SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:58 Page 32

PAPER

Evening meeting
To be held at IStructE, 11 Upper Belgrave Street London SW1X 8BH, on 10 June 2008 at 18.00h

Shear design of concrete structures

Synopsis 1962 a committee of senior American engineers concluded2 that Michael P.

Deficiencies in the shear design of concrete structures are for shear a ‘fully rational design approach does not seem possi- Collins
inherently more dangerous than deficiencies in flexural design ble at this time.’ As an interim measure they proposed a semi-
BASc, MASc,PhD,
because shear failures can occur with no prior warning and empirical shear design procedure where the form of the
FRSC, PEng
with no possibility for redistribution of internal forces. While expression was based on a logical hypothesis but the coefficients
University Professor and
assessing the shear capacity of a reinforced concrete structure were found by curve fitting to available experimental results. Bahen-Tanenbaum
accurately is critically important for public safety, the traditional Their interim equations still form the basis of the current Professor, University of
techniques available for this task are open to dispute. This American Concrete Institute (ACI-3183) shear provisions. Shear Toronto
paper will summarise recent research aimed at understanding design procedures in current European codes (e.g. EC 24 and BS
the basic mechanisms of shear transfer and will introduce 81105) are even more heavily dependent on empirical curve fitting Denis
simple, rational, general and accurate procedures for but they do have the advantage of being formulated somewhat
evaluating shear capacity. later when more experimental results were available.
Mitchell
As an illustration of the inconsistency of some current shear PhD, FRSC, PEng
Introduction design provisions, consider the results of four slab-strip speci- James McGill Professor,
McGill University,
One quiet Saturday September afternoon in Laval, Quebec a 35 mens6 shown in Fig 3. Two of these specimens were large (L) with
Montreal
year old, 1.25m thick reinforced concrete slab suddenly failed in thicknesses comparable to that of the failed Laval slab while the
shear causing the collapse of a major overpass and killing five other two were small (S) with thicknesses similar in size to tradi-
motorists. See Figs 1 and 2. After a major public inquiry1 into this tional laboratory specimens. For each pair of specimens one Evan C. Bentz
2006 failure and a detailed evaluation of 135 concrete structural contained no shear reinforcement (as was the case for the Laval BASc, PhD
slab bridges, the Government of Quebec identified 28 additional slab) and 0.83% of longitudinal reinforcement, while the other Associate Professor,
bridges that will need to be demolished because of similar struc- contained 0.50MPa of shear reinforcement. To avoid flexural fail- University of Toronto
tural deficiencies and a further 25 which will need to be substan- ures the specimens with links contained somewhat more longi-
tially strengthened. How could the design procedures of 40 years tudinal reinforcement (1.33%). All four specimens failed in shear.
ago have been so inadequate and are our current shear design The table in Fig 3 compares the observed failure shear stresses
procedures much better? of these four specimens with the failure shear stresses predicted Keywords: Overpasses,
Collapse, Failures, Shear,
By the 1950s the problem of accurately determining the flex- by current Canadian7, American3, European4 and British5 design
Design, Reinforced
ural capacity of reinforced concrete sections had been solved. codes. Note that for each slab-strip specimen there are consider- concrete, Research,
Combining the simple and general ‘plane sections remain plane’ able differences between the four code predictions. Thus for the Testing, Codes of
theory with appropriate stress block factors, engineers could esti- large specimen without links, L-10H, the highest predicted practice,g
mate flexural failure moments with considerable precision. failure shear stress is 2.56 times the lowest predicted failure
© Michael P. Collins, Denis
Accurately estimating shear capacity, which is more critical for shear stress and both the ACI and EC 2 predictions are seriously Mitchell & Evan C. Bentz
public safety, turned out to be a much more difficult problem. In unconservative. Further, the codes also give very different predic-

Fig 1. Collapse of Laval Overpass (photo: courtesy of La Presse, Montreal/ Robert Skinner) / Fig 2. Details of Laval Bridge

32 The Structural Engineer 20 May 2008

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:58 Page 33

PAPER

tions as to how the failure shear stress will change as the depth is made larger a catastrophic failure.
or links are added. The ACI code predicts the same failure shear stress for the
thick slabs as for the thin slabs and predicts that adding 0.5MPa of links will Theory-based shear design versus empirical equations
increase the failure shear stress by 0.5MPa. EC 2 predicts that the thick slab In his keynote lecture at the 1970 FIP Congress Professor Leonhardt9
with links will have a failure stress only 0.27MPa higher than the thick slab suggested that one of the prime reasons for the poor quality of shear design
without links. BS 8110 predicts this increase to be 0.88MPa while the Canadian provisions for reinforced concrete was that ‘in shear more than twenty param-
code (CSA A23.3) predicts 1.20MPa. The observed increase in the experiments eters influence the ultimate strength and also there are so many tests of little
was 1.12MPa. value.’ Like the 1962 ACI committee he called for future tests not to stray too
In contrast to the major inconsistencies that exist between the shear far from practice and suggested that to ensure accurate results, the depth of
strength predictions of these four codes, the predictions for flexural failure loads laboratory shear specimens should be at least 500mm. Fig 4 shows charac-
are remarkably consistent. Thus if the magnitude of the central point load teristics of a recently assembled10 database of 1849 reported shear tests on
required to cause flexural failure of the large specimen with links (L-10HS, members not containing shear reinforcement. It can be seen that only 144 of
b = 300mm, fy = 452MPa) is calculated, the four different codes predict values these specimens have depths exceeding 550mm and that results from most of
of 1610kN, 1598kN, 1626kN and 1616kN. Thus while the ratio of the highest these larger tests have only become available since 1996. An additional aspect
to lowest predicted shear failure load can reach 2.56, the ratio of the highest influencing the usefulness of available shear tests is that while the majority
to lowest predicted flexural failure load is only about 1.02. of actual shear designs involve continuous members subjected to uniform
If the British code of practice of 50 years ago, CP 1148, was used to predict loads only about 1% of laboratory shear tests address this situation.
the flexural failure load of specimen L-10HS a value exactly the same as the In discussing the application of ‘laboratory experimentation’ to structural
current British code would be obtained. However because the current BS 8110 design Professor Hardy Cross11 quoted with approval the paradoxical state-
uses a lower flexural factor of safety than CP 114 (1.68 versus 2.0 if the live ment of the Cambridge astrophysicist Sir Arthur Eddington that ‘No experi-
load is half of the dead load) the permitted magnitude of the service load ment is worthy of credence unless supported by an adequate theory.’ For a
moment on this thick slab is about 20% higher than it was 50 years ago. For complex phenomenon such as shear failure, influenced by perhaps 20 param-
shear the situation is radically different. In evaluating the permissible shear eters, understanding the meaning of particular experiments and the range of
on, say, L-10H, BS 8110 uses a total factor of safety of 1.83 giving a permissi- applicability of the observed results is extremely difficult unless the researcher
ble shear stress at service loads of 0.64/1.83 = 0.35MPa. CP 114, on the other is guided by an adequate theory. The shear design procedures of the 2004
hand, specified a permissible shear stress of 0.78MPa for members without Canadian code are based on such a theory, namely the modified compression
links made from high quality concrete. Thus over the last 50 years British codes field theory12 (MCFT). This paper will explain the background of this theory
have increased the allowable flexural moment on a given section by 20%, but and will demonstrate how the resulting Canadian procedures can be used to
have decreased the allowable shear force by up to 55%. Hence many reinforced accurately estimate the shear strength of reinforced or prestressed concrete
concrete structures for which the original design indicated that shear was not structures.
critical, will now be predicted to be critical in shear.
In 1970 the designers of the Laval bridge believed that the thick cantilever Modified compression field theory
slabs in their bridge were not critical in shear. Based on comparing the expres- Perhaps one of the reasons it has taken so long to develop an adequate theory
sions for allowable and failure shear stress, they would have deduced that the for shear is that the traditional type of shear test, such as that shown in Fig 3,
factor of safety against a brittle shear failure was at least 1.81. However, while simple to perform, yields results which are difficult to use as the basis
Canadian code changes since 1970 have resulted in a 36% reduction in permis- of a theoretical model. Thus if a relationship is sought between the magnitude
sible shear stress for this thick slab. This would suggest that when the bridge of the applied shear stress and the resulting strain in the shear links it will be
opened, the actual factor of safety against a brittle shear failure was only found that every link along the length of the beam exhibits different strains
about 1.33. As will be shown in this paper, the shear strength of members and that for any one link the strains change significantly from the top to the
without links is strongly dependent on crack widths and if due to long-term bottom of the link.
degradation of the concrete crack widths increase and aggregate interlock The modified compression field theory was developed by testing reinforced
properties decrease, this 1.33 factor of safety at construction may not prevent concrete elements in pure shear using the membrane element tester shown in

3 4

Fig 3. Four slab strip specimens / Fig 4. Shear tests of members without links

20 May 2008 The Structural Engineer 33

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:58 Page 34

PAPER

Fig 5. While these experiments were more complicated to perform, they led to 6, was derived from the aggregate interlock experiments of Walraven14.
results which were much simpler to interpret and thus could more easily form The MCFT is a procedure to predict the load-deformation response of a rein-
the basis of a theoretical model. By comparing the principal compressive stress forced concrete element subjected to bi-axial stresses, fx, fz and v. To use the
in the concrete, f2, with the corresponding principal compressive strain, ε2, it theory to predict the load-deformation response of a reinforced concrete beam
was found that diagonally cracked concrete was weaker and softer than the such as that shown in Fig 7 the beam could be represented as an assemblage
same concrete in a standard cylinder test and that the difference between the of elements with the response of each element being predicted by the MCFT.
two responses increased as the magnitude of the co-existing principal tensile This is the basis of non-linear finite element programs such as VecTor215 which
strain, ε1, increased. Further it was found that even after extensive cracking have been developed at the University of Toronto over the past 20 years. If such
of the reinforced concrete element, the Mohr’s circle of average stress did not a program is used to analyse the beam it will be found that in zones extend-
pass through the origin but showed a consistent and predictable principal ing for a distance of about d away from the point loads and the reactions there
tensile stress in the concrete, f1. Accounting for these tensile stresses in the will be significant vertical compressive stresses in the concrete. These clamp-
cracked concrete changed the previously developed compression field theory13 ing stresses will enhance the shear strength of the elements in these zones
to the modified compression field theory, MCFT, and allowed the shear behav- making it probable that the shear failure will occur outside of these zones. For
iour of both members with and without shear reinforcement to be predicted. beams with short shear spans the zones with significant clamping stresses will
The equilibrium equations, geometric conditions and stress-strain relation- overlap and the shear strength of the beam will be considerably increased. It
ships used in the MCFT are summarised in Fig 6. The central simplifying is important to recognise that in these ‘disturbed regions’ the shear stress distri-
assumption of the theory is that the average direction of principal compressive bution over the depth of the beam is influenced by the distribution of the
stress in the cracked concrete corresponds to the average direction of princi- clamping stresses and near the loads and reactions, plane sections do not
pal compressive strain and that the critical cracks are also inclined in this direc- remain plane.
tion. In addition to considering average stresses and average strains in the Outside of the disturbed regions discussed above it is appropriate to assume
cracked concrete and the relationships between them, the theory also consid- that plane sections remain plane and that the clamping stresses are negligi-
ers how the local stresses are transmitted across the critical cracks. As an ble. With these two assumptions a beam cross-section can be modelled as a
example, an element which contains no ‘shear’ reinforcement (i.e. ρz = 0) and stack of biaxially stressed elements with the response of each element being
is subjected to shear and uniaxial tension in the X direction must transmit a predicted by the MCFT. This is the basis of program Response-200016 which
shear stress, vci, across the crack interface, which is equal to the applied shear can be used to predict the shear stress distribution over the depth of the beam
stress, v. See equilibrium Eq 5 in Fig 6. The ability of the crack to transmit this and the complete load-deformation response of concrete sections subjected to
shear stress depends on the width of the crack, w, the maximum aggregate size, shear, flexure and axial load. See Fig 7. If only the shear strength of a beam
ag, and the concrete cylinder strength, fc′. The relationship used, Eq 15 in Fig cross-section is required then the web of the beam can be approximated by just

Fig 5. Membrane element tester / Fig 6. Equations of the modified compression field theory / Fig 7. Models of cracked T-Beam

34 The Structural Engineer 20 May 2008

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:59 Page 35

PAPER

one biaxial element located at mid-depth and the shear stress on the element MCFT shear strength for flexural regions
can be assumed to be V/(bwdv) where bw is the web width and dv is the flexural Regions of members where it is reasonable to assume that plane sections
lever arm which can be taken as 0.9d. The longitudinal strain, εx, at mid-depth remain plane and where the shear stress distribution does not significantly
of the beam can be found from the calculated strain in the longitudinal flex- differ from that calculated by traditional beam theory can be called flexural
ural reinforcement and the assumption that plane sections remain plane. For regions. The unfactored shear strength, Vn, of a section within such a region
a given value of εx the failure shear stress can then be calculated from the as calculated by the Canadian code can be determined from the following
MCFT as the sum of two terms, Vc and Vs, see Fig 7. The first term depends on equations where the unfactored failure shear stress vn is defined as Vn/(bwdv).
the ability of the cracks to transmit shear stress, characterised by the coeffi-
cient β, while the second term is a function of the clamping stress provided by vn = b fcl + tz fy cot i # 0.25fcl ...(1)
the shear links, ρzfy, and the angle of inclination of the principal compressive
stresses, θ. Both β and θ depend on the strain εx and the equivalent crack The aggregate interlock parameter β primarily depends on the width of the
spacing sxe. This MCFT based sectional design model for shear will be explained cracks and size of the aggregate ag. As crack width depends on both the average
in more detail below. tensile strain in the cracked concrete and the crack spacing it is not surpris-
ing that the expression for β derived from the MCFT18,19 has a strain term and
Distribution of shear stress over member depth a crack spacing term. The equation is
About 100 years ago Mörsch17 deduced that to maintain beam action a rein-
0.4 1300
forced concrete member must transmit a vertical shear stress equal to V/(bwdv) b= : ...(2)
1 + 1500fx 1000 + s xe
over the flexurally cracked depth. Fig 8 shows the post-failure appearance of
slab strip specimen L-10N26 which was identical to L-10H except it was made The longitudinal strain at mid-depth, εx, can be conservatively taken as one
from concrete with a cylinder strength of 40MPa. It failed at an almost iden- half of the tensile strain in the flexural tensile reinforcement. Allowing for the
tical load. The very narrow crack widths shown were measured when the tension in the longitudinal reinforcement caused by the shear and assuming
applied central point load had a value of 450kN which was 99% of the failure that there is no axial load or prestressing gives
1 + M/]V $ dvg
load. For this member Mörsch would have predicted that prior to failure 79%
of the shear force is transmitted across the flexural cracks while only 21% is fx = v n ...(3)
2E s tl
carried in the uncracked compression zone. Strain measurements in the
uncracked zone above the critical crack showed that for this specimen 24% of where M/V is the ratio of bending moment to shear at the section being
the shear was carried in the uncracked compression zone, which indicates that considered and ρl is the geometric ratio of the area of longitudinal flexural
Mörsch’s predicted shear stress distribution is a very accurate estimate of how tension reinforcement to the shear area. That is: As/(bwdv).
the shear force is carried. If the member contains more than the specified minimum amount of shear
A shear stress, V/(bwdv), of 0.64MPa needed to be transmitted across the reinforcement (i.e. ρzfy > 0.06√fc′) then it can be assumed that the crack spacing
flexural cracks shown in Fig 8 to maintain the beam action associated with will be well controlled and hence sxe can be taken as 300mm which reduces the
the 450kN applied load. Wide cracks have less ability to transmit this shear crack spacing term in Eq (2) to unity. If the member contains only concentrated
stress than narrow cracks. While the highest longitudinal strains occurred longitudinal reinforcement then the spacing of vertical cracks near mid-depth
near the bottom face of the member, the cracks were more widely spaced near of the member, sx, is assumed to be equal to 0.9d. See Fig 8. To allow for the
mid-depth of the beam, and hence the cracks with the largest measured influence of aggregate size the effective crack spacing, derived from Eq(15) in
width (0.4mm) occurred near this location. When the cracks became too Fig 6, is taken as
wide to transmit the required shear stress the ‘tooth’ of concrete between two
35s x
adjacent flexural cracks began to act as a vertical cantilever subjected to a s xe = $ 0.85s x ...(4)
15 + a g
horizontal force (going to the left in this case) at the level of the flexural
tension reinforcement. The vertical tension stresses in the top right corner For high strength concrete the cracks will go through the aggregate rather
of this tooth caused the nearly horizontal final failure crack to propagate from than around the aggregate particles leading to smoother crack surfaces with
this location. For a member without links, the MCFT sectional model predicts less aggregate interlock capacity. To account for this, if fc′ exceeds 70MPa the
the shear stress at which the aggregate interlock capacity of the cracks at term ag in Eq (4) is taken as zero. As fc′ goes from 60MPa to 70MPa, ag is linearly
mid-depth of the section can no longer transmit the required stress and reduced to zero. As an additional allowance for the low aggregate interlock
beam action begins to break down. capacity of high strength concrete the term √fc′ in Eq (1) is not allowed to exceed
8MPa.

8 9

Fig 8. Failure of slab strip L-10N2 / Fig 9. Failure of slab strip L-10HS

20 May 2008 The Structural Engineer 35

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:59 Page 36

PAPER

The MCFT predicts that the angle of the principal compressive stress, θ, at will be significantly different from that assumed by Mörsch, and the clamping
shear failure for members with shear reinforcement depends primarily upon stresses will have a significant influence on shear strength. In these regions,
the longitudinal strain, εx, while for members without shear reinforcement the strut-and-tie models more accurately describe the flow of forces. In these
crack spacing, sxe, is also important. The general equation19 is models, the flow of compressive stresses in the concrete is represented by
concrete struts while the tension reinforcement is represented by ties. Such
i = ]29c + 7000fxgc0.88 +
s xe m
# 75c ...(5) strut-and-tie models are given by the CSA, EC 2 and ACI codes. The principal
2500
difference between these three implementations concerns the expressions
Note that for members with at least a minimum amount of links the second used for predicting the crushing strengths of the struts. The CSA code uses
term in this equation reduces to unity. compatibility conditions to calculate the principal tensile strains in the concrete
struts where they cross or intersect with tension ties and uses equation 13 of
Example calculations using MCFT sectional model Fig. 6 to define the crushing strength. Assuming that the compressive strain
As an example of the use of the MCFT procedure, the failure shear stress of in the concrete at crushing of the strut is 0.002, and that the tensile strain in
the large specimen with no shear reinforcement described in Fig 3, will be the reinforcing steel tension tie is εs, the principal tensile strain in the strut is:
calculated. Because of the beneficial effects of the clamping stresses in the vicin-
ity of the load or the reaction the critical section for shear is taken to be located f1 = fs + ]fs + 0.002g cot2 is ...(6)
at a distance of dv from the face of the load or the support. For the simply
supported beam shown in Fig 3 the section dv from the load has the higher and the crushing strength of the strut is given by:
moment and hence is more critical. At this section the M/V ratio is 4050 – 75
fcl
– 0.9 × 1400 = 2715mm. fcu = # 0.85fcl ...(7)
0.8 + 170f1
As the concrete cylinder strength is 75MPa the term ag is taken as zero and
√fc′ in Eq (1) is taken as 8MPa. From Eq (4) sxe = 2940mm and hence Eq (1)
becomes: Thus if a tie has a tensile strain of 0.002 and the angle between the strut
and the tie is 25° the crushing stress of the strut is predicted to be 0.23fc′. By
vn = 1.056/]1 + 1500fxg MPa ...(1a) contrast, EC 2 and ACI assume that the crushing stress of the strut depends
only on the concrete strength (about 0.5fc′ for EC 2 and ACI). In addition, the
while Eq. (3) becomes ACI code requires that the angle between the strut and tie not be less than 25°.
1 + 2715/]0.9 $ 1400g
Fig 10 shows the results of experiments by Kani20 in which he investigated
v
= n
2 $ 200000 $ ^0.0083/0.9h
fx = v n ...(3a) the influence on shear strength of changing the shear-span-to-depth ratio (a/d)
1169
for a series of beams without links. For larger a/d ratios, the beams failed imme-
diately upon the breakdown of beam action as discussed above and shown in
Substituting Eq (3a) into Eq (1a) and solving the quadratic we find that the Fig 8. For beams with shorter shear spans, after the breakdown of beam action,
predicted value for vn is 0.599MPa and the corresponding value of εx is 0.511 the tension in the longitudinal reinforcement became essentially constant
× 10–3. For this value of εx the predicted value of the angle of inclination of the from support to support and an arch could form in the concrete enabling the
failure crack at mid-depth is given by Eq (5) as 67º. Note that the CSA failure member to carry substantially higher loads. The magnitude of this arch-action
shear stresses listed in Fig 3 are 90% of the values calculated above because can be estimated by using a strut-and-tie model as shown in Fig 10. From this
the shear stresses in the table are calculated as V/(bwd). figure it can be seen that the CSA procedures lead to accurate estimates of
As a further example, consider the large specimen in Fig 3 with links. For shear strength both for the members governed by breakdown of beam action
this specimen, the crack spacing can be taken as 300mm and hence Eq (1) and for members in which strut crushing is predicted to govern shear capac-
becomes ity. While we have seen that the shear stress at breakdown of beam action is
strongly influenced by the absolute size of the member, shorter span members
vn = 3.20/]1 + 1500fxg + 0.5 cot i ...(1b) whose capacity is controlled by arch-action are predicted to show no size effect.

Because there is 1.33/0.83 more longitudinal reinforcement in this specimen Current empirical shear design equations
Eq. (3a) becomes A British engineer examining the four predicted failure shear stresses by BS
f x = vn /1870 ...(3b) 8110 in Fig 3 might conclude that the empirical equation for vc in this code
seems to be satisfactory. After all, the least conservative prediction is only 1.12
From Eq. (5) times the observed failure shear and presumably such a small error can be
i = 29c + 7000fx ...(5b) accommodated within the total factor of safety which was given above as 1.83.
That is, for this thick lightly reinforced slab the factor of safety is now down to
If εx is assumed to be 0.001, Eq (5b) gives θ as 36º and then Eq (1b) predicts 1.83/1.12 which is 1.63.
the failure shear stress to be 1.968MPa. For this value of vn, Eq (3b) predicts It needs to be appreciated that the current BS 8110 empirical equation is
εx to be 1.05 × 10–3 indicating that the estimated value of εx was reasonably accu- the same as that developed for the 1985 edition of this code. While very few
rate. Convergence is reached when the estimated value of εx is 1.034 × 10–3 results from large specimens were available, Fig 4, the authors of this equa-
giving a calculated failure stress of 1.937MPa and an angle of 36.2º. The tion captured well the influence of most of the major parameters. However from
appearance of this section as it failed is shown in Fig 9. In this figure the shear the data available to them they believed that the equation was rather conser-
slip on the wide failure crack can be seen from the displacements of the grid vative and hence they could justify a less severe partial factor of safety for shear
of horizontal and vertical lines spaced at 100mm apart. The crack widths failures of members without links. Is it really appropriate that the total factor
shown in photograph (Fig 9) were measured when the load was at 87% of the of safety against such a brittle shear failure (1.83), which would occur with
failure value. It can be seen that this slab with links displayed crack widths essentially no warning, is just 9% greater than that used for ductile flexural
prior to failure that were 10 times larger than those measured for the slab failures (1.68) which give ample warning of approaching failure? In compari-
without links shown in Fig 8. Note that ACI and BS 8110 assume that the son, for ACI the factor of safety for these brittle shear failures is 20% greater
failure crack shown in Fig 9 will be at 45º to the longitudinal axis, while the than that for ductile flexural failures, while for EC 2 and CSA the increases
EC 2 calculation assumes it will be at 22º. It can be seen in Fig 9 that near mid- are 30% and 31%.
depth, the angle of the failure crack is about 32º which is not far from the value A further difficulty is that, as we have seen, while different international
predicted by the CSA equation. codes agree very closely about the magnitude of the load required to cause a
ductile flexural failure the codes disagree very strongly about the more criti-
MCFT shear strength for disturbed regions cal question of what load will cause a brittle shear failure and how changing
In regions of members close to applied point loads or significant changes in particular parameters will change this failure load. For the slabs in Fig 3 the
geometry, plane sections will not remain plane, the shear stress distribution ACI equations gave the least safe predictions while for the beams in Fig 10 it

36 The Structural Engineer 20 May 2008

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:59 Page 37

PAPER

was the EC 2 predictions that were the least safe. With such a degree of scatter of the effect of changing ρl in this practical way. The EC 2 and BS 8110 empir-
in the accuracy of the shear strength predictions given by these empirical equa- ical equations are notably less conservative for members with large amounts
tions, it is necessary to have a total factor of safety against shear failures which of longitudinal reinforcement subjected to high moments (See Fig 10) and
is considerably higher than the factor used for flexural failures even without hence the codes need to restrict the amount of such reinforcement which can
considering the additional factor that should be present to account for the fact be counted in the shear strength equations. On the other hand for lightly rein-
that brittle shear failures are inherently more dangerous. forced slabs the predictions of the empirical equations are extremely conser-
vative.
Effect of amount of longitudinal reinforcement
One of the difficulties with determining a shear design procedure by fitting an Effect of axial load or prestressing
equation to a large set of laboratory tests is that the way the parameters vary The ACI and BS 8110 shear design procedures for members with links are
in these tests may not be representative of the way the parameters vary in based on the assumption that the shear strength can be estimated as the shear
actual designs. For example, in practice if a member contains a large quantity at which beam action breaks down plus the shear strength predicted by
of flexural tension reinforcement, it is almost certain that it will be subjected Mörsch’s 45° truss model. While it is difficult to see why the diagonal cracking
to large moments. In laboratory shear experiments, however, often a very large shear should happen to equal the correction required to make the 45° truss
percentage of longitudinal reinforcement is provided in order to ensure that a predictions agree with experiments, for a large number of reinforced and
flexural failure does not ‘prematurely’end the shear test. In real designs, on the prestressed concrete beam tests, this simple approach was found to be reason-
other hand, it is desirable to have the member capacity governed by a ductile ably accurate. However calculations based on the MCFT predict that for
flexural failure. The 1962 basic shear expression of the ACI code is based on members subjected to axial load, the ACI approach is excessively conservative
the hypothesis that failure shear stress will decrease as the tensile stress in for members subjected to tension, but can be very unconservative for members
the longitudinal flexural reinforcement increases. Even though the empirical subjected to compression.
coefficients chosen by the ACI engineers are now known to be unconservative, To account for the effect of axial load and prestressing, the equation for εx
the concept itself is correct. Because breakdown of beam action is related to loss in the CSA code is:
of aggregate interlock capacity on the cracks, increase of longitudinal strain,
M f /dv + V f - V p + 0.5N f - A p f po
2 ^ E s A s + E p A ph
which increases crack widths, will reduce failure shear stress. The purely fx = ...(8)
empirical equations of EC 2 and BS 8110 relate the failure shear stress not to
strain in the longitudinal reinforcement but to the percentage of longitudinal Where Ap is the area of prestressing steel on the flexural tension side and
reinforcement provided. Which of these two approaches is more correct? fp0 is the stress in this steel when the strain in the surrounding concrete is equal
In Fig 11, the results of 44 experiments20,21 on simply supported beams all to zero. Thus for a given magnitude of the factored moment, Mf, and factored
with similar depths and concrete strengths, but with different values of shear shear Vf, axial tension increases εx and, hence, reduces shear strength while
span and percentage of longitudinal reinforcement are shown. These experi- prestressing decreases εx and therefore increases shear capacity.
ments were chosen to represent the practical situation where the values of In designing the required shear reinforcement near a point of contraflexure,
(a/d)/ρl for such point loaded members lie within a reasonably narrow range. the term Vc in the ACI and BS 8110 codes will be greatly increased as the level
The shear stress multiplied by this ratio equals the maximum stress in the of prestressing is increased. The plasticity based EC 2 shear design model for
longitudinal reinforcement which in practical designs will be close to yield members with links has no Vc term and specifies the maximum value of cotθ
stress at ultimate loads. The lines showing the ACI and CSA predictions in Fig to be 2.5. Hence the predicted shear strength of such members is not affected
10 have been plotted for the average value of (a/d)/ρl (namely 260) and indi- by the level of prestress. The empirical EC 2 equations for members without
cate that the failure shear stress should remain essentially constant as ρl links, however, do show an increase in shear strength with increasing prestress
increases. The trendline for the experimental points agrees with this predic- and, hence, at high levels of prestress, the capacity of a member without links
tion. In contrast, the empirically derived equations provide a very poor estimate is predicted to be higher than that of a member with links. The MCFT-based

10 11

Fig 10. Arch action versus beam action / Fig 11. Influence of changing amount of longitudinal reinforcement

20 May 2008 The Structural Engineer 37

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:59 Page 38

PAPER

CSA code predicts that prestressing will increase the shear strength of such a Overall accuracy of the four codes
member but not to the extent that ACI or BS 8110 indicate. While the dangers of relying too heavily on the entire set of available labora-
Fig 12 shows the results of a recent series of experiments22 on tory experiments have been demonstrated above, it is still of interest to
prestressed concrete continuous beams reinforced and loaded so that shear compare how the empirical design equations that were derived by fitting to
failures would occur near the point of contraflexure. Also shown in the such a body of data compare with the CSA equations which were based on the
figure are the predictions from the four codes as to how increasing the level modified compression field theory.
of prestressing influences shear strength. As these experiments were Fig 13 shows the values of the ratios of the predicted failure shear stresses
conducted in 2007 and are some of the first such experiments available, the to the observed failure shear stresses of the large database of members
code estimates are true predictions for what happens in a common practi- without links shown in Fig 4. It can be seen that the CSA provisions provide
cal situation for which laboratory experiments are not available. Notice the lowest scatter. The dashed horizontal lines indicate the total factor of
that in this case the shear strengths predicted by BS 8110 are up to 1.41 safety for shear failure of such members used in the different codes for a
times the experimental values. This figure provides strong evidence of the representative ratio of live load to dead load. Experimental points lying
value of basing shear design provisions on an adequate theory which in this above these lines indicate that failure at service loads is more probable for
case correctly predicts not only the effect of changing the level of prestress this set of parameters. It is a troubling aspect of both the ACI code and the
on the failure shear, but also predicts the decrease in the angle of inclina- EC 2 code that so many points are above these lines. For members with short
tion of the failure cracks. shear spans, BS 8110 uses an empirical enhancement factor, and in this
region gives very conservative results.

12 14

13

Fig 12. Influence of changing level of prestress / Fig 13. Comparative accuracy of four codes in predicting shear strength of members without
links. Total factor of safety listed for case where live load is 50% of dead load / Fig 14. Influence of minimum links on member ductility

38 The Structural Engineer 20 May 2008

SE10 Shear des \ Conc Struct:Layout 1 14/5/08 15:59 Page 39

PAPER

Concluding remarks on the assumption that the longitudinal reinforcement will be designed to be
The recent collapse of the Laval bridge has emphasized the importance of accu- just less than the yield stress at factored loads, the CSA code also permits the
rate shear design provisions both for the design of new structures and for the use of a simplified method for evaluating εx wherein εx is conservatively taken
evaluation of older structures, many of which were designed using shear provi- as 0.43 times the yield strain of the longitudinal reinforcement.
sions which are now known to be rather unconservative. How could the shear One of the key advantages of a theoretically-based shear design procedure
design provisions of 40 years ago been so inadequate and if they were, why was is that it is capable of making accurate predictions for the impact of changing
the problem not identified more quickly? The basic problem was that these different design parameters. Thus when faced with the design of a thick slab
provisions were developed at a time when there was no adequate theory for such as that shown in Fig 1423, engineers may question whether the enhance-
the shear strength of reinforced concrete members and the available labora- ments in shear behaviour that will result from adding minimum links will
tory experiments were nearly all conducted on rather small specimens with a justify the cost involved. If they use traditional, empirical design equations, they
rather narrow range of parameters. Because predicting shear strength is a may conclude that the strength enhancements do not justify the cost. As shown
more complex problem than predicting flexural strength and is influenced by in Fig 14, however, the addition of minimum links to the thick slab strip shown
many more parameters, developing accurate shear design provisions using only has increased its shear strength by a factor of about three and has increased
the results from a limited range of shear experiments is a very difficult task. the energy required to cause collapse by a factor of about 30. The structure
If flexural design provisions were seriously unconservative, the resulting without links gave no warning of failure and failed prior to yield of the longi-
concrete structures would display significant signs of distress and, hence, the tudinal reinforcement. With minimum links, yielding of the longitudinal rein-
problem would be quickly rectified. Unfortunately, for the case of shear, it is forcement governed the failure load and caused significant deformations and
possible for a member to be at 99% of the failure load and yet show no signif- cracking prior to failure. The bridge which collapsed in Laval is similar to the
icant signs of distress. See Fig 8. Thus unconservative shear design provisions member without links. Had the current Canadian shear provisions been in
can be used for many years without the resulting very low factor of safety being force in 1970, the slab would have required minimum links and the failure
discovered. would have been prevented.
Given the large number of existing concrete structures designed using
unconservative shear provisions and the limited resources available to increase Acknowledgments
public safety, it is essential that the shear design provisions used to evaluate The authors would like to express their gratitude to the Natural Sciences and
the structures be as accurate as possible. It is evident from the comparisons Engineering Research Council of Canada for a series of grants that have made
in Fig 13 that of the four codes studied, the shear provisions of the CSA code possible the long-term research project on the shear design of reinforced
provide estimates with the least scatter. This low scatter is despite, or perhaps, concrete at the University of Toronto and at McGill University.
because of the fact that these Canadian provisions are theoretically-based, not
fitted to the experimental database.
It should be appreciated that one of the advantages of shear design provi- Michael P. Collins is a Professor of Civil Engineering at the University of
Toronto where he has led a long-term research program aimed at
sions that are based on a theoretical model is that for important evaluations,
improving shear design procedures for concrete structures. The results
more refined analyses such as those described in Fig. 7 may be used. For of this work form the basis of the shear provisions of the Canadian
typical designs, a structural engineer may be concerned that based on the building and bridge codes, and the US bridge code. He has received 15
example calculations given above, the theoretically based CSA shear design international awards for technical publications and contributions to the
procedures may be too complex for everyday use. It should be appreciated that practice of structural engineering. He has participated in a number of
while iteration is required to evaluate the shear strength of an existing struc- major failure investigations and in evaluating and strengthening
ture, no iteration is necessary for design. The longitudinal strain at mid-depth, concrete structures in distress. He is a graduate of the University of Canterbury, New
εx, can be directly calculated from Eq 8 using the factored design loads. Based Zealand, and the University of New South Wales, Australia.

REFERENCES

1. Wood, J. G. M.: ‘Implications of the collapse of the de la Concorde Overpass’, 12. Vecchio, F. J., and Collins, M. P.: ‘The modified compression field theory for rein-
The Structural Engineer, 86/1, 8 January 2008, Institution of Structural Engineers, forced concrete elements subjected to shear,’ ACI J., Proc., 83 /2, Mar.-Apr. 1986,
London, pp. 16-18 pp. 219-231
2. ACI-ASCE Committee 326, ‘Shear and diagonal tension,’ ACI Journal, Proc., 59/1, 13. Collins, M. P.: ‘Towards a rational theory for RC members in shear,’ J. Struct. Div.,
2, 3, Jan., Feb., Mar., 1962, pp. 1-30, 277-334, and 352-396 and discussion ASCE, 104/ST4, April 1978, pp. 649-666
and closure, Oct 1962 pp. 1323-1349 14. Walraven, J. C., ‘Fundamental analysis of aggregate interlock,’ J. Struct. Div.,
3. ACI Committee 318, ‘Building code requirements for reinforced concrete (ACI ASCE, 107/ ST11, Nov. 1981, pp 2245-2270
318-08) and commentary (318R-08),’ American Concrete Institute, Farmington 15. Vecchio, F. J.: ‘Nonlinear finite element analysis of reinforced concrete
Hills, 2008, 465 pp membranes,’ ACI Struct. J., 86/ 1, Jan.-Feb. 1989, pp. 26-35
4. EN 1992-1-1:2004: Eurocode 2: Design of Concrete Structures- Part 1-1: General 16. Bentz, E. C.: ‘Sectional analysis of reinforced concrete members,’ Ph.D. Thesis,
rules and rules for buildings, European Committee for Standardization, CEN, Department of Civil Engineering, University of Toronto, 2000, 198 pp.
Brussels, Belgium, 2004, 225 pp 17. Mörsch, E.: ‘Der Eisenbetonbau (Reinforced Concrete Construction),’ Verlag von
5. BS 8110-1:1997 Structural Use of Concrete – Part 1: Code of Practice for Design Konrad Witwer, Stuttgart, Germany, 1907, 368 pp
and Construction, 1997, British Standards Institute, London, 159 pp 18. Bentz, E. C., and Collins, M. P.: ‘Development of the 2004 CSA A23.3 Shear
6. Sherwood, E. G., Bentz, E. C., and Collins, M. P.: ‘Effect of aggregate size on Provisions for Reinforced Concrete,’ Canadian J. Civ. Eng., 33/5, May 2006, pp.
beam-shear strength of thick slabs,’ ACI Struct. J., 104/ 2, Mar.-Apr. 2007, pp. 521-534
180-190 19. Bentz, E. C., Vecchio, F. J., and Collins, M. P.: ‘The simplified MCFT for calcu-
7. CSA Committee A23.3, ‘Design of concrete structures’, Canadian Standards lating the shear strength of reinforced concrete elements’, ACI Struct. J., 103/
Association, Mississauga, Ontario, Canada, 2004, 214 pp 4, July-Aug. 2006, pp. 614-624
8. Scott, W. L., Glanville, Sir William, and Thomas, F. G.: ‘Explanatory Handbook 20. Kani, M. W., Huggins, M. W., and Wittkopp, R. R.: ‘Kani on Shear in Reinforced
on the B.S. Code of Practice for Reinforced Concrete No. 114: 1957,’ Concrete Concrete’, Dept. Civ. Eng., University of Toronto, Toronto, 1979, 225 pp
Publications Limited, London, 1957, 154 pp 21. Krefeld, W. J. and Thurston, C. W.: ‘Studies of the shear and diagonal tension
9. Leonhardt, F.: ‘Shear and torsion in prestressed concrete,’ Proc. 6th FIP strength of simply-supported reinforced concrete beams,’ ACI J., 63/4, Apr.
Congress, Prague, June 1970 1966, pp. 451-476
10. Collins, M. P., Bentz, E. C., and Sherwood, E. G.: ‘Where is shear reinforcement 22. Collins, M. P., Bentz, E.C., Sherwood, E. G., Xie, L., ‘An adequate theory for the
required? A review of research results and design procedures,’ ACI Struct. J., in shear strength of reinforced concrete structures,’ Magazine of Concrete Research,
Press In Press
11. Cross, H.: ‘Limitations and applications of structural analysis’, Engineering News- 23. Lubell, A., Sherwood, T., Bentz, E. C., and Collins, M. P.: ‘Safe shear design of
Record, 17 Oct. 1935, pp. 535-537 large, wide beams,’ Conc. Int., 26/1, Jan. 2004, pp. 66-78

20 May 2008 The Structural Engineer 39