Академический Документы
Профессиональный Документы
Культура Документы
November 2005
coordinated by:
J.P. Grimault
ARCELOR Tubes
August 2005
Page 2 of 56
TABLE OF CONTENTS
Foreword........................................................................................................................................... 3
Part I
Recent developments for the derivation of an analytical formulation for CHS wall
components........................................................................................................................ 4
I.1 Introduction .........................................................................................................................4
I.1.1
Summary of the interim report (5BP-4/04 August 2004) .......................................... 4
I.1.1.1 Types of the plastic mechanisms............................................................................. 4
I.1.1.2 Two contributions to the yield mechanisms ............................................................. 5
I.1.1.3 Ring model mechanism ......................................................................................... 6
I.1.1.4 External mechanisms .............................................................................................. 6
I.1.2
Objectives expressed in the second interim report ..................................................... 8
I.1.3
Research strategy ....................................................................................................... 8
I.1.4
Research steps ........................................................................................................... 8
I.2 Direct analytical approach.................................................................................................10
I.3
I.3.1
Validation of the FEM tool ......................................................................................... 12
I.3.2
Definition of the numerical model for the study of the CHS component .................... 12
I.3.2.1 FEM mesh ............................................................................................................. 13
I.3.2.2 Loading .................................................................................................................. 13
I.3.2.3 Plastic model ......................................................................................................... 13
I.3.2.4 Influence of the thickness of the plate ................................................................... 14
I.3.2.5 Influence of the length of the tube ......................................................................... 15
I.3.3
Parametrical study..................................................................................................... 16
I.3.3.1 Objectives .............................................................................................................. 16
I.3.3.2 Description of the parametrical study .................................................................... 16
I.3.3.3 Results of the parametric study ............................................................................. 17
I.3.3.4 Comparison with Eurocode 3 formula.................................................................... 21
I.3.4
Correlation with Gomes model .................................................................................. 22
I.3.4.1 Application of the Gomes model to CHS components........................................... 22
I.3.4.2 Comparison numerical results and Gomes model ................................................. 23
I.3.4.3 Determination of the limit value ............................................................................. 25
I.3.5
First conclusions of the parametrical study ............................................................... 26
I.4 Conclusions further developments .................................................................................26
Part II
II.1
II.2
II.3
II.4
References...................................................................................................................................... 51
August 2005
Page 3 of 56
Foreword
The application of the component method for the characterisation of the mechanical properties of
structural joints in tubular construction has been initiated into CIDECT few years ago, within the
project 5BM. As an outcome of this first project, the possibility to extend to hollow joints this
method initially developed for joints between open sections has been demonstrated and the
advantage of this use has been pointed out.
Present project 5BP aims at developing further the concept, at indicating how the component
method may be implemented in daily practice but also at identifying any lack of information which
could limit the application. Finally developments were expected to be achieved so as to increase
the scientific knowledge and progress in the preparation of appropriate answers to some of the still
pending technical questions.
In the present report, the reader will find the outcome of the works performed within the 5BP
project in the last two-and-a-half years in the form of two separate volumes, respectively entitled:
Volume 1 : Practical guidelines
Volume 2 : Progress of the scientific activities on joint components and assembly
Volume 1 gathers all the information available to the designer and helpful for the design of a wide
range of structural steel joints connecting hollow and/or open sections. The use of this material
requires anyway an experience which is not necessary widely shared at the moment and therefore
simple design aids (called design sheets) more appropriate to daily practice have been prepared
for some selected joint configurations. These ones are complemented by worked examples.
Should these design sheets be appreciated, additional design sheets covering other joint
configurations would have to be drafted by referring to the material made available in the first part
of Volume 1.
Volume 2, on the other hand, reflects the progress of the scientific works achieved during the
project. These ones contribute to a better understanding of the behaviour of CHS components and
of their assembly. Further research activities which could not have been achieved in the 5BP
project are required in this field so as to come to fully validated models. These works are planned
to be achieved at PSP and at Lige University in the next months; they could be reported on at the
occasion of the next ISTS Symposium to be held in Quebec in 2006.
Part I
August 2005
Page 4 of 56
I.1 Introduction
In parallel with the work presented in the volume 1, some scientific developments have been
performed in Lige in order to derive a new analytical formulation, based on physical aspects and
plastic theory, for the resistance of CHS wall component. In the interim report BP-4/04, the first
elements of this study have been presented. Since August 2004, this study has well progressed,
some tools have been developed, a good understanding of the physical behaviour of the
components has been obtained and some key steps have been crossed in view of the derivation of
analytical formulation for component characterisation.
Hereunder the main elements of the interim reports are briefly reported and the more recent
developments performed this year are presented. Further research steps have still to be
performed. They are planned to be achieved in the next months at Lige University and at PSP; as
this work has been initiated in the 5BP project, the two mentioned institutions will report in the next
CIDECT meetings about the progress of these research works.
When CHS profiles are subjected to transverse compression and tension forces, plastic yield line
mechanisms, membrane effects and possible instability effects develop. If second order effects are
disregarded (membrane effects, instability effects, ), the resistance of the CHS section is a fully
plastic one and the failure loads in compression and in tension are equal. The plastic mechanisms
which form in the CHS section have been analysed through experimental tests and numerical
simulations and, as a result, two types of plastic mechanisms may be identified (Figure 1):
global plastic mechanisms involving the whole section;
local plastic mechanisms involving a part of the whole section.
: yield line
(a) One global mechanism
I.1.1.2
August 2005
Page 5 of 56
In the interim report, the form of the mechanisms has been studied. And it has been shown that the
mechanisms may be divided into two different contributions (see Figures 2 and 3): a ring
model mechanism (Npl(rm)) and two external mechanisms (Npl(em)).
Npl
lp
dm
A
Bp
tm
Npl
Plane view
Cut
A- A :
Beff
Bem
Brm
Bem
Bp
Rectangular brace
CHS chord
Ring Model
External mechanisms
August 2005
Page 6 of 56
It has also been demonstrated that a good estimation of the plastic resistance Npl (tot) of the chord
may be derived by summing the resistance of the separate contributions:
Npl (tot) = Npl(rm) + 2*Npl(em).
I.1.1.3
(1)
An analytical formula for this part of the mechanism has been presented in the previous report.
This formula has been validated by numerical tests. The plastic resistance of the Ring model
contribution for a CHS loaded in transverse tension or compression may be determined by the
following expression (Figure 4):
2
2
f y0
B
t0
(2)
N pl(rm) =
(d t0 ) b
b
with:
d = diameter of the CHS
b = width of the loading area
B = length of the loading area
B
d
I.1.1.4
External mechanisms
The form of the external mechanisms observed in CHS in tension and compression is seen in
Figure 4 in the case of a global mechanism. But an identical shape is observed in local
mechanisms. Furthermore the yield line pattern shown in Figure 4 is quite similar to the one
predicted by the so-called Gomes model for I or H column webs in transverse compression or
tension (Figure 5); the main difference is obviously the nature of the surface where the yield lines
develop:
- a shell in CHS profiles.
- a plate in I or H column;
Gomes proposes an analytical expression to evaluate the plastic resistance of I or H column webs
in transverse compression or tension (Formula 3 and Figure 6):
F pl =
4 m pl
b
1
L
1 b + 2c
L L
(3)
August 2005
Page 7 of 56
L
d
August 2005
Page 8 of 56
(4)
Frd (shell)
Frd(plane)
Direct analytical approaches: the aim is to derive a formulation of the factor through pure
analytical investigations. This work is presented in Section I.2.
Numerical simulations: the FEM approach is used to understand how the actual plastic
mechanism forms on a curved surface and to find a correlation with the in-plane mechanisms.
The numerical approach requires several steps:
validation of the FEM tool;
calibration of the model;
parametrical study;
interpretation and analysis of the results;
This activity is reported in Section I.3.
-
August 2005
Page 9 of 56
CHS
Tension compression
Plastic resistance
Analysis of the
plastic mechanisms
Numerical
investigations
Experimental
results
Two different
mechanisms
Global
mechanism
Local mechanism
similar
two contributions
External
mechanism
Report 2005
Numerical
investigation
Parametrical study
Correlation with
Gomes model
Similar to Gomes
plastic model
Analytical projection
on curve surface
Analytical formulation
for plastic resistance
of CHS wall
Ring model
Analytical model
validated by
numerical
simulations
OK
August 2005
Page 10 of 56
L' =
x
R
E arcsin 0 , sin 2
cos
R
(6)
E [] = elliptical function
x0
August 2005
Page 11 of 56
This general equation is rather complicated. The derivation of a single factor from this equation is
not so easy.
But before trying to go further, it would be important to validate the simplifying assumptions made
and therefore to check whether a direct analytical projection may provide an adequate response.
So the approach has been applied to a very simple model for which the plastic mechanism is
known. The model tested is an cantilever shell element and the corresponding projected plate
subjected to a concentrated load applied at its extremity (Figure 10).
For this case, the problems of projection becomes much more simple. The yield line in the plate
case is equal to the width of the plate: L = b
The projection of this yield line is L = 2R, with R = radius of curvature and = angle of the arc
length.
In this particular situation, the factor is equal to:
Pcurve L' 2R
= =
Pplane
L
b
(7)
R
b
August 2005
Page 12 of 56
800
700
600
P [kN]
500
400
Test results
300
200
100
0
0
10
15
20
25
s [mm]
Figure 11: Load displacement curves of X-joint XP1-T-1 under tension (Y. Makino et al.)
I.3.2 Definition of the numerical model for the study of the CHS component
A T-joint configuration has been chosen as numerical model to study the CHS component in
tension or compression. This T-joint is composed of a tubular element and of a plate fixed
perpendicularly to the tube.
The plate is submitted to an axial force (Figure 12). The nodes are blocked in the three directions
at the extremities of the tube.
August 2005
Page 13 of 56
FEM mesh
The model uses shell elements. Simulations have been computed with different meshes and the
results have been compared. In Figure 14, it is seen that the adopted mesh gives same results
then a more refined one.
I.3.2.2
Loading
The joint is subjected to a point load at the extremity of the plate. In order to avoid a local plasticity
effects at the load introduction point, the edge of the plate has been reinforced by very stiff beam
element. The material law used for the elements of the plate is a linear law. Therefore, only the
face of the CHS is influenced by the loading.
I.3.2.3
Plastic model
In order to study the plastic resistance of the component, all the second order effects (membrane
effects, instability, ) have to be disregarded. For this purpose, the steel material properties used
in the model have been selected in such a way that no second order effect develops. The Young
modulus has been taken equal to 2,1 1010 MPa. That doesnt affect the maximum resistance. It
only changes the initial stiffness and it prevents appearance of large displacements and therefore
of second order effects.
As far as the plastic resistance is concerned, the numerical response should be the same in
tension and compression. This is confirmed in Figure 13, where appearance of a horizontal plateau
in all the four curves proves that the plastic resistance is reached.
August 2005
Page 14 of 56
200000
183418
180000
160000
140000
P (N)
120000
122131
100000
80000
60000
40000
20000
0
0,00E+00 1,00E-04 2,00E-04 3,00E-04 4,00E-04 5,00E-04 6,00E-04 7,00E-04 8,00E-04 9,00E-04 1,00E-03
depx (mm)
tension S235
compression S235
tension S355
compression S355
The thickness of the plate can not be taken into account in the numerical model as shell finite
elements are used. Figure 14 shows that the thickness of the plate has no influence on the
numerical plastic resistance. To take this parameter into account a model with solid elements
should be built, what would lead to much heavier simulations.
In the Gomes analytical model (Formula 3), the thickness of the plate is explicitly taken into
account. But in reality it will be shown in paragraph I.3.4 that this parameter has not a great
influence on the value of the plastic resistance as long as realistic thickness values are considered.
The numerical study can therefore be achieved without taking into account this parameter and a
good estimation of the plastic resistance may be anyway expected.
For the numerical study, all the simulations are run with a plate thickness equal to 1mm.
August 2005
Page 15 of 56
250000
200000
~190000
P (N)
150000
100000
50000
0
0,00E+00
1,00E-03
2,00E-03
3,00E-03
4,00E-03
5,00E-03
6,00E-03
7,00E-03
8,00E-03
9,00E-03
depx (mm)
t=1mm
t = 10mm
t = 5mm
t = 0,1mm
Figure 14: Influence of the thickness of the plate on the plastic resistance
in the shell FEM model
I.3.2.5
A particular attention has to be paid to the length of the tube in the numerical model. Indeed, if the
tube is too long, the failure occurs by a plastic mechanism involving three plastic hinges in the CHS
profile (Figure 16) and the numerical resistance does not correspond to the studied phenomena. If
the tube is too short, the yielding extends all olong the length of the CHS profile and the studied
mechanism can not be isolated.
In fact, For each simulation, the failure mechanism has to correspond to a local mechanism close
to the plate (Figure 15).
August 2005
Page 16 of 56
As shown in Figure 16, for high values of the length, the beam mechanism appears; if the length
decreases, the whole face of the tube is yielded. An intermediate value of the CHS length has
therefore to be selected.
250000
200000
P (N)
150000
100000
50000
0
0
0,005
0,01
0,015
0,02
0,025
dep x (mm)
300mm
600mm
1200mm
1800mm
3000mm
Objectives
The first objective of the parametrical study is to analyse the evolution of the plastic resistance
according to the geometrical parameters of the joint configuration.
The main parameters influencing the plastic resistance are the following ones:
t0 = thickness of the tube
d0 = diameter of the tube
b = width of the plate
The thickness of the plate, as it has been demonstrated on Figure 14, can not be taken in
consideration through the numerical simulations. This parameter will therefore not be integrated in
the study.
In Eurocode 3, use is often made to the two following values:
2 = t0/d0
= b/d
In fact, these two values involve the three main geometrical parameters listed here-above and will
therefore be used as references in the parametrical study.
The second objective is to compare the numerical results to the Gomes formula and to derive an
expression of the factor defined by Formula (3).
I.3.3.2
A set of 36 simulations has been performed. They are grouped into 4 series according to the
thickness of the CHS profile:
Set 1
Set 2
Set 3
Set 4
August 2005
Page 17 of 56
t0 = 8mm
t0 = 6,3mm
t0 = 10mm
t0 = 12mm
I.3.3.3
Figures 18 to 2& present the results of the FEM simulations (P and d are respectively the load
applied to the transverse plate et the corresponding displacement at the CHS surface)
August 2005
Page 18 of 56
August 2005
Page 19 of 56
August 2005
Page 20 of 56
Figure 22 presents the evolution of the plastic load obtained numerically versus the parameter.
When 0, Npl ; this is logical as =0 corresponds to a CHS with an infinite thickness. For
high values of , the plastic resistance again tend to a limit value.
400000
350000
t=8mm
300000
t=6,3 mm
t=10 mm
Npl (N)
250000
t=12 mm
200000
150000
100000
50000
0
0
50
100
150
200
250
300
20
y = 484159x + 197114
t=8mm
350000
t=6,3 mm
Npl (N)
300000
t=10 mm
y = 379377x + 140751
t=12 mm
250000
y = 289710x + 93524
200000
y = 200850x + 62822
150000
100000
50000
0
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
August 2005
Page 21 of 56
The numerical results also allow to visualise the yield mechanisms (Figure 24). For high diameters
models, these numerically obtained mechanisms may be compared to results of plastic plate yield
theories. Finally they should allow to better understand the correlations between shell and plate
yield line patterns.
The numerical results have been compared to the current design formulae for CHS connections
implemented in Part 1.8 of Eurocode 3 or in the design Guide book published by CIDECT [41].
The relevant formula is the following :
N* = f() . f() . f(n) . fc,y . tc
with f() =
5.0
1 0.81
f() = 1
f(n) = 1
Figure 25 shows how the plastic resistance varies with the parameter.
(8)
August 2005
Page 22 of 56
400000
350000
300000
Npl (N)
250000
200000
150000
100000
50000
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
0,45
FEM t=8mm
FEM t=6,3mm
FEM t=10mm
FEM t=12mm
Annex K t=8mm
Annex K t=6,3mm
Annex K t=10mm
Annex K t=12mm
The Gomes model has been developed to study the behaviour of minor axis joints between I or H
profiles. Present research focuses on another component. Therefore, the parameters used by
Gomes in his formulation have to be interpreted differently to be in agreement with the CHS
problem.
The Gomes model for minor axis joint is:
Npl =
4 m pl
b
1
L
b 2c
1 +
L L
(9)
August 2005
Page 23 of 56
L is the most complicated parameter to define. In the Gomes model, the parameter L is fixed
by the geometrical layout of the joint (L = width of the column web). In the case of the CHS
chord, an equivalent parameter is not so obvious to define. Indeed, the plastic mechanism is
not limited by flanges, as for an I profile, and the length of the mechanism is, at first sight,
unknown. Therefore, for the developments of the next paragraph, the Gomes model will be
used with L equal to the length of the yield area defined by FEM simulations projected on a
plane surface. This will allow to compare and understand the behaviour of the yield mechanism
in comparison with the Gomes model. But to derive the final analytical formula, a fixed
reference has to be chosen and L will be taken equal to the diameter of the CHS. Obviously, a
corrective factor will have to be used.
c is the thickness of the plate but as it is showed in Figure 14, this parameter can not be taken
into account by the numerical simulations. Therefore, as a first approach, this parameter is
neglected in the developments of the analytical formulation. The following example proves that
the error made by neglecting this term is not important.
A case with a b/L and a very thick plate very high would be the most severe. For this example, L
is taken as the diameter of the tube.
Diameter of the CHS: L = 100mm
Width of the plate: b = 80mm
Thickness of the plate: c = 12mm
b
= 0,447
1
L
2c
= 0,076
L
The second term
I.3.4.2
2c
can therefore be reasonably neglected in comparison with
L
b
.
L
In Figures 27 and 28, the results of the numerical simulations have been compared to the Gomes
2c
model. As explained in the previous paragraph, the term
has been neglected.
L
L = L projected on a
plane surface
August 2005
Page 24 of 56
For this comparison, the parameter L has been taken equal to the length of the yield mechanism
obtained numerically, projected on a plane surface (Figure 26). The Gomes curves on Figures 27
and 28 are not pure analytical curves as they depend on a numerically determined parameter (L).
The value of L obtained numerically has a physical meaning and this allows to draw interesting
conclusions and see whether the application of the Gomes model for CHS walls has a sense.
The analysis of the graphs shows that the shapes of the numerical curves and the shapes of the
Gomes curves are the same. That proves that the Gomes model is a reliable background from
which the analytical formulation for the CHS wall component can be derived.
400000
350000
300000
Npl (N)l
250000
200000
150000
100000
50000
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
0,45
Gomes t=8mm
Gomes t=6,3mm
Gomes t=10mm
Gomes t=12mm
FEM t=8mm
FEM t=6,3mm
FEM t=10mm
FEM t=12mm
Npl (N)
250000
200000
150000
100000
50000
0
0
50
100
150
200
250
20
Gomes t=8mm
FEM t=8mm
Gomes t=6,3mm
FEM t=6,3mm
Gomes t=10mm
FEM t=10mm
Gomes t=12mm
FEM t=12mm
300
August 2005
Page 25 of 56
The graph Npl() also shows that the plastic resistance tends to a limit value when increases. It
can therefore be supposed that, for an infinite diameter, a single value of the plastic resistance, not
depending of any parameter, is reached.
In conclusion, the Gomes model can be considered as an adequate basis for the new analytical
formulation. The correcting factor, which will allow to modify the Gomes model so as to get results
close of the numerical resistances has now to be found.
Remark: Obviously for the final analytical formulation, the parameter L can not be equal to a value
taken from numerical simulations. The formula relies on a known value of L. L could be taken as
equal to the diameter of the CHS in a first stage.
I.3.4.3
It would be interesting to define the limit of the plastic resistance in the extreme case where the
diameter of the CHS section is equal to infinite.
If the diameter is equal to infinite, the CHS becomes an infinite plate.
The analytical solution may be found by referring to the Johanssen plastic theory (Figure 28):
Ppl = mpl [2b 1/X + 2L 1/X + 2 + 2 + 4(L/2-b/2) 1/X]
(10)
P
L = 0
=0
(11)
L
Figure 29: Plastic mechanism on an infinite plate
The limit of the plastic resistance when the D = equals 4mpl.
In order to illustrate this limit value, the evolution of the plastic resistance with the diameter of the
tube have been reported in Figure 30. In this case, the Gomes model has been used with L=
diameter of the tube. The values of the resistance have been divided by the plastic moment (mpl).
In such a way, all the curves calculated with the Gomes model do not depend on the thickness
August 2005
Page 26 of 56
anymore. Therefore one single reference curve represents the Gomes model. The value
4 m pl
mpl
= 4 has also been reported on the graph and it can be observed that all the values of
Npl/mpl
30
25
20
15
10
5
0
100
300
500
700
900
1100
1300
1500
1700
1900
FEM t=8mm
FEM t=6,3mm
FEM t=10mm
FEM t=12 mm
4pi
Part II
II.1
August 2005
Page 27 of 56
Volume 1 presents a full consistent approach for the design of joints between open and/or hollow
sections. This approach is based on the component method, i.e. for each individual component the
main properties (stiffness, resistance) should be determined and in a second step these properties
will be assembled to obtain the joint resistance and stiffness respectively. For quite a number of
components detailed rules for the determination of the component properties are given. As these
rules are taken from existing standards or other technical publications the validity of the rules has
already been proved. However, if rules for new components will be developed, it is necessary to
validate such new models at the level of the component and at the level of the joint.
Part I of the present Volume 2 presents the progress of the scientific works achieved during the
project with respect to the behaviour of CHS components. Other components have been studied in
other CIDECT research projects as for example 5BH [54], 5BS [55], 8D/E [56], 8G [57].
In order to be able to validate such mechanical models at the joint level, i.e. the use of a new
component in any joint configuration, a more general tool would be required to validate the
mechanical model with test results.
This Part II of Volume 2 presents the progress achieved during this project with respect to the
development of such a tool.
In order to validate the mechanical models both test results from literature and numerical
simulations can be used. To compare the joint characteristics with test results, a database on test
results has been created. This work is briefly presented in section II.2. Several test results were
added since the second interims report [3].
Finally a study with FE simulations can be performed in order to validate those configurations
where no experimental tests are available. With the aim to cover a large number of different joint
configurations, a specific software tool has been developed. With this tool FE models of hollow
section joints with different dimensions of the members and different material laws can be
generated in a very flexible way. Section II.3 contains more information concerning this FE model
generator.
In a next step various FE models have to be calibrated against experimental tests, to ensure the
accuracy of the numerical models (see section II.4). In this step the test results of the database can
be used.
II.2
August 2005
Page 28 of 56
In the literature quite a number of test results can be found. However, in most cases the
description of the individual data of such experiments is not complete. In order to validate design
models with experimental tests, it is required that the following data are available:
For the present research project, literature containing documentations on test results have be
reviewed. Tests which are sufficiently complete documented have been collected and a database
on test results was created. In order to prepare an more automatic handling of these data, some
software tools have been developed to create a visual representation of the joint configuration of
the tests on one side and to generate a preliminary FE model from the basic data on the other
side. An overview of the tests collected is given in the following tables. In total 320 tests have been
introduced in the database. Of course, not all test are fully in line with the scope of the project, but
it is hoped that the consideration of those tests may also be helpful for the validation of the models
in development.
August 2005
Page 29 of 56
Column
EHS
Beam(s)
Tests
F--diagram
Loading
Compression
Tension
EHS
=6
Table 2: Joint configurations of experimental test from reference [35]:
Joint
Column
CHS
Beam(s)
CHS
Tests
F--diagram
Loading
Compression
Tension
10
Bending
Compression +
bending
12
Tension + bending
= 41
Table 3: Joint configurations of experimental test from reference [36]:
Joint
Column
Beam(s)
Tests
F--diagram
Loading
CHS
CHS
Compression
CHS
CHS
Compression
Joint
August 2005
Page 30 of 56
Column
Beam(s)
Tests
F--diagram
Loading
CHS
CHS
Compression
= 12
Column
CHS
CHS
CHS
CHS
Beam(s)
Tests
F--diagram
Loading
1x
Compression
1x
Tension
14
Compression
Tension
12
Compression
21
Tension
2x
Bending
Compression
Tension
Bending
flat steel
(horizontal)
flat steel
(horizontal)
flat steel
(vertical)
flat steel
(vertical)
Joint
August 2005
Page 31 of 56
Column
CHS
CHS
CHS
Beam(s)
Tests
F--diagram
Loading
Compression
Tension
Compression
Compression
Tension
2x
Bending
cross
cross
I-profile
CHS
I-profile
Compression
CHS
RHS
Compression
CHS
RHS
Compression
= 124
August 2005
Page 32 of 56
F--diagram
Loading
Combinations of
compression
and tension
Compression
Tension
8
(4 with
floor)
Bending
Different
combinations of
compression
and tension
Compression
Tension
8
(4 with
floor)
4x
Bending
Column
Beam(s)
Tests
CHS
flat steel
CHS
CHS
RHS
RHS
RHS
double
flat steel
I-profile
flat steel
double
flat steel
I-profile
= 38
August 2005
Page 33 of 56
Column
RHS
Beam(s)
Tests
F--diagram
Loading
86
15 x
Bending
13
Bending +
compression
RHS
= 99
II.3
In order to prepare a parametrical study a flexible tool to generate various FE models of hollow
section joints had to be developed. For the FE-simulation the program system ABAQUS was
chosen. This FEM simulation software is suitable because of the interface of the interactive
version, which allows to generate FE models with an integrated program language called Python.
8 different types of T-joint configurations with hollow section members were considered, because
of their typical, often used basic components. Examples of the chosen configurations are given in
Figure 31.
Figure 31: T-joint configurations, which were modelled with the FE model generator
August 2005
Page 34 of 56
For the 8 given joint configurations all geometrical properties, as length of the column, height and
width of the beam, thickness of profiles, thickness of the welds, etc., can be modified. If an
examination of the influence of 1 parameter is required, all the other parameters can be fixed, while
the parameter of interest is varied in a given increment. The tool generates all the joints with the
different geometrical properties and the simulation to get stress-displacement curves can be
started.
The material behaviour of the column and the beam in the simulation can be described by a bilinear or a tri-linear stress-strain curve (see Figure 32). The properties of the material have to be
provided.
August 2005
Page 35 of 56
II.4
August 2005
Page 36 of 56
To be sure to get reliable results in the FE simulations and to ensure realistic results of the
simulations, which are used to replace real test results, the numerical models have to be
calibrated. For this purpose the influence of the shell type, element size, solver algorithm etc. on
the simulation results have to be examined, and the results of the numerical simulations have to be
compared with test results. It is the aim, to create robust FE models, that reproduce the loaddisplacement curves of real test results as exact as possible.
The first test results that were used to calibrate the FE model are based on the work of the
university of Karlsruhe (see [40]; Table 6). In this study T-joints with RHS columns and RHS beams
were tested. The specimen were fixed at the ends of the column and the loads were applied at the
end of the beams. The displacements were measured at the point, where loads were applied, in
direction of the loads. A principal sketch of the experimental setup is given in Figure 36.
Figure 36: Setup for T-joint with RHS members test to calibrate the FE models
August 2005
Page 37 of 56
1300
The test M21, that was reproduced by FE modelling, has the following geometrical properties:
460
[mm]
h [mm]
b [mm]
t [mm]
Column
140
140
8,8
Beam
100
100
4,0
where:
Joint member
fy [N/mm]
fu [N/mm]
f [-]
Column
335,7
528,9
0,2885
Beam
424,6
527,1
0,33375
August 2005
Page 38 of 56
As the strain when the material reaches the ultimate tensile strength fu is not given, as an
approximation different material laws for the beam and for the column are used, see Figure 38 and
Figure 39. An initial stiffness of E=210000 N/mm is assumed.
600
500
stress [N/mm]
400
Curve 1
300
Curve 2
Curve 3
200
Curve 4
100
0
0
0,05
0,1
0,15
0,2
0,25
0,3
strain [-]
Figure 38: Different approximations of the material law for the column of model M21
600
500
400
stress [N/mm]
Curve 5
Curve 6
300
Curve 7
Curve 8
200
100
0
0
0,05
0,1
0,15
0,2
0,25
0,3
strain [-]
Figure 39: Different approximations of the material law for the beam of model M21
0,35
August 2005
Page 39 of 56
8 node shell elements (S8R-elements in the ABAQUS-library) and 4 node shell elements (S4Relements in the ABAQUS-library) with reduced integration were tested. The side length was varied
between 8 and 4 mm.
By taking thicker shell elements with the width of 0 ,8 a 2 , the welds are considered.
All displacements and loads are plane-symmetric. That means, that modelling only one half of the
joint should lead to correct results, if the required boundary conditions are set.
As an example one model with an approximate shell element length of 8 mm is given in Figure 40.
In Table 9 the main parameters for the different FE models of test M21 are summarized and the
changes are marked. The - curves are in accordance with those given in see Figure 38 and
Figure 39.
Table 9: Main parameters of the FE model M21
Model name
FEM 21-5
FEM 21-6
FEM 21-7
FEM 21-8
FEM 21-9
FEM 21-10
FEM 21-11
Element type
S8R
S4R
S4R
S8R
S8R
S4R
S4R
Approximate size
length [mm]
8
8
5
8
8
5
5
- curve
column
Curve 1
Curve 1
Curve 1
Curve 2
Curve 3
Curve 3
Curve 4
- curve
beam
Curve 5
Curve 5
Curve 5
Curve 6
Curve 7
Curve 7
Curve 8
Further
modifications
-
August 2005
Page 40 of 56
The comparison of the test results and the results of the numerical simulations of M21 is given in
Figure 41, where P is the applied load, see Figure 37, and s is the deformation at the point of load
introduction.
45
40
35
30
Test results
P [kN]
25
FEM M21-5
FEM M21-6
20
FEM M21-7
FEM M21-8
15
FEM M21-9
FEM M21-10
10
FEM M21-11
5
0
0
10
15
20
25
30
35
40
s [mm]
August 2005
Page 41 of 56
The measured properties of test specimen M10 are given in Table 10 and Table 11.
Table 10: Geometrical properties of specimen M10
Joint member
h [mm]
b [mm]
t [mm]
Column
100
100
6,3
Beam
80
80
3,6
fy [N/mm]
fu [N/mm]
f [-]
Column
242,5
353,1
0,3725
Beam
298,5
392,9
0,3706
350
300
stress [N/mm]
250
Curve 1
Curve 2
200
150
100
50
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
strain [-]
Figure 42: Different approximations of the material law for the column of model M10
0,4
August 2005
Page 42 of 56
450
400
350
stress [N/mm]
300
250
Curve 3
200
Curve 4
150
100
50
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
strain [-]
Figure 43: Different approximations of the material law for the beam of model M10
The main parameters of model M10and their variation are summarized in Table 12.
The load displacement curves and the comparison to the test result are given in Figure 44.
Element type
S8R
S4R
S4R
Approximate size
length [mm]
6
6
5
- curve
column
Curve 1
Curve 1
Curve 2
- curve
beam
Curve 3
Curve 3
Curve 4
Further
modifications
-
August 2005
Page 43 of 56
20
18
16
14
P [kN]
12
10
8
Test results
FEM M10-1
FEM M10-2
4
FEM M10-3
2
0
0
10
20
30
40
50
60
s [mm]
h [mm]
b [mm]
t [mm]
Column
180
180
14,2
Beam
100
100
6,3
fy [N/mm]
fu [N/mm]
f [-]
Column
227,7
380,1
0,324
Beam
242,5
353,1
0,3725
August 2005
Page 44 of 56
In this model a stress-strain curve with a reduced at the ultimate strength level (1/4 of f) is used
directly, because of the results of the former tests.
The used curves for the column and for the beam are given in Figure 45 and Figure 46.
400
350
300
Curve 1
stress [N/mm]
250
200
150
100
50
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
strain [-]
350
300
stress [N/mm]
250
200
Curve 2
150
100
50
0
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
strain [-]
0,4
August 2005
Page 45 of 56
FEM 78-1
FEM 78-2
FEM 78-3
S4R
S4R
S8R
Approximate size
length [mm]
10
5
10
FEM 78-4
S4R
Model name
Element type
- curve
column
Curve 1
Curve 1
Curve 1
- curve
beam
Curve 2
Curve 2
Curve 2
Curve 1
Curve 2
Further
modifications
The whole
joint was
modelled
(not only one
half with axis
symmetric
boundary
conditions)
60
50
P [kN]
40
30
Test results
FEM M78-1
20
FEM M78-2
FEM M78-3
FEM M78-4
10
0
0
10
15
20
25
30
35
40
s [mm]
August 2005
Page 46 of 56
August 2005
Page 47 of 56
The next test, that are reproduced by FE simulation are part of the work presented by Y.
Makino, Y. Kurobane, J.C. Paul, Y. Orita and K. Hiraishi in 1991 (see [37] and Table 4).
In this tests X- and T-joints were examined under normal- and bending loads of the beam. The
column was always a CHS-profile with different geometrical properties, while the profile types of
the beams were diversified.
The joint that was reproduced by FE modelling was a X-joint with horizontal gusset plates under
tension and compression. The geometrical properties of the joints are given in Figure 50 and
Figure 51.
August 2005
Page 48 of 56
400
350
300
P [kN]
250
Test results
200
FEM XP1Compression
150
100
50
0
0
10
15
20
25
30
s [mm]
Figure 53: Load displacement curves of X-joint XP1-C-1 under compression (FEM and test
results)
With regard to the ultimate resistance the FE model doesn't fit exact with the test results. But as
said above, there is poor information about the material behaviour. The buckling of the CHS profile,
can be identified in the FE simulation. The deformed specimen is given in Figure 54.
August 2005
Page 49 of 56
700
600
P [kN]
500
400
Test results
300
FEM XP1Tension
200
100
0
0
10
15
20
s [mm]
Figure 55: Load displacement curves of X-joint XP1-T-1 under tension (FEM and test
results)
25
August 2005
Page 50 of 56
After analysing all FE simulations carefully, the results lead to the conclusion, that real test results
can be reproduced by FE modelling in a sufficient way. If there is enough information about the
material behaviour and the test setup, the load displacement curves derived from numerical
simulations are nearly identical with those ones measured at real tests.
Effects like buckling and yielding can be simulated, and the computation runs reliable, if the
ABAQUS shell elements S4R are used. In its final version the FE model generator uses for all
required joint configurations shell elements S4R. The size length has to be specified adequate.
August 2005
Page 51 of 56
References
[1]
[2]
[3]
[4]
P. ANSOURIAN
Rigid-frame connections to concrete-filled tubular steel columns
CRIF report, MT 86, Belgium, January 1974.
[5]
[6]
D. VANDEGANS
Liaison entre poutres mtalliques et colonnes en profils creux remplis de bton, base sur la
technique du goujonnage (goujons filets)
CRIF report, MT 193, Belgium, October 1995.
[7]
[8]
F.C.T. GOMES
Etat-limite ultime de la rsistance de lme dune colonne dans un assemblage semi-rigide
daxe faible
Universit de Lige, MSM Department, Internal report N 203, Belgium, August 1990.
[9]
Y.W. KIM
The behaviour of beam-to-column web connections with flush end plates
Ph. D. Thesis, University of Warwick, Department of Engineering, U.K., July 1988.
August 2005
Page 52 of 56
August 2005
Page 53 of 56
[24] R. PUTHLI
Hohlprofilkonstruktionen aus Stahl
Werner Verlag, 1998 ISBN 38-04129-75-7.
[25] F. GRIMM
Konstruieren mit Walzprofilen
Verlag Ernst und Sohn, 2003 ISBN 3-433-02840-0.
[26] F. GRIMM
Konstruieren mit Hohlprofilen
Verlag Ernst und Sohn, 2003 ISBN 3-433-02833-8.
[27] L.F.C. NEVES, F.C.T. GOMES
Semi-rigid behaviour of beam-to-column minor axis joints
September 1996 Proceedings of the LABSE Colloquium, Istanbul, Turkey, p. 207 216.
[28] C. BRASSIER
Caractrisation du comportement en rotation dassemblages souds de construction
mtallique en poutre en I et poteau tubulaire de forme circulaire Dveloppement dune
mthode analytique
June 2003 Diploma work, Department M&S, University of Lige CUST Clermont-Ferrand.
[29] J. MARRA
Assemblage dans les poutres en treillis tubulaire de section elliptique
June 2003 Diploma work, Department M&S, University of Lige.
[30] C. PIETRAPERTOSA
Etude du comportement des assemblages souds de poutres en treillis constitus de
profils mtalliques tubulaires de forme elliptique
June 2002 Diploma work, Department M&S, University of Lige.
[31] J. WARDENIER
Hollow section joints
1982 Delft University Press, Delft (The Netherlands).
[32] J. WARDENIER
Hollow section in structural applications
CIDECT Bouwen met Staal, 2002 ISBN 90-72830-39-3.
[33] Experimental test EHS joints, Internal report, Univ. Lige 2003.
[34] S. WILLIBALD, J.A. PACKER, R. S. PUTHLI
Bolted connections for RHS tension members
University of Toronto
Draft Final Report, CIDECT Project: 8D / 8E, Report 8D / 8E 8/02.
[35] Y. KUROBANE, Y. MAKINO, K. OGAWA, T. MARUYAMA
Capacity of EHS-joints under combined OPB and axial loads and its interactions with frame
behaviour
Kumamoto University, Japan
Tubular Structures, 4th International Symposium, Delft 1991.
[36] J.C. PAUL, T. UENO, Y. MAKINO, Y. KUROBANE
The ultimate behaviour of circular multiplanar TT-joints
Kumamoto University, Faculty of Engineering, Japan
Tubular Structures, 4th International Symposium, Delft 1991.
August 2005
Page 54 of 56
August 2005
Page 55 of 56
August 2005
Page 56 of 56