Sie sind auf Seite 1von 17

Journal of Constructional Steel Research 103 (2014) 327–343

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Interaction behaviour of steel I-girders Part I: Longitudinally


unstiffened girders
B. Kövesdi a,⁎, J. Alcaine b, L. Dunai a, E. Mirambell b, B. Braun c, U. Kuhlmann d
a
Department of Structural Engineering, Budapest University of Technology and Economics, Műegyetem rkp. 3., 1111 Budapest, Hungary
b
Construction Engineering Department, Universitat Politècnica de Catalunya, C/Jordi Girona 1–3, 08034 Barcelona, Spain
c
Space Structures GmbH, Max-Planck-Strasse 3, 12489 Berlin, Germany
d
Institute for Structural Design, Universität Stuttgart, Pfaffenwaldring 7, 70569 Stuttgart, Germany

a r t i c l e i n f o a b s t r a c t

Article history: If steel structures are subjected to the combination of bending, shear and patch loading, the interacting stability
Received 13 December 2013 behaviour in the design should be taken into consideration. The combined loading situation can often occur in
Accepted 28 June 2014 case of bridge girders during launching, therefore the determination of the load carrying capacity under the com-
Available online 23 July 2014
bined loading situation is an important aspect of the bridge design. In the current version of EN1993-1-5 [1] there
is no standard design method to take the interaction of these three effects into account and there are a very small
Keywords:
Interaction
number of previous investigations in the literature about this topic. The current paper focuses on the investiga-
Bending tion of the interaction behaviour of longitudinally unstiffened steel I-girders under the combination of bending,
Shear shear and patch loading. The companion paper (Part II) Kövesdi et al., 2014 [25] deals with the M–V–F interaction
Patch loading behaviour of girders with longitudinal stiffeners. Based on a verified numerical model, the load carrying capaci-
Steel I-girders ties are determined on various girder geometries for various loading situations, and the applicable interaction
Bridge launching equation is investigated in the current paper.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction transverse force and shear and bending is not negligible, and also the in-
teraction of all the three effects (M–V–F interaction) should be consid-
From the modern bridge erection methods the incremental ered in the design of steel structures.
launching process is nowadays widely used due to its numerous advan- In the current version of EN1993-1-5 [1] the calculation methods of
tages considering its time and cost efficiency. This building process, the bending, shear and patch loading resistances are given. Interaction
however, involves a problem with buckling of the thin steel web equations for bending and patch loading (M–F) and bending and
under combined bending, shear and transverse force. During launching shear buckling interaction (M–V) are included in the standard. On the
nearly all cross sections come at least once over a support where a con- other hand previous studies have been executed in the field of bending
centrated reaction force, large bending moment and shear forces are in- and transverse force interaction (M–F) and shear and transverse force
troduced and hereby buckling problems may arise in the slender web interaction (V–F). Interaction equations have been developed for these
panel. Bearing stiffeners give no solutions in case of moving loads, there- two design situations by Braun and Kuhlmann [2]. Meanwhile there is
fore it is necessary to determine the resistance of the web panels under no design interaction equation for the M–V–F interaction behaviour in
the combined loading situation. The magnitude of the three internal EN1993-1-5 [1], and there is a very small number of previous investiga-
forces changes during the launching process and their ratio depends tions in literature dealing with the combined loading situation of all
on the self-weight of the girder and the span. Usually the worst case, these three effects. The only investigations in this field were made by
which gives the maximum utilisation of the girder during launching Braun and Kuhlmann [2] in 2010 and Graciano and Ayestarán in 2013
arises when the girder is in the position directly before reaching a pier [3]. The original aim of Braun and Kuhlmann was the investigation of
(“cantilever”), but another launching phase can also give governing the bending and patch loading interaction (M–F) and the shear and
situations. patch loading interaction (V–F) separately. Based on the results of the
Based on previous investigations, preliminary calculations and pre- M–F and V–F interaction behaviours a combined M–V–F interaction
vious research activities published in the literature, it is obvious that equation has been proposed to consider the combined loading situation.
the interaction between shear and transverse force, bending and But the proposed new 3D interaction equation (M–V–F) has been devel-
oped mainly based on the calculations in the M–F and V–F domains.
⁎ Corresponding author. Tel.: +36 1 463 1998; fax: +36 1 463 1784. The aim of the current research is an in-depth investigation of the
E-mail address: balazskovesdi@googlemail.com (B. Kövesdi). applicability of the proposed M–V–F interaction equation in the 3D

http://dx.doi.org/10.1016/j.jcsr.2014.06.018
0143-974X/© 2014 Elsevier Ltd. All rights reserved.
328 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

domain. The interaction equation is studied in a large parameter range parameter range which was previously not analysed. The current inves-
for various internal force distributions with various girder geometries. tigations showed the reliable application possibility of this design meth-
A similar investigation was made by Graciano et al. in 2013 [3] parallel od. The research results on longitudinally unstiffener I-girders are
to the current research activities. The results of these two research presented in the current paper. Results on longitudinally stiffened
activities complement each other, because the analysed geometrical pa- girders can be found in the companion paper Kövesdi et al., 2014 [25].
rameter ranges are not the same. Graciano et al. [3] analysed only longi-
tudinally unstiffened girders, the current investigations cover a wide
2. Literature overview
parameter range for longitudinally stiffened and unstiffened girders as
well. Furthermore Graciano et al. made their investigations indepen-
2.1. General
dent from resistance models in standards, which means, that the bend-
ing, patch loading and shear resistances are determined by numerical
Due to a very small number of previous investigations in the com-
calculations. The current investigation is executed in the same way
bined M–V–F interaction behaviour a comprehensive literature over-
(standard independent), but also the applicability of the Eurocode resis-
view is made in the surrounding interaction planes (M–V; M–F and
tance models in the M–V–F interaction equation is studied.
V–F planes). The previous experimental, numerical and theoretical in-
In the current paper the existing experimental, analytical and nu-
vestigations are collected for all the three interaction planes. The current
merical investigations in the topic of the interaction behaviour of the
EN1993-1-5 [1] rules are studied in detail and the newest improve-
steel girders are analysed, evaluated and compared. To extend the pre-
ments are collected in this section.
vious investigations and to analyse the applicability of the previous
design proposals a numerical model is developed. The basis of the nu-
merical model development is the experimental investigations execut- 2.2. Previous studies on the M–V interaction
ed in the frame of the COMBRI research project [4]. Based on the
developed numerical model the bending, shear and patch loading resis- An in-depth literature overview about the M–V interaction can be
tances of the analysed girders are determined and the structural behav- found in the PhD dissertation of Sinur [5]. The theoretical background
iour under the combined loading situation is also investigated. The of the M–V interaction is the following. Assuming that shear force in a
research work is completed according to the following research girder is carried only by the web, the ultimate shear resistance is
strategy: reached when the web is yielded uniformly and a full tension field is de-
veloped. These values are independent from the bending moment in the
– Literature overview on the topic of (i) previous investigations on M– panel as long as the moment is less than the bending capacity of flanges
V; M–F and F–V interaction behaviours, (ii) previous analytical and alone. In case of larger bending moments, the moment should be also
numerical investigations of the M–V–F interaction behaviour of carried by the web, which reduces the shear resistance of the girder
steel I-girders. and the M–V interaction should take this into account. If the flange con-
– Development of an advanced numerical model based on shell tribution is taken into account in the shear resistance model, the reduc-
elements with variable geometry and variable loading conditions. tion of axial force in the flange as a consequence of bending moment has
Investigation of structural behaviour under different loading condi- to be also considered in the design. The first formulation of the M–V in-
tions and under a combined loading situation. teraction equation was proposed by Basler in 1961 [6], as presented in
– Verification of the numerical models based on test results. Fig. 1.
– Numerical parametric study to investigate the effect of the different
geometric parameters on the interaction behaviour. In the first step where:
the interaction behaviour in the M–F; M–V and V–F planes is
analysed, and finally in the 3D interaction domain. Mf,R is the flange bending resistance,
– Comparison of the numerical results with the M–V–F interaction Mpl,R is the plastic bending resistance of the girder,
equation proposed by Braun and Kuhlmann [2]. Vw,R is the shear resistance of the web.
– Statistical analysis based on the large number of numerical calcula-
tions to determine the safety level of the new M–V–F interaction The interaction of shear and bending in the web is defined when the
equation. bending exceeds the flange bending resistance. These investigations
give the theoretical background of the current M–V interaction equation
Based on the current numerical investigations the M–V–F interac- of the current EN1993-1-5 [1]. The M–V interaction curve has been
tion equation developed by Braun and Kuhlmann [2] has been verified. thoroughly investigated by several researchers in the past, and the
The numerical calculations proved its applicability in an extended Eurocode rules have been verified by experimental results and

Fig. 1. a, Basler's proposal [6]; b, EN1993-1-5 [1] proposal for M–V interaction.
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 329

numerical calculations. According to EN1993-1-5 [1], provided that the


shear utilisation does not exceed 50%, the design resistance to bending
moment and axial force does not need to be reduced to allow for the
shear force. If the shear utilisation is larger than 50% the combined ef-
fects of bending and shear should satisfy the following criteria:
" # !2
M Ed M f ;Rd V Ed M Ed M f ;Rd
þ 1−  2 −1 ≤1:0 for ≥ ð1Þ
Mpl;Rd Mpl;Rd V bw;Rd Mpl;Rd M pl;Rd

where:
– Mf,Rd is the design plastic moment of resistance of the section
consisting of the effective area of the flanges;
– Mpl,Rd is the design plastic resistance of the cross section consisting of
the effective area of the flanges and the fully effective web irrespec-
tive of its section class.
– Vbw,Rd is the shear buckling resistance of the analysed web panel,
– MEd, VEd are the design values of the applied bending and shear
forces.
The newest improvements and maybe the most comprehensive in-
vestigation in this field was made by Sinur in 2011 [5]. Large scale test Fig. 3. Combined M–V–F interaction surface [2].
specimens have been analysed under combined shear and bending,
and the interaction behaviour of longitudinally unstiffened and stiff-
ened girders is studied. Beside the experimental investigations a large Elgaaly in 1983 [10]
number of numerical simulations have been made to analyse various  3  3
girder geometries and loading situations (different M–V ratios). The F M
þ ≤1:0 ð4Þ
final conclusions on the M–V interaction was that the shape of the FR MR
interaction diagram according to the current version of EN 1993-1-5
[1] generally does not follow the numerically obtained results for longi- Ungermann in 1990 [11]
tudinally stiffened girders. Therefore a new M–V interaction equation is    
proposed and verified. The current interaction formula of EN1993-1-5 is F M
þ ≤1:4 ð5Þ
quadratic, while the obtained response of numerical simulation is for FR MR
most cases linear. With the recommended new equation most of the
numerical results are on the safe side. Johansson and Lagerqvist in 1995 [12]
   3
2.3. Previous studies on the M–F interaction F M
þ 1:25  ≤1:0 ð6Þ
FR MR
The interaction of bending moment and transverse force is an im-
portant aspect of the load bearing capacity as proved by several re- where:
searchers in the past. Research in this topic was started by Bergfelt in
1971 [7]. Numerous tests were executed to determine the interaction M bending moment acting on the girder,
behaviour between patch loading and bending and several interaction F transverse force acting on the loaded panel,
equations have been proposed. A comprehensive overview about the MR bending resistance of the girder,
previously developed interaction equations is made by Lagerquist in FR patch loading resistance of the loaded web panel.
1994 [8]. In this comparison the following interaction equations are
collected: Further numerical investigations are performed by Vigh [13]
analysing the flange to web connection on the interaction behaviour
Bergfelt in 1971 [7] in case of hot-rolled I-sections. Longitudinally stiffened I-girders sub-
jected to combined bending and patch loading were analysed by
 8  2
F M Graciano et al. in 2003 [14]. The main focus of their investigation was
þ ≤1:0 ð2Þ
FR MR to improve the understanding of the interaction behaviour. The im-
provement is achieved by numerical simulations based on experimental
Roberts in 1981 [9] results. The influence of the position of the longitudinal web stiffener
 2  2 was studied in the frame of a numerical parametric study. In the case
F M of cold-formed steel channel sections the bending and patch loading in-
þ ≤1:0 ð3Þ
FR MR teraction was investigated by Ren et al. in 2006 [15]. The main focus of

Fig. 2. Subdivision of the combined load into two basic load cases [2].
330 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

and

F Ed
η2 ¼ ≤1:0 ð9Þ
f yw  Leff  t w
γ M1

In the subject of the M–F interaction behaviour of plated structures


Braun and Kuhlmann [2] executed the latest research work. The aim
of their investigation was to collect all the previous experimental and
numerical results in the topic of M–F interaction and to evaluate the
previously developed M–F interaction equations compared to the ex-
perimental and numerical results. Finally a new interaction equation
Fig. 4. Finite element mesh of the developed numerical model. was proposed, which does not have knick points on it, as the current
EN1993-1-5 [1] interaction equation. The importance of the quadratic
shape of the M–F interaction equation was to be able to combine the
the previous investigations was the ultimate limit state under the com- M–F interaction equation with the V–F interaction equations into a 3D
bined loading situation. Numerical calculations are executed by Granath M–V–F interaction equation. Numerous previous results are collected
et al. in 1998 [16] analysing the serviceability limit state during the trav- and evaluated statistically according to different design proposals. Final-
elling patch load interacting with bending moment. A rough estimation ly the following interaction equation was developed instead of the cur-
of the load level, which should be allowed in order to avoid residual de- rent Eurocode equation.
formations during launching of a bridge girder, was made by Lagerqvist
[8] using the linear part of the load–deflection curves from patch load-
ing tests. These investigations are extended and refined by Granath   !3:6
F M
et al. [16]. þ ≤1:0 ð10Þ
FR Mpl;R
According to the rules of EN1993-1-5 [1], if the girder is subjected to
a concentrated transverse force acting on the compression flange in
conjunction with bending and axial force, the resistance should be ver- where:
ified using the following interaction expression:
Mpl,R is the design plastic resistance of the cross section consisting
of the effective area of the flanges and the fully effective
η2 þ 0:8  η1 ≤1:4 ð7Þ web irrespective of its section class.
FR patch loading resistance of the loaded web panel.
where:

2.4. Previous studies on the V–F interaction


NEd M þ NEd  eN
η1 ¼ þ Ed ≤1:0 ð8Þ
f y  Aeff f y  W eff All studies on the interaction between shear and patch loading have
γ M0 γM0 in common that the combined load is subdivided into two basic load

Fig. 5. a) General layout of the test girder, b) the used notations [4].
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 331

Table 1 In 2001 Roberts et al. made experiments on 24 welded I-girders and


Geometry of the girders used in the test programme [4]. developed an interaction equation for V–F interaction [21]. Based on the
Specimen Web Flange Loading length test series the interaction equation Eq. (14) was proposed:
a [mm] hw [mm] tw [mm] bf [mm] tf [mm] ss [mm]    
V−0:5  F 2:0 F
SP600 2390 600 6 450 20 200 þ ≤0:8 ð14Þ
VR FR
SP1200 2390 1200 6 450 20 200

Kuhlmann and Braun [4] made an extensive research in this topic in


2006. Experimental investigations are made to investigate the structural
behaviour of I-girders under combined shear and patch loading. Large
cases (“pure patch loading” and “pure shear force”). Fig. 2 shows the number of numerical investigation is executed to define the load carry-
separation methodology. The shear stresses due to “pure patch load” ing capacity of girders with different geometries under combined shear
are already included in the patch loading resistance model and a reduc- and patch loading. The results of the numerical simulations of Braun [2]
tion of the load carrying capacity is caused only by the additional shear showed that the proposal developed by Roberts et al. [21] could be used
stresses coming from “pure shear force”. for girders subjected to short loading lengths, it means that the equation
A concise literature overview is made by Braun [2] about the shear may be used when the limit of ss/hw ≤ 0.25 applies. The short loading
and patch loading interaction of steel I-girders. The first available exper- length means that the proposal of Roberts et al. cannot be used for the
imental investigation of this topic was made by Elgaaly in the 1970s investigation of bridge launching situations where the loading length
[17]. Based on the test results the interaction equation Eq. (11) was is significantly larger than the above mentioned limit. Braun and
proposed: Kuhlmann developed a new interaction equation (Eq. (15)) based on
the numerical parametric study executed to investigate the interaction
   1:8 behaviour of stiffened and unstiffened girders. This new interaction
V−0; 5  F 1:8 F curve is fitted as lower bound to the numerical and experimental
þ ≤1:0 ð11Þ
VR FR database.

   
V−0:5  F 1:6 F
where: þ ≤1:0 ð15Þ
VR FR
V shear force acting on the loaded panel,
F transverse force acting on the loaded panel, This interaction equation is based on a larger analysed parameter
VR shear buckling resistance of the loaded web panel, range while the previous interaction equations were based only on ex-
FR patch loading resistance of the loaded web panel. perimental results or only on a limited number of numerical calcula-
tions. The interaction curve developed by Kuhlmann and Braun is
The experimental research activity is continued by Oxfort et al. [18] expressed in the form of Eq. (15) and it is stricter than the previous
on three welded test specimens in 1981. Based on the test results Oxfort ones and covers a wider application field.
et al. [19] proposed to avoid high levels of shear force according to
Eq. (12) when transverse patch loading is present. From the test results 2.5. Previous studies on the M–V–F interaction
it was concluded that a high shear force has no influence on the
transverse patch loading resistance, it was however questioned if Based on the previous investigations on the M–F and V–F interaction
in turn a high level of transverse patch loading has an influence on fields and based on the results of the previous numerical calculations
the shear resistance. About this last point and due to the small num- Braun and Kuhlmann developed a combined interaction equation for
ber of tests and since no tests with a high level of transverse patch the M–V–F interaction behaviour in the form of Eq. (16).
loading were done, it becomes obvious that an assertive conclusion
!3:6  1:6  
cannot be drawn. M V−0:5  F F
þ þ ≤1:0 ð16Þ
  M pl;R VR FR
V−0:5  F
≤0:8 ð12Þ
VR
The interaction surface is illustrated in Fig. 3. This interaction sur-
face fits the requirements in the M–F and V–F planes, thus it has been
Zoetemeier conducted special experiments on hot-rolled I-girders in
presumed to be applicable to 3D interaction surface by Braun and
1980 [20]. Based on the test results the following interaction equation
Kuhlmann. But no verification has been done in the 3D domain.
was proposed:
   2:0
V−0:5  F 2:0 F
þ ≤1:0 ð13Þ
VR FR

Table 2
Mechanical steel properties of the test specimens [2].

Cross-sectional element Yield strength Ultimate strength

fy [MPa] fu [MPa]

Flange, stiffener 354 519


Web 383 543
Fig. 6. First eigenmode shape for SP600 girder.
332 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

Fig. 7. Real geometric imperfection and numerical model with the imperfection shape defined by hand.

Fig. 8. Comparison of the measured and calculated failure modes.

Therefore the aim of the current research is the verification of the 3. Numerical model development and verification
new M–V–F interaction surface in all the three surrounding planes
(M–F; V–F; M–V) and also in the 3D domain for various girder 3.1. Geometrical model and applied analysis method
geometries.
An independent research activity has been executed by Graciano In the current research investigations the structural behaviour and
et al. [3] on the same research field. Based on different experimental re- the load carrying capacity of the analysed girders under the combined
sults found in the literature a numerical model was developed and ver- loading situation is studied by numerical models. The analysed numer-
ified. The applied finite element model was a full shell model. The ical models are developed using the finite element software Ansys 11.0.
verification has been done for all the three different failure modes. On The numerical modelling is based on a full shell model using four node
the verified numerical model 104 numerical simulations have been per- thin shell elements. A detailed overview on the numerical modelling
formed on girders with 5 different geometries. The ratio of the bending and the finite element calculations can be found in the MSc thesis of
moment, shear force and the transverse force has been varied in the ex- Alcaine [22]. Fig. 4 shows the finite element mesh of the developed nu-
ecuted numerical parametric study. merical model.
The final conclusion of the numerical calculations of Graciano et al. The ultimate loads are determined by geometrical and material non-
was that the new M–V–F interaction curve developed by Braun and linear analysis using equivalent geometric imperfections (GMNIA). The
Kuhlmann [2] gives a good approach of the calculated resistances; all Newton–Raphson approach is used in the nonlinear analysis. The
the results of this study were on the safe side.
It has to be mentioned that in the investigations of Graciano et al. [3]
the interaction equation has been evaluated on the bases of the pure
bending, patch loading and shear resistances determined by numerical
calculations. No comparison is made with the standard resistance
models of the EN1993-1-5 [1]. Furthermore all the calculations are
made on longitudinally unstiffened girders, no data are provided for
longitudinally stiffened girders.
Therefore the aims of the current research are to extend the previous
investigations made by the authors mentioned above at the following
points:

– execute a large numerical parametric study to verify the new M–V–F


interaction equation on a much larger parameter range (much more
than 104 numerical simulations),
– evaluation of the new interaction equation on the bases of the pure
resistances calculated by GMNI analyses, and by the resistances
models of the relevant EN1993-1-5 standard. Fig. 9. Comparison of measured and calculated load–displacement diagrams.
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 333

a) shear buckling failure b) bending failure c) patch loading failure


Fig. 10. Applied numerical models to determine bending, shear buckling and patch loading resistances and observed failure modes.

background of the numerical model development is the test results of hardening material model, the real structures will be always on the
the COMBRI research project [4]. The tests were carried out at Luleå safe side. The second aim to apply non-hardening material model is
Technical University (Sweden) having the aim to intend to quantify that the results are compared by the manually calculated resistance
the effect of shear force on the patch loading resistance for welded I- models according to the rules of EN1993-1-5 [1]. All the design methods
girders. In the case of these tests the analysed girder was also subjected are using only the yield strength, and the effect of the hardening is not
to combined bending, shear and patch loading, therefore the results of included in the design models. To ensure the comparability of the nu-
these tests are optimal for the verification of the developed numerical merical results to the standard design values the non-hardening
model. The test layout and the geometry of the test specimens can be model is used.
seen in Fig. 5.
In the experimental test programme two different girders (SP600 3.3. Applied imperfections
and SP1200) have been analysed which geometries are shown in
Table 1. The meanings of the notations can be seen in Fig. 5. The aim of this research work is the determination of the patch load-
ing, shear buckling and bending resistances of the analysed girders. In
3.2. Applied material models the case of all three failure modes buckling can govern the structural be-
haviour, therefore the application of the imperfections is a key point of
The material properties of the test specimens have been determined the calculation procedure. There are different possibilities to define the
by coupon tests. Three individual coupon tests were conducted both equivalent geometric imperfection, all of these are allowed according to
along and transverse the rolling direction of each steel plate which EN1993-1-5 [1].
was used for flanges, stiffeners and webs. The averages of the measured The first possibility is the usage of hand-defined imperfections.
values of the used steel materials are given in Table 2 for the flange and These imperfections represent the commonly observed imperfection
web plates respectively. shapes after fabrications. The manually defined imperfection shape is
In the presented research work two different material models are given often by a functional description, like sine function, or other func-
used. The first model is used for the numerical model verification and tions. Its definition is more complex than the other imperfection types.
the second for the numerical parametric study. The first material The second possibility is the application of eigenmode imperfections.
model follows the measured material properties observed in the tests, The standard allows using the lowest eigenmode as an imperfection
to ensure the best agreement of the numerical calculations with the shape, what corresponds to the buckling shape of the analysed panel.
measured resistances. In the experiments the girder manufacturer pro- These imperfections are defined on a mathematical basis and the
vided the yield and the ultimate stresses of the flange and stiffener ma- modelling technique is also well clarified. In the case of the main struc-
terials at 354 MPa and 519 MPa, respectively. The yield and ultimate tures the application of an eigenmode imperfection shape leads to a safe
stresses of the web plates are 383 MPa and 543 MPa, respectively. design. The third possibility is the application of the collapse-affine im-
These material properties are used in the model verification. The char- perfection shapes. The basis of this imperfection shape is a previous
acter of the applied material model is a linear elastic–plastic material GNMI analysis, where the collapse shape of the analysed girder is de-
model using von Mises yield criterion. This material model has in the fined and the original geometry is overwritten by this imperfect
plastic domain an isotropic hardening behaviour. The material is as-
sumed to behave in a linearly elastic manner and to obey Hooke's law
with a Young's modulus of elasticity equal to 210,000 MPa up to the
yield stress. Thereafter and until it reaches the ultimate stress, the mate-
rial is assumed to behave linearly with a reduced modulus. The ultimate
strength is defined by ε = 10%, which fits with the material tests. The
material was assumed to be perfectly plastic when it reaches the ulti-
mate stress.
In the numerical parametric study the characteristic value of the
yield strength is used in the calculations instead of the measured values.
Only girders made from S355 steel material are analysed in the frame of
these calculations, where the yield strength (fy) is equal by 355 MPa. To
ensure, that all the results of the numerical calculations are on the safe
side the material is assumed to behave linearly after reaching the yield
strength by a reduced Young's modulus of E/10,000. The applied
Young's modulus is 210,000 MPa. The small increase is important to en-
sure the numerical stability in the finite element calculations, but the re-
alistic hardening effect is not taken into account. The reason for it is that
the aim of the calculations is the verification of the M–V–F interaction
surface. If the calculated resistances are on the safe side using non- Fig. 11. Applied numerical model to analyse the interaction behaviour.
334 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

shear buckling failure combined V-F failure (V/F=0,8)

combined V-F failure (V/F=0,6) combined V-F failure (V/F=0,4)

combined V-F failure (V/F=0,2) patch loading failure.

Fig. 12. Typical failure modes for V–F interaction.

shape. This imperfection type is not a commonly used imperfection, Usually this imperfection shape could not be used for the determination
therefore it is used in the current investigations. Three different imper- of the ultimate load, but the effect of the imperfection shape on the
fection shapes are used in the current calculations and the results are buckling behaviour is wanted to analyse using this imperfection
compared to each other. shape. Therefore the numerical model of the test girder SP1200 is built
Hand-defined imperfection shapes having a sine wave form are ap- with the same geometric imperfections than that observed in the
plied according to EN1993-1-5 [1]. This imperfection form is the recom- tests. The imperfect shape of the test specimen and the scaled-up im-
mended imperfection shape of the EN1993-1-5 for sub-panels, where perfect shape in the numerical model can be seen in Fig. 7.
buckling is the governing failure mode. This sine function imperfection The results showed that there are no large differences (smaller than
shape is used for both web panels used in the analysed geometry. The 2%) between the ultimate loads using the different imperfection shapes.
magnitude of the applied imperfection is hw/200 as recommended in Because in the current work different failure modes will be analysed
the standard. separately and in a combined way (interaction behaviours), the best
The second applied imperfection shape is the first eigenmode shape, and always a safe side solution is considered using eigenmode imperfec-
which contains the relevant failure mode. Fig. 6 shows the shape of the tions. This imperfection type can handle the change of the analysed fail-
first eigenmode for the test specimen SP600. The amplitude of the im- ure mode. Maybe the usage of the first eigenmode imperfection shape
perfection is scaled to hw/200 according to the recommendation of leads to a conservative design, but if the applicability of the new M–
EN1993-1-5 [1]. V–F interaction curve can be verified using it, it can be assured that all
The third analysed imperfection shape is also a hand-defined imper- of the calculations are on the safe side. Therefore all the further
fection shape having the same geometry as measured in the tests.
2

1,2 1,8
1,6
1 1,4
F/FR.Chacón

1,2
0,8
F/Fnum,R

1
0,6 0,8
0,6
0,4
0,4
numerical results
numerical results 0,2
0,2 proposed interaction curve (Braun and Kuhlmann)
proposed interaction curve (Braun and Kuhlmann) V / VR,EN3
0
V / Vnum,R 0 0,2 0,4 0,6 0,8 1 1,2 1,4
0
0 0,2 0,4 0,6 0,8 1 1,2
Fig. 14. Results in the V–F plane, evaluation according to resistance models of the EN 1993-
Fig. 13. Results in the V–F plane, evaluation according to FEM based resistances. 1-5 [1].
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 335

a) patch loading failure b) combined M-F failure c) bending failure


Fig. 15. Typical failure modes for M–F interaction.

numerical parametric studies are based on the first eigenmode imper- 3.5. Applied numerical models in the numerical parametric study
fection shape. The necessary imperfection sensitivity analysis is per-
formed also using the eigenmode imperfection shape. The difference Based on the verified numerical model a numerical parametric study
between the lowest and the biggest ultimate load is 1.45%. The differ- is conducted. To be able to validate the new M–V–F interaction surface
ence between hw/500 and hw/200 magnitudes is very small; therefore the numerical simulations have two different aims. The first aim is to
the recommended value of EN1993-1-5 (hw/200) is used in the further determine the pure bending, shear and patch loading resistances of
calculations. the analysed girders. These values are the bases of the evaluation of
the results under the combined loading situation. Expediently, the de-
termination of the pure resistances is made with simplified models,
3.4. Model verification which are developed only for the pure resistance development. The ge-
ometry of these simplified models and the observed typical failure
For the validation procedure the numerical models of both SP600 modes are shown in Fig. 10. The models for the combined loading situ-
and SP1200 specimens are built and the verification is made by the ations are not always applicable for the investigation of the pure resis-
comparison of the results obtained in the experimental programme tances; therefore simplified models are developed to determine these
and the numerical calculations. “pure resistances”. On these models the bending, shear and patch load-
During the test procedure displacements, strains and the applied ing resistances of the analysed specimens are determined.
force were recorded. At certain load increments displacements of the The second aim of the numerical simulations is the study of the steel
full panel are measured with a movable LVDT device. Vertical displace- I-girders under the combined loading situations. All the analysed inter-
ments are also measured during the tests with LVDTs at six different lo- action types, namely the M–F; V–F; M–V and M–V–F interaction behav-
cations. Strains are measured and recorded during the whole test iour can be analysed on the same model geometry by varying the
process continuously. The detailed description of the test procedure applied internal forces. The applied loads on the girders are varied ac-
and the measured results can be found in [4]. In the current investiga- cording to the internal force diagrams to be analysed. The applied
tions the test results measured at the LVDTs are compared to the results loads are the bending moment at the end of the girder, concentrated
of the numerical calculations. The measured and calculated ultimate force at the mid-span (over the internal stiffener) and concentrated
loads are also compared as well as the deflected shape of the loaded force at the 1/4 of the span. The intensity of these forces is varied in
panel. the numerical parametric study.
For the test specimen SP1200 the ultimate load determined by the
numerical model is 1013.2 kN, and the ultimate load in the tests was 4. Investigation of the V–F, M–F and M–V interaction planes
1030 kN. The ultimate shape and the failure mode observed in the
tests and in the numerical simulations are also compared. Fig. 8 shows 4.1. Research strategy
the measured and calculated ultimate shape of specimen SP600. As it
can be seen, the comparison showed for the load–displacement curve Before investigating the M–V–F interaction surface in the 3D do-
and also for the ultimate shapes a good agreement. main, the interaction behaviour in the surrounding planes is analysed.
The difference between the measured and calculated ultimate load is The aims of these calculations are to clarify the applicability of the
1.6% which is a very good agreement. The measured and calculated new M–V–F interaction surface in the surrounding planes and to com-
load–displacement diagrams are also compared which can be seen in pare the new interaction surface to the previously developed 2D inter-
Fig. 9. action curves.

1,2
1 range 2
1 range 2
0,8
0,8
F/Fnum,R

F/Fnum,R

0,6
0,6 range 1

range 1 0,4
0,4 numerical results
numerical results interaction curve (Braun and Kuhlmann) using M.num,R
0,2
interaction curve (Braun and Kuhlmann) using M.num,R interaction curve (Braun and Kuhlmann) using M.pl,R
0,2
interaction curve (Braun and Kuhlmann) using M.pl,R interaction curve EN1993-1-5 M / Mnum,R
0
interaction curve EN1993-1-5 M / Mnum,R 0 0,2 0,4 0,6 0,8 1 1,2 1,4
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4
Fig. 17. Results in the M–F plane, evaluation according to FEM calculated resistances with
Fig. 16. Results in the M–F plane, evaluation according to FEM based resistances. modified interaction curve.
336 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

2
1,8
1,6
1,4
F/FR.Chacón

1,2
1
0,8
0,6
0,4
numerical results
0,2
proposed interaction curve (Kuhlmann/Braun) M / Mpl,R
0
0 0,2 0,4 0,6 Meff/el.R/Mpl.R 1 1,2

Fig. 20. Results in the M–V plane, evaluation according to resistances determined by FEM
Fig. 18. Results in the M–F plane, evaluation according to resistance models of EN 1993-1- simulations.
5 (with Mpl,R) and Chacón et al. patch loading resistance model.

interaction equation, if necessary. Secondly, all numerically calculated


The strategy of the investigation in each plane is the same. The same resistances are compared to the resistance models of the EN1993-1-5
numerical parametric study is carried out in all the three planes on the [1] for the “pure” resistances. This evaluation method is useful to
same girder geometries, only the loading and supporting conditions are study whether the interaction curves using the standard resistance
different. The investigated parameters in the numerical parametric models are on the safe side, or not. Based on this evaluation method
study are the web depth/thickness ratio (hw/tw), the flange width/ the statistical evaluation of the resistance models and the interaction
thickness ratio (bf/tf) and the loading length (ss). The investigated geo- equations is made, to determine the safety level of the interaction sur-
metrical parameter ranges are the following: face. All the results of the numerical calculations are given on diagrams,
the actual values are available in tables in the MSc thesis of J. Alcaine
– hw = 1200–1000–900–800–600–400–300 mm,
[22].
– tw = 6–4 mm,
– bf = 450–350–300–250–200 mm,
– tf = 20–16–15 mm,
4.2. Investigation of V–F interaction plane
– ss = 600–200 mm.
At the first step of the calculations the “pure” resistances of the In the numerical parametric study 11 girders with different geome-
analysed girders are determined by both FEM simulations and by ana- tries are studied (Fig. 11). On each girder 10 calculations have been car-
lytical calculations. The interaction domains are divided by 10 equally ried out varying the ratio of the applied loads. Therefore a total number
spaced points and the internal force distributions are tried to set in of 110 calculations are executed to investigate the V–F interaction be-
such a way that the calculated resistances cover the whole interaction haviour. Fig. 12 shows the observed typical failure modes. The failure
domain as uniformly as possible. It means that for one girder geometry modes under shear and transverse force are the well-known shear
10 calculations are executed by varying the ratio of the applied internal buckling and patch loading failure modes. The combined failure mode
forces. The evaluation method is the same in all the three investigated is the combination of the shear buckling and patch loading failures,
planes. Firstly, all the numerically calculated resistances under the com- what matches the failure mode observed in the tests of Braun [2]. A uni-
bined loading situations are compared to the numerically calculated form transition is observed between the shear buckling and patch load-
“pure” (bending, shear buckling or patch loading) resistances, and the ing failure modes, no strict limit could be observed in the interaction
results are evaluated based on the interaction curve of EN1993-1-5 (if behaviour.
available), and by the new interaction surface proposed by Braun and Fig. 13 shows the results of the numerical parametric study. The cal-
Kuhlmann [2]. This evaluation method makes the comparison of the culated resistances under the combined loads are divided by the results
pure resistances and interaction behaviour independent of the resis- of the “pure” FEM based shear buckling and patch loading resistances.
tance models. This evaluation method is useful to check that the physi- The vertical axis shows the calculated patch loading resistance under
cal background of the interaction equation is correct. This evaluation the combined loading situation divided by the “pure” patch loading re-
method can be also used to make necessary modifications on the sistance. The horizontal axis shows the calculated shear buckling

Fig. 19. Typical failure modes observed in the M–V interaction analysis.
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 337

resistance under the combined loads divided by the “pure” shear buck-
ling resistance.
It can be seen from the diagram that the new interaction equation
presents a good fit of the lowest bound of the numerical simulations.
The smallest ratio calculated from the 110 calculations carried out in
the parametric study is 0.985, as the results show the equation presents
a very good agreement in the V–F domain with the numerical calcula-
tions, all the results are on the safe side (except one calculation).
Fig. 14 presents the results of the same numerical calculations but
the calculated resistances are divided by the shear buckling and patch
loading resistances which are determined according to the EN1993-1-
5 [1] shear buckling resistance model and to the Chacón et al. [23,24]
proposed patch loading resistance model, which is the newest improve-
ment of the Eurocode patch loading resistance model. Fig. 21. Results in the M–V plane, evaluation according to the plastic bending resistance
It can be seen from the diagram that all the 110 calculations are on and shear resistance based on FEM simulations.
the safe side using the EN1993-1-5 [1] resistance models, and the new
interaction equation. Furthermore it can be seen, that the lowest
bound of the results on the vertical axis (patch loading resistance) is
about 1.2. This value means that the patch loading resistance model be seen, that the proposed new interaction equation (yellow dashed
gives 20% smaller results than the calculated resistances according to curve) gives lower resistances that the current Eurocode interaction
the FEM simulations. The lowest value of the results on the horizontal equation. In this equation the bending resistance should be determined
axis (shear buckling resistance) is 1.1. It means, that the Eurocode de- according to the relevant cross section class, but using the factor 0.8 the
sign method gives 10% smaller resistance than the numerical simula- effect of the plastic design is also taken into account in an explicit way. It
tions. It should be also noted, that the scatter of the results on the means that both the Eurocode M–F interaction equation and the new
vertical axis is larger than on the horizontal axis. The same tendency M–F interaction curve proposed by Braun take the effect of the plastic
was observed in the PhD thesis of Braun [2]. The reason for it is the con- resistance into account, but using different ways. Comparing the two
sideration of the flange contribution in the patch loading resistance dashed curves it can be established, that the new interaction curve
model. It is observed in the calculations that the effect of the flange con- gives smaller resistances than the current EN1993-1-5 [1] interaction
tribution resulted in this large scatter. If the flange is more dominant, equation.
the ratio of the calculated results is larger than with smaller flanges. The results also show, that using the plastic bending resistance as the
reference and the cutoff limit by 1.0 on the horizontal axis (according to
4.3. Investigation of M–F interaction plane cross section class), the new interaction equation can give a good lower
bound limit of the calculated resistances. But there are a small number
In the numerical parametric study 11 girders are studied with of specimens which are below the dashed yellow interaction curve.
different geometries, therefore a total number of 110 calculations are These values are also very close to the interaction curve, the minimum
executed to analyse the M–F interaction behaviour. Fig. 15 shows the ratio is 0.958, which can be appointed as a good fit without any safety
observed typical failure modes. The failure modes under bending and considerations. In the marked “range 1” in Fig. 16 there are several
transverse force are the well-known bending and patch loading failure points where current EN1993-1-5 equation results in points which are
modes. The combined failure mode is the combination of them. A uni- on the unsafe side, but these points are on the safe side using the new
form transition is observed between the bending and patch loading fail- interaction equation. It means that in the range of the large bending mo-
ure and no strict limit could be observed in the interaction behaviour. ment, where the difference between the continuous line and the calcu-
Fig. 16 shows the results of the numerical calculations. All the results lated resistances is large, no points are inside of the dashed curve, all the
are evaluated based on the FEM calculated bending and patch loading calculation results are on the safe side.
resistances. Three interaction equations are plotted on the diagram. It has to be also noted, that in the “range 2” (Fig. 16) there are several
The red continuous line shows the new interaction curve using the points, which are inside of both interaction equations (blue and yellow
bending resistance of the girder according to the relevant cross section dashed lines). It means that in this region, both interaction equations
class. The results show that all the numerical calculations resulted in give results on the unsafe side, but the new interaction curve is still bet-
larger resistances than the new interaction curve. It means that all ter than the previous one.
points are on the safe side considering the bending, shear buckling
and patch loading resistances calculated by numerical simulations.
At the same time the results showed, that the new interaction equa-
tion (red line) fits the results quite well by large transverse force and
small bending moment. But in the range of large bending the difference
between the interaction curve and the numerical simulations is signifi-
cant. Although all the results are on the safe, it is worth to modify the in-
teraction equation and use the plastic bending resistance as a reference.
The yellow dashed line illustrates the interaction curve if the plastic
bending resistance is used as a reference in the interaction equation. It
means that the current calculations prove the assumption of Braun
and Kuhlmann to use the plastic bending resistance in the interaction
equation. It has to be mentioned, that the interaction equation of
EN1993-1-5 [1] for M–V interaction considers also the plastic bending
resistance, therefore the application of it in the M–F interaction ensures
the compatibility of the different interaction equations.
The third interaction curve (blue dashed curve) plotted in Fig. 16 is Fig. 22. Results in the M–V plane, evaluation according to the resistance models of EN
the interaction curve of the current version of EN1993-1-5 [1]. It can 1993-1-5 (using Mpl.R).
338 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

Fig. 23. Standard interaction equation development possibilities.

These points could be also on the safe side, if the index of the bend- the index of 3.6 can be also applied and the new interaction equation
ing utilisation ratio in the interaction equation would be changed from gives a lower bound of the numerical results. The only way where the
3.6 to 3.0. The results of this modified curve can be seen in Fig. 17 and interaction equation in its original form gives not the lower bound
the modified interaction equation is shown in Eq. (17). curve is if the patch loading resistance is determined by FEM simulation
and the plastic bending resistance is used in the interaction consider-
!3:0 !
ation. If the interaction curve is intended to be used this way, the
M F
þ ≤1:0 ð17Þ index of the bending utilisation should be changed to 3.0.
Mpl;R F num;R

Fig. 18 shows the evaluation according to the resistance models of 4.4. Investigation of M–V interaction plane
EN1993-1-5 [1] using the plastic bending resistance of the analysed
cross section (independent from cross section class). The results of the The new M–V–F interaction equation was previously verified mainly
numerical simulations show, that using the standard resistance models, in the M–F and V–F planes by Braun [2]. Therefore special attention is
all the results are on the safe side. It can be also stated, that the evalua- given to the calculations in the M–V plane in the current investigations,
tion method using the plastic bending resistance represents the tenden- because nobody analysed the applicability of the new interaction equa-
cy of the numerical results much better, than the evaluation according tion in this plane before. This is relevant when the patch loading force
to the elastic one. Furthermore it can be seen from the diagrams, that
the scatter of the results on the vertical axis is much larger than on
F
the horizontal axis. It means that the bending resistance could be deter- FR
mined with a much larger accuracy, than the patch loading resistance.
The results shown in Fig. 18 mean that using the Eurocode plastic
bending resistance model and the patch loading resistance model of
Chacón et al. [23,24], which is the newest improvement of the Eurocode
patch loading resistance model all the results are on the safe side using
both indices 3.6 and 3.0.
There is another aspect to be considered. The standard resistance
models neglect the effect of the flange resistance from the patch loading
resistance model, or at least the flange part is smaller than that in the M
real resistance. If the patch loading resistance is determined by numer- MR

ical simulations, the web and flange resistances are also taken into ac-
count. Therefore if the interaction equation is used with resistance V
VR
models, where the flange part is neglected, the index of 3.6 can be
used in the interaction equation. If the interaction equation is used by
bending and patch loading resistances determined by FEM simulations, Fig. 24. Research strategy of the M–V–F interaction surface validation.
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 339

Fig. 25. Front view of the interaction surface, evaluation according to FEM based bending, shear and patch loading resistances.

becomes small i.e. tends to zero. EN1993-1-5 [1] contains also an inter- investigation of the interaction plane M–F, where the index of the bend-
action equation for M–V interaction, therefore the comparison of this ing ratio is proposed to change to 3.0, if the interaction curve is used this
equation, and the new proposal is also carried out in this section. way.
Fig. 19 shows the typical failure modes observed in the numerical If the interaction curve (Eq. (18)) with the index of 3.0 is plotted on
simulations. In Fig. 19.a the failure mode is the shear buckling of the the diagram in Fig. 21, it can be seen, that the interaction curve gives a
analysed web panels. The failure mode in Fig. 19.d is the typical bending lower limit of the numerical results.
failure. Fig. 19.b–c represents typical failure modes under the combined
!1:6 !3:0
loading situation, where the failure mode is the interaction of shear
V M
buckling and bending. No strict limit could be observed in the paramet- þ ≤1:0 ð18Þ
V num;R Mpl;R
ric study between these two failure modes, there is a uniform transition
from the bending to the shear buckling failure modes.
Fig. 20 shows the calculation results. The vertical axis shows the The evaluation of the results is made by using the standard resis-
shear force acting on the girder divided by the shear buckling resistance tance models. Fig. 22 shows the results of the numerical calculations
determined by numerical simulations. The horizontal axis shows the evaluated on the bases of the plastic bending resistance (Mpl,R) and
bending moment divided by the bending resistance calculated also by the shear buckling resistance according to EN1993-1-5 [1]. The diagram
the numerical simulations. The diagram shows, that the new M–V–F in- shows, that all the points are outside of the interaction curve, it means
teraction equation gives a good lower bound of the calculation results. that all the points are on the safe side using the index of 3.6. The results
Most of the results are on the safe side and inside points in “range 1” shows that the plastic bending resistance can be used in the new M–V
are also very close to the interaction equation. The smallest ratio is interaction equation, and still all the results remain on the safe side,
0.985 compared to the new interaction curve proposed by Braun and no points are inside of the interaction curve.
Kuhlmann [2].
If the plastic bending resistance is used in the evaluation procedure 4.5. Conclusions according to the results in the boundary planes
with the shear buckling resistance determined by FEM simulations, sev-
eral points can be found inside of the interaction curve using the index Summarising the results which are observed, it can be stated that in
of 3.6 as shown in Fig. 21. This is the same situation as described by the the V–F plane the interaction equation proposed by Braun and

Fig. 26. Back view of the interaction surface, evaluation according to FEM based bending, shear and patch loading resistances.
340 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

Fig. 27. Front view of the interaction surface, evaluation according to FEM based shear and patch loading resistances and using Mpl,R.

Kuhlmann [2] gives a very good lower bound of the numerical Fig. 23 shows a schematic drawing about the different evaluation
simulations. methods of the interaction equation. The first diagram shows the eval-
In the M–F and M–V interaction planes if the bending and patch uation according to shear resistance model considering only the web
loading resistances are determined by FEM simulations, or the standard part and using the plastic bending resistance as reference values. In
resistance models are used as a reference in the interaction equation, this case, the interaction equation with factor 3.6 can be used.
the equation with the index of 3.6 gives appropriate results, all the cal- If the flange part is also considered in the shear buckling and patch
culation results are on the safe side. loading resistance models used in the interaction equation, the index
The calculations show that using the FEM based patch loading and of 3.0 seems to be applicable if the plastic bending resistance is also
shear buckling resistances and the plastic bending resistance as a refer- used in the interaction equation.
ence, the interaction equation with index of 3.0 gives the lower limit of If the flange part is also considered in the shear buckling and patch
the real interaction behaviour. loading resistance models used in the interaction equation (usually cal-
The difference between the two evaluation methods is the patch culated by numerical methods), and the bending resistance is deter-
loading and shear buckling resistance models. In the case of FEM simu- mined also by numerical calculations (according to relevant cross
lations the shear buckling and patch loading resistances take the web section class), the interaction equation with factor 3.6 can be also used.
and flange resistance parts also into account in the resistance model. If
standard design methods are used to determine the shear buckling 5. Investigation of the combined M–V–F interaction behaviour
and patch loading resistances, the flange parts are usually neglected.
The results show, that the interaction equation with index of 3.6 gives 5.1. Research strategy
a lower limit curve if the flange part is neglected from the shear buckling
and patch loading resistance models, what is the case in the Eurocode In the previous sections the compatibility of the new interaction
design method. The previously developed Eurocode interaction equa- curve is checked by comparison to other interaction curves, and the ap-
tions for M–F and M–V interaction take also the fact into account, that plicability of the new interaction surface in each of the three boundary
only the web part is considered in the shear buckling and patch loading planes is studied. This section deals with the verification in the 3D do-
resistance models. Therefore the interaction equations are calibrated to main. These investigations have a special importance in this topic, be-
this evaluation method. If it is considered by the development of the cause only one previous investigation was dealing with the analysis of
new standard interaction curve, the index of 3.6 fits a good lower the new M–V–F interaction equation in the 3D domain.
bound of the numerical calculations, therefore it can be used as a stan- It is intended to cover the whole interaction domain uniformly by
dard interaction equation. calculation results. The strategy of the investigations is the following.

Fig. 28. Back view of the interaction surface, evaluation according to FEM based shear and patch loading resistances and using Mpl,R.
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 341

In the first step for each model the bending, shear buckling and patch resistance on the model used for the “pure” and the combined shear
loading resistances are determined. These are the corner points of the buckling resistance determination can result in points inside of the
3D interaction equations. After it the applied shear force is fixed by a curve.
constant value, and the ratio of the bending moment and transverse The evaluation of the results is also made considering the plastic
force is varied. This calculation method results in points along the red bending resistance as a reference in the interaction equation. The results
curves in Fig. 24. In the next step, the applied transverse force is fixed of this evaluation method can be seen in Figs. 27 and 28.
as a constant value and the ratio of the bending and shear forces is The results show that the assumption of Braun and Kuhlmann to use
changed. In this way calculation results are produced along the blue the plastic moment resistance in the interaction equation is appropriate,
curves in Fig. 24. In average 30 calculations are carried out for each gird- and the interaction equation gives a better approximation of the numer-
er geometry in the 3D interaction domain. Using this strategy the whole ical calculations than using the bending resistance calculated by numer-
interaction domain is covered quasi-uniformly. Based on this database ical simulations.
the applicability of the new M–V–F interaction curve can be evaluated However there are a small number of points in “region 1” and “re-
with adequate reliability. gion 2” (Fig. 28) which are inside of the interaction surface. Almost all
Eleven types of girders with different geometries are investigated in of these points are the same calculation results, which were also inside
the frame of the numerical parametric study. 33 numerical calculations the M–F or M–V interaction planes (see Sections 4.3 and 4.4) and sever-
are carried out to define the “pure” resistances and the points along the al new (pink) points are also inside of the diagram. Because the calcula-
blue and red lines (according to Fig. 24) for each girder. A total of 363 tions in the M–F and M–V planes showed that a refinement of the
calculations are carried out in the frame of the whole parametric study interaction equation is possible by using index of 3.0 for the bending
carried out in the MSc thesis of Alcaine [22]. From the previous studies ratio instead of 3.6, therefore the evaluation is also made by the index
the database of Braun [2] is also used and evaluated in the 3D domain, of 3.0.
to get a larger database for the evaluation of the numerical calculations The results (Fig. 29) show that almost all the calculation results in
(175 calculation results). The evaluation of these results has a certain the 3D domain are outside of the interaction curve. Several points are
importance because in the numerical calculations a small co-existing inside, where the slope of the interaction curve is small and a very
bending moment is always applied to the girders, which is eliminated small difference results in points inside the curve. But the smallest
by the evaluation of the results in the M–F interaction plane. Therefore ratio is also larger than 0.99. Therefore the new interaction curve with
the calculation results of Braun are also plotted in the 3D diagrams, and the index of 3.0 can handle as a good lower bound the calculation results
compared to the 3D interaction surface. A total number of 815 numeri- according to this evaluation method.
cal calculations are evaluated to analyse the new M–V–F interaction It is also interesting, that in the 3D domain the effect of index 3.6 or
equation. 3.0 has a much smaller effect than in the M–F or M–V planes. It is obvi-
ous because the results in the third plane (V–F) fit the lower bound of
the numerical simulations. Therefore a smaller number of points are in-
5.2. Evaluation of the results according to FEM based resistances side of the interaction curve using the index of 3.6 in the 3D domain,
than in the M–F or M–V planes.
The evaluation of the results is made in the same way as described in
the previous sections. Figs. 25 and 26 show the results of the numerical
calculations in the 3D domain. Fig. 25 shows the front view, and Fig. 26 5.3. Evaluation of the results according to standard resistance models
the back view of the interaction surface.
It can be seen from Figs. 25 and 26 that almost all the points are out- The results of the numerical calculations are also evaluated accord-
side of the diagram. There are several points in region 1 (marked in ing to the standard resistance models of EN1993-1-5 [1] using the plas-
Fig. 26) where the results seem to be located inside of the interaction tic bending resistance and the latest improvement of the patch loading
equation, but these are not inside when looking from another perspec- resistance proposed by Chacón et al. [23]. The results are shown in
tive. In region 2 there are several points inside of the interaction surface, Fig. 30.
but all of these points are very close to the interaction surface, the ratios Fig. 30 shows that no points are inside of the interaction curve (ex-
are about 0.98–0.99. This region is located close to the M–V plane and cept of the cut of range). It means that using the resistance models of
close to the small bending moment region. In this region the slope of EN1993-1-5 [1] and the latest proposal of the patch loading resistance
the interaction diagram in the M–V plane is very small. Therefore a model of Chacón et al. [23] and the plastic bending resistance in the
small difference (1–2%) between the calculated shear buckling interaction equation, all the points are on the safe side, the new

Fig. 29. Back view of the interaction surface, evaluation according to FEM based shear and patch loading resistances and using Mpl,R (with index 3.0).
342 B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343

Fig. 30. Evaluation according to resistance models of EN 1993-1-5 (using Mpl,R and patch loading resistance model of Chacón et al.).

interaction equation can be used with adequate safety using index of 3.6 relatively large. This evaluation proves the applicability of the original
for the bending utilisation in the interaction equation. interaction equation proposed by Braun and Kuhlmann [2].
To investigate the approximation of the new interaction equation of
the real physical behaviour of the combined loading situation, the data-
5.4. Statistical evaluation base using the FEM calculated patch loading and shear buckling resis-
tances and the plastic bending resistance is also evaluated. The
In this section the results of the statistical evaluation are calculation results showed, that there is not a large difference between
summarised for four different evaluation methods. The first evaluation the statistical quantity of the interaction equations with indices of 3.6
method uses the FEM based resistance models (bending, shear and and 3.0. The equation with index of 3.6 (Method 3) gives also very
patch loading resistances according to numerical calculations). The sec- good approximation of the real structural behaviour. Using the index
ond evaluation method uses the resistance models of EN 1993-1-5 of 3.0, almost all points are outside of the interaction diagram, therefore
(using Mpl,R) and the latest proposal for the patch loading resistance the mean value and the lower 5% fractile values are also slightly higher.
model of Chacón et al. The third and fourth calculations use the FEM cal-
culated shear and patch loading resistances together with the plastic
bending resistance of the analysed cross sections with the index of the
Table 3
bending ratio in the interaction equations 3.6 and 3.0, respectively. Statistical evaluation for girders without longitudinal stiffeners.
It can be seen from the statistical evaluation (Table 3), that the calcu-
lations according to the FEM based resistance models (Method 1) and Method 1 Method 2 Method 3 Method 4

according to the Eurocode based design methods (Method 2) result in Mean value 1.2959 1.4146 1.1959 1.2282
relatively large safety for the new M–V–F interaction equation using Standard deviation 0.2107 0.1734 0.1497 0.1605
Lower 5% fractile 0.9503 1.1303 0.9504 0.9651
index of 3.6. The mean values (1.296 and 1.414, respectively) are
B. Kövesdi et al. / Journal of Constructional Steel Research 103 (2014) 327–343 343

6. Conclusions The new interaction equation is developed for use with the shear
buckling, bending and patch loading resistance models according to
The following conclusions are based on the results of the current nu- EN1993-1-5 [1]. Using these resistance models all the calculation results
merical investigations. are on the safe side, therefore the applicability of the new M–V–F inter-
action equation is validated for longitudinally unstiffened girders. The
– The M–V–F interaction curve (Eq. (19)) gives a good lower bound new M–V–F interaction curve can be used with current EN1993-1-5
interaction surface to the numerical results, if the bending, shear [1] resistance models with adequate safety.
and patch loading resistances are calculated based on EN1995-1-5
[1] resistance models using the plastic bending resistance in the in-
References
teraction equation. In this case all the calculation results are outside
of the interaction surface for longitudinally unstiffened girders. It [1] EN 1993-1-5:2006: EUROCODE 3—Design of steel structures Part 1–5: Plated struc-
means that all the results are on the safe side and the accuracy of tural elements; 2006.
[2] Braun B. Stability of steel plates under combined loading. [PhD thesis No. 2010-3]
the new M–V–F interaction equation is adequate. Based on the eval- Institute for Structural Design. Universität Stuttgart; 2010.
uation of more than 700 numerical calculations, the applicability of [3] Graciano C, Ayestarán A. Steel plate girders under combined patch loading, bending
the interaction equation is proven in the current investigations. and shear. J Constr Steel Res 2013;80:202–12.
[4] COMBRI 2006: competitive composite bridges by improved steel plated structures.
!3:6     TGS8 “Steel products and applications for buildings, construction and industry”
M V−0:5  F 1:6 F Technical report No 7, Contract Nr: RFS-CR-03018; 2006.
þ þ ≤1:0 ð19Þ [5] Sinur F. Behaviour of longitudinally stiffened girders under combination of high
Mpl;R VR FR
bending and shear loading. [PhD thesis] University of Ljubljana, Faculty of Civil
and Geodetical Engineering; 2011.
– The calculations proved that the assumption of Braun to use the [6] Basler K. Strength of plate girders in shear. J Struct Div 1961;87(7).
[7] Bergfelt A. Studies and tests on slender plate girders without stiffeners—shear
plastic bending resistance in the interaction equation is adequate. strength and local web crippling. Proceedings of the IABSE Colloquium, London,
The results give a good fit to the interaction equation in the main UK; 1971 67–83.
part of the interaction region. Based on the statistical evaluation [8] Lagerqvist O. Patch loading—Resistance of steel girders subjected to concentrated
forces. [PhD dissertation No. 1994:159] Lulea University of Technology Department
the mean value and the standard deviation are also smaller if the of Civil and Mining Engineering, Division of Steel Structures; 1994.
plastic bending resistance is used in the interaction equation. [9] Roberts TM. Slender plate girders subjected to edge loading. Proc Inst Civ Eng
– The results of the numerical parametric study also showed that a 1981;71:805–19.
[10] Elgaaly M. Web design under compressive edge loads. Eng J 1983;20(4):153–71.
small number of points can be found inside of the interaction sur- [11] Ungermann D. Bemessungsverfahren für Vollwand- und Kastenträger unter
face, if the M–V–F interaction equation is used with the patch load- besonderer Berücksichtigung des Stegverhaltens. [PhD thesis] RWTH Aachen;
ing and shear buckling resistances determined by numerical 1990.
[12] Johansson B, Lagerqvist O. Resistance of plate edges to concentrated forces. J Constr
simulations and the plastic bending resistance is used in the interac-
Steel Res 1995;32(1):69–105.
tion equation. The points inside of the interaction surface are located [13] Vigh LG. Influence of curved flange-to-web connection on the transverse load resis-
in the M–F and M–V interaction planes, or close to it and just a very tance of extruded or hot-rolled I girders. Thin-Walled Struct 2012;60(C):127–36.
small number of points are found inside the 3D domain. [14] Graciano C, Casanova E. Ultimate strength of longitudinally stiffened I-girder webs
subjected to combined patch loading and bending. J Constr Steel Res 2005;
– The results show that the interaction equation with an index of 3.6 61:93–111.
gives a lower limit curve if the flange part is neglected from the [15] Ren WX, Fang SE, Young B. Analysis and design of cold-formed steel channels sub-
shear buckling and patch loading resistance models, what is the jected to combined bending and web crippling. Thin-Walled Struct 2006;44:314–20.
[16] Granath P, Thorsson A, Edlund E. I-shaped steel girders subjected to bending mo-
case in the Eurocode design method. The previously developed ment and travelling patch loading. J Constr Steel Res 2000;54:409–21.
Eurocode interaction equations for M–F and M–V interaction take [17] Elgaaly M. Failure of thin-walled members under patch loading and shear. Proceed-
also the fact into account that only the web part is considered in ings of the 3rd International Specialty Conference on Cold Formed Structures,
Missouri, 1. 1975. p. 357–81.
the shear buckling and patch loading resistance models. Therefore [18] Oxfort J, Weber N. Versuche zum Beul- und Krüppelverhalten von Träger-
the interaction equations are calibrated to this evaluation method. stegblechen bei zentrischer und exzentrischer Belastung. Technical report, 6/1979.
– If the flange part is also considered in the shear buckling and patch Institut für Stahlbau und Holzbau, Universität Stuttgart; 1979.
[19] Oxfort J, Gauger HU. Beultraglast von Vollwandträgern unter Einzellasten. Stahlbau
loading resistance models used in the interaction equation, the 1989;58(11):331–9.
index of 3.0 can give lower limit curve if the plastic bending resis- [20] Zoetemeijer P. The influence of normal-, bending- and shear stresses on the ultimate
tance is also used in the interaction equation (Eq. (20)). compression force exerted laterally to European rolled sections. Report 6-80-5.
Delft: Department of Civil Engineering, Delft University of Technology; 1980.
!3:0     [21] Roberts TM, Shahabian F. Ultimate resistance of slender web panels to combined
M V−0:5  F 1:6 F bending shear and patch loading. J Constr Steel Res 2001;57(7):779–90.
þ þ ≤1:0 ð20Þ [22] Alcaine J. Interaction behaviour of steel I-girders under bending, shear and trans-
Mpl;R VR FR
verse force. [MSc thesis] in cooperation of the Department of Structural Engineer-
ing, Budapest University of Technology and Economics and Construction
– If the flange part is also considered in the shear buckling and patch Engineering Department, Universitat Politècnica de Catalunya; 2013.
[23] Chacón R, Braun B, Kuhlman U, Mirambell E. Statistical evaluation of the new resis-
loading resistance models used in the interaction equation (usually
tance model for steel plate girders subjected to patch loading. Steel Constr
calculated by numerical methods), and the bending resistance is 2014;4(1):10–5.
determined also by numerical calculations (according to relevant [24] Chacón R, Mirambell E, Real E. Hybrid steel plate girders subjected to patch loading.
cross section class), the interaction equation according to Eq. (19) Part 1: numerical study. J Constr Steel Res 2010;66(5):695–708.
[25] Kövesdi B, Alcaine J, Dunai L, Mirambell E, Braun B, Kuhlmann U. Interaction behav-
can be also used. This statement is also proven by Graciano et al. in iour of steel I-girders. Part II: Longitudinally stiffened girders. J Constr Steel Res
2013 [3]. 2014;103:344–53.

Das könnte Ihnen auch gefallen