UDC624.073.012.45:620.176.

The dependence of punching resistance

upon thegeometry of the failure surface
P. E. Regan" BSC,PhD, DIC

T H E P O L Y T E C H N I C O F C E N T R A LI O N D O N

SYNOPSIS
Punching tests were made on forty concrete slabs
with the distances between loads and supports short
enough for the shapes of the failure surfaces to be
defined by specimen geometry. The results obtained
are used to derive a simple empirical expression for
punching strength in terms of the geometry of the
failure cone. The empirical expression is shown to be
similar to the results of aplastic analysis and the
relationship of the results to the design of more normal
slabs is discussed briefly.
Figure I : Failure surfaces for calculating nominal stressesaccording
to equation I .
Introduction
Conventional calculations of punching resistance
are madein terms of a nominal shear stress Plud on a cone or cone/pyramid with a base angle
0, as shownin
control perimeter of length U at a defined distance Figure 1. For a circular load:
from a concentrated load or support. The definition
P
of the control perimeter providesno information on ff= (1)
the actual form of the punched cone, although the .rrhdTEZ% (B+hcotO) """~*.'

calculated resistance is intendedtobethat cor- where h = height of the truncated cone;

responding to the least favourable potential failure B = diameter of the loaded area.
surface.Theapproachhasanumberof practical The height h might be taken as theeffective or the
disadvantages. It cannot be applied sensibly in cir- over-all depth of the slab; the former definition is
cumstances where the failuresurface is forced into a adopted in reference 1.
shape different from that giving the normalminimum For slabs of normal dense concrete, without shear
resistance, asmay occurif a loadis close to a support. reinforcement and with loads and supports not in
It is unsuitable for slabs with shear reinforcement close proximity.it is recommended thatcot 0 be taken
where a number of possible failure surfaces crossed as 2-5 andtheultimate stressbeestimatedcon-
by different amounts of shear steel may need to be servatively as:
investigated. Itis also inapplicable to curvedplates.
An alternative approach, still in terms of nominal a,,= 0.13 qWrn
............... (2)
stresses, is to calculate stresses referred to a realistic
failure surface. Such a method is given in reference 1
where €,= d m w i t h d i n mm;
p = ratio of flexural tension reinforcement;
and the nominal stress adopted there is the concen-
trated load dividedby the surface area of a truncated
f,,= cube strength of the concrete.
It is suggested that, for other cases, the value of a,,
*Reader in CivilEngineering. The Polytechnic of Central London, should be related to the angle 0 and that the cal-
35 Marylebone Road, London, N W 1 5LS. culated resistance should be minimized with respect

Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.
 Magazine of Concrete Research : Vol. 36. No. l26 :March 1984

to possible failure angles. However. no guidance is r-----1

given on the relationship betweenr,,and 8. The tests
reported here were made to provide information on
this relationship.

Test programme
Two series of testsweremade. l n the first. the
specimensweresquareslabs withuniformly dis- 50
tributed reinforcement equal in twodirections. As
shown in Figure 2, the specimens were supported at
square perimeters and loaded by square punches.
The otherslabs were tested by the author in 1983.
Theywerealsosquare.but werereinforced only
around their edges as in Figure 3. The supports and
punches were circular. 15

In most tests, the slabs failed by punching at sur- REINFORCEMENT TEST ARRANGEMENT

faces which basically ran directly from the loads to the

supports, although there was some rounding off of Figure -7:Typical detni1.s of slabs. Series I
the square corners at the tension faces of the first
series. The exceptions tothis general pattern were a specimens which occurred in the second set when the
few flexural failuresin the first group of tests, mostly very short(8 = 70"). The data
clear shear spans were
in unreinforced slabs which areomittedhere,and for the specimens and the
test results are summarized
failures involving the almost total destruction of the in Tables 1 and 2.

T A B L E 1 : Summary of data, Series 1.

__
Square sidelength
kinforcement'
Test PU
No. P (Nimm') W )
(%) Outside of
support support
~

1 14 0 28.7 150 850 lo00 197

2 14 0 31.6 1S0 600 750 227
3 14 0 36.0 150 350 500 235
4 14 0 35.2 100 350 S00 185
5 14 0 35.2 200 350 S00 338
8 44.6 0.50 100 3.50 500 172
9 0.50 44.6 350 200 500 284
10 1.00 48.1 200 350 S00 42 1
11 1.00 48.1 100 350 S00 182
12 1.00 15.4 200 350 500 22 1
13 1 .00 15.4 100 350 500 109
14 1.00 47.8 200 300 500 623
15 1.00 47.8 200 400 500 368
16 1.00 47.8 100 200 500 45 1
17 37.6 0.75 200 400 600 1099
18 0.75 37.6 100 200 300 142t

*All reinforcement was of hot-rolled round deformed barswith yield stresses of approximately
480 Nimm'.
p = 1 .0%, Y IO at 70 mm centres bothways. dmean= 75 mm
p = 0.5%. Y10 at 140 mm centres both ways, dmean = 75 mm
p = 0.75%. Y12 at 75 mm centres both ways. dmean= I60 mm in slab 17
p = 0.75%. Y10 at110 mm centres bothways, dmean = 75 mm in slab 18

Values of p given above are based on the over-all slab depth (not the
effective depth.)
+Cube strengthf,, average value from tests of three 150 mmcubes per slabor group of slabs.
All concrete was made with rapid-hardening Portland cement. natural sand and irregular
natural gravel (rnax sixe20 mm).
$Flexural failure.

Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.
 The dependenceof punching resistance upon the geometry of thefailure surface

under
steel platewith
hole
central
' ~ ~ O O O ~ ~ O'txaring
O ]
rings split
ontwo
of
hardboard separated
by a layer of grease

a
*
-
- -- .. grease

t
15

Fi,qure 3: Tvpical details of slabs. Series 2.

T A B L E 2 : Summary of data, Series 2.

Test
No.
I Diameter of
reinforcement
(mm)
1 I fcut
(Nimm')
Diameter of
load
(mm
1 Diameter of
support
(mm)

I - I 84 300
2 8 42.2 170 300
3 250 300

I1 - I 84 300 0.74
-7 IO 38.9 1 10 300 0.84
3 170 300

111 - I 84 300
2 12 40.0 110 300
3 170 300

IV - I 84 300
2 10 55.0 170 300
3 250 300
1

v - 1 84 300 0.74 85
2 10 23.4 170 300 172 1.23
3 250 300 3.20 341:

VI - 1 200 300 1.60 323

2 10 41.7 230 300 2.29 418
3 250 300 3.20 447:

v11 - 1 84 400 0.51 130

2 10 48.7 200 400 0.80 216
3 250 400 1.07 284

VI11 - 1 55 400 0.46 72

2 IO 49.9 110 400 0.55 140
3 170 400 0.70 182

*All reinforcement wasof hot-rolled round deformed barswith yield stresses of approximately
500 N/mm'.
+Cube strength fCu average value from tests of three 150 mm cubes per group of slabs. All
concrete was madewith rapid-hardening Portland cement. natural sand and irregular natural
gravel (maximum size 20 mm).
*Ultimate loadsso marked correspond to failureby general destruction of the slab rather than
by ordinary punching.

Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.
 Magazine of Concrete Research :Vol. 36, No. 126 :March 1984

Analysis of results
In addition to geometric factors, the
tests involved
some variations of concrete strength andof reinforce-
1
ment. Results for groups of specimens in which the
concrete strengthwas the only variable are plotted in
Figure 4 and can be seen toconform reasonably well
with the relationship:
P, x fir......................... (3)
This relationship has been used to allow for minor
variations of concrete strength in Figure5, which 0 0-5 l
J
1.5
represents the effects of reinforcement. For the first p = A,lbh-%

series, the reinforcement can be described in terms of ( a ) Series I

the ratio p and is so treated in Figure 5a. For the
second series, such a description is not possible and
the horizontal axis of Figure 5b is the areaA , of one
bar, the number and arrangement of bars having been
the same in all the specimens. In both sets of tests,
there was some increase of strength with increasing
area of reinforcement. Approximately
p, x p”:’ for series 1 ...............(4a)
and
P, ,A,”:’ for series 2 ..............(4b)

It is probable that the scale or absolute depth of the

slab also influences the nominal stress at failure, but
no information on this effect is available from the
J
present tests. 0 20 40 60 80 100 120 140
The main interest in the present work was in the CROSS-SECTIONAL AREA OF ONE BAR-mm’
influence of the geometry of the failure surface. To ( b )Series 2
present this, experimental nominal stresseswttesthave
been calcuIated from equation(1) with h taken as the Figure 5: Influence of reinforcement upon punchingresistance,
over-all slab thickness and cot 8 set equal to the ratio
of the clear shear span to theover-all thickness (see
Figures 2 and 3). For the square slabs, B has been tan 8. Some scatteris to be expected, as allowance
no
taken as theside dimensionof the loading plate. has been made for variations of reinforcement, but
In Figure 6, values of wtest/f:kl are plotted against the data correspond very well to therelationship

Series 2, slabs lV3and V3

for tan 8 2.0. For higher values of tan 8, the data
are insufficient for any relationship to be proposed
Series 1, slabs 5.10 and 12 and the mode of failure was changed.
5
400 Relationship with plastic theory
Series 2, slabs 11-1. IV-2
The theory of plasticity has been applied to the
problem of punching by various The
, Series 1, slabs4.11 and 13 approach usedis that of an upper-bound analysis and,
if the failuresurface is truncated
a cone,the
expression obtained for theultimate load can be put
Series 2, slabsl-1.V-l
as:

0 4 8 12 16
CONCRETESTRENGTH,
1
20
tu’’
l
24
128
P =
u 2
fc
nh(B +h cot B)(h d m -p cot 6 )
..............(6)
where f , = plasticcompressivestrength of the
Figure 4: Influence of’concrete .stren,qthupon punching resistance. concrete:
h
Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.
 The dependenceof punching resistance upon the geometry of the failure surface
B
20 30 35 40 45 50 55 M) 65 70 75
l I

J
4
tan fi

Figure 6: Comparison oftestresults wirh equation 5 .

ft = plastic
tensile
strength of the The straightconicalfailuresurfacedoesnot
concrete; actually give the correct minimum upper-bound solu-
X = 1 -3&/f,; tion according to plastic theory. The correct failure
P = 1- 4 m surface is curved and its geometry canonly be deter-
tan 8 i 1.33. mined iteratively. However, for relatively high values
If this expression is divided by the surface area of the of tan 8, the errorsinvolved are small.
truncated cone, as in equation 1, the plastic theory Given that relatively similar results are produced
formula for the nominal ultimate stress becomes by plastic theory and by the empirical method, it is
appropriate to considerwhetherthere is a justifi-
CP,= fc(A-pcos8) ............(7) cation for the empiricalapproach. In fact, it does
2 seem tohave some advantages:
which may be compared with the empirical version of (1) the plastic theory is itself at least semi-empirical
equation 5. in its choice of fc, which normally deviates from
As shown in Figure 7, the two formulae are very the theoretical linear relationship with f,,and
similar in theirtreatments of variations of 0. In probably requires adjustments to allow for the
absolute terms, their relationship dependsupon the influences of scale and reinforcement;
value given toft, and toa lesser extent f t . (2) in its correct form, the theory is difficultto apply;
(3) if, as would normally be the case, ft is taken as
2
zero, equation 6 for P, has no minimum and the
calculated value decreases indefinitely as cot 8
increases - this does not happen in practice andis
not predicted by the empirical equation.

Application where the failure surface is

3 1 not predetermined
t'
If equations 1 and 5 are combined, the punching
strength of a slab is given by:

P,, = 0.177f~,%hd m ( B + h cot 8) tan:''' 8

............(8)
0 10 20 30 40 50 60 70
If the geometry of the loadandsupportdoes not
impose limits on 8. a slab may be expected to fail at
the surface offering the least resistance, i.e. that for
which ;IP,,/d8 = 0.
7
Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.
 Magazine of Concrete Research :Vol. 36, No. 126 :March 1984

5
1 Conclusions
The test results presented illustrate the depend-
ence of punching resistance uponthe angle 8 between
4
the failure surface and themean plane of the slab. If
the resistance is expressed in terms of a nominal
vertical stress, equal to the
failure load divided by the
3 (curved)area of the failuresurface,the nominal
stress is approximately proportional totan:3in8 and to
Y
p.
r
e
f ;c.
As with plastic theory, there is some uncertainty
I

2
whether the effective or the over-all value of the
depth of slab should beused in determining the area
of the failure surface. Effective depths are used in
1
most calculations forreinforced concrete butseem to
have little meaning where the surface is not crossed
by any reinforcement, asin the present series 2 or in
J pull-out failuresof embedments in concrete.
0 1 2 3 4 5
Blh Theamount of flexural reinforcementalso in-
fluences the punching load. For slabs with distributed
Figure 8: Variation of the inclination of the failure surface
with the
ratio Blh. reinforcement crossing the failure surface,
the
nominal stress can be taken as proportional to the
cube root of the ratio of reinforcement, but in other
cases a general description of this influence is
2
uncertain.
The ability to relate resistance to the geometry of
\\! \
the failure should be useful in treating punching of
curved surfaces (e.g.pipes and voided slabs), punch-
ing by loads close to supports and punching in the
?= 1
presence of shearreinforcement, wherevarious
a' failure surfacesmay need to be considered.
The empirical relationship between resistance and
the angleof failure is similar to what can be obtained
from asimplification of plastic theory.It hasthe
advantage of mathematical simplicity and avoids the
I l I I I I prediction given by plastictheorythat resistance
0 1 2 3 4 5 6
decreases indefinitely with increasing distance from a
Cot H
load to a support, i.e. that the minimum strength
Figitre 9: Variotions of colculoted punching resistance with cot 0. corresponds to the smallest value of 8 which is geo-
metrically possible.

REFERENCES
The variation of the critical value of cot 8 so cal-
l. Behaviourofreinforcedconcreteflatslabs,
R E G A N . P. E .
culated with Blh is shown in Figure 8 and does not
London.ConstructionIndustryResearchandInformation
seem altogetherrealistic for high values of Blh. How- Association. 1981. pp. 9 0 . CIRIA Report 89.
ever. the variation of P, with cot 8 is very modest for 2. HESS. U . JENSEN. B. C.. BRAESTRUP. M . W . , N I E L S E N . M. P. and
all but thesmallest values of cot 8. Figure 9 illustrates BACH. F. Gennemlokning af jernbetonplader. (Punching of
this variation for Blh = 1,2 and 4 and it can be seen reinforced concrete slabs.) Lyngby. Technical University of
Denmark,StructuralResearchLaboratory. 1978. pp. 63.
use of a fixed value of
that it is not surprising that the
Report Nr R90.
2.5 for cot8 gave good resultsin reference 1. For very 3. B R A E S T R U P , M. W . . NIEI.SEN. M . P,. JENSEN. B . c . and BACH. F .
small loaded areas, the use of a constant cot 8 = 2.5 Axisvmmetricpunching of plainandreinforcedconcrete,
gives theoretically unsafe results. A simple and con- Lyngby. Technical University of Denmark. Structural
servative approximation is obtained if the resistance Research Laboratory. 1976. pp. 33. Report Nr R75.
calculated for cot 8 = 2.5 is multiplied by (0.8+0.2 Contributions discussing the above paper shouldbe in the handsof
Blh) when Blh is less than unity. the Editor not later than30 September 1984.

x
Downloaded by [ York University] on [17/09/16]. Copyright © ICE Publishing, all rights reserved.