Journal of Constructional Steel Research 82 (2013) 7287

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Experimental tests and numerical modeling of cellular beams with

sinusoidal openings
S. Durif a, b, 1, A. Bouchar a, b,, O. Vassart c
a
Clermont Universit, Universit Blaise Pascal, Institut Pascal, BP 10448, F-63000 Clermont-Ferrand, France
b
CNRS, UMR 6602, Institut Pascal, F-63171 Aubire, France
c
ArcelorMittal Belval&Differdange S.A., Research Centre, Esch/Alzette, Luxembourg

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

Article history: New type of cellular beams with sinusoidal shape of openings shows a specic behavior in comparison with
Received 29 April 2012 the standard circular openings. Full scale tests were realized on three beams representing various dimensions
Accepted 27 December 2012 of the openings. The aim was to observe the failure modes of these beams and to obtain the ultimate values of
Available online 22 January 2013
strength. The specimens were heavily gauged to clearly identify the local failure modes of the opening zone.
Two specic failure modes were observed. For the opening with large height, the failure is reached by the for-
Keywords:
Cellular beam
mation of four plastic hinges at the corners of the critical section (Vierendeel bending). This mode is similar
Sinusoidal opening to that of the rectangular opening. Whereas with the small opening, failure arises by the local instability of
Experimental destructive tests the compressed out stand panel in the sinusoidal parts of the opening. A numerical model is developed
Numerical analyses and calibrated on the basis of the experimental results. The numerical model is used to analyze with more
Vierendeel bending details the behavior of the critical opening including the stress distribution in its different parts. The exper-
Local instability imental results provided useful qualitative and quantitative information to understand the behavior of the
cellular beams with sinusoidal openings. The numerical model showed a good accuracy in the prediction of
the experimental results. It can be used as a tool to generate complementary results to develop an analytical
model.
2013 Elsevier Ltd. All rights reserved.

1. Introduction cellular beams. These beams produced by cutting and re-welding of

hot rolled sections are made of regularly spaced circular openings.
Long span beams commonly used in steel construction started to Thus, for the same weight of steel, the cellular beam can reach 1.5
use large web openings during the last century in order to reduce times the height of the parent prole [6]. Nowadays cellular beams
the oors depth by passing all services through the web heights. The are widely used in steel construction and have become the most pop-
presence of the large openings changes the local transfer of the inter- ular long span system.
nal forces mainly the shear force. In 1978, Redwood [1] presented one This new type of beam showed a mechanical behavior similar to
of the rst experimental studies on beams with large isolated rectan- that of isolated rectangular openings mainly regarding the Vierendeel
gular web openings described in Fig. 1. These tests showed a new local bending [7]. However, unlike rectangular or hexagonal openings
failure mode due to Vierendeel bending [25]. In fact, the shear trans- where the critical sections are in the corners, it is not easy to dene
fer around the opening is equilibrated by the local bending of both the position of the critical section around circular openings [8]. Nu-
upper and lower members of the opening represented by tee section. merous investigations and especially the work of Ward [9] based on
This local bending creates local hinges at the four corners of the nite element study provided substantial results that were used to
opening. develop a reliable design method to check the resistance of cellular
Later, castellated beams composed of regularly spaced hexagonal beams with circular openings. The method proposes to check incre-
openings were proposed. As continuation, following the architectural mentally all the inclined sections around the opening starting from
and technical demands, constructors developed circular shapes of the straight section at the mid span of the opening. Then, the internal
openings instead of the hexagonal shapes to obtain the well known load of each inclined tee section is compared to its resistance as shown
in Fig. 2 where VEd and MEd are the global shear force and bending
moment.
Corresponding author at: Clermont Universit, Universit Blaise Pascal, Institut Pascal, Moreover, several works on castellated and cellular beams showed
BP 10448, F-63000 Clermont-Ferrand, France. Tel.: +33 473407521.
E-mail addresses: [email protected] (S. Durif),
a new local failure mode due to the buckling of the intermediate
[email protected] (A. Bouchar). web-post. This failure mode is observed for closely spaced openings
1
Tel.: +33 607903470. and slender proles [10] (Fig. 3). The reference [11] describes

0143-974X/$ see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2012.12.010
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 73

Fig. 4. Specimen of cellular composite beam with wide openings (RFCS project) [6].

experimental and numerical studies led by the CTICM to understand

the behavior of these beams and to propose analytical formulae
predicting the resistance of the intermediate web-posts of cellular
beams.
Many other researches have been conducted to develop design
Fig. 1. Failure modes of beams with isolated large rectangular opening [1]. methods and to propose expressions for standard codes on the basis
of experimental and numerical studies [1114]. Extensive experimen-
tal studies were led as part of a European research project to investi-
gate the evolution of composite action near the supports of a cellular
beam, the inuence of the local load due to secondary beams, the in-
uence of the asymmetry of cross-section and the inuence of the
elongated openings in cellular beams (Fig. 4) [6,14]. The buckling
of web-post in cellular beams exposed to re was also analyzed
(Fig. 5) [13] to propose an analytical model predicting the resistance.
The analytical model was validated on the basis of numerical models
and added to the software ACB+. An experimental study combined
with numerical approach concerned the cellular beams with closely
spaced openings and different opening shapes close to the circle
[15]. This study analyzed the web-post buckling and an empirical
formula predicting this failure mode is proposed for various opening
shapes.
A new type of castellated beam made from hot rolled proles
called Angelina with sinusoidal openings has been developed [16].
The Angelina beam is constructed from a hot rolled prole with
only one oxycutting line instead of two for the standard cellular
beam with circular openings. This new opening shape has some simi-
larities with the circular and the hexagonal openings. However, the
main common point between the circular and the sinusoidal openings
is the variation of the tee sections to be checked around the opening,
which makes the critical section position difcult to predict. Further-
Fig. 2. Local distribution of global forces MEd and VEd. more, due to the fabrication process, having large openings implies

Fig. 3. Failure modes of composite cellular beams: Vierendeel bending (left), web-post buckling (right).
74 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

Fig. 5. Web-post buckling failure mechanism at elevated temperature [13].

Fig. 6. Loading conguration of the tested beams.

large intermediate web-posts which avoid any web-post buckling. The tested specimens are well equipped by many measurement
However, preliminary numerical investigations showed that new fail- devices such as LVDT, camera targets and strain gauges to measure
ure modes, different from those of the common openings, can arise in displacements and deformations in various points around the critical
the sinusoidal part of the opening with a local buckling of the most openings (between openings numbers 1 and 2, in Fig. 6). The experi-
compressed opening panels. This is mainly due to the great length of mental results are used to validate the nite element model devel-
the sinusoidal part. oped using the software SAFIR. In this paper, the main experimental
Thus, the existing analytical models initially developed for circular and numerical results are presented and compared.
or hexagonal openings need to be updated to take account of the real
behavior and the associated failure modes of the sinusoidal openings. 2. Exeprimental program
In order to observe the real behavior of cellular beams with sinusoidal
openings experimental and numerical programs are conducted in co- 2.1. Detail of specimen
operation between Blaise Pascal University (ClermontFerrand) and
Arcelor Mittal. The tests concern three full scale Angelina beams load- The tests are conducted on three specimens named AS012, AS016
ed in four point bending (see Fig. 6). In reality, the various existing and AS018. In order to cover a wide range of Angelina beam cong-
studies showed that the behavior of beams with large web openings urations, it has been chosen to test three beams with three different
depends on the opening shapes. Thus, it seems not realistic to obtain opening sizes representing small, medium and large opening. The
universal analytical formulae to predict the failure modes and the ul- beams with small and large openings are made from IPE 450 prole
timate loads of cellular beams with various shapes of openings. Be- and the medium beam is made from IPE 400 prole. The expected fail-
sides, the majority of existing studies used an approach based on ure modes according to the preliminary modeling are either yielding
the equilibrium of internal forces around the openings. To obtain
data giving the possibility to check the validity of this approach and
Table 1
its adaptation to sinusoidal opening the experimental and the numer-
Geometric congurations of the tested beams.
ical studies performed in this study are presented.
Tested beam AS012 AS016 AS018

Nominal elastic limit (MPa) 460 235 460

Parent prole IPE 450 IPE 400
Prole height: hp (mm) 450 400
Flange width: bf (mm) 190 180
Flange thickness: tf (mm) 14.6 13.5
Web thickness: tw (mm) 9.4 8.6
Filet radius: r (mm) 21 21
Final beam height: Htot (mm) 675 735 507.5
Opening height: a0(mm) 450 570 215
Sinusoid length: ls (mm) 638 420 325
Web-post length: w 205 350 202
Beam openings number n 6
Opening length: L0 (mm) 1481 1190 852
Span: P (between supports) 10,500 9750 6700
Total length of the beam: L 10,778 10,028 6978
Fig. 7. Main geometric parameters of a sinusoidal opening.
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 75

Table 2
Geometrical characteristics of the openings (beam AS012).

Opening a0 (mm) Delta (%) L0 (mm) Delta (%)

1 449 0.22 1418 4.25

2 450 0 1411 4.73
3 451.5 0.33 1421 4.05
4 449 0.22 1423 3.92
5 450 0 1425 3.78
6 449.5 0.11 1409 4.86

or local buckling around the critical opening. As a result, steel with

different elastic limit values and openings with extreme values of
openings are chosen to obtain the various failure modes expected.
Fig. 7 and Table 1 describe the geometric parameters of the elementa-
ry opening for the three tested beams.
Table 1 presents the nominal values of the geometric characteris-
tics of the tested beams. The real characteristics measured for all the
specimens show some differences in comparison with the measured
values. This concerns the initial imperfections, the dimensions, the Fig. 9. Position of the points where the imperfection is measured.
thicknesses of the anges or the webs and the openings dimensions.
The initial geometric imperfections are measured on the beam length
and the beam height. The main results are presented hereafter. 2.2.3. Local geometric initial imperfection
To obtain the local geometric imperfection, the out-of-plane im-
perfection of the web is measured at various cross sections. Fig. 9
2.2. Geometric and material characteristics
shows the points a to f where the lateral imperfections are mea-
sured on the cross-section corresponding to the web-post between
2.2.1. Opening characteristics
two openings. The measurements were made on three web post
For all the beams, the geometric characteristics are measured on
along the beams length. Those measures permit to identify the real
various points along the length of the beam. The maximum relative
shape of the vertical full web section of the intermediate web posts.
deviation between the nominal and the measured values is less than
The measurement made on all intermediate web-posts show an
5%. Table 2 shows the example of the measured values of the geomet-
initial imperfection with a C shape of maximum magnitude equal to
ric characteristics of the openings (beam AS012). The deviation is
H/430 whereas the Eurocode proposes for numerical calculations a
observed considering the following ratio: Delta (%) = [(measured
value of H/200 [17]. As an illustration, the measurements for beam
nominal) / nominal] 100.
AS012 are given in Table 3.
The negative values of Delta show that the dimensions of the open-
Along the length of the beam, the cross sections show the same
ings are generally slightly smaller than the nominal values. Thus, the
shape of initial imperfection with different magnitudes. The maxi-
numerical model based on the nominal values is rather conservative.
mum value of the imperfection magnitude is considered for all the
The measurement of the geometrical characteristics on the parent
cross sections in the nite element model to obtain a general but re-
prole (web thickness, ange width) showed the same tendency
alistic approach of modeling.
but with positive values of Delta. However, in all observed cases, the
deviations are small enough to be neglected in the numerical model.
2.2.4. Material characteristics
To obtain the real material characteristics of the beams, tensile
2.2.2. Global geometric initial imperfection tests are realized on coupons taken from the anges and the web of
The initial imperfections are used as input parameter in the nu- the beams. Fig. 10 presents an example of the results obtained from
merical model considering the large displacement approach. In fact, a coupon tensile test (beam AS016) with the real curve and the equiv-
the shape and magnitude of the initial imperfection can inuence alent bi-linear curve used in the FEM model. The results for all beams
the failure mode of the cellular beams. Global imperfections are need- are summarized in Table 4, it gives the mean value of the measured
ed in order to represent global instabilities of the beam as lateral elastic limit (fy), the ultimate limit (fu) and the strain hardening mod-
torsional buckling. Thus, the measurements of the global initial imper- ulus (E') considered for the bi-linear law. This value of strain harden-
fections are performed at the mid width of the upper ange. The real ing modulus represents the straight line between the measured values
shape of the initial imperfection is shown in Fig. 8 where the blue con- of the elastic limit and the ultimate limit of the material. The measured
tinuous line is the reference line and the cross points represent the values of the elastic limits are higher than the nominal ones, especially
measured positions of the mid width of the upper ange. for beam AS016 which has a nominal elastic limit of 235 MPa.
Fig. 8 shows that the global out-of-plane imperfection can be
approximated as an arc of maximum amplitude of = 11 mm 2.3. Test set-up
(for beam AS012) which corresponds to a ratio of L/950. It has been
chosen to use those measured values even if the Eurocode 3 part 1.5 The test set-up consists of a four point bending test with simply
annex C proposes a value of the imperfection magnitude of L/400 supported beams as shown in Figs. 6 and 11. Each beam tested is
[17]. provided with the 6 openings. The applied loads are located on the

Fig. 8. Example of global out-of-plane imperfection (beam AS012).

76 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

Table 3 Table 4
Measured values of local out of plan imperfection (mm). Measured values of the material characteristics of the tested beams (MPa).

Web-post a b c d e f Beam Elastic limit fy Ultimate limit fu E'

1 0 0.20 0.12 1.06 0.60 0.18 AS012 499 624 938.7

3 0.07 0.74 0.77 0.97 0.78 0.04 AS016 320 454 619
5 0.11 0.70 1.49 2.02 0.99 0.05 AS018 468 590 847

intermediate web-post between the openings 23 and 45. Thus, two main steps. The rst step is an elastic range test realized on the
the critical openings combining the global shear load and the global beam at its initial stage. After the rst step, reinforcement is realized
bending moment are the numbers 2 and 5. using welded sinusoidal stiffeners on the openings (5 and 6) of the
As the study is focused on the specic failure of the cellular beams tested beam (Fig. 22 (c)). Due to the symmetry of the beam, this rein-
around openings, lateral supports are used to avoid the lateral buck- forcement guarantees that the failure will occur in the gauged openings
ling of the beams. These supports are located at the ends, the mid- (between 1 and 2). The second step corresponds to the test until failure
span and the loading points of the tested beam (see Fig. 11). characterized by the descending branch of the loadvertical displace-
The loading system is a hydraulic jack with a stiff header to obtain ment curve. To obtain the descending branch, the load is applied
two equal loads (Fig. 6). Each beam is equipped with different measur- through a displacement control of the jack.
ing instruments: dial gauges, strain gauges, strain rosettes and targets
combined to video measurement. The dial gauges are used to measure 2.5. Summary of the experimental results
the vertical and lateral displacements on different points. The strain
gauges and the rosettes give the possibility to analyze the evolution The comparison of the loaddisplacement curves between the
of the strains and stresses in some critical points around the openings rst and the second step shows that the use of local sinusoidal stiff-
dened according to the preliminary numerical studies (for example, eners after the rst step does not inuence the initial stiffness of the
see Fig. 12 for beam AS012 and Fig. 15 for beam AS016). The displace- beams. As this study is focused on the analysis of the resistance of
ments are measured with the combination of two systems, LVDT dial the beams, only the results of the second step are considered. The
gauges and video measurement. The video measurement uses two three congurations of tested beams are chosen with different dimen-
cameras to obtain the 3D displacements of the targets xed on some sions of the openings (height and length) to obtain various failure
points around the studied opening. The main advantage of this video modes. As the beam is provided with lateral supports to avoid the
system is its non limited number of targets and the measurement of global lateral buckling modes, the observed failure modes are the
3D displacements without contact. Fig. 13 shows one tested specimen local buckling or the plastic yielding due to Vierendeel bending in
with different targets around the web-post. Each target is a black circle the panels around the critical opening. In fact for cellular beams with
on white support. sinusoidal openings, the web-post is large enough to avoid the local
Due to the loading system and the beam conguration, the failure web-post buckling. Therefore, as expected for all tests the ultimate
is expected between the openings 1 and 2 where the maximum effect failure mode concerns the Vierendeel mechanism.
of the combination of global shear and bending moment is reached Beam AS016 has the highest opening size and as a consequence
(the openings 5 and 6 are stiffened for the test at ultimate load). the lowest dimensions of the tee sections on the four corners of the
This critical opening is heavily gauged with various measuring instru- openings. Thus, yielding is observed on these four corners where, at
ments and some redundancy to guarantee the reliability of the mea- failure, the deformed shape of the opening is very similar to that of
sured values. Thus, vertical displacement, horizontal displacements the equivalent rectangular opening (see Fig. 14).
and strains are measured in various points around the critical opening For this test (specimen AS016), only the plastic hinges are ob-
to understand the global behavior of the beams with sinusoidal open- served around the opening. No local buckling mode is observed or
ings regarding, the stiffness, the ultimate load and the failure modes. measured according to the out-of-plane displacement (horizontal
dial gauges) or the strain gauge values. Fig. 15 shows an example of
2.4. Test procedure two strain gauge results (1 left, 1 right) localized on the same point
of one opening corner on two sides of the web panel (strain gauge 1
Before each test, preliminary loadings are applied in order to check in Fig. 15). The two curves, representing the evolution of the strains
the response of the measuring instruments. Each test is conducted in versus time, are very similar conrming that there is no local out-of-

Fig. 10. Example of bilinear modeling of the stressstrain curve (beam AS016).
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 77

Fig. 11. Positions of the lateral supports.

gauges and by the strains measured on both sides of the critical section
in the sinusoidal part.
The examples of measuring the local buckling for the specimens
AS012 and AS018 are presented hereafter. For beam AS012, Fig. 17
shows the location where the vertical and horizontal displacements
are measured using dial gauges or video system. To illustrate the dif-
ference between the two measurement systems, Fig. 17b shows a
comparison for the vertical displacements. The comparisons show
Fig. 12. Example of strain gauges positions (beam AS012). that the vertical displacements are very close between the two mea-
surement techniques. The measurement of the horizontal displace-
ments of the intermediate web-post between the openings 1 and 2
(dial gauge H04 in Fig. 17c) show that there is no web-post buckling.
However it can be seen in Fig. 17c that there is a signicant out of
plane bending of the web as there is no difference of strain between plane displacement of the opening panel in the location of the dial
both sides. gauge H07 and the video target M12 (see Fig. 17). The combination
The evolutions of the stresses at the four corners of the opening, of both measurement techniques indicates the position of local buck-
given by the strain gauges numbered 1 to 4 (Fig. 15), are shown in ling zones around the opening.
Fig. 16. The stresses are obtained from the measured strains using Furthermore, Fig. 18 shows the evolution of the out-of-plane dis-
the Young modulus for the values lower than the measured elastic placement of the sinusoidal panels (dial gauge H07 and video target
limit (320 MPa). When the elastic limit is reached, a plateau is con- M12) and the applied load versus time. The position of the horizontal
sidered for stresses, neglecting the strain hardening. This representa- dial gauge H07 is given in Fig. 17. It can be clearly seen in Fig. 18 that
tion illustrates the yielding sequences of the four sections around the beam failure, represented by the decrease of the applied load, arises
opening. in combination with the local buckling of the sinusoidal opening
For the two other congurations (AS012, AS018), a different behav- panels represented by the evolution of the out-of-plane displacements
ior is observed with the presence of a local buckling in the sinusoidal (H07 and M12). Thus, the failure mode of the opening can be linked
part of two quarters of the opening. This local buckling is conrmed to the local buckling of the panels around the opening due to the
by the out-of-plane displacement measured by the horizontal dial Vierendeel mechanism.

Fig. 13. Test set-up with the targets positions (beam AS012).
78 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

from strain rosettes xed at the zones where the maximum out-of-
plane displacement is observed. In Fig. 19, the evolution of the applied
load versus the loading time is also represented. It shows that the load
reaches its maximum value when the ignition of local buckling is ob-
served. Thus, as for beam AS012, local buckling observed doesn't rep-
resent the global failure of the beam and it shows the additional
resistance provided by the other parts of the opening that didn't fail
yet. Fig. 20 presents the observed out-of-plane displacement. It can
be seen on this photo that the target used to measure the out-of-
plane displacement was not well adapted and has been peeled off
due to out-of-plane bending. This accidental situation arises some-
times with large displacements. In general, the measures done with
the video system show good correlation with dial gauge measure-
ments. However the users should be aware of possible problem of
out-of-plane measurements in the case of local buckling and should
prefer painted targets than bonded targets.
These experimental observations show that the failure of the cel-
Fig. 14. Formation of four plastic hinges at ultimate load (beam AS016). lular beams AS012 and AS018 cannot be considered as the conse-
quence of local buckling around the opening of the sinusoidal part
of the most compressed quarter as it has been noticed a reserve of
The nite element model developed hereafter conrmed that the strength in both cases. For both tests, the failure of the beam can be
ultimate failure is due to the combination of local buckling of two directly linked to the local failures around the opening and extensive
quarters followed by plastic hinges in one or two other quarters (see numerical study will help to clearly identify that the causes of the
Fig. 22b). global failure of the beam can be the yielding of other parts in the crit-
The local buckling is also observed considering the evolution of the ical opening.
measured strains on the sinusoidal part of the openings. The example Furthermore, according to the theoretical and numerical prelimi-
of beam AS018 is illustrated in Fig. 19. The curves show the evolution nary studies of the sinusoidal opening, it has been expected two fail-
of the principal compressive strains (%), versus the loading time, on ure modes: yielding of the opening corners similarly to a rectangular
both sides of the critical section. The compressive strains are obtained opening and local buckling in the sinusoidal part of the opening. Both

Fig. 15. Strain-time evolutions in a corner of opening (beam AS016) at ultimate load.

Fig. 16. Stress measurements on the web of the four corners of the second opening of beam AS016.
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 79

Fig. 19. Principal strain measured at both sides of the plate versus time (beam AS018).

failures are observed experimentally and the rst hypothesis is that

the failure modes may depend on the web height of the tee sections
at the mid span of the opening.
Indeed, the yielding of the opening corners has been observed on
beam AS016 with the highest opening size (a0/Htot = 0.78). For the
same parent prole, a high opening is associated with a small tee sec-
tion at the mid span of the opening. For beam AS016, the tee web
height is hw = 67.9 mm whereas for beam AS012 the tee web height
is hw = 97.9 mm. For beam AS018 with the parent prole IPE400, the
tee web height is hw = 119 mm. The failure mode of an opening
seems to be mainly linked to this tee height. For high openings the
tee height is small and the opening is expected to fail as a rectangular
opening. However, for smaller openings, the exural resistance of the
straight part of the opening is too important in comparison with the
stability of the sinusoidal part and the opening fails by elasticplastic
buckling of the out stand compressed elements of the opening. The pa-
rameter hw is the web height of the tee section at the mid span of the
opening. It can be approximated as (hw = Htot a0 2 tf).
The experimental results are used rst to understand the behavior
of the cellular beams with sinusoidal openings. Besides, these results
are used to validate a numerical model developed to predict in a pre-
Fig. 17. a. Position of the vertical (V1 and V2) and horizontal (H04, H07 and M12) dis- cise manner the behavior of these cellular beams. The modeling ap-
placement dial gauges and video measurements (beam AS012). b. Comparison of vertical
displacement measurements using dial gauge and video system (beam AS012). c. Com-
proach to complete the experimental one is presented hereafter.
parison of out-of-plane measurements using dial gauge and video system (beam AS012).

(a) Lateral displacement measurements

Time (s)
0 600
0 2000 4000 6000 8000 10000 12000 14000
-1
Horizontal displacement (mm)

500
-2

400
-3
Load (kN)

-4 300
H07
-5 M12 (video)
Load 200
-6 (b) Lateral displacement of
100 the compressed panel near
-7 the dial gauge H07
-8 0

Fig. 18. Measurement of the out-of-plane displacement around the sinusoidal opening (beam AS012).
80 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

Fig. 20. Local buckling observed at the critical section of beam AS018.
Fig. 23. Loaddeection of the whole beam AS012.

3. Numerical model
In the experimental part, the measured values of geometric charac-
A nite elements model is developed to analyze the behavior of teristics did not show a signicant difference with the nominal ones.
the tested specimens using SAFIR software with shell elements taking Thus, the nominal characteristics are considered in the numerical
into account geometrical and material non-linearities [18]. This ele- model except for the initial imperfections where the measured values
ment is particularly efcient to represent local instability with limited are used. Bi-linear law is used for the material with an elastic limit
resources. Previous numerical studies on cellular beams showed that taken from the measured ones and different values of strain hardening
the shell elements represent well the behavior of cellular beams in- modulus are used to evaluate the inuence of this parameter (see
cluding the local instabilities as the web-post buckling for closely Table 4). A non-linear analysis, with elasticplastic behavior and
spaced circular openings [10,1922]. Previous analyses made with large displacements, is performed with displacement control to obtain
the software SAFIR to study the behavior of cellular beams under the loaddisplacement curve of the specimens including the descend-
re conditions [13,23] gave good results and proved the capacity of ing branch to represent the post-critical range. The boundary condi-
this numerical model to represent the behavior of cellular beams. tions represent the real conditions of the experiments. Thus, vertical

Fig. 21. Model of a beam with symmetry conditions (left). Meshing of an opening (right).

1 2 3 4 5 6

(a)

(c)
(b)
Fig. 22. Finite element model: (a) whole beam, (b) iso-values of lateral displacement (opening 2, maximum load), (c) sinusoidal stiffeners.
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 81

Fig. 24. Deformed shape of the critical opening at maximum load (FEM and test).

supports are considered on the lower ange of the beam and lateral For both quarters, with maximum out-of-plane displacement; the -
supports are considered at the same locations as described in Fig. 11. nite element model shows an elasticplastic failure with a local out-of-
The mesh density around the critical opening has been chosen in plane bending which characterizes the local buckling. The comparison
order to dene with good accuracy the behavior of the opening re- between the loaddisplacement curves from the numerical model and
garding the large displacement and the stress distribution around the tests shows the good accuracy of the numerical model (Fig. 23).
the opening (see an example in Fig. 21). The vertical displacement is measured at the mid-span of the whole
The model is validated on the basis of the experimental results beam.
considering the ultimate loads, the failure modes and the global To evaluate the effect of the strain hardening, the value of E' =
loaddisplacement curves of the tested beams. Once the model is val- 938.7 MPa from tests (Table 4) is compared with E' = 0 MPa (elastic
idated, it will be an interesting tool to study in details the behavior of perfectly plastic law). The inuence of these parameters on the
the opening of each tested beam. A summary of the results given by loaddisplacement curve is not very high (Fig. 23). The experimental
the numerical study is proposed hereafter. curve is situated between the two curves corresponding to the two
values considered for E'.
4. Results of the numerical model and comparison with tests The deformed shape of the critical opening under the maximum
load is well described by the numerical model (Fig. 24). This shape
4.1. Beam AS012 represents the local buckling of the compressed panels around the
opening.
Beam AS012 corresponds to a medium size of opening (a0/Htot =
0.67). The numerical results show that failure is mainly induced by 4.2. Beam AS016
local buckling. This local instability appears on both upper and lower
quarters of the critical opening as described by the iso-values of Beam AS016 is characterized by a high opening size (a0/Hbeam =
out-of-plane displacements (see Fig. 22(b)). The iso-values are repre- .78). The experiments showed a failure mode similar to that of a rect-
sented in color so that maximum horizontal displacements appear angular opening with four plastic hinges localized at the corners of the
in dark colors in Fig. 22(b). As indicated before, the other critical open- opening. The curves representing the evolution of the applied load
ings 5 and 6 are reinforced using local stiffeners to exclude their local
failure (see Fig. 22(c)).

Fig. 25. Loadvertical displacement curves of beam AS016. Fig. 26. Loadvertical displacement curves for beam AS018.
82 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

Table 5
Summary of the results (model and tests).

Tested beam Fult exp. Fult FEM FEM/test

(kN) (kN)

AS012 494.3 529 1.07

AS016 274 242.6 0.89
AS018 621.8 614.9 0.99

versus the mid-span vertical displacement given by the numerical

model and the tests are shown in Fig. 25. The FEM results are obtained
considering for steel different bi-linear stressstrain laws with differ-
ent values of the strain hardening modulus E'. Indeed, as the failure Fig. 28. Critical tee sections around the opening (beam AS016).

mode observed was the signicant yielding of some parts around the
opening, it can be expected a non-negligible inuence of this strain
hardening modulus on the nite element results. Thus, three values not reach its total yielding before the development of the local insta-
of E' are considered: 0, E/100 and the measured value (619 MPa). bility (lateral displacement).
The results are compared regarding the loaddisplacement curves
(Fig. 25). The case with E' = E/1000 showed no difference in compar- 4.4. Summary of the results
ison with E' = 0.
The loaddisplacement curve from the model exhibits a plastic Table 5 summarizes the ultimate loads obtained from tests and -
plateau that characterizes the complete yielding of the opening and nite element model with the elastic perfectly plastic law of material.
the formation of the four plastic hinges as it has been observed exper- These values are obtained considering the maximum values of load
imentally (Figs. 14 and 16). reached during test or by the nite element model. They correspond
A plastic plateau is observed on the loaddisplacement curve to the plateau of the loaddisplacement curve or the higher value be-
(Fig. 25) with the elastic perfectly plastic law of material (E' = 0). fore the descending branch.
However, with strain hardening, the curves show a continuously The nite element model gives acceptable results for the ultimate
nonlinear parts and the nite element model represents with better load with values slightly lower than those given by experiments
accuracy the experimental curves. In this case, by opposition to the except for the specimen AS012 (Table 5). The most important differ-
other tested beams, the elastic perfectly plastic law of material gives ence that has been noticed is for the test AS016. From the observation
signicant differences in comparison with the experimental curves. of Fig. 25, this difference can be explained mainly by the difference
In fact, the plastic deformation is the failure mode of this beam and between the real material law and the bilinear used. Indeed it has
its loaddisplacement curve is more sensitive to the shape of the been shown that preferring a strain hardening modulus of E/100 rath-
stressstrain curve. er than E' = 619 (Table 4) give a much more reliable result. However
this difference remains acceptable as the nite element model has
4.3. Beam As018 predicted the total yielding of the opening as it has been measured
experimentally.
Beam AS018 is characterized by a small size of the opening (a0/ The nite element model is used to analyze some local aspects of
Htot = 0.42) and as a consequence the panels around the opening are the tested beams to understand the behavior of their critical sections
characterized by high dimensions. The failure mode observed in (failure mode, stress distribution ).
tests is similar to that of beam AS012 with a local failure in the sinusoi-
dal parts of the opening. The numerical model of beam AS018 repre- 4.5. Local numerical analysis of the tested beams
sents well the real behavior observed during the test including the
evolution of the applied load versus the mid-span vertical deection The stress analysis shows that the quarters of the opening are
(Fig. 26). To take account of the strain hardening effect, two values loaded in tension or compression regarding the global bending mo-
of E' are considered: E' = 0 and the measured value (847 MPa). The ment of the beam. Besides, a local bending of the tee sections due to
elastic perfectly plastic law is slightly conservative and the law with Vierendeel action is added to the tension or compression of the differ-
strain hardening describes well the behavior observed in tests. How- ent opening parts. Thus, each section is loaded in combined bending
ever, the difference between the two laws is lower than with the and axial loads with tension or compression on the web part of the
case of beam AS016. In fact, the critical opening of beam AS018 does tee depending on the position around the opening (Fig. 27).

Fig. 27. Distribution of equivalent internal forces in the upper and lower member of an
opening due to MEd and VEd. Fig. 29. Positions of the stress analysis on the web height of the tee section.
S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 83

Fig. 30. Evolution of the stresses on the height of the tee section (a).

Fig. 31. Evolution of the stresses on the height of the section (b).

Fig. 32. Evolution of the stresses on the height of the section (c).

Fig. 33. Evolution of the stresses on the height of the section (d).
84 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

The local behavior for each beam around the critical sections is
observed. This concerns the evolution of the stresses during the load-
ing. For example, for beam AS016, the failure observed during tests
is the development of four plastic hinges at the four corners of the
opening (Fig. 14). In the numerical model, the stress distribution is
observed at four locations of the critical sections in the opening quar-
ters where the yielding is observed (Fig. 28). The stress evolution of
each critical tee sections (a), (b), (c) and (d) around the opening is ob-
served on various points of the web height of the tee (Fig. 29). For all
this study, elastic-perfectly plastic laws are considered for the mate-
rials, to well illustrate the failure modes and to compare easily the ul-
timate values of the critical sections. Besides, this type of law is more
convenient to be used to justify the analytical models to be developed.
Regarding the loaddisplacement curves of beam AS016 (Fig. 25) Fig. 36. Loadvertical displacement diagram for beam AS012.
representing the evolution of the applied load versus the mid-span
vertical displacement, it can be observed that the plastic plateau
from the numerical model starts at a vertical displacement close to of the out-of-plane displacements in the critical opening where the
41 mm. Therefore, this instant can be taken as a reference to dene maximum values dene the critical sections (a) and (d). The locations
the beginning of the yielding plateau for which there is not any in- of maximum out-of-plane displacements correspond to the areas in
crease of load carrying capacity of the beam. dark color and allow dening both critical sections (a) and (d). The
To observe the evolutions of the yielding zones during the test, the evolution of the maximum out-of-plane displacement of the node, sit-
evolution of the equivalent stress (von Mises stress), at the points 1 to uated at the border line of each section (a) and (d), is plotted versus
5 on the height of the tee sections, is plotted versus the applied vertical the vertical displacement. The exponential evolution of these curves
displacement (Figs. 3033). The curves show that all the nodes reach conrms the developments of local instability in the two critical sec-
their elastic limit before reaching the plateau of the global load tions. Furthermore, the acceleration of the out-of-plane displacement
displacement curve corresponding to 41 mm (Fig. 25). This conrms in the two sections corresponds approximately to the maximum load
that the plastic hinges develop successively in the four corners around of the beam (vertical displacement of 91 mm, see Fig. 36). This con-
the opening. Thus, the global failure of the beam is not reached before rms that the global failure of the beam is linked to the local buckling
the development of a mechanism corresponding to the formation of of both sections (a) and (d) (Fig. 37).
plastic hinges in the four critical sections. However, the study of the other quarters of the opening can show
The behaviors and the failure modes of beams AS012 and AS018 an interesting behavior and reveals the formation of plastic hinges for
observed experimentally and from the numerical model are rather sections that are subjected to tension which can explain the experi-
similar and different from the AS016 beam. The deected shape of mental observations that showed the reserve of load carrying capacity
beam AS012 representing its failure mode shows a local instability in of the beams after the ignition of local buckling in the compressed
the panels around the opening (Fig. 34). Fig. 35 shows the iso-values parts of the critical opening.

Fig. 34. Deected shape of beam AS012 failure and zoom on the opening 2.

Fig. 35. Iso values of the out-of-plane displacement (left), positions of the critical sections positions (right).
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 85

The evolutions of the equivalent von Mises stresses on each criti-

cal section are used to observe in details the local failure of each quar-
ter of the opening. These evolutions on the height of each tee are
shown in Figs. 3841.
It can be observed that the two quarters corresponding to the sec-
tions (a) and (d) are subjected to compression at the out-stand panels
of the tee (web part) which induces a local buckling. However, the
out-stand panels of the tee in the two other quarters (b) and (c) are
mainly subjected to tension. Therefore, no local buckling can appear
in sections (b) and (c). The evolution of the equivalent von Mises
stresses in each critical section of both quarters (a) and (d) shows a
decrease before reaching the elastic limit of the material. This conrms
the redistribution of stresses due to the out-of-plane bending of the
compressed part of the tee section which reveals the existence of
local buckling. In parallel with this instability, the sections of the quar-
ters (b) and (c) form plastic hinges. However, the plastic hinge of the
section (b) appears after reaching the beam mid-span vertical dis-
Fig. 37. Out-of-plane displacement of sections (a) and (d) versus the global vertical placement of 91 mm which correspond to the maximum applied
displacement. load. Therefore, hinge in the section (b) seems to be post-critical. Con-
sequently, the ultimate failure of the critical opening is reached by
The vertical displacement corresponding to the maximum load of local failures combining buckling and plastic hinges in the four quar-
beam AS012, from the numerical model, is equal to 91 mm (Fig. 36). ters depending on their respective positions. This numerical observa-
As for beam AS016, this value is used to analyze the stress distribution tion showed that the global failure of the beam did not arise with the
in the critical sections around the critical opening. The positions of local failure of one or two quarters. The same observations can be
these sections are shown in Fig. 35 (right). made on the model of beam AS018 with similar failure mode.

Fig. 38. Evolution of the stresses in the height of the section (a).

Fig. 39. Evolution of the stresses in the height of the section (d).
86 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287

Fig. 40. Evolution of the stresses in the height of the section (c).

The numerical study showed that for the beams with large open- precisely identied experimentally. The experimental data have been
ings (beam AS016), the failure appears by the formation of four plastic used also to validate the nite element model developed on the soft-
hinges at the critical section of the quarters. Thus, the behavior is sim- ware SAFIR. The good correlation between the experimental and the
ilar to that of rectangular opening. Furthermore, in the case of small numerical results validated the model as predicting tool. Therefore,
openings (beams AS012 and AS018), the global failure of the beam is this numerical tool has been used to closely study the stress distribu-
linked to the local buckling of two quarters combined with two plastic tion around the opening. It conrmed that for a high opening size a0/
hinges in the post-critical stage. This failure can be assimilated to a Htot = 0.78 it can be observed a total yielding of the four opening cor-
mechanism with three or four critical sections. These numerical obser- ners. Then, for the two other beams, the failure observed experimen-
vations will open the way to develop realistic analytical models to pre- tally was considered as the local buckling of two parts of the opening.
dict the resistance of such beams. Work in progress is focused on the Furthermore the numerical studies could prove that the global fail-
denition of the resistance of these critical sections considering vari- ure is due to the combination of local failures around the opening until
ous geometrical and material congurations of beams. The goal is to the formation of plastic hinges in the quarters of opening subjected to
dene an analytical model using the common simplications for cellu- tension. These interesting observations could prove that the critical
lar beams such as that shown in Fig. 27. section position cannot be easily located and need to be determined
by incremental calculation as for circular openings. However, the
5. Conclusion check of the minimum resistance of the opening quarters should be
too conservative to predict the ultimate load because the global failure
The actual available approaches predicting the behavior of cellular appears to be caused by the combination of local failures. Therefore
beams are not sufcient in order to precisely cover the innovative the analytical model should consider precisely the resistance of each
opening with sinusoidal shape. Thus experimental and numerical in- opening quarter in the critical opening and their combination to
vestigations have been conducted. The experimental tests focused on obtain the failure mechanism of the opening. Thus, the validated nu-
the opening failure pointed out two failure modes, the opening yield- merical model can be used for further extensive parametric study
ing similar to that of the rectangular shape or the local buckling of the with various geometrical congurations of openings to study more ex-
sinusoidal part of the opening. Both of these failure modes have been tensively the beams and the local failure mode so as to develop and

Fig. 41. Evolution of the stresses in the height of the section (b).
 S. Durif et al. / Journal of Constructional Steel Research 82 (2013) 7287 87

validate a new analytical model more adapted to this new large open- [12] Mangin P, Bureau A. ArcelorMittal Cellular Beams Logiciel ACB+: Descriptif tech-
nique, Rapport technique n7.072-01. Centre Technique Industriel de la Construc-
ing shape. tion Mtallique CTICM; 2009.
[13] Nadjai A, Vassart O. Performance of cellular composite oor beams at elevated
References temperatures. Fire Saf J 2007;42:48997.
[14] Lawson RM, Lim J, Hicks SJ, Simms WI. Design of composite asymmetric cellular
[1] Redwood RG. Analyse et dimensionnement des poutres ayant des ouvertures beams and beams with large web openings. J Constr Steel Res 2005;6(62):61429.
dans les mes. Rev Constr Mt (CTICM) 1978;3:1527. [15] Tsavdaridis KD, D'Mello C. Web buckling study of the behaviour and strength of
[2] Lagaros ND, Lemonis DP, Papadrakakis M, Panagitou G. Optimum design of steel perforated steel beams with different novel web openings shapes. J Constr Steel
structures with web openings. Eng Struct 2008;30:252837. Res 2011;67:160520.
[3] Lawson RM. Design for openings in the webs of composite beams. Steel Construc- [16] Durif S, Vassart O, Bouchar A, Muzeau JP. Modle mcanique pour les poutres
tion institute Publication; 1987. p. 068. larges ouvertures d'mes de section variable. 29mes Rencontres Universitaires
[4] Darwin D. Design of steel and composite beams with web openings, design guide de Gnie Civil, AUGC, Tlemcen, vol. 1; 29-31 Mai 2011. p. 34555.
2, AISC. 3me edition. 65 pages; 2003. [17] EN 1993-1-5, Eurocode 3. Design of steel structurespart15: plated structural
[5] Chung KF, Lawson RM. Simplied design of composite beams with large web elements, annex C; 2004.
openings to Eurocode 4. J Constr Steel Res 2001;2(57):13564. [18] Franssen JM. A thermal/structural program for modeling structures under re.
[6] Veljkovic M, et al. Large web openings for service integration in composite oors. Eng J AISC 2005;42(3):14358.
120 pagesCECA projects; 2006. [19] Soltani MR, Bouchar A, Mimoune M. Nonlinear FE analysis of the ultimate behav-
[7] Bitar D. Vrication l'ELU des poutres mtalliques avec ouvertures d'me, ior of steel castellated beams. J Constr Steel Res March 2012;70:10114.
exemple de calcul et recommandations. Rev Constr Mt (CTICM) 1998;1:8098. [20] Chung KF, Liu TCH, Ko ACH. Investigation on Vierendeel mechanism in steel
[8] ENV 1993-1-1, Eurocode 3. Design of steel structurespart 1-1: general rules and beams with circular web openings. J Constr Steel Res 2001;5(57):46790.
rules for buildings, annex N; 1995. [21] Liu TCH, Chung KF. Steel beams with large web openings of various shapes and
[9] Ward JK. Design of composite and non-composite cellular beams. SCI Publication; sizes: nite element investigation. J Constr Steel Res 2003;59:115976.
1990. p. 100. [22] Hagen NC. Shear capacity of steel plate girders with large web openings, part I:
[10] Ellobody Ehab. Nonlinear analysis of cellular steel beams under combined buck- modeling and simulations. J Constr Steel Res 2009;65:14250.
ling modes. Thin-Walled Struct 2012;52:6679. [23] Vassart O. Analytical model for cellular beams made of hot rolled sections in case
[11] Bitar D, Martin P-O, Gala Y, Demarco T. Poutres cellulaires acier et mixtes: Partie of re. PhD thesis. Clermont-Ferrand: University Blaise Pascal; 2009.
1, proposition d'un modle pour la rsistance des montants. Rev Constr Mt
(CTICM) 2006;1:1539.