Beam - Column Connections Subjected to

Static and Dynamic Loading

Herman Frich

Civil and Environmental Engineering

Submission date: June 2014
Supervisor: Arild Holm Clausen, KT
Co-supervisor: Arne Aalberg, KT
Erik Grimsmo, KT

Norwegian University of Science and Technology

Department of Structural Engineering
 ACCESSIBILITY
Department of Structural Engineering
Faculty of Engineering Science and Technology OPEN
NTNU- Norwegian University of Science and Technology

MASTER THESIS 2014

SUBJECT AREA: DATE: NO. OF PAGES:

Computational Mechanics 11 June 2014 57 + 18

TITLE:

Beam – Column Connections Subjected to Static and Dynamic Loading

Bjelke-søyle – forbindelser påkjent av statisk og dynamisk last

BY:

Herman Frich

SUMMARY:

In recent years, an increased awareness has been on the reliability of bolted connections in
extreme events, such as the loss of a load bearing column in a terrorist attack. The ability
to transfer the forces through the joints is key to maintain the structural integrity and
prevent a progressive collapse of buildings. Sudden dynamic loading may cause a shift in
the response behavior that is not captured by common design methods, which are often
based on static conditions.
This thesis investigates the behaviour of a bolted steel connection in a column removal
scenario. Experimental tests have been conducted under rapid, non-cyclic loading
conditions and simplified methods based on European design standards (Eurocode) as well
as advanced numerical analyses have been compared to the experimental results. The goal
was to reveal possible implications on design of joints to improve the safety of structures.
The experimental tests showed that the assembly failed in flexure, typical for moment
connections. This was true for both quasi-static and dynamic loading conditions, and was
predicted by the simplified design method and the numerical models. However, the
numerical simulations revealed that rapid loading caused a transition from the typical
bending- to shear action due to inertia effects.

RESPONSIBLE TEACHER: Professor Arild H. Clausen

SUPERVISOR(S): Arild H. Clausen, Arne Aalberg and Erik L. Grimsmo

CARRIED OUT AT: SIMLab, NTNU

NORGES TEKNISK-NATURVITENSKAPELIGE UNIVERSITET
INSTITUTT FOR KONSTRUKSJONSTEKNIKK

MASTER THESIS 2014

Herman Frich
Beam – Column Connections Subjected to Static and
Dynamic Loading
(Bjelke-søyle – forbindelser påkjent av statisk og dynamisk last)

It is important to understand the behaviour of connections subjected to extreme loads, e.g.

dropped objects and sudden column-removal scenarios. Connections are often the critical
component in a structure, and if they fail, progressive collapse is imminent. Bolted steel
connections are common in various structures such as office buildings and off-shore
platforms. There are established design rules for these connections, e.g. Eurocode NS-EN
1993-1-8. However, these design rules are based on various assumptions and
simplifications. An important one is that quasi-static conditions are assumed. It is therefore
interesting to investigate the behaviour of these connections through experiments and
numerical analyses and compare them with the Eurocode.

In this master thesis, a double-sided beam – column connection will be tested and
analysed. The beams have an endplate welded onto one of the ends. The endplates are
bolted to the column flanges. The research project has three main objectives: (1)
determine the material properties of the various components of the connection, (2)
investigate experimentally the behaviour of the bolted steel connection, and (3) compare
the experimental results with analyses from the finite element method and Eurocode.

Some keywords for activities related to this master thesis project may include:
• Literature survey: Behaviour of bolted connections (articles, codes, text books).
• Material tests: Uniaxial tension tests. Identification of parameters for numerical model.
• Component tests: Quasi-static and possibly dynamic tests on a bolted end-plate
connection. The connection will be tested with a direction of the load such that tension
will occur in the lower beam flanges.
• Numerical analyses: FEM simulations of component tests. Validation of model.
• Design code: The capacity and the stiffness of the connection determined from
experiments and numerical analyses should be compared with estimates from Eurocode
NS-EN 1993-1-8.

The candidate may agree with the supervisors to pay particular attention to specific parts of
the investigation, or include other aspects than those already mentioned.

The thesis is to be organized as a research report, recognizing the guidelines provided by

Department of Structural Engineering.

Supervisors: Erik Grimsmo, Arne Aalberg and Arild Holm Clausen

The report is to be handed in not later than 11 June 2014.

NTNU, 13 January 2014

Arild Holm Clausen

Preface
This thesis is finalized at the end of the spring semester of 2014 and concludes my
Master of Science degree at NTNU. The work has been a part of an experimental
program carried out at the Structural Impact Laboratory (SIMLab).

Over the past years I have developed an interest in the practical application of
the finite element method in structural analysis problems. Therefore, I would
like to thank professor Arild Holm Clausen for introducing me to the subject of
Computational Methods in Structural Mechanics. His encouraging approach to
teaching has been much appreciated.

Gratitude is also extended to my supervisor, PhD candidate Erik Løhre Grimsmo,

whose guidance and support have helped me a lot in writing this thesis, the results
would not have been the same.

Finally, I would like to honor my fellow students, Gjermund Båsen and Torger
Nordgård, for valuable discussions throughout this semester.

Trondheim, June 2014

Herman Frich

I
II
Abstract
Bolted beam-column connections are commonly used in office buildings and off-
shore platforms in Norway. The use of pre-fabricated components in frame struc-
tures is popular due to the cost effective and quick erection of buildings, where
engineered members can be manufactured with great accuracy in a controlled en-
vironment at a fabrication shop.

In recent years, an increased awareness has been on the reliability of these con-
nections in extreme events, such as the loss of a load bearing column in a terrorist
attack. A lack of study on bolted connections under dynamic loading has been
revealed, and a number of experimental programs have been initiated.

The ability to transfer the forces through the joints is key to maintain the structural
integrity and prevent a progressive collapse in buildings. In addition, sudden
dynamic loading may cause a shift in the response behavior that is not captured
by common design methods, which are often based on static conditions.

Therefore, a test program was initiated to investigate the behavior of a bolted

connection under rapid, non-cyclic loading in a column removal scenario. Exper-
imental tests of a beam-column assembly were conducted under quasi-static and
dynamic loading conditions.

Simplified calculations based on European design standards (Eurocode) as well

as advanced numerical analyses were performed and compared directly to the
experimental findings. The goal was to reveal possible implications on design
of joints to extreme loads to improve the safety of structures.

The experimental tests revealed that the assembly failed in flexure, typical for
moment connections. This was true for both quasi-static and dynamic loading
conditions, and was predicted by the simplified design method and the numer-
ical models. However, an overly safe estimate of the capacity was obtained by
Eurocode’s design method.

Furthermore, the numerical simulations revealed that rapid loading caused a tran-
sition from the typical bending- to shear action due to inertia effects. Further
investigation into what type of load regime that will cause a change from flexural
to shear failure has therefore been proposed for further work.

III
IV
Sammendrag
Skrudde bjelke-søyleforbindelser er mye brukt i næringsbygg og oljeplattformer i
Norge. Prefabrikkerte komponenter i rammekonstruksjoner er populært på grunn
av en kostnadseffektiv og rask oppføring av nybygg fordi stålkomponentene kan
produseres i et kontrollert miljø på et mekanisk verksted.

Det har i de senere årene vært en økende oppmerksomhet på påliteligheten til disse
forbindelsene i ekstreme lasttilfeller, for eksempel ved tap av en bærende søyle i et
terrorangrep. Relativt få studier har blitt gjort på skrudde forbindelser utsatt for
dynamiske laster, og har ført til at en rekke eksperimentelle testprogrammer har
blitt iverksatt.

Evnen til å overføre krefter gjennom knutepunktene er avgjørende for å oppret-

tholde bæreevnen og hindre en progressiv kollaps ved tap av søyler i bygninger.
I tillegg kan en plutselig dynamisk belastning forårsake endringer i responsen
som ikke fanges opp av vanlige beregningsmetoder, som ofte er basert på statiske
forhold.

Et testprogram ble derfor startet opp for å undersøke oppførselen til en skrue-
forbindelse under plutselig, ikke-syklisk belastning. Fullskala tester av en bjelke-
søyle forbindelse ble gjennomført under kvasistatiske og dynamiske belastnings-
forhold.

Forenklede beregninger basert på europeiske standarder (Eurocode) samt avanserte

numeriske analyser ble utført og sammenlignet med de eksperimentelle resultatene.
Målet var å avdekke mulige mangler i disse beregningsmodellene og dermed øke
sikkerheten til bygninger.

De eksperimentelle testene viste at forbindelsen gikk til brudd i bøyning, typisk

for momentstive forbindelser. Dette gjaldt for både kvasistatiske og dynamiske
belastningsforhold, og ble forutsagt av den forenklede beregningsmodellen og de
numeriske beregningene. Imidlertid ga metoden presentert i standarden et for
konservativt estimat av kapasiteten.

Resultatene viste også at en rask belastning forårsaket endringer i lastfordelin-

gen. Skjærkreftenes relative påvirkning økte på grunn av treghetskrefter i de
dynamiske simuleringene. Derfor ble det foreslått å gå videre med å undersøke
hvilke dynamiske lastforhold som vil føre til skjærbrudd.

V
VI
Contents
Preface I

Abstract III

Sammendrag V

List of symbols X

1 Introduction 1
1.1 Bolted connections subjected to extreme loads . . . . . . . . . . . . 1
1.2 Previous work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Experimental program . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Scope of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Theory 5
2.1 Yield line design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Uniaxial tension test . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Digital image correlation (DIC) . . . . . . . . . . . . . . . . . . . . 12

3 Design of joint to Eurocode 3 14

3.1 Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 Material tests 19
4.1 Experimental program . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Quasi-static tensile tests . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Work hardening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.4 Rate sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.5 Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5 Experimental tests of joints 28

5.1 Quasi-static experiment setup . . . . . . . . . . . . . . . . . . . . . 29
5.2 Dynamic experiment setup . . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Quasi-static results . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.4 Dynamic results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

6 Finite element simulations 35

6.1 Finite element model . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.2 Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

VII
 6.3 Quasi-static results . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.4 Dynamic results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

7 Discussion 49
7.1 Experimental findings . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.2 Assessment of the Component method (Eurocode 3) . . . . . . . . . 49
7.3 Finite element predictions . . . . . . . . . . . . . . . . . . . . . . . 51

8 Conclusions and suggestions for further work 54

8.1 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.2 Future studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

References 55

A Capacity according to NS EN 1993-1-8 59

B Stiffness according to NS EN 1993-1-8 70

C Additional pictures of the experimental tests 75

VIII
List of symbols
∆L Change in gauge length of tensile specimen

∆ Vertical displacement of column

δ Opening of end-plate relative to column flange

ṗ Plastic strain rate

µ Coefficient of friction

ν Poisson’s ratio

φ Joint rotation

ρ Mass density

σ0 Initial yield stress

σavg Average stress in tensile specimen ( FA )

σeq Equivalent stress

σij Stress component

σi Principal stress component

εf Fracture strain

εij Strain component

εi Principal strain component

ζ Triaxial stress correction function

A Cross section of tensile specimen

c Viscous stress exponent, see Eq. (6)

E Elastic modulus

e Engineering strain

F Axial force in tensile specimen

f Yield function

IX
fu Ultimate tensile strength
fy Yield strength
Ft,Rd Tension resistance of bolt
Ib Area moment of inertia of the beam cross section
ki Linear spring stiffness
kφ,ini Initial rotational stiffness of joint
kφ Rotational stiffness of joint
keq Equivalent stiffness
L Gauge length of tensile specimen
Lb Length of beam, see Figure 13
Lc Beam span used in design, see Figure 13
Lφ Length of connection, see Figure 13
Ls Beam span used in experimental tests, see Figure 26
MEd Design moment
Mj,Rd Moment resistance of joint
P Vertical column force
p Plastic strain
pu Plastic strain at maximum load in tensile specimen
peq Equivalent plastic strain
R Isotropic hardening
s Engineering stress
t Thickness of tensile specimen
VEd Design shear force
w Width of tensile specimen
Wc Internal strain energy at fracture
z Lever arm

X
1 Introduction
1.1 Bolted connections subjected to extreme loads
Bolted steel connections are commonly used in office buildings and off-shore plat-
forms in Norway, usually to ensure the vertical load bearing of floors. Standardized
sections and plates can be welded together with great accuracy in a fabrication
shop using specialized machinery. The manufactured assemblies allows for a quick
erection of a complete frame structure at the construction site, therefore reducing
the uncertainties regarding the costs.
The structural integrity is usually achieved in the design process through sim-
plified models of the structural system, where each component must resist the
applied loads in an ultimate limit state [1]. Abnormal loads are included in the
accidental limit state design.

A column removal scenario

An example of an undesired incident that may be included in an accidental limit
state is depicted in Figure 1. An explosive blast or collision has lead to the removal
of a central column in a frame structure.

Figure 1: Frame structure under a design load case (top) and under a column removal
scenario (bottom).

Loss of vertical load bearing causes a sudden transfer of forces, and the bending
moment diagram reveals that the loading direction is reversed. Thus, joints that
are optimized to transfer the design forces may function poorly.

1
 Furthermore, structural components such as hollow core slabs can loose support
and fall onto subjacent floors. If a column is hit suddenly, the acceleration of the
surrounding beams will generate inertia forces that acts in the opposite direction
as illustrated in Figure 2. The inertial resistance will have a positive effect on
the bending moment at midspan, but the shear forces will have a relatively higher
impact. Shear failure may therefore be of great importance when loads are applied
suddenly.

Figure 2: Dynamic equilibrium in a central column removal scenario.

1.2 Previous work

An extensive test program of bolted end-plate connections under static loading was
undertaken by Coelho et al. [2]. One of the objectives was to evaluate the design
methods proposed by Eurocode [3], which is used to determine the capacity and
stiffness of joints. They found that increasing the thickness of the end-plate in a
beam-column connection will generally increase resistance, but decrease ductility
and hence the rotational capacity of the joint. Eurocode gave safe estimates for
joint resistance and rotational capacity, but overestimated the initial stiffness.
They proposed that new procedures must be developed for estimation of rotational
capacity, which was overly conservative according to the standard.
Experimental tests and numerical simulations of various moment connections
under a central column removal scenario have been conducted by Yang and Tan
[4, 5] and Sadek et al. [6, 7]. The focus has been on studying the effects of catenary
action. An initial flexural behavior and a gradual plastification typical for moment
connections were observed. However, the assemblies failed under a combination of
bending- and axial stress. An increase in the vertical load capacity by as much as
100 % was observed for flush end-plate connections due to catenary action [5]. A
sufficiently ductile behavior was essential, and many welded connections did not

2
show a significant increase in capacity. Response characteristics and failure modes
were accurately captured by non-linear finite element models [4, 7].
While a great deal of studies have been carried out on steel moment-connections
subjected to cyclic loading, relatively few publications have been on moment con-
nections subjected to short impulse loading. Most of these focus on blast loading
of frame structures, not the individual joints. However, numerical studies such as
the one carried out by Subawala et al. [8] revealed that the finite element method
can be used to assess the stress distribution in moment connections under blast
loading.
In despite of efforts, no research studies of bolted beam-column connections
subjected to short impulse loads have been found.

1.3 Experimental program

This thesis is part of an ongoing experimental test program carried out at the
Structural Impact Laboratory (SIMLab) to study the behavior of a bolted end-
plate connection under rapid, non-cyclic loading. The double-sided connection
consists of two beam members with welded end-plates bolted to a column as de-
picted in Figure 3. The beams (HEA180), column (HEB220) and plates are of
standard structural grade steel S355, while the bolts (M16) are of class 8.8.

Figure 3: Double sided connection investigated in the experimental test program.

The joint is designed to fail in flexure due to large tensile forces in the top
flange, which is the typical load case in a frame structure as illustrated in Figure
1. The joint is therefore optimized to transfer the design moment by having
an extended end-plate in the upper part. The performance of the joint in this
particular design configuration was investigated in a parallel Master’s thesis by
Baasen and Nordgaard [9].

3
1.4 Scope of thesis
In the extreme event of a sudden loss of vertical support, the load direction is
reversed as illustrated by the arrows in Figure 3. This is the basis of the ex-
perimental tests studied in this thesis. In addition, ordinary design methods as
well as advanced numerical analyses have been performed and compared to the
experimental results. The simplified methods used in design codes rely on various
assumptions to allow for hand calculations. An important one is static conditions,
therefore neglecting any dynamic effects. Non-linear finite element analysis can be
a powerful tool when used properly, and is indispensable when the response is to
be determined until failure. However, it is essential to be aware of the limitations
in these numerical models.

Objective
The thesis objective is twofold: (1) investigate how typical design methods manage
to predict failure when load direction is reversed and (2) reveal if sudden dynamic
loading will significantly change the characteristics in response and failure mode,
in particular shear failure of bolts or welds.

Outline
The thesis is divided into 8 sections as shown below:
• Section 2, Theory: most of the underlying theory used in this thesis is pre-
sented.
• Section 3, Design of joint to Eurocode 3: the capacity and stiffness are
determined according to methods proposed by Eurocode.
• Section 4, Material tests: the material test program is presented. The focus
is on how important material properties were extracted from the test data.
• Section 5, Experimental tests of joints: the experimental test program is
explained and important results are given.
• Section 6, Finite element simulations: the numerical model used to simulate
the experimental tests is established and tested for sensitivities.
• Section 7, Discussion: the simplified design method as well as the finite
element predictions are compared to the experimental results. The changes
in the response characteristics due to sudden dynamic loading have been
emphasized.
• Section 8, Conclusions and suggestions for further work: the main findings
are given and possible improvements are proposed for future studies.

4
2 Theory
Simple hand calculations and non-linear finite element simulations are used to
predict the response of the connection in Figure 3. A summary of the most impor-
tant underlying theory used throughout the thesis is presented in the subsequent
sections.
First, in Section 2.1, the principle of virtual work applied to yield line design
is given. This is the basis of the Component method in Eurocode 3.
The constitutive model of a rate dependent elasto-plastic material used in the
finite element model is presented in Section 2.2.
Finally, in Section 2.3-2.4, the development of a finite stress-strain measure is
given. The focus is on how an elasto-plastic material model can be calibrated from
tensile tests.

2.1 Yield line design

Yield line design is a method used to find the ultimate load capacity of transversely
loaded plates. The theory is a generalization of the plastic hinge method used for
beams in frame structures, and can be found in the book by Larsen [10].

Basic principle
All deformations are localized along specific yield lines, which forms a failure mech-
anism. All other parts remain elastic and move as rigid bodies. It follows from
the principle of virtual work that the work done by rotation along the plastic yield
lines (WP ) must be equal to the work done by the externally applied load (WE ):

WP = WE (1)
It can be shown that the work done by rotation around the yield line is the
same as the work done by rotation along the projected length of the yield line onto
the support axis (see Figure 4), and a simple expression for the internal plastic
work can be found:
WP = ΣmP · φi · li (2)
where mP = fy · t2 /4 is the moment resistance per unit length of a plate with
thickness t and a yield stress of fy . Furthermore, li is the length of the projected
yield line onto the support axis with angle φi .
The external work is taken as the force multiplied with displacement:
Z
WE = q(x, y) · w(x, y) · dA (3)

5
where q(x, y) is the applied load per unit area and w(x, y) is the transverse dis-
placement.
The critical load is obtained by finding the mechanism which requires the least
amount of work to develop. Therefore, an infinite number of different mechanisms
must be checked.

Simple problem
An example is presented to illustrate how the method can be used in design. A
simply supported plate with sides a is loaded by a point load P and the capacity
is determined for the yield line mechanism shown in Figure 4.

Figure 4: Simply supported plate subjected to point load.

The projected yield line length is Σli = 4a with a rotation angle φ = 2∆/a.
The internal work is calculated from Eq. (2):
2∆
Wp = mp · 4a · = 8 · mp · ∆
a
The external work is obtained from Eq. (3) by multiplying the load with the
vertical displacement:

WE = P · ∆
Demanding energy conservation (Eq. (1)) and solving for the applied load gives
the limit state capacity:

P = 8 · mp

6
Yield line design applied to beam-column connections
Due to complex geometries and loading conditions in real life structures, the critical
failure mechanism can be difficult to find. One example is bending of a column
flange in a bolted beam-column connection as shown in Figure 5.

Figure 5: Possible yield mechanism for a column flange in bending.

Simplified T-stub models have therefore been developed and implemented in

design codes. These are based on the underlying principle of yield line design and
validated by experimental tests. A limited number of failure mechanisms must be
checked as shown in Figure 6. The ultimate capacity will depend on the thickness
of plates, the placement of bolts and the material strength.

Figure 6: Failure mechanisms for T-stub model used in Eurocode [3]. Yielding of plate
(left), bolt failure and yielding of plate (middle) and bolt failure (right)

7
2.2 Constitutive model
An outline of the theory behind a rate-dependent plasticity model with isotropic
work hardening will be given. The presented theory is based on lectures given by
Hopperstad and Boervik [11], and can be found in books by Irgens [12], Lubliner
[13] and Lemaitre and Chaboche [14].

Figure 7: Rheological illustration of constitutive model.

The basic principles of the constitutive model are illustrated for a uniaxial
stress state in Figure 7. There is an elastic spring corresponding to the reversible
deformation, coupled in series with a dashpot and a friction element, which allows
for plastic dissipation and irreversible deformation.

Elasticity
Under small deformations, metals will generally show a linear stress-strain rela-
tionship. Hooke’s law for isotropic material in three spatial dimensions is given as:
    
σ11 (1 − ν) ν ν 0 0 0 ε11
σ22 
 
 ν
 (1 − ν) ν 0 0 0   ε22 
 
σ33 
 = E
 ν
 ν (1 − ν) 0 0 0   ε33 
 
σ12 
 
(1+ν)(1−2ν)  0
 0 0 (1 − 2ν)/2 0 0  2ε12 
 
σ23   0 0 0 0 (1 − 2ν)/2 0  2ε23 
σ13 0 0 0 0 0 (1 − 2ν)/2 2ε13

Yielding
Yielding is initiated when the material is stressed beyond its yield limit, and atoms
begin to slide relative to each other. Mathematically, the yield criterion (f ) can
be expressed as:
f = σeq − σ0 = 0

8
where σeq is the equivalent stress (scalar) and σ0 is the initial yield stress of the
material.
The von Mises yield stress is most commonly used for isotropic materials and
can be written as:
r
(σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ1 − σ3 )2
σeq =
2
where σi is the principal stress components. In the case of uniaxial stress (σ2 =
σ3 = 0), the Mises stress will be reduced to:

σeq = σ1 (4)

Plastic flow
A material undergoes irreversible plastic deformations during yielding. The von
Mises yield criterion together with an associated flow rule gives:

3σij0
ε̇pij = ṗeq
2σeq

where σij0 denotes the deviatoric stress and the equivalent plastic strain (peq ) is
taken as an accumulated plastic flow:
r
3 p p
ṗeq = ε̇ ε̇
2 ij ij
Zt
peq = ṗeq dt
0

Work hardening
Additional strength due to movement of dislocations in material is often observed,
and the yield stress will increase when material exhibits plastic flow. Isotropic
hardening is suitable for problems involving large plastic deformation without
abrupt change in load direction, where the Bauschinger effect become important.
The yield criterion can therefore be taken as:

f = σ0 + R(peq ) (5)

where R is the strain hardening function.

9
Viscous stress
Ductile metals show an increase in yield strength and flow stress when rate of
straining increases. Strain rate dependency is often included on a multiplicative
form:
ṗ
σeq = (σ0 + R)(1 + )c (6)
ṗ0
where c is the parameter which determines the rate sensitivity and ṗ0 is a reference
strain rate. Observe that the equivalent stress simply scales with viscous term in
Eq. (6) and do not depend on the plastic strain.

Fracture
A ductile fracture criterion proposed by Cockcroft and Latham [15] is adopted
in this thesis. Fracture occur when the strain energy per unit volume reaches a
critical value:
Zεf
Wc = max(σ1 , 0)dε1 (7)
0

where the subscript 1 denotes the maximum principal direction and εf is the
fracture strain. Observe that only tensile stress is contributing to fracture, hence
tensile separation is the underlying concept.

2.3 Uniaxial tension test

The uniaxial tension test is a widely used mechanical test to determine material
properties such as strength and ductility. This section gives an outline of the theory
going from force and geometry measurements of a specimen to a true stress-strain
relationship used as input in finite element codes.

True strain and stress relations

Force (F ), displacement (∆L), width (w) and thickness (t) are continuously mea-
sured during a uniaxial tensile test.

Figure 8: Typical coupon used in a uniaxial tension test.

10
 Engineering stress-strain measures are calculated based on the undeformed
geometry as shown in Eq. (8a) and Eq. (8b) simply by dividing the elongation by
the initial length and force by the initial cross sectional area.
∆L
e= (8a)
L0
F
s= (8b)
A0
Their validity is restricted to small strains, and will not produce accurate results
in a tensile test where both specimen length and cross sectional area are severely
altered. Thus, if large strains are expected, new measures are needed.
A common approach is to introduce a logarithmic stress-strain relationship
given in Eq. (9a) and Eq. (9b), which can be established by integrating the strain:
Z ε Z L
dL dL L
dε = ⇒ dε = ⇒ ε = ln L − ln L0 = ln
L 0 L0 L L0

and by assuming plastic incompressibility (volume preservation):

A L0
=
A0 L
which leads to:
L
ε = ln
= ln(1 + e) (9a)
L0
F F L
σ= = = s(1 + e) (9b)
A A0 L0
The logarithmic stress-strain relationship can also be expressed solely by the
cross sectional area:
L A0
ε = ln = ln (10a)
L0 A
F
σ= (10b)
A

Stress triaxiality
Necking is a geometric instability that occur when the increase in strength due to
strain hardening is less than the increase in stress due to the reduction in cross
sectional area. Strain localization in the necked region of the specimen introduces
a multi-axial state of stress as depicted in Figure 9, and the relationship given in
Eq. (4) is therefore not valid.

11
 Figure 9: Principal stress components in neck region.

An estimate of the equivalent stress - plastic strain relationship can be obtained

by multiplying the average stress in the longitudinal direction (σavg = FA ) with a
triaxial stress correction function (ζ):

σeq = ζσavg (11)


= 1 peq < pu
ζ
< 1 peq > pu

where pu is the plastic strain at time of neck initiation (maximum load). Therefore,
the triaxial stress correction is only active during necking.
Empirical relations have been found for a variety of width-to-thickness ratios
and materials for rectangular cross sections. However, a study by Yazzie et al.
[16] revealed that a general neck correction has not been found for rectangular
specimens.
By utilizing axis symmetry and by making basic assumptions about the neck
geometry, Bridgman [17] found an analytical solution for circular specimens:
1
ζBridgman =
(1 + 2R/a) ln(1 + a/2R)

where a is the cross sectional radius and R is the radius of the neck contour,
assumed to be circular.

2.4 Digital image correlation (DIC)

Various optical instrument techniques have been developed to capture rigid body
motion and local strains during mechanical tests. One such method, digital image

12
correlation, has been increasing in popularity over the past decades due to easy
setup and use [18]. The basic principles explained in this section are based on
planar deformations, but the theory can be extended to three dimensional analyses.

Concept
A series of high resolution bit-map images are captured with a camera pointing
perpendicular to a flat surface. Each pixel is represented as a gray-scale value in
a matrix making up the picture. Two such matrices F and G of the undeformed
(x, y) and deformed (x∗, y∗) configuration are shown in Figure 10.

Figure 10: Conceptual representation of a DIC analysis. Bitmap images (left) and
displacement field (right).

The cross correlation coefficient (rij ) describes how the two matrices are re-
lated around a specific pixel. If perfectly correlated, the coefficient will have the
maximum value of 1.

Σi Σj [F (xi , yj ) − F̄ ][G(x∗i , y∗j ) − Ḡ]

rij = 1 − p
Σi Σj [F (xi , yj ) − F̄ ]2 Σi Σj [G(x∗i , y∗j ) − Ḡ]2
A change in position can be described by linear deformation theory with dis-
placements u and v:
du du
x∗ = x + u + ∆x + ∆y
dx dy
dv dv
y∗ = y + v + ∆y + ∆x
dy dx
Varius search routines have been proposed to find displacements (u,v) which
maximizes the correlation function (r). By discretization of the displacement field
into subset of pixels and by using Newton-Raphson iterations Sutton et al. [19]
managed to create search algorithms which were at least 20-times faster than
available in 1986.

13
3 Design of joint to Eurocode 3
The bolted beam-column joint studied in the experimental test program (Figure
3) has been analyzed using methods established by Eurocode. Moment capacity
and rotational stiffness are calculated according to the Component method in NS-
EN-1993-1-8 [3]. See Appendix A and B for complete calculations.

3.1 Capacity
The connection depicted in Figure 11 is idealized by a set of components, which are
designed to transfer the external forces. Each component is checked for sufficient
capacity for forces acting at the periphery of the column web panel. To allow for
hand calculations, an equivalent T-stub model is used in Eurocode as described in
Section 2.1. Material- and load coefficients are set to 1.0, and material parameters
such as yield stress (fy ) and ultimate tensile strength (fu ) are obtained from tensile
tests.

Figure 11: Joint geometry.

The center of compression is set to the top flange, while bolt row 1 is the
only row considered active in tension. Tensile forces in bolt row 2 are neglected
because of a relatively small lever arm. Compressive flange force and tension forces
in row 1 are therefore carrying the external moment MEd . Bolt row 3 is designed
to carry the shear force VEd .

14
 Yielding of the end-plate and bolt failure in row 1 is calculated to be the crit-
ical failure mechanism with moment resistance Mj,Rd = 31.2 kNm. A description
of the three most critical failure mechanisms found are given in Table 1. The shear
capacity VRd is found to be 182 kN.

Table 1: Critical failure mechanisms found by the Component method

Failure mechanism Utilization

Bolt failure and yielding

of end-plate. Non-circular 1.00
pattern with an effective T-
stub length of 310 mm.

Bolt failure. Ultimate ten-

0.92
sile capacity reached for
bolts in row 1 .

Bolt failure and yielding of

column flange. Non-circular 0.82
pattern with an effective T-
stub length of 198 mm.

15
3.2 Stiffness
Rotational stiffness
A simplified stiffness model has been created as shown in Figure 12, where the
structural components are represented as linear springs coupled in series.

Figure 12: Stiffness model used in the Component method.

The stiffness contribution from each component is based on elastic deformation

of specific zones in the connection. Simple equations have been established in
Eurocode, where the elastic modulus, cross sectional area and geometry determines
the stiffness value. The results are listed in Table 2.

Table 2: Stiffness of components

Spring Component Stiffness [k · 10−6 N/mm]

k1 Stiffened column web in compression ∞
k2 Column web in tension 1.78
k3 Column flange in bending 5.11
k4 End-plate in bending 1.20
k5 Bolt row 1 in tension 1.35

Some calculations are needed to establish a moment-rotation relationship. The

following equations are derived from Eurocode. Note that a different notation has
been used.

16
The equivalent stiffness (keq ) for the tensile springs coupled in series is found by:
1
keq = 1 1 1 1
k2
+ k3
+ k4
+ k5

The initial rotational stiffness (kφ,ini ) can be established by considering the

lever arm (z) in Figure 12:
kφ,ini = keq z 2
To account for a reduction in stiffness due to plasticity, the initial stiffness is
decreased gradually until the moment capacity is reached:

MEd ≤ 32 Mj,Rd

 kφ,ini
kφ =
MEd −2.7 2
 k
φ,ini (1.5 Mj,Rd ) 3
Mj,Rd < MEd ≤ Mj,Rd

Column displacement
For an easier comparison of results to experimental tests, it is convenient to estab-
lish a relationship between the column force (P ) and the displacement (∆) based
on the rotational spring stiffness (kφ ).
The column force is calculated based on the length from the support to the
periphery of the column web panel (Lc ):
2MEd
P =
Lc
The rotational spring has been chosen to act in the center of the connection as
depicted in Figure 13. Furthermore, the beam is included in the stiffness model
to account for the additional flexibility due to beam deflection.

Figure 13: Static system showing one side of the two-sided connection.

17
 The total displacement (∆tot ) is calculated according to linear beam theory.
Shear deformations are neglected.

P L2φ P L3b
∆tot = ∆φ + ∆b = + (12)
2kφ 6EIb

The force-displacement relationship obtained is shown in Figure 14, where

beam span is identical to the experimental test setup described in Section 5.1.

100

80
Force, P [kN]

60

40

EIb = ∞
20
Corrected for beam deflection

0
0 5 10 15 20 25
Displacement, ∆ [mm]
Figure 14: Force - displacement based on Eurocode’s stiffness model.

An additional column displacement of 1 mm at maximum load is added due

to beam deflection, softening the response slightly. The maximum load carrying
capacity (P ) is calculated to be 87 kN.

18
4 Material tests
A material test program was initiated to determine the mechanical properties of all
the components in the connection studied in the experimental test program (Figure
3). The goal was to calibrate the necessary parameters in the rate-dependent
plasticity model presented in Section 2.2.

4.1 Experimental program

Tensile tests under quasi-static conditions were performed for flange, web, plates
and bolts.
Tests at increased strain rates were conducted on a selected set of the compo-
nents to capture the rate-dependent behavior of the structural steel (S355) and
bolt material (M16-8.8). Tests at low and medium strain rates were conducted
in a servo hydraulic test machine, capable of imposing strains at 0.1 s−1 . The
Split-Hopkinson tension bar test [20] was performed for specimens at high strain
rate at approximately 100 s−1 .
Replicate tests were performed for each specimen for validation purposes.

4.2 Quasi-static tensile tests

Test coupons
Rectangular coupons were taken from web, flange and plate material in longitu-
dinal and transverse directions, while circular coupons were lathed from bolts as
illustrated in Figure 15.

Figure 15: Test coupons used in the material test program.

19
 All the test specimens were carefully measured by a digital apparatus before
testing. Maximum deviation from nominal geometry was found to be 0.5 mm.

Loading apparatus and instrumentation

Two high resolution cameras were aligned perpendicular to the width and thickness
surface for the rectangular coupons as shown in Figure 16. The cameras were
calibrated to track movement in three dimensions for the circular coupons.

Figure 16: Quasi-static tensile test setup.

The specimens were pulled in tension by the actuator at a constant speed

corresponding to a strain rate of 10−3 s−1 . Reaction force (F ) was synchronized
with pictures taken by the cameras and collected at a frequency of 1 Hz.

Digital image correlation (DIC)

DIC analysis was performed to determine the width (w), thickness (t) and longi-
tudinal displacement (∆L) through the use of eCorr V3.0 [21].
Cross sectional area was simply calculated as A = w · t for rectangular test
specimens. Positional tracking in three dimensions was performed on the circular
specimens as shown in Figure 17. A least square fit to a 3D circle in the neck region
could therefore be used to find the cross sectional area for the circular specimens.

20
Figure 17: DIC analysis of circular coupon. Bitmap image (left) and 3D mesh (right).

4.3 Work hardening

As discussed in Section 2.3, simple finite stress-strain relations can not be obtained
from tensile test during necking in a specimen. Inverse modeling was therefore used
to quantify the isotropic work hardening function (R) defined in Eq. (5).

Self Consistent Method

Before neck initiation, equivalent stress and strain were calculated from the test
data using Eq. (9a) and Eq. (9b). During necking, average axial stress and strain
were obtained from Eq. (10a) and Eq. (10b). Finally, the triaxial stress correction
function (ζ) defined in Eq. (11) was found by a self consistent trial and error
method established by Yazzie et al. [16].
The basic principle behind the Self Consistent Method is the use of finite el-
ement simulations to iteratively seek the stress correction function which best fit
the experimental data. A simple flow chart can be seen in Figure 18.

21
 Figure 18: Flow chart of the Self Consistent Method.

Two symmetry planes were utilized in modeling of the test coupons, and a very
small initial imperfection was added to initiate necking. A typical finite element
model showing the mesh density used in calibration is shown in Figure 19.

Figure 19: Typical finite element model used in the Self Consistent Method.

22
 Three iterations were usually sufficient for each test coupon and gave the ma-
terial curves shown in Figure 20. A minor scatter in the structural steel material
(S355) is observed between web, flange and plate specimens.

1400
True stress, σeq [MPa]
1200

1000

800

600 M16-8.8
S355

400
0 0.5 1
Plastic strain, peq [-]
Figure 20: Work hardening curves obtained by the Self Consistent Method.

Validation
In order to assess the Self Consistent Method’s ability to simulate the material
behavior, the tensile test simulations are compared to the experimental tests in
Figure 21.

100 100
Force, F [kN]
Force, F [kN]

80 80

60 60
FEA FEA
40 Bolt 40 Bolt
Plate Plate
20 Flange 20 Flange
Web Web
0 0
0 1 2√ 0 10 20
Cross section reduction, ∆A [mm] Displacement, ∆L [mm]

Figure 21: Self Consistent Method compared to tensile test results.

23
 A good fit is observed until neck initiation (maximum load), after which longitu-
dinal displacement starts to deviate from the experiment. However, the reduction
in cross section is simulated accurately until fracture.
This behavior can be explained by considering the underlying principle of the
Self Consistent Method, which is the calibration of average stress (σavg = F/A),
and not the engineering stress (s = F/A0 ).
According to studies done by Khoo et al. [22], cross sectional reduction is a more
accurate comparison approach. It is therefore concluded that the Self Consistent
Method gives satisfactory results.

Mesh dependence
Mesh dependence in the finite element models has been studied by varying the
element size in the dog-bone specimen shown in Figure 19.

900
1
True stress, σeq [MPa]
Stress correction, ζ [-]

800
0.9 700

0.8 600
2 mm 2 mm
1 mm 500 1 mm
0.7 0.5 mm 0.5 mm
400
0.6
0 0.5 1 0 0.5 1
Plastic strain, p [-] Plastic strain, peq [-]

Figure 22: Effect of changing the element size on a rectangular dog-bone specimen.

As Figure 22 indicates, the hardening curves obtained show a strong mesh de-
pendence. Mesh refinement does not seem to give a converging solution; decreasing
the element size will increase the equivalent stress. Mesh size dependence is there-
fore expected in the use of the hardening curves in a finite element simulation.

4.4 Rate sensitivity

Additional tension tests were performed at four different strain rates in order to
determine the viscous material behavior. The flow stress was found to increase as
shown in Figure 23.

24
 1150 800
1100

True stress, σ [MPa]

True stress, σ [MPa]

700
1050
1000 600

950 102 500 102

10−1 10−1
900 10−2 10−2
10−3 400 10−3
850
10−4 10−4
800 300
0 0.02 0.04 0.06 0 0.05 0.1
Plastic strain, p [-] Plastic strain, p [-]

Figure 23: Stress at elevated plastic strain rates for M16-8.8 (left) and S355 (right).

Calibration method
The viscous exponent (c) determining the rate-dependent yield behavior in the
constitutive model was calibrated from the test data. By considering the viscous
stress defined in Eq. (6):
ṗ c
σeq = (σ0 + R)(1 + )
ṗ0
and taking the logarithm:
σeq ṗ
log [ ] = c · log [1 + ]
(σ0 + R) ṗ0
reveals that the viscous exponent can be found as the slope of the curve in a log-log
plot shown in Figure 24. A reference strain rate ṗ0 = 0.01 s−1 was used.

0.2 0.4

0.15 y=0.01362x-0.004656 0.3 y = 0.02567x + 0.01147

]

]
log [ (σ0 +R)

log [ (σ0 +R)

0.1
eq

eq

0.2
σ

0.05
0.1
0
0
−0.05
0 5 10 0 5 10
ṗ ṗ
log [1 + ṗ 0 ]
log [1 + ṗ 0 ]

Figure 24: Least square fit of the viscous exponent for M16-8.8 (left) and S355 (right).

25
4.5 Fracture
The Cockcroft and Latham fracture criterion (Wc ) defined in Eq. (7) was calibrated
from tensile tests performed under quasi-static conditions. Fracture energy was
found not to be strongly dependent on strain rate for low and medium strain rates
below 0.1 s−1 .

Calibration method
The total energy absorbed by the specimen is given as:
Z uf
E= F · du
0

where uf is the axial displacement at fracture. Fracture was defined when the
energy obtained by finite element simulation matched the experiment:

E F EA = E exp
The critical strain energy per unit volume was then found by integrating prin-
cipal stress and strain at integration points for all the elements in the critical
section:

Zεf
Wc = max(σ1 , 0)dε1
0

As illustrated in Figure 25, the principal stress is greater for elements at the
center of the specimen due to triaxiality. The fracture criterion (Wc ) will therefore
vary over the thickness of the specimen.

Figure 25: Principal stress for elements over the thickness in critical section.

26
 In addition, the fracture strain (εf ) will be strongly dependent on the element
size; a finer mesh will increase the local straining of an element. Due to such
element size dependencies, the same mesh density as used in simulations (Section
6) was used in the calibration.
The results can be seen in Table 3. Note that the scatter is because of the
triaxial stress state in the neck region.

Table 3: Scatter in Cockcroft and Latham fracture criterion

Wc [Nmm−2 ]
M16-8.8 713 - 954
S355 554 - 611

27
5 Experimental tests of joints
Experimental tests were carried out to investigate the behavior of the joint shown
in Figure 3 in a column removal scenario.

Load scenarios
Monotonic displacement of the column due to loss of vertical support is the basis
of the quasi-static loading conditions and serves as a reference for the dynamic
experiment.
The scenario behind the dynamic experiment is a sudden impact by a falling
object, with kinetic energy corresponding to a one story drop of a 400 kg mass.

Specimens
Quasi-static and dynamic experiments have been conducted on the specimen shown
in Figure 26.

Figure 26: Experimental test specimen.

A torque wrench was used to tighten the bolts with a moment of 80 Nm to

ensure contact between the end-plate and the column flange.
The geometry of the assembled connections was measured by a tape measure.
Maximum horizontal deviation between end-plate and column flange was found to
be 3 mm. The thickness of the end-plate deviated from nominal values by 1 mm.

28
5.1 Quasi-static experiment setup
Setup and loading
The test specimen was placed in an upside-down position under a portal frame
structure depicted in Figure 27. For additional pictures, see Figure C.56.
Steel angles with a circular support surface were placed loosely between the
portal beam and the specimen. The beam span (Ls ) defined in Figure 26 was
measured to be 685 mm.
A 1000 kN hydraulic actuator pulled the column end upwards at a constant
speed of 0.05 mm/s. The experiment was stopped after failure was observed on
one side of the connection.

Figure 27: Setup for quasi-static experiment.

Instrumentation
A linear variable differential transformer (LVDT) was used to measure the dis-
placement of the column end relative to the floor. Additional LVDT’s were placed
under each support in order to correct for any deflections in the portal beams.
Strain-gauges placed onto the column web and the beam flanges were primarily
used to assess the degree of symmetry during loading.
Two cameras were calibrated to perform three dimensional position tracking of
the central part of the connection using DIC analysis.

5.2 Dynamic experiment setup

Setup and loading
A pendulum impactor (kicking machine) was used in the dynamic crash tests [23].

29
It consists of an arm that swings around a bearing, which is accelerated by an
hydraulic piston at one end and attached to a trolley at the other end. The
moving trolley is then guided on rails into a direct impact with the specimen.
The specimen was placed in a vertical position next to an extended support
structure, which was bolted to a concrete reaction wall as shown in Figure 28.
Additional pictures of the setup are shown in Figure C.57. Steel forks were used
to restrict any lateral beam movements. A 20 mm thick steel plate was taped to
the column end in order to distribute impact forces.
The horizontal offset of the impactor relative to the centerline of the column
was 7 mm. The beam span (Ls ) defined in Figure 26 was measured to be 687 mm.

Figure 28: Setup for dynamic experiment.

The trolley with a total mass of 726.7 kg was accelerated on the rail system up
to a speed of 6.02 m/s by the kicking machine.

Instrumentation
A load cell on the trolley was used to measure the impact forces. A second load
cell was recording the reaction forces at the upper support.
A laser device was continuously measuring the distance of the trolley relative
to the reaction wall during the experiment. Velocity and acceleration were calcu-
lated by numerical differentiation, and allowed for a validation of the load cell by
multiplying acceleration by the mass of the trolley.

30
 One camera was used to capture local displacement in the central region of the
joint using DIC. A second camera was capturing a wider view of the specimen,
including the extended support structure.

5.3 Quasi-static results

General observations and failure mode
Two identical experiments were conducted for validation purposes. Experiment 1
failed due to thread stripping in bolt row 1 . A second nut was therefore added
in experiment 2 to allow for full utilization of the net section capacity.
A flexural behavior with bending of the end-plate and elongation of the bolts
in row 1 lead to failure as shown in Figure 29. A simple visual inspection reveals
that the plate opening remained closed in the position of bolt row 2 .

End-plate opening at time of fracture in Thread stripping (left) and tensile frac-
experiment 2. ture (right).

Figure 29: Pictures of quasi-static tests.

Some plastic deformation of the end-plate was observed after unloading, while
the column flange remained in the elastic range.
In retrospect, it is questioned whether the steel angles with the circular support
surface (see Figure 27) could have moved outwards during the experiment due to

31
friction, lengthening the beam span. Unfortunately, this was not actively observed
during the experiment or measured afterwards, leaving the issue unresolved.

Displacement measurements
The vertical column displacement is presented in Figure 30. The initial stiffness is
similar for the two experiments, with a gradual reduction due to plastification of
the end-plate and the bolts. Maximum column load (P ) measured by the actuator
was 138 kN.

150

Thread stripping
Force, P [kN]

100

Movement in test rig

50
Experiment 1
Experiment 2

0
0 10 20 30
Displacement, ∆ [mm]
Figure 30: Column force and displacement in quasi-static experiments.

5.4 Dynamic results

General observations and failure mode
Unfortunately, a replicate test had not been performed at the time of writing this
thesis. It is therefore difficult to assess the degree of randomness in the obtained
results.
The trolley impacted and caused a flexural behavior with an opening of the
end-plate which ultimately lead to tensile fracture in bolt row 1 as shown in
Figure 31.
Careful observations of the high speed camera footage revealed that the sup-
port structure moved upwards due to friction between the support and the beam
specimen, which was estimated to be approximately 10 mm.

32
 End-plate opening at time of fracture. Tensile fracture.

Support structure at time of fracture.

Figure 31: Pictures of dynamic test.

Force measurements
A closer look at the force measurements reveals that the specimen was hit by the
impactor in a series of elastic collisions as shown in Figure 32. Each collision lasted
about 1 ms and transfered short impulse forces of 1200 kN, which is approximately
10 times that of the quasi-static experiment.
After the initial collision, a 2 ms delay is observed before any reaction was
registered at the support. Afterwards, the beam started to oscillate with a period

33
of about 1 ms.

1200 Trolley
Support reaction
1000
Force [kN]

800
600
400
200

0 1 2 3 4 5 6 7 8
Time [ms]

Figure 32: Force measurements in dynamic experiment showing a 2 ms delay from

initial impact to support reaction.

Comparing quasi-static and dynamic results

The general response characteristics did not change in the dynamic experiment.
A flexural behavior and tensile fracture of bolts were compatible to findings under
quasi-static conditions.
However, using DIC analysis to measure the opening of the end-plate at the
position of bolt row 1 reveals some minor differences. A 10 % reduction in the
plate opening at time of fracture in the dynamic experiment can be seen in Figure
33.
Plate opening, δ [mm]

5
Quasi static
4 Dynamic
3

0
0 10 20 30 40
Displacement, ∆ [mm]
Figure 33: Opening between end-plate and column flange at position of bolt row 1.

34
6 Finite element simulations
Three dimensional explicit analyses have been performed in Abaqus V6.12 [24] for
both quasi-static and dynamic simulations of the experimental tests discussed in
Section 5. The constitutive model was implemented through SIMLab Metal Model
[25].
In Section 6.1, the finite element model is presented. A great deal of assump-
tions were made in establishing the model. Therefore, a parametric study was
carried out in Section 6.2 to investigate the effect of changing important variables.
Finally, the results are given in Section 6.3 - 6.4.

6.1 Finite element model

Geometry
The nominal geometry has been used for all the parts in the assembly shown in
Figure 34. Local buckling of column web in compression is assumed to be restricted
by the stiffener, and two symmetry planes were utilized in order to reduce the
computing time. No initial imperfections in the geometry were used.

Figure 34: Model geometry showing quarter symmetry, rigid support and trolley with
impactor (top).

35
 The trolley has been modeled as a rigid plate with a point mass, while the im-
pactor is modeled as a deformable solid. The mass of the trolley and the impactor
correspond to one quarter of the total measured mass of 726.7 kg.
The bolt head, shank and nut have been modeled as one solid part as shown in
Figure 37. The threaded region of the shank is idealized as circular with diameter
corresponding to the cross sectional area As = 157 mm2 .

Mesh
Linear brick elements with reduced integration (S4R) were used throughout. De-
fault hourglass control was used in Abaqus for the quasi-static simulation, while
viscous hourglass control was added in the dynamic simulation, which is recom-
mended for high impact analysis [26].
A medium dense mesh was chosen after a sensitivity study covered in Section
6.2. The mesh, pictured in Figure 35, is refined around a central region of the con-
nection for a more accurate representation of the stress field where high gradients
are expected. A finer mesh was also applied to the impact region in the model
used in the dynamic simulations.

Figure 35: Mesh used in quasi-static- and dynamic simulations (left) and dynamic
simulations (right).

36
Material
Hardening curves were found by an inverse modeling technique as described in
Section 4.3, and can be seen in Figure 36. The weld material is obtained by
dividing the yield stress of the plate material by a factor β = 0.9 according to
NS-EN-1993 [27]. Scatter in the yield stress of the strucural steel material (S355)
is accounted for by applying separate hardening curves to flange, web and plates.

1400
True stress, σeq [MPa]

1200

1000

800
M16-8.8
600 Weld
S355
400
0 0.5 1
Plastic strain, peq [-]
Figure 36: Isotropic hardening used in the plasticity model.

Physical properties of steel assumed to be representative for all materials used

are given in Table 4. Additional material parameters describing the viscous stress
(c) and fracture (Wc ) are listed in Table 5. Note that an average value has been
chosen for the fracture criterion, which was found to vary over the thickness as
discussed in Section 4.5.
Table 4: Physical constants [27] Table 5: Material parameters

E [MPa] ν [-] ρ [kg m−3 ] ṗ0 [s−1 ] c [-] Wc [N mm−2 ]

210000 0.3 7800 M16-8.8 0.01 0.01362 834
Weld 0.01 0.02567 648
S355 0.01 0.02567 583

Interactions
The geometry is divided into sub-assemblies using tie-constraints as shown in Fig-
ure 37. A general contact formulation with finite sliding is used in Abaqus for
interaction between them. The friction coefficient is taken as µ = 0.2 for un-
treated steel surface according to NS-EN-1090 [28].

37
 Beam with welded end-plate.

Column with stiffener.

Force distributor (top).

Bolt. Trolley (rigid plate) with

Rigid support.
impactor.

Figure 37: Sub-assemblies used in the finite element model. Trolley with impactor and
force distributor are not used in quasi-static simulations.

38
 Special attention has been payed to the contact conditions at the support. A
sensitivity study described in Section 6.2 revealed that spurious reaction forces
may develop. Contact without friction has therefore been chosen between the
support and the beam flange.

Boundary conditions
Quarter symmetry (see Figure 34) is obtained by xz- and yz symmetry planes,
which restricts movement in y and x direction respectively. A fixed boundary
condition is placed on the support, while the trolley traveling on rails is restricted
from moving in x- and y direction.

Loading conditions
A constant velocity of 0.5 m/s was applied to the column end in the quasi-static
simulation. The velocity was applied smoothly to reduce oscillations that occur
when motion is introduced suddenly in an explicit analysis. The initial velocity of
the trolley was set to the measured value of 6.02 m/s in the dynamic simulation.
The applied velocities are illustrated by the arrows in Figure 38

Figure 38: Velocity conditions used in quasi-static simulation (left) and dynamic sim-
ulation (right).

Special attention needs to be given to explicit analysis under quasi-static con-

ditions. A real time scale is generally not possible due to very small time steps
needed in explicit time integration. Examinations revealed that the kinetic en-
ergy was below 1 % of the internal energy, indicating that conditions were in fact
quasi-static.

39
6.2 Sensitivity
A series of simulations have been conducted in order to discover sensitivities in the
finite element model. The focus has been on a few parameters that are considered
to have a relatively large impact on the results.

Element size
Hardening curves are obtained by the use of an inverse modeling technique as
described in Section 4.3. A very fine mesh was used in material calibration, and
element size dependence in the global analyses was therefore expected.
A coarse, medium and fine mesh was created as seen in Figure 39.

Coarse mesh Medium mesh Fine mesh

(10 000 elements) (25 000 elements). (50 000 elements).

Figure 39: Different element mesh used in sensitivity simulations.

Both quasi-static and dynamic simulations were performed. The results can be
seen in Figure 40. A medium dense mesh appears to provide a converging solution.
Velocity of trolley [m/s]

150 6 Fine mesh

Medium mesh
Force, P [kN]

5.5 Coarse mesh

100
5
Fine mesh
50 Medium mesh 4.5
Coarse mesh

0 4
0 20 40 0 2 4 6 8
Displacement, ∆ [mm] Time [ms]

Figure 40: Mesh convergence of quasi-static (left) and dynamic simulation (right)

40
Prestress of bolts
The importance of applying an initial stress state to simulate the tightening of the
bolts to a moment of 80 Nm in the experimental tests has been investigated.
The axial bolt load is calculated according to NS-EN-1090 [28]:
M 80Nm
=
F = = 40kN
k·d 0.13 · 16mm
where k is assumed to be the mean value of the valid range of 0.10−0.16 according
to NS-EN-14399 [29].
Quasi static simulations have been conducted for an assembly with and without
prestressed bolts. The results can be seen in Figure 41.

150
Force, P [kN]

100

Pre-stressed
50 Not pre-stressed

0
0 10 20 30
Displacement, ∆ [mm]
Figure 41: With and without prestressed bolts (40 kN).

The assembly with prestressed bolts shows a slightly greater initial stiffness
without any change in maximum load, in effect translating the force-displacement
curve horizontally. It is therefore concluded that the model is not heavily depen-
dent on the prestress.

Beam span
The initial length to the support was measured carefully before the experimental
test, but was observed to be lengthening (aproximately 10 mm) due to frictional
forces between the specimen and the support as discussed in Section 5.4. Therefore
it is interesting to study the effect of changing the beam span in the finite element
model.

41
 Simulations have been conducted under quasi-static conditions with a shorten-
ing and lengthening of the distance to support by 15 mm. The results can be seen
in Figure 42.

Force, P [kN] 150

100

Shortened support
50 Reference
Extended support

0
0 10 20 30
Displacement, ∆ [mm]
Figure 42: Length to support (685 ± 15 mm).

An increase of 15 mm correspond to a 2 % change in maximum force and a

softer response. The opposite is observed when the distance is shortened. This
behavior can be explained by considering the cross sectional moment:
P
M= · (Ls ± ∆Ls )
2
An increase in the distance to the support will reduce the load:
2M
P =
Ls ± ∆Ls
P Ls 685
= = = 1.00 ± 0.02
P0 Ls ± ∆Ls 685 ± 15
The 2 % change can therefore be explained by considering the cross sectional
moment. The uncertainty in force can be as much as 1-2 % when compared to the
experimental results.

Friction
A change in the friction coefficient between the steel surfaces in the central part
of the connection µjoint and between the beam flange and the support µsupport has
been investigated. The results are shown in Figure 43.

42
 150 150
Force, P [kN]

Force, P [kN]
100 100

µjoint = 0.1 µsupport = 0(ref)

50 µjoint = 0.2(ref) 50 µsupport = 0.2
µjoint = 0.4 µsupport = 0.4

0 0
0 10 20 30 0 10 20 30
Displacement, ∆ [mm] Displacement, ∆ [mm]

Figure 43: Effects of changing the friction coefficient in the central region of joint (left)
and at the support (right).

Increasing the friction coefficient for central parts of the connection gives a
slight increase in the initial stiffness and vice versa. The maximum load remains
unaltered.
Increasing the friction coefficient at the support however will significantly
change both stiffness and maximum load. Very large horizontal reaction forces
and compressive axial stress in the beam may develop in the simulation. In the
experimental setup, this was not the case because the support was not sufficiently
fixed (see Section 5.1).
It is therefore concluded that the finite element model is sensitive to the value
of friction coefficient at the support.

6.3 Quasi-static results

Response and failure mode
Bolt row 1 failed in tension due to an excessive opening of the end-plate. The
maximum vertical load was found to be 144 kN.
The stress state at time of failure shown in Figure 44 reveals which of the
components that are active in the transfer of forces. The bending moment is
primarily carried by bolt row 1 in tension, with row 2 not utilized beyond initial
yield. The shear force is carried by bolt row 1 and 2 . Bolt row 3 seems not to
be part of any significant load transfer.

43
 Axial stress [MPa] Shear stress [MPa].

Figure 44: Smoothed stress field at time of failure in quasi-static simulation. Stress
component Sij is defined according to coordinate system in Figure 34.

Compared to experimental tests

The flexural behavior and the gradual plastification of the end-plate and bolts are
in compliance with the experimental findings discussed in Section 5.3.
A direct comparison is obtained by considering the force-displacement curves
in Figure 45. The oscillation around the static solution is because of stress waves
generated due to the suddenly applied motion in the explicit analysis. Increasing
the total analysis time will reduce the significance of such oscillations at the cost
of an increase in computing time. However, this is not a major issue because the
oscillations are damped out when maximum load is reached.

150
Force, P [kN]

100

Experiment
50 Simulation

0
0 10 20 30 40
Displacement, ∆ [mm]
Figure 45: Column force obtained by quasi-static simulation

44
 A stiffer response is obtained in the simulation and the model overestimates
the maximum load by 4 %. The vertical displacement at failure is 15 % greater
than what was measured in the experiment.
A smaller opening of the end-plate also suggests a stiffer response obtained by
the finite element model as shown in Figure 46. Note that the deflection in the
portal frame structure used in the experimental setup is not considered in the DIC
analysis of the plate opening. Therefore, the maximum vertical displacement can
not be compared to the results in Figure 45.
Plate opening, δ [mm]

4 Experiment
Simulation
3

0 10 20 30 40
Displacement, ∆ [mm]
Figure 46: Opening of the end-plate in quasi-static simulation.

6.4 Dynamic results

Response and failure mode
The general stress state in the critical region at failure shown in Figure 47 is similar
to the quasi-static simulation shown in Figure 44. A larger part of the shear force
is carried by the bolts in row 3 , which were not active under quasi-static loading
conditions.
An approximate 8-9 % increase in the stress values is observed in the dynamic
simulation. Local plastic strain rates of the order of 10 per second correspond to a
10 % increase in the yield stress in the visco-plastic material model. The viscous
stress can therefore explain the slightly higher stress values.

45
 Axial stress [MPa] Shear stress [MPa].

Figure 47: Smoothed stress field at time of failure (9 ms after first impact). Stress
component Sij is defined according to coordinate system in Figure 34

Compared to experimental tests

The simulated force measurements are compared to the experimental results in
Figure 48. The peak impact force of approximately 1400 kN and the 2 ms delay
until reaction forces are captured at support are in compliance with experiment.
However, the response deviates over time. Two collisions were observed in the
experiment, while simulation shows three hits by the trolley before failure.
Force on trolley [kN]

1400 Experiment
1200 Simulation
1000
800
600
400
200
0 2 4 6 8 10
Time [ms]
Support reaction [kN]

1400 Experiment
1200 Simulation
1000
800
600
400
200
0 2 4 6 8 10
Time [ms]

Figure 48: Forces obtained by dynamic simulation.

A comparison of the opening of the end-plate is shown in Figure 49. As for

the quasi-static simulation, a smaller plate opening is found, further indicating an
overly stiff response obtained by the finite element method.

46
 Plate opening, δ [mm]
5
Experiment
4
Simulation
3

0 10 20 30 40
Displacement, ∆ [mm]
Figure 49: End-plate opening in dynamic simulation.

Response to an impulse load

It is interesting to further investigate how the connection responds to sudden
dynamic forces. An exaggerated scale factor has therefore been used in Figure 50
to display the response behavior during one impact.

0.5 ms after impact 1 ms after impact. 2 ms after impact.

Figure 50: Simulated response during a sudden impact. Deformation scale factor 15.

The initial hit is causing rigid body movement of the column. All three bolt
rows are therefore hit simultaneously and must transmit shear forces to inertial

47
resistance in the beams, which occur due to the sudden acceleration of the mass.
Shear waves require time to travel outwards to the support before reaction
forces can arise. In fact, 2 ms after impact, the support is hit by the beam. The
observed delay in both the simulation and the experimental test can therefore be
explained.
The stress state 0.5 ms after impact is pictured in Figure 51, showing the
uniform distribution of shear
√ forces to the bolts. Stress value of 396 MPa is 80 %
of yield stress (τyd = fy / 3).

Axial stress [MPa] Shear stress [MPa].

Figure 51: Smoothed stress field 0.5 ms after first impact by trolley. Stress component
Sij is defined according to coordinate system in Figure 34

48
7 Discussion
7.1 Experimental findings
A flexural behavior of the joint due to bending was observed under both quasi-
static and dynamic loading conditions. The experiments showed an initial linear
response with a gradual decrease in stiffness due to plastification of the end-plate
and bolts. Ultimately, the outermost bolt row failed in tension due to the excessive
opening of the end-plate. After unloading, some plastic deformation was observed
in the end-plate, while the column flange remained in the elastic region.
The inertial resistance that develops in the beams will undoubtedly affect the
response in a dynamic load scenario. Specifically, the shear forces will have a
relatively larger impact on the response compared to the bending moments as
discussed in Section 1.1. However, the short impulse loads on the column in the
dynamic crash test did not cause shear failure of bolts or welds, even though the
peak load of 1200 kN was tenfold that of the maximum load under quasi-static
conditions. Overall, the particular bolted end-plate connection investigated in the
experimental program was resistant to any brittle fracture that can occur when
the forces are applied suddenly.
On the other hand, the relatively small plate opening at time of failure and the
absence of any plastic deformations in the beam flange suggests a poor rotational
capacity. Even though this is somewhat outside the scope of this thesis, the joint
would likely have performed better if a reduction in the end-plate thickness or an
increase in the bolt diameter had been made. A large degree of plastic deformation
is desirable and is a key feature in the development of catenary action and therefore
the prevention of progressive collapse in a column removal scenario.

7.2 Assessment of the Component method (Eurocode 3)

Design rules given in current European standards have been followed to determine
the strength and the stiffness of the joint. A relatively complex model is used in
the Component method, where mechanical properties can be assigned individually
to components such as flange, web, end-plate and bolt.
Nominal values of the material properties are typically used in design. In
reality, the steel supplier is only concerned whether the minimum requirements are
met, therefore delivering steel grades that often exceeds the nominal values. All
the material parameters used were therefore taken from the tensile test program.
The force-displacement curve obtained is compared to the finite element sim-
ulations and the experimental results in Figure 52.

49
 150

Force, P [kN] 100

Experiment
50 EC3
Simulation

0
0 10 20 30 40
Displacement, ∆ [mm]
Figure 52: A comparison of the quasi-static results.

In despite of efforts to not make any conservative assumptions, the maximum

load capacity was greatly underestimated by a factor of 1.6. In real design, load-
and material coefficients would have been included, further increasing the safety.
The initial stiffness was slightly overestimated. However, there are some uncer-
tainties in the simplified stiffness model established in Figure 13.
To better understand the limitations in the Component method, a represen-
tation of the critical yield mechanism is compared to the simulated plastic strain
field in Figure 53.

Figure 53: Critical yield mechanism obtained from Component method (left) and sim-
ulated plastic strain field (right).

50
 Only one of the inner bolt rows (between the beam flanges) is considered in
the yield mechanism. In the "normal" design case, with an extended end-plate
in the tensile zone, two bolt rows are included. The performance of the joint in
this particular design configuration was investigated in the paralell Master’s thesis
by Baasen and Nordgaard [9]. They found a significant underestimation of the
capacity by a factor of 1.5 and a slightly stiffer initial response, in accordance with
findings in this thesis. Therefore, the load direction is not an important factor in
the Component method for this particular joint.
Another consideration is the yielding of the end-plate around the upper beam
flange. The simulated strain field in Figure 53 shows an additional resistance in
bending of the end-plate that is not part of the yield mechanism. This may be a
contributor to the underestimation of the capacity.
It is evident that some simplifications must be made for it to be feasible to
calculate by hand. Increasing the complexity of the method will not only lead to a
time consuming and expensive design process, but also increase the possibility of
making errors. Another important point is that the method must work in general
for all types of bolted and welded connections.
Although complete calculations of the rotational capacity have not been per-
formed, it turned out that the joint do not satisfy the basic requirements for rigid
plastic global analysis, preventing the designer from utilizing the increased ca-
pacity due to catenary action (see Appendix B). This is in compliance with the
experimental findings, where only limited plastic deformation was observed at time
of failure. The requirements in the code can be met either by increasing the bolt
diameter to 24 mm or by reducing the end-plate thickness to 8 mm.

7.3 Finite element predictions

Quasi static simulations
Finite element analyses gave a more complete understanding of the load transfer
through the joint until failure. The force-displacement curve in Figure 52 reveals
that the maximum load was overestimated by only 4 %, while the initial stiffness
was found to be slightly greater. The opening of the end-plate during loading
was underestimated, also suggesting a stiffer response obtained by finite element
model.
The deviations may partially be explained by the uncertainty in the beam span.
As shown in Section 6.2, an increase in the beam span of only 15 mm will reduce
the maximum force by 2 % and reduce the stiffness. The experimental setup
allowed for a lengthening of the distance to the support because of friction in the
tests as discussed in Section 5.3. In addition, the flexibility of the portal frame
structure itself may have contributed to horizontal displacement of the support.
Another possible source of error is the geometry used in the finite element

51
model. In particular, the thickness of the end-plate, which was measured to deviate
as much as 1 mm from nominal thickness of 12 mm.
Simplifications in the modeling of the bolt were made, a critical component in
the assembly. Thread interaction was not included, therefore preventing the ob-
served thread stripping fracture to be simulated. However, the Cockcroft Latham
criterion performed reasonably well in predicting tensile fracture of the bolts. A
smaller opening of the end-plate indicates that the value of the critical strain en-
ergy used in the constitutive model (Wc ) was too low. As discussed in Section 4.5,
Wc will vary over the thickness in the neck region due to triaxiality and an average
value was simply chosen.

Dynamic simulations
The general response characteristics were simulated in accordance with observa-
tions. A series of elastic collisions by the trolley and the ultimate flexural failure
were captured by the finite element model. As for the quasi static simulations, an
overly stiff response was found.
Furthermore, it was difficult to reproduce the response exactly. The additional
dimension of time adds to the complexity, and small variations can lead to a very
different outcome over time.
Plate opening, δ [mm]

Plate opening, δ [mm]

5 5
Quasi static Quasi static
4 Dynamic 4 Dynamic
3 3

2 2

1 1

0 0
0 10 20 30 40 0 10 20 30 40
Displacement, ∆ [mm] Displacement, ∆ [mm]

Figure 54: Opening between end-plate and column flange at position of bolt row 1.
Experiments (left) and simulations (right).

As Figure 54 indicates, a reduction in the end-plate opening in the dynamic

experiment was not predicted correctly. Contrary to the experimental findings, a
larger plate opening is obtained in the dynamic simulation at time of failure. This
was also found in the paralell Master’s thesis by Baasen and Nordgaard [9]. As
discussed in Section 6.4, the stress values are generally increased in the dynamic
simulation due to viscous effects. A closer look at the stress state in bolt row
1 at time of failure in Figure 55 further support this finding. The maximum
principal stress (used in Cockroft and Latham fracture criterion) is increased by

52
approximately 12 % in the dynamic simulation. Therefore, one would expect the
strain energy in Eq. (7) to increase by the same factor and hence reducing the
plate opening at time of failure. Unfortunately, the cause of this unexpected result
has not been found.

Figure 55: Maximum principal stress [MPa] in bolt row 1 at time of failure. Quasi-static
simulation (left) and dynamic simulation (right).

The change in the response characteristics due to inertial resistance in a dy-

namic impact was investigated in Section 6.4. The simulation revealed that very
high shear stress develops at the time of impact, and all three bolt rows are active
in the load transfer.
Additional simulations have been conducted to try to impose the expected
change from flexural to shear failure. An increase in the velocity of the trolley
from 6 m/s to 20 m/s did not cause shear failure of bolts or welds. A possible
explanation is that tensile separation is the underlying concept of the Cockcroft
and Latham fracture criterion used in the finite element model.

53
8 Conclusions and suggestions for further work
8.1 Concluding remarks
1. Vertical displacement of the column lead to failure in flexure as predicted by
Eurocode and the finite element models.

2. Thread stripping fracture was observed in one of the experiments, but was
not captured by any of the design methods.

3. As expected, the simplified methods given in design codes gave safe estimates
of the capacity, while numerical simulations allowed for a more accurate
description of the response characteristics and failure mode.

4. As previous research has shown, the initial stiffness was overestimated by

both Eurocode’s method and the finite element models.

5. The Component method in Eurocode 3 showed no limitations in predicting

the quasi-static response when the force direction was reversed. However, an
overly safe estimate of the capacity was obtained.

6. The numerical simulations revealed that the relative impact of shear forces
is increased when the column is subjected to rapid, non cyclic loading.

8.2 Future studies

Possible subjects for further work:

• Further investigations into what load scenario will cause the expected change
from flexural to shear failure when forces are applied suddenly. The intro-
duction of an additional failure criterion based on shear glide should be
implemented in the finite element model.

• Parametric study: Investigate the effect of catenary action in a structural

collapse scenario by numerical simulations.

• Parametric study: Optimization of bolted end-plate connections. Can in-

clude placement of bolts, thickness of end-plate, extended vs flush etc.

54
References
[1] European Committee for Standardization (CEN). Eurocode - 1990 - basis of
structural design, February 2008.

[2] A. Coelho, F. Bijlaard, and L. da Silva. Experimental assessment of the

ductility of extended end plate connections. Engineering Structures, 26:1185–
1206, 2004.

[3] European Committee for Standardization (CEN). Eurocode - 1993 - design

of steel structures - part 1-8: Design of joints, May 2005.

[4] B. Yang and K. H. Tan. Numerical analyses of steel beam-column joints

subjected to catenary action. Journal of Constructional Steel Research, 70:
1–11, 2011.

[5] B. Yang and K. H. Tan. Experimental test of different types of bolted steel
beam-column joints under a central-column-removal scenario. Engineering
Structures, 54:112–130, 2013.

[6] F.H. Sadek, H. S. Lew, J. Main, S.D Robert, and V. P. Chiarito. Performance
of steel moment connections under a column removal scenario. i: Experiments.
Journal of Structural Engineering, 139:98–107, 2013.

[7] F. Sadek, J. Main, H. Lew, and S. El-Tawil. Performance of steel moment con-
nections under a column removal scenario. ii: Analysis. Journal of Structural
Engineering, 139:108–119, 2013.

[8] T Subawala, D Linzell, and T Krauthammer. Finite element analysis of steel

beam to column connections subjected to blast loads. International Journal
of Impact Engineering, 31:861–876, 2004.

[9] Gjermund Baasen and Torger Nordgaard. Beam-column connections sub-

jected to static and dynamic loading. Master’s thesis, Norwegian University
of Science and Technology (NTNU), 2014.

[10] P. K. Larsen. Dimensjonering av staalkonstruksjoner (Design of steel struc-

tures). Tapir Akademiske Forlag, 2010.

[11] O.S. Hopperstad and T. Boervik. Lecture notes, TKT4135 Materials Me-
chanics. Norwegian University of Science and Technology (NTNU), spring
2013.

[12] F. Irgens. Continuum Mechanics. Tapir Akademiske Forlag, 2008.

55
[13] J. Lubliner. Plasticity Theory. Macmillan Pub Co, 1998.

[14] J. Lemaitre and J.-L. Chaboche. Mechanics of Solid Materials. Cambridge

University Press, 1994.

[15] M. G. Cockcroft and D. J. Latham. Ductility and the workability of metals.

Journal of the Institute of Metals, 96:33–39, 1968.

[16] K.E. Yazzie, H. Fei, H. Jiang, and N. Chawla. A self-consistent approach for
necking correction intensile specimens with rectangular cross-section using
a novel mirror fixture. Transactions A: Physical Metallurgy and Materials
Science, 43A:5058–5066, 2012.

[17] P. W. Bridgman. Studies in Large Plastic Flow and Fracture. McGraw-Hill,

1952.

[18] M. A. Sutton, S. R. McNeill, J. D. Helm, and Y. J. Chao. Advances in

two-dimensional and three-dimensional computer vision. Topics in Applied
Physics, 77:323–372, 2000.

[19] M. A. Sutton, M. Cheng, W. H. Peters, and Y. J. Chao. Application of an

optimized digital correlation method to planar deformation analysis. Image
and Vision Computing, 4:143–150, 1986.

[20] Y. Chen, A. H. Clausen, O.S. Hopperstad, and M. Langseth. Application of

a split-hopkinson tension bar in a mutual assessment of experimental tests
and numerical predictions. International Journal of Impact Engineering, 38:
824–836, 2011.

[21] E Fagerholt. ecorr v3.0 digital image correlation tool by simlab. NTNU, 2013.

[22] H.A. Khoo, M. Cheng, and T.M. Hrudey. Determine steel properties for large
strain from a standard tension test. In 2nd Material Specialty Conference of
the Canadian Society for Civil Engineering, 2002.

[23] A. G. Hanssen, T. Auestad, T. Tryland, and M. Langseth. The kicking ma-

chine: A device for impact testing of structural components. International
Journal of Crashworthiness, 8:385–392, 2003.

[24] Abaqus 6.12 User Documentation. Dassault Systemes, February 2012.

[25] SIMLab. Theory Manual: SIMLab Metal Model. Norwegian University of

Science and Technology (NTNU), 2014.

[26] Abaqus 6.12 Theory Manual. Dassault Systemes, February 2012.

56
[27] European Committee for Standardization (CEN). Eurocode - 1993 - design
of steel structures - part 1-1: General rules and rules for buildings, May 2005.

[28] European Committee for Standardization (CEN). Execution of steel struc-

tures and aluminium structures - part 1: Requirements for conformity assess-
ment of structural components, May 2012.

[29] European Committee for Standardization (CEN). High-strength structural

bolting assemblies for preloading - part 7: System hr - countersunk head bolt
and nut assemblies, May 2007.

57
A Capacity according to NS EN 1993-1-8
Two beams (HEA 180) are connected to a column (HEB 220) using an end-plate and 6 bolts (M16).

The capacity of the joint is checked for a point load P = 87 kN.

Point load: Partial factors: Elastic modulus:

P  87.2kN γM0  1.0 Es  210000MPa

LP  697mm γM1  1.0

γM2  1.0
 1. INPUT DATA

Beam data (HEA 180) Column data (HEB 220)

fy.b  442MPa fy.c  400MPa

b b  180mm b c  220mm

h b  171mm h c  220mm

tf.b  9.5mm tf.c  16mm

tw.b  6mm tw.c  9.5mm

rb  15mm rc  18mm
3 2 3 2
Ab  4.53 10 mm Ac  9.10 10 mm
3 3 3 3
Sy.b  162  10 mm Sy.c  414  10 mm

fu.b  554MPa fu.c  539MPa

( rolled section) sb  rb  15 mm sc  rc  18 mm

End plate Welds

Bolts
af  5mm
tp  12mm
d  16mm (bolt class 8.8)
aw  5mm
b p  220mm
d 0  18mm
βw  0.9
fu  562MPa
k 2  0.9
fy  412MPa
αv  0.6 (full threaded)
e1  30mm
fub  946MPa
e2  60mm
fyb  865MPa
2 p 1  70mm
As  157mm
p 2  100mm
h b.n  12mm (heigth of nut)
e5  35mm (from row 1 to upper flange)
h b.h  10mm (heigth of head)
p 5  91mm (from row 1 to row 2)
 2. BOLT DESIGN

Centre of compression in line with mid-thickness of the

Figure 6.15(c)
compression flange.

Bolt row 2 is located close to centre of compression and is

neglected in all calculations. Thus bolt row 1 are the only bolts
active in tension.

Bolt row 3 is designed to take shear force VEd .

SHEAR FORCE CHECK

Design shear force:

P
VEd   43.6 kN
2
Shear resistance:

αv  fub As
Fv.Rd   89 kN (per bolt)
γM2

For bolt row 3:

Fv.Rd'  2  Fv.Rd  178  kN

Bearing resistance:
Flange beween bolt row 3 and bolt row 2 prevents shear tearing
between bolts in end plate. Local failure mode assumed.

k 1  2.5

 fub 
αb  min 1  1
 fu 

 
t  min tf.c tp  12 mm

k 1  αb  fu  d  t
Fb.Rd   270  kN (per bolt)
γM2

For bolt row 3:

Fb.Rd'  2  Fb.Rd  540  kN

Shear capacity for bolt row 3:

  
FRd  min Fv.Rd' Fb.Rd'  178  kN

Shear bolt.req  "OK" if FRd  VEd  "OK"

"NOT OK" otherwise

TENSION CHECK

The moment, MEd , is carried by flange force, Nf. Lever arm, z,

Figure 6.15(c): is the distance between the compression flange to bolts in
tension (row 1).

z  e5  p 5  126  mm

The connection elements should be designed for the forces at

5.3 (2) and (4)
the periphery of the column web panel.
P
M Ed 
2
 
 LP  tf.c  31.1 kN m
M Ed
Nf.Ed   247  kN
z

Tension capacity:

k 2  fub As
Ft.Rd   134  kN (per bolt)
γM2

Tension capacity for bolt row 1:

Ft.Rd'  2  Ft.Rd  267  kN

Tensionbolt.req  "OK" if Ft.Rd'  Nf.Ed  "OK"

"NOT OK" otherwise

3. TENSION FLANGE CHECK

3 2
Af  b b  tf.b  1.71  10  mm
fy.b
fd   442  MPa
γM0

Yield capacity of beam flange:

Nfb.Rd  Af  fd  756  kN

Tensionflange.req  "OK" if Nfb.Rd  Nf.Ed  "OK"

"NOT OK" otherwise

 4. COMPRESSION FLANGE CHECK
Yield capacity of compressed flange is at least the same as
6.2.6.7
for tension flange.

See formula 6.21

5. CAPACITY OF WELDS

FLANGE WELD

Given, af  5  mm

Design throat area:

4.5.3.2.(2)  
leff.f  b b  2  tf.b  b b  2  rb  tw.b  343  mm

3 2
Aw.f  af  leff.f  1.715  10  mm

Design tension strength of the weld:

fu.b
fvw.fb.d   435  MPa
4.5.3.2.(6) 2  γM2 βw

Design resistance of flange weld in tension:

Fw.fb.Rd  Aw.f  fvw.fb.d  746  kN

Tensionweld.req  "OK" if Fw.fb.Rd  Nf.Ed  "OK"

"NOT OK" otherwise

WEB WELD

Given, af  5  mm

Design throat area:

4.5.3.2.(2)  
leff.w  h b  2  tf.b  2  rb  2  244  mm

3 2
Aw.w  af  leff.w  1.22  10  mm

Design shear strength of the weld:

4.5.3.2.(6)
fu.b
and fvw.wb.d   355  MPa
3  γM2 βw
4.5.3.3.(3)
 Design resistance of web weld in shear:

Fw.wb.Rd  Aw.w fvw.wb.d  434  kN

Shear weld.req  "OK" if Fw.wb.Rd  VEd  "OK"

"NOT OK" otherwise

6. CAPACITY OF COLUMN FLANGE IN BENDING

Bending capacity according to 6.2.6.4. An equivavalen T-stub

model with an effective length, leff, is used.

Geometry:
p2 tw.c
m'    0.8 rc  31 mm
Figure 6.8 2 2

emin  e2  60 mm

e'  emin  60 mm

Effective lengths for circular (cp) and non-circular (nc)

patterns:
leff.cp.fc  2  π m'  194  mm
Table 6.4
leff.nc.fc  4  m'  1.25 e'  198  mm

Design resistance of T-stub representing the column flange:

n''  emin if emin  1.25 m'  39 mm

( 1.25 m') otherwise

Failure mode 1  
leff.1.fc  min leff.cp.fc leff.nc.fc  194  mm
(complete
yielding of plate) 2 fy
M pl.1.fc.Rd  0.25 leff.1.fc tf.c   5.111  kN m
γM0
4  M pl.1.fc.Rd
FT.1.fc.Rd   663  kN
m'

Failure mode 2 leff.2.fc  leff.nc.fc  198  mm

(bolt failure and
yielding of plate) 2 fy
M pl.2.fc.Rd  0.25 leff.2.fc tf.c   5.23 kN m
γM0
 2  M pl.2.fc.Rd  n'' 2 Ft.Rd
FT.2.fc.Rd   299  kN
m'  n''

Failure mode 3
FT.3.fc.Rd  2  Ft.Rd  267  kN
(bolt failure)

Design resistance of column flange in bending:


FT.fc.Rd  min FT.1.fc.Rd FT.2.fc.Rd FT.3.fc.Rd  267  kN 

7. CAPACITY OF END-PLATE IN BENDING

Bending capacity according to 6.2.6.5. An equivavalen T-stub

model with an effective length, leff, is used.

Geometry:
p2 tw.b
m''    0.8 aw 2  41 mm
Figure 6.11 2 2
emin  60 mm

e'  60 mm
tf.b
m2  e5   0.8 aw 2  25 mm
2

Effective lengths for circular (cp) and non-circular (nc)

patterns:

Table 6.6 leff.cp.ep  2  π m''  260  mm

m''
λ1   0.41
m''  e'

m2
λ2   0.24
m''  e'

NOTE: "alpha"
α  7.5
coefficient read
manually from
figure 6.11. leff.nc.ep  α m''  310  mm

Design resistance of T-stub representing the end-plate:

n'''  emin if emin  1.25 m''  52 mm

( 1.25 m'') otherwise

Failure mode 1  
leff.1.ep  min leff.cp.ep leff.nc.ep  260  mm
(complete
yielding of plate) 2 fy
M pl.1.ep.Rd  0.25 leff.1.ep tp   3.85 kN m
γM0
 4  M pl.1.ep.Rd
FT.1.ep.Rd   373  kN
m''

Failure mode 2 leff.2.ep  leff.nc.ep  310  mm

(bolt failure and
yielding of plate) 2 fy
M pl.2.ep.Rd  0.25 leff.2.ep tp   4.6 kN m
γM0
2  M pl.2.ep.Rd  n''' 2 Ft.Rd
FT.2.ep.Rd   247  kN
m''  n'''

Failure mode 3
FT.3.ep.Rd  2  Ft.Rd  267  kN
(bolt failure)

Design resistance of end-plate in bending:

 
FT.ep.Rd  min FT.1.ep.Rd FT.2.ep.Rd FT.3.ep.Rd  247  kN

8. CAPACITY OF THE COLUMN WEB IN TENSION

6.2.6.3.(3) beff.t.wc is taken as the effective length of equivalent T-stub

representing the failure mechanism for column flange in
bending.

b eff.t.wc  leff.1.fc  194  mm

Symmetry in loading (Mb1,Ed =Mb2,Ed ):

β  0
Table 5.4
ω  1

Design resistance of column web in transverse tension:

ω b eff.t.wc tw.c fy.c

Ft.wc.Rd   737  kN
γM0

9. CAPACITY OF THE COLUMN WEB IN COMPRESSION

Stiffener between column flanges in compression prevents

local buckling and resistance is at least the same as for web
in tension.

10. CAPACITY OF THE BEAM WEB IN TENSION

6.2.6.8.(2) beff.t.wb is taken as the effective length of equivalent T-stub

 representing the failure mechanism for end-plate in bending.

b eff.t.wb  leff.1.ep  260  mm

Design resistance of beam web in tension:

b eff.t.wb tw.b fy.b

Ft.wb.Rd   689  kN
γM0

11. CAPACITY OF BEAM FLANGE AND WEB IN

COMPRESSION

Plastic moment resistance:

5 3
Wpl.b  2  Sy.b  3.24  10  mm

M pl.b.Rd  fy.b Wpl.b  143.2  kN m

Design resistance of beam flange and web in compression:

M pl.b.Rd
6.2.6.7 Fc.fb.Rd   887  kN
h b  tf.b
12. SUMMARY

Dimensioning forces P  87.2 kN VEd  43.6 kN M Ed  31.1 kN m Nf.Ed  247  kN

Shear capacities
VEd
Fv.Rd'  178  kN (Shear, bolt row 3)  0.24
Fv.Rd'

VEd
Fw.wb.Rd  434  kN (Shear, web weld)  0.1
Fw.wb.Rd

VEd
Fb.Rd'  540  kN (Bearing, end-plate)  0.08
Fb.Rd'

Tension/Compression capacities
Nf.Ed
Ft.Rd'  267  kN (Tension capacity without prying  0.92
effect, bolt row 1) Ft.Rd'

Nf.Ed
Ft.wb.Rd  689  kN (Beam web in tension)  0.36
Ft.wb.Rd

Nf.Ed
Nfb.Rd  756  kN (Yield capacity of beam flange)  0.33
Nfb.Rd

Nf.Ed
Fw.fb.Rd  746  kN (Flange weld in tension)  0.33
Fw.fb.Rd

Nf.Ed
Fc.fb.Rd  887  kN (Beam flange and web in  0.28
compression) Fc.fb.Rd

Capacity of end-plate in bending due to tension

Nf.Ed
FT.2.ep.Rd  247  kN (Bolt failure and yielding of end-plate )  1.00
FT.2.ep.Rd

Nf.Ed
FT.3.ep.Rd  267  kN (Bolt failure)  0.92
FT.3.ep.Rd

Nf.Ed
FT.1.ep.Rd  373  kN (Complete yielding of end-plate)  0.66
FT.1.ep.Rd
Capacity of column flange in bending due to tension

Nf.Ed
FT.3.fc.Rd  267  kN (Bolt failure)  0.92
FT.3.fc.Rd

Nf.Ed
FT.2.fc.Rd  299  kN (Bolt failure and yielding of flange )  0.82
FT.2.fc.Rd

Nf.Ed
FT.1.fc.Rd  663  kN (Complete yielding of flange)  0.37
FT.1.fc.Rd

Capacity of column web

Nf.Ed
Ft.wc.Rd  737  kN (Column web in tension)  0.33
Ft.wc.Rd

Design moment resistance of joint

(one bolt-row active in h 1  z  126  mm

tension)

(bolt failure and yielding  

Ft1.Rd  min FT.fc.Rd FT.ep.Rd  247  kN
of end-plate)

Design moment resistance:

M j.Rd  h 1  Ft1.Rd  31.2 kN m

6.2.7.2.(1)
 B Stiffness according to NS EN 1993-1-8
Two beams (HEA 180) are connected to a column (HEB 220) using an end-plate and 6 bolts (M16).
The stiffness properties of the joint are represented by a set of linear springs. One side of the double
sided joint are depicted below.

The rotational stiffness of the joint is established and the rotation capacity is checked.

1. ELASTIC STIFFNESS COEFFICIENTS

For two sided bolted joints with equal and opposite moments, the
Table 6.10
components to be taken into account are:

 k 2 column web in compression

 k 3 column web in tension
 k 4 column flange in bending
 k 5 end-plate in bending
 k 10 bolts in tension

COLUMN WEB IN COMPRESSION

Table 6.11 The stiffness contribution for a stiffened column web is neglected:

k2 = ∞

COLUMN WEB IN TENSION

Table 6.11 beff.t.wc is the effective width of column web in tension, and is
taken as the effective length of equivalent T-stub representing the
failure mechanism for column flange in bending.

6.2.6.3.(3) b eff.t.wc  194  mm

Shear length, column web:

 d cw  h c  2  tf.c  2  rc  152  mm

Column web thickness:

tw.c  9.5 mm

Stiffness coefficient for column web in tension:

0.7 b eff.t.wc tw.c

k 3   8.5 mm
d cw

COLUMN FLANGE IN BENDING

leff is to be taken as the smallest effective length used in

Table 6.11 equivavelent T-stub representing the column flange in bending.

Table 6.4  
leff.4  min leff.cp.fc leff.nc.fc  194  mm

Figure 6.8 m'  31 mm

Stiffness coefficient for column flange in bending:

3
0.9 leff.4 tf.c
k 4   24.3 mm
3
m'

END-PLATE IN BENDING

leff is to be taken as the smallest effective lengths used in

Table 6.11 equivavelent T-stub representing the end-plate in bending.

Table 6.6  
leff.5  min leff.cp.ep leff.nc.ep  260  mm

Figure 6.11 m''  41 mm

Stiffness coefficient for end-plate in bending:

3
0.9 leff.5 tp
k 5   5.7 mm
3
m''

BOLTS IN TENSION
 Only one bolt-row considered active in tension (row 1).

Lb taken as the bolt elongation length, taken as the grip thickness

Table 6.11 (material and washers), plus half the heigth of bolt head and nut.
Washers are not used in connection.

h b.h h b.n
Lb  tf.c  tp    39 mm
2 2

Stress area, bolt:

2
As  157  mm

Stiffness coefficient for bolts in tension:

As
k 10  1.6  6.4 mm
Lb

2. ROTATIONAL STIFFNESS

EQUIVALENT STIFFNESS IN TENSION

For case with only one bolt row in tension:

1
k eq   2.044  mm
6.3.3.1.(1) 1 1 1 1
  
k3 k4 k5 k 10

INITIAL ROTATIONAL STIFFNESS

Lever arm:
z  126  mm

Initial rotational stiffness:

2
Es z 3 m
Sj.ini   6.82  10  kN
6.3.1.(4) 1 rad
keq

ROTATIONAL STIFFNESS

Design moment:

M j.Ed  0 0.01kN m  60kN m

 Coefficient for bolted end-plate:

Table 6.8 ψ  2.7

Stiffness ratio:

2
6.3.1.(6) 
μ M j.Ed  1.0 if M j.Ed  M j.Rd
3
ψ
 Mj.Ed  2
 1.5 M  if  M j.Rd  M j.Ed  M j.Rd
3
 j.Rd 

Rotational stiffness:

2
Es z
6.3.1.(4)  
Sj M j.Ed 
1

μ M j.Ed 
k eq


Rotation (rad):

M j.Ed

ϕ M j.Ed  
Sj M j.Ed 
if M j.Ed  M j.Rd

100 otherwise

Rotational stiffness (EC3)

30
Moment [kNm]

Mj.Ed 20

1000

10

0
0 0.01 0.02

ϕ M j.Ed 
Rotation [rad]
 3. ROTATION CAPACITY

A joint can be modeled as a plastic hinge in the global analysis

when the rotation capacity is sufficient. For a bolted joint with
end-plate, both of the following conditions must be satisfied:

6.4.2.(2) a)  The moment capacity of the joint is governed by resistance in

column flange or end-plate.

The moment capacity is limited by failure of end-plate in

bending => OK

6.4.2.(2) b)  The thickness, t, of column flange or end-plate is limited by:

fub
t  0.36 d   9  mm
fy

Minimum plate thickness:

 
t'  min tf.c tp  12 mm (end-plate)

fub
Requirement  "OK" if t  0.36 d   "NOT OK"
fy

"NOT OK" otherwise

The joint do not satisfy the requirements for rigid plastic global
analysis. => Either increasing bolt diameter (M24) or reduce
end-plate thickness (8 mm).
C Additional pictures of the experimental tests

Portal frame structure and the 1000 kN hydraulic actuator

Steel angle with circular support. Fin plate component bolted to the col-
umn web.

Figure C.56: Pictures of quasi-static test setup.

 Kicking machine used to accelerate the trolley.

Trolley and the test specimen mounted Side view showing the extended support
to the reaction wall. structure.

Figure C.57: Pictures of dynamic test setup.