PP Periodica Polytechnica Numerical Investigation into the

Civil Engineering
Ultimate Strength of Steel Cellular
Beams with Semi-Rigid Connections
OnlineFirst (2018) paper 11304
https://doi.org/10.3311/PPci.11304
Creative Commons Attribution b Abdelhalim Madjour1*, Mohamed Rédha Soltani1, El Haddi Harkati1

research article Received 24 July 2017; Revised 24 October 2017; Accepted 27 November 2017

Abstract 1 Introduction
Over the past several years, extensive theoretical, numerical and Throughout the last several decades, cellular beams have
experimental research has been carried out on the structural been used extensively in steel construction to pass under-floor
behavior of steel beams with web openings of various shapes. services ducts (water and sewage pipes, air ducts, cables etc.)
However, in all these studies, cellular beams have always been and for their aesthetic and design advantages.
treated or designed as simply supported at their ends. In other The process of manufacturing cellular beams consists of
words, the beam to column connections are assumed as nomi- cutting the web of the parent profile by flame in a zigzag pattern
nally pinned. The main aim of the present paper is to perform along its centerline. The two obtained T-sections are shifted
numerical simulations to determine whether the use of end plate and rewelded producing beams with hexagonal, circular or
connections, instead of simple connections, could enhance the sinusoidal openings and lead to an increase in height compare
global bending and/or shear capacity of cellular beams with to the parent profile. Cellular beams with sinusoidal openings
circular and sinusoidal web openings. called Angelina are fabricated only by using one oxycutting
Interaction between a circular or sinusoidal opening in the line instead of two for the standard perforated beams with reg-
beam web and endplate connection has also been studied. ular circular or hexagonal openings. By increasing the depth
Data obtained from a number of experimental research pro- of the beam, the strong axis bending strength and stiffness are
grams on steel solid beam to column connections and simply improved without increasing the weight of the beam allow-
supported cellular beams, available in the literature, were uti- ing this type of beams to be used for medium to long spans
lized to validate the numerical models developed in this paper. constructions. The digital and parallel fabrication processes
The numerical simulations using ABAQUS software were per- reduces considerably the cost of cellular beams.
formed on full scale cellular steel beams connected to the col- In the current structural European design codes [1–3], steel
umns with seven types of endplate connections. The results were or composite cellular beams have always been designed as
compared with those obtained for similar simply cellular beams. simply-supported at their ends. Simple connections, which are
classified as nominally pinned that are assumed to transmit end
Keywords shear only and to have negligible resistance to rotation, are used.
endplate connection, cellular beams, pinned connections, cir- Fin plates, double angle web, and partial end plates beam to col-
cular openings, sinusoidal openings, ultimate load umn connections are commonly classified as nominally pinned.
Typical examples of these connections types are shown in Fig. 1.

a) b)

c)

1 Department of Civil Engineering Laboratory of Mines ,

Faculty of Science and Technology,
University of Larbi Tebessi , Tebessa (12002), Algeria a) Partial depth end plate b) Double angle cleats c) Fin plate
* Corresponding author, e­mail: [email protected] Fig. 1 Commonly used simple beam-to-column connections

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 1

 The main aim of the present paper is to determine through The initial rotational stiffness of the end-plate bolted joints
numerical simulations whether the use of rigid or semi-rigid is given by a mechanical model composed of several exten-
joints, instead of pinned joints, could enhance the global bend- sional springs and rigid links, each one representing a compo-
ing and/or shear capacity of cellular beams with circular and nent of the joint.
sinusoidal openings. The effect of the web openings on the The moment resistance of bolted end plate joint is governed
connection moment-relative rotation curves is also studied. by the resistance of one of the following:
Analysis simulations were performed using the commercial - the column web panel in shear; the column web in com-
finite element analysis software ABAQUS, version 6.14. Data pression; the column web in tension; the column flange in
obtained from a number of experimental research programs on bending; the beam end plate in bending; the beam flange and
steel beam to column connections and simply supported cellu- web in compression; the beam web in tension; bolts in tension;
lar beams, available in the literature, were utilized to validate bolts in shear; welds.
the numerical models developed in this research. Joints are classified by stiffness and by strength according
The behaviour of a joint is represented by moment-rota- to the global analysis used. When elastic analysis is used, the
tion curve relating the moment transmitted by the joint to the joints are classified according to their stiffness. When rigid
relative rotation between the members it connects [4]. The plastic analysis is used, the joints are classified according to
numerical or experimental nonlinear moment-rotation curves their strength (moment capacity). When elastic-plastic analy-
are generally in practice approximate with bilinear response sis is used, both classifications are used.
as shown in Fig. 2. They are characterized by the following A joint in a frame able to preserve the original angle between
features [11,12]: the knee-range of the M-ϕ curve, the plas- the connected members is classified as rigid. The connection is
tic flexural resistance, Mj.Rd, the maximum bending moment, classified as simple if it rotates without increasing moment. The
Mmax, the initial stiffness, Sj.ini, the post-limit stiffness, Sj.p-l, the connection stiffness between these two boundaries is catego-
rotation corresponding to the maximum load level, Mmax and rised as a partially restrained or semi-rigid connection. In this
the rotation capacity, ϕCd. The main features of this curve, as respect, clause 5.2.2 of EC3-1-8 [5] classifies as rigid if Sj.ini ≥ kb
specified in Eurocode 3, part 1.8 [5], are: the initial stiffness, EI b /Lb where EI b is the flexural rigidity of the beam connected,
Sj.ini, the moment resistance of the joint, Mj.Rd, and the capacity Lb its length and kb = 8 for frames where the bracing system
of rotation (ductility). reduces the horizontal displacement by at least 80%; kb = 25 for
other frames. A joint may be classified as nominally pinned if
Sj.ini ≤ 0.5 EI b /Lb.
A joint which cannot be classified as either nominally
pinned or rigid is classified as rigid.
Clause 5.2.3 of EC3-1-8 [5] classifies frame joints as full-
strength, nominally pinned or partial strength by comparing
their design moment resistance with the design moment resis-
tances of the members that it connects. A joint may be classi-
fied as nominally pinned if its design moment resistance is not
greater than 0.25 times that required for a full strength joint,
provided it has sufficient rotation capacity. A joint which can-
not be classified as either nominally pinned or full strength is
Fig. 2 Moment-rotation curves [12]
classified as partial strength. The purpose of this classification
The current component method introduced in EN 1993-1-8 is to determine whether the joint or one of the connected mem-
[5] can be used for the characterization of beam-to-column bers will limit the resistance of the frame.
joints subjected to bending moment and axial forces. The Concerning the seismic behaviour of joints, numerous stud-
main principle of the component method consists of dividing ies [6–10] have shown that weakening plain beams by mak-
the end plate connection into basic components, including the ing a single or a number of openings through their webs in the
panel zone, bolt, end plate and column flange. Its application vicinity of joints move the plastic hinge in the beam away from
requires the evaluation of the force-deformation response of the joint improving the connection ductility (rotation capacity).
each individual component which depends on mechanical and
geometrical properties of the joint components. The properties 2 Experimental tests used for the validation of the
of each component allow the specification of the corresponding numerical models
moment-rotation curve and the associated properties (stiffness Numerical models are required to be carefully validated
and moment resistance). against experimental data in order to determine the accuracy
of the modeling. As mentioned above, the aim of this study is

2 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

to investigate whether steel perforated beams connected to a of IPE 240 and HEA 120 hot rolled profiles, respectively. The
column with extended end plate will have higher resistance endplate had a thickness of 15 mm. The steel of all compo-
than their simply supported counterparts. Since all experi- nents is of class S235 steel grade. The beam is connected to
mental test results reported in the literature concern simply the column with three rows of bolts. The connection had two
supported cellular beams, selected experimental data found in high tension bolts M16 per horizontal row with grade 8.8 in 18
the literature on plain-webbed beam to column endplate bolted mm drilled holes. The material properties of the steel and bolts
connections and simply supported cellular beams with circu- were obtained from tensile coupon tests taken from the main
lar and sinusoidal openings were used to validate the numeri- components of the test specimen. A vertical load was applied
cal model used to simulate the behavior of partially restrained to the upper end of the column by one hydraulic actuator and
cellular beams considered in this study. was monotonically increased under displacement control at
The finite element model proposed for the simulation of the rate of 0.02 mm/s. Different devices were used to mea-
bolted end plate connections was validated by comparing the sure displacements and rotations in the connection zone. The
numerical results, in terms of moment-rotation curves, with test setup and the location of the measuring devices used by
those of tests, performed on four bolted connection test spec- Abidelah and al [15] are illustrated in Fig. 4.
imens.
The test specimens J1.1, J3.1 and J4.1 consisting of external
double extended bolted endplate single-sided beam to column
joints were tested by Nogueiro [13] at Coimbra University. The
specific details of the testing program were also described in
reference [14]. The configurations of the joints differ mainly
in the column and beam cross sections. For all cases, 18 mm
thick endplates were chosen, connected to the beam-ends by
full strength 45° continuous fillet welds. The column is 3 m
long, hinged at the bottom to a concrete block and had a free
vertical displacement at the top. A typical joint’s detail is Fig. 4 Test setup and location of the measuring devices [15]
shown in Fig. 3.
The measured mechanical properties of the joint compo- For perforated beams, the outputs of the numerical simulations
nents were obtained from uniaxial tensile tests performed were validated using experimental measurements published by
on coupons, extracted from the beams, the columns, the end Warren [16] and Durif [17,18]. The test specimens denoted 1A,
plates, and from some bolts. The loading was applied mono- 2A, and 2B were part of eight cellular beams tested by Warren
tonically in a displacement-controlled procedure. Equations [16] at Natal University. These specimens were fabricated from
were used to estimate the experimental rotations from the ver- hot rolled I-sections UB 203 × 133 × 25. Specimens 1A and 2A
tical displacements of the beam and the horizontal displace- were subjected to a single point load, whereas specimen 2B was
ments of the end plates which were measured by means of dis- loaded with two equal and symmetrical point loads. The spans
placement transducers fixed at appropriate locations. and the ratios of the opening spacing to the opening diameter of
the tested beams, 1.5 for beams 1A and 1.33 for beams 2A and
2B, were chosen to avoid failure by premature web-post buck-
ling or shearing. Lateral supports were provided to the compres-
sion flange to prevent lateral-torsional buckling. A typical test
setup used by Warren [16] is shown in Fig. 5

Fig. 3 Detail of the joint J1.1 [14]

The fourth specimen BC1 is an interior sub-assemblage

consisting of two beams connected to opposite flange of a sin-
gle column with two extended end plates. This specimen is one
of the four specimens tested by Abidelah and Bouchair [15] at
Blaise Pascal University. The beam and the column were made Fig. 5 Example of test setup used by Warren [16]

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 3

 Three full scale cellular beams with six regular spaced
sinusoidal web openings (Angelina beams), denoted ASO12,
ASO16 and ASO18, were tested by Durif and Bouchair [16,17]
at Blaise Pascal University. Beams ASO12 and ASO16 were
made from a hot rolled IPE 450 in S460 and S235 steel grades
respectively, while beam SO18 was made from an IPE 400 in
S460 steel grade. In order to obtain different modes of failure,
various sizes of web openings were adopted. Initial geometric b) end plate detail
a) Global
imperfections were measured for the three test beams. A four
point bending test set-up was used as illustrated in Fig. 6.

c) bolt detail
Fig. 6 Loading and lateral support conditions [17] Fig. 7 Typical finite element mesh of end plate steel connection

3 Development and validation of finite element

models for joint connections and simply supported
castellated beams
The purpose of this section is to develop and validate two
numerical models that have the ability to simulate accurately
the moment-rotation response of bolted end plate connections Fig. 8 Typical Finite element mesh of cellular beam with circular openings

and the flexural response of simply supported cellular beams.

The numerical simulations were performed for the specimens,
tested in the experimental programs described above, by using
the general purpose finite element software Abaqus, ver. 6.14
[19]. Material nonlinearities, large deflection, large strain and
contact between different joint surfaces were included in the
models.
Fig. 9 Typical Finite element mesh of cellular beam with sinusoidal openings
All the models were meshed with eight-node linear brick
elements with reduced integration (1 Gauss point) and an hour- In addition to the meshing, the proposed numerical mod-
glass control using the artificial stiffness method and hybrid els have to be calibrated according to the nonlinear properties
formulation (C3D8RH). This element has three translational of the contact surfaces and the stress-strain constitutive law
degree-of-freedoms per node and has the ability to deal with which describes the uniaxial material response of each of the
problems that involve contact, plasticity, and large deforma- components of the specimens.
tions. For the endplate joint specimens, the nut, the bolt-head, The interaction between the column flange and the back of
and the two washers placed on both ends of the bolt (nut and the end plate was modeled using surface-to-surface contact
bolt-head) were meshed as a single body. The density, the con- 3-D elements having tangential and normal behaviors. The
figuration and the number of layers across the thickness of the tangential contact was defined with a frictional contact of 0.3
finite element mesh for each specimen were determined based using the penalty stiffness formulation. The normal contact
on the results obtained from convergence studies in order to was defined with hard contact using augmented Lagrange
provide a reasonable balance between accuracy and compu- formulation. The welded regions were not explicitly modeled.
tational expense. For each specimen, different mesh densities Nevertheless, they were modeled as part of the beam or the
were adopted according to stress concentration. Four layers column material. Tie constrain algorithm was used to simulate
of solid elements were needed to obtain results with enough the behavior of the welding regions that connects the end beam
accuracy for the bolted joints, while convergence was achieved to the endplate and eventually the stiffeners to the column. The
by using one layer of solid elements for the simply supported bolt head, completely fixed to the bolt shank, and the nut were
cellular beams. Overall views of typical finite element models modeled as steel cylinders. They were tied constraint to the
used in this study are shown in Fig. 7, 8 et 9. back of the end plate/column flange by using the tie constraint

4 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

 a) Columns, beams and end plates b) Bolts
Fig. 10 Material stress-strain curves used for numerical analyses

parameter. Tie constraints are used to tie together two surfaces was used to define the failure load.
for the duration of a simulation. They are particularly useful To account for the large displacements (P-∆ effect), “large
for rapid mesh refinement between dissimilar meshes. Contact Displacement static analysis” was adopted. The kinematic
between the bolt shank and the inside of the bolt hole was con- description of the elements in nonlinear geometrical analysis
sidered as frictionless. is based on the updated Lagrangian formulation. Within each
A tri-linear elastic-plastic uniaxial stress-strain curve for the increment, the nonlinear equilibrium equations are solved by
connection components and for the cellular beams was used to means of the Newton–Raphson approach.
simulate the material behaviour of steel in both tension and com- Appropriate boundary conditions, that are consistent with
pression as depicted in Fig. 10(a). A bilinear stress strain rela- the actual constraints which the specimens undergo in the test
tionship was utilized for the high-strength bolts as shown in Fig. setups, were applied.
10(b). The points defining the stress–strain curves are taken from Displacements values recorded for each step load at pre-
the test data given in references [14–18]. The Von Mises yield defined joints were used to calculate the relative joint rotation
criterion was used in conjunction with the Prandtl–Reuss associ- connection. Displacements values recorded for each step load
ated flow rule to determine the onset of yielding and the amount at predefined joints as illustrated in Fig. 11 were used to cal-
of incremental plastic strain. In Abaqus, the material response to culate the relative joint rotation connection by the following
unidirectional stress is input through σt vs. εt, the true stress and formula used in Augusto et al [14]:
the true strain respectively which account for large strains, where:
 U − Uc2 
σ t = σ n (1 + ε n ) (1) φ j = arctan  c1  − θc (4)
 hb 
ε t = ln(1 + ε n ) (2) where:
and εn and σ n are the engineering (nominal) strain and stress Uc1 and Uc2 are the horizontal displacements at the reference
respectively. points C1 and C2.
Furthermore, Abaqus requires the input of the plastic true hb is the distance between points C1 and C2
strains instead of the total true strains in the plastic range. The θc : is the column elastic rotation,
plastic true strain is defined by the following equation:
σt
ε pl = ε t − ε te = ε t − (3)
E
where εte is the true elastic strain.
The constitutive model is integrated by means with a robust
radial return mapping algorithm.
For the endplate joint specimens, the loading was applied
in two steps. In the first step, pretension effects in the bolts
caused by the tightening of each bolt were simulated by apply-
ing compressive forces to bolts. In the second step, monotonic
loading was applied by assigning vertical displacements at the
top of the column for the specimen BC1 and by vertical load at Fig. 11 Reference points and dimensions used to calculate the joint rotation
the end of the beam for the specimens J1.1, J3.1 and J4.1. For
the simply supported cellular beams, the loading was applied For the simple supported perforated beams, vertical dis-
monotonically under force control. The simulation terminates placements for different load levels were plotted at the points
when divergence in the calculation occurs. The last increment where deflections were recorded during the tests.

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 5

 Fig. 12 shows the comparison between the experimental stages, are due to the numerical model simplifications such as
moment versus rotation curves and the numerical ones for the simplified stress-strain curves and to the non-consideration of
specimen J1.1, J3.1, J4.1 and BC1. It can be seen that in general initial geometric imperfections and residual stresses.
the response of the numerical moment–rotation curves agree For the purpose of this research, it can be concluded that the
well with the experimental ones. numerical model described herein can predict with reasonable
In the initial stages, the experimental and the FEA accuracy the global behaviour of bolted end plate connections.
moment-rotation curves are very close. Therefore, the finite To evaluate the accuracy of the nonlinear finite element
element simulations can predict the initial connection stiff- model results of the tested cellular beams described above,
ness Sj.ini with satisfactory accuracy. Though, the finite ele- the numerical load-deflection curves are compared with the
ment simulation can lead to slightly more stiff connection experimental ones, as illustrated in Fig. 14. It can be observed
in the elastic range when compared with the experimental that the load–deflection curves obtained from the beam
results. In the nonlinear range of response, the experimental model agree reasonably well with the experimental results.
and the numerical curves are also close, but not as close as in The noticeable difference between the experimental and the
the initial stage. For a value of rotation, the corresponding numerical results is the ultimate deflection which is certainly
experimental bending moment is slightly higher than pre- due do the same simplifications mentioned previously for the
dicted by the 3D finite element model. Nonetheless, the FEA connections.
values are always on the safe side and the moment resistance
Mj.Rd of the joint is accurately predicted. For the four exper- 4 Effect of openings in beam webs on the
imental specimens, the finite elements simulations slightly characteristics of bolted end plate connections
over-predict the capacity of rotation. Furthermore, Fig. 13 In this section, the joint finite element model is utilized to
shows that the numerical model can reproduce the deflected study the effect of circular and sinusoidal web openings on
shape of bolted end plate connections at failure. the characteristics of bolted endplate connections. Twelve end
All the minor discrepancies observed between the experi- plate connections were analyzed. Four of the connection con-
mental and the FEM curves, and principally in the nonlinear figurations, with solid web beams, as depicted in Fig. 15, were

a)Specimen J1.1 b)Specimen J3.1

c) Specimen J4.1 d) Specimen BC1

Fig. 12 Comparison of the moment rotation curves between numerical and experimental data

6 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

 a) Specimen J1.1 b) Specimen J3.1

c) Specimen J4.1 d) Specimen BC1

Fig. 13 Comparison between numerical and experimental failure modes

a) Specimen A1 b) Specimen A2

c) Specimen B2 d) Specimen AS012

e) Specimen AS016 f) Specimen AS018

Fig. 14 Comparison of experimental and numerical load-deflection curves

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 7

 a) Joint 1 b) Joint 2

c) Joint 3 d) Joint 4

Fig. 15 Geometrical parameters of the solid beam-to-column connections

Fig. 16 Geometrical parameters of openings in the beam webs

utilized as reference specimens. The moment-rotation curve For comparison purposes between the characteristics of the
of each of the four connections is compared with two similar joints with solid beams and those with perforated beams, it
connections configuration but in which the solid beam is sub- was assumed that the total height of a cellular beam and the
stituted with perforated beam with circular or sinusoidal web solid beam are identical. The geometric configurations of the
openings. Geometric limits of the openings adopted for each web openings were adopted in order to minimize the mechan-
specimen were within the limits recommended in the Annex ical characteristics of the connections. This would be realized
N of ENV 1993-1-1 [1] and the design guidelines for Angelina by considering the largest allowed web openings positioned as
beams published by ArcelorMital [20]. Geometric parameters close as possible to the column face.
of the perforated beams adopted in the present study are shown The dimensions of the openings in the web beams are given
in Fig. 16. The diameter of the circular openings d0 is equal to in table 1.
80% of the web depth. Dimensions that define the shape and
Table 1 Dimensions of the openings in the beam webs
the layout of the sinusoidal opening geometric parameters are
as follows: Joint d0 a0 (mm) W (mm) ls

• The depth of web opening a0 to the clear height of the web 1 284.1 275.2 488.8 250

ratio is 0.75 2 319.5 300.8 530.6 250

• W the length of the intermediate web post is equal to 250 3 350.7 359.2 611.6 250

mm 4 385.5 359.2 611.6 250

• ls the length of the sinusoids were determined using prede-

sign charts developed by Arcelor Mittal for non-composite Typical finite element models of the joints with perforated
roof cellular beams [20]. beams are shown in Fig. 17.

8 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

 a) Joint 1

Fig. 17 Typical meshing for perforated beam models

The moment-rotation curves shown in Fig. 18 indicate

clearly that both circular and sinusoidal web beam openings
have little effects on the characteristics of the connection. The
moment versus rotation curves of the connections with solid
beams virtually coincide with those of the connections with
perforated beams. However, a slight difference between the
curves of the joints with perforated beams and those with solid
beams is observed as the height of the beam increases.
b) Joint 2
5 FE modeling of cellular beams connected to the
columns with endplate connections
5.1 Description of the models
The finite element procedures discussed above for both
endplate beam to column joints and simply supported cellu-
lar beams are extended in this section to assess the impact of
various endplate connection types on the global behavior of
cellular beams with circular and sinusoidal web openings.
For this purpose, numerical simulations were performed
on one hundred and twenty-eight full scale one-storey one-
bay steel planar frame models built up from eight cellular
beams, varying the beam to column connection configuration
as shown in Table 2. Four of the reference cellular beams have c) Joint 3

circular openings, while the other four have sinusoidal open-

ings. The parent section of the cellular beams and the columns
are hot-rolled profiles of steel grade S235.
The geometric parameters of the cellular beams (i.e. cross
section and openings) are shown in Table 3. The geometric
properties of cellular beams, columns and joint components
were selected and varied based upon a comprehensive prac-
tical ranges. Various end plate configurations were consid-
ered for the connection between the beam and the columns
flanges: partial depth, flush, extended on both ends, stiffened
or unstiffened in order to obtain joints with various rigidities
and moments of resistance. Furthermore, three configurations d) Joint 4
of web panels were considered: unstiffened, stiffened with Fig. 18 Comparison of moment-rotation curves for solid beam model with
perforated beam models
conventional longitudinal stiffeners and/or diagonal stiffeners.

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 9

For partial depth end plates, the height of the plate was taken 5.2 Results and discussion
as the half of the total beam height. Further parameters which 5.2.1 Ultimate strength
have been varied within the studied frames are the beam and The effects of the beam to column joint upon the global
the column cross sections. behavior of the simulated cellular beams were firstly assessed
Nevertheless, a constant height of 3 m was adopted for all in terms of load-deflection curves at mid-span of the beams.
columns. Two kind of monotonic loading pattern were con- Only selected load-deflection curves were presented in Figs 20
sidered: midpoint concentrated load and uniformly distributed and 21 because they virtually follow the same trend. In these
load. All nodes at the column bases were fully fixed to repre- figures, It can be noticed that the load-deflection curves of
sent a fixed base. Because of symmetry, only one half of the the simply supported cellular beams and those of the cellular
frames were considered with appropriate boundary conditions beams connected to the column with partial depth connections
along the plane of symmetry. A typical meshing of a full scale are almost superimposed. The behavior of the cellular beams
frame is shown in Fig. 19. connected to the column with partial endplate connection
Table 2 Geometry and type of the investigated frame joints
Parent beam End plate Rib Stiffener trs Horizontal stiffeners Diagonal stiffener
Specimens Type of end plate
section tp (mm) (mm) ths (mm) Tds (mm)
IPE 300
FPD300 Partial depth 10 - - -
FF300 Flush 10 - - -
FE300 Extended 10 - - -
FER300 Extended 10 10 - -
FEHD300 Extended 10 - 10 10
FEH300 Extended 10 - 10 -
FEHDR300 Extended 10 10 10 10
SS300 - - - - -
IPE 450 -
FPD450 Partial depth 15 - - -
FF450 Flush 15 - - -
FE450 Extended 15 - - -
FER450 Extended 15 15 - 20
FEHD450 Extended 15 - 20 -
FEH450 Extended 15 - 20 20
FEHDR450 Extended 15 15 20 -
SS450 - - - -
IPE 500 -
FPD500 Partial depth 20 - - -
FF500 Flush 20 - - -
FE500 Extended 20 - - -
FER500 Extended 20 15 - 20
FEHD500 Extended 20 - 20 -
FEH500 Extended 20 - 20 20
FEHDR500 Extended 20 15 20 -
SS500 - - -
IPE 550
FPD550 Partial depth 20 - - -
FF550 Flush 20 - - -
FE550 Extended 20 - - -
FER550 Extended 20 20 - -
FEHD550 Extended 20 - 20 20
FEH550 Extended 20 - 20 -
FEHDR550 Extended 20 20 20 20
SS550 - - - - -

Table 3 Table Geometrical parameters of the analyzed frames

Beam Profile Column Profile Span (m) Circular openings Sinusoidal openings
D W a W Is
IPE 300 HEA 200 14 350 417.2 417.2 250 691.3
IPE 450 HEB 240 22 475 677.6 677.6 250 1072.2
IPE 500 HEB340 24 525 772 772 250 1204.3
IPE 550 HEB340 24 580 772 772 250 1204.3

10 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

is almost identical to that of the simply supported ones. For this service or ultimate loads when compared to the cellular beams
reason, this type of connection is usually classified as nominally connected to the columns with flush endplate connections. It
pinned which validates the numerical model used in this study. was also observed that for a number of the considered cellu-
lar beams using flush endplate connections can considerably
increase their stiffness at service load levels. The bar plots
shown in Figs 22 and 23 shows the ultimate loads obtained for
the simply supported cellular beams and the cellular beams con-
nected to the columns with various endplate connections and in
which a number of the column web panels are stiffened by both
longitudinal and/or diagonal stiffeners. Joints with column web
panels stiffened by both longitudinal and diagonal stiffeners
possess the maximum rigidity and moment resistance among
those considered in this study. The profiles indicated in these
figures are the original parent sections of the cellular beams. It
can be observed from these figures that the use of flush endplate
will improve significantly the ultimate load and deflection of the
cellular beams when compared to the simply supported ones. It
is clear that this strength improvement illustrates the semi-con-
tinuous behavior of the cellular beams connected to the column
Fig. 19 Typical meshing of cellular beams connected to the columns with with flush endplate connections. The increase in the ultimate
flush endplate connections failure load varies between :
1). 315% and 455% for cellular beams with circular open-
The load-deflection curves show that the beams connected ings subjected to uniformly distributed load
to the columns with flush endplates and those connected to the 2). 227% and 489% for cellular beams with circular open-
columns with different types of extended endplate configura- ings subjected to mid-span concentrated load
tions (with or without endplate stiffeners, with or without col- 3). 374% and 624% for cellular beams with sinusoidal open-
umn stiffeners) are almost identical. The use of extended end- ings subjected to uniformly distributed load
plate, stiffened end plate, or stiffeners in column web panel does 4). 339% and 538% for cellular beams with sinusoidal open-
improve neither the ultimate load capacity nor the deflection at ings subjected to mid-span concentrated load.

Fig. 20 Typical load-mid-span deflection curves of a selection of full scale cellular beams with circular openings and for various types of beam to column
connections

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 11

Fig. 21 Typical load-mid-span deflection curves of a selection of full scale cellular beams with sinusoidal openings and for various beam to column connections

The increase in deflection at failure load varies between: as red in the contours. Plastic strains are observed around the
1) 174% and 642% for cellular beams with circular open- opening and the joint.
ings subjected to uniformly distributed load The two specimens demonstrated high plastic deformations
2) 203% and 393% for cellular beams with circular open- around the openings and the joint. For cellular beams with cir-
ings subjected to mid-span concentrated load cular openings, in the regions subjected to high global bending
3) 156% and 466% for cellular beams with sinusoidal open- moment and low global shear force (Fig. 24(b)), it can be shown
ings subjected to uniformly distributed load that the three sections above and below the openings are fully
4) 326% and 468% for cellular beams with sinusoidal open- or partially yielded in compression and in tension. The plas-
ings subjected to mid-span concentrated load. tic zones above and below the openings expanded to the upper
These results also show that the percentage increases in both and the lower flanges as the global moment is increased. As
ultimate load and maximum deflection change significantly the global moment decreases and the global shear increases,
with the geometry of the cellular beams and the type of load- four plastic zones appears at inclined cross sections of the tee
ing. The values of the ultimate loads and maximum deflections web situated at the second opening as can be shown in figure
obtained from the numerical simulations are widely scattered 24(a). A similar yield pattern is also revealed around the sinu-
limiting the possibility of predicting accurately the increase in soidal openings as illustrated in figure 25(a). At the first circu-
strength of cellular beams connected to the columns with flush lar opening from the support, the yield zones overlap due to a
end plates when compared to the simply supported one higher global shear force and interaction effects between the
first opening and the joint. Those yield patterns are observed in
5.2.2 Failure Modes simply supported cellular beams and they lead to the conven-
In this study, the yield criterion is used to predict the failure tional flexural and Vierendeel mechanisms. Because Vierend-
modes of cellular beams. Failure mechanisms associated with eel mechanism is likely to occur in cellular beams with large
instabilities are not considered. Like the von Mises stresses, web openings, four plastic hinges are formed at each corner of
the equivalent plastic strains (PEEQ in ABAQUS) at integral all sinusoidal openings including those subjected to low global
points are scalar values that are often used to define the yield shear force as depicted in figure 25(a). Yielding is also observed
surface size of isotropic ductile materials such as metals. Figs. at the mid-height of the web-post between two adjacent sinusoi-
24 and 25 show the surface plots of the equivalent plastic strain dal openings. This type of yielding pattern may lead to a mecha-
of the two cellular beams with circular and sinusoidal open- nism provoked by the fracture of the mid-depth weld joint. This
ings FPD500 when subjected to uniformly distributed load at failure mode is likely to occur for short horizontal joints of the
failure. All the elements that exhibit plastic strain are shown web-post when they are subjected to high horizontal shear.

12 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

 Fig. 22 Comparison of ultimate loads of cellular beams with circular openings and various types of beam to column connections

Fig. 23 Comparison of ultimate loads of cellular beams with sinusoidal openings and various types of beam to column connections

(a) (b)

(c) (d)

Fig. 24 Plastic equivalent strain (PEEQ) contour plot of specimen FPD500 with circular openings under uniformly distributed load at failure

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 13

 (a)

(c)
(b)

Fig. 25 Plastic equivalent strain (PEEQ) contour plot of specimen FPD500 with sinusoidal openings under uniformly distributed load

Furthermore, the numerical results have also shown that at beam having the same span and subjected to the same load as
failure, all the components of the connection (endplate, bolts the real beam.
and shear panel) and the end-posts of all models exhibit fully Thus, if the bending moment transferred by the connection
or partially plastic deformations as illustrated in Figs. 24 and is known, the mid-span moment would be calculated from:
25. The fact that using more resistant joints does not improve
Mspan = M0 – Msupp (8)
the ultimate load of the cellular beams indicates that the cellu-
lar beams fail before the joints develop their full plastic resis- The mid-span moment of the simply supported beam due
tance or their maximum plastic rotations or they fail simulta- to the uniformly distributed load qu or the mid-span concen-
q QL
neously. Therefore, the flush endplates used in this study can trated load Qu at failure are respectively, ul and u .
2

8 4
be classified as full strength joint. The ultimate loads qu and Qu are obtained from the above
According to EN-1993-1-8 [5], the beam to column end- numerical analyses.
plate connections behave elastically for up to at least 2/3 of The ratio λ of the predicted hogging to sagging moments
its moment resistance. Therefore, it can be concluded that at at failure for the cellular beams connected to the columns
failure, the beam end moments MEd should satisfy: with flush endplate subjected to mid-span concentrated load
and uniformly distributed load are presented in table 3 and
Mj,ela ≤ Msupp ≤ Mj,Rd (5) 4, respectively. The moment resistant of each joint configu-
where: ration was calculated using commercial software developed,
Mj,ela is the maximum elastic moment of the connection specifically for steel joint design, by Autodesk in accordance
Mj,Rd is the moment resistance of the connection. with the component method of EN 1993-1-8 [5]. The ratio
of hogging to sagging moments at failure was calculated by
5.2.3 Moment distribution at failure considering for the support moment the two extreme values,
In this section, the numerical results obtained above are adopted by EN 1993-1-8 [5], which may be experienced by
used to evaluate the moment at the support and the mid-span endplate joints beyond the elastic limit, Msupp = 2/3Mj,Rd and
cross sections in the cellular beams connected to the columns Mj,Rd .It can be observed that the ratios λ of hogging to sagging
with flush endplate at failure. In order to achieve this objective, moments at failure are quite different, depending on the geo-
the following assumption has been assumed: metrical parameters of the cellular beams and the type of the
applied load.
2Mj,Rd /3 ≤ Msupp ≤ Mj,Rd (6)
Equation (7) and the numerical results presented in Table
This assumption was verified for the cellular beams con- 4 and 5 can be used to estimate the ultimate load capacity of
nected to the column with flush end plates considered in this cellular beams with large closely spaced openings.
study. For cellular beams subjected to uniformly distributed load:
At all stages of loading, the sum of the absolute values of 8 (1 + λ )
the mid-span and the supports moments must satisfy the fol- qu = Msupp (9)
λ12
lowing equilibrium equation:
2 M j ,Rd
If we assume M supp = , hence
Msupp + Mspan = M0 (7) 3
0.11 ≤ λ ≤ 0.16 for cellular beams with circular openings
where: 0.06 ≤ λ ≤ 0.12 for cellular beams with sinusoidal openings
M0 is the maximum positive moment of a simply supported

14 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati

If we assume Msupp = Mj,Rd rigidity can be used instead of the conventional simple con-
0.15 ≤ λ ≤ 0.20 for cellular beams with circular openings nections. In this paper, numerical investigations have been
0.09 ≤ λ ≤ 0.16 for cellular beams with sinusoidal openings conducted to study up to failure the behavior of cellular beams
For cellular beams subjected to mid-span concentrated load: connected to the columns with various configurations of
bolted endplate connections. Simulations were also performed
4 (1 + λ )
qu = Msupp (10) on beam to column steel joints in order to assess the effect of
λ1
web openings on the behavior of endplate connections. From
2 M j ,Rd
If we assume Msupp = , hence the numerical results of this investigation, the following con-
3
0.11 ≤ λ ≤ 0.20 for cellular beams with circular openings clusions can be drawn:
0.15 ≤ λ ≤ 0.22 for cellular beams with sinusoidal openings (1) Large circular and sinusoidal web beam openings
If we assume Msupp = Mj,Rd located close to the beam to column joint have little
0.15 ≤ λ ≤ 0.24 for cellular beams with circular openings effects on the characteristics of the connection. Thus,
0.2 ≤ λ ≤ 0.27 for cellular beams with sinusoidal openings the characteristics of the endplate connections with per-
However, experimental and more numerical investigations forated beams can be de calculated by the same methods
are required to determine more precisely the values of the hog- used for joints with solid beams.
ging to sagging coefficient in terms of the joint and the cellular (2) Cellular beams connected to the column with partial
beams properties. endplate connection can be designed as simply sup-
ported.
6 Conclusions (3) The adoption of flush endplate or extended endplate
Cellular beams have always been designed as nominally cellular beam to column joints improves significantly
pinned to the columns. To improve the load capacity of cellu- the ultimate load and deflection of cellular beams when
lar beams, connections which have some degree of rotational compared to simply supported ones.

Table 4 Predicted ratio of hogging to sagging moments for cellular beams subjected to midspan concentrated load at failure
Beam designation M0 (kN.m) Mj,Rd (kN.m) Mspan (kN.m) λ = Msupp /Mspan (kN.m)
M0 - 2/3Mj,Rd M0 - Mj,Rd 2/3 Mj,Rd / Mspan Mj,Rd / Mspan
FF300
(1) 560 86.68 502.21 473.32 0.11 0.15
(2) 480.96 96.88 416.37 384.08 0.15 0.20
FF400
(1) 1506.12 258.24 1333.96 1247.88 0.13 0.17
(2) 1471.25 318.88 1258.66 1152.37 0.17 0.21
FF450
(1) 2009.62 494.55 1679.92 1515.07 0.20 0.24
(2) 2088 571.66 1706.89 1516.34 0.22 0.27
FF550
(1) 2886 608.51 2480.32 2277.49 0.16 0.21
(2) 2465 641.64 2037.24 1823.36 0.20 0.26
(1) Cellular beams with circular openings
(2) Cellular beams with sinusoidal openings

Table 5 Predicted ratio of hogging to sagging moments for cellular beams subjected to uniformly distributed load at failure
Beam designation M0 (kN.m) Mj,Rd (kN.m) Mspan (kN.m) λ = Msupp /Mspan (kN.m)
M0 - 2/3Mj,Rd M0 - Mj,Rd 2/3 Mj,Rd / Mspan Mj,Rd / Mspan
FF300
(1) 502.25 86.68 444.46 415.57 0.13 0.17
(2) 1058.91 96.88 994.32 962.03 0.06 0.09
FF400
(1) 1721.22 258.24 1549.06 1462.98 0.11 0.15
(2) 2719.14 318.88 2506.55 2400.25 0.08 0.12
FF450
(1) 2455.88 494.55 2126.18 1961.33 0.16 0.20
(2) 3633.12 571.66 3252.01 3061.46 0.12 0.16
FF550
(1) 3600 608.51 3194.32 2991.49 0.13 0.17
(2) 4373.2 641.64 3945.44 3731.56 0.11 0.15
(1) Cellular beams with circular openings
(2) Cellular beams with sinusoidal openings

Numerical Investigation into the Ultimate Strength of Steel Cellular Beams 15

 (4) The adoption of extended endplate cellular beam to col- [11] Weynand, K. “Sicherheits-und Wirtsschaftlichkeits- unter suchungenzuran
umn joints instead of flush endplate does not improve wending nachgiebiger Anschlüsse im stahlbau”. “Safety and economic in-
vestigations on the application of resistant connections in steel construc-
either the ultimate load or the stiffness of cellular beams.
tion”. PhD thesis. University of Aachen, Aachen, 1996. (in German).
(5) For a number of cases, endplate connections increase
[12] Girão Coelho, A. M., Bijlaard, F. S. K., Da silva, S. L.”Experimental as-
significantly the stiffness of cellular beams at service sessment of the ductility of extended end plate connections”. Engineering
load level. Structures, 3(9), pp. 1185–1206, 2004. 10.1016/j.engstruct.2000.09.001
(6) Yield patterns distribution around circular and sinu- [13] Nogueiro, P. “Cyclic Behaviour of Steel Connections”. PhD Thesis, Uni-
soidal web openings of the partially restrained cellular versity of Coimbra, Portugal, 2009. (in Portuguese). https://bibliotecadigi-
tal.ipb.pt/handle/10198/4334
beams are similar to those obtained for simply sup-
[14] Augusto, H., Da Silva, L. S., Rebelo, C., Castro, J. M. “Characteriza-
ported ones at failure
tion of web panel components in double-extended bolted end-plate steel
(7) Connections and the web-post between the opening and joints”. Journal of Constructional Steel Research, 116, pp. 271–293, 2016.
the connection experience a high yielding at failure due 10.1016/j.jcsr.2015.08.022
to interaction effects. [15] Abidelah, A., Bouchaïr, A., Kerdal, D. E. “Experimental and analyti-
(8) Values of hogging to sagging ratio were proposed in cal behavior of bolted end-plate connections with or without stiffeners”.
Journal of Constructional Steel Research, 76, pp. 13–27. 2012. 10.1016/j.
order to estimate the load carrying capacity of cellular
jcsr.2012.04.004
beams with large closely spaced circular and sinusoidal
[16] Warren, J. “Ultimate load and deflection behaviour of cellular beams”. MSc
opening subjected to mid-span concentrated load and thesis, School of Civil Engineering, University of Natal, Durban, South Af-
uniformly distributed load. rica, 2001. http://researchspace.ukzn.ac.za/handle/10413/5272
Further research work is required and especially experi- [17] Durif, S., Bouchaïr, A., Vassart, O. “Experimental tests and numerical mod-
mental one in order to fully understand the behavior of par- eling of cellular beams with sinusoidal openings”. Journal of Construc-
tional Steel Research, 82, pp. 72–87, 2013. 10.1016/j.jcsr.2012.12.010
tially restrained cellular beams.
[18] Durif, S., Bouchaïr, A., Vassart, O., “Experimental and numerical inves-
tigation on web-post member from cellular beams with sinusoidal open-
References ings”. Engineering Structures, 59, pp. 587–598, 2014. 10.1016/j.eng-
[1] ENV 1993-1-1:1992/A2:1998 Eurocode 3: Part 1.1–Amendment A2, CEN, struct.2013.11.021
1998. [19] ABAQUS User’s, Theory and Scripting Manuals, Version 6.12. ABAQUS,
[2] Large web openings for service integration in composite floors. Final Inc., Dassault Systems Simulia Corp., Providence, USA, 2012. http://ab-
Report for ECSC Research Contract 7210-PR-315, 2003 European Com- aqus.software.polimi.it/v6.12/index.html
mission, EUR 21345, ISBN: 92-79-01723-3, EU Bookshop, Catalogue [20] ACB+ Design software, ArcelorMittal. www.arcelormittal.com/sections
Number: KI-NA-21345-EN-C, 2006. https://publications.europa.eu/en/ [21] Abolmaali, A., Matthys, J. H., Farooqi, M., Choi, Y. “Development of mo-
publication-detail/-/publication/a4af7d1a-b375-4aaa-855e-4e4159737fe3 ment–rotation model equations for flush end-plate connections”. Journal
[3] Sameer, S. F., Coulson, J., Dinehart, D.W. “Castellated and cellular beam de- of Constructional Steel Research, 61(12), pp. 1595–1612, 2005. 10.1016/j.
sign”. American Institute of Steel Construction, Steel Design Guide, Series jcsr.2005.05.004
Number 31,2003(2016). https://www.scribd.com/document/353101315/ [22] Trahair, N. S., Bradford, M. A., Nethercot, D., Gadner, L. “The behaviour
Design-Guide-31-Castellated-and-Cellular-Beam-Design and design of steel structures to EC3”. Taylor and Francis, 2008.
[4] Mottram, J. T., Zheng, Y. “State-of-the-art review on the design of beam- [23] Bursi, O. S., Jaspart, J. P. “Calibration of a finite element model for isolated
to-column connections for pultruded frames”. Composite Structures, 35(4), bolted end plate steel connections”. Journal of Constructional Steel Re-
pp. 387–401. 1996. 10.1016/S0263-8223(96)00052-9 search, 44(3), pp. 225–262, 1997. 10.1016/S0143-974X(97)00056-4
[5] EN 1993-1-8: Eurocode 3: Design of steel structures - Part 1–8: Design of [24] Bursi, O. S., Jaspart, J. P. “Basic issues in the finite element simulation
joints [Authority: The European Union Per Regulation 305/2011, Directive of extended end plate connections”. Computers & Structures, 69(3), pp.
98/34/EC. 2005. 361–382, 1998. 10.1016/S0045-7949(98)00136-9
[6] Hedayat, A.A., Celikag, M. “Post-Northridge connection with modified [25] Sumner, E. A. “Unified design of extended end-plate moment connections
beam end configuration to enhance strength and ductility”. Journal of subject to cyclic loading”. PhD thesis, Virginia Polytechnic Institute and
Constructional Steel Research, 65(7), pp. 1413–1430, 2009. 10.1016/j. State University, USA, 2003. http://hdl.handle.net/10919/28078
jcsr.2009.03.007 [26] CEN, EN 1993-1-1. Eurocode 3: Design of Steel Structures — Part 1–1:
[7] Hedayat, A. A., Celikag, M. “Fracture moment and ductility of welded con- General Rules and Rules for Buildings, European Committee for Standardi-
nections”. Proceedings of the Institution of Civil Engineers Structures and zation, Brussels, May 2005.
Buildings, 162(6), pp. 405–418, 2009. 10.1680/stbu.2009.162.6.405 [27] Diaz, C., Victoria, M., Martí, P., Querin, O.M. “FE model of beam-to-col-
[8] Tsavdaridis, K. D., Papadopoulos, T. “A FE parametric study of RWS beam- umn extended end-plate joints”. Journal of Constructional Steel Research,
to-column bolted connections with cellular beams”. Journal of Construc- 67(10), pp. 1578–1590, 2011. 10.1016/j.jcsr.2011.04.002
tional Steel Research, 116, pp. 92–113, 2016. 10.1016/j.jcsr.2015.08.046 [28] Soltani, M. R., Bouchaïr, A., Mimoune, M. “Nonlinear FE analysis of the
[9] Wilkinson, S., Hurdman, G., Crowther, A. “A moment resisting connec- ultimate behaviour of steel castellated beams”. Journal of Constructional
tion for earthquake resistant structures”. Journal of Constructional Steel Steel Research, 70, pp. 101–114, 2012. 10.1016/j.jcsr.2011.10.016
Research, 62(3), pp. 295–302, 2006. 10.1016/j.jcsr.2005.07.011 [29] Madjour, A., Soltani, M. R., Harkati, E., Boutagouga, D., Labed, A. “Effect
[10] Aschheim, M. A. “Moment-resistant structure, sustainer and method of re- of web beam openings on the mechanical behaviour of bolted end- plate
sisting episodic loads”. United State Patent, Patent number: US 6012256 A, Connections”. 13TH Arab Structural Engineering Conference, University
Date of Patent: Jan, 11, 2000. http://www.google.com/patents/US6012256 of Blida, Algeria, December 2015.

16 Period. Polytech. Civil Eng. A. Madjour, M. R. Soltani, E. Harkati