Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
La Poutre, D.B.
Published: 01/01/2000
Document Version
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
• A submitted manuscript is the author's version of the article upon submission and before peer-review. There can be important differences
between the submitted version and the official published version of record. People interested in the research are advised to contact the
author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Since the first proposal for this project many people have reacted enthusiastically. I
am very indebted to all those people and institutions that have supported me to make it
a success. Special thanks go to Salzgitter AG for donating channel sections and to
Delta Staal BV for delivering them free of charge.
Many thanks to Prof. If. H.H. Snijder and Dr. if. J.C.D. Hoenderkamp for supervising
this project. I wish to record my appreciation of TNO Building and Construction
Research for its financial support for this research project. Thanks are due ir.
H.M.G.M. Steenbergen and ir. F.S.K. Bijlaard for the useful remarks.
Thanks to the people of the laboratory for helping me executing the experiments. In
particular I wish to express my gratitude to Martin Ceelen for setting up the
measurement system and patiently explaining how it works.
Dagowin la Poutré
April 1999
This report has been reprinted to meet demands and to be able to publish it in the
online electronic library of the Eindhoven University of Technology
(http://www.tue.nl/bib/).
For this version some minor errors in typing have been corrected as well as some
incorrect numbers. Specifically, the row starting with ‘ωkip’ in table 3 has been
corrected. These numbers were typed incorrectly and do not effect the rest of the table.
Non of these changes effect either the experimental results or the conclusions.
Dagowin la Poutré
October 2000
i
ii
Summary
In literature very little data on the structural behaviour of channel sections is found. To obtain data
on the structural behaviour of channel sections a serie of experiments have been set-up. Channel
sections with a span of 2.8 meters are loaded by two concentrated loads acting at a quarter span
from each support, see figure A.
s.c.
e
y
+φ e
c.g.
2.8 m
z x
To test the effect of the point of load application load has been applied at the shear centre, at the
top flange, in the middle of the web and at the bottom flange.
If the load is applied at the shear centre the section buckles suddenly when the ultimate load is
reached. The failure load is also highest when the load is applied at the shear centre.
Sudden buckling does not occur when the load is applied onto the web. Furthermore the failure
load is lower than if the load is applied at the shear centre, see figure B.
Experiments 2XX
50
2A1
30
2B2
Fu;2B = 43.7 = 88% Fu;1B 2C1
Fu;1B= 49.4 [kN] 20
1 st order
10
e
Fu;2C = 47.3 = 96% Fu;1B 0
0 10 20 30
w 1 [m m ]
Vertical deflection at m idspan
figure B: failure load in relation to load at shear centre and load-displacement graphs of experiments
The failure loads have been predicted using the Merchant-Rankine postulate. The predictions were
quite conservative and need further improvement.
iii
iv
Table of contents
INTRODUCTION...............................................................................................................................3
OBJECTIVES .......................................................................................................................................4
RESTRICTIONS ...................................................................................................................................4
BASIC ASSUMPTIONS .........................................................................................................................4
1 CHOICE OF EXPERIMENT ....................................................................................................5
1.1 SELECTED SECTION................................................................................................................5
1.2 LOAD CASES ..........................................................................................................................6
1.3 TYPE OF LOADS......................................................................................................................8
1.4 POINT OF LOAD APPLICATION ................................................................................................9
1.5 NUMBERING OF EXPERIMENTS ............................................................................................10
2 MATERIAL PROPERTIES ....................................................................................................11
REFERENCES ..................................................................................................................................50
APPENDICES .....................................................................................................................................1
A. SECTIONS...............................................................................................................................3
UNP SECTIONS ..................................................................................................................................6
1
UPE SECTIONS ...................................................................................................................................7
UAP SECTIONS ..................................................................................................................................8
PFC SECTIONS ...................................................................................................................................9
B. TENSILE TESTS ..................................................................................................................11
SPEED OF INCREMENTS ....................................................................................................................12
YIELD RANGE...................................................................................................................................13
ELASTIC RANGE (STRAIN MEASUREMENTS) ....................................................................................15
RESULTS ..........................................................................................................................................15
C. LATERAL TORSIONAL BUCKLING..............................................................................16
THEORETICAL ELASTIC LATERAL TORSIONAL BUCKLING LOAD ......................................................16
ULTIMATE ELASTIC-PLASTIC LATERAL TORSIONAL BUCKLING LOAD .............................................16
D. BENDING AND TORSION .................................................................................................18
BENDING ..........................................................................................................................................18
TORSION DUE TO SHEAR FORCE .......................................................................................................22
E. INTRODUCING LOAD ON SECTION .............................................................................29
H. CALIBRATION ....................................................................................................................49
USED EQUIPMENT ............................................................................................................................49
I. DIMENSIONS ...........................................................................................................................58
RESTRICTIONS ON DIMENSIONS .......................................................................................................58
2
Introduction
This project is part of the curriculum of the study of building and construction of the Eindhoven
University of Technology. At the ninth trimester students have to perform an exercise in either
designing and calculating a building, doing numerical research or experimental research.
This project is mainly comprised of experiments, however analytical and numerical calculations were
to be carried out as well for the execution of the experiments.
Chosen is to perform experiments on channel sections because they have quite a different structural
behaviour from other beam elements, especially where stability is concerned. Since very little has
been published about this subject the experiments are performed to obtain a better understanding of
the structural behaviour. The experimental data can also be used for comparison with finite element
analyses.
To predict the load at which failure occurs it is proposed to use the Merchant – Rankine postulate.
This postulate is used to predict the failure of frames by determining the theoretical buckling load and
the plastic capacity of the frame.
For channel sections it is possible to determine both the plastic section capacity and the theoretical
buckling load. With these two dimensions the critical load, at which failure would occur, is predicted.
In figure 1 it is shown how the failure load is predicted for a channel section. In the same picture it is
shown how failure is simulated by finite element method (FEM) simulation. The experimental data
will prove if this is a productive approach.
More on the Merchant – Rankine postulate can be found in Over Spannend Staal [7] and other
literature.
Load -deflection
40
F-plastic
35
30 F-buckling
1st order
load F [kN]
25
FEM
20 F-critical
15
10
failure
5
0
0 10 20 30 40 50
deflection [mm]
This report follows largely the chronology of the project, which is illustrated in figure 2. This time
frame depicts the progress as it took place, not as it was planned out.
3
Objectives
• Studying the structural behaviour of channel sections, specifically stability
In literature very little data on the structural behaviour of channel sections is found. The experiments
will yield load – displacement curves revealing the specific structural behaviour of channel sections.
With analytical models first order approximations can be made to the structural behaviour of channel
sections. With finite element analyses the geometrically and physically non linear behaviour can be
simulated. Both methods need to be validated by experimental data.
• Repetitive testing of each kind of experiment to obtain mean and variation on failure load
Performing one test of each kind of experiment would yield data on structural behaviour but would
leave out certainty about the failure load, i.e. the one experiment might be, due to many reasons, an
exception to normal behaviour.
It is proposed to use the Merchant – Rankine postulate to predict failure loads. These experiments
will either confirm or exclude the possibility of using this verification method.
Restrictions
• Members subjected to shear force, bending and torsion
The effect of a lateral load, causing bending and torsion, is analysed only. No axial load is applied.
• Load must be applied in such manner that no lateral support is given or failure is precipitated
The stability of the member is studied. On one had the member must be free to rotate and deflect
laterally otherwise it may be fail at a higher load. On the other hand the line of work of the load must
remain vertical as not the precipitate failure.
In beam theory sectional properties do not alter under loading. This theory is used to analyse the
behaviour of the sections and therefor the sections must retain it shape and no local buckling may
occur.
Basic assumptions
• The span will be chosen in the range of ten to thirty times the height of the section
4
1 Choice of experiment
Hot-rolled sections are available in different series: with parallel flanges (UPE-, UAP- and PFC-
series) and with tapered flanges (UNP- serie). The UNP sections have a disadvantage in making
connections, such as gables, due to the tapered flanges, see figure 3. However they are widely
available. UPE, UAP and PFC sections can make connections easier due to the parallel flanges, see
figure 5. They are more expensive than UNP sections and are usually not in stock.
figure 5: connecting a parallel flange channel (from [8]) figure 4: comparison between UPE and IPE sections (from [8])
These four series of channel sections are compared in figure 6 for sections having a height of 100
millimetres and it can be noticed that the difference is very little except for the UNP section.
In table 1 the programmes of available heights are shown: they are different for all series. The UPE-
programme has heights that correspond to IPE-sections, see figure 4.
5
table 1: heights of hot-rolled channel sections (heights with a * are available in two widths)
height [mm]
UNP 80 100 120 140 160 180 200 220 240 260 280 300 320 350 350 380 400
UPE 80 100 120 140 160 180 200 220 240 270 300 330 360 400
PFC 100 125 150* 180* 200* 230* 260* 300* 380* 430
UAP 80 100 130 150 175 200 220 250 300
The UPE section is chosen because it’s dimensions are associated with the widely used IPE-sections
and because sections with parallel flanges will probably be used more often in the future.
A B C
A’ C’
B’
A
sheeting wall
A’
section A-A’
The objective of the experiments is to test the stability of channel sections. Thus the way in which
load is applied may not facilitate lateral support. In the laboratory it is virtually impossible to
generate a distributed load that can move freely with the deformations of the section. Therefore
concentrated loads must be used.
6
In order to determine the number of concentrated loads the mechanical consequences should be
considered. The more loads the more complicated this will be and therefore only one and two
concentrated load configurations will be considered.
Single load: The maximum bending moment coincides with the maximum shear force. If torsion is
present in the member the sum of warping restraint torsion and Saint Venant-torsion
has a distribution which is identical to the shear force distribution. However, warping
restraint torsion dominates at midspan whereas Saint Venant-torsion is largest at the
supports, see figure 9.
Mx
F
warping free to occur,
X [mm] axial rotation prevented
0,5 L 0,5 L 2
Saint Venant-torsion
Double loads: The maximum bending moment coincides with the maximum shear force at the points
of load application but between those point shear force is absent. If torsion is present
the distribution is similar to that of the shear force distribution, which means that no
torsion is present between the points of load application. However, the graph shows
that the Saint Venant-torsion and warping torsion are not equal to zero. They are
equal but opposite in sign and thus they do not resist the externally applied torsion,
see figure 10.
Mx Mx
F F
X [mm]
0,5 L 0,25 L 2
Saint Venant-torsion
warping restraint-torsion
shear force distribution 0
Mx summation of torsion
[kNm]
moment distribution -2
0 700 1400 2100 2800
figure 10: distribution of forces through member with two concentrated loads
From these two load configurations the second one is selected. The constant value of the bending
moment between the points of load application makes it best suitable for testing the stability
phenomena. An extreme value might cause large deformations due to yielding before the section
buckles.
7
1.3 Type of loads
The load can be applied in two ways: one using dead weight and the other using jacks. These two
methods shall be discussed briefly.
Dead weight: By using dead weight one can be certain that the direction of the load doesn’t change,
see figure 11. However, applying loads gently and balancing them evenly over the
two points of application might be difficult. Another difficulty is the volume of
weights needed to apply relatively large loads.
Figure 12 shows the design of a test rig for experiments using dead weight. The load
bearing yokes can move freely with de deformation of the section, hinging about the
point of load application. The load is comprised of concrete blocks that are laid down
on the two yokes. A possible imbalance between the two yokes will be registered by
the load cells.
Jacks: Using jacks the problem emerges that one side of the jack is pinned. When lateral
displacements occurs the direction of the load changes, see figure 11.
A distinction can be made in tension and pressure jacks; tension jacks pull the section
back by a horizontal force and thus supporting the section against buckling, whereas
pressure jacks pushes the section outwards and thus precipitating buckling. The
magnitude of this effect depends on the length of the rod that connects the jack with
the section as well as the amount of lateral deflections.
The advantage of jacks is that loads can be applied smoothly, balancing can be done
quit accurately and relatively large loads can be generated easily.
F F
F Fvert
F Fvert
1 2 1 2 1 2 2 = deformed state
The initial choice was to use dead weight due to the apparent advantages over jacks. The maximum
amount of load that could be applied was about 6 tons. In order to obtain failure of the section three
criteria had to be met: the span couldn’t be too short, the rigidity of the section must not be too large
and the yield stress low. In the proposal for this project UPE sections with a height between 100 and
200 mm and a span between 10 and 30 times their heights have been considered for use. From these
considerations the section with a height of 160 millimetres was chosen.
Beforehand, uncertainty remains about one parameter, the yield stress. The quality indication, such as
S235JR for example, given by the manufacturer of the product is a guaranteed lowest estimated
value. The actual value is usually much higher. To by-pass this uncertainty members of 12 meters
length were ordered and were not cut until the yield stress had been determined and a decision was
made about the span of each experiment. The tensile tests to determine the yield stress will be
discussed later on in this report.
8
line of work
load cells
concrete weight
Designing of the test rig, figure 12, and a number of special fittings to apply load onto the section had
commenced. However during this process objections were raised against using dead weight and it
was decided to go-ahead with jacks instead. An entirely different rig had to be designed and a well-
founded decision to use either pressure or tension jacks had to be made. This is reported in
chapter 4 Test rig.
F F S = shear centre
F C = centre of gravity
y
w = bending deformation
⇒ w ⇒ w
S C S C φ = angle of twist
φ F
In figure 7 it can be seen that nearly all connections with channel sections are eccentric. It makes a
difference whether the load is applied at the top flange, y-axis or bottom flange, which will be
discussed later. It has been decided to apply loads at all three points of the cross section.
9
Even though centric loading is uncommon in practice, in an experiment it yields valuable information
about the stability phenomena. When the section is loaded at the shear centre it will bent and at a
certain load it suddenly buckles by torsion and lateral flexure. The point at which this happens is
called the bifurcation point. At this point an unstable equilibrium path of continued in plane bending
(bending about the y-axis, without torsion or lateral deflection) is also possible, see figure 14.
If the section is loaded eccentrically the bifurcation point will not appear in the load-deformation
graph. Therefore it was decided to include an experiment with centric loading to compare the
structural behaviour and failure loads to those of an eccentrically loaded section.
1A 2A
member 1: tensile test 2B1 2B2 2B3
1B 2B
shear centre centre of gravity member 2: 1B1 1B2 1B3 1B4
y-axis
1C 2C member 3: 2A1 2A2 2C1 2C2
3 [m]
z-axis 12 [m]
figure 16: points of load application figure 15: order in which specimens were cut from the members
10
2 Material properties
The value of the yield stress is needed in the analytical calculations and must be determined first.
The material is of quality S235JR as described by NEN-EN 10025 [4]. To determine the yield stress
specimens must be taken from the channel sections.
The proportions of the dimensions of the specimens are obtained from EN 10 002-1 [3] and are
applied on the UPE 160 section, resulting in the dimensions given in figure 17.
b/3 = 22.3
44 20
b=
Lo = 79.9 +/- 0.8 2/3 b = 47.3 70
R12
50 12 Lc = 110 12 50
Lt = 234
Four tests were carried out from which the average yield stress was determined. In figure 18 (a) a plot
of the second test is shown. The yield range is enlarged to make the fluctuations better visible.
400 305,9
Stress [N/mm ^2]
305
300 ReH =
enlarged area
302,4
200
300
ReL =
100 298.5
Min. =
0 295,9
295
0 0.1 0.2
0 0.008 0.016
strain [%] strain [%]
(a) (b)
figure 18: tensile test and yield range
Code EN 10 002-1 makes a distinction between the upper yield stress, ReH, and the lower yield stress,
ReL. The upper yield stress is described as ‘the stress at which the first clear descent of stress is
observed’ and the lower yield stress is ‘the lowest stress that occurs in the yield range, disregarding
brief irregularities.’ The upper yield stress can be determined in a straight forward manner but the
lower yield stress is a bit more ambiguous to determine. The crosses in the graph of figure 18 (b)
resemble points that have been measured by an evenly spaced time interval. Thus the number of
crosses per equal sized line segment gives the speed of yielding, and from the density of crosses the
speed can be read. In this test the lower yield stress is fitted as shown but could have been fitted
through the two lowest curves since they do have a reasonable number of crosses in comparison.
11
table 2: results tensile tests
Test fy;max fy;min ReH ReL Rm
1 298.2 290.8 298.2 290.8 443.8
2 305.9 295.9 302.4 298.5 446.9
3 310.1 300.5 310.1 302.6 452.6
4 301.6 291 298.2 291.1 447.1
average 304 294.5 302.2 295.7 447.6
deviation 5.17 4.63 5.62 5.8 3.65
The results of all four tests are given in table 2 and it is noted that there is only a small difference
between the lower yield stress ReL and the lowest yield stress fy;min. The lower yield stress ReL will be
used in the calculations in the next chapter.
Problems occurred by measuring strain and therefore the values of strain should be disregarded.
12
3 Analytical calculations
3.1 Experiment 1
The objective of this experiment is to determine the lateral torsional buckling moment. In order to do
so at least three tests must be carried out to determine a mean value and deviation. If experiments 1A,
1B and 1C are performed, nine tests have to be carried out, which is too many. Therefore only
experiment 1B, with the load applied at the shear centre, will be carried out. This is the purest
instability problem and therefore most valuable.
The experimental value on the lateral torsional buckling moment will be compared to the value
obtained from formula 12.2-3 of NEN 6771 [2].
The concentrated loads F1 and F2 should be equal. During loading some differences may occur due to
imbalance of the load. To be able to make a distinction between the two, different notations are used,
see figure 19. In order to apply the load in the shear centre, plates are welded perpendicular onto the
web.
F1 1B
F2
S
y
y-axis
L/4 S C
C
L/2 x
z plate
L/4
weld z-axis
(a) (b)
figure 19: loading scheme of experiment 1
13
1 1 1
= + Eq. 1
Fc Fpl Fke
3.1.3 Results
The results of calculations of Appendix C and Appendix D, § Bending are given in table 3 for spans
from 2.8 up to 4.8 meters.
The elastic-plastic lateral torsional buckling load, Fmax;s;d from NEN 6771, is considered to be the
most accurate prediction for the failure load. The critical load Fc, as has been determined by using the
Merchant – Rankine method, is lower and is assumed to be less accurate as prediction of the failure
load for this load case.
At an early stage the load was restricted to 25 kN and thus the span had to be 4 meters or over.
However the idea of using dead weight was abandoned and any span could be used.
Now the choice was to select a span in the range where the relative slenderness λrel was smaller or
equal to one because the failure load is furthest removed from the theoretical failure load, see
figure 20. Thus the selection was narrowed down to a span of 3.2 meters or shorter. Since the three
available sections were 12 meters each, they could be cut in four parts of three meters. Taking into
account some length for the supports the actual span came down to 2.8 meters. The predicted failure
load at that span is 41.7 kN.
Experiment 1B will be performed four times to determine the mean and standard variation of the
actual failure load.
table 3: results of calculations from Appendix D, § Bending
17.5 · h 20 · h 22.5 · h 25 · h 27.5 · h 30 · h
Span 2.8 [m] 3.2 3.6 4.0 4.4 4.8
Fke 69.2 [kN] 51.3 39.6 31.5 25.7 21.4
F el 50.6 44.34 39.46 35.54 32.33 29.65
F pl 57.3 50.6 45.3 40.9 37.3 34.3
λrel 0.91 [−] 0.993 1.069 1.14 1.205 1.267
ωkip 0.727 [−] 0.67 0.617 0.569 0.527 0.489
Fmax;s;d 41.7 33.9 27.9 23.3 19.7 16.8
Fc 31.3 25.5 21.1 17.8 15.2 13.2
Experiment 1B
70 F-ke
F-el
60
F-pl
50 F-max;s;d - [2]
F-c - M.R.
Force [kN]
40
30
20
10
0
2.8 3.2 3.6 4 4.4 4.8
Span L [m ]
14
3.2 Experiment 2
At this experiment bifurcational buckling does not occur due to torsion in the member. Therefore the
first yield stress is taken as an indication for failure.
The objective of experiments 2A, 2B and 2C is to determine the structural behaviour and failure load
in comparison to experiment 1B.
2A
F1 2B
s.c. F2 y-axis
e
y e
+φ F1 e
F2 c.g.
z
x 2C
experiment 2
(a) z-axiz (b)
figure 22: scheme of experiment 2
To analyse this load case bending and torsion are separated. To analyse the bending effects the load is
shifted to the shear centre and the same analysis as for experiment 1 is used. However, due to the
eccentricity of the load a torsional moment is present in the member. The effects of this torsional
moment is analysed in Appendix D, § Torsion due to shear force.
The elastic section capacity is determined using the Von Mises criterion. However, determining the
plastic section capacity for this load case is much more difficult than for experiment 1. To
approximate the plastic section capacity the elastic section capacity is multiplied by the section shape
factor.
3.2.2 Results
For these experiments the same span is chosen as for experiment 1 to be able to compare the failure
loads to experiment 1. In experiment 1 the elastic-plastic lateral torsional buckling load Fmax;s;d was
considered to be the failure load. For the eccentrically loaded experiments such a failure load can not
be obtained and thus the critical load obtained form the Merchant – Rankine postulate is considered
to be the failure load. table 4 through table 6 give the results of the analysis of Appendix D, § Torsion
due to shear force. These results are shown graphically in figure 24 through figure 26.
15
table 4: experiment 2A
Span 2.8 [m] 3.2 3.6 4.0 4.4 4.8
Fke;top 55.8 [kN] 42.2 33.1 26.8 22.1 18.6
F el;torsion 30.1 27.9 25.9 24.3 22.8 21.6
Fpl;torsion 35.3 32.6 30.4 28.4 26.7 25.2
Fc;top 21.6 18.4 15.8 13.8 12.1 10.7
Experiment 2A
60
F-ke
50 F-el
F-pl
Force [kN
40
F-c - M.R.
30
20
10
0
2.8 3.2 3.6 4 4.4 4.8
Span L [m]
table 5: experiment 2B
Span 2.8 [m] 3.2 3.6 4.0 4.4 4.8
Fke;S 69.2 [kN] 51.3 39.6 31.5 25.7 21.4
Fc;S 23.4 19.9 17.2 14.9 13.1 11.6
Experiment 2B
70
F-ke
60 F-el
50 F-pl
Force [kN]
40 F-c - M.R.
30
20
10
0
2.8 3.2 3.6 4 4.4 4.8
Span L [m ]
table 6: experiment 2C
Span 2.8 [m] 3.2 3.6 4.0 4.4 4.8
Fke;bottom 85.8 [kN] 62.3 47.3 37.1 29.9 24.6
Fc;bottom 25 21.4 18.5 16.1 14.1 12.4
Experiment 2C
90
F-ke
80
F-el
70
F-pl
Force [kN]
60
F-c - M.R.
50
40
30
20
10
0
2.8 3.2 3.6 4 4.4 4.8
Span L [m]
16
4 Test rig
A decision was made to use jacks to apply the load onto the sections. To minimise the effect of the
pinned end of the jacks, the jacks had to be placed as high above the section as possible. The
available height to build a test rig was restricted to 5.1 meters.
For these experiments the available test rig was originally build to test heavy concrete members. This
test rig was composed of a lot more beams than strictly would be needed for these experiments,
resulting in a tight space for manoeuvring the sections in and out of the rig. However, with a few
adjustments this test rig could be used for the experiments. In figure 27 two sections of the test rig are
depicted, showing how the load is applied and how the jacks are connected to the load bearing yokes.
In Appendix F more plans as well as photographs of the test rig are given.
In figure 28 the mechanical working of the test rig is shown.
3a Section A-A' 2a Section C-C'
scale
0 300 600 900
hinge
jack jack
hinge
anker
prestressing
cable
anker
hinge
load cell z - axis
z - axis
load bearing yoke
x - axis
support section
y - axis
Fhorz
17
4.1 Influence of pinned-end jack
In § 1.3 Type of loads the advantages and disadvantages have been discussed. In this paragraph the
influence of the pinned end of the jacks is investigated.
The influence of a point fixed in location depends on the length of the connecting rod and the lateral
displacement. The effect of these two parameters on the horizontal load in relation to the absolute load
is given by equation 2. Fhorz has been plotted for 2 < h < 4 metre and v < 30 millimetre in figure 29.
v
Fhorz (v, h ) = sin arctan ⋅ F Eq. 2
h
F1 F2
v,u = 0
F
h = 4260 mm tf = 10
connecting element
load acts at corner of flange and web
v,w = 0
tw = 6.5
h0 = 150
y,v v=0
u,v,w = 0
z,w
x,u
(a) (b) b0 = 66.75
v=0
18
The simulation has been carried out for both a pressure and a tension load and the effect on the load
and displacement at failure are compared. The results are given in figure 31, the dots resemble the
steps that are taken in the calculation. The load deflection curves are almost identical. A closer look
at the numbers shows that the failure load for tension is 1.2 % larger than for pressure whereas the
lateral displacement at failure is 0.8 % smaller.
20000
10000
0
0 10 20 30
Lateral displacem ent [mm ]
figure 31: finite element simulation of the difference between tension- and pressure jacks
Since the difference in the effect of using tension or pressure is very small it has been decided to use
tension jacks. They make the experiments easier, i.e. all points at which the connection rod can hinge,
see figure 27, will align when using tension.
h=
< h /10
h /2 =
1B 2A 2B
2C
figure 32: solutions for applying load at the different points in the cross section
One of the demands for the design of the fittings was that the section must be able to rotate freely
about a longitudinal axis so that twisting is not restrained. Because the section bends as well, the
fittings must also allow a small angle of rotation about the y-axis of the section (perpendicular to the
longitudinal axis). In Appendix F plans and photographs of the specially designed fittings are given.
19
1B: In the fitting of experiment 1B a bolt is used to carry the load across from the fitting to the
weld-on plate of the section. This bolt fits exact into the fitting, but to allow rotation about
the y-axis it has been notched at where it sticks through the plate.
2A: For experiment 2A a ball and socket joint is used, ensuring free rotation about all axes.
2B: In experiment 2B a solid pin is used to apply the load onto the web at the line of symmetry. In
this pin a straight groove has been cut to allow the section to rotate about the longitudinal
axis, see figure F11. Because the groove is straight it does not allow rotation about the y-axis,
but it is assumed that the section will come to rest on one side of it and thus rotation is not
prevented.
2C: For experiment 2C a shackle is used to apply the load to the section. On the section a plate is
welded in which the shackle is hooked. In this plate a rounded notch, see figure F18, makes
sure it stays in position, while it can rotate freely about the longitudinal- and y-axis.
Ffriction = Mfriction / α L
α L = 700 [mm]
If the section deflects laterally it will have to overcome the friction in the joint; the friction moment is
not an active force acting on the section. To get an idea of the amount of support it is assumed that a
horizontal load, Ffriction, works on the section at the position where the vertical load is applied, see
figure 33.
• In experiment 1B no lateral deflection will occur until buckling occurs. This means that the
friction might support the section in an unstable path of equilibrium, see figure 14, and buckling
could occur at a higher level of loading then would be the case if the joint was without friction.
20
• In case of experiments 2A, B and C the section deflects from commencement of loading on, so
friction must be overcome during the whole process. Since pinned-end jacks are used a horizontal
component of the applied load will arise, as has been discussed in § 4.1. The effect of this
component will be countered by the friction in the joint. In figure 34 (a) the maximum and
minimum force due to friction (equations 3 and 4) are plotted as well as the horizontal component
of the vertical load. To plot the horizontal component as a function of the applied load and lateral
displacement, equation 2 has been rewritten, in which the relation between ‘v’ and Fvert from
figure 34 (b) is used.
v v
Fhorz (v, F ) = sin arctan ⋅ F = sin arctan ⋅ F Eq. 5
h 4000
0
0 12.5 25 37.5 50
v [mm]
(a) applied load [kN] (b) 30
figure 34: influence of supports and jacks
The horizontal component of the applied load, Fhorz , is largely within the upper and lower bounds of
the friction load, see the shaded area of figure 34 (a). Therefore it can be concluded that for
experiments 2A, B and C much of the effect of friction is cancelled out by the fact that pinned-end
jacks are used.
For experiment 1B the effect can not be specified in the same way as for experiment 2 and thus
uncertainty remains.
21
z-axis
socket
ball
1 dw (x = 0 )
u = 56 ⋅ sin (θ ) = 56 ⋅ sin −
2 2
( )
= 56 ⋅ sin − F ⋅ βL ⋅ L − (βL ) − F ⋅ αL ⋅ L − (αL )
2 2
( )
EI y dx 6 ⋅ EI y L 6 ⋅ EI y L
u≈
56 ⋅ F
6 ⋅ EI y L
( ( ) (
⋅ − βL ⋅ L2 − (βL ) − αL ⋅ L2 − (αL )
2 2
)) Eq. 6
A displacement ‘u’ of the neutral axis must be due to a normal force in the member. The normal force
is the horizontal reaction force from the joint and acts at the top flange, thus adding a bending
moment to the section as well, see figure 36.
z-axis
Mreact = Freact·h/2 u
Freact
Mreact
h/2
56
= Freact
+θ
x-axis
To determine the reaction force the displacement at the centre point of rotation of the joint must be
equal to the displacement from equation 6. An expression is derived for this displacement due to the
reaction force:
Freact ⋅ L F h ⋅L
u= + 56 ⋅ sin react Eq. 7
2 ⋅ EA 4 ⋅ EI y
Equation 7 cannot be solved for Freact. Since the angle of rotation at the support is very small
(about 1º) the argument is nearly equal to the sine of the argument and thus equation 7 can be
approximated by just the argument. The reaction force Freact can now be factored from it:
u
Freact = Eq. 8
L 56 ⋅ h ⋅ L
+
2 ⋅ EA 4 ⋅ EI y
22
Entering equation 6 for ‘u’ in equation 8 yields an expression for the reaction force. The reaction
force can be written as function of the applied load, see figure 39.
56 ⋅ F
6 ⋅ EI y L
[ ( ) (
⋅ βL ⋅ L2 − (βL ) − αL ⋅ L2 − (αL )
2 2
)]
Freact (F ) = Eq. 9
L 56 ⋅ h ⋅ L
+
2 ⋅ EA 4 ⋅ EI y
deflection w(x)
x=L
x=0
x-axis Freact;2nd Freact;2nd
u u u u
The displacement of both supports can be found by solving equation 10 for the value of ‘t’ and then
subtracting it from L. The integral in equation 10 is the definition of the length of a function. The
function is the bending displacement of the member, which is given by equations D3 through D5 in
Appendix D..
t 2
dw ( x )
∫
0
1+
dx
dx = L Eq. 10
Since the deflection in the experiments is small compared to the overall length of the member, the
displacement of the supports can be approximated by equation 11.
L 2
dw ( x )
2⋅u = ∫
0
1+
dx
dx − L Eq. 11
The supports do not allow the horizontal displacement ‘u’ and so the second order reaction force is
induced. This force can only be introduced at the top flange where the joint is resting on the section.
Thus not only a normal force is induced but a bending moment as well, see figure 38.
z-axis
Mreact;2nd = Freact;2nd·h/2 u
Freact;2nd
h/2 Freact;2nd
= −θ
x-axis
Mreact;2nd
23
The reaction force due to second order effects is equal to the force that is necessary to resist the
displacement ‘u’ from figure 38. An expression for this load has been derived in equation 12, in
which the first part of the denominator describes the effect of the supports moving closer and the
second part the effect of the rotation θ of the section.
u (F)
Freact;2 nd (F) = Eq. 12
L h 2L
+
2 ⋅ EA 8 ⋅ EI y
This reaction force works in opposite direction of the first order reaction force (equation 9) and can
therefore be subtracted from it. Plotting equation 9 gives the graph of figure 39.
0
4 4 4 4
0 1.25 10 2.5 10 3.75 10 5 10
The reaction force, as calculated for supports solidly fixed in place, is nearly four times as great as
the applied load, as is illustrated by figure 39. The second order effects lowers the reaction force only
slightly and will therefore be disregarded.
It is obvious that such a force would causes deformations is the supports. If the supports deform the
reaction force declines. Therefore the stiffness of the support-member will be investigated.
24
4.4.2.2 No slip at support
The ball and socket joint is placed between the support-member and the section. The joint can slide
over both surfaces, but will be held in place by the normal force acting at the support.
support-member
(a) cross section at support (b) conveyance of reaction force (c) cross section at test rig
If it is assumed that the joint can not slip, then the support-member carries the full reaction force. The
bottom flange of the support member is clamped onto the test rig, see figure 41 (c). This connection is
modelled as a hinge, see figure 42 (b), but probably can slide as well due to the gap that emerges
during loading, see figure 41 (c).
Further more it will be assumed that only the bottom flange will carry the horizontal reaction force.
25
150
The stiffness of the support-member is determined in figure 42 (b) and expressed as coefficient ‘k’. In
§ 4.4.2.1 the displacement ‘u’ of the support has been determined. If the sections is loaded, the
support-member will be pressed sideways, resulting in a reaction force. This force is equal to the
imposed displacement divided by the stiffness of the support-member, or Freact = u / k.
Horizontal displacement
Horizontal reaction force 1
4000
0.75
displacement u [mm]
3000
Reaction force [N]
2000
0.5
1000 0.25
0 0
4 4 4 4 4 4 4 4 4 4
0 1 10 2 10 3 10 4 10 5 10 0 1 10 2 10 3 10 4 10 5 10
applied load [N] applied load [N]
figure 44: reaction force from support-member figure 43: calculated displacement at support (Eq. 6)
In figure 44 it is shown that the reaction force that the support-member is able to give is much lower
than the reaction force determined in case of a solidly fixed joint, see figure 39.
4.4.2.3 Measurements
In the previous paragraphs it was assumed that the joint could not slide over both the section and the
support member. In figure 41 (b) it is pointed out that the joint has smooth steel surfaces resting on
other steel surfaces. In figure 49 (a) it is shown that to obtain such a surface on the section the rolling
skin has been removed.
The smooth surface ensures low friction so that the displacement ‘u’ can also be facilitated by sliding
of the joint, besides bending of the support-member. It is difficult to specify the amount of friction
and therefore difficult to determine how much of the displacement happens due to slip. To learn what
actually happens at the joint, both the sliding displacement of the ball in relation to the channel
section and the bending displacement of the support-member, see figure 45, have been measured in
four tests.
26
z-axis fixed to floor
7
socket channel 6
8 fixed to section
9 6
x-axis u
L ball section
figure 45: measurements of displacements of supports
50 50
40 40
load F [kN]
load F [kN]
30 30
channel 6
channel 8
20 20 channel 9
channel 7
10 10
0 0
-0.75 -0.5 -0.25 0 0.25 0.5 0.75 -0.15 -0.1 -0.05 0 0.05 0.1 0.15
displacem ent of joint [m m ] slip of joint over section [m m ]
(a) (b)
figure 46: measured displacement of joint
In figure 46 the measured displacements are given; the sign of the displacements corresponds with the
sign of the axes, as they are given in figure 45. The total displacement at the supports is the sum of
the absolute values of the displacements from figure 46 (a) and (b). For example: at F = 40 kN the
measured displacement is 0.5 millimetre (in figure 46 a) plus 0.075(in figure 46 b) is together
approximately 0.6 millimetre, whereas the calculated displacement is about 0.8 millimetre, see
figure 43. The discrepancy between these two values may be due to erroneous assumptions in the
above presented analysis.
4.4.2.4 conclusion
If the calculated reaction force is taken from figure 44 one can see that at a load of 50 kN the reaction
force is 4 kN. In figure 47 the effect of the reaction force on the bending moment distribution is
analysed, causing a reduction of the bending moment of 0.32/35 · 100 = 0.64%. From this it can be
concluded that the shift of the point of rotation at the ball and socket joint has an effect on the overall
experiment that can be neglected.
L/α
Msupport = Freact·h/2 = 0.32 [kNm]
F F = 50 [kN]
This conclusion is further supported by experiment 1B1. In figure 48 the load displacement graph as
it has been measured is displayed. It is clearly visible that at about 50 kN the trend of the graph
changes, which indicates yielding of the section. In table 3 the load at which first yield occurs, Fel, is
presented to be 50.6 [kN]. This is in good agreement with the measured value.
If the supports would have caused a change in the moment distribution the measured point of yielding
would be different from the calculated value.
27
Experim ent 1B1
60
50
40
average load [kN]
30
20
10
0
0 10 20 30
contact with
supportive rod
area stripped
of rolling skin
(a) (b)
figure 49: end of section with imprint
This notch might have caused the section to hook on to the support and thus impede free rotation
about y- and z-axes as well as warping. This effect is difficult to quantify thus we stick to simple
reasoning: if the section was to hook on to the support a reaction force in x-direction will occur.
Since the support-member is very slack in bending about z, which is discussed extensively in §
4.4.2.2, this reaction force will be small and thus the influence on the overall experiment, similar to
the cases discussed in §4.4.2.4, will be small.
28
5 Measurements
In construction most loads are gravity induced and cause downwards deflections. The co-ordinate
axes are chosen such that these deflections are considered to be positive, see figure 19 and figure 22.
Due to practical reasons the sections are loaded upwards. By choosing the z-axis upwards the
deflections are positive. In figure 27 it is shown how the co-ordinate system is placed.
The measurements consists of displacements in y- and z-directions, rotation about the x-axis, the
applied load and in experiments 1B4, 2A2, 2B3 and 2C2 of strain. In Appendix H all used equipment
is listed and the calibration of the equipment is reproduced.
As for the load cells: they registered different loads whereas they should have been identical. This
could not have been a mechanical phenomena since the two jacks were interconnected and thus the
load on both jacks must be equal. After all experiments had been carried out the load cell were
calibrated once more and it turned out that the calibration factor of load cell 91 had changed
considerably. After long deliberation it has been decided to use the latest calibration factors as well
as the average of both registered loads in the load – displacement curves.
Most measured values could not be used right away because they must first be corrected, for
example, for rigid body displacements.
To make a difference between the actual measured quantity and the corrected quantity the following
convention is used: wm;2 = deflection measured by channel 2 in z-direction
w2 = corrected displacement at the location of channel 2
5.1 Offset
At commencement of loading the section rests on blocks and the load bearing yokes rest on the
section. The registration of data is started and the first number acquired is taken as zero load. Once
the jacks start to go up the load bearing yoke will start to carry the section and the load cells register
an increases in load. Before contact with the supports the section moves up, but no load increase is
registered. This is visible from the horizontal part of the graph in figure 50 (b). Once the section is
pressed up against the supports the load will start to increase again. At that point the curve has a
clearly visible change in direction. The value at that point is set back to zero and is subtracted from
all registered values thereafter.
2
40 value set to zero 40
1.5
Load [kN]
Load [kN]
Load [kN]
0 0 0
0 10 20 30 40 0 1 2 3 4 5 0 10 20 30 40
vertical displacement [mm] vertical displacement [mm] vertical displacement [mm]
(a) (b) (c)
figure 50: removing offset
In figure 50 the deflection near one load point, wm;2, and at midspan, wm;1, are plotted against the two
load cells, lc90 and lc91. The two load cells clearly give different values in figure 50 (b) while they
should be equal. This might be due to the fact that the cells are not completely free of any load at
commencement.
The load bearing yoke, the fitting, one half of the section and one support together weigh 134 kg
which is equal to a load of 1.3 kN. If we look at the graph we see that the horizontal line is at about
1.3 kN. Thus it is quite certain the right offset is subtracted from the graph of figure 50 (a), resulting
in graph (c).
29
5.2 Deflection in z
The displacements near the points of load application and at mid span are measured at the shear
centre so no corrections need to be made for rotation of the section. Figure F10 in Appendix F shows
that the wires for measuring displacements (both horizontal and vertical) are hooked onto a bolt that
sticks out from the web to the location of the shear centre. In the same figure it can also be seen that it
is not physically possible to measure right at the point of load application. Therefore the
displacements are measured 150 millimetres removed from it, towards midspan at x = 850 and
x = 1950, see figure 51.
w
mid span, x = 850
point of load application
support x = 1950 x = 1400 z-axis
lc90 lc91
y-axis shear centre
x-axis
wm;71 wm;72
At the span displacements are measured from a fixed point so no distinction can be made between the
rigid body displacement of the section and the deflection of the section, see figure 52. To obtain the
sole bending displacement the rigid body displacements at the supports must be subtracted.
Experiment 2B1
Experiment 2B1
45
45
40
40
35
35
30
Load F [kN]
Load F [kN]
30
25 Wm;0 25
20 Wm;1 20 Wm; 71 vs lc90
15 Wm;2 15 Wm;72 vs lc91
10 10
5 5
0 0
0 10 20 30 40 0 1 2 3 4 5
vertical displacem ent [mm] vertical displacement at supports [mm]
figure 54: measured displacements at span figure 53: vertical displacements at supports
At the supports displacements up to about 3 to 5 millimetres are measured. The difference between
the displacements at both supports is at most one millimetre, so the section will rotate slightly.
30
Therefore the displacements measured at the load points could be corrected as is shown in equations
13 and 14.
850
w 0 = w m;0 − w m;71 +
2800
(w m;72 − w m;71 ) Eq. 13
1950
w 2 = w m; 2 − w m;71 + (w m;72 − w m;71 ) Eq. 14
2800
At mid span the measured displacement is corrected by the average displacements of the supports, see
equation 15
w m;71 + w m;72
w 1 = w m;1 −
Eq. 15
2
Now, if we look at the numbers of figure 53 we see that if the applied load is 40 [kN] the
displacements wm;71 ≈ 2.5 and wm;72 ≈ 3.5 millimetres. By entering these values in equations 13 and 14
we find:
850 1950
w 0 = w m;0 − w m;71 + = w m ; 0 − 2 .8 and w 2 = w m; 2 − w m;71 + = w m;2 − 3.2
2800 2800
The average displacement of the supports is 3 millimetres. If we look at the measured displacement of
channel 0 and 2 (these are the displacements that have been measured near the points of load
application) in figure 54 we see that at a load of 40 kN these are approximately 16 millimetres. Now
to compare the method of correcting the measured displacements of equations 13 and 14 with
correcting the displacements with just the average displacements of the supports we find the
2.8 − 3 3 .2 − 3
following error at channel 0: ⋅ 100 = −1.25 % and at channel 2: ⋅ 100 = 1.25 % . These
16 16
errors are quite small and it has been decided to correct the vertical displacements with just the
average displacements of the supports.
31
5.3 Deflection in y
The lateral deflection at the span is measured at the same points as the vertical deflection is
measured. At the supports there is no room to measure so the readings of the displacements are taken
right besides the supports, see figure 55 and figure 58.
x-axis z-axis
At the top flange the measured displacement is less than at the Experiment 2B1
bottom flange, due to the difference in stiffness of the support,
45
see figure 56. The lateral displacement of the shear centre at 40
the supports is considered to be the average of the two 35
30
Load F [kN]
measured displacements. At both supports the displacements 25
Vm;73
are about the same size therefore the average displacement of Vm;70
20
15
the two supports, see equation 16, is subtracted from the 10
deflection measured at the span (vm;60, vm;61 and vm;62). 5
0
-6 -4 -2 0 2
Lateral displacem ent at support [m m]
vm;70 + vm;73 vm;63 + vm;64
+
vaverage= 2 2 Eq. 16 figure 56: measured displacements at support
2
vm;73
reading of
horizontal d
d
displacements vm;70
reading of
vertical number of
displacement channel
figure 58: measuring of displacements at support figure 57: determining average lateral displacement at support
32
5.4 Rotation about x
The rotation is measured near the points of load application and at midspan by special rotation
recorders. In figure 61 a recorder is depicted. A metal bar is suspended from a turning point, thus
remaining upright. The angle that the bar makes with the recorder housing is registered. The housing
has been clamped to the top flange but is able to hinge about the y-axis.
Due to the stiffness of the supports small rotations can occur and thus the rotations at the span need to
be corrected. Since the rotation at both supports is about the same size the average rotation of both
supports is subtracted from the rotation measured at the span.
x-axis z-axis
top view
In figure 60 and in equation 17 it is shown how the rotation at the supports can be determined from
the lateral displacement.
vm;70 − vm;73
φsupport = [rad] Eq. 17
d
hinge
turning point vm;73 vm;73
vm;70 −φ
vm;73
vm;70 - vm;73
weight
figure 60: determining rotation at supports
figure 61: measuring of rotation
33
5.5 Strain measurements
Only in one test of the different series of experiments (1B, 2A, 2B and 2C) strain measurements have
been performed. In experiment 1B the sections bends only until lateral torsional buckling occurs. In
the other experiments the sections are subjected to torsion as well and the interpretation of the results
is slightly different.
5.5.1 Experiment 1
In this experiment strain has been measured at two
positions along the member, at midspan and at
x = 850 millimetres, see figure 63. At these points
displacements are measured as well. strain gauge
At these positions two strain gauges were placed on
the top flange, see figure 62, and two on the bottom
flange, symmetrically to the ones on top. strain distribution
My EI y
σMy = E ⋅ ε = z ⇒ My = ε Eq. 18
Iy z figure 62: position of strain gauges on section
24
25
26
27
20
21
midspan
22
23 x = 850
(a) (b)
figure 63: numbering of stain gauges and position along the member
34
5.5.2 Experiments 2
In these experiments the strain is measured rosette
identically to experiment 1B with the exception
strain gauge
that at x = 850 [mm] a rosette is added between ω = -1622
the two strain gauges on the top flange.
ω = 1900
Axial strain: ω = -213
This rosette is placed such that the axial strain is
measured at the point where the warping
function is equal to zero on the top surface. Due
to a miscalculation they are all placed just next
to that point (30 millimetres from the web
instead of 33), see figure 64. By measuring the
strain at this point a distinction can be made
between strain due to bending and strain due to
the bi-moment. If the strain measured by the
axial component of the rosette is entered in
equation 18 the contribution of bending is
found.
By subtracting the strain measured in the rosette
from the strain measured in the two strain
gauges and then using equation 19 (derived
warping function equal to
from equation D23) the bi-moment can be zero on top surface
determined.
Shear strain:
Between the points of load application no shear stress is present due to bending. However, due to
Saint-Venant torsion and warping restraint torsion there are shear stresses, but these are opposite in
sign and should cancel each other out, which is shown in figure 10. By measuring the shear strain at
x = 850 this can be checked.
26
43
45 44
24
27 25
top view
26
27
20
21
midspan
22
(a) 23 x = 850
(b)
figure 65: numbering of strain gauges and rosette, and position along the member
35
6 Results
The corrections on the measured quantities are implemented on the recorded data. This has resulted
in a veriety of load-displacement curves, which are reproduced hereafter.
6.1 Experiment 1
In figure 66 the corrected displacements and rotation of the section are given at three points along the
longitudinal axis: x = 850, midspan and x = 1950. In table 7 the failure loads are given.
table 7: failure loads
60 60 60
50 50 50
avera ge loa d F [kN]
30 30 30
0 0 0
0 10 20 30 40 -15 -10 -5 0 5 10 15 -10 -5 0 5 10
w 2 [m m ] v 6 2 [m m ] fi 5 [ º ]
V e rtic a l de fle c tion a t x = 8 5 0 la te r a l de fle c tion a t x = 8 5 0 r ota tion a t x = 8 5 0
50 50 50
avera ge loa d F [kN]
40 40 40
30 30 30
50 50 50
avera ge loa d F [kN]
40 40 40
30 30 30
w 0 [m m ] v 6 0 [m m ] fi 3 [ º ]
V e rtic a l de fle c tion a t x = 1 9 5 0 la te r a l de fle c tion a t x = 1 9 5 0 r ota tion a t x = 1 9 5 0
36
Strain measurements
Ex p e rim e n t 1B 4 Ex p e rim e n t 1B 4
stra in a t m idsp a n stra in a t x = 850
50 50
40 40
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
30 30
s train gauge 20
20 20 s train gauge 22
s train gauge 21
s train gauge 23
s train gauge 24
s train gauge 26
10 s train gauge 25 10 s train gauge 27
0
0
-2 -1 0 1 2 -2 -1 0 1 2
stra in [‰] stra in [‰]
The applied bending moment at failure is: My = Fu;1B4 · L/4 = 48.37 · 0.7 = 33.86 [kNm]
37
6.2 Experiment 2
Experiment Failure load [kN] Experiment Failure load [kN] Experiment Failure load [kN]
2A1 37.68 2B1 43.13 2C1 48.01
2A2 37.22 2B2 43.84 2C2 46.67
2B3 44.17
average 37.45 average 43.71 average 47.34
standard deviation 0.323 standard deviation 0.529 standard deviation 0.946
50 50 50
40 40 40
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
avera ge load F [kN]
2A 1 2A 1
2A 1
30 2A 2 2A 2 30 30
2A 2
2B1 2B1
2B1
2B2 2B2
20 20
2B2 20
2B3 2B3
2B3
2C1 2C1
10 2C2
10 2C1 10
2C2
2C2
0 0 0
0 10 20 30 -3 0 -2 0 -1 0 0 -15 -10 -5 0
w 2 [m m ] v 6 2 [m m ] fi 5 [ º ]
V e rtica l de fle ction a t x = 850 la te r a l de fle c tio n a t x = 8 5 0 rota tion a t x = 850
40 40 40
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
30 2A 1 30 2A 1 30
2A 1
2A 2 2A 2 2A 2
40 40 40
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
a ve ra ge loa d F [kN]
30 2A 1 2A 1 30 2A 1 30
2A 2 2A 2 2A 2
2B1 2B1 2B1
20 20 20
2B2 2B2 2B2
2B3 2B3 2B3
10 2C1 2C1
10 10
2C1
2C2 2C2 2C2
0 0 0
0 10 20 30 -30 -20 -10 0 -15 -10 -5 0
w 0 [m m ] v 60 [m m ] fi 3 [ º ]
V e rtica l de fle ction a t x = 1950 la te ra l de fle ction a t x = 1950 rota tion a t x = 1950
38
Strain measurements:
Ex p e rim e n t 2A 2 Ex p e rim e n t 2A 2
stra in a t m idsp a n stra in a t x = 850
40 40
35 35
30
25 25
20 20
15 15
s train gauge 22
s train gauge 21
s train gauge 23 10
s train gauge 20 10
s train gauge 26
s train gauge 24
5 s train gauge 27 5
s train gauge 25
0 0
-3 -2 -1 0 1 2 -3 -2 -1 0 1 2
S train [‰ ] S train [‰ ]
Ex p e rim e n t 2A 2 Ex p e rim e n t 2A 2
stra in in ro se tte stra in a t top fla ng e a t x = 850
40 40
35 35
30 30
av e rage load F [kN]
av e rage load F [kN]
25 25
20 20
s train gauge 43 15
15 s train gauge 26
s train gauge 44 s train gauge 27
10 s train gauge 45 10 s train gauge 43
5 5
0 0
-0.5 0 0.5 1 1.5 2 0 0.5 1 1.5 2
S train [‰ ] S train [‰ ]
Bi-moment at failure:
- determined from strain gauges 26: Bx = -0.17 [kNm2]
- determined from strain gauges 27: Bx = -0.07 [kNm2]
- determined from failure load: Bx = -0.145 [kNm2]
39
Ex p e rim e n t 2B 3 Ex p e rim e n t 2B 3
stra in a t m idsp a n stra in a t x = 850
45 45
40 40
35 35
average load F [kN ]
25 25
20 20
s train gauge 22
s train gauge 21
15 s train gauge 23 15
s train gauge 20
s train gauge 26
s train gauge 24 10 10
s train gauge 27
s train gauge 25
5 5
0 0
-3 -2 -1 0 1 2 -3 -2 -1 0 1 2
S train [‰ ] S train [‰ ]
Ex p e rim e n t 2B 3 Ex p e rim e n t 2B 3
stra in in ro se tte stra in a t top fla ng e a t x = 850
45
45
40 40
Force (channe l 90) [kN]
35 35
30 30
ave r ag e lo ad F [k N]
25 25
20 20
15 15
s train gauge 44
s train gauge 26
10 s train gauge 43 10
s train gauge 27
5 s train gauge 45
s train gauge 43
5
0
0
-0.5 0 0.5 1 1.5 2 0 0.5 1 1.5 2
S train [‰ ] Str a in [‰ ]
Bi-moment at failure:
- determined from strain gauges 26: Bx = -0.07 [kNm2]
- determined from strain gauges 27: Bx = -0.08 [kNm2]
- determined from failure load: Bx = -0.172 [kNm2]
40
Ex p e rim e n t 2C2 Ex p e rim e n t 2C2
stra in a t m id sp a n stra in a t x = 850
50
50
45
45
40 40
average load F [kN ]
35 35
ave r ag e lo ad F [k N]
30 30
25 25
20 20
s train gauge 20
15 s train gauge 22
15
s train gauge 21
s train gauge 24 10 s train gauge 23 10
s train gauge 26
s train gauge 25 5 5
s train gauge 27
0 0
-3 -2 -1 0 1 2
-3 -2 -1 0 1 2
S train [‰ ] Str a in [ ‰ ]
45 45
40 40
av e rage load F [kN]
35 35
ave r ag e lo ad F [k N]
30 30
25 25
20 20
s train gauge 26
15 15
s train gauge 43 s train gauge 27
10 s train gauge 44 10
s train gauge 43 (2C1)
s train gauge 45
5 5
0 0
0 0.5 1 1.5 2
-0.5 0 0.5 1 1.5
S train [‰ ] Str a in [ ‰ ]
figure 71: strain measured in experiment 2C2 (measurements with rosette are from experiment 2C1)
41
7 Discussion
In general it can be said that the objectives of this study have been achieved and that the experiments
have been performed according to the restrictions. A few points are articulated:
• The wanted experimental data on strength, stiffness and stability has been acquired
• This data can be compared to analytical models and numerical calculations, as will be hereafter.
• By repetitive testing mean and variation on failure loads have been determined. The variation
turned out to be very low, indicating conscientious execution of the experiments.
• The Merchant – Rankine postulate has been checked against the experimental data. The results
are not satisfying yet but leave room for improvement.
• According to the restrictions the sections must be loaded such that no support is given or failure
is precipitated. In chapter 4 a long investigation has been carried out to quantify the effects of the
test rig on the course of the experiments. It can be concluded that most effects are minimal. Only
the effect on the lateral torsional buckling load was not possible to quantify.
7.1 Experiment 1
7.1.1 Failure
The objective of experiment 1 is to determine the lateral torsional buckling load. In figure 14
theoretical paths of displacements and rotation are shown. The results, as presented in figure 66,
show that the experiments follow the theoretical path closely, i.e. the section hardly rotates nor
deflects laterally until failure. Thus the highest registered value of the applied load is taken as the
lateral torsional buckling load.
The fact that the sections did not rotate until failure is remarkable since there is a small eccentricity in
applying load, due to the welding on of the perpendicular plates. This eccentricity ranges form 2 to 4
millimetres, see Appendix I.
Experiment 1B1 has a failure load that is about 10 % higher than that of the other experiments. The
section also buckled in the other direction. At this experiment no teflon was used at the supports.
Whether or not this has contributed to the higher failure load is unclear. Nevertheless, this experiment
will be disregarded in determining the mean and standard deviation of the lateral torsional buckling
load.
In table 7 the final results of the failure loads are shown. The average failure load of the experiments
Fu = 49.4 is quite a bit larger than the predicted failure load of Fmax;s;d = 41.7 [kN] (see table 3), i.e.
Fmax;s;d = 84 % Fu. This might be a reason to reconsider using the approach of code NEN 6771 [2],
using the reduction curves, to predict lateral torsional buckling failure of channel sections.
Perhaps a special reduction curve can be obtained for channel sections. In figure 72 it is shown how
the theoretical lateral torsional buckling moment Mke is reduced with the a-curve of NEN 6771 [2]. At
λrel = 0.91 (slenderness of experiment 1B, see table 3) Mke is reduced with ωkip = 0.727. If we look at
the actual buckling moment Mu the reduction would be Mu / Mke = 34.6 / 48.4 = 0.7. By performing
experiments on sections with other spans a new reduction curve can be constructed.
It might be possible to construct a reduction curve based on finite element analyses as well. A
simulation of this experiment using the finite element method will show whether or not the results are
in good agreement with the experimental data.
42
M
[kNm]
Mke = 48.4
Mpl = 40 actual buckling at Mu = 34.6 ⇒ reduction ω = 0.86
0 0.91 1 M pl
λ rel =
M ke
figure 72: predicted load for lateral torsional buckling and actual load
The predicted failure load Fmax;s;d is plotted in figure 73 to compare with the actual failure loads of
experiments 1B2, 1B3 and 1B4.
7.1.2 Stiffness
In Appendix D it is shown how bending and torsion can be analysed. In this experiment only bending
occurs and thus the first order displacements at any point along the longitudinal axis can be found by
using equations D3 through D5. With equation 15, obtained form equations D4, the bending
displacement at midspan (x = 1400) can be analysed.
F. a. ( L 1400) . 2 2 2 F. a. 1400. 2 2 2
w( F ) L a (L 1400) L a 1400
6. E. Iy . L 6. E. Iy . L Eq. 20
Experiments 1BX
50
average load F [kN]
40
30
1B2
20 1B3
1B3
10 [Eq.20]
N EN 6771
0
0 5 10 15 20 25 30
w2 [mm]
Vertical deflection at x = 850
The results of the experiments and of equation 15 are plotted in figure 73 and it is clear that there is a
difference in stiffness. This is remarkable because one would expect an exact fit since only bending is
present in the member. However equation 15 does neglect shear deformations. For the members of
this experiment, with a slenderness of L/h = 17.5, this is expected to be very little.
7.1.3 strain
In experiment 1B4 strain has been measured. In figure 67 it is shown that the strain at midspan is
nearly the same as at x = 850. This is expected since the bending moment is of constant value
between the points of load application. Strain gauges 20 and 21 show a small difference, indicating a
small bending moment in the bottom flange.
To determine the bending moment from the strain seems inaccurate: a small deviation in strain means
a large deviation in bending moment. However, the values are in the order of the bending moment
determined from the applied load, see table 8.
43
7.2 Experiment 2
7.2.1 Failure
The difference between experiments 1B and 2B is that the load is applied eccentrically to the shear
centre, causing torsion. In failure load this difference becomes apparent: Fu;1B = 49.4 versus Fu;2B =
43.7 [kN], or Fu;2B = 88 % Fu;1B. If the load is applied at the top or bottom of the section then this
further influences the failure load. In figure 74 (a) the failure loads of experiment 2 are compared to
the one of experiment 1. In figure 74 (b) the failure loads of experiments 2 are compared amongst
themselves.
Fu;2A = 37.5 = 75.8% Fu;1B 79% Fu;2C
e
Fu;2C = 47.3 = 96% Fu;1B
Fu;2C = 47.3 [kN]
(a) failure in relation to load at shear centre (b) failure in relation to load at bottom flange
e
Fc;2C = 25 = 60% Fmax;s;d
Fc;2C = 25 [kN]
prediction relative to load at shear centre prediction relative to load at bottom flange
figure 75: predicted failure loads and influence of load point location on failure
The predicted failure loads, from table 4, table 5 and table 6, are shown in figure 75 (a) in a similar
fashion as the experimental failure loads. It can be concluded that they are far off from the actual
failure loads, leaving all but one conclusion: the Merchant-Rankine postulate can not accurately
predict the failure load.
The Merchant-Rankine postulate could possibly be adapted such that it can more accurately predict
failure loads.
Some considerations:
• In the expression for the Merchant – Rankine postulate (equation 1) the plastic section capacity
Fpl needs to be entered. In case of experiment 2 this capacity had been approximated at Fpl = 35
[kN]. If we look at the failure loads of experiments 2 we see that these are all higher, i.e. the
plastic section capacity is greater than the approximated value. If the value of Fpl is increased in
equation 1, the value of the critical load Fc will increase as well. Thus, by further studying how to
determine the plastic section capacity the critical load might be predicted more accurately.
44
• In figure 75 (b) the predicted failure loads are given relative to the load applied at the bottom
flange. If this figure is compared to figure 74 (b) we see that the relative prediction is in good
agreement with the failure loads in relation to the load applied at the bottom flange.
7.2.2 Stiffness
Similar to § 7.1.2 the first order deformations of the member can be analysed. In this case torsion is
present in the member and the first order rotation at midspan can be analysed by using equation D18
and entering x = 1400 millimetres in it (which is midspan).
sinh( α . λ . L) .
sinh( 1400. λ ) sinh( λ . L. α ) . cosh ( 1400. λ )
sinh( β . λ . L) sinh( 1400. λ ) tanh ( λ . L) F. e . L
φ ( F) cosh ( β . λ . L) . α .
tanh ( λ . L) λ.L λ.L G. It Eq. 16
50 50
40 40
2A 1
30 2A 1 30
2B2
2B2
20 20 2C1
2C1
1storder
1st order 10 rotation
10
displacem ent
0 0
0 10 20 30 -15 -10 -5 0
w 1 [m m ] fi 4 [ º ]
Vertical deflection at m idspan rotation at midspan
(a) (b)
figure 76: bending displacement and rotation of section at midspan
Contrary to the bending displacement the first order rotation overestimates the actual twist of the
section, see figure 76 (b). The cause of this is not clear.
7.2.3 Strain
In figure 69, figure 70 and figure 72 the results of strain measurements of experiment 2 is given.
Due to bending and bi-moment one would expect antimetric graphs. However, strain gauges 21 and
23, which are placed on the tip of the bottom flange (in compression), switch near failure from
compression to tension. Strain gauges 25 and 27, placed on the tip of the top flange (under tension)
do not shown this behaviour. This is due to the fact that, besides the bending moment and the bi-
moment, the sections also bents about the minor axis.
In § 5.5.2 it has been expected that the flanges would be free of shear strain. The shear strain is
measured by the rosette, more particularly by strain gauge 44. From the results of the measurements
with the rosette it appears that this expectation holds true only for experiment 2B3.
Determining the bending moment and bi-moment from the measured strain turns out to be very
inaccurate.
It can be concluded that, in order to get an accurate view on the strain distribution more strain gauges
need to be used in one cross section. Besides gauges on the flanges, there should also be gauges on
the web and on the inside of the section.
45
7.3 General considerations
It can be stated that it is difficult to estimate the lateral torsional buckling load of channel sections as
well as to determine the bending deflections. However, this load case is quite rare in engineering
practice because one would have to adjust the
section to apply load at the shear centre. Experiments 2XX
50
In common practice if load is applied directly
onto the section, as is illustrated in figure 7, it 40
is even more difficult to predict the failure 1B3
Ex p e r im e n ts 2A Ex p e r im e n ts 2A Ex p e r im e n ts 2A
40 40 40
a ve ra ge loa d F [kN]
30 30 30
ave r ag e lo ad F [k N]
ave r ag e lo ad F [k N]
2A 1 2A 1
2A 1
20 2A 2 2A 2
1 s t order 2A 2 20 20
1s t order
FEM FEM
FEM
10
10 10
0
0 10 20 30 0 0
-30 -20 -10 0 -20 -15 -10 -5 0
w 1 [m m ]
v 61 [ m m ] fi 4 [ º ]
V e rti ca l d e fle cti o n a t m i d sp a n
late r al d e f le c tio n at m id s p an r o tatio n at m id s p a n
In § 4.2 Influence of pressure and tension jacks, experiment 2A has been simulated. The model still
needs further fine tuning to completely comply with the properties of the actual section.
Nevertheless, the results are compared to the test results. The failure load is very close to the test
results: Fu;FEM = 36.7 versus Fu;2A = 37.5 [kN]. The displacements at failure is much larger indicating
that the stiffness of the FEM model is less than that of the actual section, see figure 78.
Further analysis using the FEM will yield accurate data on failure loads. By varying the span and the
height of the section tables can be constructed on the stability of channel sections.
46
8 Conclusions
8.1 Experiment 1
In experiment 1 the lateral torsional buckling load has been determined, by applying the load in the
shear centre.
• In three test the section buckled in the same direction. For these tests the average failure load has
been determined at Fu;1B = 49.4 [kN] with a standard deviation of 0.895 [kN].
• In one test the section buckled in the other direction. The failure load appeared to be 10 % higher
than those of the other three experiments and was recorded at Fu;1B1 = 55.6 [kN].
• For all test hardly any rotation nor lateral bending occurred until buckling. When the ultimate
load was reached the section buckled suddenly. This is what is expected in theory.
• To predict the failure load using the method of NEN 6771 [2] yields conservative values.
• The bending stiffness of the section appeared to be less than calculated by first order theory, see
figure 73.
8.2 Experiment 2
In experiment 2 load has been applied onto the section at the top flange, the web and the bottom
flange.
The failure loads decreases when the point of load application is moved from the bottom flange
towards the top flange, see figure 74 (b). This is in agreement with the predicted failure loads, see
figure 75(b). However, the experimental failure loads were much higher than the predicted ones. The
bending stiffness of the section appeared to be less than calculated whereas the rotational stiffness
was higher than calculated by first order theory, see figure 76.
47
9 Recommendations
In § 5.3 Deflection if y and in § 5.4 Rotation about x, it is shown that one has to go through a great
deal of trouble to correct the measured lateral displacement and rotation of the section due to the lack
of stiffness of the support. In follow-up experiments the supports might be designed as suggested in
figure 79 to increase stiffness. Than the displacements at the supports must be measured at the first
test. If they are small enough they can be omitted in the following test, simplifying processing the
acquired data.
test rig
support of bottom
section flange lacks stiffness
improved situation
48
In § 4.4.1 the effect of friction at the supports is investigated. It has been stated that for experiment 1
the influence can not be specified.
The joints that were used were quite old and had some damage in the PTFE-covering which might
greatly increase the amount of friction. By ordering new joints and retaining them very carefully (one
grain of sand will increases the friction) one can be certain that the friction at the supports is as low
as possible.
Experim ents 2C2
Measuring rotation of the section by the special rotation
tangent to curve
recorders caused some unexplained problems. In 50
figure 81 a load rotation curve is given. The curve should 40
In Appendix I it has been investigated if the sections was within the tolerance on shape and
dimension according to Ontw. NEN EN 10279 [5].
As for determining the deviation on dimensions one could confine to taking random checks. If the
results of these checks are satisfying it is useless to systematically check dimensions on all sections,
as has been done in these experiments.
As for checking the deviation in shape, visual inspections seems to be adequate. If one section seems
to be crooked one could still check the shape deviations on a flat table. Just in case of experiment 1,
the sections should be checked because the sections will bend due to the welding of the perpendicular
plates onto the section.
49
References
[1] STAALCONSTRUCTIES TGB 1990,
Basiseisen en basisrekenregels voor overwegend statisch belaste constructies
NEN 6770, NNI, Delft, 1991, 175 p.
50
Appendices
Page 2
Appendix Page 3
A. Sections
In this appendix copies of product information on the different hot-rolled channel sections are given.
Each manufacturer uses different notations for dimensions. To avoid confusion the dimensions of the
chosen section, a UPE 160 section, are given in the notations of this report below and on the next
page.
tr tf
( )
t r = r 2 − 1 + t f2 + t 2w
tw
figure A2: thickness at the root of the section
Page 4
position in z- warping
cross ordinate function
section [mm] [mm2]
A* - 80 ωA* = - 2604
A - 75 ωA = - 3036
A** - 70 ωA* = - 3469
B* - 80 ωB* = 2326
B - 75 ωB = 1966
E 0 ωE = 0
C 75 ωC = - ωB
C* 80 ωC* = - ωB*
D 75 ωD = - ωA
D* 80 ωD* = - ωA*
ωB*
ωA*
ωB ωA
ωA**
y-axis
x-axis,
warping function ω
centre line
z-axis
figure A3: warping function of a UPE 160 section
Appendix Page 5
UNP sections
UPE sections
UAP sections
PFC sections
Data obtained from: Parallel Flange Channels, Section Properties and Member Capacities
Ascot UK: The Steel Construction Institute, 1996, SCI-P-210
Page 10
Appendix Page 11
B. Tensile Tests
Determining the yield stress has been carried out as closely to codes EN 10 002-1 [3] and NEN-EN
10025 [4]. In this appendix a report is given of all facts concerning the execution of the tensile tests.
First the location of the samples had to be determined. Parallel flanged channels are not mentioned in
[4] so the designated location of sloped flanged channels was chosen, see figure B1. It is assumed
that it will be the same for parallel flanges. The location of the specimen is at one third of the width
from the tip of the flange.
44
Lo = 79.9 +/- 1 %
R1
2
50 12 Lc = 110 12 50
Lt = 234
The specimen are called proportional specimen because the original measuring length Lo is related to
the original cross section by: L o = k S o with k = 5.65
In figure B2 the tolerance of the width of the specimen is given and in table B1 the area of the
original cross section is determined. The width of specimen 3 has been cut too small but is used
anyway.
Page 12
Le = 50 mm
parallel section
Speed of increments
stress
3 plastic range
2 yield range
1 elastic range
strain
figure B4: different parts of the test
In figure B4 the different parts of the test are indicated For each part a stress or strain increment
speed for the parallel section is determined by [3].
1 elastic range:
The speed on the testing bench was set to 0.6 [mm/min]. From this the stress incremental speed is
0 .6
210000 ⋅
F ⋅ Lt mm E ⋅ ∆L 60 = 19.0 N
calculated as follows: = ∆L = 0 .6 ⇒ σ= L =
E⋅A min t 110 mm 2 ⋅ s
Which falls in the allowed range.
2 yield range:
During yielding the speed can not be changed, but since the yield stress was unknown before the first
test the speed of the test bench has not been changed. To determine the settings for the test bench the
given limits can be changed as follows:
mm
0.00025 ⋅ L t ⋅ 60 = 1.65 < test speed < 0.0025 ⋅ L t ⋅ 60 = 16.5
min
The test speed of 0.6 mm/min was too slow. This had not been noticed until the last test, test number
3, at which the speed was erroneously incremented up to 1.2 mm/min which is also too slow.
3 plastic range:
In this range no lower value for the strain increment is given, the upper value is considerably higher:
mm
test speed < 0.008 ⋅ L t ⋅ 60 = 52.8
min
In this range only a maximum is set down and at no time speeds were reached that exceeded this
maximum.
In table B2 the speed at which individual experiments were carried out are given.
Yield range
The specimen were tested in the 100 kN testing bench and the data was acquired by computer. The
time interval of measuring was 1 second. In table B2 the date and duration of the experiment is
shown.
T e n si le T e st 1 T e n si le te st 2
Yie ld ra n g e Yie ld ra n g e
300
Max = 305,9
Max . = 298 ReH = 298
305
Str e s s [N/m m ^2]
295
300
ReL = 298.5
ReL = 290,8
Min. = 290,8 Min. = 295,9
290 295
0 0.005 0.01 0.015 0.02 0 0.005 0.01 0.015 0.02
s tr a in [% ] s tr a in [% ]
T e n si le T e st 3 T e n si le te st 4
Yie ld ra n g e Yie ld ra n g e
312 305
Max ./ ReH =
310,1
Max . = 301,6
Str e s s [N/m m ^2]
304 295
ReL = 302,6
ReL = 291
Min. = 291
Min. = 295,5
300 290
0 0.005 0.01 0.015 0.02 0 0.005 0.01 0.015 0.02
s tr a in [% ] s tr a in [% ]
200 R2 = 0.9966
150
100
measured data
50
0
0 0.0005 0.001 0.0015 0.002 0.0025
strain [%]
Results
The results of the tensile tests are given in table B3. The average value or the lower yield stress, ReL,
will be used as value for the yields stress in further calculations. All values of measured strain will
be disregarded.
The theoretical elastic lateral torsional buckling moment, Mke, is determined using formulae from
code NEN 6771 [2]. The notations of [2] code are used in this appendix.
Table C1 gives the coefficients that apply to the chosen load case.
F F
4 1.04 0.42
0,25 L 0,5 L 0,25 L
h Ed I z
S= NEN 6771: (12.2-11)
2 GdIt
S = 539.5 [mm]
Coefficients C1 and C2 are from table C1; C2 = 0 when the load is applied at the shear centre:
πC1l g π 2S 2 πC 2S
C=
l1 l1
(
1 + 2 C 22 + 1 +
l1
) NEN 6771: (12.2-10)
C = 3.5 [-]
The coefficient kred is set down for double symmetric sections in [2] but not for channels loaded in
the shear centre. It is assumed that this coefficient will be equal to unity. The theoretical elastic
lateral torsional buckling moment can now be determined using the coefficients kred and C.
C
M ke = k red Ed I zGd I t NEN 6771: (12.2-10)
lg
This moment has a linear relation with the elastic lateral torsional buckling load. From the moment
distribution it can be derived that Fke. = α ⋅ Mke / L, with α = 0.25 in this case.
The theoretical elastic lateral torsional buckling load has been determined. However, the material is
not fully elastic, buckling may occur at a smaller load than has been determined. To determine the
ultimate elastic-plastic lateral torsional buckling load, denoted by My;max;s;d, equation 12.2-3 from
NEN 6771 is used.
Appendix Page 17
M y; max;s;d
≤1 NEN 6771: (12.2-3)
ω kip M y; u ; d
Thus M y;max;s;d ≤ ω kip M y;u ;d and for class 1 and 2 cross sections the following applies:
M y;u ;d
λ= NEN 6771: (12.2-4)
M ke
Table 23 of [1] assigns instability curves for different structural shapes. For channel sections curve c
is assigned for instability about any axis. However this table seems to apply to buckling due to
normal force. Therefore the analogy with lateral torsional buckling of an I-section is used in
determining the instability curve. The section is bent about the y-y axis and both demands h/b =
160/70 = 2.1 > 1.2 and tf = 10 < 40 [mm] are met. Thus the ‘a’ curve is used.
Using equation 12.1-14 of NEN 6770 ωkip can be determined and entered in: M y;max;s;d ≤ ω kip M y;u ;d
The ultimate elastic-plastic lateral torsional buckling load, Fmax;s;d , can be determined from the
moment distribution.
Page 18
F1 F2
α = 0.25
x β = 0.75
α·L β·L L
z
Vz = -F dM (x )
shear force V (x ) =
dx
Vz = F
Vz
dθ(x )
bending moment M(x ) = EI y
dx
My My = α·L·F
dw (x )
angular rotation about the y-axis θ(x ) = −
θ dx
displacement in z-direction w(x)
w
figure D1: distribution of shear force, bending moment and deformations.
In order to find the exact values at each point along the beam axis an expression must be found for
the bending displacement w. By differentiating expressions are found subsequently for the angular
rotation θ, the bending moment My and the shear force Vz.
For a single load an expression for the bending displacement w can be found. The load F1 is placed at
a = α·L from one support and b = β·L is the complementary distance from the other support, see
figure D2 (a).
F1 F1 F2
a = α·L b = β·L a b -a a
L L
(a) (b)
figure D2: position of concentrated load
0 ≤ x < a: w (x ) =
F⋅b⋅x
6 ⋅ EI y ⋅ L
(
L2 − b 2 − x 2 ) Eq. D1
F ⋅ a (L − x ) 2
a < x ≤ L: w (x ) =
6 ⋅ EI y ⋅ L
[
L − a 2 − (L − x )2 ] Eq. D2
Appendix Page 19
If a second load F2 is placed symmetrically from midspan to F1 the expression can be found by simply
swapping the parameters ‘a’ and ‘b’ in equations D1 and D2. If both loads are applied simultaneously
the expression for the bending displacement is found by superposing the expressions for the
individual loads, yielding equations D3 through D5.
0 ≤ x < a: w (x ) =
F⋅b⋅x
(
6 ⋅ EI y ⋅ L
)
L2 − b 2 − x 2 +
F⋅a ⋅x
6 ⋅ EI y ⋅ L
(L2 − a 2 − x 2 ) Eq. D3
F ⋅ a (L − x ) 2
a < x ≤ b: w (x ) =
F⋅a ⋅x
(
6 ⋅ EI y ⋅ L
)
L2 − a 2 − x 2 +
6 ⋅ EI y ⋅ L
[
L − a 2 − (L − x )
2
] Eq. D4
F ⋅ a (L − x ) 2 F ⋅ b(L − x ) 2
b < x ≤ L: w (x ) =
6 ⋅ EI y ⋅ L
[ ]
L − a 2 − (L − x )2 +
6 ⋅ EI y ⋅ L
[
L − b 2 − (L − x )2 ] Eq. D5
In figure D3 the bending displacements due to F1 and F2 are given with α = 0.25 and subsequently
β = 0.75. The chosen load is F1 = F2 = 50 [kN] and the span is 2.8 [m]. If the two graphs for the
individual loads F1 and F2 are added up the bending displacement for experiment 1 is found.
The angular rotation θ is found by differentiation equations D3 through D5, the moment distribution
by differentiating twice and the shear force by differentiating three times. This relation is also shown
in figure D1
The moment distribution and the shear force can be found quite easily as is shown in figure D1.
Bending displacement
15
super posed displacement
displacement due to F1
10
w displacement due to F2
[mm]
5
0
0 700 1400 2100 2800
x [mm]
figure D3: bending displacement due to F1, F2 and superposed displacement
Page 20
The axial stress is largest at the top- and bottom flange at z = ± h/2, see figure D4 c). If the shear
stress distribution along the centre lines is considered the largest stress is at point B and C (see
figure A1 for the location of point B and C) in the cross section, see figure D4 a). However, at B and
C the thickness tr is larger than the thickness tf of just the flange. This is due to the root radius, see
figure A1.
B** s s=0 σx,max
-
s = b0
s h0/2 h/2
τmax C C
C
C** +
shear stress at centre lines elastic section modulus Sy axial stress distribution
a) b) c)
figure D4: stress distribution
To investigate at which point the shear stress is largest two points are considered, point B and point
B**. The shear stress is given by equation D6, in which Sy(s) is the elastic section modulus which is a
function of ‘s’ and t(s) is the thickness at ‘s.’
Vz S y (s )
τ xs = Eq. D6
I y t (s )
This shows that the largest combination occurs at point B** and C**. The Von Mises criteria is used
to determine the elastic section capacity at these points. This criteria is defined by equation D7.
σ tot = σ M y ≤ f y
with: fy Eq. D8
τ tot = τ Vz ≤
3
in which σMy stands for axial stress due to bending and τVz shear stress due to the shear force.
If the expressions for the axial stress and shear stress are entered in equation D7 one finds:
2 2 2 2
M yz VzS y (s) α ⋅ L ⋅ F h2 F ⋅ t f (b − t w − r ) h20
+ 3 ⋅ = + 3 ⋅ ≤ fy Eq. D9
I y I y t(s) I y Iytf
Appendix Page 21
It must be know at which load the elastic section capacity is reached. Writing equation D9 explicit for
the load F yields:
fy
Fel = Eq. D10
2 2
α ⋅ L⋅ h (b − t w − r ) h0
2
+ 3⋅
2
I y Iy
With equation D10 the load at which the elastic section capacity is reached, denoted by Fel, can be
analysed for any span.
Wy,pl =
A
2
⋅ sz =
23718
2
.
⋅118.9 = 141⋅103 mm 3 [ ] Eq. D11 figure D5: plastic stress
distribution
However there is also a shear force present and the plastic section capacity must be reduced. This is
done by rewriting equation D12 to:
In which σ pl = f y2 − 3τ 2
F1
s.c. F2
e
y
+φ F1 e
F2 c.g.
z
x
experiment 2
α⋅L
β⋅L
L
In [6] a solution of the differential equation of torsion is given for a single concentrated torsional
load:
sinh (αλ L )
sinh (xλ ) − sinh (αλL )cosh (xλ )
β⋅L ≤ x < L: ϕ(x ) =
(
tanh λL ) α
+ (L − x ) M x L Eq. D16
λL L GI t
GI t
In which λ is the section parameter for torsion which is defined by: λ =
EI w
From figure D6 it follows that the angular rotation is negative with regard to the chosen co-ordinate
system, therefore the torsional moment is Mx = - F ⋅ e. The eccentricity ‘e’ is 25.23 [mm], see
Appendix A, and the force for which these figures are made is 50 [kN], thus the torsional moment is
Mx = 1.26 [kNm].
Appendix Page 23
Angular rotation
0
0.07
angular rotation due to Mx;t2
0.15
0 700 1400 2100 2800
Span L [mm]
figure D8: angular rotation due to Mx;t1, Mx;t2 and superposed rotation
By superposing equations D15 and D16 for a torsional load at α⋅L on the same equations for a load at
β⋅L the angular rotation for experiment 2 is found. The superposed equation can be given in three
parts, equations D17 through D19.
0 ≤ x < α⋅L :
. .L) .
sin ( αλ . ) sin ( λ.L.α).cos( xλ
. )
sin (xλ
φ (x) sin ( β.λ.L) cos(β.λ.L) . sin ( xλ
. ) tan( λ.L) .
α .Mx L
tan( λ.L) λ.L λ.L .
GIt Eq. D18
β⋅L ≤ x < L:
Eq. D19
sinh( α.λ.L ) . sinh( β.λ.L ) .
sinh( x.λ ) sinh( λ.L.α) .cosh( x.λ ) sinh( x. λ ) sinh( λ.L.β ) .cosh( x.λ )
tanh( λ.L) tanh( λ.L) x. Mx .L
φ( x) α β ( α β) .
λ.L λ.L L G.It
By differentiating the expression for the angular rotation expressions for the Saint-Venant torsion, bi-
moment and warping restraint torsion can be obtained. The relation with the angular rotation is given
by equation D20.
dφ
Saint - Venant Torsion M x; t = GI t
dx
d 2φ
Bi - moment B x = −EI w Eq. D20
dx 2
d 3φ
Warping restraint moment M x;w = −EI w
dx 3
In figure D9 through D11 the distribution of Saint-Venant torsion, bi-moment and warping restraint
torsion is given.
Page 24
6
Saint-Venant torsion
1.1 10
Mx;t [Nmm]
6
1.1 10
0 700 1400 2100 2800
Span [mm]
figure D9: distribution of Saint-Venant torsion
Bi-moment
0
Bx [Nmm^2]
8
2 10
6
Warping restraint torsion
1 10
Mx;w [Nmm]
6
1 10
0 700 1400 2100 2800
Span [mm]
figure D11: distribution of warping-restraint torsion
axial stress
The axial stresses are comprised of stress due to bending and due to the bi-moment. The sum of these
two components may not exceed the yield stress, see equation D21
σ x = σ M y + σ Bx ≤ f y Eq. D21
σx = - 256
σx = - 82
σx = -174 B* σx = -69
σx = 105
σx = -30
σx = 122
y-axis
+ =
σBx σMy x-axis
C*
z-axis
figure D12 : axial stress due to warping restraint torsion and due to bending. The rightmost section gives the superposed
axial stress. The depicted sections are shown in isometric projection, the proportion in x-, y- and z-direction
are 1:1:1. The relation between drawing units and stress is one to ten.
In the cross section the combination of these stresses is largest at the corners B* and C*, see
figure D12. In table A2 values for the z-ordinate and the warping function ‘ω’ are listed. The axial
stress at B* due to bending is given by equation D22 and due to the bi-moment by equation D23.
M yz F⋅α⋅L − h
σMy = = Eq. D22
Iy Iy 2
B x ω(s )
σ Bx = =
Iw
F ⋅ e sinh (αλL ) sinh (βλL ) Eq. D23
− cosh (αλL ) sinh (αλL ) + − cosh (βλL ) sinh (αλL ) ⋅ ω B*
λ tanh (λL ) tanh (λL )
σ Bx =
IW
The expression for Bx in equation D23 is found by differentiating equation D17 twice, multiplying it
by –EIw and substituting F⋅e for Mx. The expressions for the axial stress are now such that factoring
for F is quite easy which will come in handy later.
The stresses in figure D12 are obtained from equations D22 and D23 in which ‘z’ and ‘ω’ were
altered. The depicted cross section is at x = α⋅L = 0.7 m. The force for which the stresses have been
determined is 30 [kN].
fy
τ xs = τ Vz + τ M x ;t + τ M x ; w ≤ Eq. D24
3
with: τvz stands for shear stress due to the shear force Vz
Page 26
τMx;w for shear stress due the Saint Venant torsional moment Mx;t
τMx;w stands for shear stress due to the warping restraint torsional moment Mx;w.
The shear force is constant from 0 < x < α⋅L and β⋅L < x < L, see figure D1. The Saint Venant-
torsion is largest at the supports, see figure D9, and the warping restraint torsion is largest at x = α⋅L
and x = β⋅L, see figures D11. The largest combination is found at x = α⋅L and x = β⋅L.
In § Elastic section capacity the point in the cross section at which the largest combination of axial-
and shear stresses occurs depends on the shear stress distribution. However, due to the presence of
torsion, the point at which the largest combination of stresses will occur depends now on the axial
stress since it is a factor larger than the shear stress is. The axial stress is largest at point B* and C*.
In figure D13 the distribution of shear stress is given.
τmax
τ
τmax
Mx;w
Mx Mx;t
The components of equation D24 are expanded in equation D25 through D27
Vz S y (s ) Vz b 0 h 0 t f V
τ Vz = = = 2962.3 ⋅ z Eq. D25
I y t (s ) 2 ⋅ Iyt r Iy
M x;t
τ M x;t = ± t (s ) = Eq. D26
It
sinh (αλL ) sinh (βλ L )
− F ⋅ e ⋅ − cosh (αλ L ) cosh (αλ L ) + − cosh (βλ L ) cosh (αλL ) + (2 − α − β )
tanh (λL ) tanh (λL )
τ M x;t =± tr
It
The expression for Mx;t in equation D26 is found by differentiating equation D17, multiplying it by
GIt, substituting F⋅e for Mx and entering α⋅L for ‘x.’
M x ; w S ω (s )
τ M x;w = = Eq. D27
I w t (s )
− sinh (αλL ) sinh (βλL )
− F ⋅ e ⋅ + cosh (αλL ) cosh (αλL ) − − cosh (βλL ) cosh (αλL ) ⋅ S ω
τ M x;w = tanh (λL ) tanh (λL )
Iw tr
The expression for Mx;w in equation D27 is found by differentiating equation D17 three times,
multiplying it by –EIw, substituting F⋅e for Mx and entering α⋅L for ‘x.’
Appendix Page 27
5000
b0 b0 h0
5000
0 141.75 283.5
A B s C D
4
4 10
b0 b0 h0
∫ ω ds 0
3
[mm ]
4
4 10
4
8 10
0 141.75 283.5
s
5
4 10
b0 b0 h0
Sω (s ) = t (s ) ω ds
∫ 0 elastic section modulus
for warping restraint torsion
[mm4]
5
4 10
5
8 10
0 141.75 283.5
s [mm]
figure D14: warping function along s, integrated warping function and elastic section modulus for warping restraint torsion
From figure D13 it can be seen what sign the respective shear stresses must have and from equations
D25 through D27 the respective stresses can be determined.
The elastic section modulus for warping restraint torsion Sω is used in equation D27. This dimension
is found by integrating the warping function of figure A3. Thus elastic section modulus for warping
restraint torsion is the volume under the warping function. The warping function at the centre lines,
which is called contour warping, represents the average warping over the thickness. One can integrate
the contour warping as a line function instead of the area of the full warping function.
In figure D14 the warping function is given along the centre line (along the ordinate s, see figure
D4b), below that the integrated warping function is given. To incorporate the thickness of the section
this function is then multiplied by the thickness of respectively the flanges and web. Because of the
difference between the thickness of web and flanges a discontinuity occurs in the function of Sω.at
s = b0. The actual function will be fluent because the difference in thickness is bridged by the root.
Thus the average of Sω at the flange and the web is taken in equation D27, which is shown by the
dashed line in figure D14.
Page 28
Equations D21, D23, D25, D26 and D27 are written such that the load F can be factored easily and an
expression for the elastic section capacity for torsion and bending, Fel,torsion, can be obtained.
Wy;plastic 141004
α= = = 117
. Eq. D29
Wy;elastic 120670
To approximate the ultimate load for sections subjected to bending and torsion it is suggested that the
same procedure is used. This procedure is just to get an idea of the magnitude of the ultimate load.
De krachtsinleiding wordt gecontroleerd ter plaatse van de puntlasten en ter plaatse van de
reactiekrachten. De grootste voorspelde kracht is F max;s;d en is gelijk aan 41.7 [kN]. De reactie kracht
is in grootte gelijk en richting tegengesteld aan deze belasting. In de toets moet de grootste kracht ten
gevolgen van de belasting ingevuld worden en is aangeduid met Fs;d .
Fu;d is de kleinste waarde van Fu;1;d , Fu;2;d of Fu;3;d Deze waarden worden bepaald met art. 14.2.1,
14.2.2 en 14.2.3 en zijn bepaald voor de puntlast en de reactiekracht.
Toets:
Fs;d 41.7
= = 0.64 ≤ 1 voldoet
Fu ;d 65.26
Het buigend moment is nul bij de oplegging, dus σf;d = 0 ; het profiel is uit één materiaal, dus fy;f;d =
fy;w;d en d1 wordt:
bf 70
d1 = t f
tw
= 10
6.5
= 32.8 [mm]
2
bf f y;f ;d σ
d1 = 2 ⋅ t f 1 − f ;d
tw f y;w ;d f y;f ;d
De gekozen overspanning is 2800 millimeter. De langsspanning σf;d in de hoek van het lijf met de
flens wordt gegeven door:
2
70 295.7 2321
.
d1 = 2 ⋅10 1− = 40.659
6.5 295.7 295.7
Fu;1;d = (c+d1)• tw• fy;d = (23.5 + 40.659) • 6.5• 235 = 123.3 [kN]
Appendix Page 31
Eindoplegging: bef = h
h
t ½h
t c
. ⋅ t 2w E ⋅ f y;d f + 3 ⋅ w
Fu ; 2;d = 0125 = 65.26 [kN ]
t w t f h − 2 t f
x =100 F
t t c
Fu ; 2;d = 0.5 ⋅ t 2w E ⋅ f y;d f + 3 ⋅ w = 261 [kN]
t w t f h − 2 t f
c
En er moet gelden: = 017
. < 0.2 voldoet.
h − 2t f
Het lijf wordt beschouwd als een gedrukte staaf met een breedte bef en een lbuc= h
lengte lbuc = h, zie figuur E3.
h
Voor de effectieve breedte geldt: b ef = + x ≤ h
2 bef = h
Invullen geeft bef = 80 + 100 =180 > 160 dus bef =h
figuur E3: globaal plooiien
Aan de hand van NEN 6770 art. 12.1.1. kan de staaf getoetst worden: van het lijf
I y;lijf =
1
12
b ef t 3w = 3.6 ⋅ 103 [mm ] 4
. ⋅ 10 3
A lijf = b ef t w = 104 [mm ]
2
I y;lijf
i y;lijf =
A lijf
= 188
. [mm ]
l buc E
λy =
i y;lijf
= 85.27 [−] λc = π
f y;d
= 83721
. [−]
λy
λ rel = = 1019
. ⇒ knik curve c ⇒ ω buc = 0.529
λc
F. Test Rig
Plans
1 Top view 2 Front view
B'
A A'
C' B
3 Left view
scale
0 300 600 900
C
figure F1: test rig (dimensions are in millimetres on given scale)
Appendix Page 33
1 Top View
1a Section B-B'
y - axis
scale
0 300 600 900
2 Front view
scale
0 300 600 900
hinge
z - axis
x - axis
hinge
jack
hinge
anker
prestressing cable
anker
load cell
z - axis
y - axis
scale
0 300 600 900
Photographs
figure F6: bird eye view of test rig (photograph: Ben Elfrink)
Page 38
section A-A'
Load bearing yokes
A A'
figure F7: the three load bearing yokes in frontal view and cross section
line of work
Load at shear centre (1B) Load at top flange (2A) Load at line of symmetry (2B)
Appendix Page 39
Details
Experiment 1B
figure F9: plans of fitting for applying load at the shear centre
displacments measured
at shear centre
Experiment 2B
figure F11: plan of load pin for applying load at the line of symmetry
9.1.2 Experiment 2A
(a) (b)
figure F14: applying load at the ‘top’ flange (in experiments under)
Appendix Page 43
Fittings to sections
Experiment 1B
Experiment 2A
Experiment 2B
Experiment 2C
G. Supports
joint
rod
figure G3: parameters for determining friction in ball and socket joint
H. Calibration
Used equipment
In table H1 all used recorders are listed. The range in which they were used is given and the deviation
of linearity is given.
The precision of the load cells is much higher than the listed one, however, they have been calibrated
in a testing bench on which the reading has a deviation of 1 % linearity. Thus, the precision of the
load cell can not exceed that of the testing bench.
In case of displacement recorders, the displacement is imposed on the recorder and is therefore the
independent variable (the x-value) in the calibration. The current is the output (the y-value). Using
the method of least squares a linear relation between displacement and current is found. This
displacement is displayed in the graphs, in which the number preceding ‘x’ is the calibration factor.
The value of R2 indicates the precision of the linear line fitted through the observed data. The closer
R2 is to the value of ‘1’ the better the fit. The value of R2 is displayed in the graphs as well.
In case of the rotation recorders the independent variable is the imposed angle and for the load cell it
is the applied load.
Page 50
C ur r e n t [m V ]
C ur r e n t [m V ]
0 0 0
0 50 100 0 50 100 0 50 100
-1000 -1000 -1000
8000 8000
measured measured
Linear 7000 Linear 7000
6000
Cur r e nt [m V ]
6000
Cur r e nt [m V ]
5000 5000
4000 4000
y = -115.02x + 5474.8
R2 = 0.9993 y = -117.05x + 4826.6
3000 3000
R2 = 0.998
2000 2000
-20 -10 0 10 20 -20 -10 0 10 20
A n g le [º] A n g le [º]
Ca lib ra tio n o f Ch a n n e l 5
7000
measured
Linear 6000
Cur r e nt [m V ]
5000
4000
y = -108.3x + 4544.7
3000
R2 = 0.9984
2000
-20 -10 0 10 20
A n g le [º]
Page 52
8000 8000
measured measured
6000 6000
Linear Linear
4000 4000
Cur r e nt [m V ]
2000
Cur r e nt [m V ]
2000
0 0
-2000 0 5 10 -2000 0 5 10
-4000 -4000
-6000 -6000
-8000 y = -1999.9x + 8791.1 -8000 y = -2001.9x + 8605.2
R2 = 1 R2 = 0.9998
-10000 -10000
-12000 -12000
Dis p lace m e n t [m m ] Dis p lace m e n t [m m ]
measured measured
10000 10000
Linear Linear
5000 5000
Cur r e nt [m V ]
Cur r e nt [m V ]
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
-5000 -5000
-15000 -15000
Dis p lace m e n t [m m ] Dis p lace m e n t [m m ]
Appendix Page 53
60 80
meas ured
40 60 meas ured
Linear
Linear
40
20
Cu r r e nt [m V ]
Cu r r e nt [m V ]
20
0
0 10 20 30 40 0
-20 0 10 20 30 40
-20
-40
-40
y = -3.1202x + 51.5
-60 R = 0.9998
2
-60 y = 3.2325x - 72.77
R2 = 1
-80 -80
Ca lib ra tio n o f Ch a n n e l 62
80
60 meas ured
40 Linear
20
Cu r r e nt [m V ]
0
0 10 20 30 40
-20
-40
-100
Dis p lace m e n t [m m ]
Page 54
C ur r e n t [m V ]
0
20
-20 5 10 15 20 25
0
-40
-20 5 10 15 20 25
-60
-40
meas ured -80 meas ured
-60 Linear -100 Linear
-80 -120
Dis p la c e m e n t [m m ] Dis p la c e m e n t [m m ]
Appendix Page 55
Channel 70 71 72 73
Displacement Current Current Current Current
[mm] [mV] [mV] [mV] [mV]
0 0.76 9.2 -0.68 -5.12
0.5015 3.8 12.38 2.66 -2.5
1.008 6.94 15.66 6.22 1.32
1.5405 10.2 19.06 9.92 5.4
2.2015 14.24 23.28 14.5 10.48
2.9075 *) 27.8 19.4 15.06
3.601 22.82 32.3 24.24 21.14
4.318 27.2 36.96 29.1 26.54
5.0075 31.5 41.46 33.96 31.7
5.703 35.82 46 38.78 36.82
6.411 40.3 50.64 43.78 42
7.112 44.68 55.18 48.7 47.02
7.8035 49.06 59.68 53.6 51.96
8.5415 53.7 64.4 58.8 57.16
9.0065 56.62 67.22 62.1 60.44
9.5385 59.9 67.22 **) 65.9 64
10.0635 63.1 67.22 **) 69.66 67.56
Cu r r e nt [m V ]
50
40
40
30
30
20
20
meas ured meas ured
10 10
Linear Linear
0 0
0 5 10 0 5 10
50 50
Cu r r e nt [m V ]
40 40
30 30
20 20
meas ured meas ured
10 10 Linear
Linear
0
0
0 5 10
0 5 10 -10
R2 = 1
5 y = -0.1027x + 0.1126
Cur r e nt [V ]
-3 R2 = 1
4
-4
3 meas ured
2 -5
Linear meas ured
1 -6
Linear
0 -7
0 20 40 60 80 -8
L o ad [k N] L o ad [k N]
Re sid u e s o f ch a n n e l 90 Re sid u e s o f ch a n n e l 91
0.015 0.02
0.015
0.01
0.01
Cur r e nt [V ]
Cur r e nt [V ]
0.005 0.005
0
0
-0.005 0 20 40 60 80 100
0 20 40 60 80 100
-0.005 -0.01
-0.015
-0.01 -0.02
L o ad [k N] L o ad [k N]
Appendix Page 57
Cur r e nt [V ]
R2 = 1
4
-3
3
-4
2
meas ured -5 meas ured
1 Linear
-6 Linear
0
0 20 40 60 -7
L o ad [k N] L o ad [k N]
Re sid u e s o f ch a n n e l 90 Re sid u e s o f ch a n n e l 91
0.02 0.03
0.015 0.02
0.01
Cur r e nt [V ]
0.01
Cur r e nt [V ]
0.005
0 0
-0.005 0 20 40 60 80
-0.01
0 20 40 60 80
-0.01
-0.02
-0.015
-0.02 -0.03
L o ad [k N] L o ad [k N]
Page 58
I. Dimensions
Restrictions on dimensions
In draft code NEN EN 10279 [5] restrictions on tolerance and shape deviation are given for hot rolled
channel sections, see figure I5. In figure I1 it is shown on which dimensions restrictions are given and
in table I1 the tolerance levels for an UPE 160 are given. In table I2 the results of measurements of
the width and the thickness of the flanges are given. These results of width and thickness are well
within the tolerance levels.
The height of the section has not been measured systematically. However, from the observations
made it can be said that it is within ± 0.5 millimetres, well within the prescribed tolerance level of ± 2
millimetres.
From visual inspection it can be concluded that the out of squareness of the sections was well within
the set limits of table I1.
The straightness of the cut sections has been measured. It was clear that the sections were completely
straight as far as the straightness qxx was concerned. After a few random checks these measurements
were abandoned.
The straightness about the other main axis, qyy, has been measured from a flat table, as shown in
figure I2. At four positions along the section the distance from the table to the top surface of the
section, b* has been measured. The values of b* at the supports, position ‘0’ and ‘3’, are entered in the
formula depicted in figure I2. By subtracting the value of the width of the flanges, ‘b’, from the value
of the expression for y(x) the straightness can be determined, see table I3 and I4.
Appendix Page 59
x L
L/4 L/2 b*3 − b*0
L/4 y(x ) = x + b*0
L
qyy
b0
b0* b3*
flat table
position: 0 1 2 3
In table I3 the results for sections of experiment 2 are given and they are plotted in figure I3. From
the table and the graphs it can be determined that the sections are almost completely straight.
Str aig h tn e s s q -yy top flange Str aig h tn e s s q -yy top flange
1 1
0.5 0.5
offs e t [m m ]
offs e t [m m ]
0 0
0 700 1400 2100 2800 0 700 1400 2100 2800
-0.5 -0.5
-1 -1
L e n g th [m m ] L e n g th [m m ]
0.5
offs e t [m m ]
0
0 700 1400 2100 2800
-0.5
-1
L e n g th [m m ]
figure I3: straightness qyy for specimen 2A1, 2B2 and 2C1
In table I4 the results for sections of experiment 1 are given and they are plotted in figure I4. All four
sections are clearly bent one way, due to the heat brought into the section while welding on the
perpendicular plates. The curvature could also be seen quite well.
The plates were welded on at the points of load application, which corresponds with position 1 and 2
of figure I2. Most of the curvature was concentrated around the weld. One could argue that the
sections is bent about these points and that in between the section is straight.
Str aig h tn e s s q -yy top flange
Str aig h tn e s s q -yy
top flange
bottom flange
Exp e r im e n t 1B1 bottom flange
Exp e r im e n t 1B2
4 4
3 3
offs e t [m m ]
offs e t [m m ]
2 2
1 1
0 0
0 700 1400 2100 2800 0 700 1400 2100 2800
L e n g th [m m ] L e n g th [m m ]
Str aig h tn e s s q -yy top flange Str aig h tn e s s q -yy top flange
Exp e r im e n t 1B3 bottom flange Exp e r im e n t 1B4 bottom flange
4 4
offs e t [m m ]
offs e t [m m ]
3 3
2 2
1 1
0 0
0 700 1400 2100 2800 0 700 1400 2100 2800
L e n g th [m m ] L e n g th [m m ]
figure I4: straightness qyy for specimen 1B1, 1B2, 1B3 and 1B4
member 2 width of flange thickness of flange member 3 width of flange thickness of flange
bottom- top- bottom- top- Bottom- top- bottom- top-
experiment position b [mm] b [mm] t [mm] t [mm] experiment position b [mm] b [mm] t [mm] t [mm]
1B1 0 69.4 69.6 9.9 9.9 2A1 0 69.6 69.8 10 9.9
1 69.4 69.7 10 10 1 69.5 69.5 9.9 10
2 69 69.4 10 9.8 2 69.7 69.7 9.9 9.9
3 69.4 69.5 9.8 9.9 3 69.5 69.6 10 10.1
1B2 0 69.4 69.5 9.8 9.8 2A2 0 69.6 69.6 9.9 9.9
1 69.35 69.5 9.9 9.8 1 69.5 69.9 9.8 9.8
2 69.4 69.5 9.9 9.8 2 69.4 69.6 9.9 10
3 69.4 69.5 10 9.9 3 69.7 69.75 9.8 9.85
1B3 0 69.25 69.3 10 10 2C1 0 69.7 70 9.9 9.7
1 69.2 69.25 10 10 1 69.45 69.8 9.9 9.75
2 69.3 69.35 9.9 9.9 2 69.4 69.8 9.9 9.8
3 69.6 69.4 10 9.8 3 69.5 69.4 9.9 9.75
1B4 0 69.3 69.5 9.9 9.9 2C2 0 69.55 69.6 10 9.9
1 69.2 69.5 9.85 10 1 69.6 69.9 9.95 9.9
2 69.5 69.5 9.9 10 2 69.4 69.6 9.8 10
3 69.3 69.6 9.9 9.8 3 69.5 69.4 9.9 9.9
number of observations 16 16 16 16 number of observations 16 16 16 16
mean 69.34 69.48 9.92 9.89 mean 69.54 69.68 9.90 9.88
standard deviation 0.137 0.114 0.071 0.085 standard deviation 0.104 0.177 0.064 0.108
Page 62
figure I5: tolerances for parallel flange channels, obtained from [5]