Sunteți pe pagina 1din 39

Comparison between the Eurocode methodology and a finite

element model: the equivalent T-stub method

Dieter Declerck

Supervisors: Anthony Tetaert, Prof. Jean-Francois Georgin (INSA Lyon)

Master's dissertation submitted in order to obtain the academic degree of


Master of Science in de industriële wetenschappen: bouwkunde

Department of Structural Engineering


Chair: Prof. dr. ir. Luc Taerwe
Faculty of Engineering and Architecture
Academic year 2016-2017
1 Preface

This research is submitted in order to obtain the academic degree of Master of Science in Civil
Engineering Technology at the university of Ghent, Belgium. With the help of an Erasmus+ exchange,
this research was executed at the national institute of applied science of Lyon, France. This is the second
paper in a long standing research to steel connections under supervision of Prof. Jean-François Georgin.
The first research was investigated by Thomas Härmälä and is a reference point for this paper.

I would like to thank Prof. Jean-François Georgin who entrust me with the subject and guided me
through the theoretical background of the subject. Further I would like to thank my parents who made
this Erasmus+ exchange possible.

1
2 Abstract

Bolted steel junctions are a popular way to connect beams and columns. To calculate the moment
resistance of junction, the fail resistance of different components of the connection needs to be
calculated according to the component method. An important component in a bolted steel junction
is the T-stub. This research discusses the analytical fail model of a T-stub with numerical results.
As analytical model, the Kennedy bolt force model is used which include the application of the
yield line theory. This theory is found less accurate but always provides safe values. Further, the
development of a complex FE-model has been carried out in order to research the equivalent T-
stub method. Since the FE-model didn’t behave like expected, the numerical and analytical values
where hard to compare and further development of the FE-model need to be carried out.

Keywords: EN 1993-1-8: 2005, equivalent T-stub method, T-stub, yield line, stub tee, bolted steel
connection, component method, Kennedy bolt force model, finite element method,

2
3 Table of Contents

1 Preface ............................................................................................................................................. 1

2 Abstract ........................................................................................................................................... 2

3 Table of Contents ............................................................................................................................ 3

4 List of abbreviations ........................................................................................................................ 4

5 Table of figures ............................................................................................................................... 5

6 Introduction ..................................................................................................................................... 7

7 Literature ......................................................................................................................................... 8

7.1 Moment resistance of a bolted steel junction .......................................................................... 8

7.2 The fail resistance of a T stub................................................................................................ 11

7.2.1 Yield line theory ............................................................................................................ 11

7.2.2 The Kennedy bolt force model ...................................................................................... 13

7.2.3 The equivalent T-stub method ....................................................................................... 18

8 Problem ......................................................................................................................................... 20

9 Determination of the research field & Methodology..................................................................... 20

10 Research & Results ................................................................................................................... 21

10.1 Yield line Theory................................................................................................................... 21

10.1.1 First FE-model ............................................................................................................... 22

10.1.2 Second FE-model .......................................................................................................... 25

10.2 The equivalent T-stub methodology ...................................................................................... 29

10.2.1 The effective length Leff................................................................................................. 29

10.2.2 Development of a FE- model ........................................................................................ 29

11 Conclusion ................................................................................................................................. 31

12 Annex ........................................................................................................................................ 32

12.1 Results of the parametric analyses of a 2D model................................................................. 32

12.2 Results of the parametric analyses of a 3D model................................................................. 33

12.3 Results of the parametric analyses of Leff .............................................................................. 34

13 References ................................................................................................................................. 38

3
4 List of abbreviations

tf: Thickness of the T-stub flange

tw: Thickness of the T-stub web

L: length of the T-stub

m: orthogonal distance between the web and the bolts

n: orthogonal distance between the flange edge and the bolts

a: distance between the bolts

Leff: effective length of a circular or non-circular T-stub

F: tensile force applied on the web

FRd: general tensile resistance of a T-stub

Fx,Rd: tensile resistance of a T-stub to fail mode x

Ft: tensile force in the bolts

Ft,Rd: tensile resistance of the bolts

Q: prying forces

4
5 Table of figures

Figure 1 Basic bolted steel connection .................................................................................................... 7

Figure 2 Moment rotation characteristic of a bolted junction ................................................................. 8

Figure 3: different categories used in EC3-1.8: A) simple B)semi continuous C)continuous ................ 9

Figure 4 Bolted steel portal subjected to an unified distributed load ...................................................... 9

Figure 5 Bolded beam-column connection with endplate ..................................................................... 10

Figure 6 T-stubs in a bolted column-beam connection ......................................................................... 10

Figure 7 dimensions of a T-stub ............................................................................................................ 11

Figure 8 fail mechanisms: left) bolt rupture, right) flange yielding ...................................................... 11

Figure 9 ideal elastic- plastic stress strain curve ................................................................................... 12

Figure 10 Stress and Strain in a beam section ....................................................................................... 12

Figure 11 Different yield line patterns in a T-stub: A) circular T-stub B) non-circular T-stub C) beam T-
stub ........................................................................................................................................................ 13

Figure 12 Kennedy bolt force model ..................................................................................................... 14

Figure 13 fail mode 3 ............................................................................................................................ 14

Figure 14 failure mode 2 ....................................................................................................................... 15

Figure 15 virtual work mode 2 .............................................................................................................. 15

Figure 16 failure mode 1 ....................................................................................................................... 16

Figure 17 Virtual work mode 1 ............................................................................................................. 17

Figure 18 fail modes of a beam T-stub .................................................................................................. 18

Figure 19 Original and equivalent T-stub ............................................................................................. 18

Figure 20 Effective length of an equivalent T-stub for one bolt row .................................................... 19

Figure 21 T-stub with 2 edges fully constraint ...................................................................................... 21

Figure 22 determination of Fn ............................................................................................................... 22

Figure 23 first FE-model ....................................................................................................................... 22

Figure 24 mesh representation of first FE-model .................................................................................. 23

Figure 25 plastic hinges occur next to fixed constrain and web ............................................................ 24

Figure 26 Determination of the plastic hinge (gray line) ...................................................................... 25

5
Figure 27 second FE-model .................................................................................................................. 26

Figure 28 mesh presentation of second FE-model ................................................................................ 26

Figure 29 k-L plot of 0.03 beam............................................................................................................ 27

Figure 30 Ly-L plot of a 0.03 beam ....................................................................................................... 27

Figure 31 effective plastic strains .......................................................................................................... 28

Figure 32 half of a non circular pattern with 2 bolts ............................................................................. 29

Figure 33 representation of the model in Comsol ................................................................................. 30

6
6 Introduction

A bolted steel connection (Figure 1) is a commonly used joint in steel constructions. Due to the lack of
welding on the construction site and transportability, many engineers like to implement this type of
connection in their construction design. In order to make a global analyses of the construction, the
moment resistance of the junction needs to be known.

In this aspect, the Committee for European Standardization (CEN, 2005) developed an analytical design
method for bolted steel connections. This method is written down in part 1.8 of Eurocode 3, which is an
European standard for design of steel structures based on different national standards of the members of
the European union. This part is often referred to as EN 1993-1-8: 2005 but hereafter, it is referred to as
EC3-1.8.

A basic bolted steel joint exists of two connecting members (beam-column, beam-beam, foundation-
column,…), an endplate and bolts. The endplate is welded to one member and bolted to the other
member. Often are backing plates, flange cleats, column web stiffness,… used to enforce the steel joint.
Design methodologies to take this reinforcements into account are also presented in EC3-1.8.

Endplate
Beam

Column
Bolts

Figure 1 Basic bolted steel connection

7
7 Literature

7.1 Moment resistance of a bolted steel junction

The moment resistance of a steel junction can be found on a moment-rotation characteristic. A moment-
rotation characteristic is experimentally determined by increasingly apply a bending moment Mj on the
bolted steel junction until it fails. The rotation φj is the relative rotation of the centerlines of the 2
connecting members. A typical moment-rotation curve of a bolted steel junction is given in Figure 2
where Sj,ini and Mj,Rd are respectively the stiffness and the moment resistance of the junction.

Figure 2 Moment rotation characteristic of a bolted junction

Based on the values of MjRd, Sj,ini and type of analysis, the junction can be categorized as one of the
following types:

 A simple behavior in which the joint may be assumed not to transmit bending moments
 A continuous behavior in which the behavior of the joint may be assumed to have no effect on
the analysis
 A semi-contiguous behavior in which the behavior of the joint needs to be taken into account in
the analysis.

The connections are graphical illustrated as in Figure 3.

8
Figure 3: different categories used in EC3-1.8: A) simple B)semi continuous C)continuous

To illustrate the influence of these categories, a global analysis of a steel frame is carried out. A steel
frame portal is subjected to an unified distributed load (Figure 4). The bolted steel connections are
respectively categorized as simple and semi-continue. In the first case, the bending moments in the beam
are large and no bending moments develop in the columns. In the other case, bending moments occurs
in both the beam as the columns but are significantly smaller due to semi-continuous connection
behavior. Bending moments must be taken into account in the beam, column and connections design
and will result in a bigger beam and smaller columns for the first model.

Figure 4 Bolted steel portal subjected to an unified distributed load

The moment resistance Mj,Rd needs to be calculated since it is a parameter to categorize the junction in
a global analysis. EC3-1.8 uses the method of components to calculate Mj,Rd. In this method, the joint is
cut in several pieces which are studied to resistance separately. Finally, the moment resistance Mj,Rd is
calculated with the maximum tolerated tensile force FRd,i of a bold row I and the distance of the bolt row
to the center of the compression flange hi (1). FRd,i is based on the resistance of the other components

9
Figure 5 Bolded beam-column connection with endplate

(1)
, = ∗ ,

Table 1 Components of a steel junction

One of the components of a bolted steel junction is a T-stub. A T-stub has a configuration of an I-beam
which is cut in half. The flange of the T-stub is bolted and the web is subjected to a tensile or pressure
force. This research focusses only on T-stubs subjected to a tensile force but more information can be
found in EC3-1.8. In a basic beam-column connection with endplate, several T-stubs subjected to tensile
force be found. It is important to inderline that bolts can be part of several T-stubs. A widely used method
to calculate the resistance of these T-stub’s is the method of the equivalent T-stub (CEN, 2005)

Figure 6 T-stubs in a bolted column-beam connection

10
7.2 The fail resistance of a T stub

As described above, a bolted steel junction is split up in different T-stubs subjected to a tension F (Figure
7). This tensile force comes from the tensile force in the beam flange. Zoetemeijer distinguishes two
failure mechanisms in these T-stub’s: failure of the bolds and failure of the flanges(Zoetemeijer,
1974)(Figure 8). In the next chapters, the T-stub failure resistance is discussed using the yield line theory
on flanges and the principle of virtual work in the configuration of Figure 7.

tf tw
m
n
L

Figure 7 dimensions of a T-stub

Figure 8 fail mechanisms: left) bolt rupture, right) flange yielding

7.2.1 Yield line theory

The stress-strain curve of structural steel can be simplified to an ideal elastic-plastic material model
(Figure 9). An important feature of this material model is the yield stress fy which is the theoretically
maximum stress in the material. Although the stress σ is limited by fy, the strains ε can still grow if fy is
reached. These latter strains are called plastic strains and the material deforms permanently. This process
is called yielding.

11
Figure 9 ideal elastic- plastic stress strain curve

A plastic hinge is a concept in the classical beam theory. The stress and strain distributions in an ideal
elastic-plastic beam-section with 2 symmetric asses subjected to bending is shown in Figure 10. The
strain is assumed to be linear with the distance to the gravity centrum of the section and zero at the
gravity centrum. The corresponding stress can be found in the stress strain curve (Figure 9). If the
bending moment is small, only elastic strains occurs and the section is not yielding (Figure 10a). At
increasing bending moment, strains will grow and the edges of the section will firstly reach plastic
strains (Figure 10b). Now the section is partially yielding. This process can go on until a stress
distribution like Figure 10c is reached. At that moment, the section lost all resistance to bending and is
fully yielding. In this state, the section can be modelled as a (plastic) hinge since a hinge has zero
resistance to bending either.

Figure 10 Stress and Strain in a beam section

The maximum bending moment a section can take without plastic strains is called the elastic moment
resistance Mel. The maximum bending moment a section can take with plastic strains is called the plastic
moment resistance Mpl.(2)

∗ ∗ (2)
=
4

12
The yield line theory assumes the formation of a yield lines pattern in the flange. These patterns can
eventually turns the flange into a mechanism since a yield line a continuous formation of a plastic hinge
along a strait or curved line is.(Summer, 2003). The flange becomes a collection of rigid parts which
can turn relatively to each other in the yield lines. Generally , the yield line patterns can be categorized
in three families: circular, non-circular and the beam family (Figure 11). A circular pattern is likely to
develop if n/m>1.8 (De Corte, 2015) while a beam pattern develops if L is short relatively to m+n and
the bolt spacing (Santiago, da Silva, Latour, Rizzano, & Trezza, 2013). Due to the practical dimensions
of columns and endplates, the beam pattern is not likely to occurs in T-stubs found in a beam-column
connection.

Figure 11 Different yield line patterns in a T-stub: A) circular T-stub B) non-circular T-stub C) beam T-stub

Since parts between the yield lines are supposed to be rigid, the virtual displacement can only occurs in
the yield lines. This means the internal energy is concentrated in the yield line pattern and elastic
deformation is neglected. The general expression for the internal work Wint stored in a yield line pattern
is given by (3) with Lix and Liy respectively the x and y component of every yield line, N the number of
yield lines, mp the plastic moment strength of the flange per unit length, Ɵ’ix and Ɵ’iy the virtual rotation
in respectively the x and y component of a yield line(Summer, 2003).

(3)
= ( , , + , , )

Equation (4) gives the value of mp with tf the thickness of the flange and fy the yield strength of the
material.


= (4)
4

7.2.2 The Kennedy bolt force model

The failure behavior of a T-stub with a beam pattern can be discussed with the Kennedy bolt force
model. Due to the simplicity of the beam pattern (Figure 11c), the T-stub can be simplified as a Bernoulli

13
beam with plastic hinges at the yield lines and a width equal to the length L (Figure 12). The forces are
simplified to:

 The tensile force F: the force applied on the web


 Prying forces Q: contact force with the surroundings due to deformation of the flange
 Bolt force Ft: Reaction force of the bolds

Ft F Ft
n m m n
tf

Q Q

Figure 12 Kennedy bolt force model

The tensile force F is applied in the middle of the beam, corresponding to the place of the web. The
model suppose only plastic deformation and has a perfect elastic-plastic material. The plastic moment
resistance of the beam is given by (5).

∗ ∗
= = ∗ (5)
4

The relation between Mp and the bolt resistance Ft,Rd determines the failure behaviour. A first failure
mode manifest when Ft,Rd is reached before Mp. Since elastic deformation is neglected, the beam has no
deformation and no prying forces Q occur. This type of behaviour is referred to as mode 3 by EC3-1.8.
The forces and the moment distribution are given in Figure 13 and F3,Rd can be calculated out of newton’s
laws (6).

M1

Ft,Rd F3,Rd Ft,Rd

Figure 13 fail mode 3

14
, = , (6)

If M1 reaches Mp, two hinges appear next to the web, corresponding with the 2 yield lines in the T-stub
with beam pattern (schematized as 1 hinge in Figure 14). If Ft,Rd is reached before M2 reaches Mp, the
T-stub will fail by bold rupture. This behavior is referred to as mode 2 by EC3-1.8. Since these hinges
introduce plastic deformation, the beam will deform and prying forces Q will appear. The forces and
the moment distributions are given in Figure 14 and FRd should be calculated out of the deformed beam
with the yield line theory and virtual work.

Mp

M2 M2

Ft,Rd F2,Rd Ft,Rd

Q Q

Figure 14 failure mode 2

The formulas of internal and external work are respectively (7) and (8) with Li equal to L, v’= θ’1*(m+n),
u’= θ’1 *n and θ’1=θ’2 (Figure 15).

Figure 15 virtual work mode 2

= + (7)

(8)
= , ∗v − ∗ ,

15
The principle of virtual work demands the equilibrium between the internal and external work which
can be used to find F2,Rd(9).

, ∗v − ∗ , = +

, ∗ ∗( + )− , = +

2 + ∑ ,
, =
+
(9)

If M2 reaches Mp, hinges will appear corresponding to the yield lines at the bolt line(Figure 11c). These
hinges turn the beam into a mechanism and the T-stub will fail by flange yielding. This behavior is
referred to as mode 1 by EC3-1.8. Since these hinges introduce plastic deformation, prying forces Q
will appear and they reaches a maximum in this mode equal to Mp/n (Zoetemeijer, 1974) . The forces
and the moment distributions are given in Figure 16 and FRd should be calculated out of the deformed
beam with the yield line theory and virtual work.

Mp

Mp Mp

Ft F1,Rd Ft

Qmax Qmax

Figure 16 failure mode 1

16
The formulas of internal and external work are respectively (10) and (11) with Li equal to L, v’= θ’1*m
and θ’1=θ’2=θ’3= θ’4(see Figure 17).

Figure 17 Virtual work mode 1

(10)
= ( )

= ∗ , (11)

The principle of virtual work demands the equilibrium between the internal and external work which
can be used to find F1,Rd (12)

, ∗v = + + +

, ∗ ∗ = + + +

, ∗ =4∗ ∗

4∗ ∗
, =
(12)

17
To conclude, a T-stub with a beam pattern can fail in 3 different modes, categorized in 2 failure
mechanism: failure by bolt rupture and failure by flange yielding.

Bolt rupture Flange yielding

F3,Rd F2,Rd F1,Rd

Ft,Rd Ft,Rd Ft,Rd Ft,Rd


Ft Ft

Q Q Qmax Qmax

Figure 18 fail modes of a beam T-stub

7.2.3 The equivalent T-stub method

The resistance FRd to failure of T-stub out of the other two yield line families is harder to define because
of the complex shape of their yield line pattern. EC3-1.8 generalize the fail modes of a T-stub out of a
beam family to all families. The resistance to the fail modes can be calculated with the same formula1 if
L is replaced with an effective length Leff; the resistance FRd of a circular and non-circular T-stub can be
found as the corresponding FRd of a beam T-stub with a length equal to an effective length Leff. This
latter T-stub is called the equivalent T-stub and differs with the original T-stub only in length; the other
dimensions stay the same(Figure 19).

Effective length

Original T-stub Equivalent T-stub

Figure 19 Original and equivalent T-stub

1 only applicable if e<1.25m; the size and type of the weld has no influence on FRd
18
The equation of Leff depends on the configuration of the T-stub like the number of bolts, the distance of
the bolts to the edge, the proximity of stiffeners… EC3-1.8 provides Leff for T-stubs that can be found
in the column flange and the endplate. Leff of a T-stub with one bolt row and no stiffeners are given in
Figure 20. Zoetemijer discussed a way to find Leff of a T-stub with 2 boldrows and a non-circualr pattern.
However, it stays hard to connect a physical value to Leff but in case of a circular pattern, EC3-1.8 defines
it as the perimeter of the yield zone(Härmälä, 2017)

Figure 20 Effective length of an equivalent T-stub for one bolt row

19
8 Problem

The formulas of the T-stub resistance FRd are based on the Kennedy bolt force model. This analytical
model contains a lot of simplifications of reality:

 The influence of elastic deformation is totally neglected while deformation plays an important
role in the grow of the prying forces Q.
 Shear forces are neglected.
 Since the formula are derived from a Bernoulli beam model, the influence of the length L might
be underestimated.
 The Kennedy bolt force model assumes the prying forces at the ends of the beam, corresponding
with the edge of the flange, while the prying forces distribution might be more complex.
 The yield line theory divides the flange in rigid parts connected by yield lines with a yield line
a continuous formation of a plastic hinge is. These assumptions might be too simple in aspect
to reality.
 The equivalency between the original and equivalent T-stub is not that simple as EC3-1.8
presents

9 Determination of the research field & Methodology

The simplifications which are discussed in this paper are the yield line theory and the equivalency
between T-stubs out of different yield line families. This research contains a comparison between
analytical results from the Kennedy bolt force model and numerical results. In order to find numerical
results, FE-models were developed with Comsol.

20
10 Research & Results

10.1 Yield line Theory

The yield line theory assumes a yield line pattern in a T-stub flange on moment of a failure. This yield
line pattern turns the T-stub flange into a collection of rigid parts with yield lines on the edges. This
chapter investigates the validity of these assumptions.

A flange failure is numerically a challenge to model due to the prying forces Q. These must be modeled
as contact forces and a way to take them into account is discussed in the next chapter. Since the yield
line theory is applicable on many other configurations, a T-stub of structural steel S235 with 2 edges
fully constraint is used in the research (Figure 21). If a linearly constant tensile force F is applied on the
web, the formula to determine failure resistance FRd is the same as a T-stub failure with a beam
pattern(equations (13)(14)). This model can be seen as a beam T-stub with full constrains at the bolt line

tf
2m

Figure 21 T-stub with 2 edges fully constraint

4∗ ∗ (13)
=

235 ∗
= (14)
4

The examination of the yield line theory is discussed on two domains. Firstly, FRd is compared with the
numerical result Fn of a FE-model. This is the maximum value of F, plotted to the displacement of the
web(Figure 22). Secondly, the stress distribution in the plate is examined in order to find the plastic
deformation assumed by the yield line theory.

21
Figure 22 determination of Fn

10.1.1 First FE-model

The first model exists of a two dimensional beam with a width equal to L. A displacement is applied on
the web of the T-flange. The model has a perfect elastic-plastic model in order to find a stress distribution
of a hinge and uses plane stress situation. In order to minimize the computation time, only one symmetric
part is modeled.

displacement
x
tf
m m
tw

Figure 23 first FE-model

22
Figure 24 mesh representation of first FE-model

A parametric analyses has been carried out with three parameters: m, tf and L. The results are in the
annex. The parametric analysis shows the independence of this model to the width L of the beam,
probably due to the use of ‘plain stresses’. A use of ‘plain strains’ will probably give another outcome
but in the next chapter, a more accurate model is used to analyze L. Further, a correlation between the
tf/m ratio and the accurateness of the results is diagnosed. The ratio tf/m will be called “the beam ratio”
there a low ratio defines a classic beam where a big ratio evolves the beam in a cubes. The obtained Fn
shows a more accurateness to FRd if the beam ratio is low. The higher the beam ratio, the larger Fn is in
aspect to FRd with maximum beam ratio of 0,29. If this maximum is exceeded, the FE-calculations fails
due to convergence problems. This is shown in Table 2.

Table 2 influence of the beam ratio tf/m on Fn/Fa

tf(mm) m(mm) tf/m(-) L(mm) Mp (Nm) FRd(kN) Fn(kN) Fn/FRd(-)

5 100 0,05 252 370,125 14,8 15,005 1,01

5 48 0,10 252 370,125 30,8 31,831 1,03

10 48 0,21 252 1480,5 123,4 128,95 1,05

5 20 0,25 252 370,125 74,0 78,216 1,06

no
14,5 48 0,30 252 3112,751 259,4 -
convergence

The fact that the beam ratio needs to be low (=classical beam interpretation) in order to have better
accurateness seems logical if the reader takes in mind that plastic hinges are defined on the classical

23
definitions of Bernoulli beams. With this in mind, a model with a low and high beam ratio respectively
0.03 and 0.25 will be analyzed in order to analyze the plastic hinges.

Figure 25 plastic hinges occur next to fixed constrain and web

To find the plastic hinge, the equivalent strains are used. These strains are a value for deformation if the
material is yielding. If the equivalent strains occur along the total section, total yielding can be assumed.
The section with the most plastic strains is taken as the place of the hinge(Figure 26). Contrary to the
analytical model, the hinges doesn’t occur at the end of the beam. At the hinge, a stress distribution is
founded which reaches in every case more or less the theoretical stress distribution. The difference
between the 0.03 and 0.2 beam is the place of the hinge. The distance m2 between the hinges is smaller
in the 0.2beam compared to the 0.03beam although m is both times 200mm.

tf m tf/m L FRd Fn Fn/FRd H1 H2 m2 m2/m


Mp (Nm)
(mm) (mm) (-) (mm) (kN) (kN) (-) (mm) (mm) (mm) (-)
5 200 0,03 15 22,03125 0,4 0,4422 1,00 198,86 1,4155 197,44 0,987
40 200 0,20 15 1410 28,2 29,8 1,06 199,5 6,2 193,3 0,967

This can explain the difference between the 0.2beam and the 0.03beam explained in the previous
paragraphs. If m2 is used in the analytical formula, Fn will approximate FRd more accurately and the
difference between 0.2beam and 0.03beam is less.

tf (mm) m (mm) tf/m (-) L (mm) Mp (Nm) FRd (kN) Fn (kN) Fn/FRd (-)
5 197,44 0,03 15 22,03125 0,4 0,4422 0,99
40 193,3 0,21 15 1410 29,2 29,8 1,02

24
Figure 26 Determination of the plastic hinge (gray line)

The ratio tw/m between an 0.05 and 0.2 model (Table 3) changes becauce tw is always 2.5[mm]. Since
this could have an influence on Fn, A 0.05 beam has been made with the different tw/m ratio’s. The
results are given in Table 3 and show a slight difference in value for of Fn. Although, this the differnce
not big enough to be the only explanation for the difference between the original 0.05 and 0.2beams.

Table 3 influence of tw/m

tw (mm) tf(mm) m(mm) tw/m[-] tf/m(-) L(mm) Mp (Nm) FRd(kN) Fn(kN) Fn/FRd(-)
2,5 5 100 0,025 0,05 252 370,125 14,8 15,005 1,01
12,5 5 100 0,125 0,05 252 370,125 14,8 15,154 1,02
Since the model was independent of L, a conclusion about yield lines is hard to formulate. On the
contrary, the concept of “plastic hinges” is proved to be accurate in this model. The accurateness depends
especially on the beam ratio since it determine the value of m2.

10.1.2 Second FE-model

In order to find the influence of L, a 3D version of the previous model was developed. Since tf and m
are already discussed in the previous chapter, this chapter will carry out a parametric analyses of L with
a 0.03, 0.1 and 0.25-beam.

25
displacement
L

tf
2m

Figure 27 second FE-model

Figure 28 mesh presentation of second FE-model

The parameter L is the length of the plate, but is also assumed to be the length of the yield lines in the
yield line theory. This assumption is researched by defining a the parameter Ly which is the yield line
length based on the numerical results (15). If the length of the plate is equal to the yield line length, the
relation between L and Ly should be as (16) with k=1.

= (15)
4∗ ∗

= ∗
(16)

It seems from the results that k is depending on the beam factor but always higher than 1. It is interesting
to underline that k is not constant for a small value of L as can be seen in Figure 29. This could be caused
by the evolution of a beam into a plate. On a Ly-L plot, this evolution is nearly visible and a linear
correlation can be assumed between Ly and L(Figure 29). Since Fn is always higher than FRd, this method
can be seen as safe.

26
Table 4 value of k

beam
0,03 0,10 0,25
factor
k 1,23 1,38 1,58

1,24

1,22

1,20

1,18
k [-]

1,16

1,14

1,12

1,10

1,08
0 50 100 150 200 250 300 350
L [mm]

Figure 29 k-L plot of 0.03 beam

400

350

300

250
Ly [mm]

200

150

100

50

0
0 50 100 150 200 250 300 350
L [mm]

Figure 30 Ly-L plot of a 0.03 beam

Since mp is defined as a continuous deformation of a hinge along a line, the deformed section of the
yield line must stay a 2 dimensional plane. To find the yield line section, the effective strains are again
used as described above. It seems to be hard to find the section because the yield lines does not stay
strait as can be seen in Figure 31. The yield lines turn away at the borders. Although this discovery does

27
not fit with the theory, the “extra length” due to the side effects is not enough to explain the difference
between Ly and L.

Figure 31 effective plastic strains

To conclude, the real length of the yield line is hard to define because the exact path is difficult to find.
Instead, a theoretical derived yield line length Ly has been found and compared with the assumed yield
line length L. It was found that the linearity between L and Ly is depending on the tf/m ratio.

28
10.2 The equivalent T-stub methodology

10.2.1 The effective length Leff

EC3-1.8 don’t connect a physical value to Leff though a presumption exists that Leff is the perimeter of
the “yielding zone” of a circular and non-circular pattern(Härmälä, 2017). This makes sense for a
circular pattern since the Leff is equal to 2*Pi*m (see Figure 20) but for a non-circular pattern, this
presumption is harder to understand. In this chapter, a parametric analysis is carried out to a, n and m of
a T-stub with 2 bold rows with each 2 bolts and Leff is compared with the perimeter. Figure 32 shows
the yield line pattern of one symmetric part of this T-stub and the perimeter of the yielding zone is
L2+2*L3 while Leff = a+4*m+1.25n. The angles α and β are determined on the lowest possible internal
energy inside the yield line pattern(Zoetemeijer, 1974).

Figure 32 half of a non circular pattern with 2 bolts

The results of the parametric analysis can be found in the annex. Generally, Leff is between 75% and
100% of the perimeter. This percentage depends on n/m and rises if n/m goes low. The distance “a”
between the two bolds has no influence on the percentage. Since this yielding zone is analytically found
by (Zoetemeijer, 1974) a FE-model should be developed in order to find a more realistic “yielding zone”.

10.2.2 Development of a FE- model

A model was developed in the FE-software COMSOL and based on the article by (Santiago et al., 2013)
in order to research the mode 1 behavior. The model exists of a T-stub half and two bolds. The
dimensions and properties of the T-stub and bolds are taken the same as in the article. In order to find
the failure load Fn, a displacement is inclemently applied on the web. There is only one contact pair
present in the model: the contact pair between the T-stub and the rigid support. The bolt head is modeled

29
as continue with the T-stub flange. The bolt is constrained at the thread’s surface in order to make it
more realistic. The T-stub rests on a support body which is constrained in space. Only one half of the T-
stub is modeled to reduce computation time.

To define a contact pair in COMSOL, a source and destination surface must be selected. A source surface
has to exist of bigger elements according to the destination surface and is part of a more rigid body
according to the destination surface. In this model, the support can be modeled as a rigid body in order
to reduce the computation time.

The method to compute the contact effect is de ‘augmented LaGrange’ method. This method is more
accurate compared to the second option, the penalty method, but takes a lot of time to return a solution.
The penalty factor used in the augmented LaGrange method is reduced in order to stimulate a
convergence in the first steps of the displacement.(Pennec et al., 2007)

Both the penalty and augmented LaGrange method calculate the displacement between the contact
surfaces, which is called mapping. COMSOL advices to use a calculation based on the initial distance
between the surfaces if the distance is small and likely to increase. This leads to a more stable and faster
convergence.(COMSOL) (COMSOL/documentation)

COMSOL uses by default the double dogleg solver to solve nonlinearity but due to convergence
problems, the newton solver is used which is more robust but slower.

The T-flange and the bold are defined as an elasto-plastic material model with a linear isotropic
hardening model. The isotropic tangent modulus ETiso is set on 200e6 in order to approach perfect
plasticity.

Figure 33 representation of the model in Comsol

To test the accurateness, the model was subjected to the same test as the model in the article by (Santiago
et al., 2013). The test had as purpose to investigate mode 1 behavior of the T-stub with 4 bolds. The
comparison is based on the displacement-force diagram of both models. The force FT is plotted on the
y-axis. The distance d is the displacement of the web-flange junction and is plotted on the x-axis. Both
curves predict the plastic transition between 100kN and 120kN and before reaching 2.5mm
displacement. Also the slope before and after the plastic transition are more or less the same. Although
30
the comparison between the 2 models is hardly exact, the developed model is supposed to be realistic
until the breaking point. The developed model will not reach a breaking point and will continue to
deform.

180

160

140

120
Force F[kN]

100

80

60

40 FEM developed
20

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

11 Conclusion

The yield line theory can be assumed as safe but not accurate in the models which were researched.
Further research showed a correlation between the accurateness and the ratio of the flange thickness tf
and flange length m. It seems that FRd is more realistic if tf/m is low.

The search to the physical value of Leff is stranded on the development of the FE-model. The developed
FE-model was able to reproduce a realistic behavior but was not able to define the failure load Fn.
Further research to this needs to be carried out on this model. However, a parametrical analysis of a
yield line pattern found that Leff is between 75% and 100% of the perimeter of the yielding zone.

31
12 Annex

12.1 Results of the parametric analyses of a 2D model

Table 5 analyses to L

tf m tf/m L Mp (Nm) Fa(kN) Fn(kN) Fn/Fa

5 48 0,10 228 334,875 27,9 28,799 1,03

5 48 0,10 252 370,125 30,8 31,831 1,03

5 48 0,10 288 423 35,3 36,378 1,03

5 48 0,10 301,44 442,74 36,9 38,075 1,03

Table 6 analyses to m

tf m tf/m L Mp (Nm) Fa(kN) Fn(kN) Fn/Fa

5 20 0,25 252 370,125 74,0 78,216 1,06

5 30 0,17 252 370,125 49,4 51,722 1,05

5 48 0,10 252 370,125 30,8 31,831 1,03

5 100 0,05 252 370,125 14,8 15,005 1,01

Table 7 analyses to tf

tf m tf/m L Mp (Nm) Fa(kN) Fn(kN) Fn/Fa

5 48 0,10 252 370,125 30,8 31,831 1,03

10 48 0,21 252 1480,5 123,4 128,95 1,05

no
20 48 0,42 252 5922 493,5 -
convergence

no
40 48 0,83 252 23688 1974,0 -
convergence

2 48 0,04 252 59,22 4,9 5,039 1,02

32
14 48 0,29 252 2901,78 241,8 253,33 1,05

no
14,5 48 0,30 252 3112,7513 259,4 -
convergence

12.2 Results of the parametric analyses of a 3D model

Table 8 Analyse of a 0.03 T-stub

mp
tf (mm) m (mm) tf/m (-) L (mm) FRd (kN) Fn (kN) k Ly
(Nm)
5 200 0,03 300 1,46875 8,8 10,86 1,23 369,70
5 200 0,03 252 1,46875 7,4 9,1 1,23 309,79
5 200 0,03 200 1,46875 5,9 7,2 1,23 245,11
5 200 0,03 150 1,46875 4,4 5,4 1,23 183,83
5 200 0,03 100 1,46875 2,9 3,56 1,21 121,19
5 200 0,03 50 1,46875 1,5 1,7353 1,18 59,07
5 200 0,03 10 1,46875 0,3 0,32 1,09 10,89

Table 9 Analyses of a 0.1 T-stub

mp
tf (mm) m (mm) tf/m (-) L (mm) FRd (kN) Fn (kN) k Ly
(Nm)
5 50 0,10 500 1,46875 58,8 81 1,38 689,36
5 50 0,10 300 1,46875 35,3 48,6 1,38 413,62
5 50 0,10 252 1,46875 29,6 40,55 1,37 345,11
5 50 0,10 200 1,46875 23,5 32,2 1,37 274,04
5 50 0,10 150 1,46875 17,6 23,8 1,35 202,55
5 50 0,10 100 1,46875 11,8 15,3 1,30 130,21

Table 10 Analyses of a 0.25 Tstub

mp
tf (mm) m (mm) tf/m (-) L (mm) FRd (kN) Fn (kN) k Ly
(Nm)
5 20 0,25 500 1,46875 146,9 232,6 1,58 791,83
5 20 0,25 300 1,46875 88,1 140,23 1,59 477,38
5 20 0,25 252 1,46875 74,0 117,04 1,58 398,43
5 20 0,25 200 1,46875 58,8 93 1,58 316,60
5 20 0,25 150 1,46875 44,1 66,76 1,52 227,27
5 20 0,25 100 1,46875 29,4 42,36 1,44 144,20
5 20 0,25 50 1,46875 14,7 19,85 1,35 67,57
5 20 0,25 10 1,46875 2,9 3,73 1,27 12,70

33
34
n/m m n a (mm) beta alpha L2 (mm) L3 (mm) free
(mm) (mm) edge
(17)

(18)

1,90 30,00 57,00 30,00


1,80 30,00 54,00 30,00 1,27 0,96 116,42 88,02 169,02
1,70 30,00 51,00 30,00 1,26 0,95 114,33 85,03 166,08
1,60 30,00 48,00 30,00 1,26 0,94 112,19 82,04 163,07
1,50 30,00 45,00 30,00 1,25 0,93 110,00 79,06 160,00
)

2 = a + 2 ∗ m ∗ tan(α)

Table 11 Analyses to n
)

+
sin( )
+

1,40 30,00 42,00 30,00 1,24 0,91 107,75 76,07 156,85


sin( )
1

+
2∗
12.3 Results of the parametric analyses of Leff

= arccos(

1,30 30,00 39,00 30,00 1,23 0,90 105,43 73,09 153,62


= arccos(

3= 1,20 30,00 36,00 30,00 1,23 0,88 103,04 70,10 150,31


1,10 30,00 33,00 30,00 1,22 0,87 100,58 67,12 146,90
1,00 30,00 30,00 30,00 1,21 0,85 98,03 64,14 143,39
0,90 30,00 27,00 30,00 1,20 0,83 95,39 61,17 139,77
0,80 30,00 24,00 30,00 1,19 0,81 92,65 58,19 136,02
0,70 30,00 21,00 30,00 1,18 0,78 89,79 55,22 132,15
35
l eff perimet Leff/peri
(mm) er (mm) meter
n/m m (mm) n (mm) a (mm) beta alpha L2 (mm) L3 (mm)
248,40 248,40 1,00
217,50 292,46 0,74
1,70 29,41 50,00 30,00 1,26 0,95 112,68 83,36
213,75 284,40 0,75
1,60 31,25 50,00 30,00 1,26 0,94 115,62 85,46
210,00 276,28 0,76
1,50 33,33 50,00 30,00 1,25 0,93 118,89 87,84
206,25 268,11 0,77
Table 12 Analyses to m

1,40 35,71 50,00 30,00 1,24 0,91 122,56 90,56


202,50 259,89 0,78
1,30 38,46 50,00 30,00 1,23 0,90 126,71 93,70
198,75 251,60 0,79
1,20 41,67 50,00 30,00 1,23 0,88 131,45 97,37
195,00 243,25 0,80
1,10 45,45 50,00 30,00 1,22 0,87 136,94 101,70
191,25 234,82 0,81
1,00 50,00 50,00 30,00 1,21 0,85 143,39 106,90
187,50 226,32 0,83
183,75 217,73 0,84 0,90 55,56 50,00 30,00 1,20 0,83 151,10 113,27
180,00 209,04 0,86 0,80 62,50 50,00 30,00 1,19 0,81 160,52 121,23
176,25 200,24 0,88 0,70 71,43 50,00 30,00 1,18 0,78 172,36 131,48
36
free edge perimeter Leff/peri
l eff (mm) n/m m (mm) n (mm) a (mm) beta alpha L2 (mm)
(mm) (mm) meter
163,41 210,15 279,41 0,75 0,70 42,86 30,00 75,00 1,18 0,78 160,42
168,62 217,50 286,54 0,76 0,70 42,86 30,00 70,00 1,18 0,78 155,42
174,44 225,83 294,57 0,77 0,70 42,86 30,00 65,00 1,18 0,78 150,42
Table 13 Analyses to a
181,01 235,36 303,68 0,78 0,70 42,86 30,00 60,00 1,18 0,78 145,42
188,49 246,35 314,11 0,78 0,70 42,86 30,00 55,00 1,18 0,78 140,42
197,09 259,17 326,18 0,79 0,70 42,86 30,00 50,00 1,18 0,78 135,42
207,12 274,32 340,34 0,81 0,70 42,86 30,00 45,00 1,18 0,78 130,42
218,98 292,50 357,20 0,82 0,70 42,86 30,00 40,00 1,18 0,78 125,42
233,27 314,72 377,64 0,83 0,70 42,86 30,00 35,00 1,18 0,78 120,42
250,89 342,50 402,99 0,85 0,70 42,86 30,00 30,00 1,18 0,78 115,42
273,20 378,21 435,32 0,87 0,70 42,86 30,00 25,00 1,18 0,78 110,42
37
free edge perimeter Leff/peri
L3 (mm) l eff (mm)
(mm) (mm) meter
78,89 220,92 283,93 318,19 0,89
78,89 215,92 278,93 313,19 0,89
78,89 210,92 273,93 308,19 0,89
78,89 205,92 268,93 303,19 0,89
78,89 200,92 263,93 298,19 0,89
78,89 195,92 258,93 293,19 0,88
78,89 190,92 253,93 288,19 0,88
78,89 185,92 248,93 283,19 0,88
78,89 180,92 243,93 278,19 0,88
78,89 175,92 238,93 273,19 0,87
78,89 170,92 233,93 268,19 0,87
13 References

CEN. (2005). Eurocode 3: Design of Steel Structures - Part 1-8: Design of Joints: Slovenski inštitut za
standardizacijo.
COMSOL. Improve the stability of your contact model. Retrieved 18/05/2017, 2017
COMSOL/documentation. Contact Pairs.
De Corte, W. (2015). STAALBOUW 2: PARTIM BOUTVERBINDINGEN cursus. bouwkunde - landmeten.
Härmälä, T. (2017). COMPARING NUMERICAL EFFECTIVE LENGTH WITH MODELLED EFFECTIVE LENGTH
IN T-STUB CONNECTIONS. Tampere University of technology, INSA Lyon.
Pennec, F., Achkar, H., Peyrou, D., Plana, R., Pons, P., & Courtade, F. (2007). Verification of contact
modeling with COMSOL multiphysics software.
Santiago, A., da Silva, L. S., Latour, M., Rizzano, G., & Trezza, S. (2013). THEORETICAL AND
EXPERIMENTAL ANALYSIS OF T-STUBS WITH FOUR BOLTS.
Summer, E. A. (2003). Unified Design of Extended End-Plate Moment Connections Subject to Cyclic
Loading. (Doctor of Philosophy in Civil engineering dissertation), Faculty of the Virginia
Polytechnic Institute and State University, Blacksburg, VIrginia. (209)
Zoetemeijer, P. (1974). A Design Method for the Tension Side of Statically Loaded, Bolted Beam-to-
Column Connections. HERON.

38

S-ar putea să vă placă și