 σB
VIII2 516.97 383.57 584.82 0.88 σB
IX1 526.53 383.57 599.54 0.88 σB
IX2 527.59 383.57 599.54 0.88 σB
Mean 1.02


PFP
Pcal
M120A 654.51 414.62 404.82 1.58 σA,1
M120B 646.88 414.62 404.82 1.56 σA,1
M150A 646.43 533.22 470.63 1.21 σA,1
M150B 669.62 536.65 481.82 1.25 σA,1
M180A 607.25 638.02 535.78 0.95 σA,1
M180B 639.55 638.02 535.78 1.00 σA,1
Mean 1.26
C120A 605.52 414.12 403.35 1.46 σA,1

C120B 676.15 414.12 403.35 1.63 σA,1
C150A 637.77 529.07 457.67 1.21 σA,1
C150B 658.19 529.07 457.67 1.24 σA,1
C180A 689.76 637.38 533.72 1.08 σA,1
C180B 642.81 637.38 533.72 1.01 σA,1
Mean 1.27

Appended Papers
203
Paper I
”Tensile capacity of loop connections grouted with concrete or
mortar”
Jesper H. Sørensen, Linh C. Hoang, John F. Olsen, Gregor Fischer
Published in: Magazine of Concrete Research (2017), 69(17):892-904
205
Magazine of Concrete Research Magazine of Concrete Research, 2017, 69(17), 892–904
http://dx.doi.org/10.1680/jmacr.16.00466
Volume 69 Issue 17
Paper 1600466
Tensile capacity of loop connections Received 31/10/2016; revised 18/01/2017; accepted 02/03/2017
Published online ahead of print 06/04/2017
grouted with concrete or mortar
Keywords: joints/precast concrete/testing, structural elements
Sørensen, Hoang, Olesen and Fischer
ICE Publishing: All rights reserved
Tensile capacity of loop connections grouted

with concrete or mortar
Jesper Harrild Sørensen John Forbes Olesen
PhD student, Department of Civil Engineering, Technical University of Associate Professor, Department of Civil Engineering, Technical University
Denmark, Kongens Lyngby, Denmark of Denmark, Kongens Lyngby, Denmark
Linh Cao Hoang Gregor Fischer
Professor, Department of Civil Engineering, Technical University of Denmark, Associate Professor, Department of Civil Engineering, Technical University
Kongens Lyngby, Denmark of Denmark, Kongens Lyngby, Denmark
This paper presents a study of grout failure in symmetric U-bar loop connections loaded in tension, with focus on
the performance of two grouting materials – concrete and mortar. The study contains an experimental investigation
as well as a rigid-plastic modelling of the tensile capacity. The test specimens consisted of symmetric ‘2-on-2’ loop
connections transversely reinforced with a double T-headed rebar. The amount of transverse reinforcement was
varied, including the limiting cases of specimens with no transverse reinforcement, as well as connections with
sufficient transverse reinforcement to allow yielding of the U-bars. The experimental work showed that connections
grouted with concrete performed better than the connections grouted with mortar. In the theoretical models, the
difference in tested capacity is explained by the difference in the internal angle of friction and in the softening
behaviour of concrete as compared with mortar.
WE rate of external work
Notation
WCI rate of internal work from concrete
Ac circular area of U-bar overlap
WSI rate of internal work from steel
AI area of failure plane in U-bar overlap
w/c ratio between water and cement
Ai area of failure plane
α angle between u and yield line
Al area of tensile failure plane
β inclination of yield line to vertical
As reinforcement area of U-bars
ν effectiveness factor, concrete in compression
AsL reinforcement area of lacer bar
νt effectiveness factor, concrete in tension
At area of tensile failure plane
σc concrete stress
a distance between innermost loops
ΦL reinforcement degree, lacer bar
a/c ratio between aggregates and cement
ϕ diameter of reinforcement
b width of joint
ϕL diameter of lacer bar
D internal bend diameter of U-bar
φ internal angle of friction
dmax maximum aggregate size
fc compression strength of grout
fcm mean compression strength of grout
ft tensile strength of grout Introduction
fu ultimate strength of reinforcement This paper presents an investigation into the structural behav-
fy yield strength of reinforcement iour of so-called ‘2-on-2’ loop connections between precast
fyL yield strength of lacer reinforcement concrete elements loaded in tension. The investigation includes
H overlapping length of U-bars rigid-plastic modelling of the tensile capacity as well as experi-
K factor influencing ν mental tests.
l relation between ft and fc
m relation between ft and fc The paper has been prepared as a self-contained research
N tensile load on loop connection article on the topic of loop connections loaded in tension.
Nu ultimate capacity of loop connection However, there is a strong correlation with the development of
Ny yielding capacity of U-bars construction-friendly and ductile shear joints for connection
N0 cracking load of loop connection of precast shear wall elements. A new design of keyed shear
s distance between U-bars wall connections has recently been tested by the authors
ua displacement vector of segment IIa (Sørensen et al., 2017). The principle of the design is illustrated
ub displacement vector of segment IIb in Figure 1(a), which shows two precast wall elements with
ul displacement in the l-direction indented interfaces connected by a joint grouted with either
ut displacement in the t-direction concrete or mortar. The significance of the new design is the
V shear load orientation of the overlapping U-bar loops, which are placed
892
Magazine of Concrete Research Tensile capacity of loop connections
Volume 69 Issue 17 grouted with concrete or mortar
Offprint provided courtesy of www.icevirtuallibrary.com

Author copy for personal use, not for distribution
Horizontal section
Vertical section
Vertical
section
Strut
U-bar loop
action
V V Precast
element
Horizontal section
σc
Lacer bar
(a) (b)
Figure 1. Keyed joint between shear wall elements (see Sørensen et al. (2017)) with ‘2-on-2’ symmetrical loop connections
in the same plane as that of the wall elements. This is different according to the principles presented, the main benefits being
from the classical design (fib, 2008; Hansen et al., 1974) where the narrow connection zone required and the ability to esti-
the plane of the loops is perpendicular to the plane of the mate the capacity of the grout, leading to a ductile design gov-
wall elements. As shown in Figure 1(b), each loop connection erned by yielding of the reinforcement.
in the shear joint is reinforced transversely with a double
T-headed rebar (in the following, termed a lacer bar), in order This study focuses on two main design variables: (a) the dimen-
to enhance the ability of the overlapping U-bars to transfer tension of the transverse reinforcement placed in the loops
sion between the precast elements. As indicated in Figure 1(a), and (b) the composition of the grout material (concrete and
the ability to transfer tension is in fact a prerequisite for the mortar). Ideally, concrete should be used in practice; however,
joint to transfer shear by way of interlock action in the keys for the narrow joints typically found in building structures,
(the horizontal component of the diagonal compressive struts mortar can be a more practical and construction-friendly sol-
formed in the grout between opposite keys must be balanced ution. The behaviour of connections grouted with mortar must
by horizontal tension transfer in the U-bar loops). The test therefore be studied as well.
results reported by Sørensen et al. (2017) showed that this
new design is able to carry larger shear loads and performs in
a much more ductile manner than the classical solution, pro- Previous investigations on loop connections
vided that the U-bars are stressed to yielding. Therefore, in Regarding tensile action in loop connections, Gordon (2006)
order to ensure ductile shear behaviour of the new joint design, introduced the term symmetric unit to describe a connection
it is necessary to develop a method regarding how to detail the where the main reinforcement is placed symmetrically about
layout of the U-bar loop connections, so that the core of grout the longitudinal axis. Gordon tested symmetric ‘4-on-3’ con-
becomes strong enough to enable tensile yielding of U-bars nections as well as non-symmetric connections, and concluded
and prevent premature concrete/mortar failure. that the non-symmetric specimens tended to experience rota-
tions of the joint concrete and to fail at a load level lower
To develop such designs, it is essential to carry out tests on than expected. Therefore, when loop connections are used to
specimens where concrete/mortar failure actually takes place, transfer tensile action, the symmetry condition should be ful-
and to develop models that can estimate the tensile capacity filled. Leonhardt et al. (1973) were the first to publish results
related to failure of the grout materials. It is in this broader of tension tests of symmetric ‘2-on-2’ loop connections (i.e. two
context that the present experimental programme and the theo- U-bars overlapping two other U-bars). In their programme,
retical study should be viewed. However, the design method the overlapping U-bars were placed closely together with no
is not limited to application in shear walls only. A number spacing in between. The specimens were cast in one sequence
of tensile connections in structural concrete can be designed without casting joints, and transverse reinforcement was not
893

provided. The influence of the overlapping area of the U-bar authors’ knowledge, not been previously published in the
loops on the failure load was investigated. To obtain yielding literature. A further advance, compared to earlier tension tests
of the U-bars, Leonhardt et al. (1973) suggested a design on loop connections, is the use of a double T-headed bar as
where the overlapping concrete area is larger than a circular transverse reinforcement. The bar provides increased ancho-
area with a diameter of 15ϕ. The design proposal included rage (DeVries et al., 1999; Ghali and Youakim, 2005;
transverse reinforcement placed inside the overlapping area. Thompson et al., 2006), even in cracked concrete (Brantschen
et al., 2016), which is required as the joint is rather narrow.
Hao (2004) and Ong et al. (2006) also tested symmetric
designs. Their experimental programme consisted of ‘2-on-1’
designs, and their calculations were based on a strut-and-tie Experimental programme
model. However, they concluded that the model was only The experimental programme comprised 32 specimens, each
applicable for similar ‘2-on-1’ designs, as the model was cali- consisting of two precast concrete elements joined by a sym-
brated with these tests only. Jørgensen and Hoang (2013) metric ‘2-on-2’ loop connection, see Figure 2. The overlapping
tested ‘3-on-2’ symmetric specimens to investigate the influence area was transversely reinforced with lacer reinforcement in
of a number of factors, including the spacing of the over- the shape of a double T-headed bar. Three series with different
lapping U-bars. They proposed an upper bound plasticity compositions of grouting material were investigated, see
model for the tensile capacity of the connection, and later the Table 1. The connections in series C were grouted with con-
solution was extended to include a combination of tension and crete (maximum aggregate size, dmax = 16 mm) while in series
bending action (Jørgensen and Hoang, 2015). M and T, mortars with dmax = 2 mm and 4 mm, respectively,
were used. Table 2 provides the relative proportions of the con-
Regarding requirements for transverse reinforcement in the joint, crete and mortar mixtures. It should be noted that the batches
Gordon (2006) also examined the use of fibre-reinforced con- of mortar in series M were produced simply by removing the
crete as replacement for conventional reinforced concrete. He content of the larger aggregates from the concrete recipe of
concluded that the fibre-reinforced material could not replace series C, without considering the cement to aggregate ratio. In
steel reinforcement as some ductility was lost. Also Araújo et al. series C and M, the lacer bar was placed in the centre of
(2014) tested loop connections with fibre-reinforced concrete as the circle formed by the overlapping U-bars, see Figures 2 and
replacement for steel reinforcement; however, they only tested a 3(a). In series T, as an alternative, the lacer bar was fixed to
non-symmetric ‘1-on-1’ connection design. the looped end of the outermost U-bars, as illustrated in
Figure 3(b). The idea behind this was to consider the circular
A comparative study of the tension behaviour of symmetric core of mortar inside the connection as a small beam subjected
‘2-on-2’ loop connections grouted with concrete and with to four-point bending, having the lacer bar as a fully anchored
mortar, as presented in the following, has to the best of the tension reinforcement.
1000
450 100 450
∅8 U-bars in loop, ‘2-on-2’ Grout K25 reinforcing bar
K8/100
200
∅25 threaded bar Double T-headed bar Ribs removed for better grip
200
D = 60
Internal bend diameter of U-bar loops ∅8 U-bars in loop, ‘2-on-2’
Figure 2. General geometry of test specimen, including reinforcement details (dimensions: mm)
894

The diameter of the lacer bars ranged from 12 to 20 mm, see (0, 12, 14, 16 and 20) adjoining the series letter (C, M and T)
Figure 3(c). Some specimens were also tested without a lacer refers to the diameter of the lacer reinforcement, while ‘a’, ‘b’
bar. The cross-sectional diameter (8 mm) and the internal and ‘c’ denote the replicates.
bend diameter (60 mm) of the U-bars were kept constant
throughout all three series. The yield strength of the U-bars Test set-up and testing procedure
was tested in uniaxial tension to an average of fy = 550 MPa. The precast elements were designed with a reinforcement bar
The average ultimate strength was found to be fu = 677 MPa. centrally placed in the axial direction. In one element, ordinary
Table 3 contains the remaining properties of each specimen. ribbed reinforcement was used, where the ribs on the part pro-
For statistical reasons, three replicates of each connection truding from the element were removed for a better grip in the
design were cast. However, some unsuccessful castings had testing machine, see Figure 2. In the other element, a threaded
to be omitted because the lacer bars accidentally moved out steel rod was used in combination with a specially designed
of position during casting and ended in a position with eccen- load application device, see Figure 4. The tensile load was
tricity to the axial system line of the specimen. The number applied through a spherical nut.
Table 1. Test series in experimental programme

The tests were performed as quasi-static deformation con-
trolled tests in a servo-controlled hydraulic testing machine
No. of Grout Max aggregate with a capacity of 500 kN. The loading rate was 0·5 mm
Series specimens material size, dmax: mm
piston movement per minute. The relative displacements across
C 13 Concrete 16 the connection were measured on all four surfaces using linear
M 10 Mortar 2 variable differential transducers (LVDTs). The measured rela-
T 9 Mortar 4
tive axial displacements were thereby related to the behaviour
of the grouted connection.
Table 2. Mixture proportions (kg) used in the three series
Experimental results and observations
Series C Series M Series T
The test results are presented as load–displacement relationships,
Cement 340 340 375 where the displacement is taken as the average elongation of the
Water 180 180 170
connection measured by the LVDTs. The ultimate load, Nu, of
Aggregate 0–2 mm 0 884 0
Aggregate 0–4 mm 884 0 1400 the connection is taken as the maximum load observed within
Aggregate 4–8 mm 244 0 0 the test range of 0–10 mm relative axial displacement. Table 3
Aggregate 8–16 mm 738 0 0 contains the recorded Nu for all test specimens.
Fly ash 0 0 28·5
Plasticiser 0 0 1·0
Figure 5 shows the tested load–displacement curves for the
Water-to-cement ratio, w/c 0·53 0·53 0·45
Aggregate-to-cement ratio, a/c 5·49 2·60 3·73 specimens in series C and M. One single specimen in series
C14 experienced unloading due to an interruption of hydraulic
20 mm
16 mm
12 mm
14 mm
(a) (b) (c)
Figure 3. Positioning of lacer reinforcement before casting of grout (a) and (b) and diameter of lacer bars (c): (a) series C and M;
(b) series T; (c) lacer bars
895

Table 3. Material properties and reinforcement details of test specimens

Grout Lacer
Specimens Material fc: MPa dmax: mm ϕL: mm fyL: MPa Nu: kN
a 68·7
C0 b Concrete 44·1 16 — — 62·4
c 66·8
a 97·6
C12 b Concrete 44·1 16 12 552 89·2
c 95·1
a 108·8
C14 b Concrete 39·9 16 14 562 101·1
c 109·1
C16 a Concrete 41·3 16 16 563 112·5
b 103·1
C20 a Concrete 41·3 16 20 564 129·6
b 126·7
a 47·7
M0 b Mortar 36·6 2 — — 48·9
c 43·4
M12 a Mortar 36·6 2 12 552 76·3
b 77·9
M14 a Mortar 39·5 2 14 562 79·3
M16 a Mortar 37·4 2 16 563 82·3
a 108·6
M20 b Mortar 37·4 2 20 564 116·0
c 96·5
a 73·6
T12 b Mortar 39·5 4 12 552 77·6
c 78·4
a 76·4
T14 b Mortar 39·5 4 14 562 79·2
c 88·0
a 88·2
T16 b Mortar 39·5 4 16 563 97·2
c 103·8
pressure during the test. As the test was resumed, the specimen The test results revealed that the choice of grout material influ-
continued the expected load path of an uninterrupted test. enced the behaviour significantly: see, for example, Figure 5.
From the plots in Figure 5 it can be clearly seen that connec- The grout material with larger aggregate content and larger
tions grouted with concrete in general were able to carry a aggregate sizes leads to a stiffer response after crack initiation
higher ultimate load than a similar connection (i.e. with the and a higher ultimate capacity compared to the grouts with
same lacer bar configuration) grouted with mortar. It can also smaller aggregates. Furthermore, the concrete grout has a
be observed that only the specimens grouted with concrete slightly better ability to sustain the load after peak compared
and reinforced transversely with a 20 mm lacer bar behaved to the mortar material.
in a significantly ductile manner. For these particular speci-
mens, the ultimate load was governed by yielding of the 8 mm For both grout materials, the connections were uncracked and
U-bars. The remaining specimens experienced failure of the behaved elastically at low load levels. At a load level correspon-
grout material without yielding of the U-bars. ding to the capacity of a connection without lacer bar, a crack
became visible on the surface of the connection. The crack
For specimens without transverse reinforcement (i.e. no lacer typically developed in the axial direction – see Figure 6(a) –
bar in series C0 and M0) the load-carrying capacity was and appeared between the innermost U-bars. The appearance
governed by cracking of the grout material, and the load– of this crack on the surface was, however, not representative for
displacement curves displayed significant softening in the post- the crack pattern throughout the thickness of the connection,
peak regime. For specimens with transverse reinforcement, on as the cracking behaviour was governed by a complex state of
the other hand, cracking of the grout material activated the stress between the U-bars. In some tests, the crack parallel to
lacer bar, which enabled further increase of the load and the the axial direction was accompanied by inclined cracks on the
post-peak softening behaviour was less pronounced than same surface. The cracks were followed by an opening of the
observed for specimens without lacer bars. interface between the joint grout and the precast element
896

24 mm bolt for load application 140

C20
A–A C16
120 C14
C12
12 mm C0
bolt 100
200
A–A A–A
Force: kN
80
Spherical
nut
60
200
25 mm threaded bar
40
(a)
20
0
0 2 4 6 8 10
Axial displacement: mm
(a)
140
M20
M16
120 M14
M12
M0
100
Force: kN
80
60
40
(b) 20
Figure 4. (a) Geometry of load application device (dimensions:

mm) and (b) device with specimen mounted before test initiation 0
0 2 4 6 8 10
Axial displacement: mm
(b)
which contained the innermost U-bars, see Figure 6(b). The Figure 5. Tested load–displacement curves: (a) series C (concrete
opening was greatest at the edges of the surface and smallest at grout); (b) series M (mortar grout)
the location of the axial crack in the centre, indicating a slight
rotation of the grout cover. For a further increase of displace-
ment into the post-peak region, the grout cover broke off and outermost U-bars, Figure 3(b). It seems more important that
revealed cracks between the U-bar loops. Figure 7 shows the the lacer bars are placed correctly in the line of loading. This
typical crack pattern inside the grout at the level of the loop was indicated by trial tests performed on the above-mentioned
reinforcement. Two inclined cracks developed between the tips unsuccessful castings (specimens with lacer bar positioned
of the U-bars. These cracks were supplemented by a straight with eccentricity to the line of loading), which showed lower
crack between the tips of the innermost U-bars and two cracks capacity.
originating at the tip of the outermost U-bars. The crack
pattern near the heads of the lacer bar indicates that the head
has been utilised for anchorage once the lacer reinforcement Failure mechanisms and upper bound
was activated after cracking of the grout material. solutions
In this section, an upper bound model for prediction of
From the test results of series M and T, it was also found that the load-carrying capacity related to the failure in the grout
the tensile capacity of the connections did not depend particu- material is developed. For this purpose, the grout materials
larly on whether the lacer bar was positioned in the centre of will be treated as rigid-plastic, modified Coulomb materials,
the circular loop, Figure 3(a), or fixed to the looped end of the obeying the associated flow rule, and the reinforcement
897

(a)
(a)
(b)
Figure 6. Crack patterns on surface of specimen T12a: (a) crack

in axial direction; (b) development of inclined cracks on surface
(b)
bars are assumed to be rigid-plastic and only capable of
carrying axial loads. A kinematically admissible failure mech- Figure 7. Examples of failed specimens after test: (a) concrete
grout (b) mortar grout
anism is shown in Figure 8, where the experimental obser-
vations described in the previous section ‘Experimental results
and observations’ are captured in a simplified representation
inspired by the work of Jørgensen and Hoang (2013). The following, two solutions are developed. The first one concerns
failure surfaces formed in the grout material are simplified as the failure of a connection with lacer reinforcement where the
plane surfaces, which may be represented by the yield lines tensile strength of the grout material is neglected. The second
indicated in Figure 8(a). The geometries of the actual failure solution is developed for connections without lacer reinforce-
surfaces are three-dimensional as a result of the complex stress ment and therefore includes the tensile strength of concrete/
state between the loop bars of the connection, see Figure 7. mortar. The second solution can be considered as a lower limit
However, the simplifications introduced are necessary in order for the tensile capacity of connections containing transverse
to develop analytical solutions suitable for practical use. reinforcement. This will be further discussed in the following.
In rigid-plastic limit analysis of problems in structural con- Upper bound solutions for connections
crete, the tensile strength of concrete is usually neglected. This with lacer reinforcement
can be justified by the fact that the (rather small) tensile Figure 8 illustrates a failure mechanism for specimens with
strength of concrete will normally have vanished almost com- lacer reinforcement. The displacement field in the grout is
pletely when the reinforcement enters the state of yielding. idealised as rigid body motions of the three segments, III, IIa
This is a fair assumption for the test specimens with lacer bars, and IIb relative to segment I. The determination of the relative
where cracking activates the lacer bars, which eventually reach displacements between adjacent segments is inspired by the
yielding at the ultimate state. However, for the specimens in work of Jørgensen and Hoang (see Jørgensen and Hoang
series C0 and M0 without lacer reinforcement, the cracking (2013) for an in-depth description), and adjusted to the geo-
load is also the ultimate load and, for this reason, it is necess- metry of the tested ‘2-on-2’ symmetrical connection. As illus-
ary to take the tensile strength into account when modelling trated in Figure 8(b) the relative displacements in each yield
the load-carrying capacity in such cases. Therefore, in the line can be determined as follows. Segment IIa is displaced
898

N N N
ul III
α
IIb ut ut IIa
u b ul u l ua
H b D Ac H
ut ut
ub I ua
β
s a s
l
N
N
N t
(a) (b) (c)
Figure 8. (a) Failure mechanism, (b) relative displacements and (c) overlapping loop area considered for grout failure
with the displacement vector ua relative to segment I The rate of internal work consists of contributions from the
lacer bar and the grout material. The contribution from the

ul lacer reinforcement (which is stressed to yielding at the two
1: ua ¼
ut points where it crosses the inclined yield lines) is given by
6: WIS ¼ 2AsL fyL ut

where ul is the component parallel to the longitudinal direc-
tion, l, and ut is the component in the transverse direction, t.
The displacement vector, ub, which describes the relative A formula for the dissipation in a plane of discontinuity
motion of segment IIb relative to segment I is ua mirrored formed in a modified Coulomb material was established by
about the l-axis. This is due to the symmetric conditions and Jensen (1976). When the tensile strength is neglected and the
for this reason, |u| = |ua| = |ub|, can be introduced. Segment III is normality condition is fulfilled, the dissipation is found by
displaced by ul relative to segment I. Finally, the two segments Nielsen and Hoang (2011)
IIa and IIb are also moving transversely and relative to
1
segment III by the displacement ut. The components of the 7: WIC ¼ νfc Ai ð1 sin αÞjuj; αφ
2
relative displacements are related in the following way
2: ul ¼ juj cosðα βÞ where Ai is the area of the plane of failure (i.e. length of the
yield line times its width) and φ is the internal angle of friction
for the grout material. In order to account for the fact that
neither concrete nor mortar are perfectly plastic materials, an
effectiveness factor, ν, is introduced (Nielsen and Hoang,
3: ut ¼ juj sinðα βÞ
2011). The effectiveness factor depends on a number of factors
such as softening effects and effects of cracking. For a loop
where α is the angle between the displacement vector and the connection loaded in tension and grouted with concrete,
inclined yield line running between the tips of the U-bars, and Jørgensen and Hoang (2013) discussed in depth the choice of
β is the inclination of the yield line to the l-axis effectiveness factor. In the following, a modification of the
equation used in Jørgensen and Hoang (2013) is adopted. The
s modification, which is similar to the findings of Jørgensen
4: tan β ¼
H (2014) and Jørgensen and Hoang (2015), is introduced to
account for the fact that the softening effect in mortar deviates
The rate of external work performed by the load, N, is from that of concrete
given by
K 1
8: ν ¼ pffiffiffiffiffi 1 þ pffiffiffiffiffi
5: WE ¼ Nul fc H
899

where fc is given in MPa and H is given in m. Here the factor The solution for the load-carrying capacity in the case of
K depends on the grout material, and H expresses a character- grout failure is influenced by the geometry of the loop con-
istic length of the sliding failure/yield line, see Figure 8(c). nection, the s/H ratio, see Figure 8(a), and the internal angle
of friction for the grout material. Owing to the normality
In order to determine the rate of internal work performed in the condition, the angle of displacement will always be larger than
grout material, the areas of the respective failure surfaces must or equal to the internal angle of friction. Furthermore, the
be determined. From Figure 8(b) it can be seen that only the two angle of displacement cannot be smaller than the inclination
inclined yield lines running between the tips of the U-bars are rel- of the yield line, β, as an inwards displacement of segment
evant for this solution, as the remaining yield lines are separation IIa/IIb towards segment III is not possible. The requirements
yield lines with zero dissipation (the tensile strength is disre- are summarised in Equation 13, which describes the tensile
garded). The areas of the inclined yield lines are calculated as capacity of the loop connection including lacer reinforce-
ment and disregarding the tensile strength of the grout material
Ac π H2 π ðD þ 2ϕÞ2 (α is calculated by Equation 12 and β is calculated by
9: AI ¼ ¼ ¼
cos β 4 cos β 4 cos β Equation 4).
where Ac is the circular area delimited by the overlapping From Equation 13 it can be seen that the internal angle of fric-
U-bars (see Figures 8(b) and 8(c)). It appears that AI has the tion, φ, is an essential parameter for the tensile capacity of the
shape of an elliptical surface. The adaptation of Equation 9 joint. This parameter is dependent on the properties of the
entails that the grout outside the overlapping U-bars is neg- grout material, and especially on the aggregate content and
lected. This is primarily justified by the fact that the grout aggregate sizes of the mixture. It should be noted, that the sol-
cover more or less spalled off during testing, leaving only the ution presented by Jørgensen and Hoang (2013), with the
grout core inside the loops to carry the tensile load. A similar exception of a factor of 2, can be reproduced by Equation 13
way of determining the area of the inclined yield lines was when inserting the internal angle of friction for normal
suggested by Jørgensen and Hoang (2013). strength concrete, tan φ = 3/4.
8 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
>
> s 2 4Φ ΦL s
>
> þ
L
1 for α φ and φ β
>
> ν ν
>
> H H
N < 2
13: ¼ [1 þ ðs=H Þ ][ð1= cos φÞ tan φ] þ 2ΦL =ν[tan φ ðs=HÞ] for α , φ and φ . β
νfc Ac >
> 1 þ ðs=HÞ tan φ
> rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
>
> s 2
>
> s
: 1þ for α , β and φ β
H H
From the work equation, an upper bound solution for the It can be seen that the tensile capacity according to Equation
tensile capacity of the connection is found as 13 approaches zero when ΦL ! 0. Hence, for very small values
of ΦL, it may happen that Equation 13 predicts a capacity that
N 1 sin α ΦL is lower than the capacity calculated for a similar connection
10: ¼ þ2 tanðα βÞ; αφ without lacer bar, but taking into account the tensile strength
νfc Ac cos β cosðα βÞ ν
of concrete/mortar (see next section). In such a situation,
the solution that accounts for the tensile strength should of
where ΦL is the mechanical reinforcement degree of the lacer course be taken as the capacity – and the connection should
reinforcement, defined as then be classified as being under-reinforced with respect to
transverse reinforcement.
AsL fyL
11: ΦL ¼
Ac fc
Upper bound solutions for connections without
The optimal upper bound solution is found by minimising lacer reinforcement
Equation 10 with respect to the angle of displacement, α. It can For connections without transverse reinforcement, the capacity
be shown that the optimal angle of displacement is given as depends on the tensile strength of the grout. The dissipation in
a concrete/mortar yield line when the tensile strength, ft, is
2 3 included, is given by Nielsen and Hoang (2011)
6 1 2ðΦL =νÞ 7
12: α ¼ β þ arcsin4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5; αφ 1
1 þ ðs=H Þ2 14: WIC ¼ νfc Ai ðl m sin αÞjuj; αφ
2
900

where the effectiveness factor, ν, is still estimated according to slightly simplified failure mechanism introduced in Figure 9,
Equation 8 and the parameters l and m are given as the solution can be optimised explicitly.
νt ft sin φ This failure mechanism consists of contributions from seven

15: l ¼12
νfc 1 sin φ yield lines, namely two inclined sliding yield lines running
between the loop bars, two separation yield lines in the
l-direction at the tip of the outer loops with an area of At, two
separation yield lines in the l-direction at the tip of the inner
νt ft 1 loops with an area of At, and one separation yield line in the
16: m¼12
νfc 1 sin φ t-direction between the inner loops with an area of Al. The
areas of the four separation yield lines formed in the axial
The tensile strength of the grout may, for example, be estimated direction, referring to Figure 9(b), are considered equal. Again
by the following formula given by Nielsen and Hoang (2011) the cover is ignored, which constitutes a conservative choice.
The areas of the yield lines are estimated as follows
17: ft ¼ 026fcm
2=3
1
19: At ¼ H ðb H Þ
2
The mean compression strength of the grout, fcm, is in this
study taken as the average value found from compression tests
on cylinders. In Equations 15 and 16, νt denotes the effective-
ness factor for concrete in tension. This factor is, with inspi- 20: Al ¼ aH
ration from Nielsen and Hoang (2011), taken as
03
H The area of the inclined sliding yield line may also in
18: νt ¼ 06
01 this case be determined by Equation 9. It should be noted
that H appears as a characteristic dimension in all the con-
Again, H is a characteristic dimension, see Figure 9. sidered areas. When solving the work equation, using the exter-
Considering the failure mechanism in Figure 8, the rate of nal work introduced in Equation 5, the internal work in
internal work consists of contributions from five separation Equation 14 and with the areas given in Equations 9, 19
yield lines and two sliding yield lines inclined with the angle β and 20, the following solution for the tensile capacity, N0,
to the l-direction. For that particular failure mechanism, the without lacer reinforcement, is found
optimal angle of displacement cannot be given explicitly when
the tensile strength of the grout is included. However, for the N0 l m sin α νt ft At νt ft Al
21: ¼ þ4 tanðα βÞ þ
νfc Ac cos β cosðα βÞ νfc Ac νfc At
N N The optimal angle of displacement is found as
At 2 3

6 m 4 ν t f t At 1 7
α ul III 22: α ¼ β þ arcsin4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5; αφ
l l νfc Ac 2
ðs=H Þ þ 1
ut ut
ub ua
IIb IIa D Ac H b
I
ut ut Again the same restrictions are imposed on the angle of dis-
placement, α ≥ φ and α ≥ β. It should be noted that the solution
β is different from that of Jørgensen and Hoang (2015), as more
H separation yield lines are included in this model.
At
l
N
t
N Tensile capacity of loop connection
(a) (b) The solutions developed above are valid when failure in the
grout material is governing. Naturally, these solutions cannot
Figure 9. Simplified failure mechanism for connection without be used if the yield capacity of the U-bars renders a smaller
lacer bar: (a) relative movements; (b) areas considered for grout
failure load. Hence, the true tensile capacity of the connection is the
minimum of the capacity of the grout material and the yield
901

capacity of the U-bar loops. As discussed, the capacity calcu- 1·5

lated by inclusion of the tensile strength of the grout and
4Asfu
without any transverse reinforcement constitutes a lower limit
↓ νAcfc
for the capacity of the joint. The true shift between the two
theoretical solutions, Equations 13 and 21, is governed by a 1·0
transition phase where the U-bars are yielding and the grout
↑ 4Asfy
at the same time contributes with a small, residual strength.
νAcfc
νAcfc
N
As a conservative and simplistic approach, Equation 21 is ↑
Equation 13
introduced simply as the lower limit of the capacity. The com-
0·5
plete solution for the tensile capacity of the loop connection is
thereby given as ↑
Equation 22
8
> N Equation 13 Concrete (C), K = 0·88, ϕ = 37°
>
> N ¼ max
< N0 Equation 21 0
23: Nu ¼ min 0 0·5 1·0 1·5
>
> Failure in grout material ΦL
>
:
Ny ¼ 4As fy Yielding of U bars (a)
1·5
Comparison of tests with theory and

discussion
Besides the geometry and the parameters that were experi- 4Asfy
1·0
νAcfc
mentally determined in this study (strength of materials),
νAcfc
the tensile capacity of the loop connection (Equation 23) is

N
governed by two parameters, the factor K in the ν-factor,

Equation 8, and the internal angle of friction, φ, which was 0·5
Equation 13
not explicitly determined in this study. However, the exper-
imental results suggest a variation in these factors for the Equation 22
different grout materials tested. Jørgensen and Hoang (2013)
proposed the factor K to be taken as 0·88 for concrete, whereas Mortar (T), K = 0·75, ϕ = 34°
0
for mortar they suggested a factor of K = 0·75 based on shear 0 0·5 1·0 1·5
tests on keyed shear joints reinforced with wire-rope loops and ΦL
grouted with mortar (Jørgensen and Hoang, 2015). In the fol- (b)
lowing, these two suggestions are adopted. The difference in
1·5
this factor is mainly attributed to the difference in the
maximum aggregate size.
Regarding the internal angle of friction, φ, the aggregate 4Asfy

content as well as the maximum size of the aggregate influence 1·0 νAcfc
this parameter. For normal strength concrete, a value of φ = 37°
νAcfc
is usually adopted (Dahl, 1992; Nielsen and Hoang, 2011), but

N
only few investigations of the internal angle of friction

of mortar have been conducted. For mortars, Nielsen (1998) 0·5 Equation 13
reported tests indicating an internal angle of friction of
Equation 22
approximately 30°. In the following comparison of tests with
calculations, a value of φ = 37° is chosen for test series C,
Mortar (M), K = 0·75, ϕ = 32°
φ = 32° for series M and φ = 34° for series T. A higher value is 0
chosen for series T compared to series M because the 0 0·5 1·0 1·5
maximum aggregate size as well as the aggregate-to-cement ΦL
ratio (a/c) is larger in series T than in series M. (c)
Figure 10. Comparison of models with test results: (a) Series C;

The comparison of test results with calculations can be seen in (b) Series T; (c) Series M
Figures 10(a)–10(c). The theoretical results are generated using
the average values of the strengths given in Table 3, and the test
results are normalised with the actual strength parameters of
902

each specimen. Equation 23 accounts for initiation of yielding Conclusions

in the U-bars. However, when the concrete core is sufficiently Tensile tests on symmetric ‘2-on-2’ loop connections have been
strong, the U-bars may develop hardening with increased dis- performed. The overall aim of the study was to compare the
placement. The load corresponding to the rupture of the performance of connections grouted with concrete and with
U-bars has therefore been added in Figure 10(a) for compari- mortar, and on this basis develop models to predict the load
son with the results of specimens C20, where the reinforcement related to failure in the grout material. Special attention was
loops yielded and hardened before termination of the test, see given to the maximum aggregate size of the grout matrix, and
Figure 5(a). the amount of transverse reinforcement (lacer reinforcement).
An upper bound solution based on the theory of plasticity was
In general, the models capture the experimental results well established for estimation of the tensile capacity of the loop
and predict the tensile capacity of the connections. The test connection. The following main conclusions can be drawn.
results revealed a difference in the capacity depending on the
material properties. These differences are accounted for in the & Connections with concrete grout with a maximum
choice of internal angle of friction for the grout material and aggregate size of 16 mm performed better (in terms of
in a reduction of the effectiveness factor for the mortar grout ultimate load as well as ductility) than similar connections
compared to concrete. This is justified by the more brittle be- grouted with mortar containing a maximum aggregate size
haviour of mortar compared to concrete, which is evidently a of 2 mm and 4 mm and with similar compression
result of the difference in the maximum aggregate size as the strengths.
effect of aggregate interlock is smaller in mortar. In addition, & The tested capacity of connections without lacer
it can be discussed whether the aggregate-to-cement ratio influ- reinforcement corresponds to the cracking load of
enced the parameters as well. connections with lacer reinforcement.
& Transverse reinforcement prevented brittle failure.
From Figures 10(a) and 10(c), it can be seen that the solution & The capacity of the joint increased with increasing amount
for connections without transverse reinforcement estimates the of lacer reinforcement.
test results of series M0 well and gives a slight underestimate & The developed rigid-plastic upper bound solutions revealed
of the results of series C0. The solution accounts for the con- that the internal angle of friction for the grout material is
tribution of the grout outside the overlapping area of the an essential parameter for the capacity.
U-bars, which fails at the same time as the overlapping area & With reasonably selected effectiveness factors and estimates
between the U-bars. However, the side cover is not included, of the internal angle of friction, it is possible to calculate
see Figure 9(c), as the complex crack pattern in this region the capacity of the connections.
cannot be simplified by a simple failure plane. Besides the geo-
metry, it is recognised from Equation 21 that the capacity The comparison with test results shows that the models
without transverse reinforcement is governed by the tensile provide a tool that identifies the transition between failure of
strength of the grout. The effective tensile strength has, owing the grout material and yielding of the U-bars. Connections in
to lack of investigations of mortars, been calculated by the practice should always be designed for yielding of the U-bars,
same formulas (Equations 17 and 18) for both mortar and as this first of all provides a significant ductile response when
concrete. This, however, is not necessarily accurate, but has loaded in tension and also constitutes the basis for a ductile
been judged to be an acceptable approximation. The solution response when the loop connections are used as part of a
may be used to establish the transition from under-reinforced shear-resisting structural element. Based on the experimental
to normally reinforced connections. In practice, connections results and the theoretical analysis, some aspects need to be
should always be designed with a sufficient amount of lacer further investigated in more detail. These include
reinforcement in order to satisfy the condition for normally
reinforced connections. & internal angle of friction for grout material
& effectiveness factor for grout material.
As a concluding remark on the practical aspects, connections
should in addition always be designed to obtain yielding
of the U-bars. The models presented in this paper provide
a method to determine the necessary combination of Acknowledgements
geometry, U-bar strength and dimension with the amount The experimental research presented in this paper was
of lacer bar reinforcement required to avoid grout failure. financially supported by the Danish Association for Precast
Naturally, yielding of the U-bars preserves the highest ductility, Concrete Elements and the COWI Foundation. The test
both when the loop connection is loaded in tension and in elements were produced at CRH Concrete and testing of Series
the case of a shear wall, where an intact grout material T was conducted with support from MSc Jacob Svejgaard
improves the conditions for a ductile response in the ultimate and MSc Lasse Øvrelid. The authors gratefully acknowledge
limit state. these valuable contributions.
903

REFERENCES Jørgensen HB (2014) Strength of Loop Connections between Precast

Araújo DDL, Curado MC and Rodrigues PF (2014) Loop connection Concrete Elements. PhD thesis, Department of Technology
with fibre-reinforced precast concrete components in tension. and Innovation, University of Southern Denmark,
Engineering Structures 72: 140–151. Odense, Denmark.
Brantschen F, Faria DMV, Ruiz MF and Muttoni A (2016) Bond behaviour Jørgensen HB and Hoang LC (2013) Tests and limit analysis of loop
of straight, hooked, U-shaped and headed bars in cracked connections between precast concrete elements loaded in tension.
concrete. Structural Concrete 17(5): 799–810. Engineering Structures 52: 558–569.
Dahl KKB (1992) A Failure Criterion for Normal and High Strength Jørgensen HB and Hoang LC (2015) Load carrying capacity of
Concrete. PhD thesis, Technical University of Denmark, Lyngby, keyed joints reinforced with high strength wire rope loops.
Denmark, Report No. 286. Concrete – Innovation and Design, Proceedings of fib Symposium,
DeVries RA, Jirsa JO and Bashandy T (1999) Anchorage capacity in Copenhagen, Denmark, pp. 1–13.
concrete of headed reinforcement with shallow embedments. Leonhardt F, Walther R and Dieterle H (1973) Versuche zur
ACI Structural Journal 96(5): 728–736. Ermittlung der Tragfähigkeit von Zugschlaufenstößen.
fib (Fédération International du Béton) (2008) Bulletin 43: Structural Deutscher Ausschuss für Stahlbeton, Berlin, Germany,
Connections for Precast Concrete Buildings. fib, Lausanne, Bulletin No. 226 (in German).
Switzerland. Nielsen CV (1998) Triaxial behavior of high-strength concrete and
Ghali A and Youakim SA (2005) Headed studs in concrete: state of the mortar. ACI Material Journal 95(2): 144–151.
art. ACI Structural Journal 102(5): 657–667. Nielsen MP and Hoang LC (2011) Limit Analysis and Concrete Plasticity,
Gordon SR (2006) Joints for Precast Decks in Steel Concrete Composite 3rd edn. CRC Press, Taylor & Francis Group, Boca Raton,
Bridges. PhD thesis, Heriot Watt University, Edinburgh, UK. FL, USA.
Hansen K, Kavyrchine M, Melhorn G et al. (1974) Design of vertical Ong KCG, Hao JB and Paramasivam P (2006) A strut-and-tie
keyed shear joints in large panel buildings. Building Research and model for ultimate loads of precast concrete joints with loop
Practice 2(4): 202–215. connections in tension. Construction and Building Materials
Hao J (2004) Structural Behaviour of Precast Component Joints with 20(3): 169–176.
Loop Connection. PhD thesis, National University of Singapore, Sørensen JH, Hoang LC, Olesen JF and Fischer G (2017) Test and analysis
Singapore. of a new ductile shear connection design for RC shear walls.
Jensen BC (1976) Nogle Plasticitetsteoretiske Beregninger af Beton og Structural Concrete 18(1): 189–204.
Jernbeton. PhD thesis, Technical University of Denmark, Lyngby, Thompson MK, Jirsa JO and Breen JE (2006) Behavior and capacity of
Denmark, Report No. 111 (in Danish). headed reinforcement. ACI Structural Journal 103(4): 522–530.
How can you contribute?

To discuss this paper, please submit up to 500 words to the
editor at [email protected]. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial board, it will be published as a
discussion in a future issue of the journal.
904
Paper II
”Test and analysis of a new ductile shear connection design
for RC shear walls”
Published in: Structural Concrete (2017), 18(1):189-204
221
Received: 22 March 2016 Revised: 27 May 2016 Accepted: 30 May 2016
DOI 10.1002/suco.201600056
TECHNICAL PAPER
Test and analysis of a new ductile shear connection design

for RC shear walls
Jesper Harrild Sørensen | Linh Cao Hoang | John Forbes Olesen | Gregor Fischer
Department of Civil Engineering, Technical

University of Denmark, Lyngby, Denmark This paper presents a new and construction-friendly shear connection for the
Correspondence assembly of precast reinforced concrete shear wall elements. In the proposed
Jesper Harrild Sørensen, Department of Civil design, the precast elements have indented interfaces and are connected by a nar-
Engineering, Technical University of Denmark,
row zone grouted with mortar and reinforced with overlapping U-bar loops. Con-
Brovej, Bygning 118, 2800 Kgs. Lyngby,
Denmark. trary to conventional shear connections, the planes of the U-bar loops are here
Email: [email protected] parallel to the plane of the wall elements. This feature enables a construction-
Funding information friendly installation of the elements without the risk of rebars clashing. The core
Danish Association for Precast Concrete Elements; of mortar inside each U-bar loop is reinforced with a transverse double T-headed
COWI Foundation.
bar to ensure transfer of tension between the overlapping U-bars. Push-off tests
show that a significantly ductile load–displacement response can be obtained by
the new solution as compared to the performance of the conventional keyed shear
connection design. The influence of the interface indentation geometry was inves-
tigated experimentally and the failure modes in the push-off tests were identified
by use of digital image correlation (DIC). For strength prediction, rigid plastic
upper-bound models have been developed with inspiration from the observed fail-
ure mechanisms. Satisfactory agreement between tests and calculations has been
obtained.
KEYWORDS
concrete plasticity, digital image correlation, ductility, keyed shear connections,

robustness
1 | INTRODUCTION supplement the advantages of precast construction with in

situ cast solutions in selected zones. An example of how
Structural solutions based on precast concrete elements are current precast solutions have been pushed to the limit can
often more economically feasible than in situ cast solutions, be studied in References 1 and 2 that report on the design
because precast technology enables a reduction of construc- and construction of a landmark building in Copenhagen,
tion time as well as labor cost. When using precast solu- Denmark. The leaning characteristic of the building
tions, the on-site work mostly consists of assembling and imposed serious challenges to the design of the shear con-
connecting the precast elements into an integrated structural nections between the precast panels for insurance of overall
system. Hence, connection designs that are construction- structural stability.
friendly play an important role for the overall cost reduc- Currently, structural continuity between precast shear
tion. It is, however, a challenge to design connections that panels is established by use of narrow keyed connections
are easy to construct and at the same time have structural containing overlapping U-bars and grouted with mortar (see
performance (in terms of strength and ductility) which can Figure 1). However, with this conventional solution, which
be compared with that of in situ cast solutions. In cases with has been used since the 1960s, it is difficult to obtain full
unusual structural geometry, it may be necessary to structural continuity because the strength and ductility of
Structural Concrete. 2017;18:189–204. wileyonlinelibrary.com/journal/suco © 2017 fib. International Federation for Structural Concrete 189
190 SRENSEN ET AL.
(a)
Mortar (grout)
Precast Concrete
Element
U-bar
Locking bar Locking bar
(b)
Pre-bend U-bars
Horizontal section
Element
lowered
Vertical
section
Element
pre-installed
Vertical
locking bar
FIGURE 1 (a) Conventional shear connection design and (b) illustration of procedure for assembling of precast elements.
these connections will normally be less than that of the pre- larger than 8 mm to be used. In addition to a single longitu-
cast elements.3,4 In addition, the construction sequence is dinal locking bar, the new design also includes the use of
influenced by the design. To avoid rebar-clashing when transverse locking bars (in the following called lacer bars)
assembling the precast panels (Figure 1b), the U-bars pro- in the form of a double T-headed rebar placed inside the
truding from the precast panels have to be bent up (prior to U-bar loops. The idea here is to utilize the double-headed
installation of panel) and subsequently straightened again rebar together with the core of mortar inside the loop as a
once the panel has been placed in position. This procedure transverse dowel that enables transfer of tension between
imposes a limit on the cross-sectional diameter of the U- the overlapping U-bars. Tension in the U-bars across the
bars and hence limits the strength of the connection (nor- connection is required to ensure equilibrium when diagonal
mally bars with diameter 6–8 mm are used). The conven- compression struts develop between the keyed joint inter-
tional shear connection is therefore not feasible for use, for faces as a result of shear loading. The double T-headed
example, in shear walls of tall buildings where considerable rebar is chosen because the heads provide increased anchor-
horizontal loads have to be carried. age of the short lacer reinforcement, which otherwise cannot
This paper presents a new solution for the connection of be ensured using regular straight reinforcement.
precast shear wall elements. The aim of the new design is to To investigate the structural performance of the new
ease the construction challenges and at the same time connection design, an experimental program was conducted.
improve the structural performance compared to the conven- The investigation showed that the load–displacement
tional solution. Figure 2 schematically illustrates the new response of the new design is significantly more ductile
connection design, which differs from the conventional than that of the conventional solution. Furthermore, the tests
solution in the way the U-bar loops are oriented and in the indicated that it is possible to obtain higher load-carrying
way structural continuity is ensured in the U-bar overlaps. capacities with the new design. In addition to the experi-
The joint interfaces are keyed as in the conventional solu- mental work, this paper also presents upper-bound rigid
tion. As illustrated in Figure 2, the loop orientation in the plastic models for prediction of the critical failure mode as
new solution allows for a construction-friendly installation well as the load-carrying capacity of the new connection
(vertical lowering) of the precast panels, without clashing of design. The models furnish a simple tool to optimize the
rebars and thus without the need to pre-bend and post- geometry of the keyed joint interfaces in order to enhance
straighten the U-bars, which enables U-bars with diameters the ductile behavior of the connection.
SRENSEN ET AL. 191
(a)
Mortar (grout)
Precast Concrete
Element
U-bar
Lacer bar Locking bar
(b)
Horizontal section
Element
lowered
Vertical
section
Element
pre-installed
Vertical
locking bar
FIGURE 2 (a) New construction-friendly connection design and (b) illustration of procedure for assembling of precast elements.
2 | PR EVIOUS IN VE STIGATIONS ON Based on the experimental findings, a number of semi-

S H E A R C O NN E C T I ON S empirical formulas were suggested for the prediction of the
ultimate load-carrying capacity. Current design provisions
With the introduction of precast element construction, the for joints between concrete cast at different times are based
design and performance of on-site cast connections became on the shear friction hypothesis (see, for example, the fib
a matter of special interest. Since the 1960s, the conven- guide to good practice3 and the European code of prac-
tional keyed shear connection has been experimentally tice15). However, other approaches can also be used.
investigated, with the main interests being the behavior of Kaneko et al.16,17 proposed a fracture mechanics approach
the connection during loading, the ultimate load-carrying to predict the crack formation in indented shear joints. They
capacity, and the design aspects of the joint configuration. identified two main fracture mechanisms for shearing of
Hansen et al.4 summarized the early work on this topic in a keys, based on an experimental program that included plain
report, which constitutes the work of the CIB commission and fiber-reinforced concrete joints. The test results were
W23A. The experimental programs that served as basis supplemented with nonlinear finite element calculations.
for the commission’s report include the work of Halasz Later, Kaneko and Mihashi18 extended the investigation by
and Tantow, Cholewicki, Pommeret, and Fauchart and presenting an analytical model for determination of the tran-
Cortini,5–8 who used similar test setups as the one used in sition between the two mechanisms. However, variations in
the present study. Shear tests with other test setups to inves- key dimensions such as length and depth were not included
tigate factors that influence the load-carrying capacity have in the experimental investigation.
also been published.9–14 In all investigations, regardless of Theoretical works based on the theory of rigid plastic-
testing method, it was recognized that the ultimate capacity ity have also been proposed. Jensen19 was the first to
was influenced by a number of factors, including the num- establish an upper-bound solution for the load-carrying
ber of shear keys, the cross-sectional area of the keys, the capacity of keyed shear joints by assuming complete shear-
strength of the grout mortar, the degree of transverse rein- ing of the key area. These findings were later the basis for
forcement, and the magnitude of external transverse con- several simplified formulas, which incorporate empirical
finement stresses. factors to fit theory with test results. This includes the
192 SRENSEN ET AL.
formulas by Chakrabarti et al.20 and Abdul-Wahab and P

Sarsam.21 Later Christoffersen22 expanded the application
of plasticity theory to include both upper- and lower-bound
solutions for the shear capacity of keyed joints. Recently,
Jørgensen and Hoang23 developed an upper-bound model
for the failure of keyed shear joints reinforced with high-
strength wire rope loops by accounting for diagonal cracks Locking bar
between the shear keys. Jensen, Christoffersen as well as
Jørgensen and Hoang considered only a global failure
mechanism with complete shearing of the keyed area. A
local failure mode that involves key corner crushing has
been observed by several authors. However, the problem
has not been treated in depth, nor has an analytical solu- Lk Ak s
tion been proposed.
3 | EX PER IM ENT AL PR OGR AM hk L

t
The experimental program contained a total of 25 push-off
tests. The program included a preliminary investigation of dk
seven specimens where the performance of the conventional
design, as a reference, was compared with the new design
with identical geometrical joint properties. The remaining
18 specimens, Series I–IX, were subdivided into two inves- Lk
tigations related to the geometry of the key indentation (see
Figure 3). The tests were carried out in quasi-static deforma-
tion control.
b
3.1 | Specimens and geometry

The general geometry of the push-off test specimens can
be seen in Figure 3 and details of geometrical and material
properties are given in Table 1. Series R refers to reference P
specimens designed with a conventional reinforcement lay-
General layout of push-off test specimens, thickness equals
out as illustrated in Figure 1a (however, in the reference
FIGURE 3
200 mm (reinforcement in precast element not shown).
specimens, the loops were placed outside the keyed area).
Series P refers to pilot specimens designed with “2-on-1”
vertical loop connections. This refers to a design similar to ten specimens (Series V–IX), the depth of the key indenta-
the principles shown in Figure 2a, however, in the pilot tion, dk, was varied, while the key length and the key height
specimens, there was (for each looped connection) only were kept constant, Lk = 140 mm and hk = 200 mm.
one centrally placed U-bar that protruded from the precast Table 2 contains parameters and material properties for Series
element to the right. The cross-sectional area of this single I–IX. The diameter of the lacer bar was carefully designed so
U-bar is referred to as As in Table 1. The reinforcement that the tensile capacity of the overlapping loops would be
area in the weak side, As, as well as the geometry of the governed by yielding of the U-bars and not crushing of the
shear keys were identical for all specimens in Series R and mortar. For this purpose, the calculation model for tensile
P. The joints in Series P were not provided with longitudi- capacity of U-bar loop connections developed by Jørgensen
nal locking bars. For specimens type I–IX, “2-on-2” verti- and Hoang24 was used. The double T-headed lacer bar in
cal loop connections were used (see Figure 2a). For these each loop was positioned as shown in Figure 2a to make it
specimens, the area As listed in Table 1 should be under- function as tension reinforcement in the small transverse cir-
stood as the cross-sectional area of two U-bars. In order to cular mortar dowel, which ensures transfer of tension
eliminate the edge effect, anchorage plates were mounted between the overlapping U-bars. Each design was tested with
at each end of the longitudinal locking bar in Series I–IX. two replicates and the material properties were found as aver-
In eight specimens (Series I–IV), the length of the keys, age values obtained from tensile tests of the steel reinforce-
Lk, was varied, while the key height, hk, was kept constant ment and compression tests of ϕ100 × 200 mm cylinders of
to half of the panel thickness, t. In the remaining the mortar used for casting the joints.
SRENSEN ET AL. 193
TABLE 1 Geometrical parameters and strength properties of the joints in the experimental program, bold numbers indicating the specifics of the specimen
No. fca [MPa] hk [mm] Lk [mm] Ak [mm2] dk [mm] As [mm2] PFP [kN] PU [kN] DI [-]
R 1 34.6 85 160 13,600 16 101 282.43 – 0.42b
2 35.7 85 160 13,600 16 101 303.80 – 0.59b
3 35.7 85 160 13,600 16 101 337.42 – 0.70b
P 1 38.1 85 160 13,600 16 101 344.24 357.45 1.00b
2 38.1 85 160 13,600 16 101 347.04 368.12 0.97b
3 42.7 85 160 13,600 16 101 342.49 339.97 0.87b
4 42.7 85 160 13,600 16 101 331.42 324.49 0.95b
I 1 31.2 100 120 12,000 28 201 379.02 441.21 1.03c
2 34.2 100 120 12,000 28 201 416.59 472.92 1.00c
II 1 31.2 100 140 14,000 28 201 366.40 463.78 1.06c
2 34.2 100 140 14,000 28 201 414.46 462.48 1.00c
III 1 31.2 100 160 16,000 28 201 393.04 494.70 1.07c
2 34.2 100 160 16,000 28 201 473.52 514.87 0.98c
IV 1 31.2 100 180 18,000 28 201 439.44 470.89 0.94c
2 34.2 100 180 18,000 28 201 478.17 515.31 0.96c
V 1 31.2 200 140 28,000 10 201 475.24 488.97 0.97c
2 34.2 200 140 28,000 10 201 492.86 535.61 1.04c
VI 1 30.6 200 140 28,000 16 201 527.09 502.55 0.89c
2 30.6 200 140 28,000 16 201 523.82 550.98 0.90c
VII 1 30.6 200 140 28,000 20 201 549.17 451.58 0.78c
2 30.6 200 140 28,000 20 201 524.46 527.85 0.87c
VIII 1 30.6 200 140 28,000 25 201 507.05 528.67 0.92c
2 30.6 200 140 28,000 25 201 516.97 545.33 0.92c
IX 1 30.6 200 140 28,000 28 201 526.53 534.19 0.93c
2 30.6 200 140 28,000 28 201 527.59 527.07 0.88c
a
Compression strength of mortar.
b
Calculated using δmax = 13 mm.
c
Calculated using δmax = 20 mm.
TABLE 2 Parameters kept constant for Series I–IX software has been described by Pereira et al.,26 who studied
Description Symbol Value the cracking behavior of cement paste, mortar, concrete, and
U-bar diameter ϕ 8 mm fiber-reinforced concrete. In the present study, the analysis
Yield strength of U-bar fy 487 MPa was performed as a two-dimensional analysis using images
Lacer bar diameter ϕLacer 16 mm taken with a 36.3-megapixel digital camera. The surface of
Yield strength of lacer bar fy,Lacer 563 MPa the connection was spray painted with a white base layer,
Diameter of locking bar ϕL 12 mm followed by black dots sprayed randomly to create a unique
Yield strength of locking bar fyL 584 MPa and recognizable pattern on the surface. The area covered
Panel thickness t 200 mm by the Aramis analysis corresponds approximately to the
Internal bend diameter of loops D 60 mm area of the joint, namely L (b + 2dk) (see Figure 3). The
Width of joint b 100 mm results include the overall response of the shear connection,
Distance between loops s 300 mm failure of the joint mortar between the precast elements, and
Total length of joint L 1,280 mm also local failure of the shear keys. The results were
Strength of precast concrete fc,element 49.6 MPa dependent on the quality of the sprayed pattern, the light
Maximum aggregate size in mortar dmax 4 mm settings, and the care taken in the adjustment of the camera.
The results covered only the development of cracks on the
surface of the joint, however, the analysis provided invalua-
3.2 | Digital image correlation ble information on joint behavior during loading.
In the present investigation, digital image correlation (DIC)
was used to study the relative displacements on the surface
of the shear connection, including development of cracks in 3.3 | Test results
the joint mortar. The analysis was performed with the pro- Figure 4 presents the general characteristics of the load–
gram Aramis.25 An example of application of the same displacement response of the different tested connections.
194 SRENSEN ET AL.
The measured displacements are relative displacements a stiffness similar to that of a monolithic wall. At a relatively
(in the longitudinal direction of the connection) between the small load level, cracks develop at the interface between the
two precast elements. Figure 4a can be used as a direct joint mortar and the precast element, slightly reducing
comparison between Series R and P, where the behavior of the stiffness. At a higher load level, diagonal cracks between
the reference specimens complies with previous investiga- the corners of each pair of opposite shear keys start to emerge
tions, for example, as described in detail by Hansen et al.4 on the surface, as indications of diagonal strut action. After
The first peak also appears to be the global peak, which is diagonal cracking, the stiffness of the joint decreases until the
immediately followed by a softening branch as the shear first peak on the load–displacement curve is reached, which
displacement increases. It should be noted that the response corresponds to the value of PFP, given in Table 1. For the new
curve of specimen R1 represents a test in which the U-bar design, a drop in the response is observed after the first peak.
loops are pushed away from each other, whereas for speci- However, as the displacement increases, the load increases
mens R2 and R3, the U-bars are pushed toward each other again and reaches approximately the same level as the first
as the shear displacement increases. This indicates a very peak. The relation between the first peak load and the ultimate
unfortunate property of the conventional design because the load depends on the key design, which turns out to be one of
postpeak response is apparently dependent on the loading the main parameters that control the failure mechanism. Rup-
direction. It may very well be due to this fact that the post- ture of the U-bars starts to take place at a displacement in the
peak behavior of the conventional design in the literature is range of 12–20 mm, depending on the reinforcement configu-
reported both as brittle and as ductile. ration. The ultimate load of the joint, indicated as PU in
Figure 4a clearly illustrates the main difference between Table 1, is typically found at large displacements. The load
the two designs. The loads corresponding to the first peak are level in Figure 4b is higher than the load level in Figure 4a
comparable for specimens having identical As (as given in due to the difference in the reinforcement area, As, per loop
Table 1) and identical key configuration. However, the post- connection.
peak behavior differs significantly as the new design exhibits The first peak capacity, PFP, is governed by several fac-
a pronounced ductile behavior. It should be noted that the ref- tors, as identified in the above-mentioned literature. In the
erence design (Series R), with limited ductility, can be classi- present study, where the tensile capacity of the loop connec-
fied as a 1-on-1 connection. The total amount of looped tions was designed to be governed by U-bar yielding, the
reinforcement in the connection may have influenced the test magnitude of PFP is influenced by the geometry of the shear
results (when comparing the ductility of 1-on-1 connections keys. A larger key area generally results in a higher first
with the ductility of 1-on-2 and 2-on-2 connections). How- peak capacity. Specimen II2 has a smaller key area com-
ever, the most dominant influence on the test results is most pared with V2 and VIII2. This explains the lower first peak
probably due to the orientation of the U-bar loops and the capacity for II2, however, the response after first peak
presence of the transverse double-headed lacer bars. This duc- shows the same tendencies as that of specimen VIII2
tile behavior is especially observed for the 2-on-2 connections, because both specimens had identical loop reinforcement
see Figure 4b, which presents examples of the main findings configuration (see Figure 4b). Furthermore, it is seen that
of the test Series I–IX. Before cracking, the joint behaves with the first peak capacities of V2 and VIII2 are rather similar
(a) (b)
550 550
500 500 V2
450 450
VIII 2
400 400
P2 ↓ ↓P1 II 2
350 P3 350
P (kN)
P (kN)
300 P4 300
250 250
R3
200 200
R2
150 150
100 100
R1 II 2 Peaks
50 50 V 2 Peaks
VIII 2 Peaks
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Measured longitudinal displacement [mm] Measured longitudinal displacement [mm]
Comparison of series R to series P Load-displacement curves for the new design
FIGURE 4 Examples of performance of tested shear keyed joints.

SRENSEN ET AL. 195
as the key areas are identical. However, after the first peak, design compared to the conventional solution. To quantify
the two specimens behave differently as the governing fail- the ductility of a shear joint, the concept of relative strain
ure mechanisms are different (referring to Mechanisms B energy described by Engström27 may be considered.
and C introduced in Figure 7). The small key depth, dk, of Engström compared the maximum resistance with the aver-
specimen V2 favors shearing of the key corners, whereas age force that can be resisted by the connection during the
the larger key depth of specimen VIII2 results in complete entire displacement spectrum. This results in an average-to-
shearing of the key. These partly or complete key shearing peak ratio less than or equal to unity, where unity is the ideal
failure mechanisms are in agreement with the findings for rigid plastic behavior. In order to refine this measure, a duc-
the conventional keyed joint described by Hansen et al.4 tility index as defined in Equation 1 is introduced:
Nimityongskul and Liu9 also observed these failure ð δmax
mechanisms, and they interpreted the failure with partial 1 P ð δÞ
DI = dδ ð1Þ
shearing of the key corners as a consequence of an increase δmax − δFP δFP PFP
of the key area. A correlation between this failure mode and The idea here is to evaluate the ability of the joint to
the depth of the shear keys has not been investigated dissipate energy in the displacement regime δFP − δmax,
until now. where δFP corresponds to the shear displacement at the
occurrence of the first peak capacity, while δmax is the maxi-
3.4 | Detection of failure mechanisms mum shear displacement capacity of the connection. The
displacement capacity, δmax, can be defined as the displace-
As DIC was used to monitor the cracking process on the
ment where rupture of U-bars initiates or taken as a fixed
surface of the specimen, the experimental failure mechan-
predefined value. The index, DI, is the ratio between the
isms could be detected. For specimens with keys hidden in
dissipated energy (see the filled area in Figure 6), and the
the joint, that is, Series I–IV, where hk < t, failure of the
value PFP (δmax − δFP), which reflects the energy of a per-
keys was only observed indirectly as displacements at the
fectly plastic connection having the capacity PFP (see the
casting joint and as diagonal cracking in the joint mortar
hatched area in Figure 6). The ductility index may attain a
(also see Figure 9). For specimens with keys having hk = t,
value larger than unity. An index DI > 1.0 indicates that the
cracking of the keyed area was clear and visible when it
joint has a robust behavior because it will be able to absorb
occurred. Figure 5 shows an example of a complete shear-
the potential energy released when, for example, gravita-
ing of a single key, where it is also observed that the pre-
tional loads (applied in a load-controlled manner) reach the
existing diagonal crack closes almost completely as the key
first peak capacity, PFP. The ductility index is highly
is sheared off. From the load–displacement response of the
dependent on the total shear area of the keys, Ak, as the first
specimen, it appears that the observed first peak capacity
peak capacity increases with increasing Ak. Table 1 contains
was related to the shear failure of the keys. On this basis, it
calculated values of DI for the tested joints. The maximum
seems reasonable to conclude that failure of the shear keys
shear displacement is chosen as 13 mm for 2-on-1 connec-
also governs the first peak capacity of the specimens with
tions and 20 mm for the 2-on-2 design as rupture of the
keys hidden in the joint (hk < t). DIC measurements of the
reinforcement loops was observed around this magnitude of
cracking/failure patterns have served as inspiration when
displacement. In general, the new design has a much higher
developing collapse mechanisms used in the upper-bound
DI index than the conventional design. The most important
calculations of the first peak capacity, PFP, (see Section 4).
factor for obtaining a high ductility index is the governing
failure mechanism. For practical application, it should be
3.5 | Ductility of connections noted that a mechanism with key corner shearing leads to
It appears from the test results that a much more ductile the most ductile and robust response. This issue will be fur-
load–displacement response can be obtained by the new ther addressed in Section 4.
(a) (b)
At first peak load, PFP Just after first peak load, PFP
FIGURE 5 Example of complete key shearing (keys indicated with dashed lines) at first peak load, PFP, specimen IX2, dk = 28 mm .
196 SRENSEN ET AL.
PU material, an effectiveness factor ν is introduced into the the-

500 PFP oretical solutions.29 The ν-factor depends on the type of
problem and is usually found by calibration with tests. For
keyed joints transversely reinforced with high-strength wire
loops, Jørgensen and Hoang23,32 suggest to adopt a
400
ν-formula similar to the one used for beam shear
problems,33 but modified to the geometric layout of the
P (kN)
300
keyed shear joint. Furthermore, the factor was adjusted to
200 fit the shear capacity of joints cast with mortar. The ν-factor
for mortar joints proposed by Jørgensen and Hoang is
adopted in this work:
100
0:75 1
ν = pffiffiffiffi 1 + pffiffiffiffiffi ≯1:0 ðfc in MPa and Lk in mÞ ð3Þ
fc Lk
0 δFP
0
5 10 15 20 It should be noted that the dependency of ν on fc and Lk
δ (mm) δmax = 20 mm basically reflects softening effects and size effects, which in
the end is also due to softening. According to Equation 3, a
FIGURE 6 Example of calculation of the ductility index, DI, for specimen decrease in key length will increase the effectiveness factor,
III2, DI = 0.98. which explains why identical key areas may lead to differ-
ent tested load-carrying capacities, depending on the Lk/hk
4 | FA IL UR E M E CHAN I SMS A ND ratio. In the test Series I–IX, the effectiveness factor ranges
U PPER - BOUN D S OL UT IONS from 0.43 to 0.52.
As demonstrated by the DIC analysis (see Section 3.4), the

first peak load, PFP, is governed by failure of the shear keys. 4.1 | Failure mechanisms
In this paper, so-called first-order rigid plastic upper-bound The load-carrying capacity, Pcal, related to a specific failure
models will be developed to predict PFP. The ultimate load, mechanism is found by solving the work equation, in which
PU, of the joints is associated with large displacements and the rate of work performed by the external loads must equal
cannot be modeled in the same simple manner. Analytical the rate of internal work dissipated in the yield lines.
modeling of PU would require second-order plastic analyses Figure 7 shows the three basic failure Mechanisms A, B,
accounting for the change of geometry. The theoretical and C considered in this study. The mechanisms have been
treatment of PU is not a part of this paper. In the following, identified partly on the basis of theoretical reasoning and
concrete, mortar, and reinforcing steel are assumed to be partly with inspiration from the experimentally observed
rigid perfectly plastic materials obeying the associated flow failure modes. For all three failure mechanisms, it is
rule. Concrete and mortar are considered as modified Cou- assumed that the precast element on the right-hand side
lomb materials with zero tensile strength. For plain strain experiences a rigid body motion described by the displace-
problems, the energy dissipated per unit area of a failure ment vector u:
surface (yield line) may be determined as follows:28,29
u
1 u= t ð4Þ
WA = νfc ð1 − sin⁡αÞjuj, α ≥ φ ð2Þ ul
2
where α is the angle of the displacement vector with the
The rate of external work is then given by:
yield line and φ is the internal angle of friction. The internal
angle of friction is a material property, which depends WE = Pcal ul ð5Þ
partly on the aggregate sizes and partly on the aggregate
For Mechanisms A and B, it is more convenient to
content of the matrix.29 Triaxial tests by Dahl30 indicate that
express the components of u by |u| and the angle α as fol-
φ also depends on the confinement pressure. However, for
lows (see Figure 7):
normal strength concrete and low confinement pressures,
the internal angle of friction is normally taken to be ul = jujcos⁡α ð6Þ
φ = 37 . For normal strength mortar with confinement pres-
sures less than the uniaxial compressive strength of the mor- ut = jujsin⁡α ð7Þ
tar, Nielsen31 reported tests indicating an internal angle of The rate of internal work, WI, for the three mechanisms
friction around 30 . In this study, it is assumed that may in general be written as:
φ = 30 for the mortar material used to grout the joints.
As neither concrete nor mortar is a perfectly plastic WI = WIc, j + WIs + WIsL ð8Þ
SRENSEN ET AL. 197
(a) P (b) P (c) P
t t t
l l l
Element Element Element Element Element Element

at rest in motion at rest in motion at rest in motion
ut ut
ut
α α γ α
ul u ul u ul u
α
γ u
P P P
Mechanism A - Key cut off Mechanism B(D) - One diagonal yield line Mechanism C - Inclined key cut off
FIGURE 7 Basic failure mechanisms.
WIc, j = WA Aj ð9Þ 4.2 | Mechanism A: Key cut off
WIs = As fy ut ð10Þ To solve the work equation for Mechanism A, the rate of
internal work is found as the sum of contributions from
WIsL = AsL fyL ul ð11Þ n shear keys being sheared off (using Aj = Ak) and the con-
where WIc, j is the contribution from a concrete/mortar yield tribution from (n + 1) reinforcement loops stressed to
yielding:
line with the area Aj, WIs is the contribution from the
U-bars, and WIsL is the contribution from the locking bar. 1
WI = n νfc ð1 −sin⁡αÞAk juj + ðn + 1ÞAs fy ut ð13Þ
Yielding of the locking bar is only required in 2
Mechanism B. An upper-bound solution is established from WE = WI:
For a general description, the following parameters are
introduced and explained in Table 3: τ 1 −sin⁡α Φ
= + tan⁡α ð14Þ
νfc 2⁢cos⁡α ν
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dk
Ak = Lk hk , Ad = t b2 + L2k , Ai = hk ;
sin⁡γ TABLE 3 Symbols used in the theoretical determination of the first peak
capacity
b n + 1 As fy AsL fyL
tan⁡β = ,Φ = , ΦL = Symbol Definition
Lk n Ak fc nAk fc
As Reinforcement area per loop connection
4π4 ϕ2 for 2-on-2 connections
The parameters introduced make it easier to derive gen- 2 π4 ϕ2 for 2-on-1 connections
eral formulas for calculation of a joint with n shear keys Ak Area of one shear key
and (n + 1) pairs of U-bar loops crossing the joint interface. Ad Area of diagonal yield line
It is convenient to define the shear capacity of the joint by Ai Area of inclined yield line in a shear key
an average shear stress, τ, which is related to the total area Φ Reinforcement degree of loop connection
of the shear keys: ΦL Reinforcement degree of locking bar

β Slope of diagonal yield line
τ Pcal
= ð12Þ γ Slope of inclined yield line in a shear key
νfc nAk νfc
198 SRENSEN ET AL.
The optimal solution is found by minimizing the expres- τ dk 1 − sin⁡φ Φ

= + tanðγ + φÞ ð20Þ
sion with respect to the angle of displacement, α. The opti- νfc 2Lk sin⁡γ⁢cosðγ + φÞ ν
mal solution is found when:
The critical angle of the inclined yield line is found as:

2Φ 0 1
α = arcsin 1 − , α≥φ ð15Þ
ν B cos⁡φ C
γ = arctan@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA ð21Þ
Φ 2Lk cos⁡φ
From Equation 15, it is implicitly given that the key area sin⁡φ + 1 + ν dk 1− sin⁡φ
influences the optimal angle of displacement and thereby
the capacity of the shear connection. It should be noted that It appears that the internal angle of friction of the joint
the expression, with only slight change of notation, is simi- mortar influences the capacity significantly and to a large
lar to the findings of Jensen19 and Christoffersen.22 extent dictates, in combination with the key length to depth
ratio (Lk/dk), which of the failure mechanisms (A, B, or C)
that constitutes the critical mechanism.
4.3 | Mechanism B: One diagonal yield line
For Mechanism B, the rate of internal work consists of the
following contributions: (n − 1) times WIc, j with Aj = Ak,
one time WIc, j with Aj = Ad, (n + 1) reinforcement loops 5 | I N F L U E N C E OF K E Y G E O M E T R Y ON
FAILURE MODE
stressed to yielding, and one contribution from the locking
bar stressed to yielding. The upper-bound solution is found
From the derived expressions for the load-carrying capac-
to be:
ity (Equations 14, 16, and 20) and the corresponding
τ n − 1 1 − sin⁡α Ad 1 −sinðβ + αÞ Φ ΦL optimal angles of displacement, it is evident that the
= + + tan⁡α + geometry of the joint and, in particular, the geometry of
νfc 2n cos⁡α 2nAk cos⁡α ν ν
ð16Þ the keys, plays an important role in defining the govern-
ing failure mechanism. Figure 8 contains the results of a
which has a minimum when the angle of displacement is: theoretical comparison of the load-carrying capacity
n− 1 + htk −2n Φν
! related to the three basic failure mechanisms. The calcula-
α = arcsin , α≥φ ð17Þ tions have been performed by assuming a reinforcement
n − 1 + AAdk
arrangement similar to the one used in the experimental
From Equation 17 it can be seen that the ratio between the program Series I–IX. Figure 8a demonstrates the influence
height of the key and the thickness of the connection influ- of the key height, hk, and it appears that a higher relative
ences the optimal solution for this particular failure key height, hk/t, favors Mechanism B compared with a
mechanism. small relative key height that favors Mechanism A.
Figure 8b demonstrates the influence of the key depth on
4.4 | Mechanism C: Inclined key cut off the failure mechanism of a joint configuration similar to
the test specimens of Series V–IX, where the relative key
For Mechanism C, the rate of internal work is found as height hk/t = 1. As expected, the smaller key depths favor
n times WIc, j (with Aj = Ai) plus the contribution from Mechanism C.
(n + 1) reinforcement loops. In this mechanism, the angle The transition point (in Figure 8b) between the failure
between the l-axis and the inclined yield line is γ (see mechanisms is of particular interest because the deforma-
Figure 7c). The components of the displacement vector are tion characteristics of the joint depend on the governing
given by: failure mechanism. As shown in Figure 4b, shearing of the
ul = jujcosðγ + αÞ ð18Þ key corners (specimen VII) results in a pronounced ductile
behavior, which in turn leads to a high ductility index. In
ut = jujsinðγ + αÞ ð19Þ this context, it should be noted that the internal angle of
As the relationship between the transverse and the longi- friction for mortar is of significant interest because the
tudinal displacement is dependent on the sum of γ and α, transition point (see Figure 8b) partly depends on the mag-
the lower limit of the condition α ≥ φ is reached at lower nitude of φ. Aramis recordings of the relative displace-
degrees of reinforcement for this mechanism compared with ments at the first peak load indicate that φ = 30 is an
Mechanisms A and B. Therefore (and to simplify), it is for appropriate choice for the material used in this study, and
this particular mechanism assumed that α = φ = 30 . The furthermore, it is in accordance with the investigations by
optimization of the upper-bound solution is then reduced to Nielsen.31 The recorded relative displacements were com-
an optimization problem involving only the angle γ, which is pared with the theoretical relative displacements for test
governed by the key dimensions and the internal angle of specimens where the angle of displacement was predicted
friction φ. The load-carrying capacity is given by: to be α = φ.
SRENSEN ET AL. 199
(a) (b)
0.9 0.9
0.8 0.8
0.7 0.7
Mechanism C
0.6 0.6 ↓
Mechanism A
↓ Mechanism B
[-]
[-]
↓
0.5 0.5
υ fc
υ fc
↑ ↑
τ
τ
0.4 0.4
Load carrying capacity
h k /t = 0.25 - A Transition
0.3 0.3
h k /t = 0.25 - B
0.2 h k /t = 0.5 - A 0.2
h k /t = 0.5 - B
0.1 h k /t = 1.0 - A 0.1
h k /t = 1.0 - B
0 0
0 0.05 0.1 0.15 0.2 0.25
Reinforcement degree Φ [-]
0 5 10 15 20 25 30 35 40
dk [mm]
Load carrying capacity for varying relative height of Load carrying capacity for varying key depth, fc = 31
key, hk/t, fc = 31 MPa MPa, Lk = 0.140 m, hk = 0.200 m, Φ = 0.15
FIGURE 8 Illustrations of change in failure mechanism when changing geometry of the shear keys.
6 | FA IL UR E M E CHAN I SMS B ASED ON the calculation. For both cases, the contribution from the
EX PE RIMEN TA L OBSE RV AT ION S longitudinal locking bar is considered.
In practice, the length of a shear wall connection will at
Based on the experimental observations, the theoretical fail- least be equal to the height of one storey, and for this rea-
ure mechanisms for Series I–IX are evaluated. DIC mea- son, there will be many more shear keys in these connec-
surements have shown that a failure mechanism similar to tions as compared with the connections investigated in this
Mechanism B, but with a relatively large crack opening in study. When many shear keys are present, the significance
one diagonal crack prior to the first peak load, governs inde- of Mechanisms D and E will be limited. However, for the
pendently of the height of the key. Figure 9 shows Aramis limited geometry of the test specimens, the influence of the
recordings of the cracking process of specimen III2, which boundary effect included in these mechanisms is relevant.
according to the theoretical calculations should reach the The load-carrying capacity of Mechanism D is found to be:
first peak load-carrying capacity by development of Mech- τ n −1 1 − sin⁡α Φ ΦL
anism A. It appears that diagonal cracks develop even = + tan⁡α + ð22Þ
νfc 2n cos⁡α ν ν
before the first peak (see Figure 9a), and the relative displa-
cements during failure take place in one of the existing diag- The optimal angle of displacement is given as:
onal cracks (see development from b to c in Figure 9). The
2nΦ

crack opening of the diagonal cracks prior to first peak can α = arcsin 1 − , α≥φ ð23Þ
ðn −1Þν
be determined from Aramis measurements. Figure 10 shows
examples of recorded crack opening of the largest diagonal For Mechanism E (see Figure 11), the load-carrying
crack (crack opening only in the longitudinal direction of capacity, assuming α = φ, is:
the joint is shown). It appears that the crack opening before τ n −1 dk 1 − sin⁡φ Φ ΦL
first peak load, PFP, is approximately 0.4 mm, which is rela- = + tanðγ + φÞ + ð24Þ
νfc 2n Lk sin⁡γ⁢cosðγ + φÞ ν ν
tively large for mortar. This observation leads to the conclu-
sion that the dissipation in the diagonal yield line (i.e., the The critical angle, γ, of the inclined yield line in the
mortar contribution) must be significantly reduced and keys is found as:
thereby making a mechanism which is similar to Mechanism 0 1
B more critical than Mechanism A. B cos⁡φ C
γ = arctan@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA ð25Þ
In the following, two additional failure mechanisms are n Φ 2Lk cos⁡φ
sin⁡φ + 1 + ðn− 1Þ ν dk 1 −sin⁡φ
introduced, namely Mechanism D similar to Mechanism B
(see Figure 7b), but omitting the mortar contribution from It should be noted, that Mechanisms B/D and E are only
the diagonal yield line when calculating the rate of internal relevant for test Series I–IV, where the specimens had 2-on-
work, and Mechanism E based on Mechanism C, however, 2 loop connections. For specimens in Series P with 2-on-1
introducing a diagonal yield line (see Figure 11) and omit- connections, the asymmetric reinforcement arrangement
ting the mortar contribution from the diagonal yield line in favors Mechanism A or C. This can be seen in Figure 12,
200 SRENSEN ET AL.
(a)
Major principal strain at first diagonal cracking

(b)
Major principal strain at first peak load

(c)
Major principal strain just after first peak
FIGURE 9 Aramis record of strain localization and cracking behavior of shear connection around first peak load, specimen III2.
(a) (b)
550
500 PFP
450
400
350
P (kN)
300
wt w
250
Diagonal
200 crack wl
150
100
t
50 Specimen III2
Specimen IV1
0
0 0.2 0.4 0.6 0.8 1
w l (mm) l
FIGURE 10 (a) Measured longitudinal crack opening in diagonal crack between shear keys and (b) definition of crack opening.
where Aramis recordings show that no diagonal cracks were in Series I–IX, the theoretical capacity has been determined
present just after the first peak load. as the minimum value predicted from the five presented
failure mechanisms. For specimens in Series P, only
Mechanisms A and C are of interest. The yield strength of
7 | CO M PA R IS O N OF TE S TS WI T H the reinforcement loops in Series P was fy = 509 MPa, the
THEORY width of the joint was b = 80 mm, and the remaining prop-
erties are given in Tables 1 and 2.
Table 4 shows the obtained experimental first peak loads as Figures 13 and 14 show a graphical comparison where
well as the theoretical determined values. For test specimens the governing failure mechanisms are identified. The
SRENSEN ET AL. 201
P TABLE 4 Comparison of test results with theoretical values
Failure mechanism
PFP
No. PFP [kN] Pcal [kN] Pcal [-] (Observed/Predicted)
t
P 1 344.24 291.12 1.18 C/C
l
2 347.04 291.12 1.19 C/C
3 342.49 297.16 1.15 C/C
Element Element 4 331.42 297.16 1.12 C/C
at rest in motion I 1 379.02 395.34 0.96 B/A
2 416.59 403.29 1.03 B/A
II 1 366.40 412.67 0.89 B/A
ut
2 414.46 421.43 0.98 B/A
γ α
u III 1 393.04 427.62 0.92 B/D
ul
2 473.52 433.99 1.09 B/D
IV 1 439.44 438.33 1.00 B/D
2 478.17 455.20 1.07 B/D
β
V 1 475.24 500.73 0.95 E/E
2 492.86 508.21 0.97 E/E
VI 1 527.09 538.50 0.98 E/D
2 523.82 538.50 0.97 B/D
VII 1 549.17 538.50 1.02 B/D
2 524.46 538.50 0.97 B/D
P
VIII 1 507.05 538.50 0.94 B/D
FIGURE 11 Mechanism E, based on experimental observations. 2 516.97 538.50 0.96 B/D
IX 1 526.53 538.50 0.98 B/D
calculations are performed using the average compression 2 527.59 538.50 0.98 B/D
strength of the grout mortar and the reinforcement strengths Mean 1.01
given in Table 2. Figure 13 shows a comparison in which Standard deviation 0.08
the length of the shear keys, Lk, is varied and the height is
kept constant at half the panel thickness. The results thereby
compare with Series I–IV (Mechanism C is not critical). that a smaller value of φ changes the transition point toward
Figure 14 shows the comparison for varying key depths a larger key depth. However, an in-depth study of the inter-
with constant key area, that is, a comparison for Series nal angle of friction for mortar is needed to clarify the prop-
V–IX. It can be seen that the refined Mechanism D captures erty and perhaps also the validity of the normality condition
the behavior and predicts the load-carrying capacity of the for mortar materials.
specimens with large key depths. It can also be seen that Figure 13 shows the key area as the ratio between the
Mechanism E explains the cracking behavior of the speci- area of a single key compared with the joint area, At. The
mens with small key depths before Mechanism D becomes joint area is calculated using the center distance of the rein-
the governing mechanism for larger key depths. Table 4 forcement loops, given as s in Figure 3, and the height of
also contains a summary of the observed as well as pre- the specimen, t. It can be seen that the average shear stress
dicted failure mechanisms. If a failure mode including a can be higher for a smaller relative key area, as expected
diagonal crack was observed, the failure is regarded as B considering softening effects in the mortar material. Gener-
for the key cut off and E for the inclined key cut off. Mech- ally, good agreement is found between the test results and
anism D cannot be observed experimentally, but in fact an the calculations. In Figures 13 and 14, the capacity as pre-
observed Mechanism B might relate to a theoretical Mech- dicted by the Eurocode 2 formula for indented interfaces
anism D. It can be seen from Table 4 that both Mechanisms using average material strengths (tensile strength of concrete
B and E were observed in test Series VI. It should be noted calculated by use of the EC2 method), without partial safety
FIGURE 12 Major principal strain distribution just after first peak load, PFP, of ‘2-on-1’ specimen (specimen P4) from pilot test series.
202 SRENSEN ET AL.
1 response curve has been identified as the load that causes fail-
0.9 ure of the shear keys. Theoretical failure mechanisms have been
established and used to derive upper-bound plasticity solutions
to calculate the first peak capacity. The failure mechanisms for
0.8
the tested connections have been refined based on observations

Mechanism B
↓
0.7
↑ from the experiments and the results of DIC analysis.
0.6 Mechanism A
The significance of the key dimensions has been addressed
[-]
↑
and the influence of the key height and depth on the failure
0.5
νfc
τ
Mechanism D
0.4
mode has been outlined. The developed models predict the
0.3
EC2 approach
transition point between the two main failure mechanisms, in
↓
terms of key depth: complete key cutoff or inclined key cutoff
(see Figure 14). The refined Mechanisms D and E, relevant for
0.2
0.1
the limited geometry tested, captured, and explained the exper-
0 imental observations. For the design of longer connections, as
those found in practice, the theoretical basic Mechanisms A,
0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32
Ak/At [-]
B, and C presented in Figure 7, will be sufficient.
Comparison of theory with results for series I-IV, fc,
FIGURE 13
It can be concluded that the new connection design is a
average = 33.0 MPa.
feasible and promising practical solution that should be sub-
jected to further investigation with the perspective of replacing
factors has been included as well. It is clearly seen that the
the conventional solution. To adapt the new design for practi-
empirical formula of Eurocode 2 is too conservative when
cal use, it is necessary to clarify a number of issues, including:
applied to the new connection design. In this context, it
should be noted that the Eurocode 2 method does not take
into account the specific key geometry. • Detailed characterization of the properties of grout mortar
• Test of a wider range of U-bar diameters and possibly a
variation of the geometry of the U-bars
8 | CON CLU SION S • Test of the tensile capacity of the connection
• Test of anchorage properties of the lacer reinforcement
A new and construction-friendly loop connection for the assem- • Investigation and modeling of the increase in load-
bly of precast shear wall panels has been developed and tested. carrying capacity after the first peak
The structural performance of the new connection, in terms of
ductility, is superior to that of the conventional design. A ductil-
ity index has been introduced in order to evaluate and compare ACKNOWLEDGMENTS
the performance of the developed design with that of the con- The experimental program was financially supported by the
ventional design. For the tested designs, the first peak on the Danish Association for Precast Concrete Elements and the
(a) (b)
0.8
0.7
Mechanism C
0.6 ↓
↑
0.5 Mechanism D Mechanism E
↓
[-]
0.4
νf c
τ
0.3 EC2 approach

↓
0.2
0.1
0
0 5 10 15 20 25 30 35 40
dk [mm] Mechanism D Mechanism C or E
FIGURE 14 (a) Comparison of theory with results for series V-IX, fc, average = 31.0 MPa, and (b) main failure modes.
SRENSEN ET AL. 203
COWI Foundation. The test elements were produced at δFP displacement at first peak
CRH Concrete and the experimental work was conducted γ slope of inclined yield line in a shear key
with support from students at DTU Civil Engineering. Rune ϕ U-bar diameter
Pedersen, B.Eng., and Mads Herløv, B.Eng., contributed ϕLacer lacer bar diameter
with test Series R and P in their bachelor thesis, ϕL locking bar diameter
J. Svejgaard, M.Sc., contributed with test Series I–IV, and φ internal angle of friction
L. vrelid, M.Sc., contributed with test Series V–IX in their Φ reinforcement degree of loop connection
master theses. Finally, the use of double-headed studs as ΦL reinforcement degree of locking bar
lacer bars instead of conventional stirrups was suggested by ν effectiveness factor
Tim Gudmand-Høyer, Ph.D. The authors gratefully τ shear stress
acknowledge these valuable contributions.
REFERENCES
NOTATIONS 1. Dahl KKB. Bella Sky Hotel: taking precast concrete to the limit. Struct
Concr. 2014;15(4):441-447.
Ad area of diagonal yield line 2. Flindt Jørgensen K. Bella Sky Hotel: exploring the potential in precast con-
crete design. Struct Concr. 2015;16(4):449-457.
Ai area of inclined yield line in a shear key
3. fib Bulletin 43. Structural connections for precast concrete buildings, fédération
Aj area of yield line international du béton ( fib), 2008.
Ak area of one shear key 4. Hansen K, Kavyrchine M, Melhorn G, Olesen S, Pume D, Schwing H.
As reinforcement area per loop Keyed shear joints: SBI rapport 97, Technical report. Danish Building
Research Institute, Copenhagen; 1976.
AsL reinforcement area of locking bar 5. Halasz R, Tantow G. Schubfestigkeit der Vertikalfugen im Grosstafelbau.
At area of joint In: Berichte aus der Bauforschung, H. 39. Berlin, Germany: Verlag Wil-
b width of joint helm Ernst & Sohn, Berlin; 1966.
6. Cholewicki A. Loadbearing capacity and deformability of vertical joints in
D internal bend diameter of loops structural walls of large panel buildings. Build Sci. 1971;6(9):163-184.
DI ductility index 7. Pommeret M. Le comportement sous charges ou déformations répétées alter-
dk depth of shear key nées des joints verticaux entre panneaux préfabriqués. Technical Report. Saint
Remy les Chevreuxe, Centre expérimental de recherches et d'études du bati-
dmax maximum aggregate size in mortar
ment et des travaux publics; 1972.
fc compression strength 8. Fauchart J, Cortini P. Étude expérimentale de joints horizonraux entre pan-
fy yield strength of U-bar neaux préfabriqués pour murs de batiments. Annales de LÍnstitut Tech-
fy,Lacer yield strength of lacer bar nique du Batiment et Des Travaux Publics, Paris, 1972.
9. Nimityongskul P, Liu HY. Vertical shear strength of joints in prefabricated
fyL yield strength of locking bar loadbearing walls. Hous Sci. 1980;4(2):137-157.
hk height of shear key 10. Chakrabarti SC, Bhise NN, Sharma KN. Failure criterion of vertical shear
L total length of joint key joints in prefabricated wall panels. Indian Concr J. 1981;55(3):63-67.
11. Abdul-Wahab HM. An experimental investigation of vertical castellated
Lk length of shear key joints between large concrete panels. Struct Eng. 1986;64B(4):93-99.
n number of shear keys 12. Serrette RL, Rizkalla SH, Heuvel JS. Multiple shear key connections for
P shear load load-bearing shear wall panels. PCI J. 1989;34(2):104-120.
13. Rossley N, Aziz F, Chew H, Farzadnia N. Behaviour of vertical loop bar
Pcal theoretically calculated shear capacity
connection in precast wall subjected to shear load. Aust J Basic Appl Sci.
PFP first peak load 2014;8(1):370-380.
PU ultimate load 14. Vaghei R, Hejazi F, Taheri H, Jaafar MS, Ali A. Evaluate performance of
s distance between loops precast concrete wall to wall connection, APCBEE Procedia 9 (ICBEE
2013); 2014;9:285–290.
t panel thickness 15. CEN. EN1992-1-1 Eurocode 2: Design of Concrete Structures – Part 1–1:
u displacement vector General Rules and Rules for Buildings. 3rd ed. European Committee for
ul longitudinal component of u Standardization, Brussels, 2004.
16. Kaneko Y, Connor JJ, Triantafillou TC, Leung CK. Fracture mechanics
ut transverse component of u approach for failure of concrete shear keys. I: Theory. J Eng Mech.
w crack opening 1993;119(4):681-700.
wl longitudinal crack opening 17. Kaneko Y, Connor JJ, Triantafillou TC, Leung CK. Fracture mechanics
approach for failure of concrete shear keys. II: Verification. J Eng Mech.
wt transverse crack opening
1993;119(4):701-719.
WE rate of external work 18. Kaneko Y, Mihashi H. Analytical study on the cracking transition of con-
WI rate of internal work crete shear key. Mater Struct. 1999;32(217):196-202.
WIC rate of internal work from concrete 19. Jensen BC. On the Ultimate Load of Vertical, Keyed Shear Joints in Large
Panel Buildings. Technical Report. Lyngby, Denmark: Institute of Building
WIs rate of internal work from U-bars Design, Technical University of Denmark, Lyngby; 1975.
WIsL rate of internal work from locking bar 20. Chakrabarti SC, Nayak GC, Paul DK. Shear characteristics of cast-in-place
vertical joints in story-high precast wall assembly. ACI Struct J.
α angle of displacement vector
1988;85(1):30-45.
β slope of diagonal yield line 21. Abdul-Wahab HM, Sarsam SYH. Prediction of ultimate shear strength of
δ longitudinal displacement vertical joints in large panel structures. ACI Struct J. 1991;88(2):
δmax displacement capacity 204-213.
204 SRENSEN ET AL.
22. Christoffersen J, Ultimate Capacity of Joints in Precast Large Panel Con- Linh Cao Hoang
crete Buldings, Series R No. 25 [Ph.D. thesis]. Lyngby: Technical Univer-
sity of Denmark, Department of Structural Engineering and Materials,
Professor, PhD
Technical University of Denmark, Lyngby; 1997. Department of Civil Engineering
23. Jørgensen HB, Hoang LC. Load Carrying Capacity of Keyed Joints Rein- Technical University of Denmark
forced with High Strength Wire Rope Loops. Proceedings of fib symposium:
Brovej, Building 118
Concrete - Innovation and Design, Copenhagen; 2015.
24. Jørgensen HB, Hoang LC. Tests and limit analysis of loop connections between 2800 Kgs. Lyngby,
precast concrete elements loaded in tension. Eng Struct. 2013;52:558-569. Denmark
25. GOM: Aramis User Manual - Software v6.1 and higher. Braunschweig: [email protected]
GOM Optical Measuring Techniques; 2009.
26. Pereira EB, Fischer G, Barros JAO. Image-based detection and analysis of
crack propagation in cementitious composites. In: Leung C, Wan KT, eds. John Forbes Olesen
Paper presented at: Proceedings of the International RILEM Conference on Associate Professor, PhD
Advances in Construction Materials Through Science and Engineering;
2011; 1–8; HongKong, China.
27. Engström B. Ductility of Tie Connections in Precast Structures [Ph.D. the- Technical University of Denmark
sis]. Goteborg: Chalmers University of Technology; 1992. Brovej, Building 118
28. Jensen BC. Nogle plasticitetsteoretiske beregninger af beton og jernbeton
2800 Kgs. Lyngby,
(English: Some Applications of Plastic Analysis to Plain and Reinforced
Concrete), Report 111 [Ph.D. thesis]. Copenhagen, Denmark: Technical Denmark
University of Denmark, Lyngby; 1976. [email protected]
29. Nielsen MP, Hoang LC. Limit Analysis and Concrete Plasticity. 3rd ed.
CRC Press, Taylor & Francis Group, Boca Raton, Florida, US; 2011.
30. Dahl KKB. A Failure Criterion for normal and High Strength Concrete, Tech-
Gregor Fischer
nical Report. Lyngby, Denmark: Technical University of Denmark; 1992. Associate Professor, PhD
31. Nielsen CV. Triaxial behavior of high-strength concrete and mortar. Department of Civil Engineering
ACI Mater J. 1998;95(2):144-151.
32. Jørgensen HB. Strength of Loop Connections between Precast Concrete
Elements [Ph.D. thesis]. Denmark: University of Southern Denmark, Brovej, Building 118
Department of Technology and Innovation, Odense; 2014. 2800 Kgs. Lyngby,
33. Zhang J-P. Diagonal cracking and shear strength of reinforced concrete
Denmark
beams. Mag Concr Res. 1997;49(178):55-65.
[email protected]
AUTHOR’S BIOGRAPHIES
Jesper Harrild Sørensen How to cite this article: Sørensen JH, Hoang LC,
M.Sc., PhD student Olesen JF, Fischer G. Test and analysis of a new ductile
Department of Civil Engineering shear connection design for RC shear walls. Structural
Technical University of Denmark Concrete. 2017;18:189–204. https://doi.org/10.1002/
Brovej, Building 118 suco.201600056
2800 Kgs. Lyngby,
Denmark
[email protected]
Paper III
”Test and lower bound modeling of keyed shear connections in
RC shear walls”
Jesper H. Sørensen, Morten A. Herfelt, Linh C. Hoang, Aurelio Muttoni
Published in: Engineering Structures (2018), 155:115-126
239
Engineering Structures 155 (2018) 115–126
Contents lists available at ScienceDirect
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
Test and lower bound modeling of keyed shear connections in RC shear MARK
walls
⁎
Jesper Harrild Sørensena, , Morten Andersen Herfelta,b, Linh Cao Hoanga, Aurelio Muttonic
a
Technical University of Denmark, Department of Civil Engineering, Brovej, Bygning 118, 2800 Kgs. Lyngby, Denmark
b
Niras A/S, Sortemosevej 19, 3450 Allerød, Denmark
c
Ècole Polytechnique Fédérale de Lausanne, ENAC, Station 18, CH-1015 Lausanne, Switzerland
A R T I C L E I N F O A B S T R A C T
Keywords: This paper presents an investigation into the ultimate behavior of a recently developed design for keyed shear
Keyed shear connections connections. The influence of the key depth on the failure mode and ductility of the connection has been studied
Precast concrete by push-off tests. The tests showed that connections with larger key indentations failed by complete key cut-off.
Push-off tests In contrast, connections with smaller key indentations were more prone to suffer local crushing failure at the key
Rigid-plasticity
corners. The local key corner crushing has an effect on the load-displacement response, which is relatively more
Lower bound solutions
ductile. In addition to the tests, the paper also presents lower bound modeling of the load carrying capacity of
the connections. The main purpose of the lower bound model is to supplement an already published upper bound
model of the same problem and thereby provide a more complete theoretical basis for practical design. The two
models display the same overall tendencies although identical results are not possible to obtain, due to differ-
ences in the basic assumptions usually made for upper and lower bound analysis of connections. It is found that
the test results, consistent with the extremum theorems of plasticity, are all lying within the gap between the
upper and the lower bound solution. The obtained results finally lead to a discussion of how the two models can
be used in practice. The primary merit of the upper bound model lies in its simplicity (a closed-form equation).
On the other hand, the lower bound model provides safe results, but is more complicated to apply. It is therefore
argued that the upper bound model may be used in cases, where calibration with tests has been carried out. The
lower bound model should be applied in situations, where the design deviates significantly from the config-
urations of the available tests.
1. Introduction The experimental results reported in [1] showed that the geometry
of the shear keys, and in particular the depth of the keys, plays an
A new design for keyed shear connections between precast wall important role for the ductility of the connection. In addition, the
elements has recently been proposed and the structural performance ability of the U-bars to develop yielding is decisive for a desirable be-
has been experimentally investigated [1]. The conceptual layout of the havior of the connection in the ultimate limit state. Inspired by ex-
design can be seen in Fig. 1. Unlike the conventional solution, where perimental observations, rigid-plastic upper bound solutions have been
the overlapping U-bars are looped in the horizontal plane [2,3], this developed for prediction of the capacity of the connection [1].
new layout consists of U-bars looped in vertical planes. The solution The aim of this paper is twofold. First and foremost, there is a need
contains double T-headed bars (lacer bars) placed perpendicular to the for additional tests to explore the behavior of the new design and in
plane of the wall elements to ensure efficient transfer of tension be- details study the influence of the key depth. Therefore, an in-depth
tween the U-bars [4]. It is also possible to add a vertical locking bar experimental investigation of the failure of the shear keys has been
similar to the conventional design of keyed shear connections. As carried out. To widen the experimental database (not only with respect
shown in [1], the new design has a much more ductile shear behavior to the number of tests but also with respect to parameter variations) U-
than the conventional solution and can in addition be more construc- bar diameters smaller than those used in the first test campaign have
tion-friendly in case of vertically lowered panels. Thus, the new design been investigated. Secondly, from an analytical point of view, there is a
has the potential to be used in e.g. high-rise buildings in cases where in need to establish lower bound solutions for the shear capacity of the
situ walls are replaced by precast concrete elements. connections in order to evaluate the already developed upper bound
⁎
Corresponding author.
E-mail address: [email protected] (J.H. Sørensen).
https://doi.org/10.1016/j.engstruct.2017.11.004
Received 13 June 2017; Received in revised form 3 November 2017; Accepted 6 November 2017
0141-0296/ © 2017 Elsevier Ltd. All rights reserved.
J.H. Sørensen et al. Engineering Structures 155 (2018) 115–126
Fig. 1. (a) Shear connection between RC precast wall

elements and (b) conceptual design of new connec-
tion solution (illustrations from [4]).
solutions [1]. The set of solutions can be used to bracket the theoreti-
cally exact solution from below and above, which is extremely useful
for practice.
For looped connections, upper bound solutions are relatively
straight forward to establish when based on observed failure modes
[1,5,6], while optimal lower bound solutions are more difficult to de-
velop. Simple lower bound solutions for keyed shear connections have
e.g. been proposed by Christoffersen [7] and further developed in
Nielsen and Hoang [8]. However, the models contain only single uni-
axial strut action and are in many cases too conservative. Recently,
Herfelt et al. [9] presented a numerical framework for obtaining op-
timal lower bound solutions based on finite element limit analysis
(FELA). The numerical tool calculates the optimal stress distribution
and the corresponding failure mechanism in the joint mortar. In this
paper the results obtained by FELA will be used as inspiration to es-
tablish analytical lower bound solutions.
Compared to the single strut solution [7], the following numerical
and analytical models utilize combinations of struts with different in-
clinations to optimize the theoretical load carrying capacity. In addi-
tion, the ability of the grout-to-panel interface to transfer shear stresses
is utilized which eventually leads to non-hydrostatic biaxial stress
conditions in the nodal zones. The stresses in the nodal zones at the key
corners are modeled by use of the concept of homogeneous stress fields.
The concept was treated e.g. in Refs. [10–12].
2. Experimental program
The experimental program comprised 12 push-off specimens with

identical loop configuration and with a variation of the key depth, dk ,
from 10 mm to 20 mm. Compared to the first test campaign [1], this
program used U-bars with a diameter of 6 mm instead of 8 mm and the
shear keys in all specimens had a height equal to the thickness of the
precast element, i.e. hk = t , which enabled identification of the local
failure of the individual shear keys by use of digital image correlation Fig. 2. General layout of push-off test specimens (illustration from [1]).
(DIC). The general geometry of the test specimens can be seen in Fig. 2
and the material properties are given in Tables 1 and 2. The specimen
identification refers to the depth of the shear keys (e.g. D10 designates a
key depth of dk = 10 mm). Each design was replicated twice, denoted A
and B. The connections were grouted with a mortar with a maximum
aggregate size of 4 mm and a vertical locking bar was included to
minimize the extent of diagonal cracking between the shear keys. The
116
Table 1 different behaviors depending on the failure mode of the shear keys.
Specifications of the test specimens, including material properties. The plots in Fig. 3(a) are the results of specimens with smaller key
depths (dk = 10–16 mm) where failure took place as a local shearing of
No. fc [MPa] Lk [mm] Ak [mm2] dk [mm] As [mm2] VFP [kN]
the key corners (as illustrated in the graph). For larger key depths
D10 A 44.6 120 24000 10 113 448.56 (dk = 16–20 mm), complete shearing of the shear keys governed the
B 44.6 120 24000 10 113 448.62 first peak load, which leads to the load-displacement characteristics
D12 A 44.6 120 24000 12 113 471.74 shown in Fig. 3(b).
B 44.6 120 24000 12 113 496.36
The transition from key corner shearing to complete key shearing
D14 A 44.6 120 24000 14 113 510.91
B 44.6 120 24000 14 113 519.16 was dependent on a number of factors, including the geometry of the
D16 A 44.6 120 24000 16 113 543.30 shear keys, the material properties of the mortar, and the strength of the
B 44.6 120 24000 16 113 541.57 reinforcement. From post-test examinations it was found that specimens
D18 A 42.0 120 24000 18 113 540.73
with dk = 16 mm could fail both by local key corner shearing (D16B)
B 42.0 120 24000 18 113 537.50
D20 A 42.0 120 24000 20 113 526.62
and by complete failure of the shear key (D16A). This indicates that the
B 42.0 120 24000 20 113 517.03 theoretical transition between the two failure modes for this particular
test series takes place at approximately this key depth. The load–dis-
fc refers to the compressive strength of the grout measured on ϕ 100 × 200 mm cylinders. placement relationships of D16A and D16B are both shown in Fig. 3. It
π
was found that both had comparable first peak loads, VFP (see Table 1).
As describes the reinforcement area per loop connection, i.e. 4 ϕ2 .
4 However, the residual load level after first peak was higher for spe-
cimen D16B which experienced key corner shearing. This was a general
Table 2
observation that specimens suffering local key corner shearing had a
Material properties and geometric values.
more ductile load-displacement relationship (i.e. residual load level
Description Symbol closer to first peak load). In contrast to this, a relatively larger drop of
the load immediately after first peak was observed for all the speci-
U-bar diameter ϕ 6 mm
mens, which failed by complete key shearing. Both types of failure were
Yield strength of U-bar fy 517 MPa
accompanied by development of diagonal cracks in the grout. Fig. 4
Lacer bar diameter ϕLacer 12 mm
shows examples of local failure of the keys in combination with diag-
Yield strength of lacer bar f y,Lacer 552 MPa
onal cracks which formed the global failure mechanism. Post-test ex-
Internal bend diameter of loops D 45 mm
Width of Joint b 80 mm
aminations also showed large plastic deformations in the U-bars. This
Distance between loops s 300 mm observation confirms that the design of the loop connection (i.e. the
Inclination of key corner θk arctan
1 overlapping length, the diameter of lacer bar and the internal bend
2
Total length of joint L 1280 mm diameter of the U-bars) did enable transfer of the yield capacity of the
Max aggregate size in mortar dmax 4 mm U-bars through the overlap.
Diameter of locking bar ϕL 12 mm
It was generally observed that the first peak load, VFP , increased
Yield strength of locking bar f yL 599 MPa with increasing key depth until complete shearing of the keys becomes
Panel thickness t 200 mm the critical failure mode. Then VFP remained practically constant and
Strength of precast panels fc,element 58.2 MPa independent of a further increase of the key depth, cf. Table 1. The
residual load level after first peak appeared somewhat constant within
the two types of failure and must thereby be related to the layout and
U-bar loops were designed according to Refs. [4,6] to transfer the full the properties of the loop reinforcement, which were kept constant in
yield force of the U-bars. this study.
Based on the experimental results, it seems that connections in
2.1. Test results practice should be designed to be governed by local shearing of the key
corners, as this minimizes the difference between first peak load and the
Fig. 3 shows examples of typical tested load-displacement re- residual load level.
lationships. The depicted displacements correspond to relative long-
itudinal displacements between the two precast elements, measured at
both ends of the connection and averaged. The figure shows two
Fig. 3. Experimentally recorded load-displacement curves.
117
Fig. 4. Identification of failure modes just after first peak

load by use of digital image correlation.
Fig. 5. Distribution of the smallest principal stress

(MPa, tension positive) in connections with
dk = 10, 14 and 20 mm, obtained by FELA,
fc = 42 MPa.
3. Rigid-plastic analysis special case of the finite element method, where a rigid-plastic material
model is assumed. For lower bound solutions, a set of constraints ensure
The first peak load, VFP , can be estimated by use of upper or lower that the equilibrium conditions and the yield criteria are satisfied, i.e. a
bound models assuming rigid-plastic material behavior. Rigid-plastic statically admissible and safe stress field, while the load is sought to be
modeling can also be used to obtain an estimate of the inelastic maximized. A detailed description of the numerical framework for
load–displacement curve. This, however, requires a second order plastic modeling of keyed connections can be found in Herfelt et al. [9] and a
analysis where change of geometry and large displacements are taken summary of the concept is given in Appendix A. Here only the main
into account [13–16]. assumptions are needed for comparison with the analytical model.
The objective of this paper is limited to the calculation of VFP . As The relevant parameters for the numerical analysis are the geometry
mentioned, an upper bound model has already been established [1]. of the shear keys, the mechanical degree of transverse reinforcement,
The model predicts the shear capacity of keyed connections in a sa- interface properties, and the grout properties. Based on the experi-
tisfactory manner provided that an effectiveness factor, ν , is introduced. mental results the transverse reinforcement degree is determined by the
The problem, however, with a pure upper bound approach is that it yield force of the U-bars. In the model, the overlapping loops are sim-
remains an open question whether the adopted effectiveness factor, plified as continuous reinforcement crossing the connection. Moreover,
obtained by calibration with tests, to a significant extent also accounts plane stress condition is assumed and the mortar is modeled as a
for the fact, that the developed upper bound model is not necessarily modified Coulomb material without tensile strength. For the interface
the exact/correct one. To answer the question and eventually to eval- between the grout and the precast concrete, a Coulomb friction cri-
uate the developed upper bound model, lower bound solutions will be terion is assumed with a friction coefficient, μ = 0.75 (corresponding to
established in the following. This includes both numerical solutions as smooth casting joints [8,17]) and a neglectable cohesion. However,
well as analytical ones. friction is only considered active in the indented areas. The main ar-
gument for this is that the normal stress required to activate friction
stems from tension in the transverse reinforcement, which represents a
4. Numerical lower bound solutions based on finite element limit
form of passive confinement in contrast to active confinement from an
analysis (FELA)
external normal force. Hence, since the load is anticipated to be carried
mainly by strut action between the shear keys (when no active normal
The main purpose of the performed numerical lower bound calcu-
force is applied), then only the indented areas (which experiences
lations is to provide inspiration for the type of stress field to be adopted
compression from the strut action) can transfer friction. The
in an analytical lower bound model. FELA can be considered as a
118
longitudinal locking bar has not been included in the numerical model,
as it is judged to entail stress fields that, at the current stage, are too
complicated for the initial establishment of analytical lower bound
models.
Fig. 5 shows the calculated distribution of the smallest principal
stress (maximum compressive stress) in the joint mortar for some spe-
cimens from the experimental program. The stress field is only visua-
lized for the joint mortar, since the precast elements in the experimental
program as well as in the numerical model had over-strength. In the
calculations, the strength of the mortar was taken as fc , according to
Table 1. This entails that the effectiveness factor, ν , was chosen as unity
(note that the absolute value of ν is not important when the results are
used only to compare with the analytical solutions).
From the optimized stress distributions, it can be seen that the load
transfer mechanism consists of a combination of compression struts
spanning over one or two shear keys, i.e. struts with different inclina-
tions. This type of stress field is obviously more complicated than the
single strut solutions [7]. The numerical results will in the following
form the basis for how to choose the stress field in an analytical lower
bound solution and finally the results of the two methods will be Fig. 7. (a) Stresses along boundaries of Triangle I (Solution 1) and (b) resultants of
stresses on boundaries.
compared.
5.1. Solution 1: Single struts spanning over one shear key

5. Analytical lower bound solutions
For the purpose of analysis, a l−t coordinate system, referring to the
Inspired by the results of FELA, an analytical model for the load longitudinal and transverse directions of the connection, is defined, see
carrying capacity of keyed shear connections will be established. Some Fig. 6. The following relations for the geometrical parameters shown in
simplifications have to be introduced in order to carry out the analytical Fig. 6 are introduced:
calculations. Two basic stress fields are considered, in the following
e1
denoted as Solution 1 and Solution 2, and the contribution from the tanθk =
locking bar is disregarded. Solution 1 consists of parallel struts span- dk (1)
ning over a single shear key (Fig. 6). Solution 2 combines struts span-
ning over one shear key and two shear keys, respectively (Fig. 8). The Lk −e
tanθA,1 =
struts carry uniaxial compression, while the nodal zones are stressed in b (2)
biaxial compression. A lower bound for the load carrying capacity will
in this context be taken as the larger of the two solutions. It turns out a = e−dk tanθA,1 (3)
that Solution 1 is optimal for smaller key depths, whereas Solution 2 is
optimal for larger key depths. For a further increase of the key depth, In Fig. 6(b) the parameter e is an optimization parameter related to
the load may be carried almost completely by struts spanning over two the strut width. The capacity obtained by this solution is governed by
shear keys without combination with struts over a single key, see e.g. either the compressive strength of the strut or the stress state in the key
Fig. 5(c). This is in accordance with the model of Christoffersen [7] (see corner (in the following denoted Triangle I), which can be assessed as a
also Nielsen and Hoang [8]). homogeneous stress field. The stress components for this solution will
be calculated in the following.
Fig. 6. (a) Distribution of struts spanning over a single key

(Solution 1) and (b) definition of geometrical parameters at
a key including the nodal zone, Triangle I.
119
5.1.1. Local equilibrium conditions for Solution 1 s−e

tanθB =
Fig. 7(a) depicts the nodal zone (Triangle I) of Solution 1 where b (12)
each of the three boundaries is assumed to be uniformly stressed by where s is the distance between the shear keys and e describes the same
normal and shear stresses. The stress field within the triangle is thus geometrical parameter as in Solution 1, see Fig. 8(b). As indicated in
homogeneous. Compressive normal stresses are taken as positive. The Fig. 8(b) the nodal zone in the indented area has been subdivided into
stress resultants on the boundaries act at the centroid of each boundary two triangular areas (II and III), which as shown later are stressed in
and are illustrated in Fig. 7(b) as forces related to the l−t coordinate biaxial compression. To describe the geometry of this zone, the distance
system. e2 is introduced as follows:
The homogeneous stress field (σt ,σl,τlt ) within Triangle I may be
expressed in terms of the stress resultants shown in Fig. 7(b): e2 = Lk −(e + dk tanθB ) (13)
Ct ,1 This relation ensures that the above mentioned assumption of stress

σt ,I = transfer over the entire indented length, Lk , is fulfilled. In this context,
(a−e1 ) hk (4)
it is assumed that e2 cannot attain negative values, and to fulfill this, a
a − e1
Al,1 −Cl,1 maximum effective key depth that can be utilized in the model is in-
a
σl,I = a−e troduced:
dk a 1 hk (5)
dk,ef = (Lk −e )cotθB (14)
Cl,1
τtl,I = Hence, for larger key depths, the maximum effective key depth, dk,ef , is
(a−e1 ) hk (6)
adopted in the calculations. This eventually means that by using Solu-
The relations between σA,1 and the stress resultants, Al,1 and At,1, are: tion 2, a capacity higher than that corresponding to a key depth of dk,ef
cannot be obtained. As for Solution 1, the capacity here will also be
Al,1 = σA,1cosθA,1sinθA,1 hk e (7)
governed by either the stress state in the struts or in the triangles.
At ,1 = σA,1cos2 θA,1 hk e (8)
5.2.1. Local equilibrium conditions for Solution 2
From the three equilibrium conditions for the triangular area, the fol- The equilibrium equations for Triangle I in Solution 1 also apply for
lowing relations between the stress resultants can be established: Solution 2 in which the outermost shear keys also contain a nodal zone
Al,1 −Cl,1−Pl,1 = 0 (9) of the same type, see Fig. 8(b). In addition, equilibrium equations for
the nodal zone bounding Strut A as well as Strut B, i.e. Triangles II and
At ,1 −Ct ,1−Pt ,1 = 0 (10) III, need to be developed. The stresses and stress resultants acting on the
boundaries of Triangle II and III are illustrated in Fig. 9. Triangle III
a−e1 d a borders on Struts A and B as well as Triangle II.
At ,1 −Cl,1 k −Ct ,1 = 0
2 2 2 (11) The homogeneous stress fields within Triangle II and Triangle III can
be expressed in terms of the stress resultants (by use of equilibrium
considerations). These stresses are:
5.2. Solution 2: Combination of struts spanning over one and two shear keys
Ct ,2
σt ,II =
In this solution, the struts are assumed to span in such a way so that (e + e2−e1 ) hk (15)
the entire key length, Lk , is utilized to transfer stresses (compare
e + e 2 − e1
Fig. 8(b) to Fig. 6(b)). Similar distributions can be seen in Fig. 5. The Fl e + e2
−Cl,2
inclination of Strut A may also in this case be described by Eq. (2). The σl,II = e + e 2 − e1
dk e + e hk (16)
inclination of Strut B spanning over two keys is: 2
Fig. 8. (a) Distribution of struts in joint mortar

(Solution 2) and (b) geometry at a key for ver-
ification of stress transfer.
120
Fig. 9. (a) Stresses along boundaries of Triangles II and III (Solution 2) and (b) resultants of stresses on boundaries.
Cl,2 for Solution 2. The relation reads:

τtl,II =
(e + e2−e1 ) hk (17)
nAl,i + (n−1) Bl = V (31)
At ,2
σt ,III = where n is the number of shear keys in the connection.
ehk (18)
In the transverse direction, where there is no external load, the
e
Bl−Fl e +2e stress resultants of the struts must be outbalanced by tension in the
2
σl,III = e transverse U-bar loops crossing the connection. This condition can be
dk e + e hk (19)
2 written as follows:
Al,2 nAt ,i + (n−1) Bt −(n + 1) As σs = 0 (32)
τtl,III =
ehk (20)
where As is the total cross sectional area of transverse reinforcement in
The relations between σA,2,σB , and their stress resultants are given by: one loop connection and σs is the stress in the reinforcement. Similar to
Al,2 = σA,2cosθA,2sinθA,2 hk e Eq. (31), Bt vanishes when applied to Solution 1. It should be noted that
(21)
the equation for global moment equilibrium may serve as a check of the
At ,2 = σA,2cos2 θA,2 hk e (22) calculated stress distribution in the joint.
Bl = σB cosθB sinθB hk (Lk −e ) (23)

5.4. Yield condition for reinforcement
Bt = σB cos2 θB hk (Lk −e ) (24)
The stresses carried by the U-bars must fulfill:
From the three equilibrium conditions for Triangle III, the following
relations are established: σs ⩽ f y (33)
Al,2 + Bl−Fl = 0 (25) By utilizing Eqs. (32) and (8), the yield condition (Eq. (33)), may be
At ,2 + Bt −Ft = 0 reformulated in terms of the stress carried by Strut A in Solution 1:
(26)
n+1 As f y
cos2 θB (Lk −e ) σA,1 ⩽
σA,2 = σB n cos2 θA,1 hk e (34)
cosθA,2sinθA,2 dk + cos2 θA,2 e2 (27)
In Eq. (27), the relationships in (21)–(24) have been used to es- For Solution 2, Eq. (33) may in a similar way be reformulated as a
tablish a relation between σA,2 and σB . Finally, equilibrium requirements requirement to σB :
for Triangle II lead to: (n + 1) As f y (tanθA,2 dk + e2)
σB ⩽
Fl−Cl,2−Pl,2 = 0 (28) cos2 θB hk (Lk −e )((n−1)(tanθA,2 dk + e2) + ne ) (35)
Ft −Ct ,2−Pt ,2 = 0 (29) This requirement in fact also contains the condition for the stress in
Strut A, cf. Eq. (27).
e + e2−e1 d e + e2
Ft −Cl,2 k −Ct ,2 =0
2 2 2 (30)
5.5. Failure criteria for joint mortar
5.3. Global equilibrium for Solutions 1 and 2 In the following, the failure criteria adopted in the analytical lower
bound model to describe the joint mortar will be discussed. Distinction
With reference to the test specimen depicted in Fig. 2, the global is made between zones with uniaxial compression and zones with
equilibrium conditions can now be used to establish relations between biaxial compression. The tensile strength of the mortar is neglected and
the external load, V, and the internal stress resultants defined above. In the uniaxial compression strength is fc . For Struts Ai and B carrying
the longitudinal direction, l, the applied load is balanced by the stress uniaxial compression, the stress level is limited to:
resultants Al,i and Bl stemming from Struts A and B, respectively, where
i = 1 for Solution 1 and i = 2 for Solution 2. The resultant Bl only exists σA,i ⩽ νfc , i = 1,2, ν ⩽ 1 (36)
121
σB ⩽ νfc , ν ⩽ 1 (37) |Cl,i | ⩽ μCt ,i, i = 1,2 (41)
where ν is the effectiveness factor. This factor normally takes into ac- For the inclined part of the shear key, the friction criterion, expressed in
count the material brittleness as well as the strength reduction due to terms of the resulting forces, can be established as:
cracking and tensile strains perpendicular to the struts. It is argued that
Pl,isinθk−Pt ,i cosθk
in the present lower bound problem, with steep direct strut actions ⩽ μ, i = 1,2
Pt ,isinθk + Pl,i cosθk (42)
within a narrow and long strip of mortar, the effect of tensile strains
perpendicular to the struts may be neglected. Hence, ν mainly accounts
for the material brittleness. In the following, ν will be taken as unity 5.7. Optimization of lower bound solutions
when the analytical lower bound model is evaluated against the nu-
merical calculations (because ν = 1 was used in the numerical calcu- By examining the geometrical and equilibrium conditions of
lations). However, when using the analytical lower bound model to Solution 1 as well as 2, it may be shown that the problems are in-
compare with its upper bound counterpart [1] and with test results, a determinate with two free optimization parameters, namely the para-
value of ν = 0.89 is adopted as a qualified estimate of the effect of meter e and one of the statical parameters. The resultant Cl,i will in the
material brittleness. This value is obtained from the formula, following be chosen as the statical optimization parameter.
ν = (fc,0 / fc )1/3 proposed in the fib Model Code [18], with fc,0 = 30 MPa.
If the interface is not utilized to transfer shear stresses, Cl,i will
Normally, concrete/mortar is identified as a modified Coulomb
vanish and the solution will eventually be identical to the single strut
material, which means that fc will also be the strength in a biaxial
solution suggested by Christoffersen [7]. Therefore, to obtain better
compression field. However, as shown e.g. by Kupfer et al. [19], the
solutions, the interface friction must be utilized as much as possible.
strength of concrete under biaxial compression may be larger than the
Hence, Cl,i should be taken as large as possible. According to Eq. (41),
uniaxial strength. Furthermore, as the areas with biaxial compression
this implies:
represent nodal zones in the model, the effectiveness factor can here be
taken as ν = 1, also in practice. This means that the failure criterion for Cl,i = μCt ,i, i = 1,2 (43)
zones with biaxial compression may be expressed as:
By choosing the relation given in Eq. (43), the only remaining para-
σ2 ⩽ cfc, c⩾1 (38) meter left for optimization of the load carrying capacity is e. The op-
timization is naturally subjected to the strength constraints established
where c can be interpreted as a strength enhancement factor and σ2 is in Sections 5.4 and 5.5. Results are presented in the following.
the largest principal compressive stress. Test results of Kupfer et al. [19]
showed that c depends of the σ2/ σ1−ratio (c = 1.27 for σ2/ σ1 = 2 and 6. Verification of analytical lower bound solutions
c = 1.16 for σ2/σ1 = 1). In the following, c is taken as unity when the
analytical lower bound model is evaluated against the numerical cal- The results obtained by FELA (which are optimal under the as-
culations (in order to be consistent with the assumptions made for the sumptions made) will in the following be used to verify the analytical
numerical model). However, for comparison with the upper bound lower bound solutions. The purpose is to investigate if the analytical
model and with test results, a qualified estimate of c = 1.15 is adopted. solutions actually capture the main tendencies of the numerical results.
This value is slightly higher than the recommendation of the fib Model Calculations have been carried out based on the properties and para-
Code [18] for biaxially compressed nodes. meters of the test specimens. The results, in terms of normalized shear
Due to the assumption of zero tensile strength, the following lim- stresses versus the key depth, can be seen in Fig. 10(a), where the result
itation also applies to the biaxially stressed areas: of the single strut solution [7] is also shown. The nominal shear stress
σ1 ⩾ 0 (39) has been determined as follows:
V
τ=
nhk Lk (44)
5.6. Failure criterion for interface
where V is calculated from Eq. (31). The full red curve in Fig. 10(a)
As an averaged description, the interface between mortar and pre- corresponds to the upper envelope of Solutions 1 and 2 and represents
cast elements with keys may be categorized as very rough, with a the optimal results provided by the analytical lower bound model. It
pseudo cohesive resistance and a high pseudo frictional coefficient (see can be seen that Solution 2 applies to larger key depths (dk > 8 mm)
e.g. Eurocode 2 [20]). However, in a detailed calculation where the whereas Solution 1 provides better results for smaller key depths. For
geometry of the keys is taken into account, the mechanical properties of key depths larger than 13 mm, Triangle I vanishes in Solution 1, as the
the real interface should be used. This means that the properties of the length a becomes smaller than e1. In this case σA,1 will act only on the
formwork have an influence on the failure criterion of the interface. inclined part of the key corner, however, as the capacity is less than
Hence, like for the numerical model, the following Coulomb friction Solution 2, it is not calculated. Contrary to the single strut solution, the
criterion is adopted for the interface: present model actually provides a significant capacity in the limiting
|τnt | ⩽ μσn case of dk = 0 mm (i.e. when the design is no longer a keyed connec-
(40)
tion). In this case, there is no nodal zone and the compression from the
where τnt is the shear stress on the boundary of the shear keys and σn is a diagonal struts is transferred through the interface by pure shear-fric-
compressive normal stress acting on the same boundary, e.g. tion. Fig. 10(b) depicts the optimal value of e versus the key depth. At
(σn,τnt ) = (σt ,II ,τtl,II ) . It should be noted that Eq. (40) does not contain a the transition point, the optimal value of e for Solution 1 is different
cohesion term. The main argument for neglecting the cohesion is that from that of Solution 2. This underlines that Solution 2 is not to be
smooth formwork was used to cast the reinforced concrete elements for considered as an extension of Solution 1, which is also seen in the fact
the experimental program (smooth formwork is commonly used also in that the transition between the two solutions is not smooth.
practice). Thus any small cohesive resistance in the interface may have Nevertheless, it is found that the analytical model provides results
(partly or completely) vanished at the ultimate limit state. The coeffi- that are almost identical to the numerical results obtained by FELA
cient of friction will in the following be taken as μ = 0.75, similar to the (Fig. 10(a)). A small deviation is observed around the transition be-
assumption of the numerical model. The friction criterion implies that tween Solution 1 and Solution 2. However, the deviation is insignificant
the stress resultants, Cl,i and Ct ,i , see Figs. 7(b) and 9(b), must fulfill the compared to the overall agreement. This strong correlation shows that
following condition: the analytical model is close to the optimal solution. In this context it
122
Fig. 11. Normalized stresses versus key depth in Solution 1, note that c = 1 and ν = 1.
Fig. 12. Normalized stresses versus key depth in Solution 2, c = 1.
Fig. 10. (a) Comparison of analytical and numerical lower bound models and (b) the
optimal distance e calculated for Solutions 1 and 2.
should be noted that optimality here only refers to the best results that
can be obtained by the assumptions made and not necessary the ‘true’
results. Note further that other geometries and reinforcement degrees
may favor other stress fields not captured by the two developed ana-
lytical solutions.
It can be seen in Fig. 10(a) that the analytical model as well as FELA
estimate a higher capacity than the single strut solution developed in
[7]. This is partly related to the inclusion of friction in the grout-to-
panel interface at the keyed areas.
Fig. 11 depicts the stresses of Solution 1 versus the key depth. It can
be seen that the maximum compressive principal stress, σ2,I , acting in
Triangle I is governing in the entire interval of dk , where Solution 1 is
optimal. Moreover it appears that the minor principal stress, σ1,I , is also
compressive and that the stress in Strut A,σA,1, is well below νfc for all
key depths. The magnitudes of the stresses depicted in Fig. 11 indicate a
local failure of the key corners (only Triangle I is critical). This is in
agreement with test results for specimens with small key depths. Fig. 13. Variation of σA,2 and σB as function of key depth in Solution 2, ν = 1.
The stresses of Solution 2 are plotted in Fig. 12. It can be seen that
σ1,II = 0 when the key depth is approximately 4 mm. This stress com-
key depths (at least not for the assumed material parameters and geo-
ponent in fact becomes negative (i.e. tension) for lower values of dk ,
metry considered). Furthermore, it can be seen that σ2,III is the gov-
which means that Solution 2 does not provide valid results for smaller
erning stress component for key depths between 4 mm and 10 mm,
123
Fig. 14. Collapse mechanisms considered in upper bound

model, (a) complete key shearing combined with diagonal
yield line and (b) local key corner failure, illustrations
from [1].
Fig. 15. (a) Comparison of upper and lower bound models with test results, (b) stresses in Struts A and B in Solution 2.
while σ2,II is governing for larger key depths. Triangle I vanishes at a key geometry and adopted material parameters ν = c = 1.
depth of approximately 21 mm. As σA,2 < νfc , the stress can be safely
transferred without considering a nodal zone. At a key depth of about
28 mm, the maximum effective depth according to Eq. (14) is in- 7. Comparison of analytical lower and upper bound models
troduced and the capacity cannot be increased further beyond this key
depth. In the following, the analytical lower bound model will be com-
Fig. 13 shows how the stresses in Struts A and B vary in Solution 2. pared with the previously developed upper bound model [1]. It is not
It can be seen that σB is larger than σA,2 in the entire interval of key expected that the two models lead to identical results (i.e. a theoreti-
depths. At the transition to the effective key depth (dk ≈ 28 mm), the cally exact solution) since they are not fully based on the same set of
stress in Strut B has almost reached the capacity of νfc , however, it assumptions. For instance, plane stress condition is assumed in the
remains slightly below. These results of course reflect the modeled lower bound model while plane strain condition is imposed in the upper
bound model. However, the comparison can be used to evaluate the
124
tendencies of the two models and in particular the gap between them. reliability of the results. However, the most correct solution would
Eventually, by comparison of both models with test results, a qualified probably be somewhere between the two models since in reality, the
discussion of the effectiveness factor adopted in the pure upper bound connection is not in a state of plane strain, nor plane stress, but
approach [1] can be carried out. somewhere in between. This is clearly seen in the fact that the test
The upper bound model presented in Ref. [1] is based on two basic results as plotted in Fig. 15(a) are all lying in between the results of the
collapse mechanisms reflecting local key corner failure and complete upper and lower bound models. Hence, an important conclusion that
key shearing, respectively, see Fig. 14. The interface properties between can be drawn from the comparison in Fig. 15 is that the effectiveness
mortar and the precast elements are neglected and the mortar is treated factor adopted in the pure upper bound approach (see Ref. [1] for de-
as a modified Coulomb material with zero tensile strength and with an tails) not only reflects the material brittleness but also partly contains
internal angle of friction taken as φ = 30°. The assumption of plane an empirical reduction to compensate for the ideal assumption of plane
strain condition implies that the angle of displacement, α , due to the strain. In this context, it is interesting to note that the width, i.e. b
normality condition cannot be smaller than φ , i.e. α ⩾ 30°. For a de- shown in Fig. 2, must have an influence on the stress and strain con-
tailed derivation of the upper bound solution, the reader is referred to ditions in the connection, but this parameter is absent in the failure
[1]. It should be noted that the upper bound solution, in contrast to the mechanisms based on the plane strain assumption that were considered
lower bound counterpart, can be formulated as closed-form equations in Ref. [1]. Hence, it is reasonable to believe that the effectiveness
and is therefore easier to use in practice. factor of the pure upper bound approach also compensates for the in-
The results of the upper bound and lower bound models can be seen fluence of b, which is not considered theoretically. The lower bound
in Fig. 15(a). As stated in the previous, ν = 0.89 is adopted for the model, on the other hand, includes the width, b, as an important
uniaxial compression struts while ν = 1 and c = 1.15 are used for the parameter. Still, if the plane stress lower bound solution should be
biaxial compression zones in the lower bound model. Such a distinction calibrated to fit the test results (Fig. 15), then artificially higher values
between types of stress field can of course not be made in the upper of ν and c would be required, most probably to compensate for the
bound model. Therefore, ν = 0.89 has been used when determining the triaxial stress state that in reality would develop locally in the joint
dissipation in all the yield lines in the upper bound model. The gap mortar.
between the upper and the lower bound solutions (Fig. 15) is however It might be too optimistic to expect that a more correct/realistic
not due to the difference in the material parameters. The gap would in solution can be developed which at the same time is as user-friendly as
fact have been larger, had the lower bound calculations also been based the closed-form upper bound solution. The upper bound model will
on ν = 0.89 and c = 1 everywhere. It is mainly the assumption of plane therefore still have preference from a practical point of view, even
strain condition versus plane stress condition that has an influence on though it has to be used in conjunction with an effectiveness factor that
the observed gap. In addition, other more sophisticated collapse me- not only accounts for the real material behavior but also functions as an
chanisms (not yet analyzed) may lower the upper bound results. adjustment parameter to compensate for the unsafe nature of upper
Another discrepancy between the two models should also be bound solutions. Awareness of this is important when applying the
pointed out. A closer look at the two solutions (in Fig. 15) in the vicinity upper bound model to practical cases, which deviate significantly from
of dk = 0 mm reveals that the upper bound solution is lying below the the experimentally tested configurations. In such situations, the lower
lower bound solution. This difference has no practical significant but is bound model presented in this paper will be useful.
apparently inconsistent with the extremum theorems of plasticity. The
reason is to be found in the fact, that the interface is assigned a friction 8. Conclusions
failure criterion, Eq. (40), in the lower bound model while this criterion
is ignored in the upper bound calculations. To obtain results consistent This paper presented push-off tests of a recently developed keyed
with the extremum theorems, the upper bound calculations should be shear connection design for precast concrete wall elements. The in-
modified to include the friction criterion for the interface and at the vestigation focused on the influence of the key depth on the failure
same time fulfill the normality condition at the interface. It may in that mode and the load–displacement response of the connections. The ex-
case be shown, that the upper bound solution will lie above the lower perimental observations include:
bound counterpart and that the two solutions will yield exactly the
same result at the limiting case of dk = 0 mm. • Connections with smaller key depths failed by local key corner
In the present calculations, the upper bound model reaches the shearing while failure by complete key shearing was observed for
upper limit (corresponding to complete key shearing) at dk ≈ 13 mm. connections with larger key depths
This is actually close to dk = 16 mm, which in the test series corre- • The maximum capacity (first peak load) was related to failure of the
sponds to the transition between the two failure modes. The lower keys and increased with increasing key depth until complete key
bound Solution 2 on the other hand, reaches an upper limit at shearing became the critical failure mode
dk ≈ 28 mm corresponding to the maximum effective key depth (Eq. • The residual load level (beyond the first peak load) is relatively
(14)). Furthermore, in Solution 2, the Triangle II is critical when dk is higher for connections suffering local key corner shearing
between approximately 9 to 17 mm while Strut B is stressed to νfc and
therefore critical for dk larger than 17 mm (see Fig. 15(b)). For The experimental observations suggest that for practical applica-
dk > 17 mm, the thickness of Strut B increases with increasing dk which tions the connection should be designed to be governed by local key
explains the increase of the capacity for dk between 17 and 28 mm. As corner crushing as this improves the ductility. In addition to the ex-
seen in Fig. 15(b), Strut A of Solution 2 is at no point critical. The fact perimental results, an analytical lower bound model was developed for
that only Strut B is critical when Solution 2 reaches the upper limit prediction of the first peak capacity. The model was validated with
makes it rather difficult to relate the result to a failure mechanism in- numerical calculations based on finite element limit analysis. The
volving complete key shearing. This simply underlines that Solution 2 model differs from existing analytical lower bound solutions for keyed
(although it may be optimal subjected to the assumptions made) is still connections in the following way:
a lower bound and thus a safe model for the real ultimate behavior of
the connection. • The load can be carried by a combination of struts spanning over
It is important for practical application that the two models (de- one or two indentations
veloped independently of each other) in fact display the same overall • Local strength increase at nodal zones and effects of interface fric-
tendencies, namely a shear capacity that increases with increasing key tion are accounted for
depth until an upper limit has been reached. This improves the • The nodal zones can be subjected to non-hydrostatic biaxial stress
125
conditions although it is more complicated to use. In conclusion, it can be stated

that the paper has contributed to a more complete theoretical basis for
The lower bound model shows similar tendencies as a previously practical design of keyed connections.
developed upper bound model and the test results fall within the gap
between the two solutions. From the comparison, it can be concluded Acknowledgment
that the lower bound solution is a safe one due to the assumption of
plane stress condition while the plane strain assumption adopted in the The experimental program was financially supported by the Danish
upper bound solution is theoretically unsafe and must be compensated Association for Precast Concrete Elements and the COWI Foundation.
for by adjustment with test results. This is reflected in the relatively low The test elements were produced at CRH Concrete and the experimental
effectiveness factor that has to be applied. The upper bound solution work was conducted with support from M.Sc. R.M. Hou and M.Sc. J.S.
should therefore only be used for configurations within the range Olsen. The authors gratefully acknowledge these valuable contribu-
covered by tests. In other cases, the lower bound model is applicable tions.
Appendix A. Lower bound finite element limit analysis (FELA)
As discussed in Section 4, FELA can be considered as a special case of the general finite element method, where a rigid-plastic material model is
assumed. Unlike the finite element method, however, FELA is a so-called direct method where the collapse load is determined in a single step and no
incremental procedure is needed. The method is therefore rather efficient and numerically stable for calculation of the collapse load of structures.
The fundamental idea of the method is to formulate a given limit analysis problem as a convex optimization problem, where the ultimate load is
maximized. In order to obtain a lower bound solution, the determined stress field must be statically admissible and safe. This is ensured by a set of
linear equality constraints, representing the equilibrium conditions, as well as a set of convex inequality constraints, representing the convex yield
conditions. The general form of the optimization problem is given below:
maximize λ
subject to BT σ = pλ + p0
f (σi )≤ 0, i = 1,2,…,m
The external load comprises a constant part, p0 , and a scalable part, pλ , where λ is the load factor, which is sought to be maximized. The external
load is balanced by the stress field described by the vector σ via the linear equilibrium equations, BT σ = pλ + p0 , where BT is the equilibrium matrix.
The yield function f (σi ) is checked in the m points to ensure a safe stress field. For concrete, the Mohr-Coulomb yield criterion is commonly used
which can be expressed as conic constraints [21,22].
The equilibrium matrix, BT , depends on the chosen lower bound finite elements and discretization. In this paper, lower bound plane stress
elements have been used together with bar and interface elements, representing the reinforcement and grout-to-panel interfaces of the keyed joint.
References [12] Pistoljevic N, Nielsen MP. HSTO-metoden. Plasticitetsteori for armerede be-
tonskiver. Anvendelse af homogene spændingstilstande i trekantformede områder.
Proc Danish Soc Struct Sci Eng 2009;80(2):19–54.
[1] Sørensen JH, Hoang LC, Olesen JF, Fischer G. Test and analysis of a new ductile [13] Calladine CR. Simple ideas in the large-deflection plastic theory of plates and slabs.
shear connection design for RC shear walls. Struct Concr 2017;18(1):189–204. In: Heyman J, Leckie FA, editors. International conference on the applications of
[2] fib. Bulletin 43: structural connections for precast concrete buildings. Federation plastic theory in engineering design, Cambridge, UK; 1968. p. 93–127.
international du béton, Lausanne, Switzerland; 2008. [14] Bræstrup MW. Dome effect in RC slabs: rigid-plastic analysis. J Struct Div – ASCE
[3] Hansen K, Kavyrchine M, Melhorn G, Olesen SØ, Pume D, Schwing H. Keyed shear 1980;106(6):1237–53.
joints – SBI rapport 97. Tech rep. Danish Building Research Institute; 1976. [15] Belenkiy LM. Upper-bound solutions for rigid-plastic beams and plates of large
[4] Sørensen JH, Hoang LC, Olesen JF, Fischer G. Tensile capacity of loop connections deflections by variation principles. J Eng Mech 2007;133(1):98–105.
grouted with concrete or mortar. Mag Concr Res 2017;69(17):892–904. [16] Sørensen JH, Hoang LC, Olesen JF, Fischer G. Testing and modeling dowel and
[5] Jensen BC. Nogle Plasticitetsteoretiske Beregninger af Beton og Jernbeton (English: catenary action in rebars crossing shear joints in RC. Eng Struct 2017;145:234–45.
Some applications of plastic analysis to plain and reinforced concrete). Report 111. [17] Dahl KKB. Construction joints in normal and high strength concrete. Tech rep.
Ph.D. thesis. Lyngby: Technical University of Denmark, Institute of Building Design; Lyngby: Technical University of Denmark, Department of Structural Engineering;
1976. 1994.
[6] Jørgensen HB, Hoang LC. Tests and limit analysis of loop connections between [18] fib. Model code for concrete structures 2010. Lausanne (Switzerland): Wilhelm
precast concrete elements loaded in tension. Eng Struct 2013;52:558–69. Ernst & Sohn; 2013.
[7] Christoffersen J. Ultimate capacity of joints in precast large panel concrete build- [19] Kupfer H, Hilsdorf HK, Rüsch H. Behavior of concrete under biaxial stresses. ACI
ings, series R No 25. Ph.D. thesis. Lyngby: Technical University of Denmark, Struct J 1969;66(8):656–66.
Department of Structural Engineering and Materials; 1997. [20] CEN. EN1992-1-1 Eurocode 2: design of concrete structures Part 1-1: General rules
[8] Nielsen MP, Hoang LC. Limit analysis and concrete plasticity. 3rd ed. Boca Raton: and rules for buildings. 3rd ed. European Committee for Standardization, Brussels;
CRC Press, Taylor & Francis Group; 2011. 2004.
[9] Herfelt MA, Poulsen PN, Hoang LC, Jensen JF. Numerical limit analysis of keyed [21] Krabbenhøft K, Lyamin AV, Sloan SW. Formulation and solution of some plasticity
shear joints in concrete structures. Struct Concr 2016;17(3):481–90. problems as conic programs. Int J Solids Struct 2007;44(5):1533–49.
[10] Nielsen MP. Om jernbetonskivers styrke (English: On the strength of reinforced [22] Bisbos CD, Pardalos PM. Second-order cone and semidefinite representations of
concrete discs). Polyteknisk Forlag; 1969. material failure criteria. J Optimiz Theory Appl 2007;134(2):275–301.
[11] Muttoni A, Schwartz Joseph, Thürlimann Bruno. Design of concrete structures with
stress fields. Basel (Switzerland): Birkhäuser Verlag; 1997.
126
Paper IV
”Testing and modeling dowel and catenary action in rebars
crossing shear joints in RC”
Published in: Engineering Structures (2017), 145:234-245
253
Engineering Structures 145 (2017) 234–245
Contents lists available at ScienceDirect
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
Testing and modeling dowel and catenary action in rebars crossing shear
joints in RC
Jesper H. Sørensen ⇑, Linh C. Hoang, John F. Olesen, Gregor Fischer
Technical University of Denmark, Department of Civil Engineering, Brovej, Bygning 118, 2800 Kgs. Lyngby, Denmark
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents a detailed study of the shear behavior of two-sided dowel joints, which includes ini-
Received 9 February 2017 tiation of dowel action at small shear displacements and development of full catenary action in the rein-
Revised 21 April 2017 forcement at large displacements. In addition to experimental results, the paper also presents a simple,
Accepted 10 May 2017
second order plasticity model to describe the non-linear regime of the load-displacement relationship. In
the model, kinematic relations and the normality condition of plastic theory are utilized to establish a
unique link between the imposed shear displacement and combinations of moment and tension that
Keywords:
develop in the rebar(s) crossing the joint. Interface friction is included in a consistent manner based
Dowel action
Catenary action
on clamping stresses induced by the tension of the rebar(s). Comparison of experimental results with
Concrete plasticity the model predictions shows satisfactory agreement. The model has, due to its simplicity, potential for
Second order modeling practical applications related to assessment of structural robustness, where estimation of the available
energy (area below load-displacement curve) is important.
Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction Due to the non-linear material behavior, a model based on the

theory of plasticity seems more appropriate to describe dowel
Dowel action in reinforcing bars is a well-known phenomenon action at the ultimate limit state. Such a model was first developed
which can be utilized as load carrying mechanism in structural for one-sided dowel action by Rasmussen [13], who assumed
concrete. Inclusion of dowel action in design is primarily relevant crushing of concrete simultaneously with yielding of the reinforce-
for problems that involve transfer of shear through casting joints ment. In contrast to the basic problem studied by Rasmussen, Fig. 1
and connections, see e.g. fib Bulletin 43 [1]. schematically illustrates a so-called two-sided dowel joint, where
Pure dowel action is experimentally most clearly observed the rebar is fully embedded in concrete on both sides of the casting
when a rebar has part of its length embedded in a large block of joint. In this case, pure dowel action (represented by rotation, h, in
concrete while being loaded by a transverse force at the concrete the plastic hinges) is only the first phase of the load transfer mech-
surface. In most practical cases, the load carrying capacity will be anism of the joint. When the shear displacement in the joint, u,
governed be development of a plastic hinge in the rebar and by increases, the dowel action will gradually be accompanied by axial
local crushing of the concrete. The first attempt to study this basic tension (represented by elongation, D, in the plastic hinges). The
problem (which in the following will be termed ’one-sided dowel load transfer mechanism is in this phase a combination of the
action’) was carried out by Friberg [2] who modeled the dowel as two actions. At the final stage, dowel action may be completely
a linear elastic beam transversely supported on elastic springs. replaced by axial tension, i.e. the load may be carried by pure cate-
Many researchers have since adopted this approach to model the nary action in the rebar. Development of full catenary action
load-displacement response of the dowel [3–10]. However, the requires relatively large shear displacements. For this reason, this
obvious shortcoming of this approach is the fact that it is a linear effect is seldom utilized in the ultimate design of structural joints
elastic model. Some attempts have therefore been made to adjust even though the load related to pure catenary action is usually
the transverse spring stiffness [11,12] and others again have higher than that corresponding to pure dowel action. However,
empirically suggested a gradual change of stiffness to fit the non- for assessment of structural robustness as well as verification of
linear test results [4–6]. structural performance under accidental load cases, catenary
action may play an important role, not only for the behavior of
shear joints but also when considering the behavior of slabs after
⇑ Corresponding author. initial bending and/or punching failure (Refs. [14–20]).
E-mail address: [email protected] (J.H. Sørensen).
http://dx.doi.org/10.1016/j.engstruct.2017.05.020
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
J.H. Sørensen et al. / Engineering Structures 145 (2017) 234–245 235
Fig. 1. Casting joint between two different concretes subjected to shear displacements and crossed by a rebar.
Several authors have recognized that for two-sided joints, the the kinematical conditions of the dowel. As will be shown, the link
development of axial tension reduces the bending capacity of eventually also allows for inclusion of interface friction in a consis-
the dowel. In order to account for this reduction, a criterion for tent manner.
the combination of axial tension and bending is needed [21–25].
Basically, this criterion is the same as the MN-interaction diagram
for the rebar cross section. The real challenge here is to establish a 2. Experimental program
unique link between the shear displacement, u, and the points on
the MN-interaction diagram. In the literature, this link is often 2.1. Material properties and test setup
established on a more or less empirical basis, e.g. based on exper-
imental measurements of the axial tension in the rebar. The experimental program was designed to examine the devel-
This paper presents an investigation into the behavior of two- opment of dowel and catenary action in rebars crossing an inter-
sided dowel joints (i.e. rebars crossing an interface between face loaded in pure shear. For this purpose, push-off tests were
concretes cast at different times), exposed to large shear displace- carried out on specimens with one single rebar, two rebars and
ments. The investigation covers both an experimental program as four rebars (i.e. n = 1, 2 or 4) crossing the casting joint. In most
well as a theoretical study. Some few initial tests as well as basic cases, three replicates of the same layout were successfully tested.
ideas have been presented in a previous work-in-progress paper The general specimen layout is shown in Fig. 2. The rebars were
(Ref. [26]). One of the focus points of the experimental program placed symmetrically about the principal axis of the specimen
has been to study the efficiency of rebar groups in comparison with cross section. In the case of n = 2 and n = 4, the rebars were placed
single dowel behavior. This is relevant for practical applications with a mutual distance of 42 mm. The shear load was applied in
because rebars crossing casting joints between precast concrete the direction of the z-axis shown in Fig. 2. Specifications and mate-
elements are often lumped in groups, e.g. in the form of overlap- rial properties have been summarized in Table 1. As indicated in
ping U-bars. Furthermore, in addition to tests of the classical Fig. 2, confinement reinforcement in the form of rectangular stir-
concrete-to-concrete interface, the program also includes rups was used to prevent premature splitting failure of the con-
concrete-to-mortar interfaces. Tests of such combinations have, crete block. The rebars crossing the casting joint had threads at
to the best knowledge of the authors, not been published before. both ends for installation of anchorage plates (to ensure develop-
The combinations are nevertheless important in practice, for ment of full tensile capacity when catenary action developed).
instance when dealing with shear connections between precast The specimens were cast in two sequences. At first, half of each
concrete wall elements, which are often grouted with mortar. specimen was cast in plywood formwork with smooth surfaces.
The experimental results have been used to calibrate and verify The second half was cast the day after. Before the second cast,
a theoretical model of the load-displacement response of two- grease was applied to the casting joint in order to reduce friction.
sided dowel joints. The primary motivation for developing such a A regular concrete (denoted C in Table 1) with a maximum
model is to provide a more accurate calculation of the displace- aggregate size of 16 mm and a commercial mortar (denoted M in
ment dependent resistance of shear joints under e.g. accidental Table 1) containing aggregate sizes of 0–2 mm were used to obtain
load cases, as this information may facilitate an estimate of the three combinations of casting joints. The combinations were
overall robustness for the structural system. The model is based (see also Table 1): concrete-to-concrete (C/C), mortar-to-mortar
on a second order rigid-plastic approach, where change of struc- (M/M), and mortar-to-concrete (M/C). The main differences
tural geometry has to be considered in order to correctly model between the two materials are the aggregate composition and
the response at large shear displacements. The adopted approach the compressive strength, see Table 2 for the proportions of the
differs from most of the previous works on this topic and has the concrete mixture. All specimens were provided with rebars with
advantage of being able to provide a simple and unique link diameter d = 8 mm.
between the relative displacement in the joint and the MN- The specimens were tested in a classical push-off setup, where
interaction diagram for the rebar cross section. The link is estab- the thrust line of the applied load coincides with the plane of the
lished by combining the normality condition of plastic theory with casting joint in order to simulate pure shear loading, see Fig. 3.
236 J.H. Sørensen et al. / Engineering Structures 145 (2017) 234–245
Fig. 2. General geometry of test specimens for push-off shear tests and indication of position of rebars in specimens with n = 1, 2 and 4 rebars.
Table 1
Specifications and material properties for test series (C = concrete, M = mortar, n = number of rebars).
n Interface combination d [mm] f y [Mpa] f u [MPa] f c1 [MPa] f c2 [MPa] No. of rep.
1 C/C 8 614 729 48.0 48.0 3

1 M/M 8 614 729 33.1 33.1 3
1 M/C 8 614 729 33.3 47.7 3
2 C/C 8 614 729 46.5 46.5 3
2 M/M 8 614 729 34.1 34.1 2
2 M/C 8 614 729 30.3 35.3 3
4 C/C 8 551 708 43.0 43.0 3
4 M/M 8 614 729 34.5 34.5 2
4 M/C 8 551 708 37.2 45.4 3
Table 2 2.2. Test results

Proportions of the concrete mixture.
[kg/m3] Fig. 4 shows the measured load-displacement relationships for

Cement 321.4
all tests. Within each test series (i.e. interface combinations C/C,
Water 156.1 M/M, and M/C) the load level for any given value of displacement,
Superplasticizer 2.0 u, is roughly speaking proportional to the number of rebars cross-
Aggr. 0–4 mm 848.6 ing the joint. This indicates that the mutual distance (42 mm)
Aggr. 4–8 mm 330.0
between the rebars was sufficient to avoid group action. It was
Aggr. 8–16 mm 753.3
generally observed, that the specimens behaved almost linear elas-
w/c 0.49
tic in the beginning. Then, gradually, the response curves became
non-linear with loss of stiffness. By imposing further shear dis-
placement, the load-displacement curves developed in a concave
manner which led to increasing tangent stiffness. At a certain
The tests were performed in quasi-static deformation control with
point, the tangent stiffness decreased again due to a convex devel-
a constant rate of piston movement of 2.5 mm/min. Relative shear
opment of the response curves. The convex development initiated
displacements in the joint were measured on both sides of the
partly as a result of the stress–strain relation of the reinforcement
specimens using linear variable differential transducers.
Fig. 3. Test setup for push-off testing of casting joints crossed by rebar(s).
Fig. 4. Test results from all specimens.
which at this point must have experienced hardening after initial difference between the compressive strengths of concrete and
yielding and hence the tangent stiffness decreases with increased mortar was found for these tests, cf. Table 1. It should be noted,
shear displacement. This continues until the ultimate load, which however, that although the response curves of the concrete speci-
was found at a maximum shear displacement in the range of 15– mens lie above those of the mortar specimens, then, at the final
25 mm. All tests were terminated by rupture of the rebars, and stage, specimens with mortar-to-mortar interface were actually
for the specimens with the same number of rebars, larger maxi- able to carry the highest ultimate load. The results are interesting
mum shear displacement also resulted in higher ultimate loads. and indicate that while the compressive strength has a positive
From post-test examinations of the specimens, severe local crush- influence on the transition to non-linear behavior, then in the
ing of the concrete/mortar was observed near the interface as end, a higher compressive strength leads to lower ultimate load.
shown in Fig. 5(a). Fig. 5(b) shows a photo of the deflected shape The reason for this will be discussed in Section 4.
of a ruptured rebar removed from a test specimen after testing. On an overall level, the tested specimens behaved similar to
The plastic deformations of the rebar were concentrated within a tests reported by Engström [24] and Randl and Wicke [25]. How-
length that corresponds to a few rebar diameters on both sides ever, Engström tested bolts with larger diameters which induced
of the interface. Outside this region, the rebars remained straight, greater action in the concrete block and for this reason premature
see Fig. 5(b). splitting failure of the concrete specimens was observed before
When comparing the response of specimens containing the rupture of the reinforcement/bolts occurred. Randl and Wicke
same number of rebars, it is observed that specimens with C/C reported similar load-displacement curves from tests on T-
interface carried the highest load at the point where the non- headed bars embedded in concrete. They observed shear displace-
linear part of the response clearly emerges. This point is lowest ments up to 20 mm and failure by rupture of the reinforcement. In
for the curves belonging to specimens with M/M interface. As their study, the interface properties were varied which influenced
can be seen, this tendency is most pronounced for specimens the shape of the load-displacement curves, especially at the transi-
with four rebars (n = 4) which seems reasonable since the largest tion to non-linear behavior.
Fig. 5. Local crushing at rebars at the casting joint (a) and deflected shape of ruptured rebar (b).
3. Second order plastic modeling Finally, by use of Eqs. (4) and (5), the following condensed
expression for the kinematical condition of the rebar can be
The non-linear part of the load-displacement response of two- established:
sided dowel joints will in this paper be modeled by use of a simple sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
second order plasticity approach where concrete as well as rein- D_ u u
¼ 1þ ð6Þ
forcing steel are treated as rigid-plastic materials with finite defor- h_ 2 l1 þ l2
mation capacity. In the model, displacements are therefore the sole
results of accumulated plastic deformations. The model will be
3.2. Constitutive relationship and sectional forces in the plastic hinges
established for the general case, where the casting joint is the
interface between two different concretes (having different com-
The assumption of rigid-plastic material behavior implies that
pressive strengths). Initially, the model is established for perfectly
in the case of pure tension, plastic deformation in the rebar is only
smooth joints. The effect of friction in the joint interface is then
possible when the cross section is subjected to the plastic tensile
included by an extension of the model.
capacity N p . Further, in the case of pure bending, plastic deforma-
tion is only possible when the cross section is subjected to the plas-
3.1. Mechanism and kinematic relationships tic moment capacity M p . These sectional capacities are:
p 2
The starting point of the model is to assume a failure mecha- Np ¼ d fy ð7Þ
nism for the rebar, see Fig. 1, where two plastic hinges must 4
1 3
develop to allow for relative displacement, u, in the joint. The posi- Mp ¼ d f y ð8Þ
tion of the plastic hinges (defined by the distances l1 and l2 from 6
the joint) depends on the strength of the concretes and the where d is the cross sectional diameter of the rebar and f y is the
moment capacity of the rebar cross section. In order to comply yield stress of the rebar. For combinations of bending and axial ten-
with compatibility requirements, the plastic hinges must in addi- sion, plastic deformations may initiate when the yield condition of
tion to rotations also undergo elongations when u increases. The the cross section is fulfilled, i.e. when f ðM; N Þ ¼ 0. The yield condi-
rates of plastic deformations can be determined by establishing tion (or the MN-interaction diagram) can be derived by requiring
the kinematic relationship for the assumed mechanism. Then, by static equivalence between the sectional forces ðM; NÞ and the dis-
imposing the normality condition of plastic theory and by applying tribution of normal stresses shown in Fig. 6. The result appears as
the work equation for increments of displacement, the necessary follows:
equations to determine the load P as a function of displacement, 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
13 ! 13 23
u, can be derived. N 2@ M M M A
f ðN; MÞ ¼ þ arcsin 1 1¼0
The problem is treated as a static displacement controlled prob- Np p Mp Mp Mp
lem, where the relative shear displacement in the joint, u, is con-
sidered as a monotonic function of time. For convenience, a ð9Þ
displacement velocity equal to unity is assumed: As an approximation, the mathematically simpler yield condi-
uðtÞ ¼ t ð1Þ tion for a rectangular cross section is sometimes adopted when
studying dowel and catenary action in rebars, see e.g. [21–23,25].
From simple geometrical considerations, the following relation- Now, according to the normality condition of plastic theory, the
ship can be established between u and the angle of rotation, h, in rates of plastic deformations must fulfill the following constitutive
the plastic hinges: relationship:
u u1 u2 @f 8 1
tan h ¼ ¼ ¼ ð2Þ h_ ¼k ¼ k 3 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2=3ffi ð10Þ
l1 þ l2 l1 l2 @M pd f y
1 MMp
where u1 þ u2 ¼ u, see Fig. 1. To accommodate the change of geom-
@f 4
etry when u increases, it is necessary to impose elongation in the D_ ¼k ¼k 2 ð11Þ
rebar in addition to rotation of the plastic hinges. Since rigid- @N pd f y
plastic material behavior has been assumed, it is convenient to con-
sider the elongation as a plastic extension, D, concentrated in the where k is a positive constant proportional to the displacement
hinges (as indicated in Fig. 1). In this way, h and D may be regarded velocity. The constant vanishes when Eqs. (10) and (11) are used
as the general strains in the plastic hinges, which are subjected to to establish the following ratio of plastic strain rates:
general stresses in the form of bending moments, M, and normal
forces, N. The following relationship between u and D can be
established:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
l1 þ l2 l1 þ l2 u
D¼ þ 1þ ð3Þ
2 2 l1 þ l2
Based on Eqs. (1)–(3), the rates of plastic deformations in the

hinges can be determined as follows:
dh dh du l1 þ l2
h_ ¼ ¼ ¼ ð4Þ
dt du dt ðl1 þ l2 Þ2 þ u2
dD dD du u
D_ ¼ ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5Þ
dt du dt 2
2 ðl1 þ l2 Þ þ u2 Fig. 6. Plastic stress distribution in a circular cross section subjected to M and N.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2=3
D_ d M and (17) will, as expected, be identical to the results presented in
¼ 1 ð12Þ Nielsen and Hoang [28]. It should be noted that the assumed uni-
h_ 2 Mp
form distribution of contact pressures only leads to zero moment
It can be seen that Eq. (12) together with Eq. (6) provides a link in the rebar at the interface when f cc;1 ¼ f cc;2 (see moment diagram
between the kinematical conditions and the state of stresses in the in Fig. 7(a)). Hence, to maintain equilibrium when f cc;1 – f cc;2 and
plastic hinges. This means that the bending moment, MðuÞ, carried when the external action corresponds to pure shear (i.e. thrust line
by the plastic hinges for any given value of displacement, u, can be coinciding with interface plane), tension must develop in the rebar
determined by equating the right hand side of Eq. (6) to the right which eventually leads to a distribution of so-called clamping
hand side of Eq. (12). The tension force, N ðuÞ, may thereafter be stresses in the interface. The tension force together with the
determined by imposing f ðM; N Þ ¼ 0 according to Eq. (9). The clamping stresses will then be able to outbalance the (small) bend-
results are: ing moment in the rebar at the interface cross section. According to
u2 2 !!3=2 Eq. (14), tension develops in the rebar as soon as u > 0. Therefore,
M ðuÞ u strictly speaking, the assumption of uniformly distributed contact
¼ 1 1þ ¥0 ð13Þ
Mp d l1 þ l2 pressure is not able to fulfill all equilibrium requirements at
0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 1
1 ! 1 2 u ¼ 0 (i.e. initiation of pure dowel action) when f cc;1 – f cc;2 . This
NðuÞ 2@ MðuÞ 3 M ðuÞ 3 MðuÞ 3 A is, however, acceptable since the assumption primarily was moti-
¼1 arcsin 1 1
Np p Mp Mp Mp vated by the aim of obtaining a simple estimate of the position
of the plastic hinges. In reality, the rebar will probably experience
ð14Þ
combinations of tension and bending moments already in the elas-
The ratio MðuÞ=Mp in Eq. (14) may be replaced by the right hand tic range.
side of Eq. (13) in order to obtain an explicit expression for the nor- The simple stress distribution shown in Fig. 7(a) cannot be
mal force in the rebar as a function of u. adopted for analysis of the entire load-displacement response.
When dowel action is accompanied by tension in the rebar (and
3.3. Effective stress distribution in the concrete in the end completely replaced by catenary action) the average
contact pressure must decrease due to the assumed material prop-
The assumed displacement field for the rebar implies that it has erties and the equilibrium conditions. Since concrete is not a per-
to cut its way through the concrete and thereby cause local crush- fectly rigid-plastic material, there will be a softening effect which
ing failure. It is in this context not possible on the basis of the pre- in turn reduces the concrete pressure when the concrete experi-
sent simplified approach to determine in details the entire stress ences too large compressive strains. The displacement, u, and
distribution in the concrete. For triaxial stress conditions of the thereby the local deformations may become so large that the con-
type developed in the concrete at the dowel, an enhanced com- crete near the joint interface spalls off/crushes thus leaving this
pressive strength, f cc , is therefore usually assumed: zone to be stress free, see e.g. Fig. 5(a). Therefore, due to material
properties, redistribution of the contact pressure will take place
f cc ¼ cfc ð15Þ as u increases. In addition to this, the redistribution of stresses
where c P 1 is the so-called enhancement factor which has to be must take place in such a way, that equilibrium can be maintained
determined by calibration with test results. Rasmussen [13] found when catenary action starts to develop in the rebar.
c-values in the range of 3.7–5.4 from tests on one-sided dowels. It is not possible in a rigid-plastic model to theoretically
Similar c-values have also been suggested in Refs. [1,23,27]. There account for the above mentioned softening of the concrete. There-
is a close link between the average triaxial compressive strength, fore, in the following, as u increases, the effect of softening (and
f cc , and the position of the plastic hinges. Rasmussen [13] used a spalling of concrete) will indirectly be taken into account by intro-
simple plasticity approach to establish this link, which eventually ducing effective lengths, l1;ef and l2;ef , over which contact pressures
led to an estimate of the load carrying capacity related to pure f cc;1 and f cc;2 are assumed to act uniformly, see Fig. 7(b). This may
dowel action (i.e. the so-called first order plastic solution). The same be interpreted as an assumption of rigid-plastic behavior with
approach is adopted in the following for two-sided dowels in order finite deformation capacity, although there is actually no real infor-
to determine the distances l1 and l2 . As shown in Fig. 7(a), it is mation about the deformation capacity of the concrete when it is
assumed that contact pressures of magnitude f cc;1 and f cc;2 are act- subjected to contact pressure by the dowel. The only simple way
ing uniformly on the rebar over the lengths l1 and l2 , respectively, to establish the condition for l1;ef and l2;ef , as u increases, is there-
when the rebar starts to carry load by pure dowel action (see also fore through equilibrium considerations. Hence, by establishing
Nielsen and Hoang [28]). The corresponding shear and moment dia- vertical force equilibrium and moment equilibrium for the part
grams for the rebar are shown in Fig. 7(a) as well. By setting up the of the rebar between the plastic hinges, see Fig. 7(b), and by utiliz-
vertical force equilibrium and moment equilibrium for the part of ing that the plastic hinges now are subjected to MðuÞ and N ðuÞ as
the rebar between the two plastic hinges, it is possible to establish given by Eqs. (13) and (14), the following relationship between u
the following equations to calculate l1 and l2 : and the effective lengths l1;ef and l2;ef can be established:
sffiffiffiffiffiffiffiffi 0 1
rffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fy u
2 d B u M ð u Þ C
l1 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð16Þ li;ef ¼ li @1 t1 2 f f A ¥ 0 ð18Þ
3 1 þ f cc;1 f cc;1 M p þ u2 d f cc;1þfcc;2
f cc;2 cc;1 cc;2
rffiffiffi sffiffiffiffiffiffiffiffi
2 d fy where i = 1 and 2. It appears that the effective length is equal to the
l2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð17Þ
3 1 þ f cc;2 f cc;2 initial lengths, cf. Eqs. (16) and (17), when u ¼ 0 and reduces to zero
f cc;1
when u has reached a value that makes MðuÞ ¼ 0. The latter situa-
When setting up the equilibrium equations leading to Eqs. (16) tion corresponds to a transition to full catenary action where the
and (17), it has been utilized that the moment capacity, Mp , of the plastic hinges turn into moment-free hinges. Therefore, the rebar
rebar cross section is given by Eq. (8). In the case of identical mate- will no longer experience contact pressure between the two hinges,
rial properties on both sides of the joint, i.e. f cc;1 ¼ f cc;2 , Eqs. (16) but instead acts as a tie.
Fig. 7. Position of plastic hinges and extent of distribution of contact pressure at initiation of pure dowel action (a) and at combined dowel action and catenary action (b).
3.4. Deformation capacity of rebars a distance of two times the rebar diameter, 2d, within which strain
localization took place. Based on the results in Fig. 8(a), an elonga-
As described in the previous, all push-off tests were terminated tion capacity of Dmax = 3 mm is adopted. It should in this context be
when rupture of the rebars took place, i.e. when the deformation noted that the standard methods for determination of nominal
capacity of the rebar was reached. To capture this effect in the model, strain capacity of reinforcing steel cannot be used to estimate
it is necessary to express the deformation capacity of the rebar in Dmax . This is so because the nominal strain capacity is based on a
terms of the plastic elongation in the hinges. This means that an reference length, which is much longer than the necking zone as
upper limit, Dmax , must be introduced such that Eq. (3) may be used well as the characteristic length of the present problem, i.e. the dis-
to determine the displacement capacity, umax , of the system. tance between the two plastic hinges.
An estimate of Dmax can be obtained from a detailed study of the
tension tests of the reinforcement. Fig. 8(a) shows four tested 3.5. Load-displacement response of frictionless joints
stress-elongation relationships for the type of rebar used in this
study. The measured elongations represent the strain accumula- Based on the obtained results and the assumptions made, it is
tions over the so-called necking zone. The results were obtained now possible to determine the load-displacement response by
by using digital image correlation (DIC) analysis based on images use of the work equation. For this purpose, a stationary situation
taken at approximately 0.25 hertz with a 36 megapixel camera. with displacement u and load P ðuÞ is used as the starting point
An example of results of a DIC analysis, showing strain localization from which an increment of displacement, du, is considered. The
in the rebar just before rupture, can be seen in Fig. 8(b). The curves external work, W E , is then given by:
in Fig. 8(a) were determined by post-processing of the digital strain
measurements, where it was possible to isolate the elongation over W E ¼ PðuÞdu ð19Þ
Fig. 8. Local stress-elongation relationship for rebars and indication of the measurement length spanning the zone where necking occurs.
The internal work, W I , has contributions from the energy dissi-

pated in the plastic hinges as well as the energy absorbed when the
concrete crushes at the rebar. The following formula can be
derived:

1 l1;ef 1 l2;ef
W I ¼ f cc;1 l1;ef d du1 þ f cc;2 l2;ef d du2
2 l1 2 l2
þ 2N ðuÞdD þ 2M ðuÞdh ð20Þ
where the incremental displacement and deformation quantities,

du1 ; du2 ; dD, and dh, can be expressed in terms of du through the fol-
lowing relationships (with dt ¼ du according to Eq. (1)):
_
dh ¼hdu ð21Þ
dD ¼D_ du ð22Þ
l1
du1 ¼ du ð23Þ
l1 þ l2
l2
du2 ¼ du ð24Þ
l1 þ l2
Now, by inserting the right hand side of Eqs. (21)–(24) into Eq. Fig. 9. Load-displacement response of a single rebar crossing a frictionless joint.
(20) and by setting up the work equation, i.e. W E ¼ W I , the follow-
ing solution is obtained for the load-displacement response of a
perfectly smooth two-sided dowel joint:
Here MðuÞ; N ðuÞ; D_ ; h;

_ l1;ef , and l2;ef can be expressed as explicit eters given in Table 3. In the figure the different contributions to
functions of the displacement u as shown earlier. The solution is the response are indicated with dashed colored lines and the total
valid as long as u is less than the displacement capacity, umax , is represented by a solid line. The transition from combined dowel
which can be calculated from Eq. (3) by inserting the deformation and catenary action to pure catenary action is indicated with a
capacity of the rebar, Dmax . The index s in Ps ðuÞ indicates that this cross. It appears that the load-displacement response follows a
solution applies to a smooth interface without friction. How to descending branch (local drop) when the regime with in-elastic
include effects of friction will be shown in the next section. deformations initiates. The local drop is a result of the material
Solution (25a) applies to the regime, where a combination of assumptions adopted in the model which at the onset of in-
dowel action and catenary action exists in the rebar while pure elastic deformation estimates the capacity assuming a uniform dis-
catenary action (i.e. M ¼ 0 in the plastic hinge) is described by tribution of contact stresses at the dowel, see Fig. 7(a). When the
Eq. (25b). in-elastic shear displacement is increased, the equilibrium consid-
Fig. 9 shows an example of a response curve, Ps ðuÞ, as predicted erations applied in the deformed state, see Fig. 7(b), cause the load
by Eq. (25). The calculations were performed by use of the param- to drop locally. As the contribution from the normal force, N ðuÞ,
Table 3 Ptotal ðuÞ ¼ P s ðuÞ þ Pf ðuÞ ð27Þ

Parameters used for presentation of load-
displacement response for a single rebar Recently Santos and Júlio [29] gave an overview of the
crossing a frictionless interface. suggested values for the coefficient of friction related to shear-
Parameter Value friction theory. The values vary widely and some of the earlier
contributors to this area, e.g. Birkeland and Birkeland [30], sug-
d 8 mm
fy 500 MPa
gested values for artificially roughened surfaces (in the order of
c1 5.0 l ¼ 1:4) and l ¼ 0:8 1:0 for construction joints. However, it
f c;1 30 MPa should be noted that these rather high values were suggested/
c2 5.0 calibrated for a model, where all resistance in the joint was
f c;2 50 MPa assumed to be due to friction without consideration of dowel
Dmax 3 mm action. Bennett and Banerjee [31] suggested a combination of
friction (with l ¼ 0:6) and dowel action including a moment in
the interface, however, without combination of bending and
increases with increasing displacement, the load starts to increase tension in the rebar. Engström [24] introduced a model which
monotonically until a point where the deformation capacity of the combines friction and dowel action, taking combinations of bend-
rebar is exhausted and the load drops to zero. ing and tension in the rebar into account, and he suggested that
l ¼ 0:3 0:6 should be used for concrete-to-concrete interface
friction. Randl and Wicke [25] also considered combinations of
3.6. Inclusion of friction
dowel action and tension with friction and suggested l ¼ 0:5 for
smooth interfaces. These previous works did, however, not contain
The model presented in the previous applies to perfectly
a solution for how to relate the shear displacement, u, in the joint
smooth joints. However, casting joints can seldom be characterized
with the sectional forces carried by the rebar. The suggested coef-
as totally frictionless and it is therefore important for practical
ficients of friction (based on model calibration) are therefore partly
applications to take into account the contribution from interface
influenced by the estimate of the clamping forces.
friction. It turns out that the model actually provides most of the
In the experimental program of this study, the smooth inter-
information needed to include the effect of friction on the load-
faces were, as mentioned, treated with grease before casting of
displacement response. Fig. 10 illustrates a free body diagram of
the second batch to minimize friction. The friction coefficient
half of the specimen. It is assumed that the interface has a rough-
should therefore be less than that suggested by Randl and Wicke
ness that corresponds to a friction coefficient of l. At any given dis-
and most probably in the lower end of the range suggested by
placement, u, the tension force in the rebar, N ðuÞ, can be
Engström. In the following a coefficient of friction in the order of
determined by Eq. (14). The tension force causes compressive nor-
l ¼ 0:3 is adopted. This value of l has been used to determine
mal stresses to develop in the interface. The resultant of these so-
the load-displacement relationships shown in Fig. 11, calculated
called clamping stresses is equal to the horizontal projection of
by Eq. (27). The capacity without friction, i.e. Ps ðuÞ, as well as the
N ðuÞ. Then, from simple shear-friction considerations, the follow-
development of the frictional contribution, Pf ðuÞ, are also shown
ing would be the contribution from friction to the load carrying
as dashed lines. It can be seen that after the point of transition to
capacity:
pure catenary action (i.e. N ðuÞ ¼ N p ), the contribution from friction
Pf ðuÞ ¼ lNðuÞ cos h ð26Þ decreases. The reason for this is of course that the clamping force
decreases as the angle of rotation, h, increases. It appears from
Here index f indicates that the contribution relates to friction Fig. 11 that friction does not have an influence on the first order
and the angle h is given by Eq. (2). The total load-displacement plastic solution for pure dowel action. The response curve, how-
response when including the effect of dowel and catenary action ever, is shifted upwards once the rebar starts to carry tensile forces
as well as the effect of friction then amounts to: and friction is activated.
Fig. 10. Contribution of friction in the interface. Fig. 11. Inclusion of interface friction to the model response.
Fig. 12. Comparison of model with test results - 1 rebar. Fig. 13. Comparison of model with test results - 2 rebars.
4. Comparison of model with test results For concrete, a value of c ¼ 5 is adopted, similar to the findings
of Vintzeleou and Tassios [3]. To obtain reasonable agreement with
The load-displacement response predicted by the model, Eq. tests, a smaller value has been adopted for mortar, namely c ¼ 4.
(27), has been compared with the experimental results and The fact that the enhancement factor should be smaller in the case
depicted in Figs. 12–14. The basic input parameters have been of mortar may possible be explained by the relatively more brittle
obtained from Table 1. By adopting a friction coefficient of behavior of mortar (as compared to concrete) due to the relatively
l ¼ 0:3 as argued for in the previous, there is only the enhance- less aggregate content as well as the relatively smaller average
ment factor, c, left to calibrate the model with test results. aggregate size.
The model only predicts the non-linear behavior. The calculated

response curves have therefore been shifted horizontally to a dis-
placement, u0 , where the plastic behavior begins. This is shown
in the figures with a shaded light grey area and is in Fig. 12(a) indi-
cated by arrows showing ‘elastic regime’ and ‘plastic regime’. To
determine the horizontal shift, an estimate of the maximum dis-
placement in the elastic regime has to be derived. In the present
work, this estimate has been obtained by modeling the rebar as a
beam on an elastic foundation with a constant stiffness, e.g. as
shown in Refs. [2–10]. The calculations can be seen in Appendix
A. Despite the simplicity of the elastic model, Eq. (A.1) actually pro-
vides a reasonable estimate of the displacement, u0 , where the in-
elastic displacement regime begins.
It can be seen that with the adopted enhancement factors, c, and
the coefficient of friction, l, the general development of the tested
response curves is captured rather well. The predicted response
also displays a concave development followed by a convex devel-
opment, although not entirely coinciding with the tested response.
As mentioned earlier, the test responses show a dependency of the
compressive strength of the concrete. The model explains some of
these observations. As already shown by Rasmussen [13], the load
that causes transition to the non-linear behavior is proportional to
the geometric mean of f c and f y , which is observed e.g. in Fig. 4. On
the other hand, the ultimate load at a high shear displacement is a
result of catenary action, where change of geometry plays an
important role. Here, a lower compressive strength of the concrete
(or mortar) is beneficial for catenary action because this requires
plastic deformation over a longer length of the rebar and at the
same time makes it easier for the rebar to cut its way through
the concrete (or mortar). This eventually leads to larger maximum
displacement and thereby also a higher ultimate load. This is, to
some extend, captured by the model. In addition, the calculations
based on f u provide the best estimate of the ultimate peak load.
This result is as expected since the peak load corresponds to rein-
forcement rupture (as observed in tests).
As a final remark, it should be noted that although the model
does not agree with test results in all aspects, then for robustness
assessment, the model seems to have potential for practical use.
The reason for this is that in a robustness analysis, the dissipated
plastic energy (i.e. the area below the response curve) is a main
concern. In this respect, and based on Figs. 12–14, the model seems
to be able to provide a reasonable estimate of the available plastic
energy.
5. Conclusions
Push-off tests of two-sided dowel joints have been performed

and a simple, second order plasticity model has been developed
to predict the in-elastic regime of the load-displacement response.
The main experimental observations were:
The load-displacement relationships were non-linear and con-

sisted of convex followed by concavely shaped curves. The load
Fig. 14. Comparison of model with test results - 4 rebars.
increased with increasing shear displacement until the ultimate
capacity corresponding to rupture of the rebar(s) was reached
A mutual distance of 5.25d between the rebars in the interface
As can be seen in Figs. 12–14, the model predictions consist of plane did not lead to any observable group effect, as the magni-
an upper and a lower curve with the area in between shaded in tude of the results from specimens with n > 1 were propor-
grey. The lower curves represent calculations based on the yield tional to the results of specimens with a single rebar (n ¼ 1)
stress, f y , while the upper curves are based on calculations where Higher compressive strength of the concrete/mortar led to a
f y in the model has been replaced by the tensile strength f u . Calcu- higher load at the transition point between linear and non-
lations with f u have been included because hardening of reinforce- linear behavior
ment is not included in the model. The shaded area between the Specimens with a relatively lower compressive strength had a
two limiting curves thus indicates the expected range of test greater displacement capacity and a higher ultimate load than
results. specimens with a relatively higher compressive strength
Despite the simplicity and the relatively small number of [3] Vintzeleou EN, Tassios TP. Mathematical models for dowel action under
monotonic and cyclic conditions. Magaz Concr Res 1986;38(134):13–22.
required input parameters, the developed second order plasticity [4] Dei Poli S, Di Prisco M, Gambarova PG. Shear response, deformations, and
model captured the main characteristics of the tested load- subgrade stiffness of a dowel bar embedded in concrete. ACI Struct J 1992;89
displacement responses well. Calculations based on the yield stress (6):665–75.
[5] Dei Poli S, di Prisco M, Gambarova PG. Cover and stirrup effects on the shear
of the reinforcement provided an appropriate estimate of the load response of dowel bar embedded in concrete. ACI Struct J 1993;90(4):441–50.
level, where non-linear behavior initiated. A reasonable estimate of [6] Mannava SS, Bush Jr TD, Kukreti AR. Load-deflection behavior of smooth
the ultimate capacity at large shear displacements could be calcu- dowels. ACI Struct J 1999;96(6):891–8.
[7] Randl N. Load bearing behaviour of cast-in shear dowels. Beton- und
lated by use of the ultimate strength of the reinforcement. The Stahlbetonbau 2007;102(S1):31–7.
model has potential for practical assessment of structural robust- [8] Tanaka Y, Murakoshi J. Reexamination of dowel behavior of steel bars
ness, where estimation of available plastic energy (area below embedded in concrete. ACI Struct J 2011;108(6):659–68.
[9] He XG, Kwan AKH. Modeling dowel action of reinforcement bars for finite
load-displacement response) is important. element analysis of concrete structures. Comput Struct 2001;79(6):595–604.
[10] El-Ariss B. Behavior of beams with dowel action. Eng Struct 2007;29
Acknowledgment (6):899–903.
[11] Marcus H. Load carrying capacity of dowels at transverse pavement joints. J
Am Concr Inst 1951;23(2):169–94.
The experimental research presented in this paper was finan- [12] Soroushian P, Obaseki K, Rojas MC. Bearing strength and stiffness of concrete
cially supported by the COWI Foundation. The experimental work under reinforcing bars. ACI Mater J 1987;84(3):179–84.
[13] Højlund Rasmussen B. Betonindstøbte tværbelastede boltes og dornes
was supported by M.Sc. Nikolai Bach and M.Sc. Martin Hansen dur-
bæreevne, English: Resistance of embedded bolts and dowels loaded in
ing their master thesis. The authors gratefully acknowledge these shear. Bygningsstatiske Meddelelser 1963;34(2):39–55.
valuable contributions. [14] Calladine CR, Simple ideas in the large-deflection plastic theory of plates and
slabs. In: Heyman J, Leckie FA, (Eds.). International conference on the
applications of plastic theory in engineering design, Cambridge, UK, 1968, p.
Appendix A. Modeling of elastic displacement for dowel joints 93–127.
[15] Bræstrup MW. Dome effect in RC slabs: rigid-plastic analysis. J Struct Div –
ASCE 1980;106(6):1237–53.
The elastic displacement can be estimated by the analogy of a
[16] Belenkiy LM. Upper-bound solutions for rigid-plastic beams and plates of large
beam on an elastic foundation [2–10]. By solving the fourth order deflections by variation principles. J Eng Mech 2007;133(1):98–105.
differential equation, the maximum elastic displacement can be [17] Bailey CG, Toh WS, Chan BM. Simplified and advanced analysis of membrane
action of concrete slabs. ACI Struct J 2008;105(1):30–40.
formulated as given in Ref. [4]:
[18] Mirzaei Y. Post-punching behavior of reinforced concrete slabs [Ph.D.
P0 thesis]. École Polytechnique Fédérale de Lausanne; 2010.
u0 ¼ ðA:1Þ [19] Fernández Ruiz M, Mirzaei Y, Muttoni A. Post-punching behavior of flat slabs.
2k3 Es Is ACI Struct J 2013;110(5):801–12.
[20] Gouverneur D, Caspeele R, Taerwe L. Experimental investigation of the load-
where P 0 is the maximum shear force (e.g. estimated by the load displacement behaviour under catenary action in a restrained reinforced
required to form plastic hinges, i.e. Eq. (25a) with u = 0), Es is the concrete slab strip. Eng Struct 2013;49:1007–16.
elastic modulus and Is is the second moment of area for the circular [21] Dulácska H. Dowel action of reinforcement crossing cracks in concrete. ACI J
1972;69(12):754–7.
rebar. The parameter k expresses a stiffness per length and is given [22] Millard SG, Johnson RP. Shear transfer across cracks in reinforced concrete due
by [2]: to aggregate interlock and to dowel action. Mag Concr Res 1984;36(126):9–21.
sffiffiffiffiffiffiffiffiffiffiffi [23] Vintzeleou EN, Tassios TP. Behavior of dowels under cyclic deformations. ACI
Struct J 1987;84(1):18–30.
4 kc d
k¼ ðA:2Þ [24] Engström B. Combined effects of dowel action and friction in bolted
4Es Is connections. Nordic Concr Res 1990;9:14–33.
[25] Randl N, Wicke M. Schubübertragung zwischen Alt- und Neubeton. Beton- und
where kc is the stiffness of the elastic foundation. For the concrete Stahlbetonbau 2000;95:461–73. Heft 8.
material a stiffness, kc , is suggested by Soroushian [12]: [26] Sørensen JH, Hoang LC, Olesen JF, Fischer G. Catenary action in rebars crossing
a casting joint loaded in shear. Proceedings of the 11th fib international PhD
pffiffiffiffi symposium in civil engineering 2016:735–42.
127 f c
kc ¼ 2=3
ðA:3Þ [27] fib, fib Model Code for Concrete Structures 2010, Wilhelm Ernst & Sohn,
d Lausanne, Switzerland; 2013.
[28] Nielsen MP, Hoang LC. Limit analysis and concrete plasticity. 3rd ed. Boca
Raton: CRC Press, Taylor & Francis Group; 2011.
References [29] Santos PMD, Júlio ENBS. A state-of-the-art review on shear-friction. Eng Struct
2012;45:435–48.
[30] Birkeland PW, Birkeland HW. Connections in precast concrete construction.
[1] fib, Bulletin 43: Structural Connections for Precast Concrete Buildings,
ACI J 1966;63(3):345–68.
Fédération international du béton, Lausanne, Switzerland; 2008.
[31] Bennett EW, Banerjee S. Strength of beam-column connections with dowel
[2] Friberg BF. Design of dowels in transverse joints of concrete pavements. Am
reinforcement. Struct Eng 1976;54(4):133–9.
Soc Civ Eng 1938;64(9):1809–28.
In precast reinforced concrete structures, the joints and connections between the
precast components are essential for the structural performance. This thesis pre-
sents a new design for keyed connections typically used between shear wall pa-
nels. An extensive experimental program is presented in combination with sound
design methods developed on the basis of the theory of plasticity. The theoreti-
cal models include first order upper and lower bound modeling of keyed connec-
tions as well as second order modeling of the load-displacement relationship for
casting joints exposed to a simple displacement field. Design and Modeling of Structural
Joints in Precast Concrete Structures

PhD Thesis

2018
DTU Civil Engineering R-384
DTU Civil Engineering

Brovej, Bygning 118

2800 Kongens Lyngby
www.byg.dtu.dk
ISBN 9788778774798
ISSN 1601-2917

PHD Thesis Jesper Harrild S Rensen Final Reduceret PDF

Uploaded by

Copyright:

Available Formats

PHD Thesis Jesper Harrild S Rensen Final Reduceret PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PHD Thesis Jesper Harrild S Rensen Final Reduceret PDF

Uploaded by

Copyright:

Available Formats

Downloaded from orbit.dtu.

dk on: Sep 19, 2019

Design and Modeling of Structural Joints in Precast Concrete Structures

Sørensen, Jesper Harrild

Link back to DTU Orbit

Design and Modeling of Structural Joints in Precast Concrete Structures

Department of Civil Engineering

DTU Civil Engineering R-384

DTU Civil Engineering

Brovej, Bygning 118

Jesper Harrild Sørensen

Department of Civil Engineering

Design and Modeling of Structural Joints in Precast Concrete

Jesper Harrild Sørensen

Preface to Published Version

Jesper Harrild Sørensen

Precast reinforced concrete components are widely used for construction

Præfabrikerede betonelementer benyttes i stor udstrækning til nybyggeri i

3 Tests of Mortar in Triaxial Compression 23

4 Failure of a Mortar Shear Key 43

5 Tensile Capacity of Loop Connections 55

6 Tests of Shear Connections 67

8 Lower Bound Solutions for Keyed Shear Connections 111

9 Comparison of Upper and Lower Bound Solutions 131

10 Dowel and Catenary Action in Rebars Crossing a Shear

A Results of Triaxial Tests 177

B Summary of Push-off Tests on Shear Connections 183

C Upper Bound Calculations for Shear Connections 191

D Lower Bound Calculations for Shear Connections 197

2. Catenary Action in Rebars Crossing a Casting Joint Loaded in Shear

3. Tensile Capacity of U-bar Loop Connections with Precast Fiber Rein-

a Distance between innermost U-bars

1.1 Precast Concrete Structures

Department of Civil Engineering - Technical University of Denmark 1

Nielsen (summarized in (Nielsen and Hoang, 2011)). Continuous develop-

Figure 1.1: Examples of landmark buildings in Denmark, built using precast

The prefabrication process promotes quality control, precision, and pro-

2 Department of Civil Engineering - Technical University of Denmark

concrete structures (CEN, 2004)), where empirical relations dominate the

1.2 Motivation of Study

1.2.1 Development of Design

Department of Civil Engineering - Technical University of Denmark 3

Horizontal section Vertical section

(b) Assembly phase of conventional design

Figure 1.2: Conventional design of shear wall connection, illustrations from

4 Department of Civil Engineering - Technical University of Denmark

Horizontal section Vertical section

(b) Assembly phase of new design

These challenges can be overcome by a change of design. Figure 1.3(a)

Department of Civil Engineering - Technical University of Denmark 5

inforcement or a type of locking bar, however, to distinguish between the

1.3 Objectives of Thesis

• are easy to complete on the construction site

1.4 Organization of the Thesis

6 Department of Civil Engineering - Technical University of Denmark

Table 1.1: Content of the thesis divided by chapter

No. Content Contained in

Chapter 3 presents experimental results of triaxial tests on mortar and

Department of Civil Engineering - Technical University of Denmark 7