Documentos de Académico
Documentos de Profesional
Documentos de Cultura
A recent University of Toronto test of a 925 mm (36.4 in.) deep up to 1.1 m (43 in.), and by Iguro et a1. 6 and Shioya and
beam without stirrups showed a shear strength Vc that is only Akiyama7 for depths up to 3.0 m (118 in.). A very systematic
slightly below the value Vc = 2."ff~ required by ACI 318-08, and size effect for beams (of the highest brittleness number so
comfortably above the value rjJ x 2."ff~ (where rjJ = 0.75 is the strength far) was demonstrated by tests of reduced-scale beams at
reduction factor, and Vc and f~ are in psi). On that basis, and in Northwestern University.8 Recently, University of Toronto
view of the safety provisions of the Code, it is often thought that the tests 3.4 of three-point-bend beams without stirrups that were
current shear strength provisions for beams up to 0.2 m (8 in.) approximately geometrically similar and had depths ranging
deep, which neglects the size effect, are safe for beams up to 1 m from 0.11 to 1.89 m (4.3 to 74.4 in.), extended the experimental
(40 in.) deep. This is not true, however, for it must be expected that evidence of size effect and showed that the strength of the
if numerous tests of 1 m (40 in.) deep beams with different shear largest test beam was 53% less than the nominal strength
spans and steel ratios, made of different concretes and under
different hygro-thermal conditions, could be carried out, the beam according to ACI 318-08.9 To guard against such a situation,
strength would exhibit a similar statistical scatter, with approxi- Section 11.4 of ACI 318-08 9 severely penalizes any beams
mately the same coefficient of variation (Co V), as the strength of without stirrups more than 254 mm (10 in.) deep by reducing
beams up to 0.2 m (8 in.) deep, for which there are numerous test the shear strength limit from Vc =2Vtc' (where Vc andfe' are
results in the database. Based on this expected scatter, it is shown in psi) to Vc = ~c' (in effect, this implies a size effect factor
that neglecting the size effect for beams up to 1 m (40 in.) deep is of 2) (refer to Section 11.4.6.1 in ACI 318-08 9).
likely to increase the expected frequency of failures from approxi-
mately 1 in a million to approximately 1 in a thousand when the
In one test series at the University of Toronto,3.4 a single
beam depth increases from 0.2 to 1 m (8 to 40 in.). beam was tested for each size; see the diamonds in Fig. l(a),
where d is the effective beam depth (from top face to the
Keywords: deep beams; shear strength; size effect; stirrups. centroid of longitudinal reinforcement at the bottom) and vc
=Vu/bw<J is the nominal shear strength measured (VU is the
INTRODUCTION applied shear force and bw is the beam width). The figure
Although the basic theory of size effect in the shear failure also shows the horizontal line of vc =2Vtc', which represents
of reinforced concrete beams was formulated more than two the nominal strength, that is, the design shear strength, which
decades a 0 and experimental evidence has become over- must exceed the effect of design loads multiplied by their
9
whelming, ,2 the ACI 318 Code has not adopted size effect load factors and divided by the understrength factor $ for
provisions for beams of depths d up to 0.6 m (24 in.) and shear, which is 0.75 according to ACI 318-08.9 The load
even 1 m (40 in.). In support, a recent experiment (Specimen factor is in this figure considered as 1.6, which applies to the
BNl00 at the University of Toront03,4) was invoked, in which live load (refer to a following comment on the combinations of
the strength of such a beam was almost equal to the nominal live and dead loads).
strength required by the Code and was much larger than the Note in Fig. l(a) that all the data points (plotted as diamonds)
strength obtained after applying the understrength (or strength except the last one, that is, all those up to the depth of 1 m
reduction) factor </>. The purpose of this paper is to show (40 in.), lie above the horizontal line of Vc = $ x 2'lfc', where
that such suggestions are unjustified and could likely lead to 2Vtc' is the nominal strength required by the standard ACI
statistically dangerous designs with insufficient safety margins. 318-08. 9 Based on this observation, it has often been
suggested that the size effect need not be taken into account
RESEARCH SIGNIFICANCE for beam depths up to 1 m (40 in.) and that any considerations
The understanding of failure probability is essential for of size effect might simply be avoided by banning beams
improving the design provisions for shear failure of reinforced without stirrups having a depth over I m (40 in.). lithe full
concrete. The importance of this problem is demonstrated by picture is considered, however, it transpires that this suggestion
a number of disasters in which the size effect in shear failure is imprudent, in several respects (note thatfc' is taken as 70%
has recently been shown to have played a role. If the size effect of the required average compressive strengthfcr' from standard
is ignored or not predicted correctly, the failure probability tests, which approximately corresponds to ACI 318-08,9
becomes higher than what the risk analysis experts consider Section 5.3.2.2; the fact that the 70% reduction must be
as acceptable. Because of a trend to larger structures, this is
an issue of paramount significance for concrete engineering.
HOW TO INTERPRET DATABASE FOR SIZE ACI Structural Journal, V. 106, No. I, January-February 2009.
MS No. S-2006-479.R2 received August 22, 2007, and reviewed under Institute
EFFECT IN BEAM SHEAR pUblication policies. Copyright 2009, American Concrete Institute. All rights reserved,
The size effect for beams without stirrups was experimentally including the making of copies unless pennission is obtained from the copyright proprietors.
Pertinent discussion including author's closure, if any, will be published in the November-
demonstrated by KaniS for beams with an effective depth d December 2009 ACI Structural Journal if the discussion is received by July I, 2009.
(a) Toronto tests (199B2000) (b) Entire database (c) Portion of database
f~ ", SJ65 psi for small size range
{ in 12 in.
,
" spt.'Cificd compressive
strength
36.4 in.
d(in.) ' 00
Fig. I --{a) Ullil'ersity of Toronto lest~4 of sltear strenglh of beams of I'urious sizes:
(b) ACI 445F (f(llllba.~e of 198 data poims: and (e) pOT/iOIl of lite (fa /abase for
beamsfrolll4 w f2 ill. (JOO to 300 mm) deep (ve' f';. and f[ are ill psi). (Note: ! ill. =
25.4 111111: I MP(I = /45 psi.)
,"
5"",/ all the intervals in the histogram. Herein, six intervals, labeled
edf edf
0.5 0.5
by 1,2,3,4,5,6, and 7, are considered. They contain 18, 106,
, Normal 107,32, 13, and 1 data points, respectively (referto the histogram
distribution
in Fig. 2(e)). Compared with the frequencies corresponding
Dn = 0.078 Dn = 0.056 to the normal distribution (dashed curve), one obtains L(n;-
eile; = 20.95, which cannot satisfy the critical value Co 953 =
v,/g (e) Chi-square test
v,.ig 7.81 for 5% significance level. On the other hand, L(n; '- eil
0.5 Lognormal e; = 3.45 is obtained for log-normal distribution (solid curve),
distribution which satisfies the critical value for the 5% significance level.
(/)
Ql
The foregoing comparisons demonstrate that the log-
'u
C normal pdf is the best choice for the small beam data from
Q) 0.25
:::>
0'
the ACI 445F database.
~
lL.
WHAT STATISTICAL STRENGTH DISTRIBUTION
MUST BE EXPECTED FOR LARGE BEAMS?
Again, theoretical deductions based on the scatter in one
and the same material l ? are inapplicable because this scatter
Fig. 2-{a) Cumulative histogram 0/ data on normalized beam is overwhelmed by the scatter due to random variability of
shear strength/or small beams extracted/rom the ACI 445F steel ratio, shear span ratio, etc., in the ACI 445F database.
database, plotted on normal probability paper, and their As emphasized by Bazant and Yu,I,2 the database is
straight-line fit; (b) ditto on log-normal probability paper; heteroscedastic in the plot of normalized shear strength vclYrcr'
(c) K-S test/or normal distribution; (d) K-S test/or log-normal (resistance) versus size, but becomes nearly homoscedastic in
distribution; and (e) Chi-square test/or goodness a/fit. the doubly logarithmic plot; in other words, the variance or
.......
Expeeted res-stance