Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Engineering Fracture Mechanics 76 (2009) 286298

Contents lists available at ScienceDirect

Engineering Fracture Mechanics


journal homepage: www.elsevier.com/locate/engfracmech

Effect of aggregate shape on the mechanical properties of a simple concrete


C.G. Rocco a, M. Elices b,*
a
b

Area Departamental Construcciones, Facultad de Ingeniera, Universidad Nacional de La Plata, Calle 48 y 115 s/n, 1900 La Plata, Argentina
Departamento de Ciencia de Materiales, E.T.S. de Ingenieros de Caminos, Universidad Politcnica de Madrid, c/Profesor Aranguren s/n, 28040 Madrid, Spain

a r t i c l e

i n f o

Article history:
Received 4 March 2008
Received in revised form 23 July 2008
Accepted 1 October 2008
Available online 25 October 2008
Keywords:
Concrete
Cohesive model
Fracture energy
Aggregate shape
Softening function

a b s t r a c t
The inuence of aggregate shape on the fracture energy, tensile strength and elasticity modulus in concrete is considered. For this purpose, eight simple cement-based composites
were designed, manufactured and tested, with two purposes: to provide experimental data
that can throw some light on this involved problem and help in the design of future
cement-based composites, and supply information that can be used as a benchmark for
checking numerical models of concrete failure, as this simple composite is amenable to
being modelled quite easily. Thirty-six notched beams were tested and values of the fracture energy and elasticity modulus were recorded. The tensile strength was measured from
indirect standard tensile tests. Comparison with available experimental data is also
included and discussed. Fracture was modelled using a cohesive crack with a bilinear softening function; data of the softening function inferred from the experimental measurements are also provided and discussed.
2008 Elsevier Ltd. All rights reserved.

1. Introduction
The strength of concrete is assumed to depend primarily on two factors: the water to cement ratio and the degree of compactation. Even so, the shape of aggregates has an inuence on the concrete strength and this is the subject of this paper. It is
known that the stress at which cracks develop depends largely on the shape of the coarse aggregate; smooth gravel leads to
cracking at lower stresses than rough and angular crushed rock [1]. Also, as coarse aggregates act as crack arresters, so that,
under an increasing load, another crack is likely to open, the shape of the aggregates should inuence the fracture energy of
concrete.
All these matters deserve to be evaluated, as there are few references on this subject. The effect of aggregate shape on the
fracture and mechanical properties of concrete was investigated in standard concretes, i.e. concretes made with graded
aggregates: Saouma et al. [2] and Li et al. [3] conducted an investigation on dam concretes, with large aggregates; rounded
and crushed. Donza and Cabrera [4], investigated the role of the shape of ne aggregates; natural sand and crushed sand.
Research along the same lines was performed by Kim et al. [5], as the use of crushed sand as a ne aggregate has rapidly
increased due to a shortage of river sand. Tests on concretes made with rounded and crushed aggregates, between 5 and
7 mm, were also performed by Guinea et al. [6].
Here, measurements of fracture parameters and mechanical properties from very simple concretes made with spherical aggregates with different diameter for each type of concrete, and crushed aggregates obtained from the spherical
ones are reported. The difference and the advantage here with respect to the quoted references is that this concrete
is amenable to being numerically modelled, due to its simplicity and because the few parameters needed are known
(properties of the matrix, aggregates and interfaces). These results, although not directly applicable to graded concrete,

* Corresponding author. Tel.: +34 915 43 39 74; fax: +34 915 43 78 45.
E-mail addresses: claudiorocco@sinectis.com.ar (C.G. Rocco), melices@mater.upm.es (M. Elices).
0013-7944/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2008.10.010

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

287

Nomenclature
C
C6
C9
E
ft
f(w)
GF
lch
M2
PA
PBA
R
S3
S9
S14
w1
wc

w

crushed coarse aggregate (3 mm average size)


crushed coarse aggregate (6 mm average size)
crushed coarse aggregate (9 mm average size)
elasticity modulus
splitting strength
softening function
fracture energy
characteristic length EGF =ft2
mortar matrix (w/c = 0.42)
percentage of aggregates on the fracture surface
percentage of broken aggregates on the fracture surface
average roughness
spherical coarse aggregate (3 mm diameter)
spherical coarse aggregate (9 mm diameter)
spherical coarse aggregate (14 mm diameter)
initial crack opening (dened in Fig. 9)
critical crack opening
abscissa of the centroid (dened in Fig. 9)

Greek

rt

cohesive strength

can be protably used as a benchmark for checking numerical models intended to predict the fracture and mechanical
behaviour of concrete.
2. Testing programme
2.1. Materials
Eight different simple concretes were manufactured using the same mortar matrix, two types of aggregate-matrix
interface (strong and weak interfaces) and two aggregate shapes (round and angular). The features of the composite components as well as of the concrete are described below:
Aggregates: Commercial spheres of mullite were used as aggregate. The average diameter of the spheres was: 3 mm,
9 mm, and 14 mm, henceforth labelled as S3, S9 and S14. As the aggregate strength strongly depends on heat treatment, in
order to achieve similar strength for all sizes, the spheres were heat-treated for 5 h at the following temperatures; 525 C
for B3, 625 C for B9, and 800 C for B14. Additional information on chemical composition and processing can be found in [7].
Angular aggregates were obtained by crushing mullite spheres of 14 mm diameter, using a small commercial crusher. The
mean size was dened as the arithmetic mean of the sieve size at which the particle is just retained and the sieve size
through which the particle just passes. The sieves considered were: 2.0, 4.0, 8.0 and 10.0 mm and the used aggregates were
those which passed between 2 and 4, between 4 and 8, and between 8 and 10, henceforth labelled as C3, C6 and C9.
Matrix: One type of mortar matrix, identied as M2, was used. The water/cement ratio was 0.42. Mortar was made with a
type III Portland cement (according to ASTM standards) and a natural ne siliceous aggregate of a size between 0.2 and
0.4 mm. Silica fume was added to increase the strength of the mortar, and superplasticizer (Sikament 300) to improve
its workability.
The reason for choosing a water/cement ratio of 0.42 was because the shape of the coarse aggregates inuence on the
strength of concrete depends on this ratio and its effect is greater at lower values. For water/cement ratios below 0.4, the
use of crushed aggregate provided strengths up to 38% higher than when rounded aggregates are used. With an increase
in the water/cement ratio, the inuence of aggregate shape falls off and, at a water/cement ratio of 0.65 no difference in
the strengths of concretes made with crushed rock and gravel was observed [1,8,9].
Matrixaggregate interface: Two kinds of matrixaggregate interface weak and strong were used. Weak interfaces
were achieved by daubing the spheres with a release agent, whereas strong interfaces were naturally obtained from the matrix and aggregate interaction. Here, strong is used in the sense that the adherence between matrix and aggregates is stronger than when aggregates are smeared with the release agent. We refer to weak and strong interfaces as treated and
untreated types.
Concrete: The composition of the eight concretes is shown in Table 1. In all cases the volume of spherical aggregates was
25.8%. To make the concrete, the matrix was prepared rst using a mechanical mixer and then the aggregate was added and
stirred by hand to avoid any damage. Details of this process and of the aggregate interface treatment appear in [7].

288

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Table 1
Composition of the model concrete (kg/m3).
Cement

Water

Sand

Silica fume

Plasticizer

Coarse aggregate

470

200

908

30

15

500

2.2. Specimens
Beams of 100  100  450 mm were cast for bending tests. Central notches were milled with a diamond steel disc,
1.5 mm wide. The notch depth was always 30 mm. Beam geometry is shown in Fig. 1a. For each type of concrete, four beams
were tested. All in all, 36 beams were cast; eight kinds of concrete plus the mortar.
Metallic casting forms were greased with a release agent and xed to a vibration table before the start of casting. Curing
was done in the following way: watertight sheeting was placed over the specimens. They remained in this water-saturated
atmosphere at 20 C for 24 h, to avoid shrinkage and cracking. Then, the specimens were taken out of the forms, marked, and
immersed in the laboratory water pools (lime saturated water at 20 C) up to 60 days, until notching and immediate testing.
Cubes of 100  100  100 mm, for indirect tensile tests, were manufactured from the broken parts of the notched tests, as
shown in Fig. 1b.
2.3. Testing procedures
Two kinds of test were done: stable three-point bend tests with prismatic notched specimens, as recommended by RILEM
TC-50 [10], and indirect splitting tests with cubic specimens. In the bending test, possible sources of experimental errors, as
detailed in [1114] were taken into account.
Bending tests were performed in an 1MN servo-hydraulic testing machine INSTRON 1275, run in crack mouth opening
displacement (CMOD) control mode, recording the crack opening, the load and the displacement of the loading point. Load
was measured with a 5 kN load cell of 5 N resolution and 0.25% accuracy. An extensometer, of 2.5 mm and 0.2% error at full
scale displacement, was used to measure the CMOD. The load-point displacement was measured by two transducers of
2.0 mm and 0.2% error at full scale. A discussion on the stability of the notched beam test can be found in [15].
Cubic specimens cut from the tested beams were used for splitting tests which were done under displacement-mode control following the recommendations of ASTM C 496, except for specimen size and width of the load-bearing strip, which was
4 mm. Previous work by the authors [16] had shown that the standard width of the bearing strip 16% of the specimen
depth is usually over-dimensioned and can lead to an erroneous estimation of tensile strength. We kept the relative width
of the bearing strip between 4% and 10%.

Fig. 1. Specimen geometries and loading set-up.

289

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

The fracture energy of the spherical aggregates, B14, was measured using notched beam tests. Here, we used cylindrical
bars where the central part was an aggregate glued to the bar, (as shown in Fig. 1c). The stress intensity factor was computed
according [17]. Tests were performed under displacement control at a rate of 1 lm/s, while load and relative displacement
were measured during testing; loads were measured with a 1 kN load cell of 1 N resolution and 0.25% accuracy. Displacements, relative to the initial load application point, were measured with inductive transducers of 2.0 mm range and 0.2%
error at full scale.
Tensile strength and elasticity modulus of spherical aggregates were measured by means of compression tests, using the
same procedure as in Refs. [7,18].
3. Experimental results
3.1. Notched beam tests
3.1.1. Fracture topography
After testing, the proportion of debonded and broken aggregates in the fracture surfaces was measured by image analysis.
The projected surfaces of both debonded and broken aggregates on the fracture surface were also measured. From these data
we drew up the following indexes (expressed in percentage), which give an idea of the type of fracture:

projected surface of aggregatesbroken and debonded


 100
projected broken surface
projected surface of broken aggregates
 100
PBA
projected broken surface
projected surface of bonded aggregates
PDA
 100
projected broken surface
PBA DA

1
2
3

All these results, as well as the percentage of broken aggregates, BA/(BA + DA) are summarized in Table 2.
When the percentage of aggregates P(BA + DA) is above the value of a concrete with homogeneously distributed aggregates 25.8% in this simple concrete this means that the crack has bypassed the aggregates and found its way mainly
through the matrix or the aggregatematrix interfaces, indicating transgranular fracture or extensive debonding.
Topographic analysis of the crack surfaces produced by the three-point bend tests was performed by means of a laser prolemeter. Twelve proles, 5 mm spaced and parallel to the crack front, were measured on one of the cracked halves of the
specimens. The prole points were recorded every 20 lm along the path, with a resolution of 3 lm in height. Fig. 2 shows
some representative fracture surfaces for concrete made with spherical aggregates (untreated and treated) and concrete
made with crushed aggregates (untreated and treated).
The average roughness R was calculated for every concrete batch. This standard amplitude parameter is conventionally
dened as [19]

1
L

jzxjdx

where L is length of the projected prole, and z(x) is the centre-line prole, which satises

zxdx

The mean roughness of each specimen was computed by averaging the 12 values one for every prole of R measured on
the specimen. The roughness of each concrete batch was obtained as the average of the four specimens tested per batch and
shown in Table 2, jointly with the normalized value of the actual fracture surface area Aa/Ap.

Table 2
Average values of percentage of broken (B) aggregates, debonded (D) aggregates, and fracture surface roughness (R).
Aggregate

Interface

P(BA + DA)%

P(BA)%

P(DA)%

R (mm)

Aa/Ap

S3
S9
S14
C3
C6
C9
S9
C6

Untreated
Untreated
Untreated
Untreated
Untreated
Untreated
Treated
Treated

25.2
32.8
28.5
24.7
29.4
29.0
30.5
32.2

25.2
32.8
27.6
24.7
29.4
29.0
5.1
23.7

0.0
0.0
0.9
0.0
0.0
0.0
25.4
8.5

0.64
0.70
0.71
0.58
0.70
0.78
1.49
1.30

1.13
1.14
1.14
1.19
1.19
1.18
1.28
1.37

Aa/Ap: actual fracture surface/projected fracture surface.

290

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Fig. 2. Fracture surfaces for concrete made with spherical aggregates; (a) untreated (strong matrixaggregate interface) and (b) treated (weak matrix
aggregate interface). Fracture surfaces for concrete made with crushed aggregates; (c) untreated (strong matrixaggregate interface) and (d) treated (weak
matrixaggregate interface).

3.1.2. Fracture energy


Measured values of the specic fracture energy, GF, according RILEM TC-50 proposal [10], are recorded in Table 3. The
values of the specic fracture energy GF were obtained by dividing the measured work of fracture by the projected area
of fracture according to the RILEM recommendation. Inspection of the fractured surfaces showed different degrees of roughness depending on the type of aggregate and matrixaggregate interfaces and hence different values of the actual areas,
although all surfaces had the same projected area. A knowledge of the fracture energy expended per unit of actual area, instead of the projected area, may cast some light on the fracture processes at the meso-level. Additional remarks on measurement procedures and meaning of GF can be found in [14,18], as well as about the fractal nature of the actual fracture surface
in [20].
Note that concrete GF values are always higher than the matrix one. It also appears that fracture energy increases as the
aggregate size increases. It is worth noticing the high values corresponding to crushed aggregates whose surface was treated
with a release agent to decrease adherence between aggregate and matrix interface.
3.1.3. Elasticity modulus
Measured values of the initial elasticity modulus, E, deduced from the load-displacement curves during the notched beam
tests, are also shown in Table 3.
For the simple concrete with untreated round aggregates, E slightly decreases as the aggregate size increases. For the simple concrete with crushed aggregates, this tendency is not so clear. What is evident are the lower values recorded for concretes with treated aggregates, regardless of their shape.

Table 3
Average values of measured mechanical properties.
Aggregate

Interface

Tensile strength (MPa)

Elasticity modulus (GPa)

Fracture energy (J/m2)

Peak loads (kN)

S3
S9
S14
C3
C6
C9
S9
C6
Matrix
Aggregates

Untreated
Untreated
Untreated
Untreated
Untreated
Untreated
Treated
Treated

3.25
3.00
3.10
3.32
3.35
3.19
2.19
2.10
3.26
2.80

24.5
23.4
22.9
26.5
27.9
24.1
20.3
21.4
27.0
6.0

40.0
43.3
48.3
41.1
44.4
48.3
41.2
59.7
34.0
60.0**

*
**

Not available.
Measured on S14.

*
*

3.9
4.2
3.8
*

3.3
*
*

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

291

3.1.4. Peak loads


Average values of recorded peak loads for each concrete type are also summarized in Table 3. These values may be helpful
for checking numerical models of concrete when the mechanical properties of the components are known.
3.2. Splitting tests
Average tensile strength, ft, of the simple concretes as measured using the splitting-tension test, according ASTM C496,
are given in Table 3. For concrete made with untreated aggregates, a decline of ft seems to take place as the aggregate size
increases. Simple concrete, where aggregate debonding was helped, shows a clear decrease of ft by comparison with untreated aggregates.
Average values of some mechanical properties of the matrix and aggregates are also summarized in Table 3.
4. Discussion and comments
4.1. Surface topography
The fracture surfaces of concrete made with strong aggregatematrix interface (untreated aggregates) show that almost all
aggregates are broken (index P(DA) is zero, except for concrete S14 where its value is less than 1%) regardless of aggregate
size and shape (see Table 2). The projected surface of broken aggregates on the fracture plane ranges from 24.7% to 32.8%.
Fracture surfaces of concrete made with weak aggregatematrix interface (treated aggregates) exhibit two different features depending on aggregate shape; concrete made with spherical aggregates show extensive debonding (index P(DA) is
25.4% and P(BA) only 5.1%, as indicated in Table 2) whereas fracture surfaces of concrete with crushed aggregates show
extensive aggregate fracture (index P(BA) is 23.7% and P(DA) only 8.5%, see Table 2).
An interesting nding related to the fracture surface of concrete made with treated angular aggregates was that, although
almost all aggregates were broken, many interfaces between matrix and aggregate were debonded (as shown in Fig. 3). Such
nding is indicative that two toughening mechanisms may be operative, at the aggregate level; a debonding mechanism and
another due to aggregate breaking. This behaviour does not occur with spherical aggregates due to their particular shape.
Average roughness, R, of the fracture surfaces is shown in Fig. 4a as a function of aggregate size. For comparison purposes,
the scatter band of the matrix roughness and additional values of concrete made with treated spherical aggregates are also
included, taken from [21]. As can be observed, the average roughness of concretes with untreated aggregates (strong interface) is similar to the matrix roughness and does not depend on the aggregate shape (see Fig. 2), although it seems that there
is a modest increase with aggregate size. Conversely, concretes made with treated aggregates (weak interface) exhibit an
appreciable increase in roughness value, regardless of the rupture type, i.e. broken or debonded aggregates (see, also Fig. 2).
Fig. 4b shows the actual fracture surface area, normalized to the projected fracture surface area, as a function of aggregate
size, for the eight considered concretes. Again, concretes made with untreated aggregates (strong interface) display no differences with respect to the matrix and appear to be independent of the aggregate shape or size. Once more, concretes made
with treated aggregates show a substantial increase of the fracture surface.
4.2. Tensile strength
Fig. 5 shows, for the two shapes of considered aggregates, the experimental results of the tensile strength ft (normalized
to the corresponding matrix value) as a function of the aggregate size. For concretes with strong matrixaggregate interface

Fig. 3. Debonded interfaces between matrix and aggregate.

292

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Fig. 4. (a) Roughness of the fracture surface as a function of aggregate size, for spherical and crushed aggregates, both with different kinds of matrix
aggregate interfaces. (b) Normalized fracture surface area as a function of aggregate size, for spherical and crushed aggregates, both with different kinds of
matrixaggregate interfaces.

the tensile strength was almost equal to the matrix tensile strength with a very low dependence on aggregate shape (angular
aggregates give values a bit higher than spherical ones) and aggregate size (a small decrease with increasing size may be
conceived). Concretes made with weak matrixaggregate interfaces show a different behaviour; both concretes, made with
spherical or angular aggregates, gave tensile strengths much lower than the matrix and these values do not appear to depend
on aggregate shape. This gure also shows the tensile strength results as a function of size for spherical aggregates with weak
interfaces, already reported in [21].
Other authors have found a similar trend [2,3,6]; a small increase in tensile strength when using crushed aggregates instead of round ones of the same size. Additionally, this effect can be found when considering the shape of ne aggregates, as
reported by Donza and Cabrera [4]. Experimental results from Refs. [24,6] are summarized in Table 4. Some time ago, Kaplan [22] observed that the exural strength was generally lower than the exural strength of corresponding mortar. Mortar
would thus seem to set an upper limit to the tensile strength of concrete, at least when the matrixaggregate interface is
weak or for rounded aggregates [21].
4.3. Modulus of elasticity
Fig. 6 shows, for the two shapes of considered aggregates, the experimental results of the initial modulus of elasticity E
(normalized to the corresponding matrix value) as a function of aggregate size. Concretes made with a strong matrixaggregate interface gave slightly lower values than the matrix value (except concrete C6, whose value was almost equal). It seems

293

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Fig. 5. Normalized tensile strength (with respect to the matrix) as a function of the aggregate size, for spherical and crushed aggregates, both with different
kinds of matrixaggregate interfaces.

Table 4
Effect of aggregate shape on mechanical properties of concrete.
Reference

Aggregate size (mm)

Aggregate type

ft (MPa)

E (GPa)

GF (J/m2)

lch (mm)

Saouma et al. [16]

38 (max)
38 (max)
540
540
5150
5150
57
57
57*
57*
Fine (S)
Fine (G)
Fine (G)
Fine (L)
Fine (D)

Round
Crushed
Round
Crushed
Round
Crushed
Round
Crushed
Round
Crushed
Round
Crushed
Crushed
Crushed
Crushed**

2.67
3.96
1.80
2.12
1.58
1.91
3.93
4.15
4.93
4.89
3.59
4.08
4.54
4.09
4.39

16.9
23.2
24.6
17.6
43.1
40.0
39.8
33.1
40.0
34.7
40.0
39.8
39.4
39.8
36.6

223
227
420
249
490
497
94.7
136.0
87.1
127.5

529
336
3189
1047
8460
5463

Li et al. [17]

Guinea et al. [18]

Donza and Cabrera [4]

(S) siliceous, (G) granite, (L) limestone, (D) dolomite.


With silica fume.
**
Elongate.
*

that there is a tendency for E to decrease with aggregate size and that concretes made with crushed aggregates provide E
values somewhat higher than concretes made with spherical ones. Concretes made with a weak matrixaggregate interface
show lower values of the modulus of elasticity, about 20% or 25% below the matrix modulus, and no effect of aggregate shape
is apparent.
Saouma et al. [2], working with dam concretes, obtained similar results for round and crushed aggregates of 38 mm maximum size. Other authors have found different results; Li et al. [3], also working with dam concretes, measured a decrease in
E, when comparing concretes made with round and crushed aggregates. A similar trend was found by Guinea et al. [6] for
concretes made with small aggregates (between 5 and 7 mm). All these results are summarized in Table 4.
4.4. Fracture energy
Fig. 7 shows, for the two shapes of considered aggregates, the experimental results of the specic fracture energy GF (normalized to the corresponding matrix value) as a function of the aggregate size. Concretes made with a strong matrixaggregate interface always show higher GF values than the matrix, increasing with aggregate size. It seems that concretes made
with angular aggregates provide slightly higher GF values than concretes made with spherical ones. Concretes made with
weak matrixaggregate interfaces also gave higher GF values than the matrix and appear to depend on the aggregate shape.
Fracture of concrete made with spherical aggregates occurs by debonding and provides GF values similar to well-bonded
aggregates. Fracture of concrete with crushed aggregates gives higher GF values; by observing the surface fractures it seems

294

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Fig. 6. Normalized elasticity modulus (with respect to the matrix) as a function of the aggregate size, for spherical and crushed aggregates, both with
different kinds of matrixaggregate interfaces.

Fig. 7. Normalized specic fracture energy (with respect to the matrix) as a function of the aggregate size, for spherical and crushed aggregates, both with
different kinds of matrixaggregate interfaces.

that, rst, aggregates partially debond (perhaps due to the weaker interface) and, nally, break due to underpinning and
aggregate interlocking.
For dam concrete, Saouma et al. [2] and Li et al. [3] (for large aggregates) found a similar trend; a small increase when
using crushed aggregates instead of rounded ones. Guinea et al. [6] also found the same trend with concretes made with
small aggregates (sizes between 5 and 7 mm) with and without silica fume. All these experimental results are summarized
in Table 4.
4.5. Characteristic length
It is well known that the fracture energy GF alone does not sufce to characterize the brittleness of concrete and its dependence on the size of the structure. Hillerborg [23] proposed a brittleness number b = l/lch, where l is any structural dimension
and lch a material parameter, know as the characteristic length, dened as

lch EGF =ft2

The higher the brittleness number of a given structure, the lower its ductility. Over the years, there have been other proposals for brittleness numbers (see, for example, Bazant, Planas), most of them closely connected with b.

295

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Values of the characteristic length for the eight simple concretes considered in this research are shown in Table 5 and
relative values, with respect to the mortar matrix, are depicted in Fig. 8. Concrete made with strong matrixaggregate interface always shows values higher than the matrix, increasing with aggregate size. No clear trend was found with respect to
the inuence of aggregate shape. Concretes made with weak matrixaggregate interfaces display much higher values for
characteristic length, particularly those made with treated angular aggregates. In this particular case, the mechanisms
responsible for the signicant increase of GF and those that control the decrease in tensile strength all conducive to high
lch values and, hence, more ductile concretes are worth to be investigated.
4.6. The softening curve
The cohesive crack model is generally accepted as a realistic simplication of the fracture of brittle or quasibrittle materials. In this model, the whole fracture process is lumped into a line, which makes it possible to treat the whole bulk of the
body as elastic. A review of this models main aspects and relevant references appear in a paper by the authors [24] and in
[25]. For a detailed exposition, see [26].
The cohesive crack model is able to capture the main aspects of the fracture of brittle materials [27], particularly of components with blunted notches that do not exhibit a precrack or singularity. This model can be generalized in different ways:
(1) the material outside the process zone considered initially as isotropic linear elastic can behave in a more complex
manner, (2) the softening function may depend on triaxiality, or on previous loading history, and (3) the uniaxial model formulation can be generalized to a mixed-mode one. Discussions of such possible extensions were published in [23,28] and
updated by the authors in [26,29].
In concrete-like materials, the softening curve can be approximated by a bilinear function [30,31], as depicted in Fig. 9.
This simple diagram captures the essential facts: large-scale debonding, or fracture, of aggregates in the steepest part,
and frictional pull-out of aggregates and crack face bridging in the shallow tail of the diagram. This function is completely
characterized when the following four parameters are known, as shown in the gure: the tensile strength rt, the specic
 and the initial slope, measured as the horizontal interfracture energy GF, the abscissa of the centroid of the softening area w,
cept w1 of the initial segment.

Table 5
Average values of the characteristic length and critical crack opening.
Aggregate

Interface

lch (mm)

wc (lm)

w1 (lm)

S3
S9
S14
C3
C6
C9
S9
C6
Matrix

Untreated
Untreated
Untreated
Untreated
Untreated
Untreated
Treated
Treated

92.8
112.6
115.1
101.4
110.4
114.4
174.4
289.7
84.0

196
285
357
192
217
231
456
324
123

15.7
22.3
22.6
16.5
15.1
17.0
30.7
36.2
9.9

Fig. 8. Characteristic length as a function of the aggregate size for the eight simple concretes considered in this research.

296

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

Fig. 9. Bilinear softening function and notation used in the paper.

Fig. 10. Normalized (with respect to the matrix) crack opening as a function of the aggregate size, for spherical and crushed aggregates, both with different
kinds of matrixaggregate interfaces. (a) Critical crack opening and (b) initial crack opening.

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

297

The four parameters needed for the bilinear approximation can be easily computed following the tting procedure developed by the authors [32,33] which makes use of the results from stable three-point bending tests of notched beams and from
indirect splitting tests. The splitting-tension test gives results close to the actual cohesive strength, as was shown in [33].
Therefore, rt was considered equivalent to ft.
The inuence of aggregate shape (spherical or angular) on rt and GF was already discussed. Only the change of w1 and the
critical crack opening wc (see Fig. 8) will be considered here. Average values for the eight concretes and the matrix are summarized in Table 5.
Fig. 10a and b shows, as a function of aggregate size, the relative (with respect to the matrix) average values of w1 and wc,
for the eight concretes considered in this research jointly with experimental results for concretes made with treated spherical aggregates reported in [21]. As may be observed, both sets of results are always higher than the matrix openings and
display increasing values as the aggregate size increases. With respect to the inuence of the aggregate shape, it seems that
the critical crack opening related with frictional pull-out of aggregates and crack face bridging increases faster for spherical aggregates than for crushed ones. The most remarkable results are the differences between treated and untreated aggregates; untreated aggregates (where the matrixaggregate interface was weaker) provided crack opening values much higher
than those measured in concretes with stronger matrixaggregate interfaces. The nonlinear part of the stress-strain curve is
related with cracks starting from the aggregates [34] and, in this respect, the shape of the aggregates should be related also
with this part of the curve. Two parameters of the softening curve the critical crack opening, wc, and the initial crack opening, w1 are involved in the cracking process and its dependence on aggregate shape was discussed, but further research is
needed to clarify the dependence of the nonlinear part of the stressstrain curve with aggregate shape.
5. Summary and conclusions
This research is an extension of previous analysis dealing with a simple cement-based composite made with spherical
aggregates of the same diameter [7,18,21]. Here, the mechanical properties of this simple composite made now with
crushed aggregates are compared with previous ones and the main ndings are summarized in the following paragraphs.
Regarding standard properties, as elasticity modulus and tensile strength, it was found that for strong matrixaggregate
interfaces, concrete made with crushed aggregates provides values of the elasticity modulus somewhat higher than concrete
made with spherical ones, and that there is a tendency for the modulus to decrease with aggregate size for both kinds of
aggregates. The tensile strength shows almost no dependence on aggregate shape and aggregate size. For weak matrix
aggregate interfaces, the elasticity modulus is lower and no effect of aggregate shape is apparent. The tensile strength is always much lower than the matrix and does not appear to depend on aggregate shape.
Toughness is characterized by the specic fracture energy and the characteristic length. For strong matrixaggregate interfaces, concrete made with crushed aggregates provides slightly higher values of the fracture energy than for concrete made
with spherical ones, and the fracture energy increases with aggregate size for both types of aggregates. With respect to the
characteristic length, no clear trend was found regarding the inuence of aggregate shape although it increases with aggregate size. For weak matrixaggregate interfaces, the fracture energy is higher for crushed aggregates than for spherical ones.
The characteristic length also shows higher values for crushed aggregates than for spherical ones.
The softening curve a material property when using cohesive models is also sensitive to the aggregate shape. For
strong matrixaggregate interfaces the relative critical crack opening (with respect to the concrete matrix) seems higher
for spherical than for crushed aggregates and in both cases, increases with aggregate size. A similar trend was found for
the relative initial crack opening w1 (see Fig. 9). For weak matrixaggregate interfaces, the relative critical crack opening is
higher for spherical aggregates than for crushed ones and the contrary seems to happen for the relative initial crack opening
where the values for crushed aggregates are higher than the corresponding spherical aggregates.
Most of these results were already known in a qualitative way for standard concretes [25]: the merit of this research is to
provide quantitative results for a kind of simple composite amenable to be modeled numerically, particularly when dealing
with fracture predictions.
Acknowledgements
The authors gratefully acknowledge Jos Miguel Martnez for help with drawing the gures. The present work was conducted within the framework provided by the Projects Programa Ramn y Cajal, DUMEINPA (S-0505/MAT-0155), sponsored
by the Comunidad de Madrid, Spain, and SEDUREC, integrated in the Spanish national research program CONSOLIDER-INGENIO 2010.
References
[1] Neville AM. Properties of concrete. 4th ed. England: Longman Group Ltd.; 1995.
[2] Saouma VE, Broz JJ, Brhwiler E, Boggs HL. Effect of aggregate and specimen size on fracture properties of dam concrete. J Mater Civil Engng
1991;3:20418.
[3] Li Q, Deng Z, Fu H. Effect of aggregate type on mechanical behaviour of dam concrete. ACI Mater J 2004;101(6):48392.
[4] Donza H, Cabrera O. The inuence of kinds of ne aggregate on mechanical properties of high strength concrete. In: Proceedings of 4th international
symposium of high-strength/high-performance concrete, Pars, France, vol. 2; 1996. p. 15360.

298

C.G. Rocco, M. Elices / Engineering Fracture Mechanics 76 (2009) 286298

[5] Kim J, Lee C, Park C, Eo S. The fracture characteristics of crushed limestone sand concrete. Cement Concrete Res 1997;27(11):171929.
[6] Guinea GV, El-Sayed KM, Rocco C, Elices M, Planas J. The effect of the bond between the matrix and the aggregates on the cracking mechanism and
fracture parameters of concrete. Cement Concrete Res 2002;32:196170.
[7] Rosell C, Elices M. Fracture of model concrete: 1. Types of fracture and crack path. Cement Concrete Res 2004;34:144150.
[8] Kuczynski W. Linuence de lemploi dagrgats gros sur la rsistance du bton. Archiwum Inzynierii Ladowej 1958;4(2):181209.
[9] Franklin RE, King TMJ. Relations between compressive and indirect-tensile strengths of concrete. Road. Res. Lab. Rep. LR412. Berks: Crowthorne; 1971.
32 pp.
[10] RILEM 50-FMC Recommendation. Determination of fracture energy of mortar and concrete by means of three-point bend test on notched beams. Mater
Struct 1985;18:28590.
[11] Guinea GV, Planas J, Elices M. Measurement of the fracture energy using three-point bend tests: I. Inuence of experimental procedures. Mater Struct
1992;25:2128.
[12] Planas J, Elices M, Guinea GV. Measurements of the fracture energy using three-point bend tests: 2. Inuence of bulk energy dissipation. Mater Struct
1992;25:30512.
[13] Elices M, Guinea GV, Planas J. Measurement of the fracture energy using three-point bend tests. 3. Inuence of cutting the P-d tail. Mater Struct
1992;25:32734.
[14] Elices M, Guinea GV, Planas J. On the measurement of concrete fracture energy using three-point bend tests. Mater Struct 1997;30:3756.
[15] Petersson PE. Fracture energy of concrete: practical performance and experimental results. Cement Concrete Res 1980;10:91101.
[16] Rocco C, Guinea GV, Planas J, Elices M. Size effect and boundary conditions in the Brazilian test: experimental verication. Mater Struct
1999;32:2107.
[17] Murakami Y, editor. Stress intensity factors handbook, vol. 4. JSMS and Elsevier Sci. Ltd.; 2001. p. 12.
[18] Rosell C, Elices M, Guinea GV. Fracture of model concrete: 2. The fracture energy and characteristic length. Cement Concrete Res 2006;36:134553.
[19] Thomas TR. Rough surfaces. London: Imperial College Press; 1999.
[20] Saouma VE, Barton C. Fractals, fractures and size effects in concrete. J Engng Mech ASCE 1994;120:83554.
[21] Elices M, Rocco C. Effect of aggregate size on the mechanical properties of a simple concrete. Engng Fract Mech 2008;75:383951.
[22] Kaplan MF. Flexural and compressive strength of concrete as affected by the properties of coarse aggregates. J Am Concrete Inst 1959;55:1193200.
[23] Hillerborg A. Results of three comparative test series for determining the fracture energy GF of concrete. Mater Struct 1985;18:40713.
[24] Elices M, Guinea GV, Gmez FJ, Planas J. The cohesive zone model: advantages, limitations and challenges. Engng Fract Mech 2002;69:13763.
[25] Cornec A, Scheider I, Schwalbe KH. On the practical application of the cohesive model. Engng Fract Mech 2003;70:196387.
[26] Bazant ZP, Planas J. Fracture and size effect in concrete and other quasibrittle materials. Boca Raton, FL: CRC Press; 1998 [chapter 7].
[27] Stang H, Olesen JF, Poulsen PN, Nielsen LD. On the application of cohesive crack modelling in cementitious materials. Mater Struct 2007;40:36574.
[28] Elices M, Planas J. Material models. In: Elfgren L, editor. Fracture mechanics of concrete structures. London: Chapman and Hall; 1989. p. 1666.
[29] Planas J, Elices M, Guinea GV, Gmez FJ, Cendn DA, Arbilla I. Generalizations and specializations of cohesive crack models. Engng Fract Mech
2003;70:175976.
[30] Roester J, Paulino GH, Park K, Gaedicke C. Concrete fracture prediction using bilinear softening. Cement Concrete Comp 2007;29:30012.
[31] Park K, Paulino GH, Roesler JR. Determination of the kink point in the bilinear softening model for concrete. Engng Fract Mech 2008;75:380618.
[32] Guinea GV, Planas J, Elices M. A general bilinear tting for the softening curve of concrete. Mater Struct 1994;27:99105.
[33] Planas J, Guinea GV, Elices M. Size effect and inverse analysis in concrete fracture. Int J Fracture 1999;95:36778.
[34] Hsu TC, Slate FO, Sturman GM, Winter G. Microcracking of plain concrete and the shape of the stressstrain curve. J Am Concrete Inst 1963;60:20924.

You might also like