Engineering Structures 142 (2017) 67–83

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Numerical modelling and parametric assessment of hybrid flat slabs

with steel shear heads
D.V. Bompa, A.Y. Elghazouli ⇑
Department of Civil and Environmental Engineering, Imperial College London, UK

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

Article history: This investigation examines the performance of hybrid reinforced concrete flat slabs, incorporating fully-
Received 19 September 2016 integrated shear-heads at connections to steel columns, through a series of numerical evaluations and
Revised 22 March 2017 parametric studies. Validations of the adopted nonlinear finite element procedures, which employ con-
Accepted 29 March 2017
crete damage plasticity constitutive models, are carried out against experimental results on hybrid mem-
Available online 6 April 2017
bers. Complementary verifications on conventional reinforced concrete flat slabs are also undertaken to
ensure the reliability of the selected ranges for key modelling parameters. Comparison of the numerical
Keywords:
simulations against the test results shows close correlations in terms of ultimate strength, deformations
RC flat slabs
Hybrid steel/concrete
and stress levels in the constituent elements of hybrid members. This is followed by a series of parametric
Fully-integrated shear-heads assessments on key structural parameters for hybrid flat slabs with steel shear heads. The results of these
Numerical modelling investigations enable the identification of three modes of failure as a function of the interaction between
Rotational response the shear-head and surrounding concrete. The findings permit the development of improved analytical
Ultimate behaviour models for predicting the response as well as the ultimate strength of such members. In addition, recom-
mendations are given for the determination of shear-head dependent parameters, which are required for
practical design purposes, with a particular focus on the embedment length and section size of the shear-
head elements. The suggested expressions for assessing the shear-head characteristics offer a more reli-
able design approach in comparison with existing methods and are suitable for effective practical appli-
cation and implementation in codified procedures.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction A recent study by the authors examined experimentally the

effectiveness of headed stud reinforcement in hybrid slabs with
Shear-head systems are often used in reinforced concrete (RC) fully integrated shear-heads [9]. The test results enabled the devel-
floors either for strength enhancement or to enable connection to opment of analytical models that depict the rotational response,
steel columns. A number of previous studies have examined the flexural strength and punching shear strength as a function of
effectiveness of shear-heads in various configurations against the shear-head embedment length, layout and section size, yet
punching shear in RC flat slabs [e.g. 1–5]. Several recent investiga- only general suggestions for the design of the steel insert were
tions on such hybrid systems have employed fully integrated cru- given. Related studies were also carried on the shear transfer
ciform shear-heads connected to tubular columns [6–8]. The mechanisms and failure criteria in one-way hybrid members (i.e.
assemblages typically consisted of four I-shaped steel profiles with RC beams with fully-embedded shear-heads connected to steel col-
various lengths welded to the four faces of the rectangular col- umns), including numerical and experimental assessments [10–
umns. It was shown [8] that the punching shear strength can be 12]. These studies focused primarily on the ultimate behaviour of
significantly enhanced in comparison with conventional RC flat the RC-to-steel interface region, rather than on developing detailed
slabs. Hybrid steel/column assemblages employing partially- expressions for the design of the shear-heads.
embedded steel inserts, in which a gap is introduced around the Although various numerical investigations on RC flat slabs exist
column to ensure ductile bare-steel shear-head behaviour, was [13–16], very few report the use of concrete damage plasticity
also examined [8]. (CDP) models in simulations [17–19] and they are generally vali-
dated against a limited number of tests with identical or similar
geometries. For example, nonlinear finite element (FE) analysis of
⇑ Corresponding author. relatively thin RC flat slabs subjected to static and pseudo-
E-mail address: a.elghazouli@imperial.ac.uk (A.Y. Elghazouli). dynamic loading was undertaken in order to investigate failure

http://dx.doi.org/10.1016/j.engstruct.2017.03.070
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
68 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Nomenclature

kw factor for failure criterion

Greek letters l0 critical length
d displacement response lch characteristic length of the element
e strain lm mesh size
ec1 crushing strain lv shear-head embedded length
g shear-head distribution factor mi moment action per unit width
r force distribution factor mRk plastic moment of hybrid sectors
kw rotation coefficient mRc plastic moment of concrete sectors
km flexibility factor nv number of shear-heads
l steel-concrete friction coefficient rc 2bc/p and bc = (bc1 + bc2)/2 (for rectangular columns)
ql flexural reinforcement ratio re exterior slab radius
r stress rs slab radius (loading radius)
rc,max strut crushing strength wi, wmax the crack displacement
u dilation angle
w rotation Uppercase latin letters
e potential eccentricity Avv shear-head shear active area
Ec elastic concrete modulus
Lowercase latin letters Es, Ev steel elastic modulus
b0 control perimeter Gc crushing energy
bc column size Kc factor for the shape of the deviatoric plane
bv shear-head width L specimen size
cc,ck location of neutral axis Mv,i moment carried by one shear-head
d bending effective depth V load
d0 shear effective depth Ve volume of the mesh element
dg0, dg aggregate size Vflex is the flexural strength
dvfb centroid of bottom flange Vi is the shear action
fc concrete strength Vtest test ultimate strength
fct tensile strength of concrete Vu,num numerical ultimate strength
fys reinforcement yield strength Vyv,num shear-head yielding in numerical simulations
fyv shear-head yield strength Wv,pl shear-head plastic section modulus
h flat slab thickness
hv shear-head depth

modes and cracking patterns [17]; the material parameters in CDP the shear-head to the punching shear strength of the member. All
were in this case calibrated based on test results of an interior slab- existing methods [7,20] are however validated on a limited range
column connection. Another FE modelling study [18] investigating of tests with similar structural parameters that cover specific ulti-
the punching shear capacity of flat slabs strengthened with con- mate behaviour characteristics. These methods become unreliable
crete screws showed poor predictions of the structural stiffness, for design situations outside the validated ranges.
primarily due to the inadequacy of the crack model employed in This paper investigates the performance of shear-head systems
the simulations. A parametric study was also carried out [19] in in hybrid RC flat slab-to-steel column systems through a series of
which assessments of the elastic stress fields that develop within numerical parametric assessments in which key geometric and
interior slab-to-column connections were carried out, with valida- material parameters are varied. Numerical simulations are carried
tion against a specimen which failed in flexural punching. out using nonlinear finite element procedures which are validated
Detailed numerical studies on hybrid steel-concrete flat slabs against tests on conventional RC flat slabs [22] as well as hybrid
are relatively limited [20,21]. Analysis and parametric assessments configurations tested by the authors [9]. The tests considered
on hybrid and RC slab tests [1,8] were carried out to highlight the showed failure modes ranging from flexure to punching shear
influence of key parameters which govern the performance of steel and included a wide range of structural parameters and geome-
shear-heads in flat slabs [20]. Numerical simulations on hybrid tries. Detailed sensitivity and parametric investigations are under-
steel tubular column/flat slab configurations with very short taken to provide detailed insights into the physical behaviour for a
shear-heads [21] also indicated that, for assessments of punching wide range of hybrid configurations outside of the existing test
shear strength, the shear-head system can be treated as an database [6–9]. Within the parametric investigations, parametric
enlarged column. However, these previous studies were limited assessments were carried out to assess the influence of the
to specific ranges and their scope was not adequate to enable the shear-head embedment length, section size, slab flexibility, slab
development of guidance and procedures which are suitable for bending effective depth, as well as the concrete strength on the
practical application. They either assess the efficiency of existing behaviour of hybrid slabs. The numerical results obtained, enable
models for RC members [1,20], or propose equations that are a direct assessment of the ultimate behaviour in terms of both
developed from rigid shear-head behaviour characteristic for very strength and deformation characteristics.
short embedment lengths [21]. The findings from the numerical investigations undertaken in
A model, modified for connections between RC flat slabs to CFT this study, in conjunction with the experimental results, provide
columns by means of shear-heads, exists [7]. It considers the same reliable validation and further refinement for recent analytical
force distribution obtained from RC flat-slab column behaviour [1] models applicable to a wide range of hybrid slabs proposed by
and an empirically determined factor to assess the contribution of the authors [9]. The rotational response and ultimate strength
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 69

assessments obtained from the simulations are used for the cali-
bration and improvement of a bi-linear rotational model which
was initially validated based on a limited number of available tests
[6–9]. Key observations from the numerical simulations also
enable the definition of shear-head dependent parameters
required for the design of hybrid assemblages incorporating
shear-heads, with a focus on the design of shear-head embedment
length lv and section size. In conjunction with punching shear
design and assessment methods for hybrid flat slabs, which are
developed from established RC approaches [23,24], new expres-
sions for the design of the embedment length and section size of
the shear-head in terms of section moment of inertia, shear active
area, as well as shear-head width and depth, are proposed. Addi-
tionally, relationships to assess the bending moment at which
yielding occurs in the shear-head flange, as a fraction of its plastic
capacity, are also proposed. Finally, practical considerations
regarding the design of the shear-head are highlighted. Consider-
ing the wide range of salient parameters accounted for within
the parametric investigations, and which are also used for analyt-
ical validation, these relationships offer a more reliable design
approach in comparison with existing methods. The proposed
expressions for assessing the shear-head characteristics, as well
as the ultimate strength of the hybrid members, are suitable for
the practical application and implementation in codified
procedures.

2. Numerical simulations

This section describes numerical simulations and validations

carried out using the finite element program ABAQUS [25] in order
to obtain detailed insights into the structural behaviour of RC flat
slabs provided with shear heads which are welded directly to steel
columns. The section first discusses the constitutive models Fig. 1. Numerical models a) 3D overview, b) side view and mesh, c) concrete
adopted for concrete, reinforcement and structural steel. It then uniaxial model, d) steel model.
describes the validations carried out against the experimental
results from hybrid flat slabs tests carried out recently by the
authors [9], as well as results from tests on conventional RC flat
axial stress-strain r-e relationships in compression and tension
slabs [22].
and related scalar damage ratios, the ratio between the bi-axial
compressive to uniaxial compressive strength (fb0/fc = 1.16) and a
2.1. Modelling procedures factor Kc = 2/3 that controls the shape of the deviatoric plane that
represents the ratio of the second stress invariant on the tensile
Three-dimensional (3D) models were constructed in ABAQUS meridian to the one of the compressive meridian. In this model,
[25] to represent the hybrid members. Using symmetry, only a the plastic volume expansion is not proportional to the increase
quarter of the slab assemblage was modelled, as indicated in in stresses (i.e. non-associative flow rule), described by the dilation
Fig. 1a,b. Use was made of 8-nodded brick elements (C3D8R) for angle of the material u, and flow potential eccentricity e = 0.1. Sim-
concrete and steel profiles, in conjunction with 3D wire elements ilarly to previous numerical simulations for one-way hybrid mem-
(T3D2) for the longitudinal reinforcement. For RC specimens, the ber [12], a plastic stiffness degradation scheme, typical for coupled
concrete and reinforcement steel was modelled similarly to the damage-plastic concrete constitutive models [26,27] was adopted.
hybrid members. The contact between the steel profile and the In the pre-peak region, no degradation occurs and the plastic
concrete body was represented using an exponential decay friction strains are equal to the inelastic strains. In the post-peak region,
law. In all models, the reinforcement was embedded in the con- the stiffness reduction enables the development of irreversible
crete body and assigned with full bond conditions. The support plastic strains that are directly proportional to the stress decrease.
and load application plates were 3D solid regions assigned with In this investigation, the compressive stress-strain behaviour
elastic steel material properties. Member displacements were was defined using the Eurocode 2 [28] recommendations for
applied to reference points that are connected with beam multi- non-linear structural analysis. The r-e relationship was defined
point constraints to reaction plates. Similarly, pinned boundary by the concrete compressive strength fc assessed on cylinder spec-
conditions were assigned to a reference point that was tied to imens on the day of testing. The elastic concrete modulus Ec and
the loading plates. Symmetric boundary conditions were incorpo- crushing strain ec1 were determined as a function of fc. The ulti-
rated. The Newton-Raphson approach was adopted for the numer- mate compressive strain considered is ec1 = 0.0035. Close inspec-
ical integration procedure. tion of several constitutive models, including models that
The concrete damaged plasticity model (CDP) was used to rep- account for post-peak compressive behaviour as a function of the
resent the triaxial behaviour of concrete. This is an isotropic scalar crushing energy Gc of the concrete, showed that the deformational
damage model that uses a potential yield surface in the effective response of some of the validated members with brittle failures
stress space (r) derived from a combined Drucker–Prager and was characterised by unrealistic ductile behaviour. To avoid such
Rankine representation. The plastic yield surface is defined by uni- predictions, it was considered that beyond ec1 the compression
70 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

stress drops to zero (Fig. 1c). In tension, linear elastic behaviour three post-cracking strain distributions (Fig. 2c). Non-linear repre-
was accounted for up to cracking, followed by a non-linear tension sentation of the post-peak tension behaviour was chosen as
stiffening law [29], in which the tensile strength of concrete fct was opposed to linear and bi-linear laws, since the latter two produced
assessed as a function of the fc [28]. The numerical input was con- a stiffer numerical response, particularly for members with low
verted into inelastic cracking strain ect,i using the concept of equiv- amounts of ql.
alent crack length in which the crack displacement wi up to In CDP, the material dilation angle u is used to represent the
wmax = 160 lm, is divided by the characteristic length of the ele- stress increase at high confining pressures in the normal-shear
ment lch. The characteristic length of a 3D element is the cubic root stress p-q plane. In RC flat slabs, concrete in the column region is
of the volume of the mesh element Ve. Both structural and rein- under biaxial or triaxial confinement. Hence, for these cases,
forcement steels were modelled with bi-linear elasto-plastic mate- parameters such as u control the numerical predictions. Fig. 2d,e
rial properties without hardening (Fig. 1d). illustrate the numerical response with u varying from 10° to 55°.
Specimen HS13-00 from the hybrid flat slab series was selected As already noted in other studies on RC flat slabs [17,19], the most
to simulate numerical models consisting of hybrid members [9], effective response is given for u = 40°, whereas for the hybrid flat
and Specimen PG1 was chosen to validate numerical aspects for slabs for u = 49°. It shows that for hybrid members, the presence
conventional RC flat slabs [22]. The parameters varied in the sensi- of the shear-head produces higher confinement in the concrete at
tivity analysis were the dilation angle u, mesh size, and the uniax- the connection region to the steel column. The steel-concrete inter-
ial tension material post-peak representation. The steel-concrete face in hybrid members is also subjected to friction. As mentioned
interfacial friction coefficient l was also examined in the case of above, the contact between the steel profile and the concrete body
hybrid members. After achieving good agreement with test results is modelled using an exponential decay law with friction coeffi-
from HS13-00 and PG1, the parameters were applied to the cients varying between l = 0.2–0.6 [30,31]. In the sensitivity anal-
remaining members and compared in terms of load-displacement ysis depicted in Fig. 2f, l was varied from low fiction (0.2) to
response V-d, strain levels e in constituent elements, and failure relatively high friction (0.6). For l = 0.2, the load-strain response
modes. In this section, numerical results are reported only for shows softer behaviour, activated by an earlier slip, in comparison
HS13-00 and PG1 which were used for calibration. Comparative to cases with l = 0.4–0.6. Based on the comparative results shown
assessments for all other hybrid and conventional RC members in Fig. 2f, l = 0.4 was used in all the analysis cases described in the
are given in detail in Sections 2.2.3 and 2.3, respectively. validation studies and parametric assessments.
In terms of mesh density, the results were largely independent
of the size lm for both hybrid (Fig. 2a) and conventional RC (Fig. 2b) 2.2. Hybrid slabs with shear-heads
configurations, (when lm varies from h/5 to h/10, where h is the flat
slab thickness). The use of a fracture dependent post-peak tension 2.2.1. Test set-up and specimen details
behaviour for concrete allowed tension damage localisation as a More detailed validation of the numerical approaches described
function of the chosen mesh size. Close inspection of members above are carried out herein against four large scale tests carried
with low amounts of longitudinal reinforcement ql, provided with out by the authors on hybrid members with cruciform H-shaped
coarser mesh (lm = h/5), showed unrealistic rigid behaviour in the shear-head systems fully embedded into RC flat slabs and welded
cracked regime, primarily due to the delayed cracking as a result to steel column stubs [9] (Fig. 3a). The nominal thickness of the flat
of poor propagation of tension damage in the continuum. Interme- slabs was h = 225 mm, whereas the plan dimensions were
diate values for mesh sizes, around lm = h/7, were chosen as a ref- 2.2 m
2.2 m. The specimens consisted of a closed section steel
erence for further validation and parametric assessment. In column stub that had four directly welded shear heads, which
addition to this, the influence of the accounted stress transfer were fully embedded in the RC flat slab (Fig. 3b,c). The adopted col-
through cracked interfaces was investigated by accounting for umn size was HEB240 in all specimens, with two 20 mm plates

Fig. 2. Identification of constitutive and numerical parameters a) mesh size for HS13-00, b) mesh size for PG1 [22], c) tension stiffening response on PG1 [22], d) dilation angle
u for HS13-00, e) dilation angle u for RC member PG1 [22], f) friction coefficient l.
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 71

Fig. 3. General layout of hybrid members a) test setup, b) reinforcement layout, c) layout of sectors.

welded on the free edges of the profile, resulting in a closed box and Vtest = 991 kN, respectively. Specimens with intermediate
section (240 mm
280 mm). Load transfer plates with 30 mm (HS07-C0) and low (HS03-00) ql had lower ultimate strengths.
thickness were welded to the top and bottom of the column. Pen- The specimen with the lowest conventional ql failed at Vtest = 582
etration of the longitudinal reinforcement through the column was kN, whereas Specimen HS07-C0 failed at Vtest = 880 kN. Similar
allowed by 25 mm drilled gaps. European section HEB 100 shear- behaviour was observed for all hybrid members at early loading
heads, with the embedded length of lv = 370 mm, were welded stages characterised by flexural cracking. Crack widths and pat-
symmetrically to the four faces of the column. Two of the members terns depended on the amount of bending reinforcement. The gen-
were provided with continuity plates around the column. Eight eral in-plane crack path was characterised by orthogonal and
support plates of 40 mm thickness and 180 mm diameter were tied diagonal lines, similar to typical yield-line patterns. Wider cracks
through 32 mm bolts located at the slab reaction radius (rs = 964 - were observed in the region of the shear-head flanges for the weak
mm from the centre of the column). The top flexural reinforcement axis rotation. This led to a slight non-symmetric rotational beha-
ratio ql varied from 0.33% to 1.37%, and the concrete compressive viour, primarily due to the geometry of the column.
strength on the day of testing varied from 29.0 to 39.2 MPa. A sum- No plasticity was recorded by the strain gauges located on the
mary of the specimen details and material properties is given in longitudinal reinforcement as well as the flanges and webs of the
Table 1. shear-heads in Specimens HS13-00 and HS13-C0. Specimen
HS07-C0 exhibited yield strains on the top flange at the column
2.2.2. Experimental results face near the ultimate load. Specimen HS03-00 showed clear duc-
The main experimental results for the four specimens described tile behaviour due to yielding of the reinforcement. For HS13-00,
above are given in Fig. 4 [9]. Specimens HS13-00 and HS13-C0, HS13-C0 and HS07-C0, the failure, attributed to punching shear,
without shear reinforcement and with relatively high longitudinal occurred due to the development of a punching shear crack that
reinforcement (ql = 1.37%) exhibited similar behaviour throughout initiated in the hybrid sector, illustrated in Fig. 3c, at the
the loading process reaching an ultimate strength of Vtest = 1005 kN concrete-to-composite interface (about 1.0d from the tip of the

Table 1
Details of hybrid specimens.

Member L1 = L2 h (mm) c1/c2 d (mm) ql (%) fyl fc lv dv fyv,f/fyv,w Vtest Vu,num Vtest/Vu,num
(mm) (mm) (MPa) (MPa) (mm) (mm) (MPa) (kN) (kN) (–)
HS13-00 2200 225 240/280 177 1.38 536 29 370 102 457/331 1005 994 1.01
HS13-C0* 2200 225 240/280 175 1.38 536 36.5 370 102 457/331 991 975 1.02
HS07-C0* 2200 225 240/280 175 0.75 565 39.2 370 102 457/331 880 894 0.98
HS03-00 2200 225 240/280 178 0.33 547 37.5 370 102 457/331 582 599 0.97
AVG 1.00
COV 0.02
*
C – continuity plate around the column.
72 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 4. Numerical validation of a) HS13-00 and HS03-00, b) HS13-00 and HS07-00, c) reinforcement strain for HS13, d) reinforcement strain for HS03-00.

shear-head) due to the force transfer through struts supported on to failure. Similarly, the overall behaviour of HS03-00, with rela-
the bottom flange of the shear-head. Furthermore, the failure prop- tively low longitudinal reinforcement, was accurately predicted
agated towards the RC sector producing an asymmetric tri- up to the peak strength showing identical elastic and cracked stiff-
dimensional surface bounded by an octagonal pattern on the top ness to those from the test. Beyond the peak strength, the numer-
face and a rectangular pattern on the bottom face of the slab [9]. ical model was unable to predict the softening behaviour due to its
Specimen HS03-00 exhibited a ductile behaviour of the flexural intrinsic continuum approach. However the numerical results were
reinforcement, yet failure eventually occurred due to the disloca- only within a 3% difference from the test, and the failure mode was
tion of a punching cone and was thus characterised as a failure well predicted. Fig. 4d shows that at the column face (LC) the rein-
mode similar to ‘flexural punching’ in conventional RC flat slabs. forcement reached yield strains at load values around 485 kN,
whereas at the composite-to-concrete interface (LI) they remained
within elastic limits. The tension yield flag parameter, that illus-
2.2.3. Comparative results
trates the mesh regions in which the cracking strain is reached,
Fig. 4a shows the predicted numerical response for the two
depicts patterns that correspond to the cross-sectional cracks
hybrid members without continuity plates (HS13-00 and HS03-
obtained in the tests (Fig. 5). All numerical models showed yield
00). They represent the two extreme cases of the studied configu-
regions that resemble the inclined crack patterns from the tests,
ration in which the reinforcement ratio ql varied from 0.33% to
with the model of HS03-00 exhibiting the highest degradation.
1.38%. This led to two distinct failure modes. Specimen HS13-00
was characterised by brittle failure that is correctly predicted by
the numerical model. Both the elastic and cracked stiffness show 2.3. Conventional RC flat slabs
good agreement with the tests. The predicted behaviour of the
specimens with continuity plates (HS13-C0 and HS07-C0) was also To provide wider validation of the modelling procedures and
in good agreement with the tests (Fig. 4b). Although the cracked additional confidence prior to conducting parametric studies,
stiffness of Specimen HS07-C0 shows a slightly more flexible numerical simulations were also carried out tests on conventional
response and marginally higher displacement at ultimate, the dis- RC flat slabs [22]. Similarly to the hybrid tests described in Sec-
crepancy in strength is within 2%, and the failure mode is captured tion 2.2, these members were provided with flexural reinforcement
accurately. The response and strength were also faithfully pre- ratios ql that vary from relatively low (0.25%) to comparatively
dicted for Specimen HS13-C0 as indicated in Fig. 4b and Table 1. high (1.5%) values. The main aim of the tests was to assess the
It is also noteworthy that the predicted ultimate strength Vu,num influence of ql and specimen size on the punching shear strength
of HS13-00 was only 1% lower than the test strength Vtest, and the of conventional RC flat slabs. Double size and half scale members
displacement at failure was within a discrepancy of 3%. In addition, to the Reference Specimen (PG1) were reported. The plan dimen-
the predicted reinforcement strains were nearly identical to those sions of PG1 were 3 m
3 m, whereas its thickness was
recorded by strain gauges in the test (Fig. 4c). They both show that h = 250 mm. The double size specimen (PG3) was 6 m
6 m in
elastic reinforcement behaviour was maintained, both at the col- plan and 500 mm thick. Half scale specimens (PG6-PG9) were only
umn face (LC) and at the composite-to-concrete interface (LI), up 1.5 m
1.5 m in plan with thickness h = 150 mm. The load
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 73

Fig. 5. Comparison between damage zones obtained from numerical simulations against test crack patterns.

arrangement and boundary conditions were similar to the hybrid Overall, the load-displacement (V-d) curves in Fig. 4a,b and
members described in the previous section. The specimen details Fig. 6, as well as the crack patterns in Fig. 5 and strain levels in
and material parameters of the specimens considered herein are Figs. 4c,d, enabled a direct assessment of the nonlinear numerical
summarised in Table 2. models described in Sections 2.2.3 and 2.3. The strength, deforma-
For specimens with high ql (PG1, PG6, PG7, PG11), failure tional response and failure modes obtained from the numerical
occurred in punching due to the dislocation of a conical body out simulations were generally in very good agreement with the corre-
of the slab. In two cases, some reinforcement yielded over the col- sponding test measurements. Having gained confidence in the
umn. None of these specimens reached their plastic regime. On the modelling procedures through the two sets of validations
other hand, specimens with low ql (PG2b, PG4, PG5, PG8, PG9, described above, detailed parametric studies were carried out, as
PG10) showed clear ductile behaviour with a visible plastic pla- described in the following section, in order to provide further
teau. Despite the low ql, the double scale specimen PG3, failed in insights into the behaviour of hybrid slabs with shear heads as well
punching shear at a lower load than its flexural strength [22]. Iden- as to support analytical and design approaches.
tical failure modes were obtained through the numerical models
adopting the procedures described above. In most cases, both the 3. Parametric assessments
elastic and cracked stiffness were accurately predicted, and the
ultimate strength was well captured (Table 2 and Fig. 6a-f). The A total of 92 models were constructed, using the numerical pro-
predicted strength of several members with low ql = 0.25% (e.g. cedures described above, in which steel columns were connected
PG2b, PG4) was slightly overestimated by about 6% on average pri- to RC flat slabs by means of full-embedded shear-heads. The para-
marily due to the late cracking in the numerical model in compar- metric investigations described in this section are undertaken to
ison with the tests (Fig. 6a,b). On the other hand, the ultimate provide detailed insights into the physical behaviour for a wide
strength of member PG5, also provided with low ql = 0.33%, was range of hybrid configurations outside of the existing test database
underestimated with an average Vtest/Vu,num = 1.04 (Fig. 6c). Addi- [6–9]. The experimental study reported by the authors elsewhere
tionally, the cracking load of the double-size specimen was under- [9] focused on one specific RC slab geometry (h = 225 mm) in
estimated by about 30% leading to a premature development of which the reinforcement ratio ql (i.e. ql 0.33%, 0.75%, 1.38%) and
inelasticity compared to the test. However, there is a close agree- the presence of the continuity plate around the column were
ment between the numerical and the test results in terms of the directly investigated [9]. On the other hand, existing investigations
cracked stiffness and the ultimate strength (Fig. 6d). Specimens [6,7], studied two primary flat slab configurations with similar
with high ql (e.g. PG1 and PG6) showed faithful strength predic- thicknesses (e.g. h = 200 mm and h = 300 mm [6,7]) and varying
tions with a discrepancy around 3% from the test values (Fig. 6e,f), shear-head embedded lengths. Also, a single monotonic test with

Table 2
Details of RC members [22].

L1 = L2 (mm) bc (mm) ql (%) d (mm) fc (MPa) Vtest (kN) Vu,num (kN) Vtest/Vu,num (–)
PG3 6000 520 0.33 456 32.4 2153 2218 0.97
PG1 3000 260 1.50 210 27.7 1023 991 1.03
PG11 3000 260 0.75 210 31.5 763 775 0.98
PG2b 3000 260 0.25 210 40.5 440 458 0.96
PG4 3000 260 0.25 210 32.2 408 448 0.91
PG5 3000 260 0.33 210 29.3 550 528 1.04
PG10 3000 260 0.33 210 28.5 540 533 1.01
PG6 1500 130 1.50 96 34.7 238 230 1.04
PG7 1500 130 0.75 100 34.7 241 220 1.09
PG8 1500 130 0.33 117 34.7 140 149 0.94
PG9 1500 130 0.25 117 34.7 115 125 0.92
AVG 0.99
STDEV 0.06
COV 0.06
74 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 6. Validation of RC flat slabs [22] a) PG2b, b) PG4, c) PG3, d) PG1, e) PG6, f) PG5.

a slab thickness of h = 155 mm is reported elsewhere [8]. As plane strut projection d0 plus two critical lengths l0. For short
observed, these experimental studies offer a limited range of test shear-heads, the diagonal lines of the control perimeter extending
configurations. The analytical models validated against such lim- from the shear-head tip regions could join, resulting in a closed
ited number of tests show reliability only for applications within shape. It can be evaluated using Eq. (1a) as the minimum of the
the tested ranges. Hence, parametric investigations with the varia- above two conditions.
tion of main structural parameters outside of the tests range are Besides the obtained V-w, Figs. 8,10,11 assess failure criteria
required to enhance the applicability of these analytical models. (FC) for punching shear in RC flat slabs [23], which were also
Consequently, the numerical results obtained from the numerical extended to hybrid slabs [9] (Eq. (2)), and provide key observations
parametric investigations described in this section enable a direct on member behaviour. These include envelopes of the loads corre-
assessment of the behaviour of hybrid flat slab members in terms sponding to yielding in the steel reinforcing materials, namely:
of strength and deformation characteristics and provide wider val- longitudinal reinforcement (RY), the flange of shear-head (FY),
idation for practical application of the analytical models. the web of shear-head (WY) and the ultimate envelope (U). In
Considering the numerical model for Specimen HS13-00 as a the case of RY, the points on the graph correspond to the initiation
reference (see Table 1 and Fig. 4a,c, the parametric assessments of yielding in one wire element in the analysis. For FY and WY, they
(summarised in Table 3) may be grouped into five main studies correspond to the case in which yielding spreads in a band (i.e. at
in which the following key parameters were investigated: least two mesh elements for FY, typically at the edges of the
flanges; or a band of mesh elements in which WY occurs from
- Shear-head embedment length-to-depth ratio (lv/hv = 0.5–5.0): in the top flange to bottom flange).
each case, the reinforcement ratio ql was varied from 0.33% to pffiffiffi
2.20%, whereas the effective depth d, slab radius rs, concrete b0 ¼ minðpd0 þ 8l0 ; pd0 þ 4½l0 þ ðbc  bv Þ=2 2Þ ð1aÞ
strength fc and shear-head section size (hv
bv = 100 -

100 mm, from a HEB100 profile), were maintained constant l0 ¼ lv þ d0 =2 P hv ð1bÞ
- Slab radius (rs/d = 5.44–8.47): in conjunction with ql = 0.33%-
2.20%, and constant lv, d, fc, rs/d, and hv
bv d0 ¼ d  dv fb  tf =2 ð1cÞ
- Slab bending effective depth (d = 140–330 mm): in conjunction
with ql = 0.33%-2.20% with constant lv/hv, fc, rs, and hv
bv kw ¼ 0:75=½1 þ 15  w  d=ðdg0 þ dg Þ ð2Þ
- Shear-head section size (hv
bv = 60
60-120
120 mm): for
lv = 370 mm and constant fc, rs, d The results of the parametric studies are grouped below in
- Concrete strength from fc = 29 – 80 MPa and ql = 0.33%–2.20% for observations regarding the influence of the shear-head embed-
constant lv/hv, d, rs and shear-head section size ment length, shear-head section size and slab thickness, and the
concrete strength. These results are then used in Section 4 for prac-
The results of the parametric studies are plotted in terms of tical analytical and design considerations.
rotational response (V-w) in which the load V was normalised
against the control perimeter b0 as a function of the critical length 3.1. Shear-head embedment length
l0, shear effective depth d0 and concrete strength fc, as described by
Eqs. (1a)–(1c) and illustrated in Fig. 7a,b. A detailed analysis on Figs. 8a-e illustrate the rotational response V-w from numerical
stress fields [9] showed that b0, required for punching shear assess- models in which the embedment length lv was varied against the
ments in the case of hybrid slabs provided with shear-heads, can flexural reinforcement ratio ql. For low to intermediate reinforce-
be expressed in terms of l0. As shown in Fig. 7b, b0 is defined for ment ratios (ql = 0.33–0.78%) combined with short shear-heads
each shear-head by an arc length with a radius equal to the in- (lv/hv = 0.5–1.0; lv/rs = 0.05–0.10), yielding of the longitudinal
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 75

Table 3
Summary of parametric assessments.

Investigation lv (mm) Type (–) rs (mm) d (mm) d0 (mm) fc (MPa) ql (%)

Embedment length D177-L05-H100-C29 50 HEB100 964 177 117 29 0.33–2.20
D177-L10-H100-C29 100 HEB100 964 177 117 29 0.33–2.20
D177-L23-H100-C29 230 HEB100 964 177 117 29 0.33–2.20
D177-L37-H100-C29 370 HEB100 964 177 117 29 0.33–2.20
D177-L50-H100-C29 500 HEB100 964 177 117 29 0.33–2.20
Flexibility F2-D177-L37-H100-C29 370 HEB100 1500 177 117 29 0.33–2.20
Effective depth D140-L10-H100-C29 100 HEB100 964 140 92 29 0.33–2.20
D230-L10-H100-C29 100 HEB100 964 230 135 29 0.33–2.20
D280-L10-H100-C29 100 HEB100 964 280 152 29 0.33–2.20
D330-L10-H100-C29 100 HEB100 964 300 180 29 0.33–2.20
D140-L37-H100-C29 370 HEB100 964 140 92 29 0.33–2.20
D230-L37-H100-C29 370 HEB100 964 230 135 29 0.33–2.20
D280-L37-H100-C29 370 HEB100 964 280 152 29 0.33–2.20
D330-L37-H100-C29 370 HEB100 964 300 180 29 0.33–2.20
Shear-head section size D177-L37-H60-C29 370 H60 964 177 86 29 0.33–2.20
D177-L37-H80-C29 370 H80 964 177 95 29 0.33–2.20
D177-L37-H120-C29 370 HEB120 964 177 113 29 0.33–2.20
Concrete strength D177-L37-H100-C40-C29 370 HEB100 964 177 117 40 0.33–2.20
D177-L37-H100-C60-C29 370 HEB100 964 177 117 60 0.33–2.20
D177-L37-H100-C80-C29 370 HEB100 964 177 117 80 0.33–2.20

a) Bold letters depict variables in the parametric investigation.

b) Cross-sectional dimensions (b
tf/d
tw/Av/Iv):
H 60: 100 mm
8 mm/100 mm
5 mm/1180 mm2/68.9
104 mm4.
H 80: 80 mm
9 mm/100 mm
5.5 mm/1780 mm2/193.4
104 mm4.
HEB 100: 100 mm
10 mm/100 mm
6 mm/2600 mm2/449.5
104 mm4.
HEB 120: 120 mm
11 mm/100 mm
6.5 mm/3400 mm2/864.4
104 mm4.

by crushing at the composite-to-concrete interface combined with

yielding of the shear-head flange in compression (for lv = 0.5–1.0;
lv/rs = 0.05–0.10) or by local yielding of tension flanges and even-
tual crushing at the interface (for lv = 2.3–5.0; lv/rs = 0.24–0.51).
As depicted in Fig. 8d,f, an increase in slab radius from rs/
d = 5.44 (rs = 964 mm) to rs/d = 8.47 (rs = 1500 mm), for constant
lv = 370 mm (lv/rs = 0.38 and lv/rs = 0.25), leads to an increase in
rotations w at ultimate, which indicates a more flexible slab beha-
viour. In the latter case, inelastic strains develop in the flanges of
the shear-head (FY) at lower normalised strengths due to higher
bending moment carried out by the slab strips which contain the
shear-heads. Although (FY) governed the behaviour for low ql,
punching occurred at ultimate for cases with intermediate to high
ql.
As suggested by the position of (FY) envelopes in Fig. 8a-e, an
enhancement of the composite action is produced with an increase
in lv, primarily due to the higher contact surface between the
shear-heads and surrounding concrete. Although in some cases
yielding of the shear-head flanges (RY) occurs, with its web
remaining within the elastic regime, punching shear governed
the behaviour at ultimate. This is illustrated by the ultimate envel-
opes (U) which are in proximity to the punching shear failure cri-
terion (FC). It is also worth noting that the stiffness degradation
patterns in Fig. 9a-d are characterised by inclined damage zones
which correspond to punching shear failure surfaces.

3.2. Shear-head section size and slab thickness

Fig. 7. Characteristic dimensions for hybrid slabs a) cross-section view, b) top
layout.
The influence of the ratio between shear-head section size and
slab thickness was examined in three parametric studies. Initially,
the slab effective depth d was varied from 140 to 330 mm for two
reinforcement bars (RY) governs the behaviour. As lv increases, embedment length ratios lv/hv = 1.0 and lv/hv = 3.7, and each of the
inelastic strains develop at the shear-head flange edges (FY). For two ratios was also used for four longitudinal reinforcement ratios
lv/hv = 2.3 (lv/rs = 0.24) and ql = 0.78%, (RY) is triggered before ql between 0.33 and 2.20%. In addition, the shear-head cross-
(FY). For intermediate to high reinforcement ratios (ql = 1.38– section was varied from H60 to HEB120 for ql between 0.33 and
2.20%), flexural yielding (FY) is arrested and failure is governed 2.20%, whereas the lv was 370 mm (lv/hv = 0.62–0.31; lv/rs = 0.38).
76 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 8. Comparative results of parametric assessments: embedment length lv: a) lv/hv = 0.5; lv/rs = 0.05, b) lv/hv = 1.0; lv/rs = 0.10, c) lv/hv = 2.3; lv/rs = 0.24, d) lv/hv = 3.7; lv/
rs = 0.38, e) lv/hv = 5.0; lv/rs = 0.52, f) lv/hv = 3.7; lv/rs = 0.25.

For compactness, only the results of six of these variations are crete. On the other hand, as the shear-head section depth increases
depicted in Fig. 10a-f. (hv  80 mm), the behaviour at ultimate reverts to typical crushing
Similar to other observations made before, members with low at the composite-to-concrete interface (Fig. 10d). Fig. 11 illustrates
ql = 0.33% developed inelastic strains in the longitudinal bars the damage tension kinematics of a case in which web yielding
(RY). Apart from slabs with d = 300 mm, in which yielding at the developed. The numerical results show that web yielding was pro-
top edges of flanges was observed, members with short shear- duced by a sudden slip (Fig. 11a) of the shear-head from the
heads (lv/hv = 1.0) exhibited elastic shear-head behaviour with fail- embedding concrete which led to a premature punching shear fail-
ure triggered by crushing at the composite-to-concrete interface. ure (Fig. 11b).
This can be noted from the comparative assessment of the ultimate
envelope (U) and the punching shear criterion (FC) in Figs. 10a, 3.3. Concrete strength
b and 8b. For longer shear-heads (lv/hv = 3.7; lv/rs = 0.38) and rather
thin slabs (d = 140 and 177), the behaviour is governed by punch- The influence of the concrete strength on hybrid slab behaviour
ing at ultimate with inelastic strains occurring in the flexural rein- was considered by varying the concrete constitutive laws in com-
forcement (RY) or shear-head flanges (FY), depending on the slab pression and tension for compressive peak strengths of fc = 29,
configuration (Figs. 10b and 8d). 40, 60 and 80 MPa as described in Section 2.1. Fig. 12 depicts the
As the slab thickness increases (d = 230–330 mm) the ultimate normalised V against rotation w for cases in which fc was varied
envelopes (U) show trends below (FC), which identify failure against ql. For compactness, only the extreme curves for each ql
mechanisms different than those described above (Fig. 10d). For are presented. In the normalised plots, the upper curves represent
these cases, failure was triggered by yielding in the web of the predictions for fc = 29 MPa, whereas the lower curves are numeri-
shear-head (WY), which resulted in lower normalised member cal results for fc = 80 MPa. Additionally, ultimate envelopes (U)
capacities than those from (FC). Similarly, for d = 177 mm and the are plotted for each set of ql and fc, and show that member flexibil-
smallest investigated shear-head section size (H60), failure was ity increases with the decrease in ql combined with higher effec-
also triggered by web yielding (WY) as illustrated in Fig. 10e. As tiveness when low fc is used. Cases with normalised ultimate
a result, for small hv/d < 0.43 and lv/rs = 0.38, a rather undesirable strengths below (FC), with intermediate and high ql (1.38–2.20%),
behaviour of the shear-head assemblage is observed due to the showed failures generally triggered by shear-head web yielding
lower stiffness ratio between the shear-head and surrounding con- (WY).
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 77

transfer the entire load from the steel assemblage to the RC flat
slab, was proposed previously [9]. The model is based on compat-
ible yield mechanisms that divide the in-plane layout into RC sec-
tors and hybrid sectors, with shear-heads embedded in the hybrid
sectors (Fig. 3c). In each sector, corresponding constitutive, com-
patibility and equilibrium relationships apply. A more simplified
bi-linear model, derived from the axisymmetric model, was also
proposed for assessing the V-w response. The model initially devel-
oped from the tests described in Section 2.2, is employed in this
study and appraised against the numerical simulations described
in Section 3. Since the slab flexibility is the governing factor for
assessments of the flat slab punching shear strength [9,23], the
prediction of the load-rotation V-w response must have a signifi-
cant degree of reliability. Hence, a focal point in the numerical
parametric investigations described in Section 3 was the prediction
of the rotational response V-w for configurations outside of the
tested ranges. Consequently, the kw factor in Eq. (3a) that accounts
for slab flexibility through the shear-head embedment length lv
and reinforcement ratio ql is represented in this paper in an
improved form which is validated against the results from the
parametric assessments in Section 4. The remaining equations
required to assess the V-w response have the same form as pro-
posed previously [9].
The rotation w of the hybrid member determined using Eq. (3a),
depends on the section utilisation factor Vi/Vflex, in which Vflex is the
flexural strength of the slab and Vi is the shear action. Other
parameters include the slab radius rs, yield strength fys and elastic
modulus Es of the longitudinal reinforcement, slab effective depth
d, and a factor kw (Eq. (3b)) that accounts for slab flexibility
through the shear-head embedment length lv and reinforcement
ratio ql. The flexural strength Vflex given by Eq. (4a) is a function
of g (Eq. (4b)) which accounts for the in-plane distribution of the
sectors, the plastic moments of the hybrid sectors mRk (Eqs. (5a),
(5b)) and the concrete sectors mRc (Eqs. (5c), (5d)), and the slab
configuration (lv, rs, re, rc). The plastic moments per unit width in
Eq. 5 may be determined from assumptions of linear strain com-
patibility in the cross-section, in which yielding occurs first in
Fig. 9. Cross-sectional damage patterns from numerical simulations a) lv/hv = 1.0; lv/
the tension reinforcement.
rs = 0.10, b) lv/hv = 2.3; lv/rs = 0.24, c) lv/hv = 3.7; lv/rs = 0.38, d) lv/hv = 5.0; lv/rs = 0.52.
 3=2
r s f ys V i
The results of the parametric studies identified three modes of w ¼ kw ð3aÞ
d Es V flex
failure as a function of the interaction between the shear-head
and surrounding concrete: punching shear due to crushing with  1=3
or without yielding of the reinforcement or top flange of the 2 rs
kw ¼ 100ql ð3bÞ
shear-head – referred to as controlled failure (A); failure due to 3 lv
yielding of the compression flange due to inefficient strut support
   
and crushing of the strut (B) and; failure due to yielding of the lv lv re
shear-head flange and web leading to uncontrolled slip (C). These V flex ¼ p g mRk þ 2  g mRc ð4aÞ
rs rs rs  rc
observations, with the complementary key findings from the
numerical investigations described in this study, in conjunction
with those from the test results described in Section 2.2.2, enable g ¼ 8 sin1 ð0:5bv =rc Þ ð4bÞ
the validation of analytical models for a wide range of hybrid slab n 
systems. The V-w responses and ultimate strength predictions are ck  X o
mRk ¼ f ys As d  þ ½Av ij hðdv ij  ck Þðdv ij  ck =2Þi=ðd  ck Þ
used for the modification and validation of a recently proposed bi- 2
linear rotational V-w model [9]. These predictions are also used for ð5aÞ
the definition of key shear-head related parameters which are  D E D E D E
d ft ck;0 d w ck;0 d fb ck;0
required for the design for punching shear and flexure of hybrid f ys As þ Av ft vdc þ Av w vdc þ Av fb vdc
ck ¼ ð5bÞ
k;0 k;0 k;0
systems with shear-heads, with a focus on the specification of
the shear-head embedment length and section size. kf c bc

 cc 
4. Analytical models and design considerations mRc ¼ f ys As d  ð5cÞ
2

4.1. Load-rotation response f ys As

cc ¼ ð5dÞ
kf c bc
An axisymmetric rotational model that considers that relatively
stiff shear-heads in conjunction with the continuity reinforcement where hxi : ð< 0 ¼ 0; P 0 ¼ xÞ
78 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 10. Comparative results of parametric assessments: effective depth d and embedment length lv: a) d/rs = 0.15, lv/hv = 1.0, lv/rs = 0.1; b) d/rs = 0.24, lv/hv = 1.0, lv/rs = 0.1; c)
d/rs = 0.15, lv/hv = 3.7, lv/rs = 0.38; d) d/rs = 0.24, lv/hv = 3.7, lv/rs = 0.38; shear-head section size e) H60 d/rs = 0.18, lv/hv = 6.2, lv/rs = 0.38; f) H80 d/rs = 0.18, lv/hv = 4.6, lv/rs=0.38.

Fig. 13 illustrates a comparison between the V-w response from critical section is situated at the composite-to-concrete interface
numerical simulations with the predicted response from Eqs. ((3)– l0, which results from the length of the control perimeter b0,req as
(5)). Considering the limitations of a bi-linear approach, the model a function of the shear action Vi Eq. (6a) (kw assessed with Eq.
predicts reasonably well the stiffness response of hybrid members (2)). Consequently, the critical length l0,req can be evaluated with
for all reinforcement ratios ql and for practical ratios of embed- Eq. (6b) and the required embedment length lv,req using Eq. (6c).
ment length lv/hv. The predicted Vflex, attained in members with qffiffiffiffi
low amounts of ql = 0.3%, is in good agreement with all numerical b0;req P V i =ðkw f c d0 Þ ð6aÞ
results and test results of Specimen HS03-00. The analytical results
show consistency with numerical results since the full flexural pffiffiffi pffiffiffi
l0;req ¼ max½ðb0;req  pd0 Þ=8; ðb0;req  pd  2ðbc  bv Þ 2Þ=ð4 2Þ
capacity is reached only for models with very low amounts of ql.
ð6bÞ
4.2. Shear-head properties
lv ;req ¼ l0;req  d0 =2 P 2hv ð6cÞ
4.2.1. Embedment length Fig. 14 illustrates the relationship between the shear-head char-
The experimental results described in Section 2.2 showed that acteristic dimensions in the relationship with predictions of the
the critical section in hybrid slabs is located within the critical shear-head length lv,req by Eq. (6c) corresponding to the imposed
shear region at the composite-to-concrete interface, and depends lv in the numerical models and experiments. It may be observed
on the governing strut support location and its inclination, as that relatively long shear-heads, with relatively small depth hv
shown in Fig. 7. The numerical simulations in Sections 2 and 3, compared to slab thickness, are located in the bottom left corner
as partly illustrated in Fig. 9, also show that in all cases the critical of the figure. This shows that the imposed hv and bv are much smal-
shear crack originated from the bottom flange of the shear-head. ler than required for a smooth force transfer. These sizes would
The numerical and experimental results support the assump- also be ineffective since web shear failure may develop.
tion of shear transfer through inclined struts, and the definition On the other hand, short shear-heads, in the range of lv/hv  1.0,
of a critical length l0 Eq. (1b) as a function of the embedment length are unable to support the force-transferring struts, which may lead
lv and shear effective depth d0 Eq. (1c). The embedment length lv of to compression yielding of the bottom flange and slip. These
a shear-head may be determined from the assumption that the members are depicted in the top right corner of Fig. 14 and show
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 79

in the range of the failure criterion (FC), the beneficial effect of

the shear-head is limited and the behaviour becomes similar to a
RC member.
For members with intermediate lv and intermediate shear-head
depths (hv/d  0.43), controlled failures are generally obtained.
Depending on the flexural reinforcement ratio ql, one of the steel
tension members may yield. Punching shear eventually governs
at ultimate in all numerical models for these cases, and the
required lv,req is typically in the same range of the imposed lv. Sim-
ilar predictions are obtained for the tests described in Section 2.2
and those from the literature [6–8].

4.2.2. Section size

The shear-head section size can be determined by means of linear
strain compatibility in the composite cross-section by assuming that
the entire shear-head is represented by a single reinforcement bar
located at its geometrical centre. The stress in a reinforcement com-
ponent and its moment contribution to the total resistant moment is
a function of the cracked stiffness Ik,cr and geometry of the member.
The cracked stiffness depends on the geometrical characteristics of
the constituent elements of the cross-section. For RC members, the
cracked moment of inertia at ultimate varies between 30 and 70%
of the un-cracked cross-section.
For hybrid members with shear-heads, examination of the
numerical results from the parametric assessments combined with
the experimental results described in Section 2.2.2, points to an
average ratio of ck/d = 0.37 at ultimate. This leads to a ratio of the
shear-head moment of inertia Iv to the cracked moment of inertia
of the entire cross-section per unit width of Iv/Ik,cr = 0.25. The plas-
tic moment of the cross-section can be determined by the contri-
bution of longitudinal reinforcement which yields first, as well as
that of the shear-head combined. Considering that the moment
action should be limited by the plastic capacity, the shear-head
section size may be determined from Eq. (7a). For cases without
eccentricity, the average moment per unit width acting within
the support strip mi can be determined from the shear action Vi
using Eq. (7b) [24].
The numerical simulations also identified cases in which yield-
ing of the shear-head web triggered failure (Fig. 10d,e). In such sit-
uations, the hv/d ratio was below 0.43. An analysis using the
Fig. 11. Tension damage kinematics for members with low hv/d a) damage pattern Eurocode 3 [32] relationship for assessing the design plastic shear
after development of a slip crack, b) damage pattern at failure. resistance of the steel insert showed that, in these cases, the shear
active area Avv of the shear-head was not sufficient to transfer the
force from the slab to the steel column. Accordingly, besides the
cross-sectional design in terms of Iv, the shear-head section size
should be determined from Eq. (8) and accounting for a conserva-
tive hv/d  0.5 limit for standard European sections.
Considering that the force is transferred from the slab to the
column through struts supported on the bottom flange, the width
bv of the shear-head dictates the cross-sectional strut thickness
and, consequently, the amount of force transferred. In order to
ensure a smooth transfer, the bottom flange should be relatively
stiff in order to avoid failure in compression in the steel insert.
The minimum shear-head width bv may be determined with Eq.
(9a) based on the assumption that the transferring struts are sup-
ported through the entire embedment length lv, in which crushing
strength of the struts may be assessed with (Eq. (9b)) [24].

mi d  ck 
Iv P 1 ð7aÞ
20 f ys d
Fig. 12. Influence of fc on the performance of hybrid slabs.

mi ¼ V i =8 ð7bÞ
that the imposed lv in the parametric studies are below the pffiffiffi
required lv,req. Although, in these cases, the behaviour at ultimate 5 Vi 3
Avv P ð8Þ
is controlled and the ultimate envelopes (U) in Figs. 8 and 10 are 4 nv f yv
80 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 13. Comparison between numerical and analytical V – w a) lv/hv = 2.3, b) lv/hv = 3.

expressed as a function of the shear action Vi and the assumed dis-

tribution in Fig. 15 as depicted by Eq. (11b).
 
1 lv
j¼ 1 ð11aÞ
3 rs
  
Vi hv lv hv
Mv ;i ¼ j þ ð1  jÞ þ ð11bÞ
nv 8 2 8
In hybrid members with shear-heads, the full plastic capacity of
the steel insert is generally not reached since the behaviour is gov-
erned by the interaction properties between the shear-head and
embedding concrete. The numerical results showed that the
shear-head/concrete slip is dependent on the contact surface and,
consequently, the embedment length lv. Hence, the shear-head
develops inelastic behaviour, primarily at its top flange, at a frac-
Fig. 14. Comparative assessment of shear-head characteristics against predictions.
tion of its full plastic capacity (Eq. (12a)). This can be estimated
by accounting for through the km factor which was determined
Vi 1 from the numerical simulations in which yielding of the shear-
bv P ð9aÞ head flange was recorded (Eq. (12b)).
nv rc;max lv
Mv ;i 6 km W v ;pl f yv ð12aÞ
1=3
rc;max ¼ 0:55ð30=f c Þ f c ð9bÞ  
d0 lv
km ¼ kw ð12bÞ
The above equations for the design of shear-head section r s hv
moment of inertia, shear active area and width, offer a set of
requirements that should be met in the design procedure of the Fig. 16 shows the predictions of Eq. (12a) against the Vu,num/Vyv,
steel insert. In conjunction with the above expressions, supple- num ratio in which Vu,num is the ultimate strength and Vyv,num is the
mentary verifications regarding the moment capacity of the force at which yielding developed in the top flange (FY). Values
shear-head, as described in the following section, must be carried below 1.0 show that no plastic behaviour is predicted (Mv,calc/(km-
out. Mv,pl)<1.0), and elastic shear-head behaviour was developed in
simulations (Vu,num/Vyv,num < 1.0). It is depicted that the assumed
force distribution (Fig. 14) in conjunction with the considered
4.2.3. Moment capacity moment behaviour (Eqs. (12a), (12b)) shows good agreement
Fig. 15 illustrates the recorded force distribution per unit width between predictions and numerical results.
(m/mmax) at the top flange of the shear-head as a function of the
embedment length lv. The qualitative distribution (m/mmax) of the 4.3. Practical application
applied load is based on the strains from numerical simulations
and integrated over the cracked cross section. It is observed that The results from the numerical assessments in conjunction with
the shape of the diagram, which can be divided into two regions, the test results from Section 2.2.2 enabled the definition of the
varies with lv. In all cases, the peak value was recorded within a above expressions for the design of shear-heads in hybrid flat slabs.
region of hv/4 from the column face. For short and relatively rigid The required embedment length lv can be determined from the
shear-heads, the distribution takes nearly a triangular from, shear action Vi in the connection region using the modified RC
whereas for flexible steel inserts it shows a non-linear form. In a approach [30] for hybrid configurations (Eqs. (1)–(6)). This has pre-
simplified manner, it can be represented by two rectangular viously been validated on a number of tests [9], and for a wider
regions delimited by hv/4. Close inspection of the numerical results range of parameters in the numerical studies described in Section 3.
shows that the average reaction forces in each region vary with The results of the numerical simulations also resulted in expres-
shear-head flexibility, and the amount of force transferred through sions for the design of the shear-head cross-section in terms of
each region can be expressed as a function of the r factor in Eq. minimum moment of inertia Iv, shear area Avv and width bv, as a
(11a). The moment carried by each shear-head Mv,i can be function of the shear action Vi (Eqs. 7a-9b).
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 81

Fig. 15. Force distribution on the shear-head with the following embedment length lv a) 100 mm, b) 230 mm, c) 370 mm, d) 500 mm.

Fig. 16. Prediction of Eqs. (11) and (12) in relationship with numerical results (Vu,
num – ultimate strength, Vyv,num – yielding of the top flange).

Fig. 17 shows predictions for Iv and Avv plotted against details of

some practical shear-head section sizes from European HEB sec-
tions embedded in flat slabs with d = 180–330 mm. It is observed Fig. 17. Predictions of Eqs. (7) and (8) as a function of the shear action Vi.
that predictions of Eq. (7) show that the required section size
increases proportionally with the shear action Vi, apart from the
case of d = 180 mm. Generally, in this situation, the required shear ted against the shear action to flexural strength Vi/Vflex ratio. Values
area Avv (Eq. (8)) governs the shear-head cross-sectional design. In of Vi/Vflex < 1.0 indicate elastic reinforcement behaviour, whereas
cases in which the effective Avv is smaller than that required, web values above 1.0 depict situations in which flexural yielding occurs.
failure could develop; this type of behaviour was observed in It is generally observed that the low amounts of ql lead to flexural
numerical simulations in situations in which the shear-head from failures and high shear action Vi requires very long shear-heads
European HEB section type had a depth hv less than d/2. Hence, with lv > rs. Members provided with low to intermediate reinforce-
as a general guide, after the assessment of the shear-head section ment (e.g. ql = 0.75%) may also reach their flexural strength when
size using Eqs. ((7)–(9)), hv should be at least d/2 for standard Euro- small shear-head section sizes are employed (some cases were
pean sections, whilst the maximum shear-head depth would be identified in Figs. 8 and 10). Typically, irrespective of lv, intermedi-
limited by practical aspects including slab thickness, the amount ate to high ql produce elastic reinforcement behaviour. Generally,
of longitudinal reinforcement and concrete cover. for the same shear action Vi, low ql ratios require shear-heads with
A sensitivity analysis on Eq. (6) to assess lv, along the results higher lv/rs in comparison with intermediate and high ql, which
from numerical simulations and available tests [6–9] is depicted may necessitate the use of short to intermediate shear-heads. In
in Fig. 18a. The shear-head section sizes varied from HEB100 to addition to this, Fig. 13 indicates ineffective behaviour for short
HEB 140 and for each case the reinforcement ratio ql varied from shear-heads (lv/rs < 0.1), since they were not able to ensure a
0.33% to 2.20%, whilst hv/d = 0.5, re/rs = 1.2, fc = 30 MPa, fys = 500 - smooth force transfer through struts. Also, available tests [6–9]
MPa, fyv = 355 were maintained. The predicted lv/rs ratios are plot- and numerical results from Section 3 for 0.1 < lv/rs < 0.55 showed
82 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83

Fig. 18. Comparison between a) Predictions of Eq. (6) and numerical results, b) Predictions of Eqs. (11) and (12) and numerical results.

effective behaviour and seem generally practical. These observa- shear heads. To provide wider validation of the modelling proce-
tions point to an effective use in the design of shear-heads with dures and additional confidence prior to conducting parametric
embedment length-to-slab radius ratios within the range of lv/ studies, numerical simulations were also carried out against tests
rs = 0.2–0.5, mainly due to their practicality and stable structural on conventional RC flat slabs. The numerical results showed good
behaviour. agreement with both sets of test results, in terms of ultimate
Fig. 18b illustrates the predictions of Eqs. (11) and (12) against strength as well as deformations and stress levels, indicating the
Vi/Vflex ratio, including the results from the numerical simulations suitability of the adopted modelling procedures including the con-
as well as available tests results [6–9]. This figure depicts results crete damage plasticity approach for simulating the response of
from the same sensitivity analysis described in Fig. 18a with a hybrid and conventional RC flat slab members.
focus on the shear-head behaviour. It is observed that, considering A total of 92 models were considered within the numerical
the predictions of Eqs. ((7)–(9)), irrespective of the shear-head sec- parametric studies in order to obtain detailed insights into the
tion size, for low amounts of ql, both the reinforcement and shear- physical behaviour of hybrid slab configurations with shear-
head top flange yield, whereas for intermediate and high amounts heads. This enabled direct assessment of the behaviour both in
only the top flange develops inelastic strains. This correlates well terms of deformation response and ultimate strength. It also per-
with the numerical results and shows that in tests [6–9], the mitted close examination of the main failure modes, with a focus
shear-head may have developed inelastic behaviour at the edges on those developing at the composite-to-concrete interface, which
of its top flanges, which was not captured by the strain gauges was governed by punching shear with or without yielding in the
(typically located above the shear-head web). The above remarks steel. The results of the parametric investigations also provided
show that Eqs. (11) and (12) may be used for assessing the onset information on the member kinematics which complements those
of yielding of the shear-head top flange. This may be more effective obtained directly from the tests. Three primary modes of failure
for design since the shear-head full plastic capacity is generally not were identified as a function of the interaction between the
attained due to characteristic hybrid slip behaviour. shear-head and surrounding concrete: (i) punching shear due to
As noted before, the results from the numerical simulations, in crushing with or without yielding of the reinforcement and/or
conjunction with experimental results described in Section 2.2.2, top flange of the shear-head, (ii) failure due to yielding of the com-
enabled the development of relationships for the design of shear- pression flange as a result of ineffective support and crushing of
heads in RC flat slabs at interior connections to steel columns the struts, and (iii) failure due to yielding of the shear-head flange
which may be directly used in the design of such elements. The and web resulting in uncontrolled slip.
improved bi-linear rotational model described in Section 4.1, The key findings from the numerical investigations described in
alongside the design expressions for punching shear proposed pre- this study, in conjunction with available experimental results,
viously by the authors [9], cover all the key aspects required for the enabled improvements to analytical models for hybrid slabs and
design of hybrid RC flat slab-steel column systems of the form comparison against a wide range of geometries and constituent
examined in this study. Accounting for the wide range of salient parameters beyond those directly examined in tests. The results
parameters considered in the parametric investigations, the of numerical simulations combined with results from existing tests
expressions proposed in this paper offer a reliable design approach show that the critical shear regions occur at the composite-to-
for hybrid members with shear-heads. The suggested equations for concrete interface and that the critical shear crack originates from
assessing the shear-head characteristics and the ultimate strength the bottom flange of the shear-head. The findings support the
of the hybrid members are suitable for direct practical application assumption of a force transfer through inclined struts as well as
and implementation in codified procedures. the definition of a critical length which determines the control
perimeter as a function of the shear-head embedment length and
shear effective depth. The rotational responses and ultimate
5. Concluding remarks strengths obtained from the simulations were also used for the
modification and improvement of a simplified bi-linear rotational
This paper investigated the performance of shear-head systems model.
in hybrid RC flat slabs connected to steel columns. Numerical sim- Importantly, the observations from the numerical simulations
ulations were carried out, using nonlinear finite element proce- permitted the definition of shear-head dependent parameters
dures with concrete damage plasticity model, and validated required for the design of hybrid assemblages incorporating
against a series of tests on hybrid steel/concrete flat slabs with shear-heads, with a focus on the design of shear-head embedment
 D.V. Bompa, A.Y. Elghazouli / Engineering Structures 142 (2017) 67–83 83

length and section size in terms of section moment of inertia, shear [9] Bompa DV, Elghazouli AY. Structural performance of RC flat slabs connected to
steel columns with shear heads. Eng Struct 2016;117:161–83.
active area, shear-head width and depth. Additionally, equations to
[10] Bompa DV, Elghazouli AY. Force transfer mechanisms between steel columns
assess the bending moment at which yielding occurs in the shear- and reinforced concrete beams by means of shear keys. In: Proceedings of
head top flange as a function of its plastic capacity, were also pro- Eurosteel 2014, 7th International Conference on Steel and Composite
posed. Finally, practical considerations regarding the design of the Structures 2014 September 10–12; Naples, Italy; 2014. Session 6, Paper No
03–476.
shear-head were highlighted. In particular, the results indicated [11] Bompa DV, Elghazouli AY. Ultimate shear behaviour of hybrid reinforced
that more effective application can be obtained from intermediate concrete beam-to-steel column assemblages. Eng Struct 2015;101:318–36.
rather than relatively short or long shear-heads provided that the [12] Moharram MI, Bompa DV, Elghazouli AY. Experimental and numerical
assessment of mixed RC beam and steel column systems. J Constr Steel Res
depth is at least half the slab thickness. Accounting for the wide 2017;30(131):51–67.
range of salient parameters considered within the parametric [13] Hallgren M, Bjerke M. Non-linear finite element analyses of punching shear
investigations, which were subsequently utilised for validation, failure of column footings. Cement Concr Compos 2002;24(6):491–6.
[14] Mamede NF, Ramos AP, Faria DM. Experimental and parametric 3D nonlinear
the analytical expressions proposed in this paper offer a more reli- finite element analysis on punching of flat slabs with orthogonal
able design approach in comparison with existing approaches. The reinforcement. Eng Struct 2013;48:442–57.
suggested equations for determining the required shear-head [15] Shu J, Plos M, Zandi K, Johansson M, Nilenius F. Prediction of punching
behaviour of RC slabs using continuum non-linear FE analysis. Eng Struct
properties and predicting the ultimate strength of the embedding 2016;15(125):15–25.
hybrid members are suitable for direct practical application and [16] Soares LF, Vollum RL. Influence of continuity on punching resistance at edge
implementation in codified procedures. columns. Mag Concr Res 2016;68(23):1225–39.
[17] Genikomsou AS, Polak MA. Finite element analysis of punching shear of
concrete slabs using damaged plasticity model in ABAQUS. Eng Struct
Acknowledgements 2015;98:38–48.
[18] Wörle P. Enhanced shear punching capacity by the use of post installed
The authors acknowledge the financial support of the Research concrete screws. Eng Struct 2014;60:41–51.
[19] Bompa DV, Onet T. Punching shear strength of RC flat slabs at interior
Fund for Coal and Steel of the European Community within the connections to columns. Mag Concr Res 2016;68(1):24–42.
project SMARTCOCO: ‘‘Smart Composite Components – Concrete [20] Eder MA, Vollum RL, Elghazouli AY, Abdel-Fattah T. Modelling and
Structures Reinforced by Steel Profiles’’, Grant No. RFSR-CT-2012- experimental assessment of punching shear in flat slabs with shearheads.
Eng Struct 2010;32(12):3911–24.
00031, for the tests described in this paper. The many useful dis- [21] Yan PY, Wang YC. Hybrid steel tubular column/flat slab construction—
cussions with all the collaborators within the project, particularly Development of a shearhead system to improve punching shear resistance. J
Professors André Plumier and Hervé Degée, are also gratefully Constr Steel Res 2016;119:154–68.
[22] Guandalini S, Burdet OL, Muttoni A. Punching tests of slabs with low
acknowledged. reinforcement ratios. ACI Struct J 2009;106(1):87.
[23] Muttoni A. Punching shear strength of reinforced concrete slabs without
References transverse reinforcement. ACI Struct J 2008;105(4):440–50.
[24] fib – Fédération internationale du béton. Model Code 2010 – final draft, vol. 1
& vol. 2, fib Bulletins 65 & 66, Lausanne (Switzerland); 2012.
[1] Corley WG, Hawkins NM. Shearhead reinforcement for slabs. ACI J Proc
[25] DSS (Dassault Systèmes Simulia Corp). ABAQUS Analysis user’s manual 6.14-2,
1968;70(11):811–24.
DSS, Providence, RI, USA. 2014.
[2] Piel W, Hanswille G. Composite shear head systems for improved punching
[26] Lubliner J, Oliver J, Oller S, Oñate E. A plastic-damage model for concrete. Int J
shear resistance of flat slabs. Compos Constr Steel Concr V 2006:226–35.
Solids Struct 1989;25:299–326.
[3] Tuchschmid AG. Tuchschmid-WALM Dokumentation –
[27] Lee JH, Fenves GL. Plastic-damage model for cyclic loading of concrete
Stützenkopfverstärkung – Optimale Krafteinleitung auf tragende
structures. J Eng Mech 1998;124(8):892–900.
Stützen. Frauenfeld (Switzerland): Tuchschmid AG; 2007.
[28] EN 1992-1-1. Eurocode 2: Design of concrete structures – Part 1-1: General
[4] Huber K, inventor; Geilinger.AG, assignee. Method of manufacturing a
rules and rules for buildings Brussels. European Committee for
building. European patent EP0128994. 1984 December 27.
Standardization; 2004.
[5] Chana PS, Birjandi FK, Design guidance on structural steel shearheads in
[29] Hordijk DA. Local approach to fatigue of concrete (Doctorate
concrete (shearhead development tests). Report. Department of the
dissertation). Delft, Netherlands: Delft University of Technology; 1991.
Environment, Imperial College London UK; 1996 July, CRIC Report No.
[30] Baltay P, Gjelsvik A. Coefficient of friction for steel on concrete at high normal
CRIC95/001/F.
stress. J Mater Civ Eng 1990;2(1):46–9.
[6] Lee CH, Kim JW, Song JG. Punching shear strength and post-punching
[31] Rabbat BG, Russell HG. Friction coefficient of steel on concrete or grout. J Struct
behaviour of CFT column to RC flat plate connections. J Constr Steel Res
Eng 1985;111(3):505–15.
2008;64(4):418–28.
[32] EN 1993-1-1. Eurocode 3: Design of steel structures – Part 1-1: General rules
[7] Kim JW, Lee CH, Kang THK. Shearhead reinforcement for concrete slab to
and rules for buildings Brussels. European Committee for Standardization;
concrete-filled tube column connections. ACI Struct J 2014;111(3):629–38.
2005.
[8] Eder MA, Vollum RL, Elghazouli AY. Inelastic behaviour of tubular column-to-
flat slab connections. J Constr Steel Res 2011;67(7):1164–73.