Académique Documents
Professionnel Documents
Culture Documents
The size-effect method for determining material fracture characteris- spent in studying its mechanical properties and struc-
tics, as previously proposed by Bazant and extensively verified for tural behavior .1-4,6,7,9-12 Nevertheless, many aspects, such
normal strength concrete, is applied to typical high-strength con-
crete. Geometrically similar three-point bending specimens are tested
as fracture behavior, need much more detailed in-
and the measured peak load values are used to obtain the fracture vestigation.
energy, the fracture toughness, the effective length of the fracture A complicating feature in the fracture analysis of
process zone, and the effective critical crack-tip opening displace- brittle heterogeneous materials such as concrete is non-
ment. The brittleness of the material is shown to be objectively quan- linear behavior. This is due to the fact that the fracture
tified through the size-effect method. Comparing the material frac-
ture properties obtained with those of normal strength concrete shows
process is not concentrated at a point, the crack-tip, but
that an increase of 160 percent in compressive strength causes: (1) an is distributed over a zone whose size is not negligible
increase of fracture toughness by only about 25 percent, (2) a de- when compared to the dimensions of the body. The ex-
crease of effective fracture proces zone length by about 60 percent, istence of the large fracture process zone is manifested
and (3) more than doubling of the brittleness number, which may be in the size effect exhibited by concrete specimens and
an adverse feature that will need to be dealt with in design. The brit-
structures, which is considerable but not as strong as in
tleness number, however, is still not high enough to permit the use of
linear elastic fracture mechanics. The R-curves are demonstrated to linear elastic fracture mechanics (LEFM). The condi-
derive according to the size-effect law exclusively from the maximum tions for which LEFM is applicable to concrete are at-
loads of specimens of various sizes and yield remarkably good pre- tained only for extremely large test specimens, testing
dictions of the load-deflection curves. of which would be very costly. However, material frac-
ture properties which are unambiguous and, especially,
Keywords: brittleness; cracking (fracturing); crack propagation; energy; frac-
ture properties; high-strength concretes; load-deflection curve; models. size- and shape-independent, can be defined only on the
basis of a very large specimen or, more precisely, by
Concretes of strengths exceeding 80 MPa (12,000 psi) means of extrapolation to a specimen of infinite size.
are now commonly being used in the construction of The simplest method to obtain size-independent ma-
high-rise buildings and offshore structures. 1.5 The utili- terial fracture properties is perhaps provided by extra-
zation of concrete of such high strength has been polating to infinite size on the basis of the size-effect
spurred on by the superior mechanical properties of the law proposed in References 13 and 14 (see also Refer-
material and its cost-effectiveness. 2,6 The high-strength ence 15). This law approximately describes the transi-
concrete of today is a highly engineered material with tion from the strength criterion, for which there is no
several chemical and mineral admixtures. It is possible size effect, to LEFM criterion, for which the size effect
to obtain workable mixes with very low water contents is the strongest possible. This law has been shown to
by using superplasticizers and retarders. Pozzolanic ad- agree well with concrete fracture tests in Mode 1,16 as
ditives such as fly ash and silica fume are employed to well as Mode 1117 and Mode III. 18,19 A good agreement
alter the hydration reactions beneficially and also to fill was also demonstrated for certain ceramics,20 rocks,21,22
the microscopic voids between cement particles. Small, and aluminum alloys.23
round aggregates are used to achieve better mixing and An important advantage of the size-effect method is
attain higher surface areas for bonding. Typical high- its simplicity. The method requires only the maximum
strength concrete has a very high-strength matrix, is load values of geometrically similar specimens that are
more compact, and possesses well-bonded aggregate-
mortar interfaces. 7 ACI Materials Journal, V. 87, No.6, November - December 1990.
Received Oct. 31, 1989, and reviewed under Institute publication policies.
In the past, research on high-strength concrete has Copyright 1990, American Concrete Institute. All rights reserved, induding
primarily concentrated on increasing its "strength."8 In the making of copies unless permission is obtained from the copyright propri-
etors. Pertinent discussion will be published in the September - October 1991
the last decade, however, considerable effort has been A CI Materials Journal if received by June I, 1991.
P 1(0'.)
G (2)
sufficiently different in size. Such measurements can be bJd
carried out even with the most rudimentary equipment.
Neither the post-peak softening response nor the true where P = applied load, 1(0'.) is a geometry-dependent
crack length need to be determined. Another advantage function of relative crack length 0'. = old, a = crack
of the size-effect method is that it yields not only the length, g(O'.) = [f(0'.)j2, E' = E for plane stress, E' =
fracture energy (or fracture toughness) of the material, EI(1 - u2 ) for plane strain, E = Young's modulus of
but also the effective length of the fracture process elasticity, and u = Poisson's ratio. Values of 1(0'.) can
zone, from which one can further obtain the R-curve be obtained by elastic analysis techniques, such as the
and the critical effective crack-tip opening displace- finite element method, and for basic specimen geome-
ment. Determination of the fracture process zone size tries, formulas for function 1(0'.) can be found in frac-
is of particular interest for high-strength concrete be- ture mechanics handbooks (e.g., Tada, Paris, and Ir-
cause (for a variety of reasons) such concrete is sus- win,25 and Murakamj26).
pected to be more brittle and, therefore, to have a When g' (0'.) > 0 [where g' (0'.) is the derivative of g(O'.)
smaller fracture process zone than normal strength with respect to 0'.), which applies to most situations,
concrete. LEFM indicates that the maximum load occurs at in-
In view of the aforementioned reasons, the size-ef- finitesimal crack extensions. Therefore, 0'. at maximum
fect method has been adopted to present experimental load is practically the same for bodies of different sizes.
investigation of high-strength concrete. The objectives Setting G = Gf (fracture energy) or K J = KJc (fracture
will be to obtain the fracture energy and the process toughness or critical stress intensity factor) along with
zone size. The third objective will be to investigate P = P u , Eq. (2) according to LEFM yields the depend-
whether the R-curves derived solely from maximum ence of aN on size, which is aN = constant! Jd. As is
load data yield load-displacement curves that agree well known, LEFM criteria govern only the fracture
sufficiently well with measurements. behavior of very large concrete structures.
In common-size concrete structures and specimens,
REVIEW OF THE SIZEEFFECT LAW AND ITS the fracture process zone that forms in front of a prop-
IMPLICATIONS agating crack affects the behavior significantly. With
Size effect increasing load, this zone grows in size while remaining
Structures and test specimens of brittle heterogene- attached to the notch tip [provided that the specimen
ous materials, such as concrete, rock, and ceramics, geometry is such that g' (0'.) > 0]. The process zone
exhibit a pronounced size effect on their failure loads. shields the propagating crack tip, and thereby increases
This phenomenon, which is an important consequence the fracture resistance. The nonlinear fracture regime,
of fracture mechanics, has been described by the size- where the influence of the process zone is dominant,
effect law proposed by References 13 and 14 lies between behavior governed by limit analyis and
LEFM. As a result, there is a considerable size effect on
Blu d the failure of normal concrete structures and speci-
/3=- (1)
../1+/3' do mens, but it is not as strong as that of LEFM. This
transitional size effect is described by Eq. (1).
which applies to geometrically similar structures (or The size-effect law [Eq. (1), giving the approximate
specimens) of different sizes: aN = cnPJbd = nominal relation of aN to (3 (the brittleness number), is plotted in
stress at maximum load P u (ultimate load), b = thick- Fig. 1. For large (3, such as (3) 10, Eq. (1) gives (with
ness of the structure or specimen identical for all sizes an error under 5 percent) the approximation aN ex d- y, ,
ACI Materials Journal I NovemberDecember 1990 609
concrete depends on the shape of the softening stress-
displacement or stress-strain relations. For concrete, the
rati0 20 seems to be about 2. However, the actual length
of the process zone need not be known for the calcula-
tions.
Also, for an infinitely large specimen, the fracture
toughness K lc can be obtained as
(5)
X = d, Y = (jja N ) 2,
(14)
B = I/vtcanddo = CIA (10)
The size range of the specimens used in the regression The foregoing relations [Eq. (13) and (14)] are valid as
analysis must be sufficiently large in relation to the long as the fracture process zone remains attached to
random scatter of material properties and test meas- the tip of the initial crack or notch. This ceases to be
urements. For the typical scatter of concrete, the mini- true after the peak load. 22 Since the fracture process
mum size range recommended is 1:4. zone in the post-peak regime is detached from the tip
The size-effect method can also be used to determine (separated from the initial tip by a traction-free crack)
fracture properties other than those already defined. and is advancing ahead with approximately constant
Considering the infinitely large specimen again, ma- size, it dissipates roughly the same amount of energy
terial parameters such as the effective (elastically per unit crack extension. Consequently, the critical
equivalent) crack-tip opening displacement at the peak value of G after the peak load must be kept constant
load, oef' can be related to cf and K lc using LEFM and equal to the values of R reached at the peak load.
relations 22 The effective R-curves given by Eq. (13) and (14) are
size-independent until the peak load but deviate after-
(11) wards into size-dependent horizontal branches. The
size-dependence of R-curves in the post-peak regime
has been previously observed for concrete36 and other
The value of oef pertains to the effective opening when materials. 22 ,23,37 These effective R-curves tend to move
the stress at the notch (or initial crack) tip is just re- toward complete coincidence with the actual R-curve as
duced to zero, and the fracture process zone is about to the size increasesY
detach and advance away from this tip.
The size of the fracture process zone can also be EXPERIMENTAL INVESTIGATION
characterized by the length parameter Test details
High-strength concrete used in the construction in-
dustry varies widely in composition and strength. Even
(12)
the demarcation between high and normal strength
concretes is subjective and appears to be continually
rising. 2,g To study typical high-strength concrete used in
first used by Irwin 30 for the size of the yielding zone in the industry, this investigation was conducted on ma-
ductile fracture. For concrete, this length parameter terial obtained directly from a batch mixed for the con-
was introduced by Hillerborg, Modeer, and Peters- struction of a high-rise building in downtown Chicago.
son. 31 The parameter Co is a size-independent measure The concrete mix was designed to exceed a 28-day
of the intensity of the toughening or crack-tip shielding compressive strength of 83 MPa (12,000 psi). Cement
mechanisms. It can be used to compare the brittleness of much higher quality than standard (ASTM C 150)
of one material with another. A material with a rei a- portland cement was selected. Silica fume (micro silica)
ti vely low Co would be more brittle than a material with and fly ash were used as mineral admixtures. The max-
a higher Co. imum aggregate size d a in the mix was 9.5 mm (3/8 in.).
The details of the mix composition are given in Table
R-curves 1. The proportions, by weight, of cement: sand: gravel
Another important consequence of the fracture : water: fly ash: micro silica were 1 : 1.35 : 2.42 : 0.35
process zone is the fracture resistance curve or R-curve. : 0.25 : 0.04. The fresh concrete had a unit weight of
Since the fracture process zone evolves as it propa- 2410 kg/ml (15Ilb/ftl) and a slump of 235 mm (9.3 in.)
gates, the resistance R(c) to fracture growth, represent- as measured with a standard 305 mm (12 in.) cone.
ing energy dissipated per unit length and width of frac- Along with the fracture specimens, three 152 x 304
ture extension, gradually increases. The function R(c), mm (6 x 12 in.) cylinders and three 102 x 204 mm (4
called the R-curve, was initially proposed to be used as x 8 in.) cylinders were cast. These cylinders were cap-
a material property,n,33 independent of the shape of the ped with a sulfur compound and cured in a fog room.
ACI Materials Journal I November-December 1990 611
Table 1 - Concrete composition
Volume of mix Per 0.8 m' (Per I yd')
Cement, ASTM Type I 363 kg (800 Ib)
Sand, FA2 490 kg (l080Ib)
Gravel, CAI5 880 kg (1940Ib)
(crushed limestone)
Water 127 kg (280Ib)
Fly ash, Class C 91 kg (200 Ib)
Microsilica 16 kg (35Ib)
Retarder, naphthalene-based I e (350z)
High-range water reducer 5.H (l800z)
direction
l
i of casting
3000
~
//\\
d= 152.4 mm
z
-0 2000 -
-0
o
o
2000
1/ l o
o
V\ ~76.2 mm 1000
d=38'1~~. ~~
1000
mm ~ ~
+------~-,----------,-------,_
0.05 0.10
~
0.15
J
0.20
o
0.00
~-T--,------___r__------,----I-~~ ----,---~--,----------1
0.02 0.04
crock mouth opening displacement (mm)
0.06 0.08 O. Ta
Fig. 6 - Typical load-displacement curves for differ- Fig. 7 - Typical load-CMOD curves for different sizes
ent sizes
Table 2 - Summary of test results
the beams were tested 14 days after casting and with Specimen Peak Deflection Final Peak
CMOD rates such that they reached their peak loads in dimensions, * load, at peak load, deflection, t CMOD, Eo. W o,
mm X mm N mm mm mm GPa N-mm
about 10 min. Typical curves for different specimen
38.1 X 101.6 1668 NR NR 0,0127 6.0 NR
sizes are shown in Fig. 6 and 7, and the raw data are 1490 0,0610 0.1092 0,0114 7.0 74
given in Table 2. 1735 0.0914 NR 0.0146 5.5 NR
The test results from the largest beams were, how- 76.2 X 203,2 2357 0.0864 0,1727 0.0222 7,7 167
2091 0.0686 0,1626 0.0152 8,2 143
ever, excluded from the following analysis because their 2535 0.0711 0.1422 0,0292 9.0 142
load values were inconsistent with the trend of the other 152.4 X 406.4 3825 0.1143 0.2184 0.0381 9.2 362
3736 0.0940 0,3200 0.0437 6.2 418
tests. In retrospect, this inconsistency may be due to the 3825 0,1710 0.2388 0.0356 12.7 326
difference in control parameters used for testing and in 304.8 X 812.8' 7072
machine characteristics. Under stroke control, which 7442 NM NM NM NM NM
7695
was used only for the largest beams, it was not possible
All specimens 38, I mm wide,
to obtain a stable response at the peak load. The insta- , Deflection when the load drops to 10 percent of peak load.
I Stroke-controlled tests, not used in analysis.
bility of crack propagation may have caused a high Eo = Young's modulus from initial load-deflection compliance.
crack propagation rate near the peak load. This might W, = area under load-deflection curve.
NM = values not measured.
explain why the peak loads of these specimens were NR = values not recorded due to failure of instrumentation.
higher than expected.
The modulus of rupture of the material was used as mainly along the aggregate-mortar interfaces. The re-
the measure of its strength. Four specimens, each with duced crack deflection by the aggregates is due to the
a depth of 51 mm (2 in.) and a span of 152 mm (6 in.), strong aggregate-mortar bond and the fact that the
were cut from two of the 152 mm (6 in.) deep fracture strength of the matrix approaches that of the aggre-
specimens that had already been tested. These un- gates. Obviously, this near-homogeneous behavior de-
notched beams were tested to failure in three-point creases the width of the process zone as well as the
bending under stroke control in the 89 kN (20 kip) load shielding effect of the aggregates.
frame. From the peak loads the modulus of rupture of
each specimen was calculated. The average valuefu was Analysis
found to be 11.0 MPa (1600 psi), with a standard de- Linear regression analysis (Fig. 8), using Eq. (9) and
viation of 5 percent. Carrasquillo, Nilson, and Slate 7 (10), yields coefficients A = 1.58 mm- I (40.1 in.- I ) and
proposed an empirical formula for the modulus of rup- C = 51.7. The coefficient of variation of the devia-
ture of high-strength concrete: f, = 0.94vJ:'(stresses in tions from the regression line is w = Sy/xlY = 15 per-
MPa). Using the value of f:14 in their relation, f, was cent, where Sy!X is the standard error of the Yestimate
obtained as 8.7 MPa (1260 psi). The difference in the and Y is the mean of the experimental Y-values. From
values of fu and f, may partly be due to errors in the Eq. (10), the parameters of Eq. (1) are computed as B
predictions of the formula (known to be up to 15 per- = 0.139 and do = 32.8 mm (1.29 in.). Nonlinear re-
cent) but it might also be due to the significant size ef- gression analysis directly for B and do, using the Mar-
fect which is known to exist in flexural tests on plain quardt-Levenberg algorithm, results in B = 0.198 and
concrete beams. 14,38 do = 13.3 mm (0.52 in.) with w = 8.4 percent. Since
As noted previously by several researchers,7,9,10,39 the nonlinear regression provides a better fit, the latter
cracks in high-strength concrete were seen to propagate values for the size-effect parameters are used hence-
through the coarse aggregates. This is an important forth.
difference from normal strength concrete, in which Fig. 9 shows the size-effect curve [Eq. (1)] based on
cracks (in the vicinity of gravel particles) propagate the calculated parameters along with the experimental
ACI Materials Journal I November-December 1990 613
3.5,.-------------------,
300 E2l Normal Strength Concrete
250
.- 3 ................. .
High Strengtti Concrete
(f) 2.5
C1)
200
:::s
--
N
~
z co 2
b >
"",,150 C1)
~
'+-
'-' .=:: 1.5
Y = 1.58 X + 51.7 10
100
I
fu = 11.0 MPa
~ 1
50
25 50 75 100 125 150 175 0.5
specimen depth, d (mm)
o
Fig. 8 - Linear regression for size-effect parameters
f'c
0.0 ,---~--------------,
ACKNOWLEDGM ENTS
1200
,.----., This work was partially supported by the Center for Advanced Ce-
Z ment-Based Materials at Northwestern University (NSF Grant DMR-
-u 8808432). The underlying basic theory was developed under partial
0 800 support from AFOSR Contract 49620-87-C-0030DEF with North-
0
western University. Summer support for M. E. Karr was provided
under NSF Grant BCS-8818230. The authors are grateful to Jaime
400 Moreno and Arthur King of Material Services Corporation, Chicago,
Ill., for donating the concrete used in the experiments.
00000 Experiment
- - Prediction