Вы находитесь на странице: 1из 26

h o c . lnstn Ciu. Engrs, Part 2, 1991, 91, Dec.

, 679-704
PAPER 9744 BUILDING
BOARD
STRUCTURAL
AND

Partial shear connection


in composite beams
for buildings

R. P. JOHNSON, MA, FEng, FICE, FIStructE*


IR N. MOLENSTRA, PhDt

The use of partial shear connection in composite beams requires that the slip capacity of the
connectors shall not be less than the maximum slip required for the to beam
reach its design
ultimate load. Maximum slip depends on many parameters, of which degree of shear connec-
tion and span are the most significant. It is also difficult to predict slip capacities. Research
on these subjects is reported, based on non-linear numerical analyses, test data, and para-
metric studies. It is compared with extensive work done in France, and related to design rules
given in recent Britishand draft European codes ofpractice.

Notation
A cross-sectional area
b breadth
d diameter of shank of stud connector
E Young's modulus ofelasticity
F longitudinal force
f';f strength of a material; function
h depth of member or component
k factor for strengthof shear connector in profiled sheeting
L span of beam; for two spans,L , 2 L ,
M bending moment
N number of shear connectors between a point of maximum sagging moment and a
support
*f value of N that would provide full shear connection
n degree of shear connection; number of nodes
%l"
minimum permitted degreeof shear connection
P shear force per connector
p, design shear resistanceof a connector
S longitudinal slipat interface between steeland concrete
S,, maximum computed longitudinal slip in a beam when subjected to its design ulti-
mate load
W design load for a span of a beam
W design load per unit length
X coordinate along a beam
Z distance between lines of action of forcesF , and F ,
?m partial safety factor for a material
or shear connector
E longitudinal strain
Written discussion closes18 February 1992; for further details
see p. ii.
* Professor of Civil Engineering,
Delivered
University of Warwick.
by ICEVirtualLibrary.com to:
t Bylander Meinhardt Partnership, Singapore.
IP: 137.108.145.45
679
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
cp curvature

Subscripts
d structural steel
C compression;concrete; capacity
cu cube
d design
e equilibrium method ofdesign
l interpolation methodof design
k characteristic
P profiled steel sheeting; plastic
PI plastic
R resistance
t tension
U ultimate
Y yield

Introduction
The behaviour of the shear connection in a composite beam depends mainly on
the relationship between the shear force per connector P , and the longitudinalslip
S, at the interface between the top flange of the steel section and the concrete or
composite slab. This load-slipcurve, usually found from a push test, depends on
the typeof connectors, their spacing, the dimensions and reinforcement of the slab,
the rib shapeof any profiled sheeting used, the position of a connector within a rib,
and the properties of the materials. The overall behaviour also depends on the
span of the beam L, and on the degreeof shear connection N I N , , where N is the
number of connectors in the shear span considered, and N , is the number that
would provide full shear connection.
2. If premature shear failure, which can be sudden. is to be avoided, N / N ,
must exceed a value that depends on the span, the type of loading,andthe
load-slip curve for the connectors. Proposals for limiting values are given, based
on non-linear numerical analysis and parametric studies on over 160 beams.
3. Thescope includes continuous beams, with steel rolled sections orplate
girders, normal or lightweight concrete slabs, and any typeof connector for which
load-slip data are available. The conclusions are conservative for certain compos-
ite slabs, where theprofile shape is such that theslip capacity of the shear connec-
tion far exceeds thatdetermined for anunhaunched reinforced concreteslab
(referred to as a solid slab ).

Full and partial shear connection


4. For a given beam, loading, and design method, full shear connection is
defined as the least number of connectors, such that the bending resistance of the
beam would not be increased if more connectors were provided. Accordingly, N ,
can be defined only in relation to particular methods of design and of testing
connectors. Reference is made in this Paper to the methods of the British codes C P
117: Part l 3 and BS 5950: Part 3, section 3.1,4 to the 1985 draft of Eurocode 4,5
referred to as EC4:1985, and to the1990 draft of Part 1 of Eurocode 4,j referred to
as EC4: 1990.
5. Design methods using partial shear connection aregiven in the last three of
these four codes. TheyDelivered by mainly
are needed ICEVirtualLibrary.com
where profiledto:sheeting is used, because
680 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
it may not be possible to provide full shear connection using only one connector
per trough. The concrete rib formed by a trough may not be strong enough to
resist the force from two or more connectors. Partial shear connection also pro-
vides more choice forthe designer, and canreduce cost.
6. The only standardized method of testing shear connectors is the push test.
The first one tobe specified in British codes dates from19653and is for solid slabs.
It is knownto give over-conservativebecauseitsslabs arenarrow (300
mm) and have only light and poorly-anchored reinforcement. A more suitable test
specimen has beendevelopedforstud connectors in composite References
to push tests made below are tothese two typesof test, unless noted otherwise.
7. In design with full shearconnection, it is assumed that failure of shear
connectors does notoccur and that theeffects of slip canbe neglected. With partial
shear connection, the effects of slip cannot be evaluated simply, so design is based
on one of two simplified methods. These methods areapplicable only to members
with steel sections in class 1 or class 2 (plastic or compact), for which moments of
resistance M, are evaluated by rectangular-stress-block(plastic)analysis. For
connectors that are ductile (to be defined later), the methods areas shown in Fig.
L

l(a). The bending resistances M and M , relate to the composite and steel sec-
41
tions respectively. LineA D is thelnterpolntion method, given by
M, = M, + n(M,, - M,) (1)
where n = N j N , ; and curve A C D is the much less conservative equilibrium
method. In this method, one calculates the least compressive force in the slab F,,
that will provide the required bending resistance when the steel section is fully
plastic, as in Fig. l(b), andprovides N connectors, given by
N = F,/P,, (2)
where P , is the design resistance per connector. For both methods, a minimum
degree of shear connection nminis specified.

c
Section Stresses

Delivered by ICEVirtualLibrary.com
Fig. 1 . Interpolntion and equilibrium methodsof design to:
IP: 137.108.145.45
681
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
8. For a continuous beam, one could use partial shear connection for sagging
bending only, or for both sagging and hogging bending. The second alternative
was specifiedin EC4: 1985.Equation (1) was replacedby

where W is the design load for the span, and W,, and W, are the design ultimate
loads given by rigid-plastic analysis of the composite and steel members respec-
tively.
9. This relationship between design load and degree of shear connection was
used for the continuous beams studied in the present work, and the interpolation
method was also used for the simply-supported beams, except where noted other-
wise.
10. For a given design load W, and a simply-supported beam, equation (3) is
correct at both n = 0 and n = 1. But for a continuous beam with a compact (class
2) composite section at an internal support, the design load is determined by
elastic global analysis,with limited redistribution of moments. Its maximum value
(Wcd, say) isless than Wp,, so equation (3) implies that full shear connection is
never required in beams of this type. In the present work, the values of n used in
analyses of a given beam were so chosen that the values W, given by equation (3)
did not exceed W,, for that beam.
1 1 . Tests on continuous beams show that the tensile force in the reinforcement
at an internal support usually exceeds the yield value F, beforeeither of the
adjacent spans fails, because of strain hardening. Partial shear connection is not
appropriate for this reinforcement; and in beams withcomposite slabs, theforce F ,
is usually so low that full shear connection can easily be provided in the troughs
available. Partial shear connectionis now specified for sagging bending only,both
in BS 5950 and in EC4: 1990.
12. For a span continuous at both ends and with symmetricalloading,the
required number of connectors in a half span thenbecomes

+ (4)
where F , is as above, and F , is the tensile force used when calculating the full-
interaction resistance M,, in hogging bending. The difference between the two
equilibrium methods is small (exceptin cantilevers), because usuallyF , b F , .
13. The value of nminwas given in EC4: 1985 as 0.5. There was a limiting span
of 20 m, and ductile connectors were defined as headed studs in concrete with
characteristic cylinder strength,f,, > 30 N/mm*, or friction-grip bolts. These rules
were not closely related to the limited research information then available (e.g.
references 1&14), in which there were no tests on beams of span exceeding 9.75 m.
The 20 m limit was in some cases unsafe,and the 30 N/mm2 limit was too crude a
recognition of the lower strain capacityof stronger concrete.
14. In BS 5950, the limit nmin= 0.5 was reduced to 0.4, mainly to offset the
effects of changes (in comparison with C P 117) in effective width of slabs and of
design compressive stressin concrete; and therule L > 20 m was replaced by

nmin = ( L - 6)/10
where L is the span in metres.Thischange was basedmainly on preliminary
results from the work Delivered by ICEVirtualLibrary.com
now reported, to: and 1987.
obtained between 1985
682 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
15. The 1990 draft of EC4 was influenced both by the present work, which
relatesmainly to solidslabs, and by recentwork by Aribert and other^,''-^'^
in which composite floors were alsostudied.Aribertswork and the rules in
EC4: 1990 are considered later.

Mathematical models
16. The earliest studies of this subject (e.g. reference 11) used elastic theory and
a linear load-slip ( P - S ) relationship for the connectors, which gives unreliable
indications of ultimate behaviour. Yam and Chapman used exponential P-S
curves and modelled three types of shear connector, but the most flexible of these
had an ultimate slip of only 2.5 mm. Some studies were concerned mainly with
serviceability~2*22 or with developing and validating elasto-plastic analysis prog-
r a m ~ . ~ ~ - ~ ~
17. Since the mid 198Os, uplift as well as slip has been included in non-linear
partial-interaction analyses. Two of these s t ~ d i e s used ~ ~ P-S character-
~ ~ *linear
istics only.Aribert and Aziz usedtransfer matrixmethodsand non-linear
analyses that are the most comprehensive yet available.Theystudied widely-
spaced connectors and loads suspended from the steel beam, and found, as did
Robinson and Narraine, that uplift has negligible effect on the distribution of
slip along thebeam. Where loadsare applied to the slab, typical values near failure
are anuplift of 0.04 mmand a 3%increase of slip.
18. Uplift is not included in the present model. This is equivalent to neglecting
shearstrainsinthesteelbeamandconcreteslab.Thesestrainscanincrease
deflexions in service by about 10%, but havelittleinfluence on behaviour at
ultimate load of beams with above average span/depth ratios.
19. The objective of the computations now described was to determinethe
maximum longitudinal slip smaX in a beam subjected to its design ultimate load,
when itsmaterials and shear connectors had the defects that are modelled in
design by the use of the partial safety factors y,; and to do this for a range of
degrees of shear connection n. The design ultimate load varies with n, and also
depends onwhether the equilibriummethod or the interpolation method is used.

Properties of materials and connectors


20. The stress-strain curve for concrete in compression was the fourth-degree
polynomial used by Basu and Sommerville,26but including a partial safety factor
y, = 1.5

ymfcJck = - 1.865(~/~,)O
2.41(~/~,) +
~ .~~(E/E
-, 0) ~
.045(~/~,)~ (5)
This agrees closely with other data and reproduces a long falling branch. The
value of E,,thestrain at maximum stress, was taken as-0.002or-0.0025,
depending on thecylinderstrength f c k , andthe crushingstraingenerallyas
-0.0035. The effect of giving concrete a tensile strength of O.lfck was found to be
negligible.
21. For structural steel,thestress-straincurve was bilinear in both tension
and compression, with design yield strength f, (i.e. with y, = 1.0) and elastic
modulus E , = 205 kN/mm2. For convenienceinnumericalanalysis,the yield
plateau and strain-hardening were
Delivered both represented by ato:single line of appropriate
by ICEVirtualLibrary.com
IP: 137.108.145.45 683
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
slope,chosen in therange E,/70-Ea/210. The yield strength was taken as 355
N/mmz (grade 50) unless noted otherwise. Similar assumptions were made for
reinforcement in tension, but withf, = 460 N/mmz and y, = 1.15. For all types of
steel, failure was assumed to occur if I e I reached 0.04.
22. Exponential load-slip curves were used for shear connectors

with failure assumed to occur when the maximum slip anywhere in the beam
B, and S, is
reached an assumed slip capacity S,. The choice of values for P , , G(,
described later.

Program EPPIB
23. The analysis of a simply-supported beam (Fig. 2) is explained first. The
cross-section, span, materials, and type of loading are chosen, and the beamitsand
shear connection are designed, either to EC4: 1985 or to BS 5950: Part 3, section
3.1. Checks are made to ensure that stresses and deflexions at the serviceability
limit state are not excessive, and that the ratio of self-weight to imposed load is
realistic. The subsequent analysis is done for the design load at the ultimate limit
state. Propped construction is assumed, to give the maximum longitudinal shear.
The external bending momentM is a known functionof X.
24. The strain distribution at any cross-section is defined by e,, e, (extension
positive), and the sagging curvature 4, as shown in Fig. 2(b). Any compression in
concrete within the depthh,, of profiled sheeting is ignored, and the line of action

1, 2, 3, etc. TIC
L p f h

(a)

+l- dx 4
Elevation Stram Internal forces
(b)

Fig. 2. Mathematical Delivered


model forbysimply-supported beamto:
ICEVirtualLibrary.com
684 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
of the force F , in the slab (compression negative) is assumed to be at depth h,/2
below its top surface. For equal curvature of the slab and thesteel beam, compat-
ibility gives
ds
- = E, - E, - z4
dx
where S is the longitudinal slip (Fig. 2(b)), and
z = i(ha + h,) + h,
For the internalforces and moments, equilibrium at
a cross-section gives
M - M=(&,,4)- M=(E,,9) - z F = 0 (84
F - Fa(&,,4)= 0 (8b)
Fa(&, 4)+
1 9 4) = 0 (84
For given (E,, E,, 4)and stress-strain
curves, unique values of M , , M , , F , and F,
can be computed. It
follows from equations (8) that
F = f , ( ,~E,,, 4) and M =f2(&, , E,, 4) (9)
A Newton-Raphson method of successive approximation was used to computeE,,
E, and 4 for each set of values F and M . Complex procedures were needed to
overcome problems of convergence. These, and fuller details of the program, are
given This core of the
computation, which is the program,
enables
equation (7) to be replaced by
ds
- = f(F, M )
dx
For shear connectors at longitudinal spacing
p, the rateof shear transferis
d F P(s)
-
dx p
where the function P(s) is the assumed load-slip function. Differentiation of equa-
) use of equation (10)gives
tion ( l land
d2F 1 ddFp 1 dP

which is the fundamental equation to be solved, with boundary conditions F = 0


a t x = O a n d x = L.
Outline of method of solution
25. The method is essentially forward integration, using equations (9H11) in
finite-difference form.Thelength of thebeam is divided into n - l elements,
defined by nodes 1 to n. For given loading and shear connection,M , p and dpldx
are known at each node. Theslip at node 1 is guessed, S, say. As F, = 0, equation
(l 1) gives F , ; E , , E, and 4 can be computed at node 2 ; and equation (10) gives s2.
The process is continued to node n. The value F , is compared with the correct
value (zero), the initial guess S, is modified, and iteration continues until s1 is
found such thatF , = 0.Delivered by ICEVirtualLibrary.com to:
IP: 137.108.145.45 685
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
26. For a beam with two continuous spans, AB and BC, there is an additional
loop of iteration. The function M(x) is based initially on a value of the hogging
bending moment at support B found by elastic analysis. Analysis proceeds as
above, assuming F = 0 at support A, until a solution is found with F = 0 at
support C. Thedeflexion at B (6,) is then compared with zero, the moment at B is
modified, and iteration continues until M , is found such that 6, = 0. When F(x)
and M(x) are known, the algorithm represented by equation (10) enables s(x) and
hence S, to be found.

Properties of shear connectors


27. Headed studs are the only type of shear connector whose properties have
been thoroughly researched, and this onlyfor their use in solid slabs. The load-slip
curves used in the analyses were deduced from data on studs, as follows; but the
results of this research are applicable to beams with any type of connector with
load-slip properties similarto those used. The relevant properties are
(a) the ultimate shear strengthper connector P,
( b ) the slip at maximum load (in a push test) S,
slip S, at load 0.5 P ,
(c) the flexibility in service, here taken as the
( d ) the slip capacity S,, here taken as the slip when the load has fallen to
0.95 P , after reaching its peakvalue.
28. There are no standard definitions of S, and S,, and the extent to which
falling-branch behaviour can be relied on in design has not been established. The
use of 0.95 P , here is based on redistribution between studs that hasbeen observed
in tests on beams with solid slabs, before maximumload was reached. More
redistribution may be expected in composite slabs, which is why Mottram and
Johnson reported' slips until the load fellbelow 0.8 P,. However, the use in
numerical analyses of falling branches extending to orbelow 0.8 P, has not, to the
Authors' knowledge, been validated by testson long-span composite beams.
Studs in solid slabs
29. Ultimate strength. This is theonlyproperty for which thetest data are
sufficient for the calculation of characteristic values. In such work, it is necessary
to exclude many low results, attributable mainly to premature splitting, that arise
from the typeof push test used.
30. Research has shown that higher strengths are obtained with wider slabs
and more realistic reinforcement than in the standard test, and that specimens
shouldhaveconnectorsattwo or more different locationsalongthe steel
member.'.''
31. Analyses of the behaviour of studs in 110 push tests of several types led to
an expressionz' for the characteristic strength per stud of N studs in a shear span
of a composite beam

It is for studs with shank diameter d , ultimate tensile strength fu, and elastic
modulus E,, in concrete of cube strength f,, and short-term modulus E,. This
relationship is shown in Fig.3 for d = 19 mm,f, = 450 N/mmz, N + CO, assuming
the conversion fromf,, to cylinder strengthf,, given in EC4: 1990, and using E ,
from that code. Values of P , for N = 10, for example, are 8% lower. The data for
fc,> 35 N/mm2 wereDelivered
sparse and more variable than forto:
by ICEVirtualLibrary.com lower concrete strengths.
686 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR C O N N E C T I O N I N C O M P O S I T E BEAMS

t /
,Oehlers et al.,N - X

ES 5950

80

fck. Nirnm2
Fig. 3. Characteristic and design strengths of 19 mm stud connectors

32. Values of P, given in two drafts of EC4 and in BS 5950 are also shown.
Line AB is high because it was intended for use with y m = 1.50 (since reduced to
1.25), and line BC was based on results from a type of push test more favourable
than the tests that provided many of the other data. The other threecurves for P ,
have a range notexceeding about 15%, which for this subject is good agreement.
33. Three sets of values for the design shear strengths P, are also shown, based
on values of y,,, given in the codes. The values from EC4: 1985 (lines DE, EF)were
used in thepresentwork, except in a separatestudy of the influence of this
assumption, inwhich the much highervalues ABC were used.
34. Stijiness. Measurements on studs in beams have shown that their stiffness
is on average similar to their stiffness in push tests, but there is wide scatter in both
sets of data.30 Several analyses of large numbers of push-test results have been
done, leading to mean curves of P / P , against either slip or a non-dimensional
group proportional to slip. Those from M e n ~ i e s , ~ B ~ t t r yand
, ~ Oehlers and
C ~ u g h l a nare
~~shown in Fig. 4. The first two show greater stiffness than those
from reference 33, mainly because the slabs were narrow, so the results included
failures by splitting, which occur at lower loads and slips than failures relevant to
studs in beams; but alsobecause, on average, the concreteswere stronger.
35. Ultimate slip and slip capacity. Until recently, very few records were kept of
ultimate slip, and even fewer of slip capacity, however defined. Most of the pre-
1985 data have been analysed ~ t a t i s t i c a l l yleading
,~~ to the following results for
normal-density concretes with 20 <fc, < 70 in N/mm2 units, and studs of diam-
eter d in mm
mean ultimate slip: mm S, = (0.389 - 0.0023fC,)d
Delivered
mean slip by mm
capacity: ICEVirtualLibrary.com to:
S, = (0.453 - 0~0018fc,)d
IP: 137.108.145.45 687
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA

---___ Menzles
0
Buttry

Oehlers and Coughlan

0
- - - Equation (6),curve A
0 0

- - - _Equation
- (6). curve B
--- -Equatlon (6). curve C
0
-- - - -Equation (6),curve D (Aribert)

I I I I I
W
1 2 3 4 5
Sl~p:mm
Fig. 4 . Load-slip curves for 19 mm stud connectors

36. About half of the specimens considered failedby splittingand were


excluded. The mean slip capacities are plotted in Fig. 5 for the range of concrete
cylinder strengths used in the numerical analyses, 20-40 N/mm2. Subsequent work
(e.g. reference 7) fits the results for slip capacity quite well. The three values used in
the present work, shownby points on Fig. 5, are
S, = 4 . 0 mm for studs with d 13 mm
S, = 6.0 mm for studs with d = 16 mm
S, = 8.5 mm for studs with d = 22 mm
Results for 19 mm studs can be obtained by interpolation between the results for
16 mm and 22 mm studs, and correspond to S, = 7.25 mm (circle in Fig. 5).

Studs in composite slabs


37. The widespread use of profiled steel sheeting adds manynew parameters to
an already complex subject. The properties of a shear connection are found to
depend on
( a ) the shapeof the trough
Delivered the sheeting
byofICEVirtualLibrary.com to:
688 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS

/

/
/
e/

f,, = 20 Nn
r/,*\;

fck = 40 Nimm
a

a
/
-
13 16 19 22
S h a n k dlameter: mm
Fig. 5 . Slip capucity of stud connectors

(b) the position of the stud in the trough, which may have to be off-centre
because of a stiffening rib
( c ) the directionof the spanof the sheetingrelative to thatof the beam
(d) the use or non-use of through-deck welding.
38. In codes, the strength of a stud in a composite slab is given as the product
of its strength in a solid slab and a factor k , which for many situations has its
maximum value of 1.0. However, for off-centre studs, pairs of studs, or narrow
deep ribs, it can have a value less than 0.7.
39. Here,onlysituations with k = 1.0 areconsidered. Even in these slabs,
failure is usually by deformation of a rib and its encasingsheeting, not by shearing
through the stud or its weld; while in solid slabs, stud shear is almost the only
possible failure mode, unless the stud is close to a free edge, as in an L-beam or a
deep haunch,where splitting canoccur. A review of test data shows that ribs can
fail in many different ways. Some studs arefound to be bent in double curvature,
some only near the weld, and some not all. at
40. It is now possible to predict kP, for individualstuds in troughs of the
widely-used profiles, but not to predict slip capacities. These are generally much
higher than in solid slabs, except where two or more studs are placed in one
trough. The influence of off-centre positions is shown by a simply-supported beam
of span 10 m tested at the University of Warwick in 1990. It and the loadingwere
symmetrical about midspan, except that the single studs were in the same half of
each trough, over the whole span. No stud failed, and near the end of the test the
slip was 16 mm at one endof the beam when it was 4 mm at the other,34because
the shear connectors in the favourable side of the troughs (i.e. in one half of the
span) were stiffer, as well as stronger.
41. The slip at P / P , = 0.5 has been found for studs in three types of profiled
sheeting to be similar to that of the same studsin solid slabs; but thatis unlikely
to be true for some profiles now available, which have rib depth/breadth ratios
exceeding 1.0. In the parametric studies, the same load-slip curves were used for
composite slabs as for solid slabs. This is probably conservative for all composite
Delivered
slabs, and certainly for those withby ICEVirtualLibrary.com
single 19 mm studs at to:
the centres
of open-profile
IP: 137.108.145.45 689
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
troughs with ratio of depth to mean breadth exceeding 1.0, forwhich the slip
capacity canexceed 10 mm.

Load-slip curves used in parametric study


42. As explained earlier,all curves were of the form
P = P,(I - e-flsp (6)bis
with slip S in mm units and the asymptotic strength P , taken as the design value
given by either BS 5950 or EC4:1985. For most of the analyses, one or otherof the
curves labelled A and B in Fig. 4 were used, for which
a = 0.558 and p = 1.0 mm ~ for curve A
a = 0.989 and p = 1.535 mm for curve B
Three other curves were used for a few analyses. All five had slips between 0.3 mm
and 0.5 mm atPIP, = 0.50, and between 3 mm and 6 mm atPIP, = 0.99. For two
beams with 50% shear connection, a check was made on theinfluence of shape of
curve on the maximum slip, using six P-S curves, with shapes that range from
curve C in Fig. 4, to the curveof Oehlers and Coughlan. For both beams, doubling
the slip at PIP, = 0.50 increased the maximum slip at design ultimate load by less
than 5%. This increase rose to about 10% when the slip at PIP, = 0.99 was
doubled. Maximumslips are obviously notsensitive to the shapeof the P-S curve.

Validation of program EPPIB


43. In reality, transfer of shearoccursonly attheconnectors;butinthis
program, 'continuous' transferwas used, by spreading theeffect of each connector
over a length equal to the connector spacing. A program using discrete transfer
was also written, to enable the effects of this assumption tobe studied. Six different
layouts of studs and elements and spacings of studs were analysed for each of three
different beams, by both methods, and five outputs from each pair of analyses were
compared.'' The only discrepancy exceeding a few per cent was for the most
widely spaced studs (600 mm), where S, with continuous connection was about
19% higher than thatfor discrete connectionwith the end stud above a support.
44. Detailedcomparisons were alsomade between computed results (using
measured strengths of materials, and y,,, = 1.0) and the results of tests on ten
beams in four different laboratories. Typicalresults for distributions of slip along a
beam are shownin Fig. 6. This is for Aribert and Labib's beam P2,"j with dimen-
sions and loading shown. The ultimate loadswere 329 k N (predicted) and 375 kN
(test), and the degree of shear connection was about 70%. The computed values
here were based on the load-slip curve of equation (6), with a = 0.8 and p = 0.7
mm-'.
45. Here and in other comparisons, the computations were found to overesti-
mate slip over a distance of about 0.5 m adjacent to each end support. This is
thought to be on accountof shrinkage of concrete, notmodelled in these analyses.
The discrepancyis smaller for longer spans,and when profiled sheeting is used.
46. The objective of this work was to study slip in shear spans exceeding about
5 m. The only relevant tests on shear spans of this length known to the Authors
were three by Burkhardt,' in which the loading was so chosen that a 9 m length
simulated half of a beam of span 18 m with distributed loading. The degree of
shear connection wasDelivered
about 0.4
by in accordance with EC4:
ICEVirtualLibrary.com to: 1990, and failure of the
690 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS

A (Aribert) Computed
1.o- -__ - Experimental
-- -_- Computed (EPPIB)

\
E \
I I
E 0
6
Y)

360 kN

-1 .o -

Fig. 6. Longitudinal slip in beam P2, with two short spans

shear connection (13 mm studs) occurred in all three tests. The loading and deflex-
ions just before failureof beam T2 are shown in Fig. 7. Computed and measured
deflexions agree almost exactly, and there is close agreement between the values of
slip (Fig. 8). It is notable that thehighest value of slip recorded before stud failure,
in any of these tests, was only 2.4 mm, and that as soon as one stud failed, others
followed. This is attributable to the adverse combination of small studs, a solid
slab, and strongconcrete (f,,= 50 N/mm2).
47. The conclusion from the validation studies as a whole was that regardless
of the choice of load-slip curve and of the strain-hardening modulus for structural

2960 3000 3090 2940

I
17.9 kN 17.9 kN 123.7 kN I

1
Single studs at600-4 Palrs at226
5 10
0 1 I
X: m
\ /

f
Fig. 7 . Layout, loading, Delivered by ICEVirtualLibrary.com
and dejexions of beam T2, with oneto:long span
IP: 137.108.145.45 691
On: Thu, 13 Jun 2013 07:29:23
J O H N S O N AND M O L E N S T R A

Experiment

X Computed

X: m \

F i g . 8. Longitudinal slip in beam T 2

steel, the computed maximum slip for a beam at its design ultimate load was
generally a slight overestimate, with a maximum errorof 20%.

Parametric study of maximum slip


48. Themaximumslip (smaX) in a beam of given dimensions and materials,
when loaded to its design ultimate load, appears initially to be a function of about
20 independent variables. In fact, economic constraints and the design process
impose many and complex limitations on the combinationsof these variables that
need to be explored.
49. In a preliminary study of simply-supported beams, S, was found consis-
tently to increase when the value of any oneof four ratios was increased, the other
three being constant. These ratiosare
L wp - Wpa Ypl F,
h Wpa h, + hp F,
where L , h, h, and h, are as shown in Fig. 2; wP and wpa are the flexural failure
loads for the composite and steel members respectively; yp, is the depth of the
plastic neutral axis below the top of the slab; and F , and F , are the maximum
design longitudinal forces that can be resisted by the concrete and steel members
respectively (so that y,, > h, + h, when F , > F J . The designs studied were chosen
to be realistic; and most of them had ahigh valueof one or more of these ratios.
50. The procedure in each case was to select a cross-section, materials and a
span (or two spans for a continuous beam); to determine N , and find the design
loads for values N I N , in the range 0.5-1.0; and then to compute smarat each of
these valuesof N I N , .
51. Fifteen two-span beams were analysed. In their design, account was taken
of the usual loading Delivered
patterns, by
butICEVirtualLibrary.com
the minimum load on to: a span was always taken
692 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
as a value sufficient to prevent uplift at the end support. The longer span was
always taken as 10 m or more. It was found that distributed loading always gave
higher S,, than combinations of point and distributed loads. The longer span was
also analysed as simply-supported, with the same degree of shear connection (and
so a lower design ultimate load). It was found that S,, was always higher in the
simple span. The majority of the studies were therefore done on simply-supported
beams.

Results for simply-supported beams


52. Inthedesign of theshearconnection for eachbeam,theinterpolation
method was used.The equilibrium method gives higher values of smaX,as described
later. All loadings were uniformly distributed and the shear connectors were uni-
formly spaced along the beam. All cross-sections were in class 1 or class 2 to
EC4: 1990, and propped construction was assumed. As the shear connection was
modelled as continuous, it was not necessary to define the size or spacing of the
connectors; theirbehaviour was fully defined by the selected degree of shear
connection (in the range 0.5-1.0) and load-slip curve. Except where noted, the
results are for steel with a yield stressf, = 355 N/mmz, as S,, is then greater than
for weakersteel. Theother main parameters were studiedoverthefollowing
ranges
cylinderstrength of concrete 20 N/mm2 to 40 N/mmz
span/depth
ratio L/h 18 to 26
slab
thickness h, + h, 120 mm to 180 mm.
The results are given separately for each of three typesof beam.
53. Hot-rolled l-sections with solid slabs. The 45 analyses gave values of S,,,at
design ultimate load for sets of values of the parameters above. From these, the
most onerous values of N I N , were deduced for particular values of S,,, and span.
These values are shown in Fig. 9(a).For a given N / N , , slip increases with span, as
expected, but the relationshipis much influencedby the other parameters.
54. The full lines shown on Fig. 9 are tentative design rules, one for each value
of slipcapacity, which are safe approximationstothecomputed results. For
example, for a beam of span 15 m, point A shows that 50% shear connection can
be used if the connectorshave an established slip capacityof at least 8.5 mm; butif
the slip capacity is only4 mm, 100% connection should be used (point B), because
in several beams with 90% connection,smax at ultimate load reached 4 mm (points
C, D, E).
55. Hot-rolledl-sectionswithcompositeslabs. Resultsfromanalyses of 28
beams, with depths h, of the profiled steel sheetsbetween 30 mm and 60 mm, led to
tentative design rules as shown in Fig. 9(a), exceptthat line FG was replaced by the
more restrictive line FH. For beams where the plastic neutral axis falls within the
slab, the presence or absence of profiled sheeting made little differenceto S,,, , for
thesame total thickness of slab. For neutral axes in thesteelbeam, smsrwas
slightly higher when the slabwas composite.
56. The greater slip capacity of connectors in composite slabs should more
than compensate for these differences,so Fig. 9(a) canbe assumed to be applicable
to bothsolid and composite slabs.by ICEVirtualLibrary.com to:
Delivered
IP: 137.108.145.45 693
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
_ _ - Equtlibrlurn method

I 0 4 mm
(where different)

0 6rnm
x 8mrn

/
/
- 0
A
I l I I
5 10 15 20
L: m
(a)

,
,B . ,'
/ "
/'/ 0

I 1 1 I I
5 10 15 20
L: rn
(b)

Fig. 9. Results ofparametric study: ( a )sections with 1.0 A , / A , < 1.5; (b)sections
with 1.5 A,/A, 3

57. Weldedl-sectionswithunequal.flanges. Thirty beamswithunequal steel


flanges were analysed. The ratio A J A , of flange areas (bottom/top) ranged from
1.0 to 3.2. The most adverse results, for both solid and composite slabs, are shown
in Fig. 9(b).The design rules shown aremore restrictive than those in Fig. 9(a).
58. It appears that application of the rules of Fig. 9(a) should be limited to
beams with flanges such that A , / A , < 1.5. For ratios up to 3.0, Fig. 9(b) shouldbe
used.
59. Influence of grade ofstructural steel. Four beams were re-analysed with the
yield strength of the structural steel reduced from 355 N/mm2 to 275 N/mmz, and
all other propertiesunaltered, using themethod of EC4: 1985 for the shear connec-
tion, as before. Results for beams SEBC8 and SEBC17 are shown in Fig. 10, and
properties of the beams are in Table 1. In the table, h , , h , , etc., are the horizontal
and vertical dimensions of the flange and the web of the rolled I-section, followed
by the type of sectionDelivered
and its mass in kg/m.
by ICEVirtualLibrary.com to:
694 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS

A
15 - 4 \
Interpolation \
method
--_ Equrlibrrum
method

10 -
E \ i,=355 Nlrnrn'
E l

v1

5-
\

f,=275 Nlrnm' i,=275 Nlmrn'

Fig. 10. Influences of design method and grade ofsteel on maximum slip: (a) beam
SEBC8; ( b )SEBCI7

60. The design ultimate loads wp, for NIN, = 1, and wpp(50), for N I N , = 0.5,
are for,f, = 355 N/mm2. Thevalues f0r.j; = 275 N/mmZwere about 20% lower.
61. Thereductions in S, at N I N , = 0.5rangefrom 33% to48%, for a
reduction inf, of 23%. It is likely that a conservative estimate of S,,, for any yield
strength less than 355 N/mmzcould be obtained by scaling downthe value
reported here in proportion to the yield strengths.
62. U s e of characteristic load-slip curve rather than design curve. Four beams
were re-analysed with the design strength P , of the 19 X 95 mm stud connectors
replaced by the characteristic strength P,, and other properties unaltered. This
was done using the equilibrium method of design, with properties of studs both
from EC4: 1985 and BS 5950. Except where lines AB and BC in Fig. 3 were used.
the ratioP J P , was 1.25.

Table 1. Typical properties of beams analysed using program E P P I B

Fl
I
Beam SEBCl EBC8 SRM23 SRM29
Span: m 10 20 20 20
h: m 2.5 3.0 4.0 3.0
h, : mm 84 85 I25 i 75
h,, : mm 46 46 60 40 60
h,,h,:mm 171,11.5 254,19.0 171, 11.5 266, 17.5 266. 17.5 266,17.5
b2,h2:mm 7.3, 333 12.4, 646 7.3, 333
12.9, 719 12.9, 719 12.9, 719
Section UB 51 UB140 UB147
UB 51 I UB147 UB147

l
,fck : N/mmz 30 25 35 25
: N:mm2 355 355 355 355 355 355
wp : kN/m 52.83 52.41 52.35 57.65 67.45 55.90
w,,(50): kNim 42.41by ICEVirtualLibrary.com
39.12 Delivered 43.08 51.2 to:
IP: 137.108.145.45
52.10 L
46.33

695
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND M O L E N S T R A
63. The results for two beams, EBC2 and EBC8, are shown in Fig. 11, and
properties of these beams are in Table 1. The design method of EC4: 1985 was
used. The BS 5950 method gave similar results.
64. For all the results, S,, using P , was less than 1i1.25 times S,, using P,, as
would be expected from what was in effect a 25% increase in the number of shear
connectors. The relevanceof these resultsis discussed later.
65. Use oftheequilibriummethod. This method was explained earlier and is
represented by curve ABC in Fig. l(a). For a given degree of shear connection, it
gives ahigherbendingresistance, and so a higherdesignload, than doesthe
interpolation method, so the maximum slip s,,~,~ is higher too. For 20 of the
simply-supported beams, this slip and the enhanced design load wde were deter-
mined for four values of NiN, each, and compared with the values ,,s and wdi
(interpolation method) which were the basis of the results so far reported. The
design method of EC4: 1985 was used. Ten beams were also analysed using the
equilibrium method of BS 5950. The results were similar.20
66. Theproportional increases of maximumslip, (S,,,,, - s,,)/s,, were
between five and ten times as great as the proportional increases in the loading,
(wde- wdi)/wdi.The increase in S,, was, on average, 60%, but individual values of
S, Jsmax ranged from 1.0 to 2.2.
67. The conclusions from these analyses are shown by the dashed lines in Fig.
9. For a given span and degree of shear connection, it appears that use of the
equilibrium method,ratherthantheinterpolationmethod, is equivalent to
assuming that the slip capacity of the connector is between 2 mm and 3 mm
greater than before. For a spanof 10 m and slipa capacity of 6 mm, it appears that
the minimum degree of shear connection should be increased from 0.5 (pointG in
Fig. 9) to 0.84.
Prediction of smax
f o r hot-rolled I-sections with solid slabs
68. Fora given degree of shearconnection, S,,, can be assumed to be a
function of the fourdimensionless groups identifiedearlier. In asearch for a
predictive function, multiple linear regressions were carried out, using both expo-

t P=P,

NIN,
Delivered
Fig. 11. Injuence of strength ofby ICEVirtualLibrary.com
shear to:
connectors on maximum slip
IP: 137.108.145.45
696
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
nential and polynomial functions of some or all of these groups, and the resultsof
computer analyses of 45 beams.
69. The best compromise between simplicity and accuracy was found to be the
expression

where
W,, = 8 M,,/L?
so = M,, Lh,/6E, I,
and M,, is the plastic moment of resistance of the steel beam, of depth h, and of
elastic flexural stiffnessE , I,. The expression for S,, is the elastic end slipwith zero
shear connection when the steel beam carries load wpa per unit length. The results
were
for N I N , = 0.5: c( = -0.13, B = 1.03; with c , = 4.9%, rz = 0.96
for N I N , = 0.75: z = -0.24, p = 1.70; with c, = 8.0%, r2 = 0.96
where c, is the coefficient of variation of the errors, S,,, - ;,S ,
r2 is the coeffi-
cient of multiple determination, r 2 = 1 - u,/u,; U, is the variance of the errors; and
U, is the variance of the computed values S,. These results are for the inter-
polation method. When the equilibrium method is used, a safe approximation to
smaxis obtained by replacing so in equation (15) by 1.5 so.

Comparison of results with those fromINSA, Rennes


70. Reference was made earlier to the work of Aribert and others at INSA,
Rennes. This team has completed parametric studies similar in scope to those
reported here." Although no direct comparisons can be made, it is clear that the
design recommendations from Rennes are less conservative than those shown in
Fig. 9. It has been found that this results from different choices of values for the
partial safety factors y,, as shown in Table 2.
71. The values used in the design of the beams influence the calculation of the
design ultimate load and the number of connectors required for full shear connec-
tion. The values used in the numerical simulations influence the stress-strain and
load-slip curves, and hence the computations of maximum slipat the design load.

Table 2. Values of y, used in parametric studies

Structural steel Shear connectors Concrete


~ ~~

Design Numerical Design Numerical Design Numerical


simulation simulation simulation
Research at Warwick 1.0 .o 1.5 1.5 1 1.25 1.25
Research at Rennes 1.0 1.o 1.o 1 .o 1.25 1.o
Delivered by ICEVirtualLibrary.com to:
IP: 137.108.145.45 697
On: Thu, 13 Jun 2013 07:29:23
J O H N S O N A N D MOLENSTRA
72. Most of thebeamshavebeenanalysed for saggingbending, with the
neutral axis in the concrete slab. In design, a change in y, for concrete from 1.5 to
1.0 reduces the depth of the compressive stress block, but hardly alters the lever
arm,the plasticmoment of resistance, orthe designload.Both groups used
y, = 1.25 for design of the shear connection, so for a given beam, the numbers N ,
for full shear connection were the same. But in the numerical simulation, it was
assumed at Warwick that the ordinates of the load-slip curves were only 80% of
the values assumed at Rennes. Fig. 4 shows that at high loads this can more than
double theslip at agiven load per connector.
73. As reportedpreviously,a few simulations were doneat Warwick with
y, = 1.0 for connectors. Fig. 11 shows that the resulting maximum slips are much
lower. These results are comparable with those from Rennes, for beams where the
plastic neutral axis remains within the depth of the slab, so that N , is not influ-
enced by ym . For some beams where the plastic neutral axis for full shear connec-
tion falls within the steel member, the values of maximum slip at low degrees of
shear connection are significantly higher than those obtained at Rennes. This is
because the difference in 7, for concrete (Table 2) results in the use of fewer shear
connectors. It follows that if the shear connections at Rennes had been designed
(as well as simulated) with y, = 1.0, the computed results (S,,, at given n and L )
from Rennes and Warwick would havebeen almost identical for most beams.
74. This has been checked, for the simply-supported beam of span 10 m, with
distributed loading, defined in the left-hand half of Fig. 6 in reference 19. The
results from that figure are reproduced in Fig. 12, with three sets computed using
program EPPIB. They are for a limiting slipS,, = 5.0 mm. The followingdata were

X Warwlck: y m a5 at Rennes

+ Warwlck. 7 , a5 Table 2

0 Warwtck;all v,,,= 1-0

0.6
0.8 0.4 0-6 1 .o
NIN,

Fig. 12. Computed results atS, by=


Delivered 5 mm for a beam of span
ICEVirtualLibrary.com to: 10 m
698 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
S H E A R CONNECTION IN COMPOSITE BEAMS
used in all calculations
E , = 210 kN/mm2,f, = 30 N/mm2,f, = 360 N/mmz
For the shear connectors

2.
in kN, mm,units. All values of y, were taken as given in the second row in Table
75. The agreement is very close between the results from Rennes, and those
computed using EPPIB and the same values of ym. The small differences are attrib-
uted to
(a) differences in the assumed shapes for the stress-strain curves for steel and
concrete
(b) modelling of shear connection as discrete at Rennes, but as continuous at
Warwick.
The analyses were then repeated, with values y, as in the first row of Table 2. The
results (Fig. 12) show lower ratios of load W to load with full shear connection,
W,, , because smax at a given load is higher than before. These results are as would
be expected from Fig. 11.
76. It is concluded that the results from Warwick are based on a more consis-
tent design philosophy than the results from Rennes, because the same values y,
were used for both design and analysis. It is not claimed that use of y, = 1.0
throughout work of this type is incorrect. If that had been done consistently, the
results from Rennes would have been very close to those from Warwick reported
above, at least for cross-sections with the plastic neutral axis within the concrete
slab. This is shown in Fig. 12, for 70% and 90% shear connection.
77. The preceding comparison relates to the choice of some degree of shear
connection, n, say, indicated by point B in Fig. l(a), below which the curve ABC
may not be used in design. Usually, n, will be higher than the general minimum
nmin(which is 0.4 in draft Eurocode 4 and in BS 5950, whereas 0.2 is advocated in
reference 19).
78. In BS 5950, no design is allowed with n n,; but the results from Rennes
haveshown thata designwith nmin n n, is safe provided thatthe bending
resistance is calculated using line BE, not curve BC. This useful result has been
included in the draft Eurocode. The choice of nB is, of course, crucial for this
method, and should ideally depend on the proportions of the cross-section of a
beam, as well as itsspan and theductility of its shear connectors.

Design methods
79. This section relates to the equilibrium method, rather than the more con-
servative interpolation method, because it fits the computed results more closely.
The recommendations from Warwick and from Rennes, for rolled sections with or
without profiled sheeting,and for designby the equilibrium method, are compared
in Fig. 13. They are for simply-supported or continuous spans, but not for cantile-
vers.
80. Both sets of lines give minimum values for the degree of shear connection,
for given slipcapacities and spans.Molenstrasconclusions on slipcapacity
range from 4 mm for 13Delivered
mm studs,
byto 8.5 mm for 22 mm studs.
ICEVirtualLibrary.com to: Aribertc o n ~ l u d e s ~
IP: 137.108.145.45 699
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA

1.0 - /- - -----
/
- / 1 .O at 25 m
/ - - - -

--- From Rennes


------ From Warwick
Draft Eurocode 4

Span: m
Fig. 13. Design rules for point B in Fig. 1,f o r cross-sections with 1.0 A J A , < 1.5

that for 22 mm or smaller studs in concrete withf, 40 N/mm2, the line labelled
5 mm (Fig. 13) is appropriate.
81. In the October 1990 draft of Eurocode 4, headedstud connectorsare
assumed to be ductile (a condition for uniform spacing of connectors within a
shear span) if the relationship between N I N , and span lies above the continuous
lineinFig. 13, for beamswith equal steelflanges. Otherconnectorsmay be
assumed to be as ductile as studsif their characteristic slip capacity (at the charac-
teristic resistance) is not less than 6 mm. This design rule isa rough average of the
conclusions from Warwick and from Rennes, because it lies between the 8 . 5 mm
line from Warwick and the 5 mm line from Rennes. A more onerous rule is given
for beams with unequal steel flanges. In draftEurocode 4,some account is taken of
the preceding statements that the results from Rennes are considered to be uncon-
servative and that comparisons with tests have found that the maximum slips
computed at Warwick are generally slight overestimates.
82. The recommendations from Warwick in Fig. 13 are also conservative for
most beams because theyare based on predictions of S, as a function of only two
parameters, N I N , andspan.Toobtain lowscatter, it isnecessary to usesix
parameters in three groups (equation (15)).A typical regression using only two of
these three groups, so and Llh, based on results from 45 beams with hot-rolled
steel sections and solid slabs, gives errors as large as loo%, with c, > 30% and
r z = 0.24. (The notationis defined below equation (15).) The recommendations on
(NIN,), are an upper bound to the computed results(Fig. 9) and so are very
conservative for some beams.
83. The less conservative rules of draft Eurocode 4 are based essentially on
judgement of what is a realistic design. The beams SRM23 and SRM29 (Table l )
provide an example of this. The analysis of SRM23 led to point J on Fig. 9(a), a
result well on the safe side of the nearby line labelled 8.5 mm.Beam SRM29 had a
narrower and thinnerDelivered
slabof weaker concrete, and, in consequence,
by ICEVirtualLibrary.com to: alower design
700 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
load for N I N , = 0.55, the value at point J. Its maximum slip at this load was 12
mm, double the value for SRM23 at a higher load. This result was ignored when
drawing the 8.5 mm line, in the belief that this particular combination of param-
eters was unlikelyinpractice, and with theknowledge that a connector with
S, = 8.5 mm in a solid slab may well have a slip capacity of 12 mm in profiled
sheeting 60 mm deep.
84. Thus, theseconclusions are based on researchinterpretedusingjudge-
ment. Further validation of them by well-controlled tests will be most welcome,
particularly on simply-supported beams with span exceeding 18 m, solid concrete
slabs, and N I N , < 1.

Conclusions
85. These conclusions relate to composite beams of steel and concrete with
steel elements in compression that are in class 1 or class 2 (plastic or compact),
subjected to their design ultimate load, when that load is governed by flexural
failure of one or more cross-sections. It is assumed that the steel members are not
fully encased in concrete, and that failure does not occur by lateral buckling or in
vertical shear.
86. The conclusions are based on computations of maximum slip at the steel-
concrete interface for over 160 beams, on a studyof equally wide-ranging work in
France, and on reviews of data from tests on composite beams and on stud shear
connectors. The degrees of shear connection studied ranged from 100% down to
50% in the present work and to20% in France.
87. The proposed designrules(Fig. 13) are based onthecriterion that the
computed maximum slip(for connectors with an exponentialload-slipcurve)
should not exceed the slip capacity of the connectors tobe used, defined as theslip
at which the shear resistedby the connectorfalls to 5% below its peakvalue.
88. The meanslipcapacities of studconnectors in solidslabs (i.e. those
without profiled sheeting) are as plotted in Fig. 5. The value for 19 mm studs is
about 7 mm. These results are from tests such that splitting of the slab did not
occur, which excludes about half of recorded push-test data. There are too few
results for thedetermination of characteristic slip capacities.
89. Slip capacities of single studs at the centres of troughs in profiled sheeting
depend on the profile. The failure is essentially that of theconcreterib, with
yielding of the stud, and theslip capacity usually exceeds that for a solid slab. For
off-centre studs and more than one studper trough, slip capacitymay be lower. If
one of these arrangements is essential in practice, then slip capacities should be
determined.
90. Computedmaximum slips arenotmuch influenced by changesinthe
assumed shapes of the stress-strain curves for the materialsor the load-slip curves
for the connectors. Load-slip curves with falling branches were studied in France,
but notin the present work.
91. The modelling of shear transfer as continuous or discrete, and the inclu-
sion or omission of uplift and shear strains, have little influence on the computed
results.
92. Maximum slip is greater,in a beam of given span, when thatspan is
simply-supported, rather than continuous; and when the composite flange of the
steel beam is smaller than the otherflange.
93. Computed maximum slips are based mainly on the use of the interpolation
method of design. For Delivered
20 of the by ICEVirtualLibrary.com
beams, to:
the less conservative equilibrium method
IP: 137.108.145.45
701
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
was also used. The design rules are based on the use of the latter method, and are
for structural steelwith a yield strength of 355 N/mm2.For weakersteels,
maximum slip is roughlyproportional toyield strength.
94. Predictions of maximum slip using program EPPIB agree closely with those
from the program used at INSA at Rennes, in France, when the same sets of
partial safety factorsy, are used. Almost the same resultsare obtainedwhen all ym
values are taken as 1.0, as when specified values for the ultimate limit state are
used.
95. The design rules deduced from the analyses at Rennes are unconservative
because of inconsistency in the values assumed fory, . The rules deduced from the
present work may be conservative in most instances, because the maximum slipis
assumed to be a function of onlytwo parameters,spanand degree of shear
connection. Predictions based only on these are inevitably a gross simplification.
Maximum slip in fact depends on at least four non-dimensional ratios, in a way
that has been extensively studied by regression analyses, from which one set of
results is given.
96. The design rules being proposed in Eurocode 4 : Part 1, are based on both
the present work and the research at Rennes, but also on exercise of judgement in
excluding adverse results computed for combinations of parameters thought tobe
unlikely in practice. The rules give the value of nB in Fig. 1. The shape of curve
ABC is such that the bending resistance of a member is not sensitive to uncertainty
in the value of nB. Even so, further experimental validations for long spans andfor
low degreesof shear connection wouldbe useful.

Acknowledgements
97. Resources for this research were provided by the University of Warwick,
with
contributions
from
the
National Foundation for Scientific Research,
Belgium, and the Building Research Establishment,UK.
98. The value of discussions with Professor J. M. Aribert, Professor K. H. Roik
and Professor J. W. Stark is acknowledged with thanks.

References
1. BURKHARDT P. Le comportement klastique et plastique des poutres mixtes. Federal Poly-
technic of Lausanne, 1977,PhD thesis 264.
2. JOHNSON R. P. Limitations to theuse of partial shearconnection in compositebeams.
University of Warwick, 1986, Research Report CE21.
3. BRITISH STANDARDS INsnmnoN.Composite construction in structural steel and concrete:
simply-supported beams in buildings.BSI, London, 1965, CP 117: Part 1.
4. BRITISH STANDARDS INSTITUTION. Code of practice for design of composite beams. BSI,
London, 1990, BS5950: Part 3, section 3.1.
5. COMMISSION OF THE EUROPEAN COMMUNITIES. Eurocode 4. Common unijied rulesfor com-
posite steel and concrete structures, Draft for national comment. Commission of the
European Communities, Brussels, 1985, ReportEUR 9886 EN.
6. COMMISSION OF THE EUROPEAN COMMUNITIES. Eurocode 4 : Part 1. Design of composite
steel and concrete structures: general rules and rules for buildings. Editorial Panel for
Eurocode 4, Oct. 1990, Revised draft, Issue1.
7. ROIK K-H.and HANSWILLE G . Beitrag zur Bestimmung der Tragfahigkeit von Kopfbol-
zendiibeln. Der Stahlbau, 1983,10,301-308.
8. JOHNSON R.P. and OEHLERS D. J. Analysis and designfor longitudinal shear incomposite
T-beams. Proc. Instn Ciu. Engrs,
Delivered Part 2, 1981,71, Dec.,989-1021.
by ICEVirtualLibrary.com to:
702 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23
SHEAR CONNECTION IN COMPOSITE BEAMS
9. MOTTRAM T. J. and JOHNSON R. P. Push tests on studs welded through steel sheeting.
Struct. Engr, 1990,68, No. 10, May, 187-193.
10. YAML. C. P. and CHAPMAN J. C. The inelastic behaviour of continuous composite beams
of steel and concrete.Proc. Instn Ciu. Engrs, Part 2, 1972,53, Dec., 487-501.
11. NEWARK N. M. et al. Tests and analysisof composite beams with incomplete interaction.
Proc. Societyfor Experimental Stress Analysis, 1951,9, No. 1,75-92.
12. BRASINCA H. E.and STARK J. W. B. Beproeving van liggers op verkleinde schaal.TNO
Institute for Building Materials and Construction, Delft, 1972, Nov.,IBBCReport
B1-71-52.
13. JOHNSONR. P. and MAY I. M. Partial-interaction design ofcompositebeams. Struct.
Engr, 1975,53, No. 8, Aug., 305-31 1.
14. VOGELG. Shear connection behauiourfrom beam tests. University of Missouri, Columbia,
Jan. 1971, MSc thesis.
15. ARIBERT J. M. and AZIZA. K. Modele general pour le calcul des poutres mixtes hyper-
statiques jusqua la ruine. Construction Metallique, 1986, No. 4 , 3 4 1 .
16. ARIBERT J.M. et al. Etude numerique et experimentale de Iinfluence dune connection
partielle sur le comportement de poutres mixtes.Journees. Association Franqaise des
Ponts et Charpentes,1983, Mar., 68-89.
17. ARIBERT J. M.Etudecritiquepourvoienumeriquedelamkthodeproposeedans
1Eurocode 4 pour le dimensionnement des poutres mixtes acier-beton a connection
partielle. Construction Metallique, 1988, No. 1, Mar., 3-26.
18. ARIBERT J. M. and AL BITAR A. Optimisation du dimensionnement en connection parti-
elk de poutres de planchers mixtes realises avec un bac de t61e mince nervuree. Con-
struction Metallique, 1989, No. 4, Dec., 4-33.
19. ARIBERT J. M. Dimensionnement de poutres mixtes en connection partielle, in Mixed
Structures including New Materials, IABSE Symposium, Brussels. Int. Association for
Bridge and Structural Engineering,Reports, 1990,60,215-220.
20. MOLENSTRA N. Ultimate limit states of compositebeams. University of Warwick, Oct.
1990, PhD thesis.
21. PLUM D. R. and HORNE M.The R. analysisof continuous composite beams with partial
interaction. Proc. Instn Ciu. Engrs, Part 2,1975,59, Dec., 625-643.
22. KRISTEKV. and STUDNICKA J. Analysis of composite girders with deformable connectors.
Proc. lnstn Ciu.Engrs, Part 2, 1982,73, Dec., 699-712.
23. ARIZUMIY. et al. Elastic-plastic analysis ofcompositebeamswithincompleteinter-
action by thefiniteelementmethod. Computers and Structures, 1981,14, No. 5 4 ,
453462.
24. ROBERTS T. M. Finite difference analysis of composite beams with partial interaction.
Computers and Structures, 1985,21, No. 3,469473.
25. ROBINSON H. and NARAINE K. S. (BUCKNER C. D. and VIESTI. M. (eds)). Slip and uplift
effects in composite beams. Composite construction in steel and concrete. ASCE, New
York, 1988,487-497.
26. BASUA. K. and SOMMERVILLE W. Derivation of formulae for the design of rectangular
composite columns. Proc. Instn Ciu. Engrs, Suppl. 1969,233-280.
27. KABAILA A. Discussion on: Equation for the stress-strain curve of concrete, by Desayi
D. and KrishnanS. Proc. Am. Concr. Inst.,1964,61, No. 9, Sept., 1227-1229.
28. MOLENSTRAN.and JOHNSONR. P. Problems of convergenceinNewton-Raphson
analyses of composite beams with partial shear connection.
Paper in preparation.
29. OEHLERS D. J. and JOHNSON R. P. The strength of stud shear connections in composite
beams. Struct. Engr, 1987,65B, June, 4 4 4 8 .
30. MCGARRAUGH J. B. and BALDWINW.J. Lightweight concrete-on-steel composite beams.
Engng J . Am. Inst.Steel Construction, 1971,8,9&98, July.
31. MENZIES J. B. CP117 and shear connectors in steel-concrete composite beams. Struct.
Engr, 1971,49, No. 3, Mar., 137-153.
32. BUTTRYK. E. Behaviour of stud shear connectors in lightweight and normal-weight con-
crete. University of Delivered
Missouri, Columbia, 1965, MSc thesis.
by ICEVirtualLibrary.com to:
IP: 137.108.145.45 703
On: Thu, 13 Jun 2013 07:29:23
JOHNSON AND MOLENSTRA
33. OEHLERS D. J. and COUGHLAN C. G. The shear stiffness of stud connectors in composite
beams. J . Constr. Steel Res., 1986,6, Oct.,273-284.
34. LAWSONR. M. et al. Testsoncomposite beamswithlargeweb openings. To be
published, Struct. Engr, 1991.

Delivered by ICEVirtualLibrary.com to:


704 IP: 137.108.145.45
On: Thu, 13 Jun 2013 07:29:23

Вам также может понравиться